*3f,S: Food Microbiotogu ond €afetu

Measuring Mold Infestation in Raw Tomato fuiceS.I. POTTS, D.C. SIAUGHTER, AND I.F. THoMPSON

ABSTRACT: Amodified fluorescentlectin test formolded rawtomato juice was compared with both thevisual moldinspection method conducted by the California Processing Tbmato Advisory Board and the Howard mold count(HMC) conducted by 4 commercial tomato processors. The assay quantifies fungal contamination by detectingfungal chitin using FlTC-labeled lectin that selective$ binds to chitin. The mold content of l0O naturally infectedraw tomato juice samples was determined using these 3 methods. The coefficient of determination beiween thelectin assay and the HMC (r'z= 0.73) was better than the coefficient of determination between tlre Californiaprocessing tomato visual mold inspection method and the HMC (r'?= 0.38). The coefficient of determination be-tween the fluorescent lectin assay and the HMC (re: 0.73) was comparable to the coefficient of determinationbetween different qualitycontrol laboratories'HMC values, which ranged from rz = 0.69 to r2: 0.81. The fluorescentlectin assay had consistently better precision (average CV = 87o) than the HMC (average CV : 387o).

Keywords: filammtous firngi, tomato, lectin, mold detection, Howard mold count

lntroductionpaw CnuroRNtA pRocEssrNc ToMAToESI fare inspected for quality prior toprocessing at field stations locatedthroughout the state under a programadministered by the Califomia Process-ing Tomato Advisory Board (PTAB).Each 23.6 metric ton load of raw toma-toes is inspected for 7 quality factors,including mold and soluble solids con-tent. PIAB (1996) uses a visual mold in-spection method for the raw fruit, inwhich seasonal inspectors determinethe mass percentage of tomatoes con-taining visibly observable mold con-tamination. The California processingtomato industry has expressed con-cerns about the accuracy ofthis subjec-tive mold assessment method.

A fluorescent lectin test based on theselective binding of a lectin (wheatgerm agglutinin) to chitin in the fungalcell wall has been proposed as an alter-native method of mold assessment inraw tomato juice (Potts and others2000). Preliminary research conductedby Potts and others, using 4 fungal spe-cies on laboratory-inoculated fruit indi-cated the potential of this assay. How-ever, the researchers noted thatadditional assay refinement was re-quired to reduce the time to conductthe test prior to its commercial accept-ability, and that it should be evaluatedusing a wide-range of fungal specieswith naturally infected fruit in order todetermine whether the lectin assay is anappropriate replacement for the cur-rent mold assessment method.

The current standard for gradingmold in processed food products is the

@ 2002 Institute of Food Technologists

Howard mold count (HMC), a micro-scopic count conducted at processors'quality control laboratories (Howard19f 1; AOAC 1999). Many researchershave noted limitations of the Howardmold count, including poor repeatabili-ty, inaccuracy, high labor costs, and de-pendence on food processing condi-tions. Howard and Stephenson (1917)noted that a mold count of 40 might beobtained in samples having any amountof rot between2.2 and.1007o. Eisenburg(1952) conducted a limited study com-paring fungal species effects on spoiledvolume and HMC methods for mold-contaminated tomatoes. He observed,for example, that tomatoes contami-natedwith Geotrichum spp. at aTVoto177o spoiled volume level had a HMCranging from 60 to 100, while tomatoescontaminated with Phytophthora spp.at the same 7Vo to lTVo spoiled volumelevel had a HMC ranging from only 20to 30. Potts and others (2000) studiedthe repeatability of the HMC on blindtriplicate raw tomato juice samples andobserved average coefficients of varia-tion (CV) ranging from22 to 50% de-pending upon the operator.

Despite its imperfections, the HMCwas selected as the method for com-parison in this study because it is theonly official method of mold assess-ment, it can be conducted more rapidlythan alternative methods (Gouramaand Bullerman 1995), and a large num-ber of replicate measurements could bemade by 4 commercial tomato proces-sors located in Northem Califomia, im-proving its reliability.

This research compared the accura-

cy and precision of an improved fluo-rescent lectin assay and the currentPTAB visual mold inspection methodwith the Howard mold count, usingnaturally infected raw tomato samplestaken from California processing toma-to fields.

