Effects of Uncinula necator on the yield and quality of grapes (Vitis vinifera) and wine

Plant Pathology (2004) 53, 434 445 Doi: 10.1111/j.1365-3059.2004.01016.x Blackwell Publishing, Ltd. Effects of Uncinula necator on the yield and quality of grapes (Vitis vinifera) and wine A. Calonnec a *, P. Cartolaro a, C. Poupot b, D. Dubourdieu b and P. Darriet b a UMR Santé Végétale INRA-ENITA, 71 avenue Edouard Bourlaux, 33883 Villenave d Ornon cedex; and b Faculté d Œnologie, Université Victor Segalen Bordeaux 2, 351 cours de la Libération, 33405 Talence, France The effects of powdery mildew (Uncinula necator) on grape yield, juice and wine quality were quantified for cultivar Cabernet Sauvignon (CSa) in 1997 and 1999, and for Sauvignon blanc (Sa) in 1999. Analyses were carried out on batches of healthy berries to which known percentages (0 50%) of diseased berries were added, and on natural clusters that were classified into four visual classes from low (C 1 ) to high (C 4 ) disease severity. CSa diseased berries showed an average weight reduction of 12% (1997) and 20% (1999). The direct consequence of a higher percentage of smaller, diseased berries was a reduction in yield. The accompanying loss of weight in must from C 1 to C 4 clusters increased from 10 to 45%. Sugar content in diseased berries was not significantly different from disease-free berries in 1997, but was 20 21% (CSa) and 14% (Sa) higher in 1999. Severely infected batches also showed a higher total acidity than healthy ones. The total anthocyanin content of CSa was decreased by 0 91% (1997) and 0 66% (1999) per percentage mildewed berries added by weight. In Sa wines the concentration of 3-mercaptohexanol, a component of varietal aroma, was decreased by powdery mildew. Multidimensional analyses, based on all the variables studied, successfully grouped batches of CSa according to disease severity. Using directional triangular tests wine experts were able to recognize CSa wines produced from berries with 25% of powdery mildew; the threshold for nonexperts was 50%. CSa wines obtained from samples with more than 30% of diseased berries by weight were significantly classified as the worst according to preference order criteria, but below this value the preference was not significant. Sa wines with <50% mildewed berries could not be differentiated significantly by organoleptic tests performed by nonprofessionals. Keywords: anthocyanins, crop damage, 3-mercaptohexanol, organoleptic test, wine Introduction Powdery mildew (Uncinula necator) is the most widespread and destructive disease of grapevines worldwide, and is the main target of fungicides used on Vitis vinifera. There is no effective forecasting system for epidemics initiated by ascospores and detection of the disease during its initial stages is difficult, mainly because, in vineyards, it first appears on the lower surface of leaves. Thus treatments against powdery mildew are often used prophylactically, without regard to the risk of disease or damage to the crop; on average seven fungicide treatments are routinely applied. The lack of a reliable forecasting system, combined with a poor understanding of the link between disease severity and yield or wine quality, means that winegrowers are not encouraged to reduce fungicide *To whom correspondence should be addressed. E-mail: calonnec@bordeaux.inra.fr Accepted 29 February 2004 applications. Consequently, some wineries in the USA adopt a cautious attitude towards control, seeking to ensure that the harvest does not contain more than 3% of powdery mildew-infected berries. Similar concerns in Australia have led to the development of a polymerase chain reaction assay that can be used for early detection of very low levels of the fungus in vineyards (Stummer & Scott, 2000). In high-value vineyards, such as in the Bordeaux region of France, the policy is even more extreme, with most wine growers adopting a no disease strategy. Integrated pest management systems which are both economically viable and environmentally attractive (treatments are better targeted at an eventual risk and a susceptible phenological stage) are an increasingly popular strategy for crop production, but may require grower tolerance of higher levels of disease. Consequently a clear understanding of the links between disease and yield, and disease and wine quality, is crucial. Many studies have been conducted in an attempt to relate disease severity to yield loss, especially for powdery mildews (Jarvis et al., 2002), but with variable success. The effect of disease on 434 2004 BSPP

Uncinula necator, grapes and quality of wine 435 yield loss varies widely with time of infection, plant variety and production system (Gaunt, 1995; Madden & Nutter, 1995). Compounded by the difficulties associated with accurate disease quantification during the growing season, a link between disease severity and yield loss can be difficult to establish. In addition to its effect on yield, disease can also result in a loss in quality (Monson et al., 1986; Campbell & Duthie, 1990; Hoffman et al., 1998) and marketability, especially in fruit and flower crops (Jarvis et al., 2002). In viticulture, quality is a major concern, especially in highvalue vineyards where yields are regulated in favour of producing high-quality wines. Moreover, wine is somewhat unusual because the evaluation of its quality is much more complex and subjective than that of most other products. It is influenced by the fruit composition (sugars, acids, phenols and aroma), and by the terroir (climatic factors, soil composition), the vine-cropping system and wineprocessing technology. While the effects of Botrytis cinerea, the causal agent of grey mould, on wine quality are well described (Pallotta et al., 1998; Dubos, 1999), little is known for powdery mildew. Published results are equivocal: different authors have reported either an increase (Viala, 1893; Ough & Berg, 1979) or a decrease (Pool et al., 1984; Gadoury et al., 2001a) in the sugar content of the grape. The effect of the disease on the colour or anthocyanin content of the grape or wine has been described as positive (Ough & Berg, 1979; Pool