The Cntributin f Hydrlyzed Flavr Precursrs t Quality Differences in Shiraz Juice and Wines: An Investigatin by Sery Descriptive Analysis N. A. ABBOTT ~, B. G. COOMBE 2, and P. J. WILLIAMS 3. Vlatile cmpnents liberated by hydrlysis f C-18 reversed phase islates frm Vitis vinifera cv. Shiraz juice were evaluated by sery descriptive analysis. The islates were hydrlyzed at ph 3.2 r by treatment with a nn-selective glycsidase enzyme. Shiraz juices assigned as high and lw 'quality', each surced frm tw regi and sampled in 1988 and 1989, were studied. Wines made frm the 1989 juices were als subjected t descriptive analysis. The results shwed that fr ne pair f wines and all but ne pair f hydrlysates, the quality differences culd be distinguished and quantified. The glycsidic hydrlysates prepared by bth methds were fund t cntribute nn-berry attributes t wine such as 'stalky', 'earthy', and 'cigar/tbacc'. Wines cidered t be f high quality were rated higher in these nn-berry attributes than their lw quality cunterparts, and it may be deduced that glycsidic hydrlysates cntain arma cmpunds that are imprtant t high quality Shiraz wine. KEY WORDS: Shiraz, sery analysis, quality, enzyme hydrlysis, acid hydrlysis, flavr precursr The rle f glycsidic cnjugates f vlatile cmpunds as precursrs f flavr in grapes and wine is nw better understd (8,9). Analysis f vlatiles liberated by either acid r glycsidase enzyme hydrlysis f the glycsidic fractin has prvided sme iight int the flavr cmpunds respible fr the sery prperties f nn-flral varieties (9). T date, hwever, this apprach f precursr analysis has been cnfined t cmparis f the vlatile cmpsitin f white wine grape varieties. We have recently adapted this apprach t attempt a mre quantitative determinatin f'quality' in a black variety, Vitis vinifera cv. Shiraz (1). The grapes were sampled frm cmmercial vineyards under different viticultural management regimes and frm three different climatic regi. The high and lw 'quality labels' were given t the Shiraz samples by bth winemakers and viticulturists based n their experiences with the wines made frm the respective vineyards (see Materials and Methds). The study is still at a preliminary stage, but has indicated that Shiraz grapes f high quality, when cmpared with grapes f lesser quality, yield a greater cncentratin f vlatiles as enzymeliberated prducts and a greater number f vlatiles decnjugated by acid hydrlysis f the respective precursr fracti (1). In cnjunctin with the chemical analysis f the vlatile cmpnents released frm the cnjugates, it is 1,2Department f Plant Physilgy, Waite Agricultural Research Ititute, The University f Adelaide, Glen Osmnd, Suth Australia, 5064; and 3The Australian Wine Research Ititute, Glen Osmnd, Suth Australia, 5064. *Authr t whm crrespndence shuld be addressed. This research was cnducted at The Australian Wine Research Ititute. We thank the staff members f The Australian Wine Research Ititute fr participating n the sery panel, industry members wh cntributed bth wine and grape samples and the Grape and Wine Research Cuncil fr funding the research. Manuscript submitted fr publicatin 19 July 1990. Cpyright 1991 by the American Sciety fr Enlgy and Viticulture. All rights reserved. 167 necessary t investigate the flavr prperties f the vlatiles bund in the precursr fractin f Shiraz. Specifically, there is a need t determine the cntributin made by the decnjugated vlatiles t the sery prperties f Shiraz and t give sery validatin t the premise n which the research is based, i.e. that samples frm vineyards traditinally yielding fruit varying in quality can bjectively be distinguished. These aims necessitate the applicatin f sery descriptive analysis (4) t base wines spiked with the acid and enzyme hydrlysates. The aims als require cmparisn f the sery prperties f the spiked samples with thse f wines made frm the grapes frm which the precursr fracti were derived. This paper reprts the results f these sery studies. Materials and Methds Samples: The juice samples were selected frm three viticultural regi and f three different qualities. The wrd 'quality' as used in this paper relates t expected market value as wine. The assignments f 'high', 'medium', and 'lw' quality were btained frm: (a) the winemakers' evaluatin f the fruit; (b) the cmmercial and, where apprpriate, shw recrds f wines frm the vineyards; and (c) the viticultural characteristics f the vineyards. The yield frm the vineyards assigned as being f lw quality was tw t three times greater than that frm the high quality vineyards fr bth regi and years. The high quality fruit came frm vineyards which supplied grapes used in premium wine prductin. Lw quality fruit was used fr bulk wine prductin. The fur wines were prepared cmmercially frm high and lw quality grapes frm the Cnawarra and the Barssa Valley regi in 1989. They were sampled directly after fermentatin and had received n wd treatment.
