close

Вход

Забыли?

вход по аккаунту

?

Color Vision in Drosophila melanogaster

код для вставкиСкачать
Color Vision in Drosophila melanogaster
Corey G. Washington
Submitted in partial ñilfillment of the requirements for
the degree of Doctor of Philosophy
under the Executive Committee of the Graduate School of Arts and Sciences
COLUMBIA UNIVERSITY
2010
UMI Number: 3447997
All rights reserved
INFORMATION TO ALL USERS
The quality of this reproduction is dependent upon the quality of the copy submitted.
In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if material had to be removed,
a note will indicate the deletion.
UMI
Dissertation Publishing
UMI 3447997
Copyright 2011 by ProQuest LLC.
All rights reserved. This edition of the work is protected against
unauthorized copying under Title 17, United States Code.
Pro
uest
ProQuest LLC
789 East Eisenhower Parkway
P.O. Box 1346
Ann Arbor, Ml 48106-1346
©2010
Corey G. Washington
All Rights Reserved
ABSTRACT
Color Vision in Drosophila melanogaster
Corey G. Washington
A wide range of organisms have the ability to see in color, to distinguish
between lights on the basis of their wavelength and independently of their intensity.
The visual systems that underlie these capacities vary widely in the number of
receptors involved, the number of wavelength regions that they can distinguish and
the boundaries of the visible spectrum. Through this diversity, it is generally accepted
that an organism must satisfy two basic requirements if it is to have color vision: (a) it
must have at least two visual receptors tuned to different regions of the spectrum, and
(b) its visual system must compare the outputs of these receptors.
I performed behavioral experiments to determine whether the fruitfly,
Drosophila melanogaster, has color vision and, if so, what kind of color vision
system it has. I focused on Drosophila because the range of resources available for
manipulating its nervous system genetically make it potentially a very attractive
model system for studying the biological basis of color vision.
At present research into color vision in Drosophila faces a puzzle. On the one
hand, phototaxis is the only fly behavior that would appear to satisfy the
preconditions for color vision. It is the only behavior that has been shown to implicate
more than one receptor type and published behavioral studies suggest that flies may
have opponent responses to different wavelengths. There is anatomical evidence for
the existence of opponent connections in the underlying neural network that could
support comparisons of the sort thought to be necessary for color vision.
At the same time, behavioral experiments over the years have shown that
phototaxic behavior depends strongly on the intensity of the stimuli over a wide range
of stimulus strengths. Whether flies choose lights of one wavelength over another
can depend on the intensities of those lights, which is inconsistent with color vision
being at work. Moreover, the network underlying phototaxis does not appear to
segregate luminance and chromatic information. Such a convergence is consistent
with the observed intensity-dependence of the behavior but is not typical of networks
that underlie color vision.
To explore the hypothesis that flies have color vision, I carried out phototaxic
experiments with wild type flies using lights of wavelengths across the spectrum over
a range of intensities. Tests involved lights on one or both sides of the apparatus. The
results indicate that when exposed to light on one-side wild type flies exhibit
contrasting behavioral responses to 'UV (331-355 nm) versus 'blue-green' (442-515
nm) stimuli over a range of higher intensities. Higher intensity UV light enhances
phototaxis rates, while higher intensity visible light depresses rates. While this
behavior does not satisfy the definition of color vision because the flies are not
choosing between two lights, it suggests that flies possess some of the key elements
of a color vision system.
In tests with lights on opposite sides of the apparatus, fly preference depends
more strongly on intensity than in one-sided tests. In two-sided tests fly responses
violate the principle of univariance, which is typical of responses based on color
vision, but exhibit opponent behavior at only a limited range of intensities. Thus wild
type behavior in two-light tests provides evidence of significant luminance input into
elements of a color vision system.
The data on wild types suggest that Rl -6, the main photoreceptor class in the
fly retina, which respond to low intensity stimuli, may contribute to the intensitydependence of phototaxis observed in many studies. To explore this hypothesis, I
carried out phototaxic tests with RhI -mutants, which lack functional Rl-6 cells.
While wild types respond differently to higher intensity UV and blue-green stimuli,
the main spectral division in the behavior of Rh1-mutants is between 442 nm ('blue')
and 515 nm ('green'). Flies are attracted to higher intensity blue light, while they
avoid higher intensity green. Preference for blue over green is consistent over a wide
range of intensity settings, though not all. Thus, i?/zi-mutant behavior closely
approximates blue-green color vision. These results imply that Rl-6 not only
contribute to the detection of low intensity light but also have influence over the
chromatic discriminations that they flies make.
Table of Contents
Table of Contents
i
List of Figures
ii
List of Tables
vi
Acknowledgements
vii
Dedication
viii
Chapter 1 . Introduction
1
Chapter 2. The problem of color vision in Drosophila
37
Chapter 3 . Materials and methods
82
Chapter 4. Chromatic opponency in wild type flies
92
Chapter 5. Receptor contribution to opponency
128
Chapter 6. References
157
?
List of Figures:
1.1
Absorption spectra of primate photoreceptors
1.2
Univariance at a single receptor
11
1.3
Univariance atan additive integrator
14
1 .4
Bivariance at an opponent integrator
17
1 .5
The primate retina contains two post-receptor chromatic opponent mechanism..20
1 .6
Luminance input into an opponent system can result in responses of uniform
polarity for all wavelengths
1 .7
9
24
The retina consists of distinct types of ommatidia containing different receptor
cell types
27
1.8
Spectral sensitivity functions of Rl-6 cells and the opsins
30
1 .9
The early layers of the phototaxis network may contain an opponent system that
receives luminance input
1.10
32
Distinct Rl -6 cell types that view a common point in space project to the same
lamina cartridge
34
2. 1
Luminous efficiency function of rods and cones
40
2.2
Discrimination between lights drops to a minimum around estimated point of
equal brightness
2.3
Rats distinguish lights of wavelengths above 400 nm from lights of wavelengths
below 400 nm
2.4
44
46
Bee phototaxis rates increase monotonically with intensity in tests of one light
against darkness
49
2.5
Fly phototaxic response as a function of intensity in a white light versus spectral
light assay
2.6
52
Fly phototaxic responses depend less on intensity as the wavelength of the
variable source increases
54
2.7
Spectral sensitivity of simple model additive integrator
57
2.8
Behavioral response to different wavelength stimuli form parallel curves if based
on purely additive integration
59
2.9
Wild type flies prefer pure UV to UV + green at higher intensities
71
2.10
Wild type flies have opposed responses to lights of greater and less than 4 1 4 nm
at certain intensities against a bilateral UV baseline
74
2.1 1
Spatz/Jacob model of phototaxis
77
3.1
Experimental setup and T-maze
83-4
3.2
Y-maze
4. 1 .
One-light phototaxic response for dark-adapted wild type flies for 6 wavelengths
86
at 8 intensities
4.2.
94
Dark-adapted flies phototax at higher rates in response to UV than blue-green
light at higher intensities in one-light tests
4.3.
One-light phototaxic response for light-adapted flies for 6 wavelengths at 8
intensities
4.4
96
98
Light-adapted, wild type flies phototax at higher rates in response to UV than
blue-green stimuli at higher intensities in one-light tests against darkness
99
4.5.
Fly choice between 331 nm and 400 nm sources depends on light intensity
103
4.6.
Phototaxic choice between 442 nm and 5 1 5 nm light sources depends on
¡ii
intensity
105
Preference for 355 nm versus other wavelengths is intensity-dependent
108
Univariance is violated in the choice between a fixed 355 nm source and variable-
intensity UV and blue-green sources
1 09
Adding fixed-intensity UV light to a bilateral UV stimulus enhances phototaxis
rates, while adding fixed-intensity blue-green light depresses response rates
113
Phototaxic response to stimuli of variable wavelength and intensity added to a
one-sided 331 nm baseline
116
Adding UV light to a 331 nm baseline enhances, while adding blue-green light
diminishes, the attractiveness of a stimulus
117
Phototaxic response to light of variable wavelength and intensity added to a onesided 442 nm baseline
118
Adding UV to a 442 nm baseline enhances the attractiveness of the stimulus,
while adding visible light decreases the stimulus' attractiveness
119
Flies prefer added UV to added 515 nm light across a range of intensities against
a bilateral UV background
122
Phototaxic response of RhI-mutants at 6 wavelengths and 5 intensity levels
133
Phototaxic response is reduced in RhI-mutants relative to wild types at all
wavelengths
134
RhI -mutants exhibit opponent responses above and below 486 nm higher
intensities
135
Flies prefer 442 nm to 515 nm at nearly all wavelength and intensity settings in
Y-maze tests
138
IV
Phototaxic response of Rh6-mutants to one light at 6 wavelengths and 7
intensities
142
One-light phototaxic response of wild types and i?/z<5-mutants to 442-515 nm
stimuli
143
i?/ztf-mutants exhibit lower maximum responses than wild types for all
wavelengths
144
i?/z6-mutant responses at higher intensities and longer wavelengths decline less
than wild type responses
145
Phototaxic response of ortc2-shits flies is impaired relative to wild types at control
temperatures
149
Phototaxic responses oiortc2-shf flies at experimental tempertures show relative
reductions in phototaxis rates at 33 1 , 355 and 515 nm
?
150
List of Tables:
1.1.1 List of photoreceptor cell types along with their opsins and the sensitizing
pigment
Vl
29
Acknowledgements
I would like to thank my advisors, Andrew Tomlinson and Richard Axel, for initiating
this project and for support throughout the years it took to complete. I would also like to
thank the members of my dissertation committee, Claude Desplan, Mickey Goldberg and
Don Hood for encouragement, guidance and feedback. I am also indebted to my friend
and collaborator Atsushi Sato for help with devising and carrying out experiments. My
experience in graduate school was enriched by the friendship and camaraderie of present
and former members of the Tomlinson lab including Vladimir Katanaev, Jose Galan, Ron
Arias, Naureen Quibria, Hina Patel, Yannis Marvomatakis, Jason Rojas, and Suda
Kumar. I would also like to extend my gratitude to Gary Johnson and other members of
the machine shop for their help in building the experimental apparatus. I would also like
to thank John Koester, Carol Mason, Darcy Kelley, Cecil Oberbeck, Alla Kerzhner, Ken
Miller and other members of the neuroscience program for administrative support and for
fostering a nurturing environment in which to learn neuroscience. I would also like to
thank Elena Demireva for extensive help editing this manuscript and Kristen Werner for
help in preparing my thesis presentation. Finally, I would like to thank Lola Kadiri for
friendship and support over the years.
VII
Dedication
I would like to dedicate this thesis to my mother, Ruth Washington, and my father, Ernest
Washington, with love and affection.
VlIl
Chapter 1 . Introduction
Chapter 1. Introduction
Color vision is widespread throughout the animal kingdom. Primates and many
mammals, birds and some reptiles, fish and amphibians, cephalopods (squid and octopus)
and some arthropods (insects and crustaceans) see in color. The range of color vision
systems possesed by animals is also immense. Most mammals are dichromats (Jacobs,
2009), while some birds are pentachromats (Palacios and Várela, 1992; Palacios, et. al,
1990) and mantis shrimp may have up to 12 types of receptors that may function as
dichromatic pairs (Cronin and Marshall, 1989). The wavelengths organisms can perceive
ranges from 300 nm for some insects (Salcedo, et. al, 1999) up to 720 nm at moderate
intensities for humans (Le Grande, 1957). Humans can distinguish hundreds of hues
independently of intensity (Boynton, 1979), while rats can distinguish only two (Jacobs,
et. al, 2001).
Given how pervasive color vision is, it is not surprising that organisms derive
extremely useful information from it. Color vision allows animals to distinguish plants
that are suitable to consume from those that may be poisonous and to identify animals
that may be dangerous. In some cases it enables organisms to determine whether a
conspecific is agitated or ready to mate. Animals that lose the ability to see in color due
to injury or genetic deficiency are at a significant competitive disadvantage.
One can learn whether an organism has color vision only by studying its behavior. I
carried out behavioral experiments to determine whether the fruitfly, Drosophila
melanogaster, has color vision, and if so, what kind of color vision system it has. I
focused on melanogaster because of the immense possibilities for manipulating its
1
Chapter 1 . Introduction
nervous system genetically. There are null mutants for each of the visual receptors
(Yamaguchi, et. al, 2010) and a range of powerful techniques exist for altering the
function of the receptors and the pathways downstream of them. Because each visual
receptor can make a contribution to its owner's ability to see in color, a great deal can be
learned by studying individuals that lack the normal complement of receptors. When it
comes to most species, in contrast, the only option is to wait for accidents of nature to
produce such individuals.
If flies have color vision, these resources make melanogaster a very attractive
model for probing the development, form and function of the system. One may, for
example, be able to shed light on different forms of color blindness by examining
whether the discriminatory capacities of fly photorecceptor mutants is analogous to those
of humans and other animals with mutations in their receptors. One can also take
advantage of the opportunities to create receptor combinations that may not exist in
nature to perform experiments that cannot be performed in other animals. The relative
simplicity of its nervous system provides another reason for studying color vision in the
fly. A central goal of neuroscience is to understand how neural circuits produce
behavior. Because relatively few synapses link the fly visual system to its behavioral
output, it may be possible to use behavior to gain insight into the neural processing that
underlies color vision. Such inferences from behavior to circuitry may be less reliable
when carried out in more complex organisms.
Given what might be learned from the fly as model for color vision, it is not
surprising that the question of whether flies have it has long attracted the attention of
2
Chapter 1 . Introduction
researchers. In fact, people have been working on the problem for nearly 40 years
(Schlumperli, 1973). Over the decades a number of interesting experiments have been
done with mixed results. Some of the earliest, using the simplest implements and
experimental designs, have produced highly suggestive data (Heisenberg and Büchner,
1977; Fischbach, 1979). Some later ones, using flight simulators have produced negative
results (Yamaguchi, et. al, 2008). While a number of authors have argued that flies have
color vision (Quinn, et. al, 1974; Merme and Spatz, 1977; Hernandez de Salomon and
Spatz, 1983; Tang and Guo, 2001), convincing results have been elusive. After all this
time, the question of whether flies have color vision remains without a solid answer.
The experiments reported below seek to demonstate that flies exhibit features of
color vision in phototaxis, a simple reflex in which an organism moves toward a light or
chooses between two lights. Phototaxis is a robust behavior in flies, and all of the fly
visual receptors (Rl -8) have been implicated in the response (Yamaguchi, et. al, 2010).
Recent neuroanatomical studies of the phototaxic network (Gao, et. al, 2008) also show
that it includes pathways that receive input from distinct spectrally-biased receptors (R7
and R8) and may contain opponent connections. Together with previously published
behavioral findings these results suggest that flies may exhibit elements of color vision in
phototaxis.
On the basis of experiments performed using a T-maze, I will argue that flies have
contrasting responses to light from two wavelength regions, roughly UV and visible,
independently of their intensity. I will also argue that this ability rests on an opponent
mechanism that contrasts UV and visible regions. Along with multiple receptors
3
Chapter 1 . Introduction
sensitive to different regions of the spectrum, the existence of an opponent system in
which an integrating cell is excited by one receptor type and inhibited by another with a
differing spectral sensitivity is signature feature of color vision circuits (Conway, 2009).
The fly opponent system appears to be of an unusual sort, however, in that the relation
between the receptors is asymmetric. In most systems there are paired, symmetric,
opponent pathways. In flies one receptor type enhances the behavioral response, while
another diminishes it, and these roles never reverse.
It appears that luminance input enters the fly visual system through the main
photoreceptor cells, Rl -6. If this is correct, then eliminating this input may leave the fly
with a pure chromatic mechanism, i.e. an opponent system that is unadulterated by
luminance. Such a reduced visual system may also be able to support color vision. I
carried out Y-maze experiments with RhI -mutants, which lack functional Rl -6 cells, to
test this hypothesis. For most intensity settings RhI -mutants prefer blue (442 nm) to
green (515 nm) and increasing the intensity of the green light does not, in general, make
it more attractive. I conclude that while the behavior of RhI -mutants does not satisfy the
definition of color vision, they do come quite close to discriminating between blue (442
nm) and green (515 nm) independently of intensity.
4
Chapter 1 . Introduction
Definition of Color Vision: Let us begin by defining 'color vision'. It is generally
accepted that an organism has color vision ifand only ifit can distinguish between
(reflected, transmitted or emitted) lights ofdifferent wavelengths independently oftheir
intensity.
One of the virtues of this definition is that it is cast in purely physical terms that do
not depend in any way on the subjective experience of the organism. Whether an
organism has color vision is a question solely of its ability to tell signals apart using
wavelength and not intensity. So although the experience of color in humans is something
subjective and an interest in understanding this subjective element was almost certainly
part of what motivated researchers to study color vision in the first place, the essential
problem does not turn on colors, perceptions or other subjective entities. This separation
of the subjective and objective features of the topic is, of course, what makes it possible
to study color vision in non-human organisms, like flies, to whose experiences we have
no access, and that may have no conscious experiences at all.
Before proceeding further it will be helpful to explain what is meant by some of the
terms in the definition. The 'wavelengths' to which the definition refers include all parts
of the spectrum that is visible to the organism, whether narrow or broad in extent.
Second, it is important that the notion of intensity independence not be taken too literally.
It does not mean that an organism needs to be able to distinguish between lights so dim
they can barely be detected or between lights as bright as the sun. For each organism,
there an intensity level below which color vision ceases to operate, at which vision
becomes achromatic. Such low intensity signals are often described as falling into the
5
Chapter 1 . Introduction
'scotopic' region of the intensity spectrum (LeGrande, 1957). At the other extreme, some
lights are so bright, they can be blinding, overwhelming the cells that process chromatic
information in the animal. Over short periods of time, when an eye is adapted to lower
light levels, this can easily occur. For extremely bright lights it can be perpetual.
It is when the intensity of the lights falls in the middle regions in which they are
bright enough to activate, but not too bright to overwhelm, the chromatic receptors that
the comparison between them is to be made. For us, when a light's intensity falls within
these regions, it will activate the three types of cones cells that in normal subjects are
responsible for the perception of color. The distinction an animal makes does not have to
be fully independent of intensity but only independent over a wide range of intermediate
intensities. Whether this range is three, five or ten orders of magnitude will depend on the
neural system that underlies the capacity, driven by evolutionary pressures from its
environment, and ultimately to be determined by experiment. In what follows 'color
vision' assumes intensity-independence only in this limited sense. Our question is
whether there is a range of intensities, above and below certain thresholds, over which
Drosophila can distinguish two lights in a consistent fashion according to their
wavelengths.
6
Chapter 1 . Introduction
The elements of a color vision network
In this section, we consider the characteristic features of color vision systems and
the reasons for thinking that they are either required for the capacity or sufficient for its
existence. To illustrate these points we discuss a range of examples with primates as the
central case.
Multiple Receptors: Any animal with color vision must have at least two visual receptors
that process wavelength information. The number of color receptors actually possessed
varies widely between species. The best-studied insect is the honeybee, which has nine
photoreceptors (Wakakamu, et. al, 2005), three of which, sensitive to 350, 440 and 540
nm light, are involved color vision (Menzel and Blakers, 1975). Bumblebees also have
nine photoreceptors (Rl -9) and again three (R7-9) are involved in color processing. The
butterfly Papillo xuthus has six receptors, sensitive to UV, violet, blue, green, red and
broadband light (Arikawa, 2003). Aside from the broadband receptor, all of these cells
are thought to play a role in color processing (Kinoshita, et. al, 2005). Most mammals are
dichromats. Rats have one type of rod and two cones, one sensitive to the UV (359 nm)
and the other to green (510 nm; Jacobs, 2001).
The color vision systems of New World primates are often polymorphic. Male
squirrel monkeys {Saimirí sciureus) are dichromats with one receptor in the short
wavelength region and one in the medium or long region (wavelengths for M to L include
535, 550, or 563 nm). Females have the short wavelength cone and are di- or trichromats,
depending on whether they are homozygous or heterozygous for the middle/long
wavelength cone. This pattern is also found in other New World species, though it is
7
Chapter 1 . Introduction
unknown whether the distribution is universal. Old World monkeys and apes are
trichromats, possessing S, M and L cones with peaks around 420, 530 and 560 nm. They
display little evidence of polymorphism. While normal humans are also trichromats, they
show far greater polymorphism than their fellow Old World species, which accounts, in
part, for the existence of color vision anomalies in approximately 4% of the population,
mostly males. In most of these cases, the M or L cone is either missing or nearer in peak
sensitivity to its partner than normal. Some nocturnal primates, e.g. the owl monkey
(Aotus) and the thick-tailed bushbaby {Otolemur crassicaudatus) have only a single cone,
either M or L, and are completely colorblind (Jacobs, 2003).
For illustrative purposes lets us consider the Old World primate color vision system
in some detail. The retina contains four types of receptors, three cones and one rod
(figure 1.1).
8
Chapter 1 . Introduction
420
498
534
9
564
I OO
¡u
o
m
£
<tì
^
v¡
SO
«
S
)
t
1
i
1
400
¡lotef
1
1
L
\H
1—?—?—?—?—i—?—?—?—?—?—? ? ? ? ? ? g ? ? ? ? ? ? ? ? ? I
500
Bitte
l'Yüfí
600
(¡/ven
Yetioxv
700
Red
Wavelength (nini)
Figure 1.1. Absorption spectra of primate photoreceptors. The rod appears dotted in
black, while the three cones, whose labels 'S', 'M' and 'L' mark sensitivity to different parts of
the spectrum, are indicated in color. The wavelengths of the peak sensitivities appear above
their curves. Adapted from Bower and Dartnall (1980).
Chapter 1 . Introduction
10
The rod system is far more sensitive than the cone system and forms the basis for our
dim light, scotopic, vision. Rods are responsible for the black and white, or more precisely
black and bluish-white, experience we have under moonlight or just as objects become
visible. As light levels rise and cones become active vision becomes mesoptic. With further
increases in light levels rods become saturated, while the three cones continue to respond
normally. At these, photopic, light levels color has its "normal" appearance.
To understand why more than one receptor is required for color vision we need to
consider a property of receptors known as 'univariance'. Univariance lies in the fact that the
response of a photopigment to absorbing a photon is independent of the photon's wavelength.
In consequence, a photoreceptor's response to a light signal depends solely on the number of
photons absorbed, not their wavelengths (Le Grande, 1957). Each receptor responds to
photons of a given wavelength in a probabalistic manner, captured by its sensitivity profile. A
photon whose wavelength is closer to the peak of a receptor is more likely to be absorbed than
others of more distant wavelengths, but once a photon is absorbed its effect on the receptor is
the same: the conformation of the chromophore changes from eis to all-trans and the
transduction process starts. The response level of the receptor depends only on the rate at
which photons are absorbed. For a monochromatic source of wavelength ?, this rate will be a
product of the probability of the cell absorbing a photon ofthat wavelength (Pj1) and the rate
at which photons ofthat wavelength hit the receptor (??).
1.
??? = ??*??.
Chapter 1 . Introduction
11
Univariance at a Single Opsin Receptor
550
700
Wavelength (nm)
Figure 1.2. Univariance at a single receptor. The graph plots the sensitivity profile of a
hypothetical receptor whose peak sensitivity is 534 nm. 500 nm photons are more likely
to be absorbed than 600 nm photons, but all photons result in depolarization. The receptor
will have the same response to signals consisting of 500 nm and 600 nm photons,
respectively, if the intensities are suitably calibrated.
Chapter 1 . Introduction
For a source with a range of wavelengths the rate will be the sum over all of
wavelengths (?) of the product of the photon flux (I) and the probability of the cell
absorbing a photon (P) at that wavelength:
2.? = /?(?)?(?)??.
If there are sources Si and S2, consisting of photons of wavelengths ?? and ? 2,
respectively, with S2 having K times the photon flux of Si, and the receptor being K times
as sensitive to ?? as ?2, then the sources will produce identical responses in the receptor,
the higher number of photons coming out of S2 compensating for the lower probability of
responding to any given S2 photon. No receptor on its own can tell these two sources
apart.
