Food Chemistry 138 (2013) Contents lists available at SciVerse ScienceDirect. Food Chemistry

Food Chemistry 138 (2013) 1696 1705 Contents lists available at SciVerse ScienceDirect Food Chemistry journal homepage: www.elsevier.com/locate/foodchem Relationships between harvest time and wine composition in Vitis vinifera L. cv. Cabernet Sauvignon 1. Grape and wine chemistry Keren Bindon a,, Cristian Varela a, James Kennedy a,b, Helen Holt a, Markus Herderich a a The Australian Wine Research Institute, P.O. Box 197, Glen Osmond, South Australia 5064, Australia b California State University Fresno, Department of Viticulture & Enology, 2360 E. Barstow Avenue MS VR89, Fresno, CA 93740-8003, USA article info abstract Article history: Received 14 August 2012 Received in revised form 28 September 2012 Accepted 28 September 2012 Available online 10 November 2012 Keywords: Grape Ripeness Wine Methoxypyrazine C 6 alcohols Ester Volatile Aroma Tannin Phenolic Polysaccharide The study aimed to quantify the effects of grape maturity on wine alcohol, phenolics, flavour compounds and polysaccharides in Vitis vinifera L. cv Cabernet Sauvignon. Grapes were harvested at juice soluble solids from 20 to 26 Brix which corresponded to a range of wine ethanol concentrations between 12% and 15.5%. Grape anthocyanin and skin tannin concentration increased as ripening progressed, while seed tannin declined. In the corresponding wines, monomeric anthocyanin and wine tannin concentration increased with harvest date, consistent with an enhanced extraction of skin-derived phenolics. In wines, there was an observed increase in yeast-derived metabolites, including volatile esters, dimethyl sulfide, glycerol and mannoproteins with harvest date. Wine volatiles which were significantly influenced by harvest date were isobutyl methoxypyrazine, C 6 alcohols and hexyl acetate, all of which decreased as ripening progressed. The implications of harvest date for wine composition is discussed in terms of both grape composition and yeast metabolism. Ó 2012 Elsevier Ltd. All rights reserved. 1. Introduction During grape ripening, multiple biochemical processes occur at different rates, and are stage-specific (Zamboni et al., 2010), consequently the chemical composition of grapes destined for winemaking can be defined temporally. For grape-derived compounds which may positively or negatively influence wine chemistry and sensory properties, these may be in a state of increasing, decreasing or remaining constant at a given point in grape development. Hence, decision-making in terms of harvest date for commercial winemaking requires the consideration of complex factors. A harvesting decision for optimal ripeness requires adequate knowledge of grape compositional factors relevant to achieve a targeted wine style, taking into consideration the grape cultivar, climate, topography, seasonal weather conditions and vineyard management practices. Traditional measures used to determine grape ripeness include the assay of juice total soluble solids as an estimate of grape sugar accumulation, or an estimate of grape acidity decline as titratable acids or ph (Jackson & Lombard, 1992). However, there is general recognition that these measures alone are not sufficient to accurately predict wine composition, notably as many key grape-derived compounds do not track with sugar Corresponding author. Tel.: +61 8 83136190; fax: +61 8 83136601. E-mail address: keren.bindon@awri.com.au (K. Bindon). accumulation, and are also highly dependent upon the grapevine genotype and its environment (Jackson & Lombard, 1992). In terms of grape cultivars, the variety Vitis vinifera L. cv. Cabernet Sauvignon presents an interesting case study due to a frequently observed dichotomy of sensory attributes, distinguishing vegetative from fruity characters (Robinson et al., 2011). Advances in flavour chemistry have shed light on this commercially important grape variety, defining potential flavour precursors in grapes and predicting their sensorial impact in wine (Forde, Cox, Williams, & Boss, 2011; Kalua & Boss, 2009; Roujou de Boubeé, Van Leeuwen, & Dubourdieu, 2000). While vegetative characteristics in Cabernet Sauvignon have long been thought to be due to the presence of isobutyl methoxypyrazine (IBMP) which denotes a green bell pepper attribute (Rojou de Boubeé et al., 2000; Cacho & Ferreira, 2010; Robinson et al., 2011), the C 6 alcohols and their derivatives have also been implicated in this sensory attribute as herbaceous or green (Escudero, Campo, Farina, Cacho, & Ferreira, 2007; Kalua & Boss, 2009). IBMP is known to consistently decrease during grape ripening, which has potential implications for vegetative aroma in wines according to harvest date (Rojou de Boubeé et al., 2000). Grape C 6 derivatives in situ can exist as acetate esters, aldehydes or alcohols, with C 6 -alcohols predominating during the later stages of grape ripening (Kalua & Boss, 2009). However, in grape crushing and during fermentation there is the potential de 0308-8146/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.foodchem.2012.09.146

K. Bindon et al. / Food Chemistry 138 (2013) 1696 1705 1697 novo generation of additional C 6 derivatives from fatty acid precursors via the lipoxygenase pathway. The conversion of C 6 derivatives to acetates by yeast activity, may alter the expected herbaceous profile, in particular since hexyl-acetate may be associated with fruity attributes (Dennis et al., 2012; Forde et al., 2011). While impact odourants are important in conferring wine sensory attributes, Escudero et al. (2007) and Cacho (2010) have emphasised that synergistic interactions between wine volatiles can lead to the enhancement or dampening effects on aroma/ flavour attributes. For example, in terms of wine fruitiness, yeast-derived esters which are largely responsible for this attribute can either mask vegetative odours, or ester aroma may be enhanced by the presence of norisoprenoids or dimethyl sulfide in low concentrations (Escudero et al., 2007). In terms of assessing the effects of grape ripening on wine volatiles, it is therefore also important to recognise the significant role of yeast metabolites. For Cabernet Sauvignon, astringency has been found to be a significant factor which discriminates wine palate, or taste (Robinson et al., 2011). Interestingly, this has been found to be negatively correlated with vegetative aroma, and strongly associated with the concentration of condensed tannin (proanthocyanidin) in wines (Robinson et al., 2011). In red