Materials and Methods

Sample collection and PTAB visualmold assessment

Over a 6-wk period during the 1999California tomato processing season, atotal of 100 comminuted tomato juicesamples were received from Californiatomato grading stations. The sampleswere collected during the normal com-mercial tomato inspection process con-ducted by PTAB. From each 23.6 metricton tomato load, PTAB collected 45.5 kgof tomatoes from random Iocations inthe load. The sample was randomlysplit with 22.7 kg of fruit graded for visi-bly apparent mold following standardCalifornia grading station practices(PTAB 1996). From the remaining sam-ple, a random subsample of 3.6 kg ofwater-washed tomatoes were commi-nuted for 40 s under a vacuum follow-ing PTAB's standard practice using anindustrial blender (Waring modelCB-6, Hartford, Conn., U.S.A.). Fromeach comminuted sample, 350 ml ofjuice was stored in a plastic jar that wasthen placed in ice. The jars were thentransferred to the University's researchIaboratory within 4 h. At the Universitylaboratory part of the juice was setaside for the lectin assay, and the rest ofthe juice poured into a set of 15 ml cen-

1, 2OO1-JOURNAL OF FOOD SCIENCE 321Vol. 67, Nr.

Fietd Test of the Ftuorescent Lectin Assog

trifuge tubes (Falcon), and autoclavedfor 15 min at 121 'C for later analysis bythe HMC method. The autoclaved sam-ples were stored at 6 "C.

Howard mold count assessmentThree blind replicates of each auto-

claved raw juice sample were sent tothe quality control laboratory in each of4 commercial tomato processing facili-ties in Northern California for HMCanalysis (100 samples x 3 replicates :300 samples per laboratory x 4 labora-tories : 1200 total samples). Each com-mercial laboratory conducted an indus-try standard HMC (Method 965.41,AOAC 1999) on each 15 ml sample tube,where the number of fields positive formold was recorded as a percentage ofthe total fields observed.

Mold assessment using theimproved lectin assay

The improved lectin assay was con-ducted each d on the fresh (unauto-claved) samples as they were received,since we found that autoclaving affect-ed the lectin assay results considerably(Payne 2000). For each sample, 3 ml ofjuice was poured into each of 12 50-mlcentrifuge tubes (Falcon). The lectin as-say was conducted on sets of 4 tubes,where the order in which the tubeswere analyzed was chosen at randomfrom the total set of tubes collectedeach d. Five ml of a 10 mg/ml FITC la-beled wheat germ agglutinin (WGA) so-lution in a Tris-HCl solution (pH 10)was added to each tube. Preliminary ex-periments indicated that this amount ofTIis-HCl solution, combined with 3 mlof tomato juice, consistently resulted inan optimal operating pH around 8.0(Potts 2000). Itwas also determinedthattheWGA could be added with the Tlis-HCI solution to the juice in a single step,reducing time requirements without re-ducing performance when compared toPotts and others (2000) original proce-dure. These tubes were capped andplaced on a shaker for 3 min (com-pared to 40 min in Potts and othersoriginal procedure) to allow theWGA tobind to any mold present. After thisbinding time, 40 ml of distilled waterwas added to each tube, the tubes shak-en to dissolve any unboundWGA, andthen centrifuged at 3800 RPM (2250 x g)for 2 min. The supernatant was gentlypoured off, leaving the pelleted cells atthe bottom. An additional 40 ml of dis-tilled water was then added. The tubeswere again shaken and centrifuged for 2min atthe same speed, the supematantremoved, and then 15 ml of Tfis-HCl

solution (pH 10) was added to the sam-ple, which was shaken to resuspend thecells prior to fluorescence measure-menL

The quantity of bound FlTC-labeledWGA was determined by fluorescenceusing 490 nm excitation and 520 nmemission filters (TUrner model 450,Barnstead-Thermolyne, with N8490and NB520 10 nm bandwidth bandpassfilters). The fluorometer was calibratedand checked for drift using a fluores-cent standard of Oregon Green (Molec-ular Probes, Inc., Eugene, Oreg., U.SA).The average of 3 fluorescence readingswas recorded for each tube. To mea-sure the autofluorescence in each sam-ple, 4 control tubes were also tested,with the only procedural difference be-ing that no FlTC-labeledWGAwas in-cluded in the assay. The average readingfrom these 4 autofluorescence controlswas subtracted from the fluorescenceresults. As a check on the binding timerequired, an additional 4 tubes of eachsample were assayed using the sameprocedure given above, except that theWGA binding time was increased from3 to 10 min.