In the example in figure 1 .2, the response of the receptor to the 500 and 600 nm
sources will be equivalent if the intensity of the 600 nm source is just under three times
the level of the 500 nm source, matching the proportionally greater probability that the
receptor will respond to a 500 nm photon. The selection of intensities is more complex
when the sources are not monochromatic, but in principle the same results can be
obtained for sources of any two wavelength distributions. Because sources with very
different wavelengths and intensities can have precisely the same effect on a receptor, the
response of a single receptor cannot separate information about wavelength from
information about intensity.
While this argument demonstrates that having two or more receptors is necessary
for an organism to have color vision, it is important to realize that having multiple
12
Chapter 1 . Introduction
receptors is not by itself sufficient for color vision. An animal with two cones might not
be in any better position to see in color than an animal with a single receptor if the proper
downstream wiring is not present.
Opponency: If the cell downstream integrates the outputs of two receptors in an additive
manner, its response will have a sensitivity profile that is the (possibly non-linear) sum of
the profiles of the two inputs. Because the response to stimuli of different wavelengths
will be of the same character (depolarization or hyperpolarization, depending on whether
the receptors excite or inhibit the cell), one can easily choose intensities for any two
stimuli that will produce the same response in such an integrator.
13
Chapter 1 . Introduction
timi
hail if it ·?
Figure 1.3. Univariance at an additive integrator. The graph plots the sensitivity
profile of a hypothetical integrator that takes additive input from receptors with peaks at
534 and 564 nm with spectral signals of 500 and 600 nm as example stimuli. The
integrator will have a stronger response to the absorption of a 500 nm photon but the
response to both, as with all other stimuli, will be depolarization. Responses to 500 nm
and 600 nm stimuli can be equilibrated by appropriate choice of intensities.
14
Chapter 1 . Introduction
The behavior of such an integrator will still be univariant. No decision mechanism
based on such an integrator can respond differently in a systematic fashion to lights of
different wavelengths. For this reason an additive integrator taking input from two
receptors cannot perform color processing (figure 1.3).
It is sometimes said that color vision requires the comparison of the outputs of two
or more photoreceptors. Computationally, this amounts to the claim that calculating the
difference between, or ratio of, receptor activations is necessary for color vision to arise
(Conway, 2009). If differences are calculated, one receptor must excite an integrating
cell, while the other inhibits it, i.e. the integrator must be on the receiving end of
opponent input. It has been argued that opponency is necessary because it is needed to
avoid a build up of noise in the chromatic system that would otherwise overwhelm the
wavelength signal if the responses of neurons are merely added together at each
processing stage (Lee, et. al, 2002).
It is difficult to tell how strong the argument for the necessity of opponency is, but
if there is opponency the comparison should take place early in the visual pathway, as
Boynton explains in connection with the human visual system: "the initial cone signals of
vision exist in analog form and, moreover, depend on the statistical properties of light.
The farther such signals are carried through the visual system without being compared
the less reliable they become. Therefore, to keep the signal-to-noise ratio as high as
possible it is best to compare the signals early. A differencing function taking place near
the receptors constitutes such a comparison" (Boynton, 1979, p. 215).
15
Chapter 1. Introduction
In contrast to the difficulty in assessing the strength of the argument that opponency
is necessary, I think a straightforward case can be made for the idea that subtraction of
receptor activations is sufficient for color vision when the response is independent of
intensity and tied directly to behavior. If the integrator is a subtracter, the polarity of its
response will differ for different wavelengths, and this difference will not disappear with
changes in source intensity. Increasing or decreasing the intensity of an input that excites
the opponent cell will leave it as excitatory. An inhibitory input whose intensity is varied
remains inhibitory. The cell will be bivariant, i.e. have qualitatively different responses to
the two sources. If the response of such an integrator dictates the behavioral response of
the organism, then the organism will respond differently to the two wavelengths
regardless of intensity and so have color vision.
The point emerges more clearly if we consider a concrete case.
16
Chapter 1 . Introduction
No Univariance at Opponent (Subtractive) Synapses
2r
15h
1
-
* 0.5O
-O 5 I
400
'
450
¦
500
'
550
T
600
'
650
'
700
Wavelength (nm)
Figure 1.4: Bivariance at an opponent integrator. The graph plots the sensitivity
profile of a cell that is depolarized by a receptor with a 534 nm peak and hyperpolarized
by a receptor whose peak is 564 nm. So long as 500 and 600 nm inputs are detectable,
increasing or decreasing their intensity will not result in their having the same effect on
the cell. The cell's response is bivariant with the division between stimulus types
occurring at 549 nm.
17
Chapter 1 . Introduction
The cell in figure 1 .4 takes excitatory input from a receptor whose peak is 534 nm
and inhibitory input from a cell whose peak is 564 nm. Wavelengths below 549 nm excite
the cell, while those above inhibit it. Changing the photon flux of a source composed of
wavelengths above (below) 549 nm will never excite (hyperpolarize) the cell. No
arrangement of intensities can cause photons from these regions to have the same effect.
In this way, opponency separates wavelength from intensity. If an organism responded to
light stimuli in line with the response of this neuron, something that is likely to occur
only in very simple cases, its behavior would discriminate between the two wavelength
regions independently of intensity. If the contrasting responses occur when two lights are
present, then the organism has color vision.
Neural opponent mechanisms have been found in a wide range of organisms with
color vision. The honeybee appears to have two opponent mechanisms in the medulla, the
second layer of its optic lobe, one opposing UV to blue+green and the other blue to
UV+green (Backhaus, 1991). Color opponent ganglion cells opposing blue and green
have been found in the cat retina (Cleland and Levrick, 1974). The turtle retina contains
three types of horizontal cells with opponent characteristics; one hyperpolarizes in
response to UV, S and M cone (400 - 600 nm) input and polarizes in response to longer
wavelength, L cone (>600 nm), stimuli; another cell has similar properties but reverses
polarity at 540 nm; a third hyperpolarizes in response to short and long wavelength
stimuli and is depolarized by 520 - 640 nm stimuli (Ventura, et. al, 2001). Some
evidence points toward domestic chicks having a red-green opponent mechanism
(Osorio, et. al, 1999). Behavioral studies suggest that European Starlings and Japanese
18
Chapter 1 . Introduction
Quail may use UV or a short wavelength visual receptor in opponency with a longer
wavelength receptor (Smith, et. al, 2002).
There is evidence of neural opponency at different points along the primate visual
pathway. In macaques, opponency with center-surround receptive field structure may
occur in photoreceptor cells via feedback from horizontal cells (Verweij, et. al, 2003;
Orin, et. al, 2010). Trichromatic primates possess two opponent mechanisms in the postreceptor pathways, S/(LM) and M/L. The firing rate of one set of cells in the S/ML
system is raised by activation of S cones and inhibited activation of M and L cones, while
the firing rate of another set of cells is raised by M and L cone activation and inhibited by
S cone activation, similarly for the M/L system. Dichromats lacking M or L cones have
only a mechanism that opposes the S cell to the remaining M or L cell (DeValois, 1965;
De Valois and De Valois, 2003). The diagram below depicts the retinal pathways of Old
World primates.
19
Chapter 1 . Introduction
Circuitry of primate retina
¦-î'yt'^
Figure 1.5: The primate retina contains two post-receptor chromatic opponent
mechanisms. The M and L cones feed into red/green opponent midget ganglion cells,
while all three cone types contribute to S/ML opponency in the small bistratified cells
and small-bodied inner cells. Chromatic opponent cells appear in color, while the
achromatic mechanism appears in black and white. Adapted from Martin (2004).
20
Chapter 1 . Introduction
Small bistratified ganglion cells are S-ON/ML-OFF; activation of S cones raises the
firing rate of the cells, while activation of M and L cones lowers it. S-OFF/(M/L)-ON
responses have been recorded in the small-bodied inner cells (Calkins, et. al, 1998;
Percival, et. al, 2009). M/L opponency first occurs in the midget ganglion cells of the
retina (Lee, et. al, 2008; Conway, 2009).
Cells excited and inhibited by S cones are often described as 'blue' on and off,
respectively. Cells involving M and L opponency are similarly labelled 'green' and 'red'.
It is important to realize, however, that labels such as 'blue', 'green' and 'red' are merely
mnenomic and may not accurately describe the response of these cells to light of these
general hues. The fact that the S-cones are inhibited by horizontal cells, which take input
from M and L cones, means that their firing rate is depressed by M and L cone activation,
not that yellow light alone inhibits them. Rather, it appears that these cells may be excited
by blue light and inhibited not by yellow specifically but by white/broadband light
(Conway, 2009).
Segregation of pathways: One of the most striking features of the primate visual
system is the segregation of chromatic and luminance channels along the main visual
pathways. Luminance information is carried primarily by the magnocellular ('M')
pathway, which begins by taking input from M and L cones in an additive, non-opponent,
fashion. The signal from these cells passes onto ON- and OFF-diffuse bipolar cells,
which form opponent synapses onto parasol ganglion cells, contrasting light and dark,
and then moves into the four ventral magnocelllar layers of the LGN. Magnocellular
LGN cells have phasic responses to maintained achromatic contrast and show little cone-
21
Chapter 1 . Introduction
22
opponency (Martin, 2004). They project to layer IVCa of Vl, terminating in the
interblob, orientation-senstitive, regions of Vl.
Unlike the M pathway, the two chromatic pathways employ substantial coneopponency. The parvocelluar and konicocellular pathways carry M/L and S/ML opponent
information, respectively. The chromatic information carried along the P pathway begins
with the L and M cones. Within the fovea each M and L cone contacts a single ON- and a
single OFF-midget bipolar cell, which in turn contact a midget ganglion cell, via a coneopponent synapse. The pathway projects to the dorsal parvocellular layers of the LGN
and onto IVCß of Vl, before reaching the blob regions of Vl. The K pathway takes
opponent input from S cones against M and L cones, passes through interlaminer
koniocellular zones flanking the magnocellular layers in the LGN, and onto layers 3/4a of
Vl (Chatterjee, et. al, 2007). The K-pathway carries most of the of the S/ML opponent
signal, but about a third of the S/ML opponent cells found in the LGN occur in the
parvocellular layers (Martin, 2004).
The segregation of visual information into chromatic and luminance pathways
would appear to play a significant role in preserving the opponent character of the
responses of individual neurons. To see why, consider what can happen if this type of
segregation is absent. The simple model opponent neuron described above receives input
from cells with the same receptive field (no center-surround). A 534 nm receptor excites
the cell, while a 564 nm receptor inhibits it. If broadband luminance input, equivalent to
input from the two receptors combined, is added to the opponent inputs, it could push the
cell response positive or negative for all wavelengths, wiping out the opponent behavior.
Chapter 1 . Introduction
The diagram below shows the effects of excitatory non-opponent inputs of various
strengths. (Inhibitory non-opponent input would move the output downward.)
23
Chapter 1 . Introduction
Opponent Synapse with Degrees of Luminance Input
Pure Opponent
+.2 Luminance
+.4 Luminance
Luminance
Luminance
inance
550
Wavelength (nm)
Figure 1.6: Luminance input into an opponent system can result in responses of
uniform polarity for all wavelengths. The curves represent the response of a
hypothetical integrator that takes constant opponent input (peak excitatory input 534 nm,
peak inhibitory input 564 nm) and luminance input (534 nm added to 564 nm peak input)
of a variety of strengths. When the luminance input is small the integrator retains its
opponent character (luminance constant=.2 to .8). When the luminance input is strong
(luminance constant =1), the response has the same polarity for all wavelengths.
24
Chapter 1 . Introduction
25
The response properties of the original M-L neuron appear in green at the bottom of
figure 1.6. The upper curves represent responses in which luminance inputs of strength
k*(M+L), k = .2, A,. . .,1 are added to the initial M-L profile. For small additions of
luminance, the response remains opponent in character. The neuron is excited by some
wavelengths and inhibited by others. When the luminance contribution reaches k=l, all of
the L cone activated inhibition has been eliminated ((M-L)+ 1*(M+L) = 2M) and the
neuronal response is positive for all wavelengths, leading to univariance. Just as with a
single receptor, sources of different wavelengths stimulating the visual pathway will have
the same effect on the integrator provided their intensities are suitably calibrated. So if
the discriminatory capacities of an organism are tied to an integrator that takes both
opponent and strong luminance input, the organism could lack color vision. Whether it
retains opponency or not will depend on the relative strengths of the opponent and nonopponent inputs at different wavelengths as well as the intensity ranges over which they
are active. If the opponent input arrives at higher intensity levels than the luminance
input, the system will be segregated functionally if not anatomically. In such a case, an
organism with behavior based on the opponency may have color vision at higher intensity
levels.
Chapter 1 . Introduction
The Fly Phototaxic Network
In this section I review anatomic and physiological data regarding the fly retina and
the early parts of its phototaxic network. This network appears to have the two basic
elements of a color vision system; it contains multiple receptor types with differing
spectral sensitivities, and there is evidence of downstream opponent integration.
However, it appears that the pathway that processes intensity information is not
segregated from those that process wavelength, which may result in behavior that is not
opponent and not conducive to color vision.
Let us begin with the receptors. The receptors reside in the retina, which consists of
approximately 750 optically isolated ommatidia, each of which contains three basic
sensory cell types: (i) six large, outer cells, Rl-6 that extend the length of the retina, (ii) a
smaller diameter, centrally located, cell R7 covering the distal 2/3 of the retina, and (iii)
another small cell proximal to, and immediately below, R7, extending the remaining part
of the retina, R8 (figure 1 .7). Each photoreceptor cell expresses a single receptor
molecule (known as 'i?/jodopsins' -Opsins for short—and numbering T to '6'), but R7s
and R8s come in two correlated varieties, each of which expresses a different receptor.
'Pale' ommatidia ('p' for short) contain i?/z3-expressing R7 cells and .KM-expressing R8
cells, while 'yellow' ommatidia ('y' for short) contain i?/z4-expressing R7 cells and Rh6expressing R8 cells, 'pale' and 'yellow' referring to the photoreflective properties of the
ommatidia. At the extreme dorsal portion of the eye (dorsal rim), both R7s and R8s
express Rh3. All Rl -6s express RhI.
Chapter 1 . Introduction
RH3/RH5
RH3/RH3
?
Ì
*."·:»
•
»*;·.
· 2
ß T !
Ii
?
e· »
6 ? I
11/
-\
1
Pale
YiIl(W
Dorsal Rim
Figure 1.7: The retina consists of distinct types of ommatidia containing different
receptor cell types. The three types of ommatidia are 'pale', 'yellow' and 'dorsal rim'.
Pale ommatidia contain R7 cells that express the Rh3 opsin and R8 cells that exprès
express the Rh6 opsin; yellow ommatidia contain R7 and R8 cells that express Rh4 and
Rh6 opsins, respectively. At the dorsal rim of the retina, all R7 and R8 cells express Rh3.
27
Chapter 1 . Introduction
Each cell type responds preferentially to light of different wavelengths. The identity
of the opsin contained in each receptor is the most important factor determining spectral
sensitivity, but cells may also have sensitizing pigments that absorb light and transmit
energy to the opsins, thereby, extending the cell's sensitive range. In Drosophila Rl -6
cells contain the pigment (3-OH-retinol). Table 1.1 identifies the peak sensitivities of the
opsins and this pigment. Figure 1.8 shows the spectral sensitivity function of Rl -6 and
the opsins.
28
Chapter 1 . Introduction
Cell Type
Opsin (peak nm)
Rl-6
RhI (486)
R7pale
Rh3(331)
R7 yellow
Rh4 (355)
R8 pale
Rh5 (442)
Rh8 yellow
Rh6(515)
Occeli
Rh2 (420)
Sensitizing pigment
(peaknm)
3-OH-retinol (350 nm)
Table 1.1: List of photoreceptor cell types along with their opsins and the sensitizing
pigment. Each photoreceptor cell contains a unique opsin that mediates its sensitivty to
light. Rl-6 also contains a sensitizing pigment that extends its sensitive range into the
UV. Rl -8 occur in the eye, while the occeli are located on the top of the head.
29
Chapter 1 . Introduction
Spectral Sensitivity
0.8 h
0.6 h
0,2 µ
¦l'I il il I H il II' Il III Il I ¡TU» I IH(I
300
violet
UV
Rh 3
400
Rh4
500
600 nm
blue
Rh2
RhS
Rh 1
Rh<V
Normalized
Absorption
3(X)
350
4(K)
450
5(M)
550
6OO
Wavelength nm
Figure 1.8: Spectral sensitivity functions of Rl -6 cells and the opsins. Top:
normalized absorption of the Rl -6 cells as as a function of wavelength. The cells have a
peak at 355 nm contributed by the sensitizing pigment and a peak at 486 nm that is due to
the RhI opsin, giving them significant detection capability across the spectrum. Bottom:
normalized absorption of the opsins as a function of wavelength. (Both figures from
Salcedo, et. al, 1999.)
30
Chapter 1 . Introduction
Data from Calliphora indicate that the occeli are not involved in wavelenth specific
behavior in phototaxis (Kirschfield and Lutz, 1977), but in Drosophila melanogaster
occeli are required for normal phototaxis (Miller, et. al, 1981). A sensitizing pigment has
been detected in Rl -6 cells in Calliphora (Hamdorft, et. al, 1992) and a blue-absorbing
pigment has been detected yellow R7s in Musca and Calliphora (Kirschfeld, et. al,
1978). Given the high degree of homology among Diptera, it is possible that such a
pigment exists in homologous cells in Drosophila, though no evidence for this hypothesis
has emerged so far. Because its receptor cells differ in sensitivity, the fly retina satisfies
the first requirement for color vision.
These five receptors feed into pathways that have been implicated in phototaxis,
described below in a wiring diagram from Gao, et. al (2008). The network contains a
potential opponent synapse and may not segregate chromatic and luminance channels.
31
Chapter 1 . Introduction
Dm8
M
R1-6
o
Eye
L1 "*-*
Lamina
Medulla
Lobula
Figure 1.9: The early layers of the phototaxis network may contain an opponent
system that receives luminance input. The synapse between R7 and R8 indicated at the
top left may be an opponent synapse. If it is functional and typical of other synapses
involving the photoreceptors, it will mediate contrasting responses to visible and UV
signals in the R7 cell. The broad spectrum sensitivity of Rl -6 makes it highly responsive
to intensity and so a candidate luminance channel. In consequence, the convergence
between the Rl-6-to-L3 pathway and R7/8 on the downstream neurons suggests that
chromatic and intensity information might not be segregated in the phototaxis network.
(Adapted from Gao, et. al, 2008)
Chapter 1. Introduction
There are three points in this network at which signals from the receptors may be
integrated. The first is the synapse of R8 onto R7, depicted by the thin connection on the
left of figure 1.9. This should be thought of as a.potential opponent synapse, because
evidence has yet to emerge that it is physiologically functional. However, if it is
functional and the transmitter is histamine and the receptor a chloride channel, as is the
case for other synapses involving the photoreceptors, R8 would inhibit the R7, and the
two would form an opponent system. Because R7s are sensitive to UV and R8s to visible
light, the synapse would create UV-visible opponency. It is not known whether the
ommatidium depicted in the diagram is pale or yellow and so whether the synapse would
oppose 331 to 442 nm or 355 to 515 nm (I. Meinertzhagen, personal communication).
Still, wavelengths from one of these regions would excite the response, while other
wavelengths would inhibit it.
If it is functional, the synapse is well placed to serve as a chromatic opponent
connection. The cells involved belong to the same ommatidium and their receptive fields
coincide. The synapse also occurs early in the pathway when the spectral character of the
pathways is clearly defined, a perfect place, as Boynton underscores, to support responses
that differ according to the wavelengths of the inputs.
33
Chapter 1 . Introduction
GC
ZB
Figure 1.10: Distinct Rl -6 cell types that view a common point in space project to
the same lamina cartridge. Rl -6 cells that view a common point in space (4 mm in
front of the fly eye) are located at different positions (Rl in one, R2 in another, etc.)
within the ommatidia that surround a given ommatidium and project to a common lamina
cartridge (right). Within a given ommatidium, the Rl -6 cell types view different points in
space and project to lamina cartridges that surround the catridge corresponding to that
ommatidium (red circle, left, middle). As a result, cartridges are defined functionally as
lamina units taking visual input from a common point in space. (Adapted from Hardy,
1985).
34
Chapter 1 . Introduction
The second point at which signals integrate is the lamina (the first layer of the optic
lobe) which receives input from Rl -6. The lamina is divided into 'cartridges', each of
which corresponds to an ommotidium in the eye. However, cells from an ommatidium do
not project to the counterpart cartridge but rather to the cartridges in the perimeter around
the counterpart (figure 1.10, left). Each cartridge receives input from the ommatidia
surrounding its counterpart ommatidium (figure 1.10, right). One cell from each of these
perimeter ommatidia (each of a different Rl -6 type, e.g. Rl, R2, etc.) projects to the
central cartridge, and all of these aim at the same location 4 mm in front of the fly eye
(figure 1.10, right). So each cartridge takes input from cells whose receptive fields
include a common area in space. The integration that takes place in the cartridges
underlies spatial localization and discrimination, but because it involves only Rl -6 cells,
cannot support color vision. The fact that the cells are responsive across the spectrum
indicates that the pathways downstream of it carry significant information about intensity.
The third point is where the spectrally-biased pathways from R7 and R8 and the
luminance pathway from R 1-6 converge in the Tm5 and Tm9 cells in the medulla. Tm5
receives direct input from R7 and indirect inputs from Rl -6 via L3 and R7 via Dm8. Tm9
receives direct input from R8 and indirect inputs from Rl -6 via L3 and R7 via Dm8. The
R7 synapse onto Tm5 and R8 synapse onto TM9 are mediated by histamine and
inhibitory, but beyond this little is known about the polarity of the synapses onto the Tm
neurons. The Dm8 neurons are glutamatergic (Gao, et. al, 2008), but it is not known
whether glutamate is excitatory or inhibitory at the synapse with Tm5 or Tm9. Similarly,
there is no strong evidence regarding the identity of the transmitter released by the L3
cells or whether it excites or inhibits its targets (I. Meinerthagen, personal
35
Chapter 1 . Introduction
communication). So, in sum, we do not know whether the synapses from R7 and R8 onto
Tm5 and TM9 are opponent or non-opponent.1
The structure of the network indicates that luminance signals may exert influence
on the response. Rl -6 cells provide input to the the chromatic pathways via L3. If the R8
onto R7 synapse is functional, Tm5 will receive opponent input from R7 and luminance
input from Rl -6. If the R7 and Rl -6 inputs are not functionally segregated, the Tm5
neuron may resemble the simplified neuron with significant luminance input described
above (figure 1.9). If the luminance input is strong, it could overcome the bipartite
opponent input, resulting in the Tm5 response being univariant. On the other hand, Rl -6
do respond at lower light levels than R7/8, so it is possible that the two pathways may be
segregated by intensity even though they are coextensive anatomically. If the responses
of Rl -6 saturate at higher intensities, then TM5 may exhibit opponent characteristics at
higher intensities, while showing intensity-dependent responses at lower light levels.
In sum, the Drosophila phototaxic network has some of the features that one would
expect from a network that supports color vision but the lack of segregation between
luminance and chromatic channels would lead one to expect that phototaxic behavior is
also intensity-dependent under certain conditions.
The above diagram does not describe all of the anatomy that we have reason to think
exists. Null mutants in the ort locus, which codes for a histamine-gated chloride channel
expressed in Dm8, Tm5, Tm9, and Ll -3 cells still phototax weakly, which suggests that
there is an orMndependent pathway involved in the behavior (Gao, et. al, 2008).
