J Sci Food Agric 1998, 78, 349È358 Quantitative NMR Imaging of Kiwifruit (Actinidia deliciosa) during Growth and Ripening Christopher J Clark,1* Lynley N Drummond2 and Janet S MacFall3 1 HortResearch, Palmerston North Research Centre, Private Bag 11 030, Palmerston North, New Zealand 2 Department of Food Technology, Massey University, Private Bag 11 222, Palmerston North, New Zealand 3 Department of Biology, Elon College, PO Box 2225, Elon College, North Carolina 27244, USA (Received 16 May 1997 ; revised version received 9 February 1998 ; accepted 11 March 1998) Abstract : Quantitative 1H magnetic resonance (MR) imaging was used to determine relaxation changes (T , T -CPMG) at regular intervals during growth and 1 2deliciosa var deliciosa). Temporal trends and difripening of kiwifruit (Actinidia ferences between Ñesh, locule and core tissue were found for both relaxation parameters. However, no consistent associations were found between nondestructive measurements and those for individual free sugars, soluble solids content (SSC) and macronutrients and micronutrients determined on dissected companion samples. Increases of 200% in total free sugar concentration in Ñesh and 68% in SSC accompanied starch hydrolysis after harvest. Despite the magnitude of these changes, relaxation times remained unaltered. These observations were repeated in a second investigation using A arguta fruit and T , T , T 2 2a CPMG and self-di†usion image contrasts. Here, SSC increased 125%1 during compressed 15-day ripening period, while MR parameters like self-di†usion declined only 7È14% from harvest values. T -CPMG relaxation was also investi2 gated in aqueous solutions containing individual organic acids, sugars or pectate and juice from ripening fruit (4É7È15É5% SSC). Analysis of solutions and juices showed relaxation is indeed sensitive to increases in sugar composition but relatively insensitive to changes in organic acids and soluble pectin at concentrations normally found in fruit. Results imply that relaxation parameters determined from MR images may not be appreciably inÑuenced by processes that cause solution composition to vary dramatically, even though these changes are reÑected in the relaxation properties of the juice itself. Possible reasons for this are discussed with regard to the impact of cell structure and magnetic Ðeld strength on relaxation processes. ( 1998 Society of Chemical Industry. J Sci Food Agric 78, 349È358 (1998) Key words : Actinidia deliciosa ; fruit ; kiwifruit ; magnetic resonance imaging ; ripening. INTRODUCTION detailed representations of vascular architecture extracted from three dimensional data arrays (McFall and Johnson 1994). Visual inspection of morphology is typical of the way in which MRI is used to gather qualitative information from samples. Obtaining quantitative data concerning the underlying MR parameters that provide contrast in these images, on the other hand, requires a di†erent approach. This is achieved most successfully using quantitative MRI. Quantitative imaging is an approach whereby images are constructed that contain the spatial distribution of Since the seminal study on the use of proton magnetic resonance imaging (1H MRI) to investigate watercore in apple (Wang et al 1988), application of MRI to horticultural produce has evolved from simplistic descriptions of images of samples containing quality disorders, to * To whom correspondence should be addressed. Contract/grant sponsor : NZ Foundation for Research, Science and Technology. Contract/grant number : C06504 349 ( 1998 Society of Chemical Industry. J Sci Food Agric 0022È5142/98/$17.50. Printed in Great Britain C J Clark, L N Drummond, J S MacFall 350 static and dynamic characteristics of aqueous protons, such as their relaxation and di†usive properties. Signal intensity (S) in a standard NMR experiment depends on a number of factors : S \ N [1 [ 2 expM[(T [ T /2)/T N H R E 1 ] exp([T /T )] ] exp([T /T ) (1) R 1 E 2 Here, T (ms), the time between the 90¡ radiofrequency E pulse and formation of the spin-echo signal, and T R (ms), the recovery time between pulse train repetitions, are two instrumental parameters, and N , T (ms) and H 1 T (ms) are three sample-dependent parameters describ2 ing the proton spin-density, the spin-lattice relaxation time and the spin-spin relaxation time, respectively (Callaghan 1991). SimpliÐcation of this expression by choosing appropriate values of T (or T ), and analysing R E the variation in signal intensity as the other instrument parameter is varied incrementally, leads to the formation of “calculatedÏ images containing relaxation or proton spin-density data (MacFall JR et al 1987). Similar procedures enable images containing selfdi†usion and Ñow data to be constructed (Callaghan et al 1994 ; Kuchenbrod et al 1995, 1996). Evidence suggests that MR parameters reÑect physiological changes in fruit during ripening or storage (Callaghan et al 1994 ; Ishida et al 1994 ; Clark and MacFall 1996 ; Goodman et al 1996). To date, there has been no concerted attempt to link changes in MR parameters obtained from quantitative images to variation in other physicochemical properties, like carbohydrate composition, which would be expected to a†ect the magnetic characteristics of cell solutions. Di†erence in soluble solids content (SSC), for example, is currently being explored as the