et al., 1984) or strongly negative (Amati et al., 1996; Piermattei et al., 1999). The weight of clusters was not found to be significantly affected by the disease (Ough & Berg, 1979; Gadoury et al., 2001a). None of these studies was able to correlate grape and wine quality with precise levels of powdery mildew, perhaps because the authors have generally compared diseased berries with healthy berries, or the harvest from treated and untreated plots, using average levels of disease per plot. The relationship between final levels of disease on grapes and grape yield or quality is unlikely to be simple or linear, as it integrates the effects of the dynamics of host susceptibility and inoculum density. Even in untreated epidemics very variable symptoms can be observed, grape clusters being composed of different proportions of diseased and healthy berries with variation in severity between individual diseased berries. This paper analyses the relationships between (i) disease severity and yield (of berries and must); and (ii) disease severity and wine quality. Grape clusters are characterized according to different levels of disease, and how, in turn, disease affects average berry weight, yield of must, and the main indicators of grape quality (sugar content, anthocyanins) and wine quality (anthocyanins, tannins, varietal aroma, sensory taste). Artificially reconstituted clusters are compared with naturally produced clusters of grapes. Artificially reconstituted clusters allow the controlled incorporation of different proportions of highly but equally diseased berries with healthy berries. Variation in disease severity between individual berries is inherent in naturally produced clusters. The same analyses were performed on both types of cluster to quantify precisely the potential consequences of the disease on the yield and quality of grapes and the wine produced from them. Materials and methods Sites Samples of grapes were collected from two vineyards in the Bordeaux region, France, from the cultivar Cabernet Sauvignon (CSa) in 1997 and from the cultivars Cabernet Sauvignon and Sauvignon blanc (Sa) in 1999. In 1997 the CSa plot was located on a clayish silty soil, and in 1999 the CSa plot was located on a clayish gravely soil and the Sa plot on a clayish soil. The characteristics of the plots are described in Table 1. All the vines were protected against downy mildew (Plasmopara viticola) and B. cinerea, keeping the grapes free from these pathogens. Part of each plot was treated against powdery mildew in order to provide disease-free controls (Table 1). Disease assessments In the field, clusters of grapes were scored on a 0 4 visual scale reflecting the proportions of diseased berries: C 0, no visible powdery mildew; C 1, <25% diseased berries; C 2, 26 50% diseased berries; C 3, 51 80% diseased berries; C 4, >80% diseased berries. Samples studied The grapes were harvested on 9 September in 1997, and on 8 September (Sa) and 28 September (Csa) in 1999, based on the sugar content of healthy grapes. A number of bunches were sampled from each row in order to obtain a total of 50 kg for each disease class (C 1 C 4 ). Healthy clusters ( 150 kg) free from powdery mildew were sampled from treated vines. For grape analyses and wine making, subsamples of bunches for each disease category were randomly selected from this harvest. Analyses of grapes (disease characterization, weight, sugar and anthocyanin content) and of wine produced from them (anthocyanin and tannin content, acidity) were performed on samples constituted in various ways (Table 2). First, on clusters of berries classified into the different disease classes previously described (C 1 C 4, subsequently called C batches for Cluster); on healthy berries (C 0 ) mixed with various percentages of diseased berries (from 0 to 50% by weight; I 0 I 50, subsequently called I batches for Incorporated); and finally on diseased berries alone (I 100, for grape analyses only). For Sa, only I batches and diseased berries were studied. Diseased berries used to create I batches were selected from C 4 -type clusters. Batches of 10 13 kg of grapes (I 0 I 50 and/or C1 C 4 ) were vinified according to the protocols of Hatzidimitriou et al. (1996) and Darriet et al. (2001). Immediately following de-stemming and crushing of the healthy harvest, the must was weighed to determine the quantity of diseased berries to crush for incorporated batches.

436 A. Calonnec et al. Table 1 Characteristics of plots used to obtain different classes of powdery mildew severity on grape Year Cultivar a / vine age (years) Planting density (vines ha 1 ) Orientation of rows Plot size b (rows vines) Treated part c Position Size (rows vines) 1997 CSa/21 6767 NE SW 9 120 SW 9 12 1999 CSa/6 6600 N S d 11 36 N 11 12 Sa/12 3333 E W d 5 90 S 3 90 a CSa = Cabernet Sauvignon, Sa = Sauvignon. b Whole plots were protected against downy mildew (mancozeb/cymoxanil) and bunch rot (thiram/vinclozolin/pyrimethanil). c Part treated against powdery mildew (dinocap sulphur). d For ripeness sampling the CSa (1999) plot was divided into three parts: part 1 (11 rows 12 vines, oriented south) where the disease was most severe, part 2 (11 rows 6 vines) where the disease was less severe and part 3 (11 rows 12 vines, oriented north) treated against powdery mildew with no disease. The Sa (1999) plot was divided into two parts: part 1 (1 row 90 vines) with the highest level of disease, part 2 (rows 3 5 90 vines), the treated part, with a moderate level of disease. Table 2 Assessment of effects of powdery mildew on grapes of cultivars Sauvignon (Sa) and Cabernet Sauvignon (CSa) and on the resulting wines Analyses Variable Date Cultivar Samples analysed Type a Origin Grapes Average weight 1997 CSa Berries sorted out from C batches (C 0 C 3 ) 67 clusters 1999 CSa Berries sorted out from C batches (C 0 C 4 ) Berries sorted out from C 1 C 3 123 clusters 68 clusters Average amount of sugar 1997 CSa I (I 0 I 20 ) and C batches (C 1 C 3 ) Must from vats of 10 12 L 1999 CSa I (I 0 I 50 ) and C batches (C 1 C 4 ) I 0 ( 2), I 3 10 20, C 1, C 2 ( 2), C 3 ( 2), C 4 Must from 700 g grapes per batch Must from