36
Chapter 2. The Problem of Color Vision
37
Chapter 2. The Problem of Color Vision
Methods of Demonstrating Color Vision
In this section we consider three classic studies on color vision that convey
important lessons for how to approach the problem in flies. By following these guidelines
it may be possible to avoid some of the difficulties that have plagued previous attempts to
uncover the role of color vision in fly behavior.
The most famous demonstration of color vision in any organism may be Von
Frisch's experiment on honeybees (Von Frisch, 1971). To make his case Von Frisch
trained bees to distinguish between a colored card and a series of gray-scale cards at an
artificial foraging site. He placed a blue card along with 15 others, ranging from black to
white through gradations of gray, on a table in a 4x4 array, under a sheet of glass to block
olfactory cues. On the glass at the center of each card, he put a clear dish. The dish over
the blue card contained nectar. All others were empty. The bees would fly to the table and
feed off the dish on the blue card. Bees have good spatial memory and to ensure that they
were not associating the nectar with a particular location the position of the blue card was
changed periodically. If bees lacked color vision, the blue card would "appear" to them as
it would to us on a black and white television as one of 16 gray-scale squares. Because of
the number of cards used, Von Frisch could be fairly certain that the bees would confuse
the blue card with one of the others if they could only perceive differences in intensity.
The bees had no problem returning to the blue card and never confused it with the others.
He concluded that they could detect and respond to differences in wavelength.
Chapter 2. The Problem of Color Vision
38
By varying the intensity of the signals of different wavelengths, Von Frisch
illustrates one approach to showing that an organism sees in color. The test compared
signals of two different wavelength distributions, the relatively narrowband light from the
blue card and the relatively flat, "white light", reflected from the gray cards. Given that
the test was run during daylight, Von Frisch could be fairly sure that the light reflected
from the blue card would be intense enough to activate color vision, if bees had it.
Having chosen the lightness of the card to fall within the range of the flowers bees are
likely to encounter when foraging, he could use one card to represent a range of
ecologically significant intensities. What he actually varies are the intensities of the
distractors, which range from very low in the black card to significantly higher than the
level of the blue card in the white one. The test is whether flies can separate the
wavelength signal from the intensity noise. The fact that they can means that they satisfy
the criterion for color vision.
Von Frisch's method can be mimicked in flies. One needs to choose a signal of a
specific wavelength and a relatively high, but ecologically reasonable, intensity,
analogous to the blue card, and pair it against a range of signals of a different
wavelengths, whose intensities run significantly above and below those of the target. If
the animal can consistently choose the target, then we have reason to say it has color
vision. We will follow this approach in some of the experiments reported below.
Another way to separate wavelength from intensity is to control for the principal
effect of intensity on the subject, otherwise known as perceived "brightness". Brightness
is what primarily changes in your subjective experience as you turn up a variable
Chapter 2. The Problem of Color Vision
39
intensity source. It is that facet of experience that at its lowest level makes visual
experience an indeterminate gray with no part of the environment standing out and as it
increases allows objects and colors to be delineated before they become enveloped in a
blinding whiteness.
Human experiments in hetrochromaticflicker photometry illustrate how it is
possible to control for brightness. A subject is shown a bipartite disk, a different color on
each side (Le Grande, 1957). The hemi-disks are alternately turned on and off. If they
differ in perceived brightness, the subject will experience a flicker between the two
halves. The subject is able to reduce or increase the intensity of one side until the flicker
disappears. At this point of equiluminance brightness has been controlled for, and the
disks appear equally bright. If any perceivable difference remains, it will be in the
wavelength distribution, and anyone who can perceive this difference has color vision.
A key point to bear in mind here is that equalizing brightness does not mean
equalizing photon flux, the rate at which photons pass through the lens of the eye. Photon
flux is a physical and objective quantity, while brightness is a subjective characteristic of
the observer's experience. Because organisms are more sensitive to some wavelengths
than others, lights that put out the same number of photons/unit area may not appear
equally bright to a subject. At photopic light levels, humans are most sensitive to 555 nm,
where the luminance channel from the cones has its peak response. In dim light, we
detect 498 nm light most easily, peak point for the rods. These facts are captured by the
luminous efficiency curves, which plot sensitivity as a function of wavelength.
Chapter 2. The Problem of Color Vision
40
Human Scotopic (left) and Photopic (right) Luminance Efficiency Curves
1.4
1.2
>-»
U
.1
1
O
S=
LU
S 0.8
CB
^ 0.6
?
co
I 0.4
0.2
350
400
450
500
550
600
650
700
750
800
Wavelength (nm)
Figure 2.1: Luminous efficiency function of rods and cones. The graphs plot the
normalized luminosity of a monochromatic light as a function of wavelength under
scotopic conditions, in which only rods are active (left curve), and photopic conditions, in
which rods saturate and cones are active (three data sets, right curves, showing some
variance and noise). The peak of the scoptopic function is 498 nm, the peak of the
photopic curves are 555 nm. (Data are from LeGrande, 1957)
Chapter 2. The Problem of Color Vision
41
The rod curve is on the left of figure 2.1, while three estimates of the collective
peak of the cones are clustered together on the right. For rods photon fluxes that make
brightness matches result in stimuli that are indistinguishable. For cones a brightness
match for stimuli of wavelengths that differ by more than the wavelength discrimination
threshold leaves the subject with a perception of equally bright, but distinct, colors.
When the subject is human, it is relatively easy to control for brightness. When
non-humans are at issue, it becomes more difficult. Does it even makes sense to attribute
brightness to the 'experience' of non-humans? Do non-humans even have 'experiences' to
which brightness can be attributed? It is not clear that these questions can be answered
with confidence and the answer given may depend on the organism under discussion.
What is more feasible is to determine the role brightness plays in human behavior and ask
whether the same role exists in the behavior of non-humans. A positive answer might
justify using of the term in connection with situations in which an animal responds to
light stimuli. This is the approach taken by Jacob, et. al, (2001) as part of an effort to
controling for brightness in rats. The authors note that brightness is intrinsically
connected to an organism's capacity to detect additions of small amounts of light. The
brighter a light is the easier it is to tell when it has been added to an existing light. The
authors use sensitivity to increments of light of different wavelengths, a quantity that can
be estimated from an animal's behavior, as a way to estimate when two lights look
equally bright to a rat.
Rats possess only two cones, with peak sensitivities of 509 nm and 359 nm. Jacob's
group began with the hypothesis that rats have color vision and distinguish between UV
Chapter 2. The Problem of Color Vision
42
and green lights independently of their intensity. They sought to remove brightness from
the experimental situation by controlling for it in a process that takes place in five stages.
1 . Find a behavior that varies monotonically with intensity.
Rats were trained to make intensity discriminations in an arena in which three
lighted panels appeared on one wall, illuminated equally from the outside by a tungsten-
halogen lamp with relatively broadband output. Additional monochromatic light was
added to one panel, making it brighter, and rats were given a pellet if they touched the
wall below this panel. Once the rats could reliably choose the more intense light, the
authors confirmed that the response rates varied monotonically with the intensity
difference between the standout panel and the others.
2. Use the monotonie behavioral response to determine increment sensitivity to particular
wavelengths.
As with humans, rats are more sensitive to some wavelengths than others.
Increment sensitivity is the inverse of the increment detection threshold, the lowest
intensity at which additional light of a wavelength can be perceived. The authors used
measurements of the thresholds for detecting added light of various wavelengths from the
initial experiment to calculate the rat's increment sensitivity to added light.
3. Convert increment sensitivity to an estimate of perceived brightness and use this
estimate to establish provisional equal brightness levels.
Chapter 2. The Problem of Color Vision
43
The authors used increment sensitivity as an indirect indicator of brightness levels
for the rat. They assumed that if an organism is more sensitive to one wavelength over
another, lights ofthat wavelength will appear brighter to it than those of the other, when
the two have the same of photon flux. If sensitivity correlates with brightness, one can
use the ratio of the detection thresholds to calculate when two lights appear equally
bright. In general, a light that requires Y times as many photons to detect compared to
another will appear equally bright when it has Y times the photon flux of the other.
4. Validate estimate of equal brightness experimentally.
Evidence that increment sensitivity is a valid proxy for brightness comes from an
experiment in which estimates of the relative fluxes at which stimuli would appear
equally bright were tested in a discrimination study. The researchers hypothesized that
when possible rats use both wavelength and intensity to distinguish between sources and
that by using both they perform better than when using only wavelength. Rats were
trained to discriminate between a constant 510 nm source and sources of 380 nm and 510
nm of variable intensities. After extensive training their performance on the tests was V-
shaped (figure 2.2). It increased fairly consistently as the calculated difference in
perceived brightness increased. The bottom ofthe V occurred around the estimated point
of equal brightness. For the variable 510 nm source, choice dropped to chance levels,
since all potential cues were now gone. Rats could still identify the 380 nm source, but
the success rates were at a minimum.
Chapter 2. The Problem of Color Vision
44
100
90
80
70
Ö
8 60
50
40
30
-0.6
-0.4
__L
-0.2
0
0.2
0.4
Relative test liolit intensitv
0.6
Figure 2.2: Discrimination between lights drops to a minimum around estimated
point of equal brightness. The proportion of correct responses is plotted as a function of
the difference between estimated brightness levels. Lights were estimated to be equally
bright at the zero point on the x-axis. 510 nm source, dotted line with triangles; 380 nm
source, solid line. Discrimination between 510 nm sources drops to chance as the
estimated difference in brightness goes to zero. Rats retain the ability to discriminate
between 380 nm and 510 nm sources across all brightness levels, but discrimination
reaches a minimum as the intensities of the lights approach levels at which they are
estimated to appear equally bright. (Jacobs, et. al, 2001).
Chapter 2. The Problem of Color Vision
45
These results are consistent with the hypothesis that sensitivity provides a valid
estimate of subjective brightness.
5. use validated measure to control for brightness in color vision tests.
To determine how many wavelength regions (hues) rats can distinguish, the authors
performed an experiment in which they controlled for subjective brightness, while
successively reducing the distance in nanometers between the lights until the rats could
no longer distinguish between the sources. First, the rats were trained to discriminate
between a 510 nm light and a variable wavelength source of intensities calculated to be
equally bright.
Chapter 2. The Problem of Color Vision
U-H-*i,
400
*»?
•
· versus 370 nm
&
¿S versus 510 mn
450
46
500
Test wavelength (nm)
Figure 2.3: Rats distinguish lights of wavelengths above 400 nm from lights of
wavelengths below 400 nm. Rats were tested in discrimination trials in which a 370 nm
and a 510 nm source were paired against a variable wavelength source. The graph
depicts the proportion of correct responses as a function of the wavelength of the variable
source. Rats discriminate between the 370 nm source and the variable source, when the
wavelength of the latter is greater than 400 nm, but discrimination drops below chance at
wavelengths less than 400 nm. They discrimiante between the 510 nm source and the
variable wavelength source, when the wavelength of the latter is less than 400 nm, but
response rates fall below chance when the wavelength of the variable soure is above 400
nm (Jacobs, et. al, 2001).
Chapter 2. The Problem of Color Vision
47
Rats could discriminate between the 510 nm and 370-390 nm sources, but once the
variable wavelength reached 400 nm discrimination dropped to chance and remained
there for all visible wavlengths (figure 2.3). In mirror image experiments in which a 370
nm light was paired against lights of other wavelengths, flies could learn to distinguish
370 nm from wavelengths above 400 nm, but discrimination at or below 400 nm was
never greater than chance. The authors conclude that rats can distinguish two regions of
the spectrum UV (370-400 nm) and visible (400-600 nm). This experiment provides a
useful guide as to what one has to do to properly control for brightness in a non-human
organism. As we will below, not all of these steps have been taken in fly research.
While many papers focus on whether an organism's abilities satisfy the standard
definition of color vision, it is possible to make a less direct case by seeking evidence for
a chromatic opponent system. Menzel and Greggers (1985) take this approach when
asking whether color vision is at work in honeybee phototaxis. Honeybees phototax in the
course of normal foraging. When searching for a feeding place, they prefer dark areas
(are negatively phototaxic), and after arriving at a potential foraging site will run into
crevices or hollows to search for nectar or pollen. After feeding, they prefer lighted
regions (are positively phototaxic) and will seek out a bright exit, head for it, and fly out
into the open. The authors exploited these drives in setting up experiments into whether
an opponent system might underlie the behavior.
Given that opponency is a neural phenomenon, behavioral evidence for it must be
indirect. Phototaxis being a reflex, no learning is involved, and it is quite possible that the
circuitry connecting the simulus to the behavioral response is relatively short. If the
Chapter 2. The Problem of Color Vision
48
pathway is short, opponent activity might appear relatively untransformed in the
behavior. Different wavelengths may have different effects on behavior with some
enhancing phototaxis and some inhibiting it. So there could be signs of spectral
opponency in the behavior, giving one reason to think that there may be underlying
neural opponency and color vision. In what follows I will use the term 'behavioral
opponency' to describe contrasting behavioral responses to stimuli from different regions
of the spectrum.
The authors first looked for signs of behavioral opponency in the bees' responses to
monochromatic sources. Specifically, they tested for whether a single wavelength
activates an inhibitory response. They allowed bees to fly to a test chamber containing a
tube leading to a foraging station. The station contained a Y-maze with one dark branch
and the other marked by a source of variable intensity and wavelength. The authors found
that the fraction choosing the lighted arm increased consistently with intensity and
approached 100% at the highest intensities (figure 2.4). The responses were sigmoidal.
Thus, there was no sign of behavioral opponency in phototaxis to single monochromatic
sources.
Chapter 2. The Problem of Color Vision
49
100 P
Sí37nra
439
341
ml·
slope ?
1 .12
4OS" t. 56
43© : S.QQ
thresîiûia
S37pim:
83 -W
a
489: 1-18
S37: 130
¦ft
10 leg.
Figure 2.4: Bee phototaxis rates increase monotonically with intensity in tests of one
light against darkness. Bees made a choice between a lighted arm and a dark arm of Y-
maze. The graph plots the proportion exiting the lighted arm as a function of the photon
flux of the light for different wavelengths. Response rates increase with increasing
intensity and never decrease.
Chapter 2. The Problem of Color Vision
50
The second experiments looked for behavioral opponency in the bee response to
mixed stimuli. The assay paired a mixture of UV (341 nm) and visible light (409, 439,
489, 537 nm) on one side ofthe Y-exit with a green source lighting the other. The authors
calculated what effect to expect, based on previous unmixed trials, if the lights
contributed in a purely additive fashion. At the very least the response to the combined
source should be greater than the response to each of the components taken on their own
(Abney's law). In a minority of cases the results differed from what was expecred, but
there was no pattern of inhibition, systematically related to wavelength, as one would
expect if opponency were involved. The authors conclude that purely additive
mechanisms are at work in honeybee phototaxis and that color vision is not. (This
conclusion rests on the assumption that behavioral opponency is necessary for color
vision.)
For our purposes, the value of Menzel and Greggers' work is as an illustration of
how an animal behaves during phototaxis when it does not exhibit behavioral opponency.
As we will see below, flies behave quite differently both on tests of one light against
darkness and in color mixing experiments.
In summary, we have seen two ways of showing that intensity is irrelevant to
discrimination: Von Frisch's method of varying the intensity of the stimuli and showing
that discrimination based on wavelength persists and the Jacobs group's method of
controlling for the brightness of the stimuli on the basis of behavior. Menzel and
Greggers' work offers an illustration of behavior in phototaxis that is not opponent.
Chapter 2. The Problem of Color Vision
51
Color Vision Experiments in Flies
Let us now consider some of the experiments that have been conducted over the
last four decades into color vision in flies. The types of experiments that have been
performed reflect the dilemma described at the outset. Early experiments focusing on
innate phototaxis show the behavior to be highly dependent on intensity and so,
apparently, inconsistent with color vision. As a result, researchers have either attempted
to control for the effects of intensity on the phototaxic response through training or have
sought evidence for color vision in other fly behaviors. Later, more sophisticated,
phototaxis experiments that make use of color mixing suggest, however, that color vision
may be at work in innate (untrained) phototaxis.
In 1973 Schlumperli published one of the earliest studies in the literature on fly
color vision (Schlumperli, 1973). The experiment illustrates the fact that phototaxic
responses are intensity-dependent for a wide range of stimulus settings, but also provides
some evidence for color vision. His experimental apparatus was a simple two-armed
maze in which a choice chamber, accessible from a starting box, was illuminated through
arms on both sides. On the one side was a white light of variable intensity and on the
other a narrowband source whose spectral output could be varied from UV to red and
whose intensity could be controlled with neutral density filters. During a test, a fly in a
starting box was introduced into the choice chamber and allowed to move in the direction
of either source. After the choice was made, the fly was collected. The total number
going in each direction was determined and the proportions calculated.
Chapter 2. The Problem of Color Vision
m
*0©
OO y*
Í39/
S7B /
/
W. 5800
367
/
332
551
/
Sr^
367
3SS
/
/
/
/
/
y
/
/
?
/
/
709
095
Z
/
/
?
y
/
652
634
/
K*- N1-WOO
^
^
/*x
V
^
SiVJ?
40e/
s8F4
!
SlT
S/
/
/
x»
S
?
1
'?
52
^
/
¿
cu
/
>
^
seo
570
435
<^
367
406
/
/
N:5400
7
ÍL1L?,
/
311
/
J38?
939
/
534
504
^
B34
406
?. 50
? >100
. ? ,150
• ? :20?
504
/
SU
¿.
?:3950
430
533
57«
634
?
'°9 'rei >
?
Figure 2.5: Fly phototaxic response as a function of intensity in a white light versus
spectral light assay. Flies were given a choice between a white light of irradiance equal
to .05 erg/sec/cm2 (panel a), .1 erg/sec/cm2 (panel b), 1 erg/sec/ cm2 (panels c and d) and
a spectral light of different wavelengths (31 1 to 709 nm). Each point represents the
proportion of flies going toward the spectral source as a function ofthe intensity ofthat
source. A single run, plotted as a line, consists of a series of trials in which the white light
at a fixed intensity is paired against the spectral source set to a particular wavelength
(indicated next to the curve) at different intensities. In (a)-(c) the flies were light adapted;
in (d) they were dark adapted. Log Irea = 0 corresponds to an absolute intensity of 8.8
erg/cm2/sec ofthe spectral source (from Schlumperli, 1973).
Chapter 2. The Problem of Color Vision
53
Summarizing his data, Schlumperli computes the slopes of each run line along three
of its segments. The first slope is between the points at which 20% and 40% ofthe flies
go toward the spectral source, the second between 40% and 60%, and the third between
60% and 80%. The slopes in each of these groups were collected and fitted with separate
linear regression lines in (a)-(d) in figure 2.6, corresponding to the same panels in figure
2.5. For panels (b) and (c), the slopes decrease at all points along the curves as the
wavelengths ofthe spectral light increases. In (d) the slopes between 40% and 60% and
between 60% and 80% response rates decrease with increasing wavelength of the spectral
source, but the slope between 20% and 40% response rates increase. While the flies'
responses always increase with increasing intensity for a given wavelength, those based
on longer wavelengths do so more slowly than those based on shorter ones at all points in
the response under conditions in panels (b) and (c) and for the latter part of the response
curves under conditions in panel (d).
Chapter 2. The Problem of Color Vision
•
• I
21t
¡Br
54
30
î
ia-
ti
1ft-
?1
10
Oi-
ns
U
ta
Î.0
>
??- o
1V,
?>
OSr
>v
—j
»oo
¦
soo
• 05
,
eoo
il
too
.
too
soô"
Figure 2.6: Fly phototaxic responses depend less on intensity as the wavelength of
the variable source increases. Regression analysis of slope of run lines from figure 2.5.
Panels (a)-(d) plot the slope ofthe run lines in (a)-(d), respectively, in figure 2.5. Slopes
ofthe response rates as a function of the intensity of the spectral light are computed
between three pairs of points along the curves: between 20-40% (?), 40-60% (·) and 6080% (o) response rates. Points corresponding to the three slope cohorts in each panel are
fitted with individual linear regression lines. The slopes decrease with increasing
wavelength ofthe spectral source for all cohorts in (b) and (c) and for the 40-60% and the
60-80% cohorts in (d) (Schlumperli, 1973).
Chapter 2. The Problem of Color Vision
55
Schlumperli maintains that these results show that flies have color vision. His
argument relies on the following criterion: "spontaneous behavioral tests show color
vision if the results demonstrate wavelength-specific reactions which are inexplicable
with the properties of receptors alone" [p. 76]. The definition has two parts. The first
requires that an organism exhibit "wavelength-specific reactions"; the second, that these
reactions not be "explicable" on the basis of the properties of the "receptors alone". The
first part is similar to the standard definition of color vision, i.e. that an organism
demonstrate wavelength specific responses over a range of intensities. The second part is
new. Schlumperli does not explain why he does not use the standard definition nor does
he offer any justification for the criterion he does use. He also does not elaborate on the
second condition. As we will see below, it is possible that the requirement that the
response not be "explicable" on the basis ofproperties of "receptors alone" is referring to
the principle ofunivariance, the suggestion being that ifthe behavior violates
univariance, then the organism has color vision.
To begin let us examine the data according to the standard definition. Does the
behavior exhibited in figure 2.5, and summarized in figure 2.6, demonstrate that flies
have color vision, that they distinguish between white light and light of different
wavelengths independently of intensity? To answer this question we need to determine
what to expect from the assay if flies have color vision. What does the hypothesis of
color vision imply about the results of two-choice phototaxic experiments?
Schlumperli's experiment bears some resemblance to Von Frisch's. Von Frish fixed
the intensity ofthe signal from the specific wavelength source (the blue card) and varied
the intensity of broadband source (the gray-scale cards). Bees went toward the blue card
Chapter 2. The Problem of Color Vision
56
regardless ofthe intensity of the gray cards. For a given experiment, Schlumperli fixes
the intensity of the broadband source (white light) and varies the intensities of the
specific wavelengths. To have color vision the flies must distinguish between the
broadband and specific sources independently ofintensity. At 50-50 choice rates the flies
are clearly not distinguishing between the two sources. Thus, to distinguish between two
sources flies must go toward one at a rate that is significantly greater than 50% for a
significant range of intermediate intensities. By this standard Schlumperli's experiments
do not show that flies have color vision. With the exception of the longest wavelengths in
(a) and (b), each ofthe runs crosses the 50% choice rate at some intensity. Flies go from
distinguishing between the two sources to not distinguishing between them as the
intensity rises. Even the two lines that do not cross the 50-50 threshold fail to show
evidence of color vision because the trend is toward the 50% line, and the curves might
have crossed it if the experiment were continued to higher intensity levels.
Schlumperli's data do not show that flies have color vision but they may provide
some evidence against the hypothesis that flies are responding purely to the intensity of
the stimuli. If there are no chromatic opponent mechanisms in the fly, the receptors will
combine additively in an integrator. Such an integrator will be univariant, having the
same type of response regardless ofthe wavelength ofthe impinging photons (Rl -6
treated as one cell, figure 2.7).
Chapter 2. The Problem of Color Vision
57
Activation Level of Integrator on Linear Model with No Opponency
i 0.6
442
466
Wavelength (nm)
Figure 2.7: Spectral sensitivity of simple model additive integrator. The level ofthe
model cell's response is plotted as function of the wavelength of the stimulating light.
The cell is more sensitive to certain wavelengths than others as indicated by the higher
relative activation for those wavelengths. Vertical arrows represent effects of stimulating
lights of wavelengths corresponding to the peak sensitivities of the opsins in the fly eye.
Chapter 2. The Problem of Color Vision
58
If an integrator ofthis kind drove the phototaxic behavior, it would produce
responses that have the same slope for all wavelengths. If phototaxis rate is a sigmoid
function of the number of photons absorbed (the photon flux of the light hitting the fly
eye times the sensitivity ofthe integrator), the response rates to different wavelengths
will form parallel curves displaced along the X-axis (figure 2.8). The behavioral response
will be more sensitive to wavelengths to which the integrator is more sensitive, but once
the response reaches the same level equivalent log increases in intensity result in the
equal increases in relative activation. These curves all have the same slope when they
reach a given height. When they cross the line, say PI=5, they are all rising at the same
rate.