The precision ofthe lectin assay andof the HMC assay were evaluated bycalculating the coefficient of variation(CV) across blind triplicate measure-ments conducted on each juice sample.A Kruskal-Wallis one-way analysis ofvariance test was used to determine ifthe CV values for the lectin assay andthe HMC assay conducted by the 4commercial laboratories were signifi-cantly different. It was not possible toconduct blind triplicate measurementsusing the PIAB visual mold method dueto economic and logistical constraints.

Total solids and soluble solidscontent measurement

The total solids content of each ofthe 100 samples was determined on theautoclaved samples according to AOACMethod 920.151 (AOAC 1999), with thefollowing exceptions: 10 g of materialwas used instead of 20 g, and the mate-rial was dried at 55 "C instead of 70 'C

to avoid caramelization of sugars. Thesoluble solids content of the 100 auto-claved samples was measured using arefractometer (Bellingham + StanleyInc., Model RFM100, Lawrenceville,Ga.,U.SA).

Lectin assay of crabshell chitin-spiked tomato juice

Mold-free tomato juice adulteratedwith crabshell chitin was used to verifythe chitin detection accuracv of the im-

proved lectin assay. Purified crabshellchitin (Sigma Chemical Co., St. Louis,Mo., U.S.A.) was finely ground in a mor-tar and pestle. One hundred ml of dis-tilled water was added to 60 mg of theground chitin. Defect-free processingtomatoes were handpicked from Cali-fornia tomato fields, and comminutedfor 40 s in an industrial blender. Chitinsolution was added to tomato juice toobtain t0 levels of chitin concentrationranging from 0 to 400 Fg dry chitin/mltomato juice. Six ml of each solution waspipetted into clean, 50 ml centrifugetubes (Falcon), to obtain 4 replicates foreach solution (4 replicates x 10 chitinlevels : 40 samples). Four controls(without lectin) at the highest chitin con-centration (400 pg chitin/ml tomatojuice) and 4 at the lowest chitin concen-tration (0 pg chitin/ml tomato juice)were also measured using this proce-dure. The 48 samples were analyzed forchitin content using the 3 rnin improvedlectin assay described previously.

Results and Discussion

Iectin assay of crabshell chitin-spiked tomato juice

The regression of fluorescence read-ing compared to crabshell chitin con-centration shows high linearity and lowstandard error (Figure 1). Unlike theHMC and chemical assays for quantifi-cation of chitin, the lectin assay is linear,with an r2 : 0.994 and a standard errorof 11 pg dry chitin / ml tomato juice.This test verified that the lectin assay islinear and quite precise. Any differenc-es in response ofthe lectin assay to dif-ferent fungal species is likely to be dueto differing amounts of chitin in themold mass or to variabiliw in chitin ac-cessibility.

Howard mold count resultsThe individual HMC scores, as deter-

mined by the 4 quality control labora-tories, for the juice samples rangedfrom 0 to 90%. The average HMC scorefor the 100 samples ranged from a lowoflTVo for quality control laboratory Dto a high of 4l%for quality control lab-oratory B, Table 1. While the HMC val-ues between the quality control labora-tories were significantly correlated(Table 2), only 2 of the 4 laboratories (Band C) had mean HMC values that werenot significantly different from each an-other (Table 1).

In addition to differences in overallHMC levels, there was considerablevariability between the blind triplicatevalues at each quality control laborato-

322 JoURNAL of FooD 9cl€NC€-Vol. 67, Nr. 1, 2002

Fietd Test of the Ftuorescent Lectin Assog

ry. The precision of the HMC was de-fined as the coefficient of variation (CV)of the blind triplicates (Table 2). In cal-culating the CV for the HMC, a value ofI was first added to all HMC scores.This addition was needed to allow theCV to be determined for samples withmean HMC scores of zero. The additionof one HMC unit does not greatly affectthe CV of nonzero HMC values. Theusual analysis of variance assumptionsof a normal distribution and equalvari-ance across groups were not valid forthe coefficient ofvariance values ana-Iyzed. Both the modified Levene (Lev-ene 1960; Brown and Forsythe 1974)and Bartlett (Bartlett and Kendall 1946)tests showed that the variance of coefE-cient of variation values were signifi-cantly (ct = 0.01) different between as-says. A Kruskal-Wallis one-way analysisof variance test was used in place of theusual analysis ofvariance technique be-cause it does not rely on these assump-tions. The HMC CV for the blind tdpli-cates ranged from 0 to 1567o. Theaverage CV of blind triplicates for the 4commercial laboratories were not sig-nificantly different (q : 0.0f ) with anoverall average HMC CV for blind trip-licates of387o for the 4 laboratories.