168- HYDROLYZED FLAVOR PRECURSORS Extractin f vlatiles frm juice precursr fracti: Precursr fracti islated frm Shiraz juice (500 ml) by retentin n C-18 reversed phase silica gel were subjected t acid r enzyme hydrlysis as previusly described (1) with the exceptin that n internal standards were added. The enzyme added was Rhapect C, a cmmercial pectinase preparatin cntaining glycsidase activities (3). The hydrlysis prducts were islated by cntinuus extractin with dichlrmethane. Slvent was remved frm these extracts by distillatin thrugh a clumn f Feke's helices t give a cncentrate f ca 100 BL. The cncentrate was taken up in water (10 ml), and the aqueus slutin was rtary evaprated in vacu at rm temperature fr 10 minutes t remve the slvent residue. Each sample was adjusted t a vlume f 20 ml with water and stred at-20 C. Sery studies, general aspects: The samples and reference standards were prepared fresh each mrning. All samples were presented t the panel in 20-mL aliquts in cded, clear, tulip-shaped glasses. Panelists assessed the samples by arma nly, befre midday, in islated bths under red light at 22 C + 2 C. Fr bth du-tri difference tests and descriptive analyses, the precursr hydrlysates were presented t the panel in a neutral base Shiraz wine at a cncentratin equivalent t that at which the precursr glycsides were present in the riginal juice. Fr the purpse f cmparing spiked samples, a neutral base wine exhibiting minimal backgrund arma was required. This base wine was chsen and prepared t generate a matrix as clse t a red wine medium as pssible. It was prepared by rtary evaprating 300 ml f cmmercial bulk Shiraz wine in vacu at rm temperature fr three hurs t a final alchl strength f ca 8%. The wine samples were taken frm a fresh bttle fr each replicate. Du-tri difference tests: These tests were cnducted by the prcedure described by Amerine et al. (2). The neutral base wine was used as the reference in difference tests fr the precursr hydrlysates. The reference used in difference tests invlving wine samples was ne f the pair under evaluatin. Tw sets f dutri tests were evaluated at each sessin, each sample being assessed nly nce by a panel f 26 staff members f The Australian Wine Research Ititute. The significance f the du-tri tests were determined frm detailed statistical tables fr ne-tailed tests, p = 1/2 (7). Descriptive analyses: Furteen staff members frm the Ititute were selected t take part in the descriptive analyses accrding t their availability and interest in the prject. All panel members had participated in the du-tri tests and were familiar with the samples. The initial training sessin invlved the assessment f several cmmercially available Shiraz wines. The arma descriptrs suggested in this sessin frmed the basis fr the develpment f a range f descriptive standards. Wines and precursr hydrlysates were presented at a further twelve training sessi during which these descriptive standards were assessed and mdified. Twelve attributes were cidered by the Table 1. Descriptive attributes and cmpsitin f stck mixtures and reference standards. Attribute cde H/R BER STR R/V CIT SPC PEP STK EAR C/T LIC CHOC Attribute Hney/raisin Berry Strawberry Rse/vilet Citrus Spice Pepper Stalky Earthy Cigar/tbacc Licrice Chclate 1Prtin f stck mixture made up in base wine (60 ml). Cmpsitin f stck mixture Raisi (10 g) were steeped in water (20 ml) fr 48 h and cmbined with a 1:2 mixture f hney (Adelaide Hills) and water (20 ml) Blackberry jam (Beerenberg, 15 g), black current jam (Cttee's, 15 g), Ribena crdial (2 ml), and blackberry juice (Jhn West, 2 ml) were added t 15 % ethanl (23 ml) Strawberry jam (hme made) 20 #g/l ~-Inne plus 2-phenethanl (0.1 ml)in ethanl (20 ml) E30817 VMane Fils natural passin fruit flavr (0.1 ml)in ethanl (10 ml) with 99.9% lemn juice (Berri, 0.2 ml) Whle clves (20) steeped in ethanl (7.5 ml, 24 h), remved, and liqur made up t 20 ml with ethanl Whle piments (10) steeped in ethanl (7.5 ml, 24 h), remved, and liqur made up t 10 ml with ethanl Black peppercr (10) crushed and steeped in ethanl (10 ml, 24 h), then diluted 1:1 with ethanl Geranium stalks Cut grass Mushrm cmpst (1 g)in water (20 ml) Earth/bark Shrt panatella cigars (Henri Winterman, 700 mg) steeped in a mix f water (20 ml, 24 h) and ethanl (1 ml) Cigarette tbacc (500 mg) steeped in water (20 ml, 24 h) Licrice strand Cking chclate (Cadbury Burneville, 23.8 g) and cca( Cadbury Burneville, 1 g) suspended in water (40 ml) Prtin used in reference standard 2mL 5mL 4g 0.1 ml 0.05 ml 1 ml 0.01 ml 0.2 ml 0.2mL lg 200 mg 1 ml 500 mg 4mL 4mL 4g 2mL