grape varieties like Cabernet Sauvignon, the assessment of compositional changes in grape phenolics during ripening, denoted phenolic ripeness has been emphasised as an important consideration (Jackson & Lombard, 1992; Perez- Magarino & Gonzalez-San Jose, 2006). To date, a clear relationship between the tannin concentration of grapes, and the resulting tannin concentration in wines has not been demonstrated, in particular with respect to grape ripeness (Fournand et al., 2006). Work by Ristic, Bindon, Francis, Herderich, and Iland (2010) has demonstrated a strong relationship between grape skin tannin concentration and wine tannin concentration, while seed tannin showed a poor relationship. This finding, and the observation that a higher tannin concentrations together with a higher proportion of skin tannin are correlated with increases in commercial wine allocation grading (Kassara & Kennedy, 2011), highlights the need for a greater understanding of tannin partitioning from grape solids to wine as it relates to grape ripening. Further, accumulation of the coloured pigments anthocyanins occurs following the onset of ripening (veraison), and increased grape colour has been correlated with enhanced wine colour and ageing capacity (Perez-Magarino & Gonzalez-San Jose, 2006). Anthocyanins have the capacity to bind to tannin under the conditions of vinification (Hayasaka & Kennedy, 2003), and changes in this class of phenolics therefore has significant implications for both wine astringency, and colour stability as bisulphite-resistant (polymeric) pigments. The current study provides a holistic overview of key components of the variety Cabernet Sauvignon, tracking composition from grape to wine. While limited to a single season, we have aimed to present a case study whereby wine composition can be inferred from the temporal changes in grape composition during ripening. This study has sought to incorporate the impact of both grape biochemistry and yeast metabolism on wine composition in order to account for differences in classes of compounds. 2. Materials and methods 2.1. Instrumentation For HPLC analyses an Agilent model 1100 HPLC (Agilent Technologies Australia Pty. Ltd., Melbourne, Australia) was used. For analysis of grape and wine volatiles an Agilent gas chromatograph (GC) (6890 series) coupled with an Agilent 5973/5973N/5975A mass selective detector (MS) (Agilent Technologies Australia Pty. Ltd., Melbourne, Australia) was used. For the analysis of monosaccharides as their alditol acetates a Hewlett Packard GC (6890 series) coupled with a Hewlett Packard 5973 MS (Hewlett Packard Australia, North Ryde, NSW, Australia) (GC MS) was used. Methods used Agilent Chemstation software for data analysis (Agilent Technologies Australia Pty. Ltd., Melbourne, Australia). 2.2. Harvesting, grape sampling and preparation for analysis Vitis vinifera L. cv. Cabernet Sauvignon samples were obtained from a commercial vineyard in the Langhorne Creek growing region of South Australia, Australia which has a latitude 35 16 11.56 00 S and a longitude 139 00 14.47 00 E, with an elevation approximately 28 m above sea level. The grapevines were 12 years old, planted 2.5 m (row) 1.8 m (vine); on own roots, spur pruned, with a single-wire trellis and sprawling canopies. Irrigation for the 2009/2010 growing season was 272 mm (0.5 ML/ha). The average yield was 6 kg/vine, and winter pruning weight for 2010 was 0.9 kg/vine, giving a pruning:yield ratio (Ravaz index) of 6.7. Grape samples were obtained at five different stages of ripeness in the 2009/2010 growing season. Harvest dates were on the 16th and 23rd February, and on the 2nd, 10th and 17th of March, and are designated as sequential treatments H1 to H5 in this study. To obtain a representative sample, approximately 3 70 kg of grapes were harvested from the same three rows distributed within the vineyard block. The rows had approximately 100 grapevines/row. For each harvest date, grapevines were selected at staggered points across each row, and completely harvested. In order to minimise within-vineyard differences in total soluble solids (TSS), the grape lots were well mixed, and sub-divided into 3 50 kg lots for small-scale winemaking. Three replicate 100-berry samples of each harvest were processed fresh, and manually separated into juice, skin and seed components, while kept on ice. Skins and seeds were then frozen in liquid nitrogen and stored at 80 C until processed. A further 2 200-berry samples were collected for each winemaking replicate. One was frozen at 20 C, and the second was pressed by hand in a small plastic bag to express the juice. Fresh juice from the 100- and 200-berry samples and was then centrifuged at 1730g for 5 min. Juice from the 100-berry sample was frozen at 20 C until processed. The fresh juice from the 200-berry sample was directly analysed. 2.3. Small-scale winemaking Three 50 kg replicate lots of grapes harvested at different ripeness stages were crushed and de-stemmed with the addition of 50 ppm SO 2. Based on the minimum ph of the first harvest date of 3.2, musts from subsequent dates were adjusted to approximate this ph using tartaric acid. Based on the measured YAN content of each must, total assimilable nitrogen was adjusted to 250 mg/l with diammonium phosphate (DAP). The yeast used was S. cerevisiae PDM (Maurivin, Sydney, Australia) which is analogous to EC1118 (Novo et al., 2009). Yeast was inoculated at 200 mg/kg, and fermentation was carried out on the skins for 7 days in a temperature-controlled room at 15 C, and plunged 20 times twice daily. Thereafter, the ferments were drained and pressed and the combined free-run juice and pressed wine fermented to dryness. Wines were then racked off gross lees, and 60 ppm SO 2 added, and were acid-adjusted to ffi ph 3.5 with tartaric acid. Wines were then cold-stabilised at 0 C for a minimum of 21 days, and thereafter racked off fining lees. The final SO 2 level was adjusted to a total of 80 ppm (free ffi40 ppm), since no malolactic fermentation was performed. Wines were filtered through a 0.8 lm membrane and bottled in 750 ml bottles under screw-cap. 2.4. General analysis of fresh juice and wine Juice TSS as Brix was determined using a digital refractometer. Juice and wine ph and titratable acidity (TA) were determined