vats of 10 12 L Sa I 0 3 10 20 30 40 50 Must from vats of 10 12 L Wine Amount of total and 1997 CSa I 0 3 10 20, C 1 2 3 Must from 700 g extractible anthocyanins grapes per batch 1999 CSa I 0 5 15 30 50, C 1 2 3 4 Must from 700 g grapes per batch Amount of anthocyanins 1997 CSa I 0 3 10 20, C 1 2 Wine from flasks and tannins in wines 1999 CSa I 0 5 15 30 50, C 1 2 3 4 Wine from flasks Thiol concentration 1999 Sa I 0 3 10 20 30 40 50 Wine from flasks Wine tasting 1999 CSa Sa Triangular test: I 30, I 50, C 2, C 3 compared with I 0 I 0, I 15, I 30, I 50 and I 0, C 2, C 3, C 4 sorted by preference order Triangular test: I 5, I 15, I 30, I 50, C 1, C 2 compared with I 0 I 0, I 5, I 15, I 30, I 50 sorted by preference order Triangular test: I 10, I 20, I 30, I 40, I 50 compared with I 0 I 0, I 10, I 30, I 50 sorted by preference order Panel of 14 people Panel of six experts Panel of 14 people a C 0 C 4 : clusters showing different levels of disease with C 0 the least diseased and C 4 the most diseased; I 0, I 3, I 5, I 10, I 15, I 20, I 30, I 40, I 50 : healthy berries to which proportions (0, 3, 5, 10, 15, 20, 30, 40, 50%) of diseased berries were incorporated by weight. Disease characteristics Clusters were characterized according to the percentage of infected berries after separating diseased and healthy berries from 15 30 CSa clusters for each class. To obtain a more precise characterization of disease severity in 1999, the infection levels of berries were assessed at bunch closure (8 July) and berries were classified as follows: (i) slightly diseased, with only traces of mycelium (LD); (ii) more severely attacked, with at least half the surface area covered by mycelium (HD); and (iii) fully covered by mycelium and visibly smaller (FD). At harvest, due to a shortage of time berries were identified only as healthy (H), diseased (LD + HD) or severely attacked (FD). Diseased berries used to create I batches were HD and FD types in proportion 60/40. Weight and yield For red CSa grapes the average fresh weight of healthy and diseased berries was assessed from the dissected clusters at harvest. In 1999, at bunch closure, the diameter of berries was also measured for 10 berries per cluster for each class of berries (H, LD, HD, FD). Potential yield

Uncinula necator, grapes and quality of wine 437 losses were assessed from the proportion of diseased berries per cluster combined with the mean weight of healthy and diseased berries. Another assessment of yield losses was calculated from the weight of vinified clusters and the proportion of cluster types on the plot. Sugar content during ripening As healthy and diseased clusters could not be found on the same vines, in 1999 the plots were subdivided into several parts with different levels of disease in order to identify and control at harvest the potential variation of maturity (Table 1). The CSa plot was divided into three parts (1, untreated, high level of disease; 2, untreated, moderate level of disease; 3, treated, no disease) and the Sa plot into two parts (rows 3 and 5, treated, moderate level of disease; row 1, untreated, high level of disease). CSa berries were sampled three times before harvesting and on the day of harvest. The healthy berries were taken from the treated part and two untreated parts. Diseased berries were also collected on part 1. For the Sa cultivar, berries were sampled in the same way from the three rows 12 and 5 days before harvest. Berry samples were then crushed in plastic bags to obtain the must, and the sugar concentration (g L 1 must) was estimated after measuring the potential alcohol content using a hand refractometer. Sugar, acidity and anthocyanins in grapes at harvest Sugar concentration was measured, as described above, in the juice from 700 g of berry samples taken from the different batches of grapes (I and C batches and diseased berries) crushed in plastic bags (S g ), or on must taken from vats just before vinification (S v ). On the same juice samples, acidity was measured by titration with 0 1 m sodium hydroxide and anthocyanin contents were estimated by the method of Glories (1998). Extractable anthocyanins (A e ) and total anthocyanins (A t ) were determined using the principle of discoloration with sulphur dioxide (Ribéreau- Gayon & Stonestreet, 1965). Chemical analyses on wine 3-mercaptohexanol, a compound that characterizes the Sauvignon aroma, was analysed by GC-mass spectrometry (Tominaga et al., 1998). Anthocyanins and tannins were measured on clarified wines 3 months after the end of malolactic fermentation (Ribereau-Gayon & Stonestreet, 1965). Organoleptic analyses Two-directional triangular tests were carried out in 1999. These tests are based on the recognition of two identical wines between three samples. The test was based on appearance and olfaction. A panel of 14 people, including wine experts and nonexperts, carried out the first test, and a panel of six wine experts carried out the second test. They were also requested to rank wines in order of preference. Statistical analyses anova and Tukey s mean comparison tests (α = 0 05) were used to compare the different classes of clusters visually classified for their numbers of berries, their percentages of diseased berries, and their weights. A nonparametric Student s t-test was used to test the significance of the difference between the average weight of a healthy and a diseased berry per cluster. Linear regressions (least-squares method) were used to test the relationships between the percentage of powdery mildew (%PM) and the variables: sugar, anthocyanin, tannin and thiol contents. The different batches I 0, I 5, I 15, I 30, I 50, C 1, C 2, C 3 and C 4 in 1999 were also compared using multivariate methods: a principal components analysis (PCA) followed by a hierarchical clustering analysis (HCA). The aim of this analysis was to compare the different batches for several variables at the same time, and to include in the same results analyses from 700 g samples of grapes and those from the vats (100 150 clusters). Principal components analysis was carried out on the correlation matrix. Each batch or individual was described by