basis for an MR sensor to assess sweetness of fruit (Zion et al 1995). To further understanding of how MR parameters vary in biological systems, and to determine whether these might be a useful non-destructive sensor of physiological processes in fruit, investigations where quantitative imaging was used to monitor changes in MR measurements in the fruit and juice of kiwifruit (Actinidia deliciosa) during growth and ripening are reported. Additional experiments were also carried out on an A arguta fruit, a small-fruited relative of kiwifruit where the ripening period is compressed, and binary aqueous solutions containing compounds commonly found in fruit juices. Reasons for selecting Actinidia as a model exploratory system were twofold : Ðrstly, both species accumulate large amounts of granular starch during development, then hydrolyse it extensively to soluble free sugars following harvest (MacRae et al 1992) ; secondly, a limited preliminary study imaging a surrogate, a small hermaphrodite male kiwifruit, indicated that signiÐcant changes in MR properties and strong associations with compositional measurements might be anticipated (Callaghan et al 1994). EXPERIMENTAL Samples Commercial fruit Fruit were obtained from 11-year-old vines (cv Hayward) grown on a T-bar trellis system in a research orchard near Hamilton (North Island, New Zealand). Commencing 13 December 1993, approximately 2 weeks after full bloom, between 45 and 160 fruit were harvested regularly throughout the growing season. At commercial harvest (9 May 1994 ; 6É2% SSC), twice as many fruit were sampled. One half was analysed immediately, while the other was allowed to ripen (2 weeks storage at 4¡C followed by ethylene-induced ripening for 2É5 weeks at 20¡C). At each sampling date, fruit were sliced laterally to divide the length of the stem axis into three equal parts. These basal, median and distal sections were further dissected into the skin plus outer pericarp (Ñesh), the inner pericarp (locules plus seeds) and core. Depending on stage of development, segments from 13 to 45 fruit were bulked to provide sufficient mass for analysis. This procedure was repeated to give three replicates of each sample type at every harvest data. Fresh-weight samples were subsequently weighed, frozen, lyophilised, reweighed and ground to a powder. Small fruit Fruit from an A arguta selection (J5I3) (\2É5 cm diameter) were harvested on 18 March 1996. To prolong storage and prevent shrivelling, the mature fruit were sealed in small plastic containers. SSC (Atago refractometer) was determined 1, 4, 7, 10, 15 and 21 days after harvest to coincide with imaging, and three replicates consisting of four fruit each were retained for compositional analysis. Whole-fruit were processed for analysis in the same way as larger fruit. However, because of limited fruit numbers and their small size, individual tissues were not dissected in this experiment. Juice and solutions At weekly intervals prior to commercial harvest, juice was squeezed from 10-mm segments cut from the top and bottom of individual fruit (cv Hayward). After commercial harvest, changes were monitored over 4 months using fruit held in storage (0¡C). On each occasion, juice from 10 fruit was bulked to provide sufficient solution (10 ml) for spectroscopic analysis. Solutions were frozen at [ 80¡C until required. Samples were subsequently centrifuged prior to use to remove chlorophyll and denatured protein. Solutions containing soluble compounds typically found in fruits were prepared in distilled water. Concentration ranges were decided on the basis of a survey of compositional analyses of a number of fruits and at different stages during their development. Thus, the ranges chosen reÑect upper limits encountered in a sample NMR imaging of kiwifruit spectrum, not just those in kiwifruit or at harvest. Materials and concentration ranges used included the following : sugarsÈglucose, fructose and sucrose, 5, 10 and 20% (w/v) ; polyolsÈinositol 5 and 10% (w/v) and sorbitol 5, 10 and 20% (w/v) ; organic acidsÈmalic 1 and 2É5% (w/v), citric and quinic 2É5 and 5% (w/v) ; and pectinÈNa CO -soluble kiwifruit pectin 0É5 and 2É0% 2 3 (w/v). Pectin was kindly supplied by Robert Redgwell. Details of the extraction and composition of this material are in Redgwell et al (1992). Analytical methods Measurements on freeze-dried powders included the following : determination of macronutrients and micronutrients in nitric/perchloric-acid digests by standard atomic absorption spectroscopy ; starchÈcolorimetric detection of glucose after hydrolysis with amyloglucosidase (Smith et al 1992) ; and individual sugars after extraction and derivatisation for analysis by GC (Bieleski 1994). Imaging techniques Commercial fruit Proton NMR experiments on commercial fruit were performed in a 2 T horizontal widebore magnet at Duke University, NC, USA. Four fruit were imaged at each harvest date. These were despatched by international courier within 2 h after picking. Because of the length of time in transit, however, NMR experiments did not begin until 1 week later and were generally completed within 2 or 3 days thereafter. Three small glass capillary tubes containing CuSO 4 internal standards were taped along the top and along the side of the fruit, so that they would appear in images of the median longitudinal and transverse