1698 K. Bindon et al. / Food Chemistry 138 (2013) 1696 1705 using a ph meter and combination electrode. The TA was determined by titrating with 0.33 M sodium hydroxide solution to a ph end-point of 7. The TA is expressed in g/l of tartaric acid equivalents. Ammonia concentration in fresh juices was determined using the Glutamate Dehydrogenase Enzymatic Bioanalysis UV-method test (Roche, Mannheim, Germany). Juice free a-amino acid nitrogen (FAN) was determined by the o-phtaldehyde/n-acetyl-l-cysteine spectrophotometric assay procedure, which excludes proline (Dukes & Butzke, 1998). Juice yeast-assimilable nitrogen (YAN) was calculated by adding the nitrogen present in ammonium to the FAN concentration. Wine ethanol concentration was determined using a Foss WineScan FT 120 as described by the manufacturer (Foss, Hillerød, Denmark). The concentrations of residual sugar, glycerol, malic acid, succinic acid and acetic acid were measured by high-performance liquid chromatography (HPLC) using a Bio-Rad HPX-87H column as described previously (Varela, Pizarro, & Agosin, 2004). 2.5. Analysis of methoxypyrazines in grape juice and wine A frozen ( 20 C) 200-berry sample was homogenised at 8000 rpm for 2 20 s in a Retsch Grindomix GM200 (Retsch GmbH & Co, Haan, Germany). A 40 g sub-sample of the homogenate was centrifuged to obtain clarified juice. A 20 ml aliquot of either juice or wine was taken and extracted for analysis of methoxypyrazines, to which a 50 ll aliquot of deuterated isobutyl methoxypyrazine (IBMP-d 3 ) (CDN Isotopes, SciVac Pty. Ltd., Hornsby, NSW, Australia) was added in a 50 ml centrifuge tube and gently mixed. The solution was adjusted to a ph < 1 using a sulphuric acid solution and centrifuged at 3200 g for 3 min. The acidified solution was eluted through a pre-conditioned cation-exchange (Varian Bond Elut SCX) solid-phase extraction cartridge (Agilent Technologies Australia Pty. Ltd., Melbourne, Australia). The cartridge was washed with two volumes of Milli-Q water and air-dried under vacuum. The cartridge was eluted with 2% (w/v) NaOH and collected in a 20 ml solid-phase micro-extraction (SPME) vial. Sodium chloride (2 g) was added and the vial capped and analysed using GC/MS with headspace SPME equipped with a DB-5MS (30 m 0.25 mm 0.25 lm) capillary column (Agilent Technologies Australia Pty. Ltd., Melbourne, Australia) and a standard splitsplitless injector. The SPME fibre was a J&W DVB/Carboxen/PDMS (StableFlex, Agilent, Australia) which was preconditioned in the GC injector at 270 C for 60 min. Samples were equilibrated 5 min at 50 C prior to SPME adsorption at 45 min at 50 C, with periodic agitation. The SPME desorption time was 600 s with the inlet temperature at 250 C. The initial oven temperature was 50 C held for 5 min, and the oven ramp program was for 5 C/ min until 100 C, followed by 25 C/min increments until 250 C which was held for 10 min. The column flow rate was constant at 1.5 ml/min, with helium as the carrier gas. IBMP, isopropyl methoxypyrazine (IPMP) and sec-butyl methoxypyrazine (SBMP) were quantified using a multi-level set of matrix-matched standards (Sigma Aldrich, St. Louis, MO, USA; Pyrazine Specialties, Atlanta, GA, USA) and IBMP-d 3 as a surrogate standard. The MS was operated in EI mode with a multiplier gain of 400 mv and in selective ion monitoring (SIM) mode. Target ions and qualifier ions (in parenthesis) were m/z 154 (127/94) for IBMP-d 3, m/z 124 (94/ 151) for IBMP, m/z 152 (124/137) for IPMP, and m/z 138 (124/ 151) for SBMP. Linearity for the concentration range 5 200 ng/l was >0.99, the limit of detection was 5 ng/l, and the coefficient of variation was <10% for all methoxypyrazines assayed at both 5 ng/l and 100 ng/l. The uncertainty of measurement was estimated at 20% [2 the coefficient of variation] or an absolute amount of 2 ng/l, whichever was the greater. 2.6. General analysis of wine volatiles Grape-derived volatile compounds, norisorpenoids (rose oxide, b-damascenone and b-ionone) and monoterpenes (linalool, a-terpineol, nerol and geraniol) were analysed using headspace SPME coupled with GCMS (HS/SPME-GC/MS), with polydeuterated internal standards for stable isotope dilution analysis (SIDA) as described elsewhere (Pedersen, Capone, Skouroumounis, Pollnitz, & Sefton, 2003; Ugliano, Siebert, Mercurio, Capone, & Henschke, 2008). C 6 compounds (hexanol, E-2-hexenal, Z-3-hexen-1-ol and E-2-hexen-1-ol) were also analysed by HS-SPME/GCMS using hexanol-d 13, E-2-hexenal-d 9 and E-2-hexen-1-ol-d 4 as internal standards as previously described (Capone, Black, & Jeffery, 2012). Yeast-derived sulphur-containing volatiles (carbon disulfide, diethyl disulfide, dimethyl sulfide, ethanethiol, ethyl thioacetate, hydrogen sulfide, methanethiol and methyl thioacetate) were determined by using headspace cool-on-column gas chromatography coupled with an atomic emission detector (HS-COC-GC AED), with ethylmethyl sulfide and propyl thioacetate as internal standards (Siebert, Solomon, Pollnitz, & Jeffery, 2010). Yeast volatile fermentation products were analysed using HS-SPME GCMS, with SIDA as described previously (Siebert et al., 2005). Fifteen compounds, including ethyl- and acetate esters and higher alcohols, were quantified. 2.7. Grape and wine colour and tannin analysis For grape tannin extraction, frozen skins and seeds were extracted in 70% v/v acetone in water for 18 h. An aliquot of the extract was dried under a stream of nitrogen and reconstituted in 50% v/v ethanol in water. Grape tannin extracts and wines were analysed by the methyl cellulose precipitable tannin (MCPT) assay using the high-throughput method described in (Mercurio, Dambergs, Herderich, & Smith, 2007). For analysis of total grape anthocyanin, a frozen ( 20 C) 200-berry sample was homogenised at 8000 rpm for 2 20 s in a Retsch Grindomix GM200 (Retsch GmbH & Co., Haan, Germany). A 1 g sub-sample of homogenate was extracted according to (Mercurio et al., 2007) and anthocyanin concentration and content determined using the original method outlined in (Iland, Bruer, Edwards, Weeks, & Wilkes, 2004). Wine total anthocyanin and colour parameters were determined according to (Mercurio et al., 2007). Wine tannin was analysed using phloroglucinolysis and gel permeation chromatography following isolation by solid-phase extraction according to the methods described in (Kassara & Kennedy, 2011). Pre-veraison skin tannin fractions of known mean degree of polymerisation (mdp) (by phloroglucinolysis) were used as standards for calibration. For GPC calibration, a second order polynomial was fitted with the tannin elution time at 50% for each standard. For the MCPT assay, phloroglucinolysis and GPC analysis, ( )-epicatechin (Sigma Aldrich, St. Louis, MO, USA) was used as the quantitative standard. 2.8. Grape and wine polysaccharide analysis Grape juices and wines were concentrated ffi 4 times. Frozen grape juice ( 20 C) was concentrated by lyophilisation in order to prevent degradation by endogenous enzymes, and wine was concentrated under a stream of nitrogen at 30 C. Thereafter, grape and wine soluble polysaccharides were isolated from each experimental replicate through precipitation in ethanol as described previously (Ayestarán, Guadalupe, & León, 2004; Vidal, Williams, Doco, Moutounet, & Pellerin, 2003). Neutral monosaccharides were analysed following hydrolysis in 2 M trifluoroacetic acid for 90 min at ffi100 C. Hydrolysates were concentrated under a stream of nitrogen, and then analysed either as their alditol acetates by GCMS on a BPX70 capillary column (SGE, Ringwood, Victoria,