six variables: the sugar concentration of grape samples (S g ) and of the must in vats (S v ); the extractable (A e ) and total (A t ) anthocyanin contents of the grape samples; and anthocyanin (A w ) and tannin (T w ) contents of the wine. From the matrix of coordinates of the individuals on the PCA axes, the euclidean distance between individuals was calculated and an HCA performed on this distance matrix. A dendrogram was constructed to show the similarities between the individuals and group them into homogeneous classes. Correlations between the variables and similarities between individuals are discussed. For organoleptic analyses, the directional triangular tests used the binomial distribution with a one-third probability of giving the correct answer by chance. To determine which differences were significant at a probability of 5%, nine and five correct answers were necessary from panels of 14 and six people, respectively. Non-parametric Friedman rank tests, which model the ratings of n judges on k treatments (wines), were performed to establish an overall significant difference between the sum of ranks obtained for each wine after their classification in order of preference. Multicomparison tests were used to determine which of two wines was significantly preferred. The statistical processing of the results was done using SAS (SAS/STAT, 1990) for anova (glm procedure) and Tukey s test, and s-plus (Statistical Sciences, 1995) for regression analyses (lm function), PCA (ca function), and hierarchical classifications (hclust function method = average). Friedman s rank tests, multiple comparison tests and Student s t-tests were computed using excel (Microsoft). Results Characterization of the disease The percentage of diseased berries per cluster in CSa, assessed at harvest, varied significantly with the disease

438 A. Calonnec et al. Figure 1 Percentage of the different types of mildewed Cabernet Sauvignon (CSa) berries in the four different types of clusters. LD, slightly diseased berries with only traces of mycelium; HD, more severely diseased berries with at least half of the surface area covered by mycelium; FD, berries fully covered by mycelium and visibly smaller. Table 3 Effects of powdery mildew on numbers and weight of Cabernet Sauvignon grapes 1997 1999 Harvest Bunch closure Harvest Variables studied Type a N b Mean N Mean N Mean Average weight H 67 1 50 98 0 31 61 0 98 of a berry D 39 1 32 123 0 28 68 0 78 per cluster (g) Average number C 0 22 126 55 27 133 78 of berries per C 1 15 112 00 21 133 52 17 138 71 cluster C 2 15 111 47 28 151 32 21 115 43 C 3 15 107 13 26 127 65 30 117 70 C 4 25 127 92 Average weight C 0 22 191 67 27 42 24 of a cluster (g) C 1 15 165 74 21 41 89 17 123 92 C 2 15 151 26 28 45 27 21 96 96 C 3 15 140 61 26 33 86 30 95 12 C 4 25 31 57 a Berry state: healthy (H) or diseased berries (LD + HD + FD); or cluster type (C 0 C 4, C 0 = no diseased berrries and C 4 = more than 80% of diseased berries on visual scale). b Number of clusters used in analysis. class (df = 63, F = 141 2, P = 0 0001 in 1997; df = 65, F = 41 8, P = 0 0001 in 1999) and was on average higher in 1999 than in 1997 (Fig. 1). In 1997, as a consequence of a less severe epidemic, there were no FD berries. At bunch closure in 1999, the higher proportion of diseased berries in classes C 3 and C 4 was associated with a higher proportion of more severely attacked berries (HD and FD) and a lower number of berries with only traces of mycelium (LD) (Fig. 1). For C 4 clusters, the proportion of FD berries was >40%. Consequently the severity of disease assessed as the area of the cluster covered by mycelium was not directly proportional to the severity assessed as the percentage of diseased berries, being much higher in C 3 and C 4 clusters. Weight and size of diseased grapes Diseased berries from C 1 to C 3 clusters were significantly lighter than healthy ones, averaging 1 32 compared with 1 50 g per berry in 1997 [P(T t) < 0 0001] and 0 78 compared with 0 98 g in 1999 [P(T t) < 0 0001], corresponding to weight decreases of 12 and 20%, respectively (Table 3). FD berries showed the greatest reduction with an average weight of 0 43 compared with 0 81 g for LD + HD berries. In 1999, at bunch closure 80 days before harvest, diseased berries were on average 9 7% lighter than healthy ones (39% for FD berries). At this date the mean diameter of diseased berries was 6 6 compared with 7 3 mm for healthy berries in cluster classes C 2 C 4. Weight loss mainly occurred in diseased berries, and there was no significant difference in weight between healthy berries coming from the different classes of clusters, except in 1997 when the healthy berries of infected class C 3 were 13% lighter than those from C 1. In 1999 the mean weight of a C 3 cluster was significantly lower than that of a C 1 cluster (95 1 vs 123 9 g, df = 65, F = 3 27, P = 0 04). The number of berries per cluster was not significantly different between the samples studied. Based on the average weight of H, HD and FD berries, their frequencies within the different clusters and the total number of berries per cluster, yield losses were estimated to be about 4, 25 and 32% in 1999, and 13, 15 and 22% in 1997, for C 1, C 2 and C 3 clusters, respectively. At vinification in 1999 the loss of must produced from an average of 210 diseased compared