planes. Fruit were positioned in a 10-cm-diameter “birdcageÏ coil with the long axis of the fruit parallel to the direction of the bore. At every imaging session, 28 single-slice, spin-echo images were obtained from each fruit. Six images using a standard spin-echo spin warp pulse sequence were sequentially acquired through each transverse and longitudinal slice plane for calculation of T values (echo 1 time (T ) \ 10 ms ; repetition time (T ) ranging from 100 E R to 3200 ms). A further eight images were acquired through each slice plane for calculation of T values 2 using a CarrÈPurcellÈMeiboomÈGill (CPMG) pulse sequence (T \ 20È160 ms in 20-ms increments ; T \ E R 3200 ms). Individual image slices were all 2-mm thick with an in-plane resolution of between 0É12 and 0É33 mm (256 ] 256 data arrays)Èresolution decreased as fruit became progressively larger. Relaxation times were obtained from the “calculatedÏ images arising from the Ðtting procedures of MacFall 351 et al (1987). Measurements in the transverse plane were from circular regions of interest (ROIs) randomly selected within the Ñesh, locules (but not seeds) and core. Those in the longitudinal plane consisted of ROIs from fresh in basal, median and distal regions of the fruit. Relaxation measurements for a particular tissue were calculated as the mean of six ROI measurements, each in turn the average of relaxation data accumulated over 137 (Ñesh), 21 (locule) or 80 (core) pixels. All measurements were normalised according to the T and T 1 2 values calculated for two of the reference tubes associated with each image, to remove any likelihood of instrument variability over the course of the study. Small fruit A single fruit was selected for the 3-week experiment involving A arguta. At each imaging session, MR experiments utilising a 7É1 T 300 MHz Bruker AMX-300 NMR spectrometer were carried out to produce a range of image contrasts. These included standard spin-echo spin warp (T and T -SE), a T 1 2 2 CPMG sequence, gradient echo (T *), and di†usion con2 trast. The pulse sequences and instrument parameters used are described in Callaghan et al (1994), as are the techniques employed subsequently to produce the “calculatedÏ images containing the quantitative data for each contrast. Images, obtained from 1-mm-thick slices, were all reconstructed and displayed on a 256 ] 256 pixel array giving an in-plane resolution of 98 lm. Juice and solutions Transverse relaxation was determined as a function of pulse-width spacing (PWS) in a CPMG pulse sequence following the method of Hills and Duce (1990). Data were acquired using the 300-MHz AMX spectrometer at Ðxed 90È180¡ pulse spacings varying between 44 ls and 20 ms. The T time associated with each PWS was 2 determined from linear regression by modelling the logarithmic transform of that region of the data on the free induction decay curve associated with aqueous protons. All measurements were on sample volumes of about 1 ml at 25¡C. Transverse relaxation data are presented as the reciprocal of transverse relaxation (or relaxation rate) vs the reciprocal of PWS (see Figs 6 and 7). In this form the dispersion data are able to be modelled using the two-site chemical exchange model of Carver and Richards (1972), where the shape of the relationship is expressed as a function of the concentration of exchangeable solute protons, amongst other factors. RESULTS Commercial fruit (A deliciosa) Variation in the intensity of individual tissues in calculated images (Figs 1 and 2) indicated the existence of 352 C J Clark, L N Drummond, J S MacFall Fig 1. Calculated spin-lattice relaxation (T ) images through a longitudinal equatorial plane of kiwifruit (A deliciosa) during 1 development and ripening. Numbers correspond to weeks from fruitset, with commercial harvest occurring at week 23. Image resolution varies from 0É12 in the youngest fruit to 0É33 mm in the older samples (256 ] 256 pixel data arrays). The circular internal standard tubes (4 mm inner diameter) serve as a scale marker for each panel. Image intensity is based on a continuous grey-scale setting ranging from 0 ms \ black to 2500 ms \ white. Note, images at 23 and 27É5 weeks are of the same fruit at harvest and after 4É5 weeks ripening. distinct di†erences between the relaxation properties of the Ñesh, locule and core tissues at particular sampling dates, as well as trends during the season. Digital information extracted from discrete ROIs from images such as these formed the basis for quantitative analysis of these di†erences and trends. The shortest relaxation times (either T or T -CPMG) 1 2 were those measured in the core (Fig 3). T times for 1 Ñesh and locule tissue increased during fruit development and were of a similar magnitude at individual harvest dates. In contrast, T -CPMG times di†ered 2 widely among the three tissues from around 12 weeks after fruitset, proving this to be a more sensitive indicator of changes in magnetic environment than T . 