Contributions of Hanseniaspora species to Pinot Noir microbial terroir in Oregon’s Willamette Valley wine region

ABSTRACT

The apiculate yeast genus Hanseniaspora has appeared frequently in enological research for more than 100 years, mostly focused upon the species H. uvarum due to its notable capacity to cause spoilage. Recently, there has been increased research into the potential benefits of other Hanseniaspora species, such as H. vineae, in producing more complex wines. Furthermore, large-scale DNA sequencing-based (metabarcoding) vineyard ecology studies have suggested that Hanseniaspora species may not be evenly distributed. To address potential differences across geographical areas in Oregon, we sampled extensively from 12 vineyards within the Willamette Valley American Viticultural Area (AVA), across 2 sub-AVAs (Eola-Amity Hills and Yamhill-Carlton). Metabarcoding was then used to assess the contribution of Hanseniaspora to the grape berry fungal community and the impact of wine processing on diversity. While 6 of the 23 recognized Hanseniaspora species were present on Pinot Noir grapes in the Willamette Valley AVA, differences between vineyards were driven by the abundance of H. uvarum. Significant positive correlations between the amount of H. uvarum present in must and at cold soak, and then cold soak to early ferment were observed. While intuitive, it is worth noting that no prior studies have observed this across such a large number of grape samples from different vineyards. Our results provide clear evidence that the abundance of H. uvarum on grapes may be an important predictor of potential impacts on wine quality, particularly if performing cold soak, which acts as an enrichment step.

IMPORTANCE

Hanseniaspora yeasts are frequently found in uninoculated wine fermentations, and depending upon the species present, their contributions to the wine may be positive or negative. We found that in Oregon’s Willamette Valley, the most common species of Hanseniaspora in Pinot Noir vineyards was the known spoilage organism, H. uvarum. This species was one of the strongest contributors to differences in fungal communities between different vineyards and was enriched during typical Pinot Noir processing. These results support Hanseniaspora as an integral and functional component of vineyard “microbial terroir” within Oregon.

INTRODUCTION

Fermented foods and beverages have become attractive systems to study microbial community dynamics and functionality (1). In addition to their relative simplicity and tractability, foods such as bread and cheese are produced around the world providing researchers with opportunities to observe microbial community structures resulting from divergent starting materials and spanning myriad variations in processing (2–5). The grape-to-wine production system is also globally replicated and is characterized by minimal processing between harvest of raw materials and initiation of fermentation. Consequently, the qualities of wine are thought to directly reflect the impact of the grape-growing environment upon grape composition, a concept known as “terroir” (6). While a microbial contribution to terroir had been previously hypothesized (7), studies during the past decade utilized high-throughput DNA sequencing (metabarcoding) to demonstrate that grape bacterial (microbiome) and fungal (mycobiome) communities are complex and vary in composition according to factors, such as vineyard climate, grape varietal, and geography [e.g., see references ( 8, 9)]. The concept of “microbial terroir” (2, 4) in wine was rapidly adopted, supported by correlations between grape must microbial variation and wine composition (10). However, some researchers have been more circumspect on whether biogeographical variation in microbial community structure translates to a functional contribution to terroir (11), and the strongest direct evidence for microbial terroir concerns regional intra-specific variation among the dominant wine fermentation yeast Saccharomyces cerevisiae (12).

In the grape-wine system, there are other fungi known to have impacts on quality. For example, Botrytis cinerea and Erysiphe necator are filamentous fungi commonly found on grapes which can cause off flavors in wine and degradation of desired polyphenols (13). As for yeast, one of the most frequently observed and studied non-Saccharomyces genera is Hanseniaspora. This genus comprises 23 species, 11 of which have been isolated from grapes and wine (14, 15). Hanseniaspora yeasts are typically only present during initial stages of wine fermentation, as they do not tolerate ethanol to the same degree as S. cerevisiae (16), but some species are nevertheless well known for their impacts upon wine quality. For example, H. uvarum has historically been associated with negative aromas such as “vinegar” (acetic acid) and “nail polish remover” (high concentrations of ethyl acetate) (17), while H. vineae has been shown to impart positive aromas such as “fruitiness” (low concentrations of ethyl acetate) and “banana” or “pear” (3-methylbutyl acetate) (18, 19). Understanding which Hanseniaspora species are present in grape must may inform impacts on wine quality that occur during the early stages of fermentation.

In addition to the impact of vineyard environmental factors on grape microbial terroir, winemaking practices can impact the mycobiome and microbiome of a wine fermentation. For example, pre-fermentation addition of the preservative sulfur dioxide (20) selects against H. uvarum but not H. osmophila, while cold-soaking of crushed grape must prior to fermentation (21, 22) enriches for H. uvarum. The cold-soaking process is traditionally used in red wines to increase the concentration of anthocyanins by keeping crushed grapes chilled for a period of time prior to fermentation (23) . Pinot noir producers commonly use this process due to the naturally light color of the wine; however, the cold environment selects for cryotolerant organisms such as H. uvarum and can increase the concentration of unwanted volatile compounds in the wine (17, 22). Additionally, inoculation of a wine with S. cerevisiae can induce competition with other microbes due to its dominance in fermentation (24), especially Hanseniaspora due to its low ethanol tolerance compared to S. cerevisiae (15). However, when the cold-soak process is used, inoculation with S. cerevisiae typically occurs post-chilling (25), which may allow the more cryotolerant species to grow during the cold-soak process.