Uncinula necator, grapes and quality of wine 439 Table 4 Sugar, anthocyanin and 3-mercaptohexanol concentrations in juice and wine from Cabernet Sauvignon (Csa) and Sauvignon (Sa) grapes in 1997 and 1999, and their relationship with percentages of mildewed berries in different batches Anthocyanin (mg L 1 ) c Year and cultivar Batch %PM a Sugar (g L 1 ) b In must In wine S g S v A t A e A w 3-mercaptohexanol (ng L 1 ) in wine 1997 I 0 0 182 1213 735 332 CSa I 3 3 183 1067 653 315 I 10 10 202 1061 587 315 I 20 20 189 791 595 302 HD 100 240 C 1 12 1 191 874 501 C 2 31 2 183 784 445 C 3 62 9 185 520 355 1999 I 0 0 176 173 1251 583 759 720 d CSa I 5 5 181 185 1346 599 766 I 15 15 187 187 1243 560 728 I 30 30 195 190 1092 506 681 I 50 50 198 197 686 422 631 HD 100 215 518 353 5 FD 100 220 C 1 19 1 (4 7) 185 175 1463 612 5 783 C 2 48 (23 9) 188 181 1171 579 3 778 723 C 3 75 2 (63 6) 203 192 1110 581 698 706 C 4 100 (96 8) 220 218 607 3 346 5 476 1999 I 0 0 182 360 Sa I 3 3 182 410 I 10 10 187 453 I 20 20 184 334 I 30 30 193 252 I 40 40 191 220 I 50 50 196 138 a Percentage of diseased berries by weight assessed at harvest including all diseased berries (from traces of mycelium to berries fully covered with mycelium, LD + HD + FD). In brackets: percentages of diseased berries at least half covered by mycelium at bunch closure (HD + FD). b S g, sugar content in must extracted from 700 g samples of grapes; S v, sugar content in must sampled from vats before vinification. c A t = total anthocyanins in must; A e = extractable anthocyanins in must; A w = anthocyanins in wine. d Results from two vats. with healthy clusters was 10% (C 1 ), 26% (C 2 ), 34% (C 3 ) and 45% (C 4 ). From the mean weight of the different cluster classes at vinification and their relative frequencies (0 81% C 0, 18 4% C 1, 20 5% C 2, 25 6% C 3, 34 6% C 4 ) the total loss of yield in the untreated area was estimated to be 23 7%. Total acidity and sugar, anthocyanin and tannin contents In 1997 at harvest, there was no significant difference in sugar content between sample batches in vats. Conversely, in 1999 in both CSa and Sa grapes, diseased berries had a higher sugar content than healthy ones (Table 4). The difference could not be attributed to a location effect as it was observed on infected and healthy berries from the same parts of the plot (Fig. 2). At harvest, a linear relationship was observed between the proportion of diseased berries added and the increase in sugar content both in 700 g CSa grape samples (S g = 0 36 %PM + 179 9, R 2 = 0 97) and in samples taken from vats before vinification (S v = 0 38 %PM + 138 5, R 2 = 0 81 for CSa vat samples; S v = 0 26 %PM + 182 1, R 2 = 0 83 for Sa vat samples). Diseased berries contained 20 21% (CSa) and 14% (Sa) more sugar than healthy ones, corresponding to an increase in sugar content of 0 2 and 0 14% per percentage diseased berries added. In 1999 the sugar content of C batches also differed with disease class (C 1 < C 2 < C 3 < C 4 ). However, the sugar content of these batches was most closely related to the percentage of HD and FD berries assessed at bunch closure. Models describing the relationship between sugar content and percentage of diseased berries for I plus C batches were very similar to those determined for I batches only, both for the grape samples (S g = 0 38 %PM + 180 1, R 2 = 0 97) and for the vat samples (S v = 0 39 %PM + 176 1, R 2 = 0 86). These results demonstrate a good correlation between the sugar content of the must extracted from a few clusters (700 g) and that extracted from larger samples of clusters and sampled at vinification in 10 L vats. In 1999 I 50 and C 4 CSa clusters had a higher total acidity than healthy ones: 5 7 g L 1 for I 50 and 7 2 g L 1 for C 4 against 4 9 g L 1 for I 0.

440 A. Calonnec et al. Figure 2 Sugar content of healthy (H) and mildewed (D) berries sampled before and at harvest from cultivars Cabernet Sauvignon (CSa) and Sauvignon (Sa) in 1999., H part 1 or H row 1; H part 2 or H row 3 5;, H part 3, healthy berries sampled in different parts of the row. In the treated part (part 3 for Csa and rows 3 5 for Sa) two berries per vine on every other vine were sampled. In untreated parts (parts 1 and 2 for Csa and row 1 for Sa) one berry per stock for all stocks were sampled., D part 1 or D row 1, diseased berries sampled on part 1 (Csa) and row 1 (Sa), with the highest level of disease. A significant decrease in total and extractable anthocyanins was related to the percentage of diseased berries (Table 4). In 1997 the loss in total anthocyanins was equivalent to 0 91% per percentage diseased berries added by weight to healthy berries (A t = 10 01 %PM + 1102 3, R 2 = 0 86). In 1999, for I batches, the decrease in anthocyanin content was significant, with a loss in anthocyanins in wine of 0 66% for each 1% of diseased berries added by weight (Table 4; Fig. 3). C batches had a higher anthocyanin content according to the percentage of diseased berries assessed at harvest. However, most C batches fell within the bands of simultaneous confidence limits for the regression determined for I batches on the basis of the percentage of diseased berries (HD + FD) assessed at bunch closure (Fig. 3). Regression lines including all batches were then very close to those with I batches only (A t = 7 9 %PM + 1357 5, R 2 = 0 82; A e = 2 4 %PM + 606 6, R 2 = 0 75; A w = 2 2 %PM + 774 3, R 2 = 0 69). In 1999 the anthocyanin content of wine (A w ) was higher than that of the extractable anthocyanin in grape samples (A e ) (Table 4; Fig. 3). This could be explained either by insufficient maceration leading to a lower level of anthocyanin in the grape samples, or by the repeated puncturing of the skins during microvinification favouring Figure 3 Relationship between total (A t ) and extractable (A e ) anthocyanins in must of grapes; anthocyanin in different vats of wine (A w ); and percentage of mildewed berries added by weight to healthy ones in 1999., Anthocyanin content of I batches (diseased berries added to healthy ones at 0, 5, 15, 30, 50% by weight) and HD berries., Anthocyanin content of C batches as a function of percentage of total diseased berries assessed at harvest (19 1, 48, 75 2, 100%). +, Anthocyanin content for C batches as a function of percentage of diseased berries (HD + FD) assessed at bunch closure (4 7, 23 9, 63 6, 96 8%). Solid line, regression line plotted for I batches alone; dashed line, simultaneous confidence bands for regression line using the F distribution. the extraction of anthocyanins from the skin and increased levels in the wine (Y. Glories, personal communication). The slopes of models for A w and A e were, however, not significantly different, so batches could be compared. In 1999 the tannin content of the wine (T w ) was tested, and shown to increase with the percentage of diseased berries (T w = 0 02 %PM + 2 64, R 2 = 0 97 for I batches and T w = 0 096 %PM + 2 7, R 2 = 0 62 for all batches).