HYDROLYZED FLAVOR PRECURSORS m169 panel as necessary t describe the sery prperties acrss the samples. The cmpsitin f the standards used t describe the twelve attributes is given in Table 1. The 20 samples selected fr descriptive analysis were presented in a randm rder with n tw samples being evaluated tgether mre than nce. All samples were presented in duplicate except the base wine which was randmly presented n five ccasi thrughut the study. At each sessin three samples were evaluated. The twelve reference standards were present in each bth alng with the samples. Panelists were asked t smell the standards individually in the rder given in Table 1 and then rate the inteity f each arma in the samples. Arma inteities were rated n a 10-pint scale 0 t 9 where 0 = n attribute present and 9 - high, i.e. equal t the arma f the standard. The arma inteity f the standards were nt sufficiently great t present any bvius interference t the test samples. Statistical analyses: Statistical analyses were perfrmed using NH Analytical sftware 'Statistix' (Rseville, MN). The culting services f Lindsay Veitch Inc. were emplyed. N judges were remved frm the panel, therefre, the mean scres f the 14 assessrs fr each attribute and fr each sample were used in all analyses. The 1988 and 1989 samples were prcessed as ne data set as the errr structure fr bth years was expected t be the same even thugh there was an added treatment, i.e., wine in the secnd year. Cequently, the experimental design was nn-r- thgnal and multiple regressin analysis (MRA) rather than the usual analysis f variance (AOV) was perfrmed n the data. If AOV had been chsen as the methd fr statistical analysis, separate analyses wuld have been necessary fr each f the tw years, leading t difficulties in any subsequent cmparis. A further benefit f the MRA was that an assessment f any carry-ver effect was pssible t determine if the inteity f the arma f any samples biased the results f anther sample assessed in the same set. This carry-ver effect culd nt be calculated using AOV. Quality High High Mid Mid Lw Lw High High Mid A design matrix was thus estab- Mid lished t enable the data t be analyzed Lw as a single data set and t allw differ- Lw ences in quality between samples t be High investigated. This design matrix was, High therefre, ctructed t include the Mid main variables: i.e., year, y (1988 and Lw 1989), regin, r (Barssa and Lw Cnawarra), quality, q (high and lw), and treatment, st 1 and st 2 (enzyme hydrlysis, acid hydrlysis, and wine) and all pssible interacti amng them. Thus, the main effects were y, r, q, (st 1 and st 2) with degrees f freedm 1, 1, 1, and 2, respectively. Additinally, allwance was made fr a pssible carry-ver effect, c, taking ne degree f freedm. The first rder interacti were y X r, y X st 1, r X q, (r st 1 and r X st 2) and (q X st 1 and q X st 2). The term y X st 2 was mitted because it was linearly dependent n y, st 1, and y X st 1. Thus, first rder interacti used 1, 1, 1, 1, 2, and 2 degrees f freedm, respectively. Secnd rder interacti were yx rx q, yxr X st 1, yx qx st 1, and(rx qx st 1 and r X q X st 2) with 1, 1, 1, and 2 degrees f freedm, respectively. The terms y X r X st 2 and y X q X st 2 were mitted because f their linear dependency n the ther variables. Finally, fr the third rder interactin there was y X r X q X st I with ne degree f freedm; y X r X q X st 2 being mitted because f linear dependence n the ther variables. Hence, the frty bservati prvided 1, 5, 1, 8, 5, 1, and 19 degrees f freedm fr the grand mean, main effects, pssible carry-ver effect, first rder, secnd rder, and third rder interacti and errr, respectively. All significance tests used the errr mean square n 19 degrees f freedm. Results and Discussin Du-tri data: The du-tri difference test results reprted in Table 2 demtrate that