K. Bindon et al. / Food Chemistry 138 (2013) 1696 1705 1699 Australia) (Lau & Bacic, 1993; Sims & Bacic, 1995) or by HPLC as their 1-phenyl-3-methyl-5-pyrazolone derivatives (Honda et al., 1989). The uronic acid content was determined colourimetrically using D-galacturonic acid (Sigma Aldrich, St. Louis, MO, USA) as the quantitative standard (Filisetti-Cozzi & Carpita, 1991). Polysaccharides were analysed by size exclusion chromatography (SEC) using the method of Vidal et al. (2003) with the following modifications. Dry samples were resuspended in 0.1 M sodium nitrate and 25 ll was injected onto a BioSep Sec 2000 column (300 7.8 mm, Phenomenex, Lane Cove, NSW, Australia) and analysed by HPLC SEC in 0.1 M sodium nitrate at a flow rate of 1 ml/ min. Polysaccharide elution was monitored by refractive index detection over 14 min. Column calibration was carried out with commercial dextrans having molecular weights ranging from 5 to 270 kda (Sigma Aldrich, St. Louis, MO, USA). In order to confirm the elution profile of the HPLC SEC method, polysaccharide isolates were fractionated by semi-preparative SEC on a glass column (Amersham Biosciences, Uppsala, Sweden) packed with Sephacryl S-300 gel (Pharmacia, Uppsala, Sweden) of bed volume 400 cm 3 using 0.1 M sodium nitrate at a flow rate of 1 ml/min. Fractions eluting at successive time points were collected, dialysed (7 kda cut-off) against Milli-Q water for 48 h, and lyophilised. Polysaccharide fractions were identified as mannoproteins (MPs), arabinogalactan proteins (AGPs) or rhamnogalacturonans (RGs) based on their relative monosaccharide composition and by comparison with published data (Vidal et al., 2003). Separation by HPLC SEC was incomplete, but elution ranges for the different polysaccharide classes were as follows: MPs = 5.88 min (185 kda); AGPs ffi6.3 8.0 min (24 48 kda); RGs = 8.1 8.7 min (5.8 kda); low molecular weight polysaccharides ffi10 min (<5.8 kda). 2.9. Statistical analysis Significant differences between harvest dates were determined from triplicate experiments using a one-way analysis of variance (ANOVA), followed by a post hoc Student s t-test. The JMP 5.0.1 statistical software package (SAS, Cary, NC, USA) was used. 3. Results and discussion 3.1. Grape berry sugar accumulation Over the five sampling stages, grape soluble solids, expressed as Brix of sugar increased continually from 20.3 (H1) to 26.0 (H5) (Table 1). The analysis of grape juice sugars as total soluble solids in order to track ripening and estimate harvest time and final wine alcohol is a traditional practice in wine production (Iland et al., 2004). Alternately, grape sugar can be expressed as sugar per berry in order to provide information on sugar loading (Deloire, 2011). This manner of monitoring the ripening process provides information on grapevine physiology, and allows grape berry sugar accumulation kinetics to be inferred. Typically, the ripening process is characterised by an initial active sugar accumulation phase followed by a plateau, notwithstanding that total soluble solids in juice may increase due to a reduction in berry size in the later stages of ripening (Bindon, Dry, & Loveys, 2008; Deloire, 2011). A divergence from this model has been suggested to be due to an imbalance in source:sink relationship within the grapevine (Deloire, 2011). For the site used in this study, the calculated Ravaz index of 6.7 suggests that the grapevines were not unbalanced (Bravdo, Hepner, Loinger, Cohen, & Tabacman, 1985) and it has been noted previously that the site was irrigated. The rate of sugar accumulation per berry (mg/berry/day) (Table 1) was determined using historical data (Bindon, Bacic, & Kennedy 2012) and revealed that sugar accumulation was initially >3 mg/berry/day for H1 and H2, and slowed by H3 and H4. According to the model developed for grapevines (Deloire, 2011) this would indicate a plateau had been reached when sugar loading is <3 mg/berry/day. However, an unexpected increase in sugar loading was observed from H4 to H5, which was noted to follow a rain event and was a period associated with moderate temperatures (Supplementary data S1). It has been noted that water deficit reduces photosynthesis, and as a result impedes sugar loading during the late stages of grape development (Wang, Deloire, Carbonneau, Federspiel, & Lopez, 2003). The later accumulation in sugar may therefore have resulted from a transient enhancement in photosynthesis following the rain event, and we suggest was not the result of an imbalance in carbohydrate partitioning within the grapevines. 3.2. Grape composition Grape juice TA decreased during the ripening period and was associated with an increase in ph, and ph did not exceed 3.5 by the final sampling date (Table 1). As discussed previously, YAN levels were low (maximum 106 mg/l), and decreased as ripening progressed. YAN decreases corresponded to drops in both a-amino nitrogen and ammonia. The grape-derived aroma compound IBMP, typical to Cabernet Sauvignon, decreased between H3 and the final two harvest points H4 and H5. The decrease in grape-derived IBMP during ripening is expected for this variety (Rojou de Boubeé et al., 2000). Grape juice soluble polysaccharides, as the sum of their respective acidic monosaccharides (primarily galacturonans) and neutral monosaccharides did not show a significant trend during ripening. The neutral monosaccharide fraction did not undergo any compositional changes, and were on average 43% galactose and 37% arabinose in molar proportion, with detectable but minor contributions of glucose, rhamnose, xylose, fucose and mannose (data not shown, Supplementary information S2). This analysis indicates that the juice-soluble neutral polysaccharides were primarily arabinogalactan proteins. The phenolic composition of the grape skin and seed components (solids) was determined in terms of total anthocyanin and tannin (Table 1). Anthocyanin, expressed on a berry mass basis (concentration) or per berry (content) increased throughout the ripening period. Seed tannin decreased, and skin tannin increased as ripening progressed, when expressed both as concentration and content. This resulted in a progressive, and significant increase in the skin to seed tannin ratio, which was doubled between H1 and H5. Since the seed tannin concentration was higher than skin tannin concentration, this resulted in a net decrease in total grape tannin during ripening. 