Uncinula necator, grapes and quality of wine 441 Figure 4 Principal components analysis on the quality components of juice and wine data from cv. Cabernet Sauvignon in 1999. (a) Correlation circle, variables (A t, A w, A e, S g, S v, T w ), are projected on the first two axes. The closer the arrows are to the circle, the better the variable is represented on the 1 2 plane. The smaller the angle between the arrow and the axis, the better the variable is represented on this axis. (b) Location of individuals relative to the others in the first two-dimensional space. Figure 5 Comparison of batches (C 0, C 1, C 2, C 3, C 4, I 0, I 5, I 15, I 30, I 50 ) for anthocyanin concentrations in grapes and wine, sugar in grapes and must, and tannins in wine, using a hierarchical clustering analysis with the average method performed on the individual coordinates of the principal components analysis. Tree grouping depending on distance (1 minus similarity) between individual batches. Numbers indicate average distance between an individual joining the group and all individuals from the previous group. When the variables studied (S g, A e, A t, S v, A w, T w ) were compared in correlation analyses for all CSa batches in 1999, strong positive correlations were observed between the sugar content of grape samples and of the must in vats (S g S v, R = 0 94); the anthocyanins in grapes and wine (A t A w, R = 0 92; A e A w, R = 0 96); and the total and extractable anthocyanins in grapes (A t A e, R = 0 95). Negative correlations were observed between anthocyanins in grapes and tannins in wine (A t T w, R = 0 93; A e T w, R = 0 94). Even when I 50 and C 4 samples were excluded, correlations between grape samples and wines for anthocyanin and sugar contents were still high (R = 0 8), showing that grape sample analysis can give a satisfactory assessment of the sugar and anthocyanin contents of wine. However, measurements on wines are generally much easier to perform. Principal components analysis was performed on all batches: the six variables, with I 50 and C 4 as supplementary individuals, accounted for 59 and 63% of inertia (Fig. 4). The first axis was well represented by the six variables (from 13 7 to 19 5%) and separated individuals with a high anthocyanin content (I 0, C 1, I 5 ) from those with a high sugar (C 3 ) and/or tannin (I 30, I 50, C 4 ) content. The second axis (12% of inertia) separated individuals with a higher sugar and extractable anthocyanin content (A e ) from those with high tannin content. The cluster tree drawn following hierarchical clustering analysis significantly differentiated individuals into four (66% of inertia) to six (83% of inertia) groups (Fig. 5). At 66% of inertia, individuals with the highest anthocyanin content and the lowest sugar content (I 0, I 5, C 1, I 15, C 2 ) were significantly different from the other four individuals (I 30, C 3, I 50, C 4 ). At 83% of inertia a distinction was made between I 0, I 5, C 1 (group 1) and I 15, C 2 (group 2); and between I 30 (group 4) and C 3 (group 3). The distance between C 3 and I 30 may be explained by the higher proportion of FD berries in I 30 (23 4%) than in C 3 (12 6% at

442 A. Calonnec et al. bunch closure, 15% at harvest), thereby decreasing its anthocyanin and increasing its tannin content. Therefore, according to the anthocyanin, sugar and tannin contents in grapes, in must before vinification or in wine, I 0 I 5 batches could be distinguished from I 15, I 30 and I 50. Clusters C 1, C 2 and C 3 could be grouped with I 5, I 15 and I 30, respectively, which suggests that disease levels in clusters were overestimated with respect to their effect on grapes and wine. 3-mercaptohexanol concentration The Sauvignon wines showed a significant linear decrease in 3-mercaptohexanol with increasing mildew (3MH = 5 4 %PM + 427, R 2 = 0 84) which reached 38% for I 30 and 63% for I 50 (Table 4). A 20% decrease is significant considering the olfactory perception threshold of this compound (60 ng L 1 ). Organoleptic test During the first triangular tests performed by 14 nonexperts with CSa wines from C 2, C 3, I 30 and I 50 batches, only I 50 and C 3 (>50% of diseased berries) were distinguished with a significance of 1% (12 and 10 people out of 14, respectively); the other two wines (C 2 and I 30 ) were not differentiated even at a 5% level of significance. The six experts carrying out the second test significantly distinguished the I 30, I 50 and C 2 batches, but not I 5, I 15 or C 1. From these tests nonexperts were able to distinguish wines with at least 50% of powdery mildew-infected berries, whereas professionals differentiated wines with 25% of mildewed berries (C 2 and I 30 ). The preference test performed with 14 people was the most discriminating. The control (I 0 ) had the lowest rank sum and was significantly preferred to I 30 and I 50 in one test, and to C 3 and C 4 in the other (Fig. 6). CSa wines with at least 30% diseased berries by weight (41% in number) could be distinguished, albeit less well. Below this level the preference was not significant. For Sa wines the panel of 14 people did not significantly differentiate the batches based on the triangular tests. The most recognizable batch was I 50 with nine people out of 14 identifying it as the worst. I 15, I 0 and I 30 had a very close rank sum for preference order and were not significantly distinguished; however, they were significantly different from I 50 (Fig. 6). The threshold of recognition was higher for Sa wines than for red CSa wines. Discussion This study analyses the effects of powdery mildew on the composition of diseased grapes, and on the resulting wines. These effects were most pronounced when disease was severe. For grapes, the modified characteristics were yield, anthocyanin and sugar content. In wine, anthocyanins, tannins and compounds affecting varietal aroma were modified. The effect of powdery mildew on grape yield was caused by an increase in the percentage of diseased berries of reduced size, and probably also of a reduction in the number of berries per cluster. Berries attacked by powdery mildew showed delayed growth (decrease in berry size) which was already significant at bunch closure, the end of the first growth period of a berry. During this phase a berry builds up its structure, develops seeds, and produces and stores large quantities of organic acids for cell division and expansion, normally leading to rapid growth (Davies & Robinson, 2000; Ollat et al., 2002). Berry set and size are mainly limited by carbon availability during early stages of development, and most of the sugar imported into the berry is metabolized, with little storage. The growth reduction observed in diseased berries could be the consequence of modification of the source sink relationships for carbohydrates in the Figure 6 Organoleptic comparison of Cabernet Sauvignon (Csa) and Sauvignon (Sa) wines in preference order tests made by a panel of 14 people in 1999. P values <0 05 indicate an overall significant difference between wines after performing Friedman s test. Wines connected by a line are not significantly different according to a multiple comparison test.