Chapter 2. The Problem of Color Vision
59
Behavioral Response on a Purely Additive Model of Receptor Activation
1
331 nm
355 ? m
0.9
442 ? m
486 ? m
515 nm
0.8
0.7
I 0.6
?
•è? 0.5
?
'S
0.4
Q0.3
0.2
0.1
0
2
4
6
8
10
12
Log Photon Flux
Figure 2.8: Behavioral response to different wavelength stimuli form parallel curves
if based on purely additive integration. The preference of the flies for the illuminated
side is plotted as a function of the photon flux of lights of different wavelengths. Each
curve models the response of the integrator to a light of a given wavelength at different
flux levels. The responses are sigmoidal, increasing monotonically with increasing
photon flux. Curves corresponding to wavelengths to which the integrator is more
sensitive are shifted towards the origin on the x-axis. Curves differ only in their degree of
translation along the x-axis. The phototaxis rate given by PI = 2x((l+e~p*sen)"1-.5), where ?
is the photon flux and sen is the response of the integrator to a particular wavelength of
the light. The threshold response is set arbitrarily to 6 log units.
Chapter 2. The Problem of Color Vision
60
Now what Schlumperli observes is that slopes of his curves at particular response
levels decrease as the wavelengths of the spectral light increases under certain conditions.
For UV the ascent is rapid, for visible wavelengths less and less so. The response
depends less on intensity as the wavelength ofthe spectral light increases. This is not
what takes place under the simple univariant model in figure 2.8. To the extent that the
model captures general features of phototaxis based on additive and linear mechanisms,
his experiments provide evidence for the existence of subtractive mechanisms or nonlinear computation prior to the integrator. This is not color vision, but it may be an
important part of it.
Schlumperli's tests pair two sources against each other. The responses he records
are clearly dependent on intensity in a way that makes clear that flies are not
distinguishing between the stimuli independently ofintensity, as color vision requires.
There have been two responses to the intensity-dependence of innate phototaxis.
Researchers have attempted to train flies in phototaxis in an effort to produce results that
do not depend on intensity, and they have looked for non-phototaxic assays that might
yield intensity-independent behavior. Up to now, neither approach has yielded
reproducible successes. Let us consider two experiments using non-phototaxic
paradigms.
Tang and Guo (2001) argued that flies have color vision based on operant
conditioning experiments in a flight simulator. In the assay, flies are tethered by their
thorax to an apparatus that allows movement in the yaw (horizontal) plane and that
measures the torque at the attachment point. When a torque is perceived, the simulator
Chapter 2. The Problem of Color Vision
61
rotates the fly in the same direction by an amount, and at a rate, calculated to simulate
actual rotation in flight. The flies "hovered" inside a cylindrical arena on which four T's,
two green and two blue, were placed at 90° intervals at the same height, green alternating
with blue. In the pre-training period, the authors tested the flies for innate preferences for
green or blue, but none emerged. When allowed to orient freely, they oriented at
statistically equivalent rates toward each color. In the assay, the flies were trained to
orient toward green Ts and away from blue ones. Heat was used as a punishment, and
flies were heated up whenever they faced a blue but not a green T.
Training took place in a pair of two-minute blocks separated by testing blocks of
two minutes with three testing blocks following the last training period. The authors
reported that after training the flies oriented toward the green Ts over the blue ones by
statistically significant margins. This result suggests that the flies can distinguish green
from blue but leaves open whether they might be doing so on the basis of effective
intensity (perceived "brightness") of the stimuli rather than their wavelengths. To rule out
intensity as a cue, they trained flies using the same paradigm to orient toward light blue
Ts and away from dark green ones. When tested with color and intensity reversed, i.e.
dark blue and light green, they preferred the dark blue Ts, suggesting that when trained
with both cues, they had learned to discriminate on the basis of wavelength rather than
intensity. Since this occurs when the intensities are markedly different, it is reasonable to
infer that it occurs when the intensities are comparable. These results suggest that flies
have color vision and can learn wavelength-temperature associations. This study has one
major problem: efforts to replicate it have failed (C. Desplan, personal communication).
Chapter 2. The Problem of Color Vision
62
Studies with the optomotor response (OR) have produced strong negative results.
The OR is an orientation behavior in which a fly moves in the yaw plane in a direction
opposite that of vertical bars moving in its visual field. The response requires a
luminance contrast and weakens before disappearing when this contrast goes to zero.
Because flies lacking functional Rl -6 cells do not respond to gray-scale bars moving in
either direction, it has long been thought that the response was mediated solely by these
cells. However, this question has not been investigated in detail. Yamaguchi and
colleagues (2008) looked into whether R7/8 might contribute to the OR when there is a
luminance contrast between colored stripes. They ran experiments using blue and green
bars of varying luminance levels. Mutants lacking functional R8 and R7 cells (sev; Rh5;
Rh6) behaved like wild types in all experiments, indicating that these receptors do not
contribute to the OR response, even when the bars are colored. The OR appears to be a
single receptor behavior in which color vision plays no role.
Let us turn now to some of the efforts to using training in phototaxis to eliminate
the dependence on intensity from the behavior. Quinn and Benzer (1974) were some of
the first authors to attempt to use training in phototaxis to control for the effects of the
intensity of the stimuli in an assay designed to look for evidence of color vision. They
studied phototaxis in a Y-maze illuminated by different wavelengths of light on each arm.
They assumed that innate phototaxis was a readout of brightness, i.e. when presented
with a choice between two sources, the flies would go toward whichever appeared
brighter. The experiment had three phases: a control period in which the light levels were
set, through trial and error, so that the flies split equally between the two sources, which
the authors took as a sign that the lights were equally effective and, therefore, that
Chapter 2. The Problem of Color Vision
63
brightness had been controlled for; a training phase in which the flies learned to associate
one color of light with a punishment, while the other was unpunished; and a test in which
they were exposed to the two lights without punishment and allowed to choose freely.
The reasoning was that if the flies learned to discriminate between equally bright lights
through training, they could not have done so on the basis of brightness/intensity and
must have color vision.
The Y-maze came apart into a holding pen consisting of a test tube covered with
black tape (a |) and a small base followed by a bifurcation (a V). One arm of the V was
illuminated by 610 nm ('red') light, while 450 nm ('blue') shined on the other. Between 40
and 100 flies were used on each trial. During the control period, the flies were placed in
the holding tube, which was plugged from each side with a plastic stopper. After 60
seconds, one stopper was removed and the open end was attached to the vertex of the V.
Flies were allowed to run into both arms and, after 30 seconds, the proportion going
toward each side was determined.
On successive trials the light intensities were adjusted until half of the flies went
into each of the arms, the point of supposed 'equal brightness'. Having controlled for
brightness, the authors trained the flies using punishment in an operant conditioning
paradigm (incorrectly referred to in the article as "negative reinforement"). One arm of
the maze was coated with quinine sulfate, which has an extremely bitter taste for humans
and which flies avoid even at low concentrations. During training runs, flies were placed
in the covered tube, held for 60 seconds, and released into the maze and allowed to
phototax down the arms for 30 seconds, after which they were shaken back down into the
Chapter 2. The Problem of Color Vision
64
tube and given 60 seconds of rest. This process was repeated twice. Finally, the maze was
cleared of quinine and the flies were tested. The tube containing the flies was reconnected
to the maze and the number going toward each side was determined. To control for any
residual bias, the flies were also trained to avoid the other wavelength and tested on that
basis.
The authors report that the flies avoided the color associated with quinine, going to
the other side 55% to 45%, a statistically significant difference under a là test. Because
of the precautions taken during the control phase, the authors ruled out "brightness" as a
cue along with other sources of error such as left-right bias, deducing that the
discriminations were made purely on the basis of the wavelengths of the lights. From this
they inferred that flies can learn to associate wavelengths with gustatory stimuli and
concluded that flies have color vision. We tried to replicate these results but were
unsuccessful. After control trials and training, flies continued to choose between the
sources at rates that did not differ significantly from chance. No replications or successes
at similar training paradigms have been reported in the literature.
One of the most sophisticated cases for color vision in flies was made by Menne
and Spatz (1977). In their article, the authors set out to train flies to discriminate between
440 nm ('blue') and 590 nm ('yellow') light, while ensuring that the learned association
was independent of intensity. The assay also had three parts. First, the untrained
responses of the flies was measured during a control period. 100 flies were placed in the
training apparatus, consisting of two culture tubes connected at their open ends and
attached to a vortex. One of the tubes was illuminated with blue light, the other with
Chapter 2. The Problem of Color Vision
65
yellow light. The apparatus was tipped to the left and the vortex activated for one second
to accumulate all of the flies on that side. Afterward, it was set back on the horizontal,
and the same sides illuminated with blue and yellow. After two minutes a photograph
was taken and the distribution of flies determined. The procedure was repeated but with
tipping to the right. The two distributions were then averaged with the mean serving as
the baseline relative to which training effects were assessed.
To train flies to avoid a color the authors used classical conditioning to associate a
punishment (shaking) with that color of light. The other color was not associated with
punishment. Flies were trained in the same apparatus used in the control phase. In one
type of training session, blue light shined on both culture tubes, while the vortex was
turned on for two seconds and then off for three followed by a five second period in
which flies are exposed to yellow without shaking. The process was repeated six times.
Testing took place immediately after training. The test protocol was the same as in the
control period. Flies were shifted to the left, shaken, and returned to the horizontal. Blue
and yellow light was shined on different sides, and after two minutes the flies were
photographed and counted. They were then shifted to the right and the procedure was
repeated. The average of the two test distributions was calculated. Finally, the proportion
moving toward the unshaken light in the test was subtracted from the number moving
toward that color during the baseline to produce a measure of learning.
To control for the effect of the direction of shifting during control and test periods,
the authors randomized the sequence so that the initial shift was to the left on half of the
trials and to the right on the other half. The position of the lights was also randomized;
Chapter 2. The Problem of Color Vision
66
blue illuminated the left tube and yellow the right in half the control and half the test
trials, and the reverse in the other half. Care was taken so that there was no correlation
between the position of the light during training and during testing. To ensure that
learning was not cued to 'brightness', the authors varied the intensity of the training lights
over six orders of magnitude while keeping the intensity of the test lights constant.
Across all intensity levels, color positions and shift directions, they report statistically
significant differences between the distribution of flies before and after training. In each
case, the punishment regime reduced the preference for the associated light.
We tried to replicate these results using an identical punishment regime. After
training, the flies in our experiments continued to respond to the lights in ways that were
not distinguishable from their responses during the control period. Others report failed
attempts at replication (C. Desplan, personal communication.) No other successful
reports of replication have appeared in the literature, and no positive results have
appeared using similar paradigms. The lack repoducibility of efforts to train the
phototaxic response was a significant factor in our decision to focus on the innate
response in the studies reported in chapter four.
Quinn and Benzer's (1974) attempt to control for brightness is typical of what has
appeared in the literature. The authors assume that any two simuli that cause flies to split
50-50 appear are equally effective in a manner relevant to controlling for brightness.
They do not explain why they make this assumption, but they may have been led to it by
results like Schlumperli's, which appear to demonstrate that the phototaxic responses in
two-light assays with wild type flies vary monotonically with changes in intensity.
Chapter 2. The Problem of Color Vision
67
This assumption also plays a role in the work of Hernandez de Salomon and Spatz
(1983), who used an apparatus and training regiment identical to Menne and Spatz (1977)
but a different approach to demonstrating that training effects are not dependent on
intensity. In addition to varying the irradiance levels during training, Hernandez de
Salomon and Spatz sought to control for the brightness of the stimuli. Like Quinn and
Benzer, they assumed that stimuli resulting in a 50-50 split in innate response must be
equally bright to the flies. Each experiment used lights of intensities that led half the flies
to migrate toward the tube lit with blue and half toward the tube lit with yellow, thus
neutralizing their supposed brightness. Training began at this point. As with Menne and
Spatz (1977) it involved six pairs of five second periods in which shaking was associated
with one light and no shaking with the other. The preference of the flies for blue and
yellow was then tested as before, and the change from the baseline 50-50 response was
determined. The authors report that training produced statistically significant decreases in
preference for the color associated with shaking for a range of training intensities.
Significantly more flies went toward the light that was not associated with shaking. These
results have not been reproducible.
There are two main problems with assuming that phototaxis is a readout of
brightness. The minor problem is that it is not something that can be assumed but an
hypothesis that must be justified. As we saw, the Jacobs group's behavioral test for equal
brightness in rats rests on the intuitively plausible connection between brightness and
increment detection thresholds. The literature on fly phototaxis does not follow this
approach and indeed makes no explicit arguments on the topic. As noted, the conclusion
appears to rest on the fact that when flies are given a choice between two lights in a
Chapter 2. The Problem of Color Vision
68
phototaxic test, increasing the intensity of one light never decreases its attractiveness.
When the flies split 50-50 between the two lights, the stimuli are obviously equally
effective in some sense, but it is not clear why one should say that the effectivness is
characteristic of brightness.
The major problem, as we will see below in Heisenberg and Buchner's (1977) and
Fischbach's (1979) studies, is that, two-light tests aside, innate phototaxis rates do not in
general vary monotonically with intensity. How flies respond to added light depends on
which wavelengths of light are being added in a way that is not consistent with the
response giving information about 'brightness'. In other words, the central premise of the
argument these authors need to give is false. Brightness cannot be controlled for using
phototaxis alone.
Let us turn to the more indirect route and approach color vision through opponency.
The same reasoning that made opponency relevant to color vision in bee phototaxis also
applies to flies: i) opponency is a defining feature of color circuits; ii) mixing
experiments can provide a behavioral counterpart to receptor opponency by showing that
one light enhances the response, while another suppresses it; iii) because phototaxis is a
reflex, the circuit may be short, and behavioral opponency can provide indirect evidence
of neural opponency. As it turns out, there is evidence for behavioral opponency in the
literature on fly phototaxic behavior.
Six years before Menzel and Greggers (1985) performed their experiments,
Fischbach (1979) performed one-light experiments in a fly-balance in which the position
of a fly could be monitored over time, and where the light was switched periodically from
Chapter 2. The Problem of Color Vision
69
end of the tube to the other. When exposed to light of 575 nm, the flies went toward the
light at increasing rates for low intensities, but as the intensities reach the middle range
(101105 quanta/mm2/sec) the response rates began to decline and eventually the response
became Photophobie. In contrast, the rate at which flies went toward UV increased with
intensity and did not show a downward trend even at the highest intensities.
Gao, et. al, (2008) also includes one-light assays in which stimuli were UV (370
nm) and green (525 nm) light of different intensities. The response curves for UV stimuli
were S-shaped; the green curves were not. The proportion of flies going toward the light
increased as the intensity of UV increased and, after reaching a plateau, leveled off, never
decreasing significantly. In contrast, the green curves rose to a peak before decreasing.
This is evidence of inhibition in the response to green wavelengths of higher intensities
that was absent for UV light. The results contrast with what Menzel and Greggers (1985)
found to be the case with one-light bee phototaxis, where there was no inhibition in
response to UV or visible light.
Heisenberg and Büchner (1977) carried out color-mixing experiments using 'slow'
phototaxis with flies. High intensity 560 nm ('green') light (??7 ? higher than the
phototaxic threshold) was shined on one side of the T-maze as a baseline. 360 nm (UV),
same energy on both sides, was added, starting at a low intensity and increasing by 1 Ox at
each step. After each intensity increment, flies were given time to settle into an
equilibrium—this is what made it slow—and the number going toward each side was
counted. The intensity of the bilateral UV was increased up to a maximum value
equivalent to the green photon flux. Flies of three genotypes were used: wild type, RhI-
Chapter 2. The Problem of Color Vision
mutant {ora, NinaE), and sevenless, which lack R7 cells and in which all R8s express
Rh6.
70
Chapter 2. The Problem of Color Vision
71
sev
**
«
ora
1
WT
1
ft
?
10* W* 10* 10"· IO"2 10
I1n, ire!, units!
1
»-»
Figure 2.9: Wild type flies prefer pure UV to UV + green at higher intensities. Flies
were exposed to high intensity green (??7 ? detection threshold) to which bilateral UV
was added at increasing intensities in a 'slow' phototaxis assay. Graphs plot preference
index (difference between the number going toward UV+green minus the number going
toward pure UV divided by the total number) as a function of the relative photon flux of
UV versus green (UV flux/green flux) for three different genotypes. Preference index
(PI) is positive when more flies go toward the combined UV+green source and negative
when more go toward the pure UV source. Sevenless (sev) and i?/*/-mutants (ora) always
prefer the combined source to the pure UV. Wild types (WT) prefer the combined source
when UV intensity is low but switch to the pure UV source as the UV intensity increases
(Heisenberg and Büchner, 1977).
Chapter 2. The Problem of Color Vision
72
When exposed to bright green alone at the start of the experiment, flies of all
genotypes moved toward the light at a rate greater than chance (figure 2.9). Flies of all
genotypes also phototax toward UV at any detectable intensity when paired against dark.
If the response were purely additive, as it was with bees, the appeal of a UV+green
source should always exceed that of a pure UV source. But this is not what the authors
found for wild types. At low intensities, added bi-lateral UV had no effect; all flies went
toward the same side (now UV + green) at the same rate as green alone. Once the
intensity ofthe UV reached ??"4 ? the flux of green, wild type flies began to move toward
the side lit only by UV. The preference for pure UV over the combined source reached a
ceiling at a UV intensity of approximately 1 log unit below the green flux. In contrast,
sevenless flies showed no change in behavior with increasing UV, while RhI-mutants
exhibited only a slight, statistically insignificant, reduction in their preference for
green+UV.
Because flies lacking functional Rl -6 or R7 cells did not show a change in
preference, the authors conclude that Rl -6 and R7 were involved in driving the
preference for pure UV at high intensities. Beyond this, it is difficult to draw firm
conclusions. The problem stems from the complexity of the experimental manipulation.
When UV light was added to both sides, two features of the situation were changed,
making it hard to tell which caused the change in the flies' behavior. The change in the
wild type behavior with the addition of UV could have resulted from a decrease in the
appeal of the UV+green side or an increase in the appeal of the pure UV side or both.
Chapter 2. The Problem of Color Vision
73
On a receptor level the picture is also murky. If R7 is dominant in phototaxis
(Stark, et. al, 1976; Hu and Stark, 1977), the relative appeal of pure UV could come from
an inhibition of R7 by Rl -6 or yellow R8 (peak 515 nm) or both. The relative appeal of
the pure source could also come from the combined effect of strong activation of all
receptors by the UV+green side, perhaps triggering global inhibition that does not occur
when receptors are activated more selectively. These are possible explanations, but one
cannot choose between them or discount others on the basis of the given data. It is clear
that an assay in which UV is added to only one side of the maze would yield results that
are more easily interpreted. Still, the data is suggestive and contrasts with Menzel and
Gregger's result, that mixing is purely additive in bees.
Fischbach (1979) performed an experiment with the fly-balance using using a
bilateral UV background and added light of fixed higher intensity on one side, ranging in
wavelength from 350 nm to 650 nm.
Chapter 2. The Problem of Color Vision
¿
?
? ?
m
?
m
St
m
¿ ?
?
» 1
P
h
G
J
«
IP»
1
\
74
A
\
N
I
I
i
I ' *>
ß
4
d
?
<
?
yy \
&
? j
4ÓQ
SOO
**
ßd?
SPECTRAL WAVELENGTH |mn] ——
Figure 2.10: Wild type flies have opposed responses to lights of greater and less than
414 nm at certain intensities against a bilateral UV baseline. Flies were given a
choice between a UV source and combined UV + variable wavelength source in a fly
balance assay. Graphs plot relative attractiveness as a function of the wavelength of the
added source. Positive deflection is toward the combined source. Wild types (o) went
toward the combined source when the wavelength of the added source was less than 414
nm and toward the pure source when the wavelength of the added source was greater than
414 nm. RhI -mutants (·) and sevenless (?) went toward the combined source for all
added wavelengths (Fischbach, 1979).
Chapter 2. The Problem of Color Vision
75
When light of less than 414 nm was added, the flies went toward the combined
source, but when the added light was of longer wavelength, the flies preferred the pure
UV side (figure 2.10). i?/?./ -mutant and sevenless flies went toward the combined source
for all wavelengths. (As we will see below, there is reason to think that the i?/z/ -mutants
used in Heisenberg and Buchner's and Fischbach's studies did not have null mutations at
the gene locus.) Fischbach concluded that flies have color vision and distinguish UV
and visible light as different hues. The conclusion is not warranted. The results of the
experiment are consistent with the existence of UV-visible opponency, but do not make a
strong case for opponency because they only record the responses for each wavelenth at a
single intensity. To make the case for a UV-visible opponent system Fischbach should
have varied the intensity of the stimuli. And the results do not demostrate color vision
because they do not show that flies prefer UV to visible light in an intensity-independent
manner. Doing so would require a test of added UV versus added visible light at a range
of intensities. Still, Heisenberg and Buchner (1977)'s and Fischbach (1979)'s results are
highly suggestive and consistent with UV-visible color vision in flies.
Returning to the topic of controlling for brightness using phototaxis, both papers
demonstrate that innate phototaxis rates do not always increase with increases in the
intensity of the source. Adding bright visible light of certain intensities decreases
attraction to a side when the maze is illuminated by bilateral UV. Such results undermine
the most general form of a key assumption—that the phototaxis rate is a readout of
brightness—on which Quinn and Benzer (1974)'s and Hernandez de Solomon and Spatz's
(1983) efforts to control for brightness rest.
Chapter 2. The Problem of Color Vision
76
The preceeding papers make it plausible that a UV-visible opponent system might
be involved in phototaxis, though the exact location of the opponency in the phototaxic
network is uncertain. Another view is present in a model of phototaxis proposed in Spatz
and Jacob (1977). Under this model, activation of all receptors (Rl -8) can enhance
phototaxis, but activity in R8 also inhibits activity in Rl -6 (figure 2.1).
Chapter 2. The Problem of Color Vision
'R7
?
R1-R6
R8
lnhib.
Figure 2.11: Spatz/Jacob model of phototaxis. Activation of all three receptor types
(Rl -6, R7, R8) combine additively to drive phototaxis. Activation of R8 reduces the
influence of Rl -6.
77
Chapter 2. The Problem of Color Vision
78
At the time the model was the developed, the division between pale and yellow had
yet to be made and R7 and R8 were each thought to consist of a single cell type. R7 was
thought to have its peak around 360 nm and R8 around 520 nm. Rl -6 was known to have
one peak around 360 nm and another in the 470-490 nm range. The model in figure 2.1 1
implies that wavelengths across the spectrum will attract flies, but that long wavelength
activation will also reduce the effectiveness of stimuli across the spectrum through its
effect on the broadband Rl -6 cells.
Though the authors do not explain which pieces of data motivate specific aspects of
their model, some connections seem obvious. It seems that data from Schlumperli's and
Heisenberg and Buchner's papers may be behind idea that R8 inhibits Rl -6. Schlumperli
(1973) showed that the slopes of curves that plot the phototaxis rate toward a spectral
light versus white as a function of the spectral light's intensity decrease as the spectral
wavelengths get longer under certain conditions. One might explain this by assuming that
the longest wavelength receptor provides inhibitory input. Stimuli of longer wavelengths
will activate this receptor to a greater degree than shorter wavelength stimuli, which
could drive down slopes at longer wavelengths. Heisenberg and Buchner (1977) showed
that high intensity 560 nm light, by itself, drove phototaxis at around 60%, but their
mixing data can be read as suggesting that it may have an inhibitory effect when
combined with high intensity UV. Jacob and Spatz may have interpreted these results as
indicating that R8, the long wavelength opsin, drives phototaxis on its own but also
inhibits the UV sensitive Rl -6 cells.