3.3. Wine ethanol content and non-volatiles Ethanol concentration increased from 11.8% (v/v) at H1 to 15.5% (v/v) at H5 (Table 2). To test consistency in the conversion of grape sugar to ethanol, yeast fermentation efficiency, defined as the concentration of sugar needed to produce 1% (v/v) of ethanol was calculated. This was found not to be influenced by grape maturity and was on average 16.85 g/l per 1% (v/v) regardless of initial sugar concentration. Practically all sugar was consumed during fermentation, nevertheless residual sugar concentration was slightly higher for wines from H5. Yeast fermentation efficiency remained unaffected by grape maturity indicating that a set fraction of carbon is destined for ethanol production in the range of sugar concentrations evaluated in this study. Wine ph and TA did not show any clear trend with grape maturity due to the acid adjustment employed as part of the vinification process. The H4 treatment had lower TA, most likely due to low tartaric acid concentration, which may reflect a minor loss during

1700 K. Bindon et al. / Food Chemistry 138 (2013) 1696 1705 Table 1 General compositional analysis, non-volatile compounds and isobutyl methoxypyrazine in grape juice and solids from different harvest points in 2010 where H1 was the earliest (16th February) and H5 the latest (17th March) sampling date. H1 H2 H3 H4 H5 Juice composition Soluble sugar [ Brix] f 20.3 ± 0.12 a 22.1 ± 0.12 b 23.1 ± 0.15 c 24.1 ± 0.1 d 26.0 ± 0.0 e Soluble sugar per berry [mg/berry] f 199.5 ± 1.1 a 226.8 ± 1.2 b 229.4 ± 1.5 b 239.6 ± 0.9 c 280.5 ± 1.3 d Rate of sugar accumulation [mg/berry/day] f 4.78 3.79 0.47 1.24 5.68 ph 3.18 ± 0.01 a 3.18 ± 0.02 a 3.33 ± 0.01 b 3.33 ± 0.01 b 3.48 ± 0.01 c Titratable acidity ph 7.0 [g/l] g 8.3 ± 0.2 a 6.9 ± 0.3 b 6.5 ± 0.05 b 5.7 ± 0.21 c 5.3 ± 0.25 c a-amino nitrogen [mg/l] 53 ± 2.0 a 57 ± 4.6 a 31 ± 3.6 bc 36 ± 1.7 b 23 ± 4.0 c Ammonia [mg/l] 57 ± 0.9 a 60 ± 9.4 a 46 ± 3.5 ab 31 ± 1.0 b 33 ± 3.9 b Yeast-assimilable nitrogen [mg/l] 100 ± 2.9 a 106 ± 12.2 a 69 ± 5.9 b 61 ± 1.9 b 51 ± 7.4 b Neutral polysaccharide [mg/l] h 149 ± 11 145 ± 8.4 135 ± 9.0 125 ± 13 150 ± 22 Acidic polysaccharide [mg/l] i 328 ± 31 329 ± 21 247 ± 21 287 ± 49 273 ± 21 Total polysaccharide [mg/l] 477 ± 29 474 ± 24 382 ± 16 412 ± 39 424 ± 39 Isobutyl methoxypyrazine [ng/l] 8.3 ± 0.3 abc 8.7 ± 0.7 ab 9.0 ± 1.2 a 6.3 ± 0.3 c 6.7 ± 0.3 bc Grape solids composition Anthocyanin [mg/g] j 1.37 ± 0.06 a 1.44 ± 0.05 ab 1.61 ± 0.09 bc 1.67 ± 0.02 c 1.87 ± 0.04 d Anthocyanin [mg/berry] j 1.49 ± 0.05 a 1.49 ± 0.03 a 1.73 ± 0.10 b 1.78 ± 0.02 b 1.88 ± 0.05 b Total tannin [mg/g] k 4.15 ± 0.10 a 3.76 ± 0.15 ab 3.63 ± 0.01 bc 3.48 ± 0.16 bc 3.26 ± 0.14 c Total tannin [mg/berry] k 4.04 ± 0.10 a 3.85 ± 0.12 ab 3.61 ± 0.12 ab 3.46 ± 0.16 b 3.51 ± 0.23 b Seed tannin [mg/g] k 2.94 ± 0.09 a 2.52 ± 0.13 b 2.37 ± 0.02 b 2.00 ± 0.10 c 1.80 ± 0.12 c Seed tannin [mg/berry] k 2.87 ± 0.08 a 2.58 ± 0.11 ab 2.36 ± 0.06 bc 1.98 ± 0.11 cd 1.94 ± 0.20 d Skin tannin [mg/g] k 1.20 ± 0.01 a 1.24 ± 0.02 a 1.26 ± 0.03 ab 1.48 ± 0.08 b 1.47 ± 0.13 b Skin tannin [mg/berry] k 1.18 ± 0.02 a 1.27 ± 0.02 ab 1.26 ± 0.06 a 1.47 ± 0.07 bc 1.57 ± 0.12 c Skin tannin:seed tannin ratio l 0.41 ± 0.01 a 0.49 ± 0.02 a 0.53 ± 0.02 a 0.74 ± 0.04 b 0.83 ± 0.11 b EGC m of skin tannin extension subunits (% w/w) 52.8 52.6 52.1 53.1 50.6 Values as mean ± standard error, significant differences between treatments are indicated by different letters in superscript determined by ANOVA, post hoc Student s t-test, n = 15. f Sugars determined as soluble solids % w/w by refractive index, sugar per berry was calculated as Brix 0.59 17 berry weight (g) according to Deloire (2011), sugar/ berry/day estimated from net accumulation between harvest dates. g Titration end-point to ph of 7 expressed as tartaric acid units. h Neutral sugars determined as their 1-phenyl-3-methyl-5-pyrazolone derivatives. i Uronic acids determined as galacturonic acid units. j Anthocyanin determined colorimetrically at 520 nmin 1 M HCl, expressed as malvidin-3-glucosideunits. k Tannin determinedby themethyl-celluloseprecipitation assay expressedin epicatechin units pergberry freshmass orper berry. l Determined as [skin tannin]/[seed tannin]. m EGC, epigallocatechin subunits as a w/w percentage of skin tannin extension subunits determined using phloroglucinolysis. cold-stabilisation. As expected, malic acid concentration decreased with grape maturity (Coombe, 1992), whereas acetic acid, citric acid and tartaric acid showed no significant change. On the other hand, trends in succinic acid and glycerol corresponded to that of ethanol, with higher concentrations in wines from later harvest dates. Both total anthocyanin and tannin in wines increased significantly from H1 to H5, and reflected increases in both bisulphiteresistant (non-bleachable) pigments and wine colour density (Table 2). By comparison with grape composition, anthocyanin and wine colour density more closely reflected changes in grape anthocyanin concentration during ripening, increasing with each successive harvest stage (Tables 2 and 3). Total wine tannin concentration, on the other hand, showed a poor relationship with the trend of decreasing grape tannin (Table 1) from H1 to H5. Interestingly, the proportion of skin tannin (as % of total wine tannin) increased progressively in wines from the earliest to the latest harvest. This is noteworthy, in that differences in extractability between grape-derived skin and seed tannins may exist, which are not reflected in changes in their total extractable levels using aqueous acetone. Cadot, Caille, Samson, Barbeau, and Cheynier (2012) have noted a similar trend for Cabernet Franc, whereby advancing grape maturity was positively correlated with wine tannin and epigallocatechin, and likewise suggested that the extractability of grape tannins is