fr mst samples the vlatiles released frm the bund fracti f Shiraz juices, either by acid r enzyme hydrlysis, cnferred n the base wine sery prperties significantly different frm thse f the cntrl (a neutral base Shiraz wine withut any additin). It is imprtant t nte that nt all Table 2. Results f the du-tri difference tests fr arma f enzyme hydrsylates (EH) and acid hydrsylates (AH) cmpared t that f the base wine. N. f N. f Regin Year crrect Sig crrect Sig. respes respes Barssa 1988 20/28 ** 21/28 ** Barssa 1989 19/24 *** 20/28 ** Barssa 1988 21/29 *** Barssa 1989 20/29 * 19/28 * Barssa 1988 18/24 ** 21/29 *** Barssa 1989 24/27 *** 20/30 ** Cnawarra 1988 21/24 *** 19/27 * Cnawarra 1989 21/24 *** 24/28 *** Cnawarra 1988 14/22 19/25 ** Cnawarra 1989 21/27 *** 17/29 Cnawarra 1988 23\27 *** 20/28 ** Cnawarra 1989 13/27 19/28 * Langhrne Ck. 1988 22/24 *** 17/30 Langhrne Ck. 1989 13/24 16/26 Langhrne Ck. 1988 16/28 20/25 *** Lang hrne Ck. 1988 22/27 *** 10/25 Langhrne Ck. 1989 18/27 17/30 *, **, *** = significance at p < 0.05, p < 0.01, and p < 0.001, respectively. = n significant difference. Where n values are shwn, sample nt analyzed. EH AH
170 m HYDROLYZED FLAVOR PRECURSORS Table 3. Results f du-tri difference tests between tw pairs f hydrsylates. Hydrsylate cmparisn AH vs. EH AH vs. AH AH = acid hydrlysis prduct. EH = enzyme hydrlysis prduct. ~Refer t Table 5 fr sample cdes. *** = significance at p < 0.001. Samples1 M and E land M Significance the samples culd be distinguished in these tests; i.e., mst f thse frm Langhrne Creek were nt distinguished frm the cntrl by the panel. Because f the almst unifrmly psitive difference respes given by the panel t the high and lw quality hydrlysates frm Barssa and Cnawarra juices, these samples were chsen fr descriptive analysis. Althugh the 1989 lw quality enzyme hydrlysis sample frm Cnawarra was nt significantly different frm the base wine, fr cmpleteness f the data set it was included in the descriptive analyses. Previus studies f the sery prperties f precursr hydrlysates frm nn-flral grapes have shwn that nly the acid hydrlysates culd be detected when these were back-added t a base wine (9). Prducts given by glycsidase hydrlysis frm Chardnnay (5,9) and frm Sauvignn blanc and Semilln (9) were nt detectable when assessed at a cncentratin near t that at which the precursr glycsides ccurred in the riginal wine. In cntrast, mst f the Shiraz samples examined gave bth acid and enzyme hydrlysates with sery prperties detectable at single strength in a base wine. As the previus investigati f the acidand enzyme hydrlysates invlved nly single samples fr each variety, this present finding fr Shiraz suggests that examinatin f the enzyme hydrlysates frm a wider range f samples f nn-flral white varieties may yet allw bservatin f psitive du-tri respes. Du-tri data in Table 3 cnfirm that the panel culd distinguish between the vlatiles released frm a precursr fractin by hydrlysis with a glycsidase enzyme, and thse released by acid hydrlysis. Furthermre, the acid hydrlysates liberated frm a high quality juice were perceived by the panel as being significantly different frm thse liberated frm a lw quality juice. The wines prepared frm the high quality grape samples were significantly different frm thse prepared frm the lw quality grape samples fr bth the Table 5. Cdes f samples evaluated by descriptive analysis. 1988 1988 1989 1989 Barssa Cnawarra Barssa Cnawarra Lw quality EH A B C D High quality EH E F G H Lw quality AH I J K L High quality AH M N O P Lw quality wine Q R High quality wine S T EH = enzyme hydrlysis prducts. AH = acid hydrlysis prducts. Barssa and Cnawarra regi (Table 4). The identificatin cdes fr the 20 samples chsen fr descriptive analysis are given in Table 5. Statistical analyses: Multiple regressin analyses (MRA) were executed n the mean scres f the furteen assessrs fr each attribute fr the entire mdel. As a pssible mea f simplifying the interpretatin f the data, principal cmpnent analyses (PCA) were run n the crrelatin matrix frm the fitted values f all samples. The first tw principal cmpnents (PCs) are described in Figure 1 and accunt fr 91.2% f the ttal variance f the data. This principal cmpnent