Chapter 2. The Problem of Color Vision
79
Gao, et. al, (2008) recognize that the phototaxic response of wild type flies depends
on both the wavelength and intensity of the stimulus. They label the phenomenon
exhibited in phototaxis "spectral preference" to distinguish it from color vision (Hu and
Stark, 1977). Central to the article are the results of two-light, two-sided, phototaxic tests
in which flies are exposed to a constant green source on one side and a variable intensity
UV source on the other. The authors observe that as the intensity of the UV is raised from
a low to high levels the proportion of flies going toward UV increases in wild type and
many mutant strains. Wild type flies have a significant preference for UV over green
light of the same photon flux.
This paper was the first to examine the effects of the second-order neurons in
phototaxis. The authors conditionally knocked-out the neurons using the GaU-IJAS
system applied to the shibirits (shits) protein under control ofthe ort gene promoter. Shits,
a temperature-sensitive mutant of dynamin, is a protein essential for vesicle release. Gal4
is the yeast transcription activator protein and UAS is the upstream activator, a short
segment of the promoter to which Gal4 binds to activate gene transcription. When placed
under control of a promoter, Gal4 is expressed only in the target neurons, where it
activates the UAS, leading to the transcription of shibire's mRNA and expression of the
temperature sensitive protein product. Below 33°c the dynamin produced by the mRNA
functions normally, but above this temperature its conformation makes it unstable. The
UAS-GaU transgenes over-produce the mutant protein, swamping the endogenous form
and, according to the theory, effectively shutting down synaptic release at the higher
temperatures. The ort gene codes for a histamine-gated chloride channel (Gengs, et. al,
Chapter 2. The Problem of Color Vision
80
2002), and its promoter drives expression in L 1-3 cells in the lamina along with a range
of medullar cells, including the Dm8, Tm2, Tm5a-c, Tm9 and Tm20 lines.
The authors targeted cells downstream of R7 and Rl -6. The flies had the genotype
UAS-shi'V+;+; UAS-shits/orf2-Gal4, where orf2 is a fragment of ort promoter that gives
rise to expression in Dm8 and L1-3 cells. When the activity of cells under control of orf2
are diabled, using shits at temperatures above 330C, the preference for UV over green is
reduced compared to wild types. The UV preference returns to wild type levels at lower
temperatures. In null mutants ofthe ort gene {ort1), phototaxis rates and UV preference
are significantly reduced. Rescuing ort using the orf2 promoter in the ort-mx\\ mutant
background returns phototaxis rates and UV preference to wild type levels. On the basis
of published reports that Ll and L2 are neither necessary nor sufficient for UV preference
(Rister, et. al, 2007), the authors conclude that Dm8 neurons are necessary and sufficient
for UV preference in phototaxis.1
In parellel experiments, the authors conditionally knocked-out and rescued cells
from another subset of or/-expressing neurons. The C3 fragment of the ort promoter
controls expression in L2, Tm2, Tm9, C2, and Mil neurons. Knocking out these cells
using shits at restrictive temperatures did not affect UV preference, though restoring ort
activity to null mutants under the promoter's control further decreased the UV preference
This inference is invalid because orf2 also expresses in L3 neurons, and the experiments
do not rule out a role for L3 in UV preference. The authors report that the response of
NinaE17/NinaEG128R ('NinaE', in the paper)—the null allele of RhI, the opsin in Rl-6
cells which feed L3, over a hypomorph—do not differ from wild type. However, they do
not report the response of the null mutant (NineE17) on its own, which would be needed
to establish that L3 has no effect on UV preference.
Chapter 2. The Problem of Color Vision
81
relative to wild types. The authors conclude that cells under this fragment's control are
sufficient, but not necessary, for green preference.
To summarize, this survey of the literature leads us to a number of conclusions. The
difficulty of training flies to specific wavelengths in phototaxis suggests that the most
productive way to study color vision in phototaxis may be to focus on the innate
response. Second, in the absense of a convincing method of controlling for brightness, it
appears that the best approach to separating responses to wavelength from intensity is to
vary the intensity of the stimuli. Finally, there is evidence supporting the hypothesis that
flies possess an opponent system that contrasts UV and visible light.
Chapter 3. Materials and Methods
82
Chapter 3. Materials and Methods
Fly StocL·: Wild type flies used for experiments were Canton S from the Bloomington
stock center. Fly stocks were grown in culture bottles, at a density of approximately 200
flies/bottle, on standard cornmeal medium, and maintained in an incubator at 250C on a
12/12 light-dark cycle. Bottles were changed every other day. After eclosión, flies were
kept in the incubator for 3-9 days.
Behavioral Equipment
Experiments were conducted using a T-maze or a Y-maze paradigm (Figure 3.1).
The T-maze is composed of three slabs of plexiglass (22.5x5.6x2. lem) placed upright
together with the outer two connected by a metal attachment on two sides. The outer two
slabs are attached by screws on their bottom side to a fourth slab that serves as a base. A
2.7 cm diameter hole, drilled 3/4ths up the height one of the outer slabs, serves as the
entry point. A hole of the same diameter drilled l/3rd up the height of the central slab
serves as an elevator to receive the flies through the entry hole when the middle slab is
raised. Lowering the middle slab traps the flies in the elavator between the two outer
slabs. 2.7 cm holes drilled l/3rd up the height of the outer two slabs serve as exit holes.
When the middle slab is lowered to the base, the flies can move freely in either direction.
The exit arms of the T-maze consist of standard fly culture tubes (22 mm inner diameter)
made of clear UV transmitting plastic with tape wrapped around the outside near the top
to bridge the gap between the culture tube outer diameter and the diameter of the exit
hole.
Chapter 3. Materials and Methods
?» »? ni iwuli »JJWJIUIWWgHP»
A
I ¦:
83
Chapter 3. Materials and Methods
84
Limht
T-maze
LmM
bumrœ
Figure 3.1. Experimental setup and T-maze. (a) T-maze, (b) experimental set up, (c)
schematic of experimental set up. (C) Light sources (orange boxes) were connected to
monochromators (black rectangles) by fiber-optic cables (black lines). Output beams
passed through neutral density filters (perpendicular lines) and plano-convex lens
(crescents) to collimate the beam. For source mixing, beams were combined by means of
a beam splitter (diagonal segment, right). The two beams hit opposing exit slits of the Tmaze (cross at center of black rectangle, center). For 1- or 2-light experiments, only the
upper light source was used and, for 1 -light experiments, the exit Slit of one of the
monochromators was blocked.
Chapter 3. Materials and Methods
85
The Y-maze consists of four parts: a plexiglass base to which two exit arms,
containing openable entry slots are attached, and a 2.54 cm diameter syringe with a cutoff
top. The syringe is used to deliver flies to a choice point in the plexiglass base (figure
3.2).
Light sources: The light for experiments requiring one or two signals was a 175 W source
(ASB-XE-1 75EX, UV/NIR Extended, Spectral Products). Two signals are produced by
splitting the output of the light with a bifurcated fiber-optic cable (Spectral Products).
When a third signal is required, a dissecting microscope light (KL1500LCD, 175 W
Xenon, Zeisse) attached to a single fiber-optic cable was used.
Wavelength selection: The fiber-optic cables connected the light sources to three
monochromators (CMl 10, Spectral Products) used to select out specific wavelength
bands. The width of the monochromator entry and exit slits determined the bandwidth
and affected the flux levels of the output beam. Slits were set to either .3 mm or 2.4 mm
to select narrower or wider wavelength bands with lower or higher photon fluxes,
respectively.
Intensity determination: After leaving the monochromator exit slit the light passed
through a series of neutral density filters (OD= .3 to 3, d = 10 mm, Thor Labs) to adjust
the intensity and a plano-convex lense (f=100 mm, d= 50 mm, Thor Labs) to collimate
the beam before it entered the maze arms. The monochromators, filters and lenses were
affixed on a breadboard (Melles Griot) by holders attached to an optical rail.
Chapter 3. Materials and Methods
86
Figure 3.2. Y-maze. The syringe appears at upper right and is connected to the exit tubes
(left and below) by the base, containing the choice point (center).
Chapter 3. Materials and Methods
87
Filters were calibrated using a spectrophotometer (RPS 900, International Light).
Ten readings of the irradiance 2 cm in front of the exist slit of the monochromator were
take with each filter and 10 without filters, for filters of T = .5, .1, and .01. For T = .001
filters measurements were made using light exiting the fiber optic without the
monochromator. The average irradiance with a filter was divided by the average
irradiance without filters to gain an estimate of the filter's actual transmission. The
irradiance readings of the spectrophotometer are accurate only to +/- 50% and estimates
of actual transmission were consistent with the manufacturers characterization of the
transmission rate.
Adaptation: For dark adaptation, flies were placed in a beaker surrounded by aluminum
foil and covered with a black card. For light adaptation, they were placed in a light-tight
wooden box. A white light emitting LED sheet (LED revolution) covered the inner
surface of the boxtop and illuminated tubes placed inside.
Initial experiments were conducted using a diffusers (Thor labs, DG20-220) placed
between the lenses and the exit tubes of the T-maze. However, the experimental results,
in particular the opponent behavior, gained using the diffusers were less significant those
acheieved without the diffusers. No diffusers were used in the experiment reported
below. The view of the stimulus from the T-maze was homogenous in its spectral
content, but the intensity in the central 2.3 cm diameter cross section at the location of the
T-maze was higher than in the .2 mm surround.
Chapter 3. Materials and Methods
88
Experimental Procedures
Experiments were conducted on 4-10 day old flies. The night before an experiment
flies were anesthetized with CO2 to facilitate removing them from bottles and counting
them for placement in tubes. Approximately 70 flies were placed in each experimental
tube and left overnight at 22 0C. The rack of tubes was transferred to the behavior room
30 min before the start of an experiment. The room, maintained at 22-24 0C, was
illuminated by a single red light. For dark-adapted tests, a tube was taken from the rack
before each test run, tapped on the desk to arouse the flies and laid horizontally for 90
seconds, where it was exposed to the red room light. The tube was then placed in a
covered beaker that was surrounded by aluminum foil for 90 seconds. For light-adapted
tests, flies were placed in the LED illuminated box for 5 minutes prior to test runs.
T-maze experiments'. After adaptation, the tube was removed from the beaker/adapting
box and flies were tapped from the culture tube into the entry hole of the maze and
trapped in the elevator. The maze was put in place in the experimental set up. The flies
were then lowered to the choice point, where they were exposed to the lights, and
allowed to run into the maze arms. After 30 seconds, the elevator was raised, ending the
trial and trapping the flies that had moved into the exit arms. The culture tubes were
removed and plugged with cotton. The number of flies going in each direction was
counted and proportions were determined.
Y-maze experiments: After dark adaptation, the tube was taken out of the beaker and flies
were tapped into the open end of the syringe, which was then inserted into the V base of
Chapter 3. Materials and Methods
89
the maze. The plunger of the syringe was depressed. When the plunger reached the
choice point, the slits closing off the arms of the maze were opened and the flies were
allowed to run into the arms. After 1 minute, the slits were closed and the exit arms with
flies inside were removed. At the end of the experiment, the number going in each
direction was counted. The arms of the maze were then cleaned with water, dried, and
reattached to the base.
At the start and the end of every experiment, the power spectrum of each
monochromator's output was measured at the maze choice point using a
spectrophotometer (RPS 900, International Light). To control for any left-right bias, the
orientation of the mazes was reversed on 50% of the trials as was the direction of the
lights.
Five types of T-maze experiments were conducted:
(1) In one light,full intensity and spectrum experiments, the maze was illuminated from
one side by a single monochromator. The monochromator was set to 6 wavelengths: 331,
355, 400, 442, 486 and 515 nm, corresponding to the peak sensitivities of the fly opsins
plus 400 nm. For each wavelength, the irradiance levels were varied from the maximum
(no filter) down to a level of 1-2 log units below the point at which the behavioral
response did not differ from chance. At least 1000 flies were used for each data point
(wavelength/intensity setting).
(2) In two-light, mixing experiments, the T-maze was illuminated on one side by two
monochromators whose outputs were combined by a beam splitter with no source on the
Chapter 3. Materials and Methods
90
other side. One source was set to 33 lnm or 442 nm at intensity levels sufficient to attract
approximately 80% of the flies when presented alone (7.1 units photon flux for 331 nm,
slit width .3 mm; 8.1 units photon flux for 442 nm, slit width of .3 mm). The other
monochromator was set to 331, 355, 400, 442, 486 or 515 nm and intensity was stepped
from the maximum output (no filter) down through 7 log units.
(3) In two-light, two-sided assays, the T-maze was illuminated by one source on each
side.
(4) In three-light, mixing experiments, each side of the maze was illuminated by a
baseline 355 nm source of the same intensity. A third light of variable wavelength and
fixed intensity was added to one side by means of a beam spitter.
(5) Color vision tests were similar to three-light, mixing experiments but the set up was
designed to simulate the addition of UV and visible light of variable intensities (on
opposite sides of the apparatus) to a constant intensity bilateral UV baseline. To acheive
this result using only 3 monochromators, I began with the experimental setup used in the
3-light mixing tests and varied the intensity of the solitary 355 nm source above the
baseline level to achieve the effect of a fixed-intensity UV baseline and an added UV
source of variable intensity. I also varied the intensity of a 515 nm source on the other
side that was added to the UV baseline on that side using a beam splitter.
Statistical Analysis
A T-maze experiment is a two-choice test; flies go in one of two directions, flies
remaining in the central chamber are not counted. The standard statistical test performed
Chapter 3. Materials and Methods
91
on two-choice data generating frequencies and proportions is a chi-squared (Jc) test.
However, ?ê tests cannot be used on the data generated using T- and Y-mazes because
the distribution is overdispersed, i.e. it is more variable than the expected binomial
distribution (Collett, 1992). As is the case with the binomial distribution, tests like Jd
assume that the mean of the data is equal to its variance (Agresti, 1996). Because the
variance of the data generated in T- and Y-maze experiments is greater than its mean,
such tests are too sensitive to be applied in these cases. Thus, if applied to this data the Jc
test will often return the result that proportions are significantly different when in reality
they are not.
To get around this limitation I assumed that the overdispersion in the data was due,
in part, to the influence of unidentified, and hence uncontrolled, variables, which
produced differences in the mean responses from day to day. When probabilities vary
across trials confidence intervals can be calculated by first determining the variance of
the probabilities across days and using this value to estimate the added variability that
may be present in each day (Box, et. al, 1978). I followed this approach in estimating
confidence intervals and hypothesis testing in the data that follows.
When the data was generated in experiments run on a single day I calculated mean
for an experiment and the variance across the results of individual trials run during the
day, which still exceeded what was expected from the binomial distribution. Hypotheses
tests were performed using i-tests applied to the arcsin of the square root of the daily
means and the variance across the individual trials.
Chapter 4. Chromatic opponency in wild type flies
92
Chapter 4. Chromatic opponency in wild type flies
In this chapter, I investigate the hypothesis that wild type flies have color vision
using four types of phototaxic experiments: (i) one-light tests, on light- and dark-adapted
flies, in which flies are exposed to a light of a single wavelength on one side of the
apparatus and darkness on the other; (ii) two-light, two-sided tests in which flies are
exposed to one wavelength on either side; (iii) color-mixing experiments in which two
wavelength are combined on one-side to face either one wavelength or darkness on the
other; and (iv) color vision tests in which UV and visible light are added on opposite
sides to a bilateral UV baseline.
As noted in the introduction, intensity-independent behavioral opponency in twosided tests is a sufficient condition for color vision. The results of these experiments lead
to the conclusion that flies have a primitive color vision-like system that operates in one-
sided tests to distinguish between two wavelength ranges, 'UV (331-355 nm) and what
we would call 'blue-green' (442-515 nm) in an opponent fashion with UV light
enhancing and blue-green light inhibiting the phototaxic response. Tests with lights on
both sides, which are required to demonstrate true color vision fail to show that flies
distinguish between sources independently of intensity, though there is evidence of
opponency or non-linearity in the computation.
One-Light Tests: One-light tests can reveal thresholds for detecting light and provide
evidence about unadulterated responses to stimulation by single light sources. These tests
were used to investigate the hypothesis that the fly response to single lights exhibits
opponent behavior. If the behavioral response is opponent in character, it should have
Chapter 4. Chromatic opponency in wild type flies
93
two features. First, stimuli from different wavelength regions should have contrasting
effects on the behavior at higher intensities. Flies should prefer stimuli in one wavelength
range to those of another at rates that differ significantly, and the responses should trend
in opposite directions over a similar range of intensities. Responses to one wavelength
range should be going up, while responses to another range are going down.
To investigate the effects of dark adaptation on wild type response, I dark adapted
the flies for 90 seconds before the start of the first T-maze experiments. Response rates
were quantified as the preference index, defined as the number of flies going toward the
light minus the number going away from the light divided by the total number, was
determined for each intensity level at each wavelength. For dark-adapted flies the
response profiles for UV wavelengths (331 and 355 nm) are roughly sigmoidal (Figure
4.1). For log fluxes below five flies move toward the light at chance rates (Preference
Index (PI)=O), but between five and 1 1 log units intensity their preference increased with
intensity and decreased only slightly at the highest flux level. The response to visible
stimuli was quite different. Blue-green wavelengths (442 nm to 515 nm) result in
preferences that are two-part, and clearly, not sigmoidal. The flies begin to respond at
approximately five log units and attraction increases up until 8.5-9.5 log units, after
which the responses decrease with increasing intensity (Figure 4.1). The behavior shows
a difference between the response to UV and blue-green stimuli at higher intensities with
no significant differences at lower intensities.
Chapter 4. Chromatic opponency in wild type flies
331
355
400
442
486
0.8
nm
nm
nm
?m
?m
94
S*
/
515 nm
?
?
0.6
/
Ol
?
J
/
tu
S
\
\
/
/
?
\
\
,
0.2
0.1
0
Bi=SiPiC?
I
?
?
?
?
?
?
1
1
1
1
2
3
4
5
6
7
8
9
10
11
12
Log Photon Flux (quanta/mm2/sec)
a—I
Figure 4.1. One-light phototaxic response for dark-adapted wild type flies for 6
wavelengths at 8 intensities. Top: graphs plot preference index ((number going toward
light - number going away)/total number ) on the y-axis and the logarithm of the photon
flux (quanta/s/mm2) on the x-axis. Each data point represents at least 1000 flies with each
fly making a single choice. Inset: experimental setup. Colored rectangle is
monochromator. Red arrow represents direction of beam. The T-maze is the white cross
on the dark square.
Chapter 4. Chromatic opponency in wild type flies
95
To test whether flies exhibit opponent behavior in this assay, I pooled the UV data
and the blue-green data and calculated 95% confidence intervals for the data at each
observation (Box, et. al, 1978). The UV and visible responses are statistically
indistinguishable at intensities below 7.7 log units, but statistically different at higher
intensities (Figure 4.2). The visible curve begins to decline at 8.5 log units, while the
reduction in the UV response rate occurs at 9.5. Thus, flies respond to UV and bluegreen stimuli at different rates at higher intensities, and for an intensity span of one log
unit (8.5 to 9.5) the slopes of the curves have opposite signs with the UV response rate
increasing, while the blue-green response is decreasing.
Chapter 4. Chromatic opponency in wild type flies
96
CS Pooled UV versus Pooled Visible Response
UV
Visible
X
OJ
13
(U
U
?
I
?
?
CL
34567
89
10
11
Log Photon Flux (quanta/mrn2/sec)
Figure 4.2. Dark-adapted flies phototax at higher rates in response to UV than bluegreen light at higher intensities in one-light tests. Graphs plot preference index of
pooled UV data (331 and 355 nm) and pooled green-blue data (442, 486, 515 nm) as a
function of the log of the photon flux. For flux levels below 7.7 units, there is no
difference between UV and blue-green responses (two-tailed p>.05). At flux levels above
7.7 units, the response to UV is higher than to blue-green stimuli (two-tailed ? < .05).
Chapter 4. Chromatic opponency in wild type flies
97
To investigate the effects of light adaptation on the phototaxic response, flies were
light adapted for 5 minutes before the second experiment. Adaptation generally reduces
response sensitivity by 1-2 log units, which translates into a requirement of -10-1 0Ox
more photons before the flies begin to respond compared to dark-adapted conditions. In
response to stimuli of all wavelengths, flies began to respond at approximately 6 log
units. The responses divided into two groups with attraction to 331, 355 and 400 nm
stimuli trending upward with increasing intensity and 442, 486 and 515 nm responses
decreasing with increasing intensity after an initial peak (Figure 4.3). I pooled the data
and calculated 95% confidence intervals for each point. The pooled UV (331 and 355
nm) and blue-green (442 -515 nm) responses do not differ below 8.7 log units intensity
but differ significantly above this level (figure 4.4). The pooled responses also trend in
opposite directions from 9.5 to 1 1.5 log units.
Chapter 4. Chromatic opponency in wild type flies
X
?
o
e
?
?
0.9
¦331 nm
355 nm
0.8
• 400 nm
0.7
442 ? m
486 nm
¦515 nm
98
0.6
0.5
0.4
¦
Q.
0.3
0.2
0.1
0
-0.1
6
7
8
9
10
11
Log Photon Flux (quanta/mm2/sec)
I
Figure 4.3. One-light phototaxic response for light-adapted flies for 6 wavelengths at
8 intensities. Top: graphs plot preference index against the logarithm of the photon flux
(quanta/s/mm2). Responses to 331-400 nm stimuli show upward trends at higher
intensities, while responses to 442-515 nm stimuli decrease after an initial rise. Inset:
experimental Setup. At least, 2000 flies/data point.
Chapter 4. Chromatic opponency in wild type flies
0.7
0.6
99
Light Adapted CS Pooled UV versus Visible Response
•Visible
UV
0.5
?
?
?
?
£ 0.2
0
-0.1
7
8
9
10
11
12
Log Photon Flux (quanta/mm2/sec)
Figure 4.4. Light-adapted, wild type flies phototax at higher rates in response to UV
than blue-green stimuli at higher intensities in one-light tests against darkness.
Graphs plot preference index of pooled UV (331 and 355 irai) and pooled green-blue
(442-515 nm) data as a function of the log of the photon flux. For flux levels below 8.7
units, there is no difference between UV and blue-green responses (two-tailed, p>.05). At
flux levels above 8.7 units, the response to UV is higher than to blue-green (two-tailed
p<.05). Preference index increases for UV wavelengths from 9.5 to 1 1.5 units and
decreases for blue-green wavelengths over the same intensity range.
Chapter 4. Chromatic opponency in wild type flies 100
These results indicate that the phototaxic response to low intensity light is largely
achromatic. Increasing the intensity of light of wavelengths across the spectrum from the
detection threshold causes light- and dark-adapted flies to go toward the light at
increasing rates. At higher intensities the phototaxic rate to UV is significantly higher
than to blue-green, and the responses to stimuli in these wavelength ranges trend in
opposite directions. Differential response strength and opposite trend directions are
established features of opponency, supporting the thesis that the one-light phototaxic
response to higher intensities has opponent characteristics. In sum, higher intensity UV
and blue-green light appear to modulate a baseline achromatic response with UV light
enhancing, and blue-green light reducing, attractiveness.
two-light TEST: Color vision is the ability to distinguish between two lights on the basis
of wavelength and independently of intensity. Therefore, a demonstration that flies use
color vision during phototaxis must involve at least two light inputs and flies must show
preference for light of one wavelength range over light of another for a range of
intensities. The substantial difference between one-light responses to UV and blue-green
light makes it plausible that flies might distinguish UV from blue-green in an intensityindependent manner in two-light tests. Even the smaller, but still significant, differences
between one-light responses to 33 1 nm versus 400 nm and 442 nm versus 5 15 nm (figure
6.1) make it reasonable to investigate whether flies discriminate between lights within
these ranges independently of intensity.