influenced by ripening. The increasing proportion of extracted skin tannin was reflected in a significant increase in wine tannin mdp and molecular mass by phloroglucinolysis in H5 relative to the other treatments. Conversely, the estimation of tannin molecular mass by GPC showed that a minor, but significant decrease occurred from H1 to H5, rather than the expected increase based on the phloroglucinolysis results. A detailed analysis of the GPC elution profile of wine tannins purified from the different harvest treatments was undertaken in order to account for these differences in molecular mass distribution (Fig. 1A). In terms of total polymeric material absorbing at 280 nm isolated from wine, there was a clear increase observed with later harvest. However, it is evident that in the wines from later harvest dates there was an increase in the proportion of later-eluting, apparently smaller tannins which may account for the shift in the elution profile to one of a lower average molecular mass (at 50% elution). The contribution of material absorbing at 520 nm, i.e. derived from the incorporation of anthocyanins or their derivatives into the tannin polymer, increased progressively from H1 to H5 (Table 2). This 520 nm-absorbing material was found to be bound to tannin, and did not co-elute with the later-eluting malvidin-3-glucoside at 16 17.5 min (Fig. 1B). In addition, the contribution of 520 nm-absorbing material was of an apparent lower molecular mass, eluting between 14.5 min and 16 min (Fig. 1B), consistent with the increase in 280 nm absorbance in this elution range for the wines produced from the later harvest dates. It is therefore unlikely that changes in the GPC elution profile between 14.5 and 16 min reflect a true shift in tannin polymer size, and may rather be the result of changes in hydrodynamic volume of the molecule due to anthocyanin incorporation, or possibly other structural modifications. Therefore, the increase in earlier-eluting material from 12 to 13.5 min in H4 and H5 may more accurately reflect the increased proportion of higher mdp tannins derived from grape skin, in

K. Bindon et al. / Food Chemistry 138 (2013) 1696 1705 1701 Table 2 Wine compositional analysis. Ethanol concentration, fermentation efficiency, acidity and non-volatiles in wines produced at different levels of grape ripeness in 2010 where H1 was the earliest (16th February) and H5 the latest (17th March) harvest date. H1 H2 H3 H4 H5 Ethanol [% v/v] 11.77 ± 0.03 e 12.87 ± 0.09 d 13.57 ± 0.09 c 14.2 ± 0.10 b 15.50 ± 0.06 a Fermentation efficiency [g/l per 1% (v/v)] f 17.03 ± 0.27 a 17.00 ± 0.15 a 16.95 ± 0.30 a 16.64 ± 0.30 a 16.63 ± 0.47 a Residual sugar [g/l] g 0.40 ± 0.00 b 0.40 ± 0.06 b 0.53 ± 0.09 b 0.40 ± 0.06 b 0.90 ± 0.00 a Glycerol [g/l] 6.97 ± 0.03 d 7.63 ± 0.07 c 8.13 ± 0.03 b 8.27 ± 0.14 b 9.47 ± 0.03 a ph 3.36 ± 0.02 d 3.50 ± 0.03 b 3.41 ± 0.02 cd 3.62 ± 0.01 a 3.46 ± 0.03 bc Titratable acidity ph 7.0 [g/l] h 6.30 ± 0.12 a 5.70 ± 0.00 bc 6.00 ± 0.15 ab 5.37 ± 0.07 c 6.33 ± 0.13 a Acetic acid [g/l] 0.27 ± 0.09 a 0.20 ± 0.06 a 0.27 ± 0.03 a 0.30 ± 0.01 a 0.37 ± 0.03 a Citric acid [g/l] 0.30 ± 0.00 a 0.33 ± 0.03 a 0.33 ± 0.03 a 0.30 ± 0.00 a 0.30 ± 0.00 a Malic acid [g/l] 3.03 ± 0.03 a 2.73 ± 0.07 b 2.40 ± 0.06 c 2.37 ± 0.09 c 2.27 ± 0.03 c Succinic acid [g/l] 1.33 ± 0.03 d 1.50 ± 0.00 c 1.60 ± 0.00 bc 1.73 ± 0.09 ab 1.87 ± 0.03 a Tartaric acid [g/l] 1.90 ± 0.00 abc 1.70 ± 0.00 c 2.07 ± 0.09 a 1.80 ± 0.12 bc 2.03 ± 0.09 ab Neutral polysaccharide [mg/l] i 84.7 ± 6.2 a 85.0 ± 3.8 a 91.5 ± 14.8 ab 121.3 ± 18.0 b 114.6 ± 4.2 ab Acidic polysaccharide [mg/l] i 137.7 ± 42.6 a 94.5 ± 5.8 ab 77.5 ± 12.8 abc 31.6 ± 5.9 bc 25.8 ± 7.9 c Total polyaccharide [mg/l] i 222.4 ± 37.1 a 179.4 ± 3.9 ab 169.0 ± 2.3 ab 152.9 ± 22.8 b 140.4 ± 4.0 b Anthocyanin [mg/l] j 411 ± 22 a 529 ± 11 b 592 ± 23 b 657 ± 1.78 c 728 ± 30 d SO 2 -resistant pigments [a.u.] k 1.04 ± 0.08 a 1.30 ± 0.05 ab 1.82 ± 0.30 c 1.74 ± 0.10 bc 2.59 ± 0.05 d Wine colour density [a.u.] k 7.64 ± 0.17 a 9.95 ± 0.26 b 11.34 ± 0.31 c 11.26 ± 0.17 c 14.51 ± 0.21 d Total tannin [mg/l] l 731 ± 61 ab 750 ± 38 ab 638 ± 8 a 786 ± 46 b 1088 ± 43 c Skin tannin [%] m 56.8 ± 0.52 a 58.8 ± 0.57 a 63.5 ± 0.82 b 64.1 ± 0.19 b 71.8 ± 0.89 c Tannin mdp m 7.15 ± 0.10 ab 7.04 ± 019 a 7.72 ± 0.06 c 7.55 ± 0.26 bc 8.39 ± 0.07 d Tannin molecular mass (g/mol by subunit) m 2148 ± 30 a 2114 ± 58 a 2317 ± 18 b 2264 ± 80 ab 2519 ± 23 c Tannin molecular mass (g/mol by GPC) n 3315 ± 75 a 3073 ± 122 ab 2987 ± 6 ab 2914 ± 207 b 2934 ± 30 b Tannin 520/280 ratio (%) n 6.77 ± 0.24 a 7.96 ± 0.15 b 8.35 ± 0.12 bc 8.73 ± 0.22 c 9.61 ± 0.13 d Values as mean ± standard error, significant differences between treatments are indicated by different letters in superscript determined by ANOVA, post hoc Student s t-test, n = 15. f Fermentation efficiency is the concentration of sugar needed to produce 1% (v/v) of ethanol. g Residual sugar determined as glucose and fructose by HPLC. h Titration end-point to ph of 7 expressed as tartaric acid units. i Polysaccharide calculated as the sum of neutral sugars from their corresponding alditol acetates determined by GCMS and uronic acids determined colorimetrically as galacturonic acid units. j Anthocyanin as malvidin-3-glucoside units determined colorimetrically. k a.u., absorbance units, wine colour density corrected for SO 2 in the presence of acetaldehyde. l Tannin determined by the methyl-cellulose precipitation assay expressed in epicatechin units. m Tannin compositional information determined by phloroglucinolysis: skintannin calculated fromtheratio of epigallocatechin in extension subunits in grape skin tannin (Table 1)as aproportion of wine tannin extension subunits, mdp, mean degree of polymerisation. n Determined by gel permeation chromatography (GPC), molecular mass average at 50% elution, absorbance ratio calculated from 280 and 520 nm peak areas of the tannin isolate. Table 3 Compositional analysis of neutral monosaccharide composition as mol% in polysaccharides isolated from wines produced at different levels of grape ripeness in 2010 where H1 was the earliest (16th February) and H5 the latest (17th March) harvest date. H1 H2 H3 H4 H5 Rhamnose 9.98 ± 0.27 a 9.98 ± 0.46 a 9.90 ± 0.51 a 5.24 ± 0.63 b 4.23 ± 1.39 b Fucose 0.62 ± 0.04 ab 0.69 ± 0.04 a 0.63 ± 0.04 a 0.42 ± 0.10 bc 0.27 ± 0.07 c Arabinose 24.34 ± 0.33 a 24.72 ± 0.78 a 23.73 ± 0.72 a 18.90 ± 1.08 b 17.84 ± 1.07 b Xylose 0.91 ± 0.05 a 0.88 ± 0.08 ab 0.88 ± 0.06 ab 0.68 ± 0.10 bc 0.60 ± 0.03 c Mannose 28.94 ± 0.47 a 28.84 ± 0.90 bc 31.14 ± 0.97 