1 There were no di†erences of any consequence amongst the relaxation properties in basal, median and distal sections of the fruit. Trends at either extremity adhered closely to those outlined for the median slice in Fig 3. Thus, while gradients in nutrient composition between the top and bottom of the fruit exist (Clark and Fig 2. Calculated spin-spin relaxation (T -CPMG) images through a transverse equatorial plane of kiwifruit (A deliciosa) during 2 development and ripening. Individual images are from the same fruit depicted in Fig 1. Numbers, resolution and scale markers are the same as in Fig 1. Image intensity is based on a continuous grey-scale setting ranging from 0 ms \ black to 350 ms \ white. NMR imaging of kiwifruit 353 Fig 3. Seasonal variation of (A) the calculated T relaxation time, (B) the calculated T -CPMG relaxation time1and (C) the distribution of total free2sugar concentration in Ñesh, locule and core tissue from a median transverse slice of kiwifruit (A deliciosa). In (A) and (B), error bars are an LSD based on a pooled estimate of variance (P \ 0É05) for between-tissue comparisons at any imaging date, and data points are the mean of six ROI measurements on each of four fruit. In (C), error bars are LSDs (P \ 0É05) for between-tissue comparisons at speciÐc times, and data points are the mean of three replicates. Arrows indicate commencement of commercial harvest. Smith 1988), they appear to have no signiÐcant inÑuence on the magnitude of relaxation properties. In fact, no consistent, strong associations between composition and relaxation times were established for any tissue components. Where strong associations were indicated, these tended to be dependent on the inÑuence of a single data point and were discounted. T relax1 ation time in the basal, median or distal regions of Ñesh, for example, was found to be negatively correlated (P \ 0É05) with moisture content (%) and the concentrations of P, Ca, Cu or Fe. However, the same relationships were absent for locule and core tissue. Other correlations were found between measurements in either a particular tissue or at particular times. T and T 1 2 CPMG were correlated with Ñesh N concentration in median sections, but not others, while T was negatively 1 correlated with total free sugars at weeks 8 and 23 only. Many of these associations appear coincidental and make no obvious physiological sense. Further analysis using stepwise regression was precluded in that measurements were only recorded on six occasions during the season. Dramatic changes in relaxation times following harvest were entirely absent (Fig 3, Table 1). Free and total sugars increased between 100 and 300% after harvest and soluble solids by 68%, as expected (Fig 3, Table 1). However, this marked alteration in solution composition was not reÑected in detectable changes in relaxation properties. Small fruit (A arguta) Visual inspection of images from the 21-day A arguta ripening experiment implied that changes in relaxation times were minimal. Examples of calculated images produced from each pulse sequence experiment are shown for day 4 data only (Fig 4). The images are notable for the lack of contrast between locule and Ñesh (cf Fig 2). The darker intensity of the core suggests a longer T 1 relaxation time or a shorter transverse relaxation time relative to the other two tissues. In the case of the selfdi†usion coefficient (D), the darker intensities are consistent with increased solution viscosity or slower di†usion. With the exception of the seeds and core, the TABLE 1 Relationship between maturation status, type of tissue and magnetic and compositional properties of unripe kiwifruit (A deliciosa) at commercial harvest (6É2% soluble solids content) and at eating ripeness 1 month latera Parameter T 1 T -CPMG 2 Fructose Glucose Sucrose Total soluble sugars Total soluble solidsb (%) Flesh L ocule Core Unripe Ripe P Unripe Ripe P Unripe Ripe P 1610 201É4 13É8 12É0 5É4 32É4 6É3 1809 215É9 45É7 41É8 16É7 106É9 10É6 NS NS ** NS ** * *** 1774 344É3 10É9 10É1 5É7 28É0 1792 354É6 34É3 32É0 11É0 79É6 NS NS ** *** * ** 874 88É9 14É2 12É3 7É6 37É9 1370 105É8 49É2 40É4 29É2 126É1 NS NS *** *** * *** a Relaxation data were collected from the same four fruit on each occasion. Probability (P) is based on a t-test with n \ 6 degrees of freedom for relaxation measurements and n \ 4 for sugars. Units are ms for relaxation and mg g~1 dry weight for carbohydrates. b Whole fruit measurement. 354 C J Clark, L N Drummond, J S MacFall Fig 4. Tissue contrasts in A arguta derived from di†erent pulse sequence experiments. Calculated images show spin-lattice relaxation (T , 0È2500 ms), spin-spin relaxation (T se, 0È200 ms), gradient echo (T *, 0È4 ms), spin-spin relaxation using a CPMG pulse 2 2 obtained from the T -CPMG sequence, and the sequence1 (T cp, 0È300 ms), pseudo proton spin-density (0È2000 arbitrary units) 2 2 self-di†usion coefficient for water (0È3É5 ] 10~9 m2 s~1), respectively. Ranges in parentheses refer to values delineating the greyscale setting in each panel. All images are of the same fruit at 4 days after harvest and are based on 256 ] 256 pixel data arrays (resolution 98 lm), except the self-di†usion image which is 128 ] 128. The scale marker (5 mm) in the Ðrst image applies to each panel. pseudo proton spin-density map (Fig 4) indicates that the distribution of water is uniform throughout the inner and outer pericarp. Relaxation times in the gradient-echo (T *) experiments proved too short 2 (\5 ms) to be adequately measured. Examination of quantitative measurements conÐrmed that changes in relaxation were small, and, where there were temporal trends, these were gradual rather than abrupt (Fig 5). SpinÈspin relaxation (T -SE) proved to 2 be the most sensitive