Chapter 4. Chromatic opponency in wild type flies 101
To test the hypotheses that flies make intensity-independent wavelength
discriminations (i) within the UV range, (ii) within the blue-green range or (iii) between
UV and visible ranges, I carried out two-light tests with one light of fixed wavelength and
intensity and the second with fixed wavelength and variable intensity.
331 nm versus 400 nm discrimination: In this experiment, I fixed the 331 nm light at a
moderately high intensity and varied the intensity of the 400 nm light. The intensity of
the 400 nm source varied over a range that included the intensity of the 331 nm source. In
the reversed experiment I fixed the intensity of the 400 nm source at a moderately high
intensity and varied the intensity of the 33 1 nm light over a range that included the
intensity of the fixed source.
When the intensity ofthe 400 nm source was low (l/10th to 1/1000? ofthe 331 nm
source) in the fixed-intensity 331 nm/variable-intensity 400 nm experiment, flies went
toward the 33 1 nm source at a nearly 90% rate (Figure 4.5). As the intensity of the 400
nm source increased, greater proportions of flies preferred the 400 nm source reaching a
peak of approximately 75% at the highest intensity. In the fixed-intensity 400
nm/variable-intensity 331 nm experiment, flies went toward the fixed 400 nm source at a
75%-80% rate when the intensity of the 331 nm source was low (3.5 to 1.5 log units
below the 400 nm source). As the intensity of the 331 nm source increased, preference
for the 331 nm source reached a plateau at just over 80% response rate. Because the
preference for one source over the other changes with intensity, flies do not discriminate
between sources of 331 nm and 400 nm wavelengths independently of intensity in these
Chapter 4. Chromatic opponency in wild type flies 102
experiments. These results suggest that flies do not exhibit color vision in two-light, twosided, tests within the UV region.
Chapter 4. Chromatic opponency in wild type flies 103
Photon Flux Fixed Source/Photon Flux Variable Source
1/1DD
1/10
1
10
-400 fixed 331 variable
-331 fixed 400 variable
-2-1
0
1
Log Photon Flux Variable Source - Log Photon Flux Fixed Source (quanta/mm2/sec)
—^>^T<
?
Figure 4.5. Fly choice between 331 nm and 400 nm sources depends on light
intensity. Top: flies were tested in two-light, two-sided assays in which a fixed intensity
331 nm source (red curve, flux = 9.13) was paired against a variable intensity 400 nm
source (flux between 6.45 and 1 1.44), and a fixed intensity 400 nm source (black curve,
flux = 9.46) was paired against a variable intensity 331 nm source (flux between 5.82 and
10.81). Preference index (y-axis) is plotted as a function of the difference between the log
of the flux of the variable source and the log of the flux of the fixed source (x-axis,
bottom) and the ratio of the flux of the variable source to the flux of the fixed source (xaxis, top). Zero on the x-axis is the point of equal flux levels. Flies change preference
from the fixed source to the variable source as the flux levels of the latter rise. At equal
flux levels flies prefer the 331 nm to the 400 nm source. Inset: experimental setup. At
least, 1000 flies/data point. Error bars denote 95% confidence intervals.
Chapter 4. Chromatic opponency in wild type flies 104
442 nm versus 515 nm discrimination: The next experiment examined discrimination
within the blue-green range in two-light tests. The intensity of a 442 nm source was fixed
at a moderately high level, while the intensity of a 5 1 5 nm source was varied over a range
that included the intensity of the fixed source. Then the intensity of the 515 nm source
was fixed at a moderately high level, while the intensity of the 442 nm source was varied
over a range that included the intensity of the 515 nm source.
The source preference changed as the intensity of the lights was varied. When the
5 1 5 nm light was at a minimum intensity, flies went toward the fixed-intensity 442 nm
source 70% of the time (Figure 4.6). Increasing the intensity of the 515 nm source
increased the proportion of flies going toward that source, until a preference level of 60%
was reached. When the 515 nm source was at a fixed intensity and the intensity of the
442 nm source was more than 1.5 log units lower, the flies went toward the 515 nm
source at a 60% rate. As the intensity of the 442 nm source increased the preference
switched to that source. Because the preference for one source over the other changed
with intensity, flies did not discriminate between sources of 442 nm and 515 nm
independently of intensity in these experiments. The results suggest that color vision is
not involved in two-light discrimination within the blue-green region in this assay.
Chapter 4. Chromatic opponency in wild type flies 105
Photon Flux Fixed Source/Photon Flux Variable Source
1/1000
1/100
1/10
1
10
100
1000
W
(N
(D
?
Q)
Q)
O)
Log Photon Flux Variable Source - Log Photon Flux Fixed Source (quanta/mm2/sec)
I—o—I
Figure 4.6. Phototaxic choice between 442 nm and 515 nm light sources depends on
intensity. Top: flies were tested in two-light, two-sided assays in which a fixed-intensity
442 nm source (flux = 9.23, green curve) was paired against a variable-intensity 515 nm
source (flux 6.55 to 1 1.58), and a fixed intensity 515 nm source (flux = 9.32, blue curve)
was paired against a variable intensity 442 nm source (flux 6.53 to 1 1.52). Preference
index (y-axis) is plotted as a function of the difference between the flux of the variable
source and the flux of the fixed source (x-axis, bottom) and the ratio of the flux of the
variable source to the flux of the fixed source (x-axis, top). Zero is the point of equal flux
levels. Flies change preference from the fixed source to the variable source as the flux
levels of the latter rise. At equal flux levels flies prefer the 442 nm to the 515 nm light
source. Inset: experimental setup. At least, 1000 flies/data point. Error bars denote 95%
confidence intervals.
Chapter 4. Chromatic opponency in wild type flies 106
Discrimination between 355 nm and wavelengths across the spectrum: The next
experiment examined discrimination between 355 nm and a range of wavelengths in twosided tests. I fixed the intensity of the 355 nm source at a moderately high level and
varied a range of sources (331, 355, 400, 442, 486, 515, 550, and 590 nm) over intensity
spans that included the flux level of the fixed source. If flies discriminate between the
355 nm and other wavelengths independently of intensity, they should prefer the 355 nm
source to the other sources, or vice-versa, through these intensity changes. If the response
is independent of the wavelength of the variable-intensity source (univariance holds),
then the response profiles should have the same shape regardless of the wavelength of the
variable-intensity source.
Changing the intensity of the 331-400 nm sources resulted in a rapid change in
preference from the 355 nm to the variable-intensity source as the intensity of the latter
increased (Figure 4.7). In contrast, none of the visible wavelength sources attracted more
flies than the 355 nm source, even at the highest intensities. Still, the slopes of the visible
response curves as a function of intensity is positive at higher intensities. One cannot
conclude that flies discriminate between 355 nm and other wavelengths independently of
intensity in these experiments.
To examine the question of whether univariance governs choice in these tests, I
pooled the UV data (33 1 nm and 355 nm) and pooled the blue-green data (442 nm - 5 1 5
nm) and plotted the results as a function of the difference between the intensity of these
sources and the fixed 355 nm source (Figure 4.8). The UV responses rose at a
Chapter 4. Chromatic opponency in wild type flies 107
significantly higher rate at comparable response levels, indicating that univariance does
not hold between the UV and blue-green region in a choice against a 355 nm alternative.
Chapter 4. Chromatic opponency in wild type flies 108
Photon Flux 355 nm Source/Photon Flux Variable Source
1/10000
1/1000
I— 590
-+- - 550
515
486
442
400
355
331
"-4
1/100
1/10
1
10
100
1000
0
1
2
3
nm
nm
nm
?m
?m
?m
?m
nm
-3-2-1
Log of Variable Source Photon Flux- Log of 335 nm Source Photon Flux (quanta/mm2/sec)
I
T-maze
I
Figure 4.7. Preference for 355 nm versus other wavelengths is intensity-dependent.
Top: flies were tested in two-light, two-sided assays in which a fixed-intensity 355 nm
source was paired against variable-intensity sources of 8 different wavelengths.
Preference index (y-axis) is plotted as a function of the difference between the log of the
flux of the variable source and the log of the flux of the fixed source (x-axis, bottom) and
the ratio of the flux of the variable source to the flux of the fixed source (x-axis, top).
Zero represents the point where fluxes are equal. Increasing the intensity of the 331-400
nm sources changes the flies' preference. Increasing the intensity of the 442-590 nm
visible sources makes the sources more attractive, but does not result in a change of
preference. At least, 1000 flies/data point. Inset: experimental setup.
Chapter 4. Chromatic opponency in wild type flies 109
Ratio of Photon Flux of Variable Source to Photon Flux of Fixed Source
1/10000
1/1000
1/100
1/10
1
10
100
0
1
2
10(
visible
.6
.4
.2
X
OJ
¦p
(?
¿
-.2
-.4
-.6
-4
-3-2-1
3'
Log Photon Flux Variable Source - Log Photon Flux Fixed Source (quanta/mm2/sec)
Figure 4.8. Univariance is violated in the choice between a fixed 355 nm source and
variable-intensity UV and blue-green sources. Graphs plot the preference indices of
pooled UV (331 and 355 nm) and pooled blue-green (442 and 515 nm) responses (y-axis)
as a function of the difference between the logarithms of the photon fluxes of the variable
and fixed-intensity sources (x-axis, bottom) and the ratio of the flux of the variable
source to the flux of the fixed source (x-axis, top). The log of the odds (logit) of the flies
going toward the variable source was fit with a linear regression model and the effect of
wavelength on the slope of the curves was found to be highly significant (p<.0001). The
slope of the UV response profile was significantly greater than the slope of the visible
profile at each level of the preference index.
Chapter 4. Chromatic opponency in wild type flies 110
If flies used color vision in two-light tests, they should prefer a source of one
wavelength to a source of another regardless of their intensities. In the experiments
pairing 331 nm against 400 nm and 442 nm against 515 nm, the flies' preference changed
with the intensity of the lights, indicating that they did not use color vision to make these
discriminations.
In the experiment pairing the 355 nm source against lights of various wavelengths,
the preference for 355 nm versus UV wavelengths (331-400 nm) changed with intensity.
When the 355 nm source was paired against visible wavelengths, preference trended
toward the visible source as its intensity increased. Both facts suggest that color vision is
not involved in the two-light choice between 355 nm and other wavelengths. At the same
time, if univariance governed the phototaxic response, wavelength should have had no
effect on the slope of the curves at equal preference levels. The fact that the collective
slope of the blue-green wavelength sources is smaller than the slope of the UV sources at
a given response level indicates that wavelength is a factor and univariance is violated.
Thus, the results indicate the fly phototaxic response is not univariant in the choice
between 355 nm and variable-intensity sources across the spectrum.
The smaller slope of the response to blue-green wavelengths may be a sign that
inhibition or non-linearities are at work in the response to these wavelengths. Because
inhibition is one half of the opponent response, the data provide some support for the
existence of opponency in the underlying network.
Chapter 4. Chromatic opponency in wild type flies 111
It is important to note the significant difference between the character of the onelight and two-light responses. The one-light responses do not vary monotonically with
intensity, while the two-light responses do. Increasing the intensity of a source in a twolight tests never makes it less attractive, whereas increasing the intensity of visible
sources in one-light tests makes them less attractive above a certain intensity. The onelight experiments suggest that wild type flies possess an opponent mechanism that
becomes active only at higher intensities and opposes responses to UV and blue-green.
The failure of univariance in the two-light experiments also provides some evidence of
UV versus blue-green inhibition. Thus, the one-light and two-light data are consistent
with the existence of inhibitory responses driven by blue-green input. Such inhibition is
part of a sufficient condition for color vision but is not sufficient for color vision by
itself.
Behavioral Opponency and Mixing Experiments: The next experiments focused on
characterizing the opponent behavior that emerges at higher intensities
Two-Sided Mixing Experiments: The first experiments set out to replicate the opponent
responses observed by Fischbach (1979). In this test, I provided equal energy UV light
of 355 nm to both sides of the T-maze at half the maximum intensity of the light-driven
monochromators. Successively, I added 355 nm to one side of the same flux as the
baseline and then 442 nm and 515 nm to the same side at 1Ox the intensity of the
baseline.
When only the baseline 355 nm sources were present, the flies split equally
between the two sides. When 355 nm light was added to one side, more flies went toward
Chapter 4. Chromatic opponency in wild type flies 112
that side. When 442 nm and 515 nm light were added, more flies went toward the other
side (Figure 4.9).
Chapter 4. Chromatic opponency in wild type flies 113
355 nm versus 355 nm + Added Source
0.2
0.15
0.05
S -0.05
-0.15
-0.25
355 nm
442 nm
515 nm
Wavelength of Added Source (nm)
Fixed
intensity
I
I
Fixed
intensity
Variable
wavelength fixed
intensity
Figure 4.9. Adding fixed-intensity UV light to a bilateral UV stimulus enhances
phototaxis rates, while adding fixed-intensity blue-green light depresses response
rates. Top: fixed-intensity UV and blue-green light was added on one side to a bilateral
UV baseline (log flux = 10.3, each side). Bar graph depicts change in response rate from
the baseline as a result of adding light of three wavelengths. When only the baseline 355
nm sources were present, the flies split 50.5 to 49.5%, which did not differ significantly
from 50-50 (two tailed, p=.69). Adding UV (flux = 10.3) to one side caused the flies to
prefer that side at a statistically significant rate of 56.2% (two-tailed ? = .0014), while
adding 442 nm and 515 nm (flux = 1 1.3) caused flies to go toward the opposite side at
statistically significant rates of 59.1% (two-tailed ? = .0001) and 60.4% (two-tailed ?
=.0001), respectively. Error bars represent 95% confidence intervals. At least 1000
flies/data point. Bottom: experimental setup.
Chapter 4. Chromatic opponency in wild type flies 114
Flies exhibit opposed responses when presented with added UV light versus bluegreen light at particular flux levels against a UV baseline. Added UV light enhances
phototaxic rates, while adding blue-green light makes the combined source less attractive.
The last experiment was performed with added lights of wavelengths that were
peaks of three opsins (Rh4, Rh5, and Rh6) at fixed intensities. The baseline wavelength
was in the UV region. The limited character of the experiment (small range of added
wavelengths, fixed intensities and single wavelength as baseline) leaves open three
important questions: (i) do the results apply when light of other wavelengths are added?;
(ii) are the modulatory effects independent of the intensity of the added light?; and (iii)
does the effect only occur with a UV baseline? Because color vision requires intensityindependence, it is important to know that the opponency holds across a range of
intensities. The extent of the baseline and added wavelengths at which opponency occurs
are also critical for assessing the character of any chromatic discrimination that may be
exhibited. The next experiments were designed to address these questions.
One-sided mixing experiments: In the these experiments I added lights of a wider range
of wavelengths and varied their intensity in a one-sided, mixing assay. I also used two
different wavelengths, 331 and 442 nm, as one-sided baselines. I performed control
experiments to determine the intensity of light needed to drive approximately 80% of the
flies toward the baseline in one-light tests and used lights of these intensities as baselines
during the experiment. By setting the baseline intensity at this level there is reason, based
on the one-light tests, to suppose that achromatic mechanisms may have been saturated
and that only chromatic responses will remain. For the experimental manipulation, I
Chapter 4. Chromatic opponency in wild type flies 115
added 5 wavelengths on the same side as the baseline at a range of intensities and
measured the effect of the added light on the phototaxic response.
Adding UV light (331 and 355 nm) of flux less than 7 log units to the 331 nm
baseline had no effect on the response (Figure 4.10). Above this level the addition of UV
light attracted a higher proportion of flies to the combined source. The addition of bluegreen light (442 - 515 nm) to the 331 nm baseline also had no effect when the flux level
of the added light was below 8-9 log units. At intensities above this level, the addition of
blue-green light resulted in a lower proportion of flies going toward the combined source.
Pooling the results of adding the UV wavelengths and pooling the effects of adding the
visible wavelengths shows that the addition of UV light results in a significant
enhancement in attractiveness at approximately 7 log units, while the inhibitory effects of
added visible light appear at between 8 and 9 log units (Figure 4.1 1). The results were
similar for the 442 nm baseline (Figures 4.12 and 4.13).
Chapter 4. Chromatic opponency in wild type flies 116
Two Light Mixture with 331 Baseline
0.4
331
355
442
¦ 486
515
0.3
0.2
nm
nm
nm
nm
nm
0.1
X
?
-?
o
O
S -0.1
¦4—
?
?_
CL
-0.2 h
-0.3
-0.4
-0.5
j
2
3
4
5
6
u
J
7
8
9
10
L
11
12
Log Photon Flux of Added Source (quanta/mrn2/sec)
331 nm
I
variable
Figure 4.10. Phototaxic response to stimuli of variable wavelength and intensity
added to a one-sided 331 nm baseline. Top: graphs plot the effect of adding different
wavelengths and intensities of light to a 33 1 nm baseline (PI going toward the combined
source - PI going toward the baseline alone) as a function of the photon flux of the added
source. At least 1000 flies/data point. Bottom: experimental setup.
Chapter 4. Chromatic opponency in wild type flies 117
Pooled UV versus Pooled Visible added to 331 nm Baseline
0.4
UV
0.3
Visible
0.2 h
0.1 h
?
?
13
-??
?
?
e
\ T
? -0.1
?
?_
CL
K
-0.2
-0.3
-0.4
-0.5
±
5
6
7
8
9
10
11
?
12
Log Photon Flux (quanta/mm2/sec)
Figure 4.11. Adding UV light to a 331 nm baseline enhances, while adding blue-
green light diminishes, the attractiveness of a stimulus. Graphs plot the pooled effect
of adding UV (331 and 355 nm) light and the pooled effect of adding blue-green (442515 nm) light in a one-sided, mixing assay as a function of the log of the photon flux of
the added source. Adding UV light of flux greater than ~7 log units raises the
attractiveness of the side, while adding visible light of greater than ~8 log units decreases
the side's attractiveness. At least 1000 flies/data point. Error bars denote 95% confidence
intervals.
Chapter 4. Chromatic opponency in wild type flies 118
Two Light Mixture with 442 Baseline
0.3 ?
0.2
0.1
1
•331 nm
355 nm
442 nm
• 486 nm
515 nm
o
QJ
O
?
i» -0.1
?
CL
-0.2
-0.3
-0.4
3456789
10
11
12
Log Photon Flux of Added Source (quanta/mm2/sec)
Figure 4.12: Phototaxic response to light of variable wavelength and intensity
added to a one-sided 442 nm baseline. Graphs plot the effect of adding different
wavelengths and intensities of light to the baseline (PI going toward the combined source
- PI going toward the baseline alone) as a function of the photon flux of the added
source. At least 1000 flies/data point.
Chapter 4. Chromatic opponency in wild type flies 119
Pooled UV versus Pooled Visible added to 442 nm Baseline
Visible
T\
?
¦7Z
(U
O
CU
I
?
a>
CL
6
7
8
9
10
11
12
Log Photon Flux (quanta/mm2/sec)
Figure 4.13: Adding UV to a 442 nm baseline enhances the attractiveness of the
stimulus, while adding visible light decreases the stimulus' attractiveness. Graphs
plot the pooled effect of adding UV (331 and 355 nm) light and the pooled effect of
adding blue-green (442-515 nm) light to a 442 nm baseline in a one-sided, mixing assay
as a function of the photon flux of the added source. Adding UV light of flux greater than
approximately 8 log units raises the attractiveness of the side to which it contributes,
while adding visible light of flux greater than 8 log units decreases the side's
attractiveness. At least 1000 flies/data point. Error bars denote 95% confidence intervals.
Chapter 4. Chromatic opponency in wild type flies 120
The results of these experiments show that once a threshold flux level is reached
UV and blue-green lights have opposing effects when added to one-sided UV and visible
baselines. Added UV enhances the attractiveness of a source at higher intensities, while
adding visible light diminishes a source's attractiveness.
The behavior exhibited in these mixing experiments is quite similar to that
described by the bivariate model opponent neuron presented in figure 1 .4. Light from one
wavelength region (UV) enhances the response, while light from another depresses it, just
as light from one region excited the neuron, while light for another hyperpolarized it.
However, having contrasting responses to signals presented alone is not sufficient for
canonical color vision. To demonstrate that an organism has color vision one must show
that it distinguishes between lights of different wavelengths over a range of intensities
when they are presented together.
Given that added UV enhances the attractiveness of a signal, while added visible
decreases attractiveness in one light tests, a color vision experiment based on phototaxis
should begin with a bilateral baseline of equal attractiveness and provide added visible
light to one side and added UV to the other at a range of intensities. If flies exhibit color
vision, they should prefer the side to which UV is added regardless of the intensities of
the added UV and visible sources.
I performed this experiment using a 355 nm baseline with 515 nm light added to
one side and 355 nm light added to the other. When UV was added to the other side, the
flies always preferred the side with pure UV. Increasing the intensity of the added 515
nm light did not make the side to which it was added uniformly more attractive.
Chapter 4. Chromatic opponency in wild type flies 121
However, it also did not make it uniformly less attractive as it did in the one-light tests
(figure 4.14). Thus, the opponency that occurred when light was presented from one side
did not appear in the two-sided tests. The lack of a trend away from the added 515 nm
light makes it difficult to conclude that flies will continue to prefer the side with added
UV to the side with added visible as the 515 nm intensity rises beyond the levels tested.
So, the experiment does not provide convincing evidence of color vision.
Chapter 4. Chromatic opponency in wild type flies 122
0.9 r
0.6
0.2
- Log added 355 nm flux = 2.9
- Log added 355 nm flux = 1 .9
- Log added 355 nm flux = 0.9
-0.1
9
10
11
Log photon flux of added 515 nm source (quanta/mm2/sec)
Variable
Intensity
355 nm
I
I
Fixed
intensity
355 nm
Variable wavelength,
variable intensity
Figure 14.4: Flies prefer added UV to added 515 nm light across a range of
intensities against a bilateral UV background. Stimuli consisted of a bilateral 355 nm
stimuli (8.83 log units photon flux) with 515 nm light added to one side and 355 nm
added to the other side. Flies were given a choice between the combined 355 nm and 515
nm source on one-side and the 355 nm source on the other (baseline 355 nm plus added
355 nm). Top: graph plots the preference index as a function of the intensity of the added
515 nm light. Positive is toward the pure UV source. Each line plots the response for
added 355 nm light of a fixed intensity. Flies go toward the pure UV source over the
Chapter 4. Chromatic opponency in wild type flies 123
combined source for all positive values of the flux of the added 355 nm light. Data points
and error bars represent means values and 95% confidence intervals. Bottom:
experimental set up.
Chapter 4. Chromatic opponency in wild type flies 124
Discussion
The responses of wild type flies in one-sided phototaxic tests exhibit signature
features of color vision. Adding higher intensity UV enhances phototaxic rates in both
one-light and mixing experiments, while adding higher intensity blue-green light
depresses phototaxis rates. Because these effects persist across the range of intensities
that can be produced by the apparatus, it seems reasonable to say that wild type flies
respond differently to UV and blue-green light independently of intensity, which is as
close as one can come to color vision when light on one side is presented against
darkness.
The fly response to bilateral stimuli is more complex. Increasing the intensity of
one light in a two-light, two-sided test, never makes the variable side less attractive.
Thus, the behavioral opponency that occurs in one-sided tests does not appear in simple
two-sided tests. However, opponent responses arise in two-sided tests when UV and
visible light of certain intensities are added on one side against a bilateral UV
background, but the effect is not independent of intensity. Added UV light always
enhances the attractiveness of the side to which it is added, while added blue-green light
decreases the attractiveness at certain intensities, while enhancing it as others. Because,
two-sided tests are standard for demonstrating true color vision, the results do not support
attributing color vision to flies, though they clearly possess elements of a color vision
system.