Comparison of the lectin assay andPTAB mold scores with the HMC

A Kruskal-Wallis one-way analysis ofvariance test showed that the precisionof the lectin assay (3 min binding time)was significantly better (a : 0.01) thanthe precision of the HMC assay per-formed by any of the quality controllaboratories (Table 2). Evaluated as av-erage coefficient of variation acrossreplicates, the precision ofthe lectin as-say is alrnost 5 times better than that ofthe HMC.

The PIAB visual mold scores rangedfrom 0.0 to 13.5%, with an average scoreof. 5,77o and a standard deviation of3.1%. Correlations between all mold as-sessment methods are shown in Table 3.As noted earlier, the literature has docu-mented the lack of precision in theHMC. In this study, 20 out of a total400HMC blind triplicates had a CV exceed-ing 150%. In order to reduce the biascaused by the lack of precision in someHMC measurements, HMC readingswith blind triplicates CVs over 150%were not included when calculating theaverage HMC across all 4 commerciallaboratories (AHMC) for comparisonwith the lectin assay or the HIAB meth-od. An important obserrration is that thecorrelation between the lectin assay andthe AHMC scorei is comparable to cor-

Table | -Dlstribution of HllG values asdetermined btr 4 commercial qualitycontrol laboratories for the naturallylnfected raw tomato juice sanplesstudied

Howard Mold Gounti (o/o)

Laboratory Mean2 Standard

Deviation

Table 2-Average coefficlent of varia-tlon (GVl of replicatest for HllG andlectin assay

Detection Method Ave. CVa

HMG, Lab. AHMC, Lab. BHMC, Lab. CHMC, Lab. DLectin assay,

3 minute binding time

39.5oloa35.4o/oa40.9o/oa36.0%a

8.0%bB

AD

3Calculations basod uoon 100 samDles. CVdstermined from blind reDlicate meisurements oneach sampleaCV values with the same grouping letter are notsignificantly (a = 0.01) difterentlCalculations based upon 100 samples, each ol

which was the averags of blind triplicatemeasuremenls2 Means with thg same grouping letter are notsignilicantly (c - .01) different

Table 3-Corrclatlons betureen HtC,

41 a3 6 a2 5 b1 7 c

26201 91 4

lectln assay, and the PTAB mold aconeg

Detectaon HMC, HMC,method Lab. A Lab. B

HMC, HMC, AHMC5 PTAB MoldLab.C Lab.D Score

HMC, Lab. AHMC, Lab. BHMC, Lab. CHMC, Lab. DPTAB Mold ScoreLectin assay,3 min binding timesAHMC is th€ av€rage HMC value of each of 100 samples lor all lour commercial laboratories

1 . 0.90 1.0.87 .88.83 .83.58 .64

.79 .85

1 ; - -.80 1.0.57 .54 .62 1 . 0

.67.78.79

relations between HMC results from Iquality control laboratory to another.Figure 2 shows the regression relation-ship between the lectin assay and theAHMC, and Figure 3 shows the regres-sion relationship between the PIAB visu-al mold score and the AHMC. The lectinassay has a much higher correlation withthe AHMC than the PTAB visual mold

score test does with the AHMC.It appears that the residual values be-

tween the lectin assay and AHMC are re-duced at lower AHMC levels (Figure 2).For analysis, these 100 samples werebroken into 3 groups by AHMC score,with Group I containing samples withAHMC values between 0 and. 20Vo,Group 2 contairfng samples withAHMC

1 000d

tr

9 o! l ow c

o> ( tE oq o9 LG O

I

teJ . =

so

800

600

400

200

050 1 00 1 50 200 250 300 350 400

Chitin concentratlon(pgrams crab shell chitin/ml tomato juice)

Flgure l-Lectln assay results for tomato luice aplked with crabshell chltln

Vol. 67, Nr. 1, 2002-JoURNAL oF FooD gcl€Nce 323

Fietd Test of the Ftuorescent Lectin Assog . . .

values between 2l and.4lVo, and Group 3containing samples with AHMC valuesbetween 4l and 607o. A Kruskal-Wallisone-way analysis of variance testshowed that Group I had significantlylower residuals than Groups 2 and 3 (o =0.05). This indicates that the relation-ship between the lectin assay and theAHMC is better at mold levels betweenO and20% HMCthanabove20 HMC.