Prior studies that have observed geographical differences in grape/must fungal communities usually identify one or more species of Hanseniaspora within their analysis (8, 26), consistent with their frequent isolation in classical studies of uninoculated wine fermentation ecology (27–30). Despite the landmark “microbial terroir” study (8) noting one Hanseniaspora species (H. guilliermondii) as a variable component of Californian (CA) vineyards, it is surprising that most of the larger-scale studies of vineyard grape fungal communities (31, 32) do not explicitly report biogeographical differences in Hanseniaspora species occurrence or abundance. As a consequence, even though Hanseniaspora is a prevalent genus associated with grapes in the vineyard and there is clear evidence of differential species impact upon wine quality, it is unclear whether the biogeographical variation of Hanseniaspora species presence/abundance exists and has the potential to have a functional impact on wine quality.

In this study, we sought to address this gap by profiling the mycobiome of Pinot Noir grape must in Oregon’s Willamette Valley American Viticultural Area (AVA). We used these data to determine which Hanseniaspora species were prevalent and abundant, and whether biogeographical differences in the must mycobiome were contributed to by this genus. In addition, laboratory-scale cold soak and micro-ferments were analyzed to determine if biogeographical differences in Hanseniaspora species abundance in grape must have the potential to lead to differential impacts on wine fermentation outcomes. Our results extend the understanding of grape berry mycobiome to include Oregon and illustrate the potential for biogeographical variation among non-Saccharomyces yeast populations to have functional impacts.

MATERIALS AND METHODS

Chemicals and microbiological media

All chemicals were from Sigma-Aldrich (details) unless otherwise specified. Yeast cultures were grown and enumerated on yeast peptone dextrose (YPD) agar [10 g/L yeast extract (Bio Basic)], 20 g/L peptone, 20 g/L D-(+)-glucose (US Biological Life Sciences), and 15 g/L agar (RPI) or in YPD broth. Grape must samples were enumerated through serial dilutions on YPD + biphenyl (10 mg/L).

Vineyard sampling

Twelve organic or biodynamically managed Pinot Noir vineyards within the Willamette Valley AVA in total were selected as sampling sites to minimize impacts of agrochemicals upon grape berry fungal communities. Six vineyards were clustered within the Eola-Amity Hills sub-AVA, while the other six were from the Yamhill-Carlton sub-AVA (Fig. 1). Distances between vineyards and sub-AVA center points, as well as distances between the two sub-AVA center points, were calculated using GPS coordinates for centroid formation (see Statistical analyses). Distances between coordinates and centroids were determined using Google Maps to calculate the average distance in each sub-AVA and between sub-AVAs.

Fig 1.

Geography of sampling within Willamette Valley AVA. Map of the state of Oregon with vineyard geolocations randomly assigned letter codes and colored by sub-AVA. Average distance of each vineyard within sub-AVA to its center point is noted along with the distance between sub-AVA center points. Vineyard dots colored by sub-AVA. The underlying geodata for the maps come from spData, which is made available under a CC1.0 license (see https://cran.r-project.org/web/packages/spData/index.html).

Within each vineyard site, nine sample locations were randomly selected from a 4 × 5 grid overlayed upon a satellite map. Five to six non-damaged bunches from two to three vines at each location were aseptically collected into 2,800 mL Whirl-Pak bags (NASCO Sampling) and transferred to Oregon State University for processing. Samples were collected when fruit was at commercial ripeness according to each vineyard’s specifications over a 3-week period in September and October of 2022. Ripeness levels of samples varied from 17.6 to 23.9 °Brix with a median value of 22 °Brix. Two samples each from vineyards C and K were excluded from the study due to sampling from vines that were not Pinot Noir, reducing the total number of discrete sampling locations (and therefore samples) to 104.

