Abstract
Wine is a highly complex alcoholic beverage and its flavour and quality depends on a multitude of factors including soil composition, climate, vintage, grapevine, viticulture, and oenology practices. Mt. Difficulty Wines in Central Otago, NZ, has 3 vineyard sites (Target Gully, TG; Long Gully, LG; and Pipeclay Terrace, PT) that tend to produce single vineyard Pinot Noir wines with higher site-typicity characteristics (typical vineyard site characteristics) than wines produces from other closely located vineyard sites. Anecdotal statements from the winemaker describe these wines as complex in flavour (taste and aroma) which differentiates them from the remaining wines produced by the company, thereby justifying a higher retail price. Despite the wide knowledge of flavour compounds in wine (volatile as well as non-volatile), detailed knowledge on the specific compounds contributing to site-typicity and clonal characters is very limited.
In order to gain a better understanding of why certain sites produce higher typicity wine than others and to identify chemical cues that may be responsible for these typicity characters, the current study investigated the volatile and non-volatile profiles of commercially produced Pinot Noir grape juices and wines (pre-barrelled and barrelled wines) from 6 Mt. Difficulty vineyard sites, including 3 high typicity, 3 low typicity expressing sites and 8 different grapevine clone mixtures. Untargeted fingerprinting methodologies were applied to analyse samples over two consecutive vintages (2012 and 2013). Up-to-date and highly advanced multi-way and multivariate data analysis methodologies were combined with the analytical techniques.
Volatile analysis was carried out using two different approaches, a) solid-phase extraction - gas chromatography - mass spectrometry (SPE-GC-MS) in combination with principal component analysis (PCA) and parallel factor analysis 2 (PARAFAC2) and b) proton-transfer reaction - mass spectrometry (PTR-MS) with prior headspace dilution in combination with analysis of variance (ANOVA) and PCA.
To address the non-volatile compounds, fingerprints were obtained by two approaches a) proton nuclear magnetic resonance spectroscopy (1H-NMR) in combination with a number of pre-processing techniques and PCA and b) front-face fluorescence spectroscopy in combination with the multi-way deconvolution technique PARAFAC.
PCA and PARAFAC2 of the data obtained from the volatile analysis by SPE-GC-MS showed that the discrimination of grape juices and wines was predominantly dependent on vintage (discriminating wines of vintage 2012 from wines of 2013) and barrel maturation (discriminating pre-barrelled wines from barrelled wines). A trend of discrimination based on site-typicity and clone mixture was found. Grape juices and wines from high typicity vineyard sites (TG, LG and PT) tended to cluster together with samples from Black Rabbit (BR, expected low typicity site) and were separated from the remaining low typicity samples. The discrimination was consistently observed across the two vintages, but the composition of volatile compounds (ratios) responsible for the clustering differed between vintages. The effect of discrimination based on the clone mixture was observed in grape juice and wine samples but was less expressive compared to the site-typicity effect. An additional finding was that grape juice samples from vineyard site PT and clone 115 had relatively higher concentrations of benzaldehyde (tentatively identified) in both vintages leading to the conclusion that benzaldehyde was a marker compound for grape juices from site PT and clone 115.
In agreement with the GC-MS results, the PTR-MS data, analysed by ANOVA and PCA, generally showed clustering of samples from high typicity sites with samples from BR and separation from remaining samples of the low typicity sites. A separation based on clone mixtures was less expressive than the discrimination based on vineyard site. The marker compound benzaldehyde for grape juices from site PT and clone 115 was also apparent from the PTR-MS data. The predominant effect that discriminated samples was dependent on the winemaking process (e.g. maceration temperatures and Brix of samples).
PCA of the 1H-NMR data predominantly discriminated samples on the basis of the winemaking process conditions. Further discriminating effects were due to differences in the pH of the samples that resulted in signal shift differences between the sample spectra. From the results obtained, it could not be definitively determined whether the shift differences were solely representative for the samples (pH related) or were additionally impacted by artefacts introduced by the alignment of the spectra. The results also showed signal shape differences between samples which were likely to be due to data acquisition discrepancies. Trends of separation of samples on the basis of vineyard site were observed, whereby samples of different clones from a single vineyard site clustered together. Discrimination of the juices or the wines based on site-typicity and / or clone mixture was not seen in the NMR data. It was apparent that the approach used for the NMR analysis (using pure wine samples, mixed with D2O and internal standard, no pH adjustment, water and ethanol suppression, covering the spectrum from -1 to 11 ppm, alignment using icoShift, PCA of large spectral regions) was not appropriate for obtaining comparable data across samples that would facilitate them being discriminated on the basis of site-typicity and / or clone. Using NMR as a profiling methodology requires compromises to be made in order to balance sample through-put (number of samples able to be analysed) and inherent sample properties against sensitivity and comparability of the data. For example, a pH adjustment of samples is highly time-consuming and leads to loss of inherent sample properties but may lead to better alignment results and hence better comparability of the data. Discrimination between juices or wine samples through NMR data may have been possible if further optimisation of sample preparation, instrumental analysis and data pre-processing had been carried out.
PARAFAC analysis of the front-face fluorescence spectroscopy data found that consistent clustering of samples from clone 115, in grape juice and wine samples in vintage 2012, was likely due to contributions from caffeic acid. A trend of clustering based on site-typicity was evident in the pre-barrelled wines in vintage 2012, but not from the juices or barrelled wines of the same vintage. Samples from BR also clustered with the high typicity PB wines, similar to the findings made with volatile measurements. The data from vintage 2013 was not obtained due to instrumental issues. To confirm the observed trends in the fluorescence data, analysis of a further vintage is recommended.
In summary, the methodologies GC-MS, PTR-MS and front-face fluorescence spectroscopy successfully fingerprinted the grape juice and wine samples in the current study. Trends of discrimination based on site-typicity and clone mixture were obtained and a few marker compounds identified. All three of these methods showed their applicability for the discrimination of juice and wine samples on the basis of very subtle differences, and suggest the usefulness of the methodologies for future wine research. GC-MS may be of less applicability for untargeted metabolomics approaches, as the instrumental runs are relatively long and data analysis is cumbersome. The NMR methodology in contrast, failed to discriminate samples on the basis of site-typicity and clone and remains to be optimised. However, whether NMR is a feasible technology for the untargeted screening of grape juice and wine samples without sample preparation (pH adjustment and / or sample extraction) is debatable.
The results obtained in this study are relevant to future research for the investigation of site-typicity and clonal characteristics. It has been shown that volatile as well as non-volatile compounds contribute to the character differences between grape juices and wines and hence the study emphasises the need to investigate both of these groups of compounds.
In order to fully elucidate the role of volatile and non-volatile compounds in the character of site-typicity and clone further research is required. Such research may include experiments that vinify wines under uniform laboratory conditions (e.g. identical yeast strain and oxygen levels), whereby the winemaking related variables can be minimised. If in the future marker compounds can be routinely identified in high value wines, it may help the wine industry to decide whether the grape juice or wine expresses typicity characters at earlier stages than currently possible by tasting the end-product. Furthermore to improve the understanding about the term site-typicity, as it is described by the winemaker, sensory evaluation of the samples by consumers as well as trained sensory panels would be of value.