diagram demtrates a separatin f the hydrlysates frm the wine samples alng PC1 and a separatin f the acid hydrlysates frm the enzyme hydrlysates alng PC2. The eigenvectrs calculated frm the crrelatin matrix describe the relative psiti and ladings f the twelve attributes as used by the panel and are illustrated in Figure 2. The vectrs indicate hw each attribute was used by the panel in the descriptive analysis ver all the samples; nte that all attributes were used by the panel as significant descriptrs fr the PC2 0 (20.3%) Enzyme hydrlysates 1 i Wine Table 4. Results f du-tri difference tests fr the arma f high vs. lw quality wine samples frm tw regi. Regin Samples 1 Significance Barssa Q and S ** Cnawarra R and T *** ** and *** indicate significance at p < 0.01 and p < 0.001, respectively. 1Refer t Table 5 fr sample cdes. Acid hydrlysates b PC1 (70,9%) Fig. 1. Prjectin f sery descriptive data fr all samples (A t T) n principal cmpnents PC1 and PC2. Refer t Table 5 fr sample identificatin cdes. The high quality samples are represented bythe shaded prtin.
HYDROLYZED FLAVOR PRECURSORS m171 PC2 - Hney/raisin Licrice ~" Cigar/tbacc Earthy! PC1 Stalky Strawberry :~R, se/v ilet.=.citrus ~Berry = ~epper Spice Chclate Fig. 2. Relative psiti and Iadings f the twelve attributes used in the descriptive analysis fr samples A t T n principal cmpnents PC1 and PC2. Refer t Table 5 fr sample identificatin cdes. samples being tested. The vectr diagram (Fig. 2) crrespnds t the PC diagram (Fig. 1); superimpsitin indicates which attributes were used by the panel t describe the sample grups. The wine samples were separated frm the juice hydrlysates alng PC 1 in terms f the attributes berry, strawberry, rse/vilet, citrus, and pepper which were cllectively imprtant t the wines and the ppsing earthy and cigar/tbacc attributes which were imprtant t the hydrlysates. The majr attributes cntributing t the separatin f the hydrlysates alng PC2 were cigar/tbacc, hney/raisin, stalky, and earthy. It is interesting t nte that fr bth hydrlysate grups the high quality samples are psitined n the left indicating that they are being displaced frm the lw quality by a cmmn attribute r attributes. Interacti f independent variables: The MRAs f the full mdel ver the twelve attributes shwed that many f the first and secnd rder interacti were significant. T simplify the descriptin f the data, it was decided t ignre any interacti which did nt exhibit significance when analyzed as an entire grup: e.g., all first rder interacti tgether, etc. T determine if the independent variables made a significant cntributin t the verall mdel, F- ratis were calculated fr each grup ver the 20 samples and fr each attribute and are presented in Table 6 with their significant differences. Year was a significant factr fr five attributes and regin fr six attributes. This result was nt unexpected: the tw years experienced different weather patter during grape ripening, and the tw regi have different climate and sil types and als different canpy management regimes. Quality was significantly different fr seven f the twelve attributes studied indicating that the panel was able t distinguish the high frm the lw quality samples. Treatment was a significant variable fr the majrity f attributes ver all the samples, the implicatin f which shall be discussed belw. The first rder interacti cntributed significantly t the mdel fr five f the attributes analyzed- hney/raisin, strawberry, citrus, stalky, and cigar/tbacc. There were n significant secnd rder interacti, and the third rder interacti were barely significant fr three attributes - berry, strawberry, and stalky. The relative magnitudes f the F-ratis in Table 6 justify cnfining attentin t the main variables, hence the significant first and third rder interacti will nt be discussed further. The surce f the significance within each grup f independent variables in Table 6 can be determined frm the pled mea f the fitted values fr the cmpnent parts f each year (i.e., 1988 and 1989), regin (Cnawarra and Barssa), quality (high and lw), and treatment (enzyme hydrlysates, acid hydrlysates and wine) grups. The pled mea f the fitted Table 6. F-ratis f significant independent