Uncinula necator, grapes and quality of wine 443 developing berry, with powdery mildew competing for the glucose and fructose. A possible increase in the invertase activity in diseased leaves with an increased partitioning of photosynthate into starch rather than sucrose, as observed in wheat leaves infected with Blumeria graminis (Wright et al., 1995), may also limit the translocation of sucrose from leaves to clusters, with an increasing effect in diseased berries. In the current experiments there was no observed significant difference in weight between healthy berries from the different cluster types, which suggests a direct effect of powdery mildew on the attacked berries rather than an indirect effect through the leaves. Although not measured in these experiments, a loss in wines produced from diseased berries would then be expected due to a decrease in the volume of extractable juice, the highly diseased berries being smaller. The economic impact of losses in wine would be highly dependent on the management of the vineyard. For growers adopting a general policy of restricting yield to favour wine quality (e.g. through a balanced cropping system, inter-row grass, green harvest), the slightest yield loss could have a high economic impact. However, more flexible cropping systems that allow adjustment of yield later in the season are more able to compensate for losses due to disease. In contrast to the results presented here, Pool et al. (1984) and Gadoury et al. (2001a) did not find any difference between the cluster weight of cv. Rosette (an interspecific hybrid) and cv. Concord (Vitis labruscana) harvested from plots sprayed or unsprayed against powdery mildew. This may have been due to more moderate levels of disease, or to an overall approach that did not account for variability introduced by differences in disease levels in the different clusters. Moreover, these studies involved very different cropping systems from those studied here. Some cultivars may be less affected by the powdery mildew sink effect, storing higher amounts of sucrose in the berry, as is the case for some cultivars derived from V. labruscana (Takayanagi & Yokotsuka, 1997). However, Pool et al. (1984) observed a reduction in the number and weight of clusters per vine stock in untreated plots with a high level of disease. An increase in sugar concentration was previously reported by Viala (1893) in some years, and once by Ough & Berg (1979) for three different cultivars (Thompson seedless, Carignane, Ribier). Results presented here suggest that this increase is the consequence of a reduction in the volume of, and an increased transpirational water loss by, diseased berries, enhanced by cracked skin. This may be related to climatic conditions affecting the structure and quality of the skin, as observed in 1999 for both white Sa and red CSa grapes. The generally lower level of phenolic compounds in the skin measured in 1999 compared to 1997 indicates a lower parietal cohesion as early as berry setting (Dubos & Roudet, 2003), which may mean that an increase in sugar is not always reported. The slight decrease in sugar content observed by some authors (Pool et al., 1984; Gadoury et al., 2001a, 2001b) was related to high yield under an unbalanced cropping system (low density of 1500 vines ha 1, with no control of vigour and yield). In these experiments the decrease in anthocyanin levels in berries reached 0 7 0 9% per percentage mildewed berries added to the healthy harvest. Amati et al. (1996) and Piermattei et al. (1999) have shown a total decrease in anthocyanins of 40 70% in diseased grapes from cv. Sangiovese, depending on the type of anthocyanin. Piermattei et al. (1999) have also observed increases in caftaric acid, a common phenolic compound, and in the phytoalexin trans-resveratrol, both of which are involved in plant disease resistance. There is, however, no evidence that the decrease in anthocyanin is a consequence of an increase in the defence compounds produced by the vine, or that powdery mildew is sensitive to them. The grape is the plant component most susceptible to powdery mildew between flowering and bunch closure, during a period when UVc is most active in stimulating phytoalexin synthesis (Bais et al., 2000). The effect of powdery mildew on the decrease in anthocyanin could be more detrimental in vintages with a low level of colour, as in 1997 when the terroir was less favourable for CSa compared to 1999. The level of anthocyanin is probably the best indicator for determining the deleterious effect of powdery mildew on wine. The incidence of powdery mildew infection on 3- mercaptohexanol, a compound that characterizes the Sauvignon blanc aroma, was studied for the first time. The concentration of 3-mercaptohexanol decreased in relation to the percentage of powdery mildew-infected berries but, at tasting, the difference between wines was clearly discriminated only at the 50% infection level. In the CSa wine tasting, a difference from the control was detected when the proportion of diseased berries reached 25% (as in C 2 clusters). Wines with 40 60% (I 30 and C 3 ) of diseased berries (30 50% by weight) were classified as the worst. This is in accordance with the analytical results summarized by the hierarchical classification that takes into account the effects on sugar, anthocyanin and tannin levels in grapes and/or wines. The fungal aroma, which is due to potent odorous compounds, was not perceived in the wines, suggesting that it was at least partially degraded during vinification (Darriet et al., 2002). This study also characterized the disease symptoms which determine the effects of the fungus on grapes and wine. The effects of C batches on quality were systematically less severe than expected when compared with those observed for diseased berries added to healthy ones (I batches) in proportions equivalent to those in the C clusters. This is because only diseased berries at least half covered by mycelium were used for I batches whereas, for clusters, even lightly diseased berries are included and their proportion depends on the disease class. The increased proportion of the area covered by mycelium in the highest disease cluster class was directly related to the epidemiology of the disease. The period of high susceptibility of the grape to the fungus is between flowering and bunch closure, and disease severity on the clusters is dependent both on the distance from the primary foci and on the earliness of the attack. Therefore C 4 clusters were those that were mainly localized around the primary foci