technique for discriminating between various tissues. As with the larger fruit, core tissue had the shortest relaxation time relative to the other tissues, and the slowest D coefficient (Fig 5). Measurements were compared 1 day after harvest and after 15 days ripening when the fruit were eatingripe (by day 21 fruit were too soft and over-ripe) (Table 2). As expected, the SSC of companion A arguta fruit increased sharply from 8É2 to 18É6% (P \ 0É001) during ripening. In contrast to this 125% increase in soluble solids, variation in the MR parameters was small : T 1 increased by 11%, but only in the core ; T -SE decreased 2 by 22È25% in the Ñesh and core ; and T -CPMG 2 decreased by 10% in the core (Table 2). Di†usion was the single measurement in which all three tissues demonstrated a consistent e†ect, declining 7È14% from their harvest values. Juice and solutions Fig 5. Variation of (A) the calculated T -SE relaxation time 2 and (B) the self-di†usion coefficient (units ] 1012 m2 s~1) in Ñesh, locule and core tissue from a median transverse slice of an A arguta fruit during ripening. Error bars are LSDs for between-tissue comparisons at individual imaging dates. Data points are the mean of eight separate ROI measurements. The appearance of the relaxation curves indicated that fast proton exchange between water and solute was the dominant mechanism involved in T relaxation in each 2 of the individual solution studies (Hills et al 1989 ; Hills and Duce 1990). The salient features have been summarised using a subsample of the experiments carried out (Fig 6). Pectates are released from cell walls during the ripening processes (Redgwell et al 1992). While their concentration in solution is not high (typically 0É08È0É5% (w/v)), they may have a profound e†ect on solution properties. At a concentration of 0É5% (w/v) and pH 4É2, the relaxation properties of pectate solution NMR imaging of kiwifruit 355 TABLE 2 Relaxation (ms) and self-di†usion (]1012 m2 s~1) in an A arguta fruit, 1 and 15 days after harvesta Parameter T 1 T -SE 2 T -CPMG 2 Self-di†usion Flesh L ocule Core Day 1 Day 15 P Day 1 Day 15 P Day 1 Day 15 P 1339 128É0 141É7 2003 1362 100É4 135É5 1860 NS ** NS *** 1352 133É9 166É4 1929 1355 139É7 141É9 1656 NS NS NS *** 949 106É1 69É3 1465 1050 79É9 62É5 1253 ** *** ** ** a The probability (P) is based on a paired t-test with n \ 7 degrees of freedom. (D1800 ms) were very similar to that of pure water (D2000 ms) (Fig 6A). Addition of acid to adjust this pectate solution to pH 3, or increasing the concentration of pectate itself to 2É0% (w/v), caused formation of a viscous gel. In both cases, relaxation times declined to around 680 and 500 ms, respectively. Spin-spin relaxation of water and pectate solutions/gels showed no dependence on PWS. Solutions of the weak acids, malic (pK \ 3É4) and 1 quinic (pK \ 3É7), were prepared at pH 4 by neutral1 isation with KOH. Equivalent concentrations (1% w/v) had di†erent relaxation times (2500 vs 2100 ms) with relaxation times declining to 2400 ms (2É5% (w/v) malate) and 1500 ms (5% (w/v) quinate) at the higher concentrations associated with fruit. In each case, relaxation was independent of PWS. The fact that the two 1% (w/v) solutions do not have the same transverse Fig 6. SpinÈspin relaxation rates (1/T ) for solutions containing compounds found in fruit juices as2a function of reciprocal 90È180¡ CPMG pulse-width spacing (PWS) : (A) Na CO 3 extractable pectin from kiwifruit ; (B) organic acids ; (C)2 fructose at di†erent concentrations ; (D) fructose, glucose and sucrose ; (E) e†ect of pH on citric acid ; and (F) e†ect of pH on sucrose. relaxation time is not surprising. With four potentially exchangeable hydroxyl protons on quinate as opposed to one in malate, on a molar basis, 1% (w/v) quinate solution has almost three times as many exchangeable protons as malate. The observed relaxation rate (1/T ) 2 would be expected to be faster in the solution with the greater number of exchangeable solute protons (Carver and Richards 1972). In contrast, relaxation properties of sugar solutions varied depending on the molecule, its concentration and the PWS selected (Fig 6C, D). For example, relaxation times in 5, 10 and 20% (w/v) fructose solutions ranged between 100È2000 ms, 50È1700 ms and 20È1400 ms, longer relaxation times being associated with shorter PWS. This pattern was evident for other free sugars, plus inositol and sorbitol (data not presented). The 20% fructose solution was unique, however, with respect to having exceptionally short relaxation times below PWS of 2É5 ms. In other simple carbohydrate systems, relaxation times were generally stable below 2É5 ms (cf Fig 6D). Samples of equivalent concentration (10% (w/v)) show clearly that relaxation is a†ected by the type of molecule present, even when they have the same number of exchangeable protonsÈas with fructose and glucose (Fig 6D). Here, relaxation times in fructose (50È 1700 ms) were shorter than those in glucose (74È 1700 ms) and sucrose (100È2000 ms). Acidity also a†ected relaxation rate (Fig 6E, F). An increase in acidity of a 1% (w/v) citrate solution from pH 4 to 3, caused relaxation times to decrease from 2400 to 2300 ms (Fig 6E). In the case of sucrose (10% (w/v)), increasing acidity from pH 5 to 3 changed relaxation times from between 100 and 2000 ms to between 520 and 700 ms, at the same time