It is well known that the HMC be-comes nonlinear at higher mold countlevels due to field saturationwhere addi-tional fungal mycelia in an alreadyposi-tive field does not increase the count.Potts and others (2000) observednonlin-

earity between mold level and HMC athigh HMC levels (that is, above 70 HMCfor Stemphylium botryosum). Nonlin-earity was tested by regressing the lectinassay score with each laboratory's HMCresults, and investigating whether thequadratic term was significant. For labo-ratories A, C, and D, the quadratic termwas not significant (a : .05). Onlylabo-ratory B had a significant quadraticterm, and this was significant at thect : .05 level but not at the ct : .01 level.The linearity observed in this study ismost likelya result of lower mold levelsthan that used by Potts and others; thehighest average HMC value in this study

was 70.4, and 907o of the AHMC valueswere below55.

Effects of method improvements onthe performance of the lectin assay

The lectin assaywils conducted usingboth 3 min and 10 min binding timetreatments. The 3 min binding timetreatment had a very high degree ofcorrelation with the l0 min bindingtime treatment (r : 0.97). For the 3 minbinding time, the average CV betweenblind replicates was 87o, with a medianvalue of 7%. For the 10 min bindingtime, the average coefficient of varia-tion between blind replicates was 117o,with a median value of 9%. The correla-tions of the l0 min binding time lectinassay and the 3 min binding time lectinassay with both the PTAB visual moldinspection method (rro,oin : 0.69, r3.in: 0.67) and with the AHMC (rro min :0.86, rrori,, : 0.85) were slightly higherfor the 10 min treatment than the 3 mintreatment. Given the general desire forrapid testing, these results indicate littlebenefit to the longer treatment.

Three fluorescence readings for eachblind replicate were averaged to obtainthe results for the correlations in Table3. While lowering variability, conductingadditional fluorometer readings addstirne and cost to an industrial assay. Todetermine whether these replicatereadings were beneficial, the AHMCscores were regressed against the 3 minlectin assay using both the average ofthe 3 fluorometerreadings, and individ-ual readings. The standard error ofcali-bration using the average of the fluo-rometer readings was 9.9 HMC units,while it was only slightly higher for indi-vidual readings: 10.0 HMC units for thelst reading, 10.2 for the 2nd reading,and 10.4 for the 3rd. The extra time re-quired to make multiple fluorometerreadings does not appear to be justifiedgiven the slight decrease in variability.

The background autofluorescence ofthe tomato juice ranged from 21 to 76,with an average reading of 43 relativefluorescent units. However, 80% of theautofluorescence values were between30 and 57 relative fluorescent units,while theWGA-treated samples (3 minbinding time) had raw fluorescencescores up to 313. It appears that thevariability in autofluorescence is smallrelative to the change in fluorescencedue to mold. To verify this, the 3 minlectin binding scores were regressedagainst the average HMC with and with-out the background autofl uorescencesubtracted. There was no improvementin the correlation to HMC with the au-

Figuro 9-Relatlonshlp betureen the PIAB mold scor€ and the averrge Howardmold count

324 JOURNAL Op FOOD SC|ENCE-Vo|. 57, Nr. 1,2002

- 8otr3 toog 60oE so

Eno'€ 30

g 206b to

1 2 = 0 . 7 3SEC=9.9 HMC unlte

oo

o oo

o o

o o' 7 9 o

I (,

o-r)

g Oo

d

q)

0 50 100 1s0 200 250 300Lectln assay (relatlve lluorescent unlts)

Flgure 2-Belatlonthlp betureen the lectin anay and the Howard mold count

- 8otr3 zooE 60oE 50

E oo=

f 30

g 20G

b to

4 6 I 1 0PTAB Mold Score ("/o)

12 1 4

r 2=0 .38-SEG=14.9 HMG unlts

i : oo o

oo

oo"?38u -/

/o

o o 0 oo

o o6 _ v

-1/ o 9o o

i oi o

oo?8o o

Fietd Test of the Fluorescent Lectin Assog

tofluorescence subtracted. Since read-ing a sample with no lectin to correctfor autofluorescence did not improvethe lectin assay results, it could bedropped from future experiments tosave time.