444 A. Calonnec et al. and that had much more severely diseased berries than C 3. The use of batches with increasing proportions of added diseased berries allowed the assessment of the average percentage of heavily diseased berries that had a deleterious effect on wine. The comparison of incorporated batches and entire clusters, and the characterization of severity, emphasize the relationship between the symptomatology of the disease and its effect on grapes and wine. The results need to be confirmed on other cultivars and vintages with larger volumes of wine. However, they offer the prospect of future integrated management programmes against powdery mildew. In such programmes it will also be necessary to examine the interactions between more than one pathogen. For example, some evidence suggests that powdery mildew may predispose the grape to infection by other fungi, bacteria and yeasts (Gadoury et al., 2002). Moreover, the proliferation of natural grape microflora and of other fungi responsible for rots may be enhanced by a general decrease in the use of fungicides, such as those used against Botrytis or downy mildew (Ducommun & Corbaz, 1982) and copper, which have an effect on bacteria. These secondary organisms, however, usually multiply under specific environmental climatic conditions, late in the season on an over-ripening grape. In the conditions in the current study (plots treated against Botrytis and downy mildew and naturally infected with powdery mildew), secondary organisms were sometimes observed at high disease levels where skins were damaged, but never where disease levels were moderate or low enough to be acceptable to the grower. Acknowledgements We are grateful to the INRA team for research on experimental fungicides who kept our experimental plots free from all undesirable pathogens and for their technical help in sorting diseased berries from kilograms of clusters. We also thank the INRA Santé des Plantes et Environnement department (programme AIP-Ecopath) and the Conseil de Région Aquitaine for their financial support. We thank Annik Lacombe (INRA, Translation and Terminology Unit, Jouy-en-Josas) for revising the manuscript. References Amati A, Piva A, Castellari M, Arfelli G, 1996. Preliminary studies on the effect of Oidium tuckeri on the phenolic composition of grapes and wines. Vitis 35, 149 50. Bais J, Murphy P, Dry I, 2000. The molecular regulation of stilbene phytoalexin biosynthesis in Vitis vinifera during grape berry development. Australian Journal of Plant Physiology 27, 425 33. Campbell C, Duthie J, 1990. Impact of leaf spot diseases on yield and quality of alfalfa in North Carolina. Plant Disease 74, 241 5. Darriet P, Bouchilloux P, Bugaret Y, Clerjeau M, Poupot C, Dubourdieu D, 2001. Effect of copper fungicide spraying on volatile thiols of the varietal aroma of sauvignon blanc, cabernet sauvignon and merlot wines. Vitis 40, 93 9. Darriet P, Pons M, Henry R, Dumont O, Findiling V, Cartolaro P, Calonnec A, Dubourdieu D, 2002. Impact odorants contributing to the fungus type aroma from grape berries contaminated by powdery mildew (Uncinula necator) incidence of enzymatic activities of the yeast Saccharomyces cerevisiae. Journal of Agriculture and Food Chemistry 50, 3277 82. Davies C, Robinson S, 2000. Differential screening indicates a dramatic change in mrna profiles during grape berry ripening. Cloning and characterization of cdna encoding putative cell wall and stress response proteins. Plant Physiology 122, 803 12. Dubos B, 1999. La pourriture grise. In: Dubos B, ed. Maladies Cryptogamiques de la Vigne: Champignons Parasites des Organes Herbacés et du Bois de la Vigne. Bordeaux, France: Éditions Féret, 55 68. Dubos B, Roudet J, 2003. Early evaluation of grape berry susceptibility to Botrytis cinerea. In: European Meeting of the IOBC/WPRS Working Group Integrated Protection in Viticulture, Volos, Greece, 2003. OILB/SROP Bulletin 26, 59 62. Ducommun P, Corbaz R, 1982. Influence of blue mould fungicides on the tobacco leaf mycoflora. Phytopathologishe Zeitschrift 104, 19 26. Gadoury D, Seem C, Pearson RC, Wilcox WF, 2001a. Effects of powdery mildew on vine growth, yield and quality of Concord grapes. Plant Disease 85, 37 140. Gadoury D, Seem C, Ficke A, Wilcox WF, 2001b. The epidemiology of powdery mildew on Concord grapes. Phytopathology 91, 948 55. Gadoury D, Seem C, Wilcox W, Henick-Kling T, 2002. Effects of powdery mildew on bunch rots, berry microflora, and juice and wine quality. Davis, CA, USA: University of California, Proceedings of the 4th International Workshop on Powdery and Downy Mildew in Grapevine, 2002, California. In: Gadonry D, et al., eds. 64 5. Gaunt RE, 1995. The relationship between plant disease severity and yield. Annual Review of Phytopathology 33, 119 44. Glories Y, 1998. Chimie du vin stabilisation et traitements. In: Ribéreau-Gayon P, Glories Y, Maujean A, Dubourdieu D, eds. Traité D oenologie, Tome 2. Paris, France: Dunod, 200 2. Hatzidimitriou E, Bouchilloux P, Darriet P, Clerjeau M, Bugaret Y, Poupot C, Dubourdieu D, 1996. Incidence d une protection viticole anticryptogamique utilisant une formulation cuprique sur le niveau de maturité et l arôme variétal des vins de Sauvignon. Bilan de trois années d expérimentation. Journal International des Sciences de la Vigne et du Vin 30, 133 50. Hoffman D, Hartman G, Mueller D, Leitz R, Nickell C, Pedersen W, 1998. Yield and seed quality of soybean cultivars infected with Sclerotiorum. Plant Disease 82, 826 9. Jarvis WR, Gubler WD, Grove GG, 2002. Epidemiology of powdery mildews in agricultural pathosystems. In: Bélanger RR, Bushnell WR, Dik AJ, Carver TLW, eds. The Powdery Mildews: A Comprehensive Treatise. St Paul, MN, USA: APS Press, 169 99. Madden L, Nutter F, 1995. Modeling crop losses at the field scale. Canadian Journal of Plant Pathology 17, 124 37. Monson W, Hanna W, Gaines T, 1986. Effects of rust on yield and quality of pearl millet forage. Crop Science 26, 637 9.