removing the dependency on PWS (Fig 6F). For most PWS, this represents an increase in the relaxation time, the opposite of the trend for citric acid. The relaxation rate of kiwifruit juice increased with increasing SSC and showed a marked dependence on PWS at SSC greater than about 8É4% (Fig 7). Fructose and glucose are the major components in these juices. Solution pH (unadjusted) averaged 3É3 and ranged between 3 and 4. Perseverance of a pulse-width dependence in these juice extracts (despite their high acidity) C J Clark, L N Drummond, J S MacFall 356 Fig 7. SpinÈspin relaxation rates (1/T ) for kiwifruit juice 2 obtained from fruit at di†erent stages of ripening (soluble solids content) as a function of reciprocal 90È180¡ CPMG pulse-width spacing (PWS). indicates that acid does not inÑuence relaxation is this more complex system to the same extent as its addition does in single component systems. DISCUSSION The results of quantitative MR imaging during growth and ripening of kiwifruit indicate that there are distinct tissue and seasonal di†erences in the relaxation measurements, but the measurements do not necessarily reÑect underlying biochemical processes. In commercial-sized fruit, it is quite clear that the relaxation parameters determined from quantitative images are insensitive to major physiological events that have a dramatic impact on solution composition. Furthermore, the lack of change in the relaxation measurements postharvest cannot be attributed to compounds present in the juice itself. The observation that shorter relaxation times are associated with increasing SSC in juices, or with increasing concentration in binary aqueous solutions, is consistent with the Carver and Richards (1972) formulation, in which T is inversely related to the con2 centration of exchangeable solute protons, and experimental results relating relaxation and carbohydrate concentration (Mora-Gutierrez and Baianu 1989 ; Birch and Karim 1992). Failure to obtain these results via the imaging process indicates that other factors must be involved. Two factors that might contribute to insensitivity of the imaging data to changes in solution composition are choice of PWS and magnetic Ðeld strength. The sugar solution studies indicate that relaxation becomes less dependent on concentration at PWS \ 100 ls. This can be discounted here, however, since the T -CPMG 2 imaging experiments utilised a PWS of 2É5 ms (1/ PWS \ 400 s~1). This is where di†erences between individual sugars or change in concentration would be expected to have a substantial impact on T relaxation 2 (Fig 6D). T dispersion e†ects also originate from 2 exchange of protons between sites with di†erent chemical shift values (eg sugar hydroxyl protons vs aqueous protons). The absolute value of the chemical shift di†erence between such sites is, in turn, strongly inÑuenced by the choice of magnetic Ðeld strength. Thus, di†erences relating to solution composition observed at 7É1 T (Fig 6) will be much less at 2 T, the Ðeld strength at which commercial fruit were imaged. This may account for lack of observation of signiÐcant T changes 2 in commercial-sized fruit postharvest. It does not account for the fact that the authors failed to observe T 2 changes during ripening of the small A arguta fruit, which was imaged at the same Ðeld strength as the binary solutions. Other possibilities relate to modulation of relaxation by a sampleÏs structural features. Intercellular air spaces, for example, produce localised magnetic Ðeld gradients at solutionÈair interfaces. With a pulse sequence like T -SE, di†usion of water molecules 2 through these gradients enhances the rate of relaxation and may produce susceptibility artefacts in the image itself (Callaghan 1991). Transverse relaxation in parenchyma of courgette, onion and apple have been described using combinations of fast proton exchange between water and solutes and di†usion through the internally generated Ðeld gradients at cellÈair interfaces (Hills and Duce 1990). Di†usion between intracellular and intercellular aqueous-Ðlled compartments also averages signal intensity. Hills and Duce (1990) combined these factors in a two-compartment cellular relaxation model based on chemical and di†usive exchange. Semi-quantitative Ðtting of their fruit and vegetable data indicated that cell morphology and membrane permeability strongly inÑuenced T relaxation. Links 2 between the multi-exponential nature of transverse relaxation in biological cells and cell geometry have also been proposed by Brownstein and Tarr (1979). From their model, based solely on the properties of bulk water and characterisation of the cell wall as a surface sink, it is possible to infer estimates of both the shape and size of the cells from the NMR data. Detailed information concerning the size, shape, number and ultra-structure of kiwifruit cells during growth and ripening have been documented (Hopping 1976 ; Hallett et al 1992). Further experiments with large fruit at both low and high Ðeld strengths, and modelling the pulsewidth dependence of transverse relaxation data from di†erent tissues during development and ripening (Hills and Duce 1990), is an obvious path forward in order to probe the compositional and structural ramiÐcations outlined here. Nevertheless, where changes in relaxation measurements are observed by MRI which are likely to