variables fr the 12 attributes ver 20 samples with degrees f freedm (df) and significance. Attribute Year Regin Quality Treatment I st rder 2 nd rder Hney/raisin 38.11 *** 15.49 *** 3.89 ** Berry 12.51 ** 6.37 * 13.07 ** 7.99 ** Strawberry 41.97 *** 27.01 *** 10.76 *** 5.34 ** Rse/vilet 7.93 ** Citrus 98.02 *** 53.82 *** 26.76 *** 4.56 ** Spice 19.57 *** Pepper 13.90 *** 4.47 * 26.49 *** 9.04 ** Stalky 14.72 *** 22.81 *** 7.93 ** 3.60 * Earthy 13.10 ** Cigar/tbacc 37.97 *** 71.90 *** 3.24 * Licrice 13.45 ** 5.86 ** Chclate 21.17 *** df 1 1 1 2 8 5 3 rd rder 5.70 ** 7.94 ** 5.26 ** 1 *, **, *** indicate significance at p < 0.05, p < 0.01, and p < 0.001, respectively. The residual mean squares have 19 degrees f freedm.
172- HYDROLYZED FLAVOR PRECURSORS Table 7. Pled mea f fitted values fr the significant main variables - year, regin, juice, and treatment - and the significant differences acrss the variables fr the 12 arma attributes. Hney/raisin Berry Strawberry Rse/vilet Citrus Spice Pepper Stalky Earthy Cigar/tbacc Licrice Chclate Mean values Year Regin Quality Treatment 1989 1988 (n = 24) (n = 16) Sig. 2.355 2.386 2.298 2.239 1.876 1.794 1.903 1.769 2.051 2.569 ** 1.391 1.194 * Enzyme Acid Wine Cw Bar Lw High tl t2 t3 (n = 20) (n = 20) Sig. (n = 20) (n = 20) Sig. (n = 16) (n = 16) (n = 8) 2.321 2.212 2.345 2.187 1.936 1.579 ** 1.852 1.662 * 1.516 1.229 * 1.489 1.255 * 1.639 1.409 * 0.968 1.232 * 1.772 1.898 1.442 1.605 1.489 2.113 * 2.032 2.467 ** 2.331 2.702 1.788 2.099 2.175 2.838 1.661 1.578 2.363 1.390 1.465 2.872 1.116 1.146 2.388 0.990 1.099 1.888 1.707 1.756 2.305 1.776 1.262 1.597 1.103 1.341 1.688... Sig. tl/t2 tl/t3 t2/t3 ** ** *** *** *** RS *** *** RS *** *** RS *** *** RS *** *** *** *** Cw = Cnawarra; Bar = Barssa. *, **, *** indicate significance at the p < 0.1, p < 0.01, and p < 0.001, respectively. = n significant difference. values and their significant differences are presented in Table 7. The results fr the cmparis amngst the three treatments, i.e., enzyme hydrlysates (samples A t H), acid hydrlysates (samples I t P), and wines (samples Q t T) ver bth regi and bth years shw that bth the acid hydrlysates and enzyme hydrlysates when cmpared with the wines were significantly different fr all attributes except stalky; the enzyme and acid hydrlysates differed nly fr three attributes, hney/raisin, stalky, and licrice. Nevertheless, the level f significant difference in the three distinguishing attributes indicated that the hydrlysate types shuld be treated separately. Differences amng the samples: Fischer's LSD test (6) was used t determine the significance f the differences between individual samples hence allwing an investigatin int the surce f the differences shwn in Table 7. Cmparis between high and lw quality enzyme hydrlysates (i.e. samples A t H) and the high and lw quality acid hydrlysates (i.e. samples I t P) (Table 8) demtrated that, fr bth the tw regi and the tw years, qualities were significantly different fr at least ne attribute fr all samples except I and M. Mst f the differentiating attributes in bth the acid and enzyme hydrlysis prducts were rated quantitatively higher fr the high quality samples. With regard t the differences between the high and lw quality wines, the panel fund that the Cnawarra pair were significantly different (Fig. 3). The high quality Cnawarra wine was rated higher fr the nnberry-type attributes, i.e., stalky, earthy, cigar/tbacc, and licrice, while the lw quality wine was rated higher in berry, strawberry, and citrus. The high and lw quality Barssa wines (samples Q and S) culd nt be differentiated by descriptive analysis even thugh they were fund t be significantly different in the du-tri tests (Table 4). Descriptive analysis data fr the Barssa wines are nt shwn. s._ n 3 "" 2.,m s_ c High quality Cnawarra wine Lw quality Cnawarra wine k "k,----. ~ "k "k H/R STR CIT STK EAR C/T LIC Significantly different attributes Fig. 3. Mean attribute scres fr seven attributes shwing significant differences between high and lw quality Cnawarra wines (samples T and R). *, ** and *** indicate that differences are significant at p < 0.05, 0.01, 0.001, respectively. Refer t Table 1 fr attribute cdes.