ab 41.18 ± 1.32 c 43.93 ± 2.32 a Galactose 27.22 ± 0.29 abc 27.03 ± 0.24 bc 26.44 ± 0.64 c 28.46 ± 1.14 ab 29.08 ± 0.29 a Glucose 8.09 ± 0.11 a 7.86 ± 0.51 a 7.27 ± 0.25 a 5.11 ± 0.37 b 4.04 ± 0.03 c Values as mean ± standard error, significant differences between treatments are indicated by different letters in superscript determined by ANOVA, post hoc Student s t-test, n = 15. accordance with the observations using phloroglucinolysis. This finding confirms observations of Cadot et al. (2012) who noted a decrease in thiolysis yield in tannins from Cabernet Franc wines produced from grapes of more advanced ripeness. They suggested that this could reflect enhanced formation of thiolysis-resistant anthocyanin-tannin adducts in wines made from riper grapes, associated with increasing levels of anthocyanin. Total wine-soluble polysaccharides decreased by harvest date only when H1 was compared with the other treatments. However, there were clear compositional changes, notably a loss in the acidic polysaccharide fraction and increases in the neutral polysaccharide fraction in wines from later harvest dates (Table 2). Analysis of neutral monosaccharide composition revealed that trends from H1 to H5 were a proportional loss in arabinose, with significant increases in mannose (by mol%) (Table 3). Trends in the proportional composition of the minor sugars, rhamnose, fucose, xylose and glucose also decreased in wines from later harvest dates, but galactose stayed relatively constant. Since the observed trends were consistent with a loss in grape-derived galacturonans through vinification, and increases in yeast-derived polysaccharides (i.e., MPs) with later harvest date, this was further assessed by size exclusion chromatography (Fig. 2). Wines from the earlier harvest dates had a higher proportion of the lower molecular mass polysaccharides, which correspond to the elution of RGs, consistent with the observation of a higher acidic polysaccharide component and rhamnose. Wines produced from advanced grape maturity stages showed a

1702 K. Bindon et al. / Food Chemistry 138 (2013) 1696 1705 Fig. 2. Elution profile determined by size exclusion chromatography of mannoproteins (MP), arabinogalactan proteins (AGP) and rhamnogalacturonan (RG) polysaccharides from wines produced from early (H1, 12% v/v ethanol) and late (H5, 15.5% v/v ethanol) harvested grapes. Fig. 1. Gel permeation chromatography analysis of (A) elution profiles at 280 nm of wine tannins from different harvest dates where H1, earliest harvest, 12% v/v ethanol and H5, latest harvest, 15.5% v/v ethanol and (B) contribution of wine tannin peak areas at 280 nm and 520 nm by comparison with monomeric malvidin- 3-glucoside (malvidin-3-glucoside had the same elution profile at both 280 nm and 520 nm). proportional loss of late-eluting, low molecular mass polysaccharide (galacturonans), and increased higher high molecular mass material, consistent with the enrichment of the MP fraction. The increased concentration of MPs may be due to enhanced yeast metabolism in higher-ethanol wines, and may also have been associated with increased secretion of endo-polygalacturonase (PGU1), an enzyme which can be expressed in the S. cerevisiae PDM strain used (EC1118) (Novo et al., 2009; Radoi, Kishida, & Kawasaki 2005). Since polysaccharide quantity and composition did not differ between harvest dates in juices, the loss in acidic galacturonans from later-harvest wines may potentially be associated with an increased PGU1 activity. A further important consideration is the role of inter-molecular interactions between wine tannins and polysaccharides. A pertinent study by Riou, Vernhet, Doco, and Moutounet (2002) observed the behaviour of wine polysaccharide classes in either enhancing or restricting the aggregation of seed tannins in model solution. High molecular weight AGP and MPs were shown to reduce the particle size of tannin aggregates, maintaining tannin solubility in colloid complexes. Conversely, an RG dimer (rhamnogalacturonan II) enhanced tannin aggregation, thereby potentially enhancing tannin precipitation. In the current study, the higher tannin concentration in later-harvest wines could lead to an increased potential for aggregate formation, but a higher MP concentration could favour the maintenance of tannin in colloid complexes. Since juice-soluble polysaccharides (Table 1) were both higher in the concentration and proportion of acidic polysaccharide residues than those in wine (Table 2), the potential exists that some loss of tannin could have occurred via aggregation and precipitation as co-aggregates with soluble galacturonan-rich polysaccharide. While this would need to be verified experimentally for wine tannin, this may offer a reasonable explanation for the lack of RGs in later-harvest wines. An additional, anecdotal observation in the winemaking process was that treatments H4 and H5 had reduced filterability. It has been observed that the presence of high molecular mass MPs reduces wine permeation flux (Vernhet, Pellerin, Belleville, Planque, & Moutounet, 1999), and the potential exists that some grape-derived polysaccharides present at the end of ferment were removed from the finished wines due to fouling of the membrane with MPs. Since the presence of a higher pectic polysaccharide to MP ratio has also been noted to enhance wine permeation flux (Vernhet et al., 1999), it is possible that this contributed to polysaccharide recovery differences between H1 to H3 and the wines produced from the later-harvested grape sources. 3.4. Wine volatiles All volatile compounds detected in more than one wine are presented in Table 4. b-damascenone and b-ionone were the only C 13 -norisoprenoids detected in wines, b-damascenone showed no significant differences by harvest date, whereas b-ionone was only detected in H1 wines at 0.5 lg/l (data not shown). The monoterpene linalool showed no significant differences between harvest treatments, and in Cabernet Sauvignon this compound has been found to be synthesised de novo by yeast metabolism, independently of grape-derived precursors available during fermentation (Keyzers & Boss, 2010). The concentration of IBMP, the only pyrazine detected in wines, decreased significantly by harvest date from H1 to H5. This decrease was a more significant trend in wine than that observed for IBMP extracted from grapes (Table 1). The extracted levels of IBMP in wines were higher than the detectable levels in grapes. This may be consistent with an incomplete extraction from the grape solids in our analysis of IBMP in grape juice. To better approximate wine IBMP using grape measurements, Ryona,