Chapter 4. Chromatic opponency in wild type flies 125
When a fly makes a choice prior to phototaxing in a T- or a Y-maze its nervous
system must process a number of comparisons. The experimental results reported in this
chapter suggest that there are at least two mechanisms that integrate chromatic signals. At
the most general level, the division of the spectrum into two basic regions, UV and bluegreen in one-sided tests, is consistent with the number and character (UV and blue-green)
of the identified chromatic channels, downstream of R7 and R8, respectively (Gao, et. al,
2008). In one-light experiments, the phototaxic rates rise in response to blue-green
stimuli before falling. This behavior would appear to result from self-inhibition (negative
feedback) along the blue-green pathway as the intensity rises. There is no evidence of a
similar self-inhibition operating along the UV pathway.
The color mixing experiments were designed to activate more than one receptor
class and pathway and to focus on the higher intensity behavioral response. Higher
intensity UV and blue-green activate R7 and R8 cells, respectively, and both stimuli
activate R 1-6 cells. The baseline stimuli in one-sided tests were intense enough to
saturate achromatic mechanisms. The finding that bright blue-green light presented
against a UV background can reduce the response rate relative to the baseline can be
explained if the pathway downstream of R8 opposes the pathway downstream of R7.
Self-inhibition by the blue-green pathway cannot account for this reduction. If blue-green
self-inhibition was the only inhibition present in the circuit, flies would respond at
minimum to combined higher intensity UV and blue-green stimuli at the same rate as
they do to higher intensity UV by itself. Therefore, an additional mechanism is required
to explain how the fly neuronal circuits process visual information in the context of
Chapter 4. Chromatic opponency in wild type flies 126
mixing of wavelengths at higher intensities. One such mechanism may involve
chromatic opponency.
The location of the neuronal substrates and the circuit networks that account for the
opponent behavior is uncertain at the present. One possible mechanism could involve the
action of a synapse of R8 onto R7 identified in the wiring diagram from Gao, et. al
(2008). If functional, this synapse will become active when an ommatidium receives
simultaneous UV and blue-green inputs. A second candidate is the TM5 neuron, which is
immediately downstream of R7 and also receives input from Rl -6.
The asymmetric nature of the identified synaptic relationship between R7 and R8 is
consistent with the observed opponent behavior. If R8 inhibits its target as it does in other
cases, it will reduce activity in the R7 cell. No synapse with opposite polarity (R7
inhibiting R8) has been found anatomically. The behavior is asymmetric in that adding
higher intensity blue-green can be inhibitory but adding UV is always excitatory. In most
systems in which opponent synaptic connections have been found, some connections are
symmetric: receptor A inhibits a downstream target, while receptor B excites it, along
one pathway, while receptor A excites, and receptor B, inhibits a target, along another
pathway. If the R8 onto R7 synapse is the only chromatically relevant opponent
connection, the fly would possess the simplest possible opponent system.
An important first step in determining whether the R8 onto R7 synapse is the locus
of opponency would be establishing its functionality. So far no physiological effects of
R8 activation on R7 have been observed. It would be informative to record physiological
responses from intact R7 and TM5 neurons, while the eye is exposed to UV and blue-
Chapter 4. Chromatic opponency in wild type flies 127
green stimuli simultaneously. Experiments that target the R7 cells more specifically than
the sevenless mutation, which causes all R8s to express Rh6, such as via an Rh3/Rh4
double mutant, may aid in dissecting the neuronal connectivity further. Alternatively, a
selective genetic elimination of the synapse could be used to further probe its influence
on behavior. However, no receptors or transmitters characteristic of such a synapse have
been identified as of yet, making this approach unfeasible at the present.
The main puzzle posed by the data is why the one-sided and two-sided behavioral
profiles differ so dramatically. Two-sided responses violate univariance but are not
behaviorally opponent. The reduction in the phototaxis rates in response to higher energy
visible stimuli does not occur in two-light tests. In these tests the preference for a visible
source never decreases as the intensity ofthat source increases. This dependence on
intensity may result from Rl -6 mediated luminance input.
Luminance and chromatic channels appear to function at different intensity levels
in one-sided tests with the luminance channel driving the response to low intensity
stimuli and the chromatic channels becoming active at higher levels and modulating the
baseline established by luminance. In two-sided tests it appears that the luminance
channel may have a larger dynamic range and affect the response up to the highest
intensity levels. Whatever integration occurs in the choice between two lights appears to
rest more strongly on Rl -6 activity than the integration involved in the one-sided
response.
Chapter 5. Receptor contributions to opponency
128
Chapter 5. Receptor contributions to opponency
The number of basic hue contrasts that a person perceives corresponds to the
number of opponent pairings that occur in her visual system. Normal trichromatic
observers have two opponent systems (M/L and S/ML) and perceive green/red and
blue/yellow as contrasting pairs. Dichromats have only S/(Mor L) or M/L opponency and
perceive only one basic contrast, either blue/yellow or green/red. Cone monochromats
and individuals who have no cones do not perceive any hues, for them the world is
colorless.
The fly eye contains five types of receptors. Because two receptors are required
for an opponent pair, it would seem likely that there is more than one opponent pairing in
the fly visual system. If the generalization regarding hue perception in humans extends to
the behavior of flies, one would expect that they would have contrasting behavioral
responses to more than one wavelength pairing. However, the evidence accumulated so
far suggests that wild type flies distinguish behaviorally between only two wavelength
regions, UV and blue-green, something that can be accomplised with only two receptors .
This raises the question of what functions the three other receptors are performing.
Substantial evidence supports the view that the R7 cells mediate the preference
for higher intensity UV in two-choice tests (Tomlinson and Ready, 1987; Gao, et. al,
2008). The roles of R8 and Rl -6 in wavelength-selective behavior are less well
understood. To examine the role that Rl -6 and R8 play in color vision and to test how the
functions of the two R8 cells might differ, I conducted a series of phototaxic experiments
using receptor mutants.
Chapter 5. Receptor contributions to opponency
129
The experiments were of four types: (i) one-light, full intensity and spectrum
studies, on RhI -mutants, (ii) two-light Y-maze tests on RhI -mutants; (iii-iv) one-light,
full intensity and spectrum studies on i?M-mutants and flies in which cells downstream
of R7 and Rl -6 are conditionally silenced using shibirits ('shits'). I found that while RhImutants have reduced sensitivity and phototaxic response at all wavelengths, they also
have opponent responses contrasting 442 nm and 515 nm at the highest intensities. In
two-light tests they prefer 442 nm to 5 1 5 nm light at most intensities and increasing the
intensity of the 515 nm source does not make it uniformly more attractive, in contrast to
wild type behavior in a T-maze (Figure 4.6).
These results suggest that Rl -6 may have an effect on the chromatic
discriminations flies make and may contribute to the intensity-dependence of wild type
preference in a choice between 442 nm and 515 nm sources. The results also provide
evidence that, in the absence of Rl -6, pR8s might drive phototaxis toward 442 nm, while
yR8s drive flies away from 515 nm at higher intensities. Finally, they are consistent with
yR8s playing a role in driving attraction toward light at moderate intensities. Results of
the shits experiments could not be interpreted because there was evidence that the Rl -6
pathway was not fully disabled by the experimental manipulation.
Rl -6 Experiments: The experiments reported in the chapter 4 and the spectral properties
of Rl -6 raise a number of questions about the role of these cells in color vision. The most
basic is whether Rl -6 cells contribute to wavelength discrimination. Because Rl -6 are
sensitive to photons across the spectrum (Figure 1.8, top), it is possible that they may
provide input to a luminance channel and not facilitate discriminations between different
Chapter 5. Receptor contributions to opponency
130
wavelength regions. On the other hand, Rl -6 differ chromatically from the other
receptors primarily in having two sensitivity peaks, UV (350 nm) and blue (515 nm),
rather than one. As a result, the cells will respond more strongly to certain wavelengths
given stimuli of equal flux rates. Such differential sensitivity can provide actionable input
into a chromatic discrimination mechanism and so may facilitate contrasting behavioral
responses to different wavelengths.
The second question concerns the spectral resolution of the color vision system.
The wiring diagram from Gao, et. al (2008, Figure 1.9) depicts Rl -6 as making
contributions, via L3, to both the main pathways involved in phototaxis and, presumably,
to the color vision system. If color vision results from a comparison of the signals
between these pathways beyond the receptor stage, i.e. the integration is not
accomplished by the R8 onto R7 synapse, then it appears that Rl -6 make confounding
inputs to the discrimination mechanism. By adding input of the same spectral character to
both pathways, Rl -6 inputs should obscure contrasts between them, reducing the
discriminatory power of the system. Under this hypothesis, removing Rl -6 should
increase the ability of flies to discriminate between signals of different wavelengths.
Third, it is plausible that Rl -6 would affect the intensity level of the background
lighting against which opponency develops. Rhl-nu\\ mutants are approximately one
order of magnitude less sensitive to light in ERGs than wild types flies (Harris, et. al,
1976), and i?/z7-hypomorphs are approximately 3Ox (1.5 log units) less sensitive than
wild types in one-light phototaxic tests (Gao, et. al, 2008). Opponent behavior appears in
the phototaxic response of wild type flies in one-light tests only after the response has
Chapter 5. Receptor contributions to opponency
131
been driven to the 75-80% level, and in mixing tests the baseline needs to be at least this
attractive before opponency develops. This profile does not fit the form of classical
opponency, in which deflections are either above or below a neutral baseline. If Rl -6
mediate the low intensity, achromatic portion of the wild type response, removing them
could give rise to conventional opponent behavior that develops in the absence of a
background. As a result, it is possible that RhI-mutants might exhibit color vision
without background stimuli.
To explore these questions, I first performed a one-light, full-spectrum and
intensity study with Rhl-null mutants over of 6 wavelengths at 5 intensities (Figure 5.1).
Consistent with the view that Rl -6 provide sensitivity to low light levels, the flies are less
sensitive to light as indicated by response thresholds that are ~3 log units higher than
wild types. The maximum amplitudes are also reduced for all wavelengths relative to
wild types (Figure 5.2). To determine whether particular response amplitudes differ from
chance I replotted the data as bar graphs with 95% confidence intervals (Figure 5.3).
Stimulation by 486 and 515 nm light at the highest intensities led a majority of flies to
move away from the light, while shorter wavelengths attracted the flies at all intensities.
Thus, RhI-mutants exhibit conventional opponent behavior at the highest intensities to
wavelengths above and below 486 nm.
If Rl -6 cells have an effect on the resolution of discrimination across the
spectrum, one would expect the size of the difference between the maximum and
minimum response levels to differ in wild types and RhI-mutants. The most attractive
stimulus for RhI-mutants is high intensity 33 1 nm light (log flux =11.0) and the least
Chapter 5. Receptor contributions to opponency
132
attractive is high intensity 515 nm (log flux = 1 1.6). The spread between the maximum
and minimum of the high intensity responses, .61, does not differ from the spread for
wild types, .60 (two-tailed ? = .82). This result suggests that Rl -6 cells do not contribute
to the overall amplitude of the opponent contrast. To test for increased discrimination in
the blue-green regions, I compared the largest difference between responses to 442 and
515 nm in Rhl-mutants and wild types. The largest difference in RhI-mutants is .38 at
the highest intensity setting, while the maximum difference in wild types is .24. The
difference between these values is not significant (two-tailed ? = .1). Thus, the presence
of Rl -6 does not affect the amplitude of the contrast between the responses to blue and
green stimuli.
Chapter 5. Receptor contributions to opponency
133
0.5
0.4
0.3
•331
¦ 355
¦ 400
442
nm
nm
nm
?m
486 nm
¦515 nm
X
?
T3
CZ
0.2
?
o
a
?
?
?
"4—
0.1
?
Q-
-0.1
-0.2
7
8
9
10
11
Logarithm of Photon Flux (quanta/mm2/sec)
I
Figure 5.1: Phototaxic response of ÄÄi-mutants at 6 wavelengths and 5 intensity
levels. Top: graphs plot preference index as a function of the logarithm of the photon
flux. Intensities range from 1-2 log units below the detection threshold, to the maximum
output of the apparatus. At least 2000 flies/data point. Inset: experimental setup.
Chapter 5. Receptor contributions to opponency
134
NinaE
<5
0.4
331 nm
355 ? m
400 nm
442 nm
48B ? m
515 nm
Wavelength (nm)
Figure 5.2: Phototaxic response is reduced in if/ii-mutants relative to wild types at
all wavelengths. Bars depict the maximum response at each wavelength for RhI-mutants
and wild types (CS). All differences are highly significant (two tailed p<.0001). Bars and
error bars represent means and 95% confidence intervals.
Chapter 5. Receptor contributions to opponency
135
0.6
331
355
400
442
486
515
0.5
nm
nm
nm
nm
?m
nm
0.4
0.3
0.2
0.1
A -,
-i
TZT=IiT-—mi
-0.1
-0.2
-0.3
Photon Flux Level
Figure 5.3: l?Ai-mutants exhibit opponent responses above and below 486 nm
higher intensities. Graphs plot the preference index as a function of the ranking of the
photon flux in the series of measurements for each wavelength (? ' is the lowest flux
level, '5' is highest). The highest response level occurs at 331 nm light, level 5 (flux =
1 1.0, PI = 42.4) and the lowest at bright 515 nm, level 5 (flux = 1 1.6, PI=-19.05).
Significantly more flies avoid 486 nm (two tailed ? < .001, level 5; ? = .0017, level 4)
and 515 nm (two tailed p<.001, level 5; p<.0001, level 4) light than are attracted to these
wavelengths at the two highest intensities. All other wavelengths are attractive at all
intensities at which the response is significant.
Chapter 5. Receptor contributions to opponency
136
The reduced response level and decreased sensitivity of RhI-mutants are
consistent with the view that Rl -6 cells provide a baseline level of activation in response
to low intensity stimuli that is modulated by the activity of R7/8 at higher intensities in
wild type flies. The response of RhI-mutants fits the pattern of a classical opponent
system in requiring no baseline and with a division between positive and negative
responses. The finding that the distinction between attraction and avoidance occurs in the
blue-green region rather than between UV and blue-green indicates that Rl -6 have an
effect on the chromatic features of the response. Finally, there is no evidence that
removing Rl -6 enhances the ability of flies to distinguish between UV and visible or blue
and green wavelengths.
The second set of experiments on iî/zi-mutants involved two-light discrimination
tests between 442 nm and 515 nm in a Y-maze assay. I argued in the Chapter 2 that an
organism exhibiting intensity-independent opponent behavior to lights presented
simultaneously will have color vision. Although the preceeding tests involved only one
source, the results suggest that RhI -mutants may show opponency, if the lights are
presented simultaneously and, therefore, may have blue-green color vision. To test this
hypothesis, I carried out two-light tests to determine whether i?/zi-mutants discriminate
between 442 nm and 515 nm independently of intensity. Flies distinguish between two
lights on the basis of wavelength and independently of intensity if they consistently chose
one light over the other as their intensities vary. They fail to distinguish between the
Chapter 5. Receptor contributions to opponency
137
lights on the basis of wavelength alone if their choice can be altered by changes in
intensity.
Preliminary tests showed that RhI-mutants did not show strong discrimination
between lights of different wavelengths in a T-maze. As a result, experiments were
conducted in a Y-maze. The intensity of the 442 nm source was set to the five flux levels
at which flies showed a positive response in one-light tests. At each level of the 442 nm
source, the intensity of the 515 nm source was varied from just below the detection
threshold up to the the maximum output of the apparatus. The attractiveness of the 515
nm source did not increase consistently with intensity (Figure 5.4) in contrast to the wild
type response in a T-maze (Figure 4.6). The proportion of flies going toward the 515 nm
source tended to rise toward the middle intensities but fall at the highest intensities. Flies
preferred the 515 nm to 442 nm source at only one wavelength/intensity setting (442 nm,
flux = 1 1 .9, 5 1 5 nm, flux = 1 0.0), a setting at which flies also go toward the 5 1 5 nm
source at statistically significant rates when it is paried against darkness in the Y-maze
(data not shown) and at which inhibitory mechanisms may begin to reduce the
attractiveness of the 442 nm source (Figure 5.3).
Chapter 5. Receptor contributions to opponency
0.2
442 flux = 12
0.1
442
442
442
442
0
-0.1
flux
flux
flux
flux
=
=
=
=
11
10
9
8
T3
C
?
'S -0.4
CL
-0.5
-o.e
-0.7
-0.8
"8
8.5
9
9.5
10
10.5
11
11.5
12
515 nm photon flux log (quanta/mm2/sec)
I
I
Figure 5.4: Flies prefer 442 nm to 515 nm at nearly all wavelength and intensity
settings in Y-maze tests. Graph plots preference index in choices between 442 and 515
nm sources as a function of the 515 nm source intensity for different intensities of the 442
nm source. Positive PIs indicate a preference for green. As the intensity of the 515 nm
source is increased, the proportion of flies does not increase consistently. Flies went
toward the 5 1 5 nm source over the 442 nm source at statistically significant rates at only
one intensity setting (442 nm, log flux = 1 1.85; 515 nm log flux = 10.0). Inset:
experimental setup. 200 flies/data point.
Chapter 5. Receptor contributions to opponency
139
RhI -mutants phototax toward 442 nm and away from 5 1 5 nm at the hightest
intensities in one-light tests. At these settings, i?W-mutant flies prefer 442 nm to 515 nm
lights at highly significant rates in two-light tests. Thus, they prefer 442 nm to 515 nm at
intensity settings at which they exhibit opponent behavior. If flies had blue-green color
vision, they would either prefer the 442 nm to the 5 1 5 nm source or prefer the the 5 1 5 nm
to the 442 nm source regardless of their intensities. Because they prefer 515 nm to 442
nm at one setting, but 442 nm to 515 nm at all others, the flies do not have blue-green
color vision, though they have a strong general preference for blue over green. Given that
the major chromatic distinction in wild type behavior is between UV and blue-green, the
addition of Rl -6 cells appears to convert a capacity for discriminating 442 nm from 515
nm light into a capacity for discriminating UV from blue-green. Thus, it appears that Rl6 make a significant contribution to chromatic discrimination in the fly.
R8 Experiments: In chapter 4 1 found that the response rates of wild type flies in onelight tests decrease significantly at higher intensities and longer wavelengths. RhImutants show a parallel reduction, leading to Photophobie behavior at higher intensities
for 486 nm and 515 nm light. Because 515 nm is the wavelength at which yR8 cells are
most sensitive, one might conjecture that these cells contribute to the response reduction.
If yR8s mediate a reduction in high intensity phototaxis rates, then i?M-mutants should
show smaller reductions than wild types at longer wavelengths and higher intensities in
Chapter 5. Receptor contributions to opponency
140
one-light tests. To test this hypothesis and examine other potential roles yR8s may play,
I carried out one-light, full-intensity and spectrum tests on i?/z6-mutants.
Rh6-mutant flies go toward the light at statistically significant rates for all
intensities. They begin to respond at -6-7 log units (Figure 5.5). At each wavelength the
response rises to a peak before declining at higher intensities. The response to 515 nm
light differs from others in having a peak at a lower intensity and a smaller decline at
higher intensities. A plot of the responses of i?M-mutants and wild types at longer
wavelengths makes it easier to see how Rh6 and wild type responses differ (Figure 5.6).
Wild types appear to have higher maximum response rates than iîM-mutants at each
wavelength, and their responses appear to decline to a greater extent at higher intensities.
A plot of the maximum responses of wild types and Rh6-mutants for each wavelength
(Figure 5.7) shows that i?/z6-mutants have smaller maximum responses at each
wavelength but that the differences between the wild type and Rh6 maxima are greater at
longer wavelengths. To determine the significance of the differences between the maxima
at each wavelength I calculated the probability that the differences were generated by
chance given the null hypothesis that the response levels were equivalent. The differences
were highly significant at each wavelength but most significant at 486 nm (two tailed
p<10-7).
The size of the reduction in the response at higher intensities is the difference
between the maximum response to a particular wavelength and the response rate at the
Chapter 5. Receptor contributions to opponency
141
highest intensity at that wavelength. The difference in the size of the reduction between
wild types and i?M-mutants is not significant for shorter wavelengths (331-400 nm) but
is significant at longer wavelengths (442-515 nm, Figure 5.8). Wild type responses have
larger declines than i?/z<5-mutants at higher intensities and longer wavelengths.
Chapter 5. Receptor contributions to opponency
142
Rh6 Response to Peak Wavelengths of Drosophila Opsins plus 400 nm
0.9
0.8
0.7
?
<x>
-a
¦331
¦ 355
¦ 400
442
nm
nm
nm
nm
486 nm
•515 nm
0.6
Î 0.5
I 0.4
?
£ 0.3
0.2
0.1
0
-0.1
4
5
6
7
8
9
10
11
12
Logarithm of Photon Flux (quanta/mm2/sec)
I
Figure 5.5. Phototaxic response of Rh6-mutants to one light at 6 wavelengths and 7
intensities. Top: graphs plot preference index as a function of the logarithm of the photon
flux. Intensity levels ranged from the maximum produced by the apparatus to 2-3 log
units below the point at which the responses did not differ from chance. The response to
515 nm differs from other wavelengths in having its peak at a lower flux level. At least,
2000 flies/data point. Inset: experimental setup.
Chapter 5. Receptor contributions to opponency
143
0.8
0.7
0.6
0.5
------ 442 nm CS
------ 486 nm CS
------ 515
H— 442
"i
486
H— 515
nm CS
nm Rh6
nm
nm
X
?
13
-?Oj 0.4
O
C
œ
Ï5
0.3
<d
L_
CL
0.2
0.1
-0.1
4
5
6
7
8
9
10
11
12
Logarithm of Photon Flux (quanta/mm2/sec)
Figure 5.6. One-light phototaxic response of wild types and i?Ä6-mutants to 442-515
nm stimuli. At least, 1000 flies/data point for wild types and 2000 flies/data point for
Rh6-mutants.
Chapter 5. Receptor contributions to opponency
144
Effect of RhB Mutation on Maximum Response
D.8
S ?.6
Qi
O)
S. 0.4
V)
03
I 0.3
<B
2 0.2
0.1
IM
331
355
400
442
Wavelength (nm)
486
515
Figure 5.7: Rh6-mutants exhibit lower maximum responses than wild types for all
wavelengths. Graphs plot the maximum phototaxic response of wild types (CS) and Rhômutants at 6 different wavelengths. The highest response levels for wild types are
significantly greater than for i?/z(5-mutants at all wavelengths but the difference is larger
at the longest wavelengths (486 and 515 nm). At least 2000 flies/data point. Error bars
represent 95% confidence intervals.
Chapter 5. Receptor contributions to opponency
145
0.6
|CS
¡Rh6
S· 0.5
S 0.4
0.3
0.2
0.1
-0.1
¦i il J
J_
331 nm
355 nm
400 nm
442 nm
486 ? m
515 nm
Wavelength (nm)
Figure 5.8. Rh6-mutant responses at higher intensities and longer wavelengths
decline less than wild type responses. Graphs plot the difference between the maximum
response at a given wavelength and the response at the highest intensity ofthat
wavelength for Rh6-mutants and the wild types (CS). The size of the reduction for Rh6mutants is smaller than for wild types at the longest wavelengths. At least, 2000 flies/data
point for Rh6-mutants and 1000 flies/data point for wild types. Error bars represent 95%
confidence intervals. [*p<.05, ** ? < .01, *** p<.0001].
Chapter 5. Receptor contributions to opponency
146
The fact that wild type flies have higher maximum responses at longer
wavelengths in one-light tests than i?M-mutants and that these maxima occur at moderate
intensities suggest that activation of yR8 cells may enhance phototaxic rates levels at
moderate intensities. This hypothesis is also supported by the observation that RhImutants phototax toward 515 nm, the peak of Rh6 sensitivity, at statistically significant
rates only at moderate intensities in the Y-maze. At the same time, the greater decline in
the phototaxis rates at the highest intensities and longer wavelengths in wild types
compared to Z?/z6-mutants can be explained if we suppose that activation of yR8 cells
reduces phototaxis rates at these intensities. The same explanation can also account for
the the decline in the phototaxis rates for i?/ji-mutants at 515 nm at these intensities.