HYDROLYZED FLAVOR PRECURSORS--173 O O ~. "! 0.5 = -0.5 0 I,.m "0 s..._ 0 0 {R t.._ 1.5 1.0 0-1.0 H/R BER STR m High quality acid hydrlysates Fl Lw quality acid hydrlysates, R/V SPC PEP STK EAR C/T High quality enzyme hydrlysates 171 Lw quality enzyme hydrlysates H/R BER STR ~ ~ E ~ SPC PEP k "k'k k "k "k LIC CHOC R/V CIT STK EAR C/T ~..~~jj Fig. 4. Each histgram shws the difference between the mean attribute scre fr a hydrlysate and the crrespnding base wine scre, fr thse attributes in which either ne r bth mean scres differ significantly frm zer i.e., the base wine. Acid hydrlysates are shwn in A and enzyme hydrlysates in B., *, ** and *** indicate differences between the high and lw quality hydrlysates are nt significant, significant at p< 0.05, 0.01 and 0.001, respectively. Refer t Table 1 fr attribute cdes. Relatihips between sery data fr precursr hydrlysates and wines: T fcus n the effects f hydrlysate additin t the base wine, the mean attribute scres f the hydrlysates were subtracted frm the crrespnding base wine scres (Figure 4). The base wine is represented in Figures 4A and 4B as zer n the y-axis. Histgrams are shwn fr the attributes where either ne r bth f the hydrlysate pairs were significantly different frm the base wine. At least ne f the attribute scres fr the quality pairs f the acid hydrlysates, presented in Figure 4A was rated significantly higher than the base wine fr seven attributes - strawberry, rse/vilet, spice, pepper, stalky, earthy, and cigar/tbacc. The high quality acid hydrlysates were given a higher mean scre fr strawberry, pepper, and the quantitatively-imprtant attribute cigar/ tbacc than their lw quality cunterparts. Five attributes were scred higher than the base wine fr at least ne f the enzyme hydrlysate quality pairs (Fig. 4B). These were rse/ vilet, citrus, stalky, earthy, and cigar/tbacc. The high and lw quality enzyme hydrlysates were significantly different fr the last fur f these attributes. Of these, stalky, earthy, and cigar/tbacc were scred significantly higher fr the high quality samples. These arma characteristics f the enzyme hydrlysates, and particularly the dminance f the nn-berry-type attributes in the high quality samples were bservable qualitatively in the PC and vectr diagrams in Figures 1 and 2. Imprtantly, when the data in Figure 3 are interpreted in relatin t thse in Figures 4A and 4B it can be seen that the nn-berry-type attributes, stalky, earthy, and cigar/ tbacc, which were quantitatively dminant in the high quality enzyme hydrlysates and, in the case f the last attribute was als dminant in the high quality acid hydrlysates, were the same as thse characterizing the high quality wine frm Cnawarra. This cmmn sery feature f the high quality samples was als bservable frm the data in Table 6 which shwed that cigar/tbacc and earthy were significant quality determinants acrss all treatments. Table 8. Sery attributes fund t be significant fr the cmparisn between high and lw quality enzyme- and acid hydrlysis prducts. Enzyme hydrlysis prducts Acid hydrlysis prducts Samples I Attributes Samples Attributes A and E Hney/raisin #, Pepper I and M C and G Stalky #, Earthy # K and O Stalky B and F Strawberry, Stalky #, Cigar/tbacc # J and N Cigar/tbacc # D and H Spicey #, Cigar/tbacc # L and P Strawberry #, Pepper # 1See Table 5 fr sample identificatin cdes. #Attribute fr which the high quality sample is mre intee. = n significant difference. Cnclusi The data presented cnfirm that the quality differences assigned t the samples were distinguishable by sery analysis and, fr all but ne pair f wines and ne pair f acid hydrlysates, these differences culd be described and quantified. The majr cntributi made by bth the acid- and enzyme hydrlysates frm Shiraz grapes t wines f this variety are nn-berry-type sery attributes. It is significant that the precursr hydrlysates frm high