be related to speciÐc physiological events (Callaghan et al 1994 ; Clark and MacFall 1996), the solution studies contain information relevant to their interpretation. For instance, organic acids at concentrations commonly found in fruits, and kiwifruit pectate solution, have long relaxation times similar to bulk water (Fig 6A, B). Hence, concentration changes involving these compounds should have little e†ect on MR data, except in NMR imaging of kiwifruit circumstances where change in pH leads to acid catalysis of proton exchange between carbohydrates and water (the present study ; Hills 1991) or gelation of pectin. The concentrations of individual free sugars, on the other hand, vary widely during the season. Unfortunately, in addition to their dependence on PWS and concentration, solutions containing the same number of exchangeable protons (fructose and glucose) do not have equivalent relaxation properties either (Fig 6D). At extremely high concentrations ([100% (w/v)) these variations are attributed to di†erences in viscosity or solubility (Suggett et al 1976 ; Lai and Schmidt 1991). At more dilute concentrations however, where viscosity differences between solutions of fruit sugars are small, variations in relaxation probably relate to conformation changes in solution a†ecting the ratio of axial/ equatorial hydroxyl protons and ease of exchange with aqueous protons (Harvey and Symons 1978 ; MoraGutierrez and Baianu 1989). Moreover, in ternary aqueous systems containing simple sugars plus neutral salts, relaxation rates are inÑuenced by ionic strength and the composition of the neutral salt (Bociek and Franks 1979 ; Lai and Schmidt 1991 ; Sattiacoumar and Arulmozhi 1993). Setting aside any e†ects structure may impose on relaxation, the array of factors capable of inÑuencing relaxation in sugarÈwater solutions serves to emphasise that variations in serial imaging data can not be interpreted unequivocally without recourse to detailed compositional analysis. Localised spectroscopy and chemical shift imaging can, of course, be used to probe carbohydrates directly in imaging experiments rather than trying to infer their presence based on changes in relaxation alone (Ishida et al 1994, 1996 ; Tse et al 1996). The failure of imaging to detect large changes in solution composition is not inconsistent with attempts to develop other NMR techniques as non-destructive sensors for sweetness in fruit. Interference from localised Ðeld gradients in two-dimensional data collection is avoided by analysing the spectral characteristics of the whole fruit. Use of surface coils to generate homogeneous magnetic Ðelds in particular regions of the sample and analysing the sugar/water ratio of their respective spectral peaks (Zion et al 1995), or using the strong relationship between self-di†usion and sucrose concentration in aqueous solution (Wai et al 1995), are the more successful approaches currently being evaluated. What is unusual about the small male fruit used in our exploratory investigation where relaxation changes at 7É1 T were observed after harvest (Callaghan et al 1994) ? The only obvious di†erence is that the concentration of free sugars in the small ripened fruit exceeded 400 mg g~1 dry weight. This is at least four times greater than that recorded in commercial fruit. Perhaps thresholds as high as this have to be exceeded before solution e†ects overcome those imposed by structural constraints ? However, even at concentrations 357 as high as this, the self-di†usion coefficient in male fruit declined by only 12% compared with the 28% that might otherwise be expected from the relationship between self-di†usion and SSC in sucrose solutions (Irani and Adamson 1958). Hence, even under circumstances where the greatest possible changes in solution composition occur, changes in MRI measurements are not especially sensitive. The results indicate that quantitative MRI is able to measure temporal trends and tissue di†erences in MR properties of kiwifruit. One can attempt to rationalise these on the basis of changes in solution composition. The tendency towards longer relaxation times later in the season, in locular tissue for instance, is certainly consistent with increasing dilution of cell contents, as dominant soluble macronutrients like N and K become less concentrated during fruit development (Clark and Smith 1988). However, this ignores other factors that have an impact on relaxation, like increase in cell size (Hopping 1978) and change in the volume of the intercellular gas space postharvest (Hallett et al 1992). Hence, interpretation of data from quantitative images of biological samples will require a more intimate, holistic view of what is occurring within the sample, and at a level of detail which will be difficult to provide for many of the systems that are currently of interest. Indeed, the results suggest that fruit systems are probably too complex for even a basic set of relaxation principles to apply across a range of crops. ACKNOWLEDGEMENTS The authors thank Rod Bieleski, Ken Marsh and Vivienne Paterson for technical assistance ; Barbara Dow for support with statistical analysis : Paul Callaghan for pulse-sequence programming ; and Allan Johnson for generous access to Duke University Radiology Department facilities. This study was funded by the NZ Foundation for Research, Science and Technology (C06504), and Ðnancial support from the NZ Kiwifruit Marketing Board is acknowledged. REFERENCES Bieleski R L 1994 Pinitol is the main soluble carbohydrate in leaves from some coastal plants indigenous to New Zealand. NZ J Bot 32 73È78. Birch G G, Karim R 1992 Apparent molar volumes and 1H-NMR relaxation values of glucose syrups. J Sci Food Agric 58 563È568. Bociek S, Franks F 1979 Proton exchange in aqueous solutions of glucose. J Chem Soc Faraday T rans 75 262È270. Brownstein K R, Tarr C E 1979 Importance of classical di†usion in NMR studies of water in biological cells. Phys Rev A 19 2446È2453. 