K. Bindon et al. / Food Chemistry 138 (2013) 1696 1705 1703 Table 4 Analysis of volatile composition of wines produced at different levels of grape ripeness in 2010 where H1 was the earliest (16th February, 12% v/v ethanol) and H5 the latest (17th March, 15.5% v/v ethanol) harvest date. H1 H2 H3 H4 H5 Isoprenoids and pyrazines b-damascenone [mg/l] 10.3 ± 0.35 b 10.5 ± 0.00 b 11.7 ± 0.17 a 9.8 ± 0.17 b 11.8 ± 0.58 a Linalool [mg/l] 3.0 ± 0.00 a 2.7 ± 0.35 a 2.7 ± 0.35 a 2.3 ± 0.35 a 3.0 ± 0.00 a Isobutyl methoxypyrazine [ng/l] 14.3 ± 0.4 a 11.7 ± 0.4 b 8.7 ± 0.4 c 8.0 ± 0.6 cd 7.0 ± 0.6 d Thiols Dimethyl sulfide [mg/l] 2.0 ± 0.00 d 3.0 ± 0.00 c 3.3 ± 0.35 c 4.0 ± 0.00 b 5.3 ± 0.35 a Methyl thioacetate [mg/l] nd nd nd 1.7 ± 1.67 b 10.7 ± 0.35 a Esters Ethyl acetate [mg/l] 28.9 ± 1.44 d 36.1 ± 1.91 c 44.1 ± 3.23 b 45.2 ± 0.23 b 55 ± 0.12 a Hexyl acetate [mg/l] 77 ± 6.35 a 63 ± 4.62 b 64 ± 2.89 ab 67 ± 1.73 ab 57 ± 2.89 b 2-Methylbutyl acetate [mg/l] 90 ± 9.2 d 119 ± 10 cd 138 ± 13 bc 175 ± 13 b 235 ± 19 a 3-Methylbutyl acetate [mg/l] 0.9 ± 0.12 d 1.0 ± 0.12 cd 1.3 ± 0.12 b 1.3 ± 0.00 bc 1.8 ± 0.06 a 2-Phenylethyl acetate [mg/l] 70 ± 11 b 68 ± 1.73 b 114 ± 35 ab 85 ± 13 ab 165 ± 43 a Ethyl propanoate [mg/l] 248 ± 14 c 284 ± 13 c 335 ± 8.7 b 357 ± 0.6 b 419 ± 19 a Ethyl 2-methylpropanoate [mg/l] 47.3 ± 4.21 bc 41.3 ± 0.64 c 56.3 ± 5.20 b 42.2 ± 0.75 c 68.2 ± 1.79 a Ethyl butanoate [mg/l] 301 ± 18 c 326 ± 23 c 386 ± 5.77 b 348 ± 2.89 bc 474 ± 19 a Ethyl 3-methylbutanoate [mg/l] 14.4 ± 0.75 c 13.8 ± 0.06 c 21.8 ± 2.71 b 14.4 ± 0.40 c 27.3 ± 1.15 a Ethyl hexanoate [mg/l] 1.0 ± 0.12 d 1.0 ± 0.06 cd 1.3 ± 0.06 ab 1.2 ± 0.00 bc 1.4 ± 0.00 a Total ethyl esters [mg/l] f 1.6 ± 0.12 d 1.7 ± 0.06 cd 2.1 ± 0.06 b 1.9 ± 0.00b c 2.4 ± 0.06 a Total acetate esters [mg/l] f 1.1 ± 0.12 c 1.3 ± 0.12 bc 1.6 ± 0.12 b 1.7 ± 0.00 b 2.2 ± 0.12 a Total esters [mg/l] f 2.7 ± 0.23 d 3.0 ± 0.17 cd 3.7 ± 0.17 b 3.6 ± 0.00b c 4.7 ± 0.12 a Alcohols Hexanol [mg/l] 5.6 ± 0.12 a 4.7 ± 0.17 b 4.4 ± 0.12 b 4.4 ± 0.00 b 3.0 ± 0.17 c Z-3-Hexen-1-ol [mg/l] 27.3 ± 0.87 a 18.7 ± 0.35 b 13.0 ± 0.75 c 9.3 ± 0.23 d 6.1 ± 0.12 e Butanol [mg/l] 490 ± 17 d 648 ± 26 c 778 ± 23 b 843 ± 13 b 1275 ± 21 a 2-Methylpropanol [mg/l] 21.7 ± 0.69 b 24.1 ± 1.85 ab 23.6 ± 0.52 ab 24.5 ± 0.17 ab 25.7 ± 0.98 a 2-Methylbutanol [mg/l] 68 ± 0.58 b 79 ± 0.58 ab 77 ± 8.08 ab 89 ± 4.62 a 88 ± 4.04 a 3-Methylbutanol [mg/l] 179 ± 2.89 a 185 ± 1.15 a 187 ± 22 a 205 ± 9.2 a 205 ± 11 a Total higher alcohols [mg/l] 269 ± 2.31 a 288 ± 2.31 a 288 ± 30 a 320 ± 14 a 320 ± 16 a Values as mean ± standard error, significant differences between treatments are indicated by different letters in superscript determined by ANOVA, post hoc Student s t-test, n = 15; nd, not detected. f Values exclude ethyl acetate. Pan, and Sacks (2009) have observed that IBMP extraction into grape juice from solids may be facilitated by the application of higher extraction temperature and extended skin contact time, giving a better approximation of the levels expected in wine. Hexanol and (Z)-3-hexen-1-ol, and hexyl acetate, all C 6 compounds originating from grape-derived precursors, also showed lower concentrations in wines with advancing harvest date. The C 6 alcohols are thought to be derived from C 18 fatty acids via the lipoxygenase pathway and alcohol dehydrogenase, either in situ during grape ripening, or under the oxidative conditions present when fruit is crushed (Dennis et al., 2012; Kalua & Boss, 2009). Hexyl acetate, on the other hand, has been found to be at low levels, or absent in grapes, and it appears that alcohol acetyl transferase activity within Cabernet Sauvignon favours the formation of (Z)-3-hexenyl acetate as the principal biosynthetic pathway (Kalua & Boss, 2009). However, (Z)-3-hexenyl acetate was absent in the wines of the current study. Recently, a direct relationship between hexanol concentration in ferment, and hexyl acetate production in wine via the action of yeast alcohol acetyl transferase has been demonstrated (Dennis et al., 2012). This most likely accounts for the observed decline in hexanol and hexyl acetate in wines produced from grapes at later maturity stages. Apart from hexyl acetate, the total concentration of all esters, excluding ethyl acetate, increased in wines from H1 to H5. This increase was caused by changes in both ethyl esters and acetate esters (Table 4). Formation of esters depends on both the concentration of their precursors and yeast metabolism (Saerens et al., 2008; Verstrepen et al., 2003). Wines produced from grapes of advancing maturity resulted in higher concentrations of most esters, namely ethyl acetate, ethyl propanoate, ethyl butanoate, 2-methylbutyl acetate and 3-methylbutyl acetate. Although ethyl 2-methylpropanoate, ethyl 3-methylbutanoate, ethyl hexanoate and 2-phenylethyl acetate did not show a clear trend with grape maturity, they exhibited higher concentrations at H5 than at H1. The concentration of the yeast-derived low molecular weight volatiles dimethyl sulfide and methyl thioacetate in wines were very low, but also increased from H1 to H5, consistent with a higher general yeast metabolism in musts of higher starting sugar content. Butanol was the only higher alcohol which showed a linear increase in concentration in wines with advancing harvest date. 2- Methylpropanol, 2-methylbutanol and 3-methylbutanol showed increased concentration from H1 to H2 and no significant differences thereafter (Table 4). 3.5. Response of yeast metabolism to increases in must sugar content Increasing must sugar concentration as a result of advanced grape maturity, resulted in significant effects on yeast metabolism. As the production of ethanol by yeast increased, there were concomitant increases in most yeast-derived metabolites, including low molecular weight sulphur-containing volatiles, esters and higher alcohols. We have also previously discussed the increased secretion of yeast-derived mannoproteins in higher-ethanol wines, with the corresponding loss of polysaccharides rich in uronic acids and rhamnose, potentially the result of enhanced yeast polygalacturonase activity. However, it is noteworthy that the acids acetic acid, citric acid and malic acid showed a different profile. Since the carbon fraction from sugar ending up as acetic acid or citric acid is very small (less than 0.15%), the relatively low resolution of the quantification method at low concentrations could explain why these acids were