174- HYDROLYZED FLAVOR PRECURSORS quality grapes, particularly thse prepared enzymatically, as well as the wine made frm these grapes were rated higher in nn-berry attributes than lw quality samples. Accrdingly, it may be deduced that the precursr hydrlysates cntain arma cmpunds which are imprtant t high quality wines f this variety. These bservati justify further studies int the chemical cmpsitin f the acid and enzyme hydrlysates f Shiraz. It is evident that these studies shuld allw bservatin f quantitative, and pssibly qualitative, differences which can be related t wine quality. It remai t be determined if the increase in nn-berry sery attributes which are bservable in Shiraz and ther red wines n maturatin can be ascribed t hydrlysis and slw release f glycsidically-bund flavr cmpunds frm the fruit. Literature Cited 1. Abbtt, N. A., B. G. Cmbe, M. A. Seftn, and P. J. Williams. Secndary metablites f Shiraz grapes as an index f table wine quality. In:Prceedings f the 7th Australian Wine Industry Technical Cnference, Adelaide, S. A. P. J. Williams, D. Davidsn, and T. H. Lee (Eds.). Australian Industrial Publishers, Adelaide, (In press, 1990). 2. Amerine, M. A., R. M. Pangbrn, and E. B. Ressler. Principles f Sery Evaluatin f Fd. p 333. Academic Press, New Yrk (1965). 3. Aryan, A. P., B. Wilsn, C. R. Strauss, and P. J. Williams. The prperties f glycsidases f Vitis vinifera and a cmparisn f their 13-glucsidase activity with that f exgenus enzymes. An assessment f pssible applicati in enlgy. Am. J. Enl. Vitic. 38:182-8 (1987). 4. Nble A. C. Analysis fwine sery prperties. In:Mdern Methds f Plant Analysis. Vl. 6. Wine Analysis. J. F. Jacksn and H. F. Like (Eds.). pp 9-28. Springer-Verlag, Berlin (1988). 5. Nble, A. C., C. R. Strauss, P. J. Williams, and B. Wilsn. Sery evaluatin f nn-vlatile flavr precursrs in wine. In: Flavr Science and Technlgy: Prceedings f the 5 th Weurman Flavr Research Sympsium. M. Marte, G.A. Dalen, and H. Russwurm Jr. (Eds.). pp 383-90. Jhn Wiley and S, New Yrk (1987). 6. O'Mahney, M. Sery evaluatin f fd. Statistical methds and prcedures. S. R. Tannenbaum, and P. Walstra (Eds.). Marcel Dekker Inc., New Yrk (1986). 7. Ressler, E. B., R. M. Pangbrn, J. L. Sidel, and H. Stne. Expanded statistical tables fr estimating significance in paired-preference, paireddifference, du-tri and triangle tests. J. Fd Sci. 43:940-7 (1978). 8. Strauss, C. R., B. Wilsn, P. R. Gley, and P. J. Williams. Rle f mnterpenes in grape and wine flavr. In: Bigeneratin f Armas. T. H. Parliment, and R. Crteau (Eds.). pp 222-42. ACS Sympsium Series 317, American Chemical Sciety, Washingtn, DC (1986). 9. Williams, P. J., M. A. Seftn, and B. Wilsn. Nnvlatile cnjugates f secndary metablites as precursrs f varietal grape flavr cmpnents. In: Flavr Chemistry: Trends and Develpments. R. Teranishi, R. G. Buttery, and F. Schahidi (Eds.). pp 35-8. ACS Sympsium Series 388. American Chemical Sciety, Washingtn, DC (1989).