358 Callaghan P T 1991 Principles of Nuclear Magnetic Resonance Microscopy. Clarendon Press, Oxford, UK. Callaghan, P T, Clark C J, Forde L C 1994 Use of static and dynamic NMR microscopy to investigate the origins of contrast in images of biological tissues. Biophys Chem 50 225È 235. Carver J P, Richards R E 1972 A general two-site solution for the chemical exchange produced dependence of T upon 2 the Carr-Purcell pulse separation. J Magn Reson 6 89È105. Clark C J, MacFall J S 1996 Magnetic resonance imaging of persimmon fruit (Diospyros kaki) during storage at low temperature and under modiÐed atmosphere. Postharv Biol T echnol 9 97È108. Clark C J, Smith G S 1988 Seasonal accumulation of mineral nutrients by kiwifruit. 2. Fruit. New Phytol 108 399È409. Goodman B A, Williamson B, Simpson E J, Chudek J A, Hunter G, Prior D A M 1996 High-Ðeld NMR microscopic imaging of cultivated strawberry fruit. Magn Reson Imag 14 187È196. Hallett I C, MacRae E A, Wegrzyn T F 1992 Changes in kiwifruit ultrastructure and cell packing during postharvest ripening. Int J Plant Sci 153 49È60. Harvey J M, Symons, M C R 1978 The hydration of monosaccharidesÈan NMR study. J Soln Chem 7 571È586. Hills B P 1991 Multinuclear NMR studies of water in solutions of simple carbohydrates. I. Proton and deuterium relaxation. Mol Phys 72 1099È1121. Hills B P, Duce S L 1990 The inÑuence of chemical and di†usive exchange on water proton transverse relaxation in plant tissues. Magn Reson Imaging 8 321È331. Hills B P, Wright K M, Belton P S 1989 Proton NMR studies of chemical and di†usive exchange in carbohydrate systems. Mol Phys 67 1309È1326. Hopping M E 1976 Structure and development of fruit and seeds in Chinese gooseberry (Actinidia chinensis Planch). NZ J Bot 14 63È68. Irani R R, Adamson A W 1958 Transport processes in binary liquid systems. I. Di†usion in the sucrose-water system at 25¡C. J Phys Chem 62 1517È1521. Ishida N, Koizumi M, Kano H 1994 Ontogenetic changes in water in cherry tomato fruits measured by nuclear magnetic resonance imaging. Sci Hort 57 335È346. Ishida N, Ogawa H, Koizumi M, Kano H 1996 Ontogenetic changes of water status and accumulated soluble compounds in growing cherry fruits by NMR imaging. J Magn Reson Anal 2 168È169. Kuchenbrod E, Haase A, Benkert R, Schneider H, Zimmermann U 1995 Quantitative NMR microscopy on intact plants. Magn Reson Imag 13 447È455. C J Clark, L N Drummond, J S MacFall Kuchenbrod E, Landeck M, Thurmer F, Haase A, Zimmermann U 1996 Measurement of water Ñow in xylem vessels of intact maize plants using Ñow-sensitive NMR imaging. Bot Acta 109 184È186. Lai H-M, Schmidt S J 1991 Proton, deuterium and 17O NMR relaxation studies of lactose- and sucrose-water systems. J Agric Food Chem 39 1921È1926. MacFall J R, Wehrli F W, Greger R K, Johnson G A 1987 Methodology for the measurement and analysis of relaxation times in proton imaging. Magn Reson Imag 5 209È 220. MacFall J S, Johnson G A 1994 The architecture of plant vascular and transport as seen with magnetic resonance microscopy. Can J Bot 72 1561È1573. MacRae E, Quick W P, Benker C, Stitt M 1992 Carbohydrate metabolism during postharvest ripening in kiwifruit. Planta 188 314È323. Mora-Gutierrez A, Baianu I C 1989 1H NMR relaxation and viscosity measurements on solutions and suspensions of carbohydrates and starch from corn : the investigation of carbohydrate hydration and stereochemical and aggregation e†ects in relation to 17O and 13C NMR data for carbohydrate solutions. J Agric Food Chem 37 1459È1465. Redgwell R J, Melton L D, Brasch, D J 1992 Cell wall dissolution in ripening kiwifruit (Actinidia deliciosa). Solubilization of the pectic polymers. Plant Physiol 98 71È81. Sattiacoumar S, Arulmozhi V 1993 Proton NMR relaxation studies in aqueous solutions of sugars with alkali halides. Bull Soc Chim Belg 102 507È516. Smith G S, Clark C J, Boldingh H L 1992 Seasonal accumulation of starch by components of the kiwifruit vine. Ann Bot 70 19È25 (Erratum 1993. Ibid 71 95). Suggett A, Ablett S, Lillford P J 1976 Molecular motions and interactions in aqueous carbohydrate solutions. II. Nuclear magnetic relaxation studies. J Soln Chem 5 17È31. Tse T Y, Spanswick R M, Jelinski L W 1996 Quantitative evaluation of NMR and MRI methods to measure sucrose concentrations in plants. Protoplasma 94 54È62. Wai W-K, Stroshine R L, Krutz G W 1995 A modiÐed Hahn echo pulse sequence for proton magnetic resonance (1HMR) measurements of percent soluble solids of fruit. T rans ASAE 38 849È855. Wang S Y, Wang P C, Faust M 1988 Non-destructive detection of watercore in apple with nuclear magnetic resonance imaging. Sci Hort 35 227È234. Zion B, Chen P, McCarthy M J 1995 Nondestructive quality evaluation of fresh prunes by NMR spectroscopy. J Sci Food Agric 67 423È429.