Effects of soluble and total solidsSoluble solids content for the 100

samples ranged between 3.7 and 7.2"Brix, with a mean of 4.9 'Brix and astandard deviation of 0.7 "Brix. Totalsolids ranged between 4.3 and, 7.8Vo(wet weight basis), with a mean of 5.37oand a standard deviation of0.77o. Therewas a high degree of correlation be-tween the soluble solids content andthe total solids content, with an r2 : .94and a standard error of.O.l7% total sol-ids. There was no significant relation-ship between either total solids or solu-ble solids and any of the moldmeasurementmethods (a : .05).

ConclusionESTS USING NATURALLY INFECTED

I processing tomatoes showed thatthe lectin assay can be utilized as a re-

placement for the current PTAB visualmold inspection method and is moreprecise than the HMC for raw juicesamples, providing that carefirl calibra-tion procedures are followed. The lectinassay is over 4 times more precise thanthe Howard mold count. It has as gooda correlation with the Howard moldcount as the correlation of the HMCconducted between different testingsites, and is considerably better corre-lated with HMC than the current PTABvisual mold inspection method.

ReferencesAOAC 1999. AOAC official method 965.41 mold in

tomalo products. In: official methods of analysis ofAOAC International. 16d Ed. sth Revision. Gaith-ersburg, MD: AOAC International. P 63-64.

Bartlett MS, Kendall DG. 1946. The statistical anal-ysis of variances-heterogeneity and the logloga-rithmic transformation. IRSS Suppl. 8:128-138.

Brown, MB, Forsythe AB. 1974. The small samplebehavior of some statistics which test the equalityof several means. Techonometrics l6:129-132.

Eisenburg, MV, 1952. Observations and suggestionson factory control of rot and extraneous mamatterin tomato products.washington DC: Nat CanAssoc,Inf. I€t. No. 1371. P 36-39.

Gourama, H, Bullerman LB, 1995. Detection of moldsin foods and feeds: Potential rapid and selectivemethods. J. Food Prot. 58(12):1389-1394.

Howard, BJ. 1911. Tomato ketchup under the micro-scope with practical suggestions to insure a clean-ly product. Washington DC: U.S. Dept. Agricultue,Bureau of Chemistry. Circular No. 68.

Howard, BJ, Stephenson CH. 1917. Microscopicalstudies on tomato products. Washington DC: U.S.Dept. ofAgdculture. Bulletin 581, October 6, 1917.

Levene, H. 1960. Robust tests for the equality ofvari-ances. In I. olkin, editor. Contributions to proba-bility and statistics. Stanford, CA: Stanford Uni-versity Press. P 278-292.

Payne, Il. 2000. Personal communication. Biologicaland Ag. Engineering Dept, University of Califor-nia at Davis,

Potts, SJ. 2000. An opto-chemical assay for mold de-tection in processing tomatoes [Ph.D. dissertation].Davis, Calif.: University of Califomia. 143 p.

Potts Sl, Slaughter DC, Thompson lF. 2000. A fluores-cent lectin assay for the rapid quantification ofmold in tomato products. l Food Sci 65(2):346-350.

PTAB. 1996. Processing tomato inspection manual,west Sacramento, Calif.: Processing Tomato Advi-sory Board. 33 p.

MS 20001600 Submitted l2lLgl00, Revised 3/9/01,Accepted 412510I, Received 8/2/01

The authors ue grateful to Jemifer Payne, EuniceTan, PaulaChou, Ryan tlammond, and IohnieYoung in the Biologicaland A8. Engineering Department at the University ofCali-fornia at Davis for theL assistance in conducting this study.We also thank the California Processing Tomato AdvisoryBoard for the support they provided.

Authors are with the Biological and AgriculruralEngineering Dept., Uniuersity of Califurnia OneShields Ave., Dauis, CA 95616. Direct conespon-denu to author Slaughter (E-mail: dalaughter@ucdauis.edu).

Vol. 67, Nr. 1, 2002-JOURNAL OF FOOD 9C|BNCE 325