The response of i?/z6-mutants to 515 nm light also gives some indication of how
Rl -6 might contribute to the phototaxic response. In i?M-mutants the response at this
wavelength is mediated primarily by Rl -6 with a small contribution from pR8s. The
phototaxic curve rises to a plateau at moderate intensities before levelling off, a trajectory
which suggests that the Rl -6 contribution to phototaxis may saturate at moderate
intensities and remain relatively stable into the high intensity region.
The next set of experiments sought to probe the influence of R8 cells further. RhImutants phototax toward 442 nm at greater than chance rates for all intensities above the
detection threshold. Given that 442 nm is the peak sensitivity of pR8 cells, one can
hypothesize that these cells may drive attraction toward 442 nm light. Alternatively, the
positive phototaxis rates of RhI -mutants at 442 nm could be due to residual R7 activity.
Chapter 5. Receptor contributions to opponency
147
At higher intensities, 442 nm phototxis rates fall between UV rates and chance, just as the
sensitivity of R7s drops between the UV region and longer wavelengths. If 442 nm light
activates R7s to a moderate degree, the attraction that results could drive flies toward that
wavelength to an intermediate degree, accounting for the observed response rates.
To investigate the role of R8s I sought to block activity downstream of R7 and
Rl -6. 1 used flies of the genotype genotype UAS- shf;; orf2-Gal4, UAS-shf ('ortc2-shf
for short), which express shits in cells downstream of R7 and Rl -6 using the c2 fragment
of the ort promoter (Gao, et. al, 2008). Ortc2 drives expression in DM8 cells, which are
the primary transducers of R7 activity, and L 1-3 cells, which are downstream of Rl -6,
but not in any TM cells, including those downstream of R8s. Thus, at experimental
temperatures synaptic transmission along the main pathways downstream of R7 and Rl -6
should be blocked. However, because ortc2 does not express in TM5 cells, which are
immediately downstream of R7, there is at least one pathway downstream of R7 that is
not silenced by this genetic approach. Significant responses to either low intensity stimuli
or higher intensity UV were taken to indicate that Rl -6 or R7, respectively, were
contributing to the behavior. If present, such response would indicate a failure of the
experimental manipulation.
I conducted a one-light, full intensity and spectrum study of ortc2-shi flies. At the
control temperatures, the maximum response of the flies is to 442 nm at 10.5 log units
intensity; the minimum response is to 355 nm in the same intensity range (Figure 5.9).
Both of these results differ from the wild type profile. At experimental temperatures, flies
begin to respond at 6-7 log units intensity and responses to all wavelengths rise up to 8
Chapter 5. Receptor contributions to opponency
148
log units intensity (Figure 5.10). Above this level, the responses divide with 400-486 nm
stimuli producing higher responses, and the response rates to 331, 355 and 515 nm
stimuli declining below the 8 log unit baseline. Flies respond to 442 nm stimuli at greater
than chance rates at all intensities above the detection threshold.
Chapter 5. Receptor contributions to opponency
149
-331 nm
- 355 nm
0.8
- 400 nm
442 nm
486 nm
0.6
-515 nm
?
?
?
o
0.4
?
?
?
Q-
0.2
-0.2
7
8
9
10
11
12
Logarithm of Photon Flux (quanta/mm2/sec)
I
Figure 5.9. Phototaxic response of orf2-shf flies is impaired relative to wild types at
control temperatures. Top: graph plots preference index as a function of of the photon
flux. The strongest response is to 400-486 nm stimuli with a maximum response of PI =
.43 for 442 nm at 10.5 log units photon flux. The response to 515 nm and 355 nm drops
below significance in the region of 10 log units. At least, 1000 flies/data point. Inset:
experimental setup.
Chapter 5. Receptor contributions to opponency
0.8
0.6 h
¦331
355
¦ 400
442
488
¦515
150
nm
nm
nm
nm
nm
nm
a>
13
(J
CZ
?
0.4
?
?
a?
CL
0.2
-0.2
7
8
9
10
11
12
Logarithm Photon Flux (quanta/mm2/sec)
I
Figure 5.10. Phototaxic responses oiorf2-shf flies at experimental tempertures
show relative reductions in phototaxis rates at 331, 355 and 515 nm. Top: graph plots
preference index as a function of the photon flux. Responses divide into two groups at
moderate to high flux rates with flies responding at higher rates to 400-486 nm stimuli.
Above 9 log units the responses to 331, 355 and 515 nm light drop to chance. At least,
1000 flies/data point. Inset: experimental setup.
Chapter 5. Receptor contributions to opponency
151
At control temperatures it appears that shi s has significant effects on phototaxic
behavior. Response rates are lower at contol temperatures relative to wild types,
especially in the UV region. The chromatic profile of the response is also changed with
orf2-shi flies having their strongest responses to middle wavelength stimuli (400 and 442
nm), while wild types respond most strongly to UV (331 and 355 nm). While not ideal,
this deviation does not invalidate the results of the experimental manipulation. It means
only that shits may function not as a mechanism for conditionally disabling synaptic
transmission in this experiment, but as a mechanism for partially disabling synaptic
transmission even at control temperatures.
At experimental temperatures, the low response rates to UV wavelengths at higher
intensities are not consistent with significant R7 activity in the remaining R7 to TM5
pathway. Given that the phototaxis rates to 442 nm are relatively robust compared to the
responses at other wavelengths, it would appear that pR8 cells may be driving the
phototaxic response to 442 nm. At higher wavelengths the response to 515 nm does not
differ from chance, indicating that yR8s may not drive phototaxis at these intensities.
However, one feature of the experimental results raises questions about their
interpretation. The flies respond at 7-8 log units intensity to all wavelengths. This
response threshold is 1-2 log units below the level at which RhI-mutants begin to
respond, which suggests that the pathway downstream of Rl -6 might not be fully
silenced in ortc2-shi flies even at the experimental temperatures. It is concievable that the
higher response threshold in RhI-mutants might be due to damage that the mutation
Chapter 5. Receptor contributions to opponency
1 52
causes to the eye and downstream pathways, damage that does not occur in orf -shi flies.
Still, the uncertainty relating to the completeness of the knockout makes it impossible to
draw firm conclusions from this experiment.
Discussion
The results of the experiments in chapters four and five suggest a tentative
solution to the question of how the fly, in spite of having five types of reeceptors in its
eye, manages to discriminate between only two wavelength regions in phototaxis. The
two R8 cells may be responsible for the most basic color contrast. If a stimulis that
activates pR8 cells attracts a fly, while activation of yR8 cells leads to avoidance, a
system consisting only of these two cells and their downstream pathways will lead to
behavior that discriminates between blue and green light.
A substantial body of research suggests that R7 cells account for the strong wild
type attraction to UV light at higher intensities (Tomlinson and Ready, 1987; Gao, et. al,
2008). If the network downstream of the R7s drives attraction to wavelengths to which
the cells are most sensitive and the above proposals regarding R8 function are correct,
then we can explain the behavioral response of Rh1-mutants in one-light tests. These flies
are attracted to wavelengths of 442 nm and below as a result of R7 and pR8 activity and
avoid higher intensity 486 nm and 515 nm light via yR8 stimulation. Under the present
hypothesis R7 cells do not alter the basic contrast established by R8 cells but merely add
UV sensitivity to the fly response. R8 cells form an opponent system. R7 cells do not.
Chapter 5 . Receptor contributions to opponency
153
Rl -6 cells make a clear contribution to chromatic discrimination. RhI -mutants
discriminate behaviorally between UV-blue (<442 nm) and (>486 nm) cyan-green, while
wild types contrast UV (331-355 nm) and blue-green (442-515 nm). How does the
addition of Rl -6 to the eye shift the dividing line between wavelengths that enhance and
depress phototaxis rates? The wiring diagram from Gao, et. al, 2008, indicates that the
L3 pathway downstream of Rl -6 converges on the pathways downstream of R7 and R8.
Since only one pair of R7/8 cells is depicted, it is not clear from the diagram whether the
downstream architecture is the same in ? and y ommatidia. However, if the structure of
the network is the same in both ommatidia types, the L3 line will converge with the
axons of both ? and y R7s and R8s at the second-order TM neurons.
Aside from the response to low intensity light, the key difference between RhImutants and wild types is in the response to 442 nm stimulation. 442 nm light is attractive
to iî/zi-mutants at all intensities above the detection threshold, while additional light of
this wavelength reduces the response of wild types at the same intensities. Assuming it
converges with lines downstream of R7 and R8 via L3, Rl -6 activity will increase the
activity on the lines downstream of all of the receptors. If the influence of the line
downstream of pR8/Rl-6 changes with the strength of the signal on the line, then we can
explain how Rl -6 activity can affect the discrimination border. If the pR8/Rl-6 line
enhances phototaxis when activity on the line is relatively low, as is the case in RhImutants, flies will discriminate between wavelengths of 442 nm and below and those at
or above 486 nm. If the pR8/Rl-6 line depresses phototaxis rates when activity on the
Chapter 5. Receptor contributions to opponency
154
line is high, as is the case in wild types, flies will discriminate between UV and bluegreen.
Finally, yR8 cells may have have a response profile similar to pR8s. At moderate
intensities, it appears that yR8s contribute positively to phototaxis, while they may help
to drive down phototaxis rates at higher intensities.
There are a variety of ways in which this activity dependence could be
implemented. The most straightforward is, perhaps, for the the downstream pathways to
divide with one branch contributing positively to phototaxis at low to moderate intensities
and the other branch, active only at higher intensities, contributing negatively to
phototaxis. At low intensities visible light will attract the flies, but at higher intensities
visible light will decrease phototaxic rates.
To test these hypotheses regarding receptor contribution to color vision, it will be
informative to carry out full-intensity and spectrum studies on flies possessing only one
functional opsin. This can be achieved by a selective rescue experiment in NorpA null
mutants. NorpA encodes PLC, an essential element in the phototransduction pathway and
mutants for this gene are blind. A selective single receptor rescue of PLC expression in a
NorpA null background would allow one to determine whether (i) both R7 cells drive
attraction to UV at all intensities, (ii) R8s form an opponent system at higher intensities
and (iii) examine the behavior of Rl -6 to determine how these cells might contribute to
chromatic discrimination.
Chapter 5. Receptor contributions to opponency
155
If (i) is correct flies expressing PLC selectively in R7 cells should be sensitive to
331 and 355 nm preferentially and be attracted to these wavelengths at all intensities
above the detection threshold. If (ii) is correct, yR8 cell PLC rescue flies should be most
sensitive to 515 nm and move away from stimuli of this wavelength at the highest
intensities. Rescuing PLC in pR8 cells should yield flies that are most sensitive to 442 nm
and are attracted to this wavelength across all intensities. Results deviating substantially
from these predictions would indicate that the above hypotheses are not correct and that
the roles of the individual R7 and R8 cell types cannot be inferred directly from
sensitivity profiles of these cells, given the responses of RhI-mutants to different
wavelenths and intensities, as was done above.
The i?/z6-mutant data suggest that the Rl -6 contribution to phototaxis rises at low
intensities and then plateaus at higher levels. This hypothesis can be tested by NorpA
rescue in Rl -6 cells. If the hypothesis is correct, then the response profile at all
wavelengths should parallel the i?/z6-mutant response at 5 1 5 nm.
The results also raise further questions about which features of the phototaxic
circuit support opponency. I suggested that the R8 onto R7 synapse could mediate the
opponency between UV and blue-green light observed in wild types. The results from
this chapter show that, if it contributes to the behavior, the connection between the
synapse and the response may not be straightforward. The synapse opposes blue-green
input from R8 cells with UV input from R7 cells, creating UV/blue-green opponency.
However, the main contrast in RhI -mutants is between blue and green rather than UV
and blue-green. Presumably, a contrast between blue and green results from a comparison
Chapter 5. Receptor contributions to opponency
156
between pathways downstream of pR8 and yR8s rather than as a result of an R8 onto R7
synapse. The UV/blue-green contrast that occurs in wild types emerges only after Rl -6
make contributions downstream of the R8 onto R7 synapse. It is unclear why the
existence of a UV/blue-green synapse is not sufficient for UV/blue-green opponency on it
own without Rl -6 present or why the appearance of UV/blue-green opponency emerges
from Rl -6 input contribution to a system that contains an R8 onto R7 synapse.
Finally, there is the question of why RhI -mutants should have the capacity to
distinguish between blue and green in phototaxis while wild types do not under the
conditions studied. One can only speculate on this point, but one possibility is that the
ability is tied to the adaptation condition of the flies. Rl -6 cells are more sensitive than
R7/8 and can be inactivated by exposure to high intensity light. Under such conditions
wild types may behave like dark-adapated iî/zi-mutants and, therefore, may discriminate
between blue and green. Thus, blue-green discrimination may emerge in wild types
under conditions of extreme light adaptation.
Chapter 6. References
157
Chapter 6. References
Backhaus, W. (1991). Color opponent coding in the visual system of the honeybee.
Vision Res 37(7/8), 1381-1397.
Booth, D., Stewart, A., and Daniel, O. (2004). Colour vision in the glow-worm Lampyris
noctiluca (L.) (Coleóptera: Lampyridae): evidence for a green-blue chromatic
mechanism. J Exp Biol 207, 2373-2378.
Bowmaker, J. K., and Dartnall, H. J. A. (1980). Visual pigments of rods and cones in a
human retina. J Physiol 298, 501-51 1.
Box, G., Hunter, H., and Hunter, J. (1978). Statisticsfor Experimenters: An Introduction
to Design, Data Analysis, and Model Building. New York, Wilely Interscience.
Boynton, R. (1979). Human Color Perception. New York, Holt-Reinhart-Winston.
Calkins, D. J., Tsukamoto, Y., and Sterling P. (1998). Microcircuitry and mosaic of a
blue-yellow ganglion cell in the primate retina. J Neurosci 18(9), 3373-3385.
Chatterjee, S., and Callaway, E. M. (2003). Parallel colour opponent pathways to primary
visual cortex. Nature 426(6967), 668-671.
Cleland, B., and Levrick, W. (1974). Properties of rarely encountered types of ganglion
cells types in the cat's retinal and overall classification. J Physiol 240(2), 457-492.
Conway, B., Moeller, S., and Tsao, D. (2006). Specialized color modules in macaque
extrastriate cortex. Neuron 56(3), 560-573.
Conway, B. (2009). Color vision, cones and color-coding in the cortex. Neuroscientist
75(3), 274-290.
Cronin, T. W., and Marshall, N. J. (1989). A retina with at least ten spectral types of
photoreceptors in a mantis shrimp. Nature 339, 137-140.
De Valois, R. L. (1965). Analysis and coding of color vision in the primate visual system.
Cold Spring Harbor Symp Quant Biol 30, 568-587.
De Valois, R. L., and De Valois, K. K. (1993). A multistage color model. Cold Spring
Harbor Symp Quant Biol 33, 1053-1065.
Fischbach, K.-F., and Reichert, H. (1978). Interactions of visual subsystems in
Drosophila melanogaster: a behavioral genetic analysis. Biol Behav 4, 305-317.
Fischbach, K.-F. (1979). Simultaneous and successive colour contrast expressed in 'slow'
phototaxic behavior of walking Drosophila melanogaster. J Comp Phys 130(2), 161-171.
Chapter 6. References
1 58
Gao, S., Takemura, S., Ting, C, Huang, S., Lu, Z., Luán, H., Rister, J., Thum, ?., Yang,
M., Hong, S., Wang, J., Odenwald, W., White, B., Meinertzhagen, I., and Lee, C-H.
(2008). The neural substrate of spectral preference in Drosophila. Neuron 60(2), 328-342.
Hamdorf, K., Hochstrate, P., Höglund, G., Moser, M., Sperber, S., and Schlecht, P.
(1992). Ultra-violet sensitizing pigment in the blowfly photoreceptors Rl -6; probable
nature and binding sites. J Comp Phys A 1 71, 601-615.
Hardie, R. C, and Kirschfield, K. (1983). Ultraviolet sensitivity of fly photoreceptors R7
and R8: evidence for a sensitizing function. Biophysics of Structure & Mechanism 9,
171-180.
Hardie, R. C. (1985). Functional organization of the fly retina. Progress in Sensory
Physiology 5. D. Ottoson. New York, 2-79.
Harris, W. A., Stark, W.S., and Walker J.A. (1976). Genetic dissection of the
photoreceptor system in the compound eye of Drosophila melanogaster. J Comp Phys
256(2), 415-439.
Heisenberg, M., and Büchner, E. (1977). The role of retinula cell types in visual behavior
of Drosophila melanogaster. J Comp Phys 117(2), 127-162.
Hering, E. (1964). Outlines ofa Theory ofLight Sense. Cambridge, MA., Harvard
University Press.
Hernandez de Salomon, C, and Spatz, H. (1983). Colour vision in Drosophila
melanogaster: wavelength discrimination. J Comp Phys 750(1), 31-37.
Hu, K., and Stark, W.S. (1977). Specific receptor input into spectral preference in
Drosophila. J. Comp Phys 121(2), 241-252.
Jacob, K. G., Willmund, R., Folker, E., Fischbach, K.F., and Spatz, H. (1977). T-maze
phototaxis of Drosophila melanogaster and several mutants in the visual systems. J Comp
Phys 116(2), 209-225.
Jacobs, G. H. (1996). Primate photopigments and primate color vision. PNAS 93(2), 577581.
Jacobs, G. H., Fenwick, J.A., and Williams, G.A. (2001). Cone-based vision of rats for
ultraviolet and visible lights. J Exp Biol 204(14), 2439-2446.
Jacobs, G. H. (2009). Evolution of color vision in mammals. Phil Trans Royal Society B
364, 2957-2967.
Chapter 6. References
159
Kinoshita, M., Shimada, N., and Arikawa, K. (1995). Color vision of the foraging
swallowtail butterfly papilio xuthus. J Exp Biol 202(2), 95-102.
Kirschfeld, K., and Lutz, B. (1977). The spectral sensitivity of the ocelli of Calliphora
(Diptera). Z. Naturforsch 52(c), 439-441.
Kirschfeld, K., Franceschini, N., and Minke, B. (1977). Evidence for a sensitizing
pigment in fly photoreceptors. Nature 269(5627), 386-390.
Kirschfeld, K., Feiler, R., and Franceschini, N. (1978). A photostable pigment within the
rhabdomere of fly photoreceptors no 7. J Comp Phys 125, 275-284.
Le Grande, Y. (1957). Light, Color and Vision. London, Chapman and Hall, LTD.
Lee, B. (2004). Paths to colour in the retina. Clin Exp Optom 57(4-5), 239-248.
Lee, B., Smith, V., Pokorny, J., and Sun, H. (2008). Chromatic adaptation in red-green
cone-opponent retinal ganglion cells of the macaque. Vision Res 48(26), 2625-2632.
Lee, T.-W., Wachtler, T., and Sejnowski, T. J. (2002). Color opponency is an efficient
representation of spectral properties in natural scenes. Vision Res 42, 2095-2103.
Martin, P. (2004). Colour through the thalamus. Clinical and Experimental Optometry
55(4-5), 239-257.
Meinertzhagen, I. A. (1973). Development of the compound eye and optic lobe of insects.
Developmental neurobiology ofarthropods. D. Young. Cambridge, Cambridge
University Press, 51-104.
Menzel, R., and Blakers, M. (1975). Colour receptors in the bee eye — morphology and
spectral sensitivity. J Comp Phys A 705(1), 11-13.
Miller, G., Hansen, K., and Stark, W. (1981). Interaction among ocellar and compound
eye receptors. Journal of Insect Physiology 27(1 1), 813-819.
Osorio, D., Vorobyev, M.,and Jones, C. (1999). Colour vision of domestic chicks. J Exp
Biol 202, 2951-2959.
Packer, O. S., Verweij J., Li, P., Schnapf, J., and Dacey, D. (2010). Blue-yellow
opponency in primate S cone photoreceptors. J Neurosci 30(2), 568-572.
Palacios, ?., Martinoya, C, Bloch, S., and Várela, F. (1990). Color mixing in the pigeon:
a psychophysical determination in the longwave spectral range. Vision Res 30, 587-596.
Palacios, A. G., and Várela, F. J. (1992). Color mixing in the pigeon (Columbia livia).
Vision Res 32, 1947-1953.
Chapter 6. References
160
Patterson, W., McDowell, A., Hughes, A., and Bilotta, J. (2002). Opponent and non-
opponent contribution to zebrafish electroretinographs using heterochromatic flicker
photometry. J Comp Phys 188(4), 283-293.
Percival, K. A., Jusuf, P.R., Martin, P.R., and Griinert, U. (2009). Synaptic inputs onto
small bistratified (blue-ON/yellow-OFF) ganglion cells in marmoset retina. J Comp
Neurol 577(5), 655-669.
Quinn, W. G., Harris, and W., Benzer, S. (1974). Conditioned behavior in Drosophila
melanogaster. PNAS 77(3), 708-712.
Salcedo, E., Huber, ?., Henrich, S., Chadwell, L.V., Chou, W.H., Paulsen, R., and Britt,
S.G. (1999). Blue- and green-absorbing visual pigments of Drosophila: ectopic
expression and physiological characterization of the R8 photoreceptor cell-specific Rh5
and Rh6 rhodopsins. J Neurosci_79(24), 10716-10726.
Schlumperli, R. A. (1973). Evidence for color vision in Drosophila melanogaster through
spontaneous phototaxic choice behavior. J Comp Phys 86(1), 77-94.
Smith, E. L., Greenwood, V. J., and Bennett, A. T. D. (2002). Ultraviolet colour
perception in European starlings and Japanese quail. J Exp Biol 205, 3299-3306.
Solomon, S., and Lennie, P. (2007). The machinery of color. Nat Rev Neurosci #(4), 276286.
Stark, W. S., Walker, J.A., and Harris, W.A. (1976). Genetic dissection of photoreceptor
system in the compound eye of Drosophila melanogaster. J Physiol 256(2), 415-439.
Stark, W. S., Ivanyshyn, A.M., and Hu, K.G. (1976). Spectral sensitivities and
photopigments in adaptation of fly visual receptors. Naturwissenschaflen 63, 513-518.
Tang, S., and Guo, A. (2001). Choice behavior ?? Drosophila facing contradictory visual
cues. Science 294(5546), 1543-1547.
Tomlinson, A., and Ready, D.F. (1987). Cell fate in the Drosophila ommatidium.
Developmental Biology 123(1), 264-275.
Ventura, D., Zana, Y., de Souza, J., and DeVoe, R. (2001). Ultraviolet color opponency
in the turtle retina. J Exp Biol 204(14), 2527-2534.
Verweij, J., Hornstein, E.P., and Schnapf, J.L. (2003). Surround antagonism in macaque
cone photoreceptors. J Neurosci 23(32), 10249-10257.
Von Frisch, K. (1971). Bees: their vision, chemical sense, and language. Ithaca, Cornell
University Press.
Chapter 6. References
Wakakuwa, M., Kurasawa, M., Giurfa, M., and Arikawa, K. (2005). Spectral
heterogeneity of honeybee ommatidia. Naturwissenschaften 92(10), 464-467.
Yamaguchi, S., Wolf, R., Desplan, C, and Heisenberg, M. (2008). Motion vision is
independent of color in Drosophila. PNAS 705(12), 4910-4915.
Yamaguchi, S., Desplan, C, Heisenberg, M. (2010). Contribution of photoreceptor
subtypes to spectral wavelength preference in Drosophila. PNAS 707(12), 5634-5639.
161
Документ
Категория
Без категории
Просмотров
0
Размер файла
5 572 Кб
Теги
sdewsdweddes
1/--страниц
Пожаловаться на содержимое документа