Grape berry processing

All vineyard samples were stored at 4°C and processed within 24 hours. Grape berries were aseptically hand destemmed and randomized for each individual location sample. Furthermore, 600 g of berries was weighed into 1,630 mL filter Whirl-pak bags (NASCO Sampling) and crushed using a stomacher (AES EasyMIX) for 1 min. From the crushed must, three separate 10 mL samples were removed for yeast enumeration, DNA extraction, and cryopreservation. Following centrifugation at 15,000 rpm in Eppendorf 5810R centrifuge, supernatant was retained and stored at −20°C for chemical analyses, while pellets were resuspended in 1 mL sterile DI water and stored at −20°C until extraction; for cryopreservation, pellets were resuspended in YPD containing 15% glycerol and stored at −80°C; for enumeration, pellets were resuspended in 1 mL 0.1% peptone water and processed immediately. The remaining must for each distinct location sample was aseptically transferred into a single sterile 950 mL mason jar fitted with a screw top lid. Mason jars were placed in a refrigerated incubator (VWR B.O.D Incubator) at 9°C for 6 days to replicate the traditional cold soak of Pinot Noir grapes. After cold soak, screw-top lids were replaced with sterile silicone air-lock lids (Fermentaholics.com) and were incubated at 27°C. Fermentations were weighed three times daily until ~10% of fermentable sugar was consumed according to initial measurements of must soluble solids (°Brix), which were measured using Anton Paar density meter (DMA 35). From the crushed must, three separate 10 mL samples were removed for yeast enumeration, DNA extraction, and cryopreservation. Following centrifugation at 15,000 rpm in Eppendorf 5810R centrifuge, supernatant was retained and stored at −20°C for chemical analyses, while pellets were resuspended in 1 mL sterile DI water and stored at −20°C until extraction; for cryopreservation, pellets were resuspended in YPD containing 15% glycerol and stored at −80°C; for enumeration, pellets were resuspended in 1 mL 0.1% peptone water and processed immediately. At the end of both cold-soak and fermentation stages, 1 mL samples were removed from each jar and centrifuged at 15,000 rpm in Eppendorf 5424 centrifuge. Supernatants were stored at −20°C until used for chemical analysis, while pellets were resuspended with 1 mL sterile DI water and processed for DNA extraction as described for must samples. Sample names and information are listed in Table S1.

Enumeration of grape must yeast populations

Resuspended must sample pellets were serially diluted up to 10⁻⁵ in sterile 0.1% peptone water in 96-well plates, then 5 uL of all dilutions was spot plated onto YPD media with added biphenyl (150 mg/L) in Omni-tray (Nunc Single-Well) plates for enumeration following incubation at 27°C for 2–3 days.

Chemical analysis

Thawed must supernatants were used to determine initial must pH (Mettler Toledo FiveEasy and Hanna Instruments FC240B probe) and basic juice chemical parameters following standard methods. YAN was measured by the primary-amino nitrogen assay (33) and by the enzymatic analysis of ammonia (Boehringer, Mannheim, Germany) using a SPICA auto-analyzer (Biosystems, Spain).

Construction of Hanseniaspora positive control community standard

Representative isolates of six Hanseniaspora species associated with grapes or wine in vineyard ecology studies (Table 1; Table S2) were grown separately in 5 mL of YPD broth at 27°C until turbid and hemocytometer counts were approximately 1 × 10⁸ CFU/mL. Cultures were diluted to 1 × 10⁶ CFU/mL and verified by repeat hemocytometer counting and volume adjustment. Each culture was combined at equal volume, with the exception of H. opuntiae, which was highly flocculant, and therefore a smaller volume was used (Table 1). Furthermore, 1 mL aliquots of the Hanseniaspora mock community were centrifuged for 2 min at 15,000 rpm (Eppendorf 5424 centrifuge). Supernatants were discarded, and pellets were resuspended in 1 mL YPD + 15% glycerol. Resuspended pellets were transferred to 2 mL cryo tubes (VWR) and stored at −80°C.

TABLE 1.

Composition of Hanseniaspora positive control community

Community type	Species	Study ID	Original ID	CFU proportion in mock community
Hanseniaspora	H. uvarum	Y0068	MUCL 31704	0.18
	H. valbyensis	Y0454	V5-84	0.18
	H. osmophila	Y0689	MUCL 028621	0.18
	H. guilliermondii	Y0702	MUCL 053843	0.18
	H. vineae	Y0688	MUCL 027768	0.18
	H. opuntiae	Y0687	MUCL 044418	0.1

DNA extraction

Extractions of pelleted grape must, cold-soak, and fermentation samples (n = 312) were carried out using the DNEasy Powerfood Microbial DNA Kit (Qiagen), according to manufacturer’s instructions. Samples were randomized and split into 15 extraction rounds. Each round contained a negative extraction control, using extraction kit buffer in place of sample, and a positive extraction control of either the custom Hanseniaspora mock community or MSA-2010 Mycobiome Genomic Cell Mix standard (American Type Culture Collection), which contained no Hanseniaspora species. Four samples were lost during the extraction process for a final total of 308 samples.

Metabarcoding PCR

PCR reactions were conducted with 12.5 mL of 2× Platinum Taq Master Mix (Thermo Fisher Scientific), 2.5 mL of each primer (10 mM), and 2.5 mL DNA extraction in a 25 mL reaction. Each PCR reaction was performed in duplicate on separate 96-well PCR plates (Eppendorf) that included one well allocated to the Mycobiome Genomic DNA Mix standard positive control (MSA-1010, ATCC) and three no template control wells. Dual-indexing fusion primers (ITDNA, Coralville, IA) were based upon BITS and B58S3 (8), including Illumina adapters and Nextera XT v2 adapters (Table S3), following Comeau et al. (34). PCR conditions were as follows: initial denaturation at 95°C for 5 min, followed by 35 cycles of 95°C for 1 min, 55°C for 2 min, 72°C for 2 min, and a final extension at 72°C for 10 min. PCR product quality and fragment size were verified by gel electrophoresis (2% agarose gel, 120 V/cm), and duplicate reactions were combined to purify and normalize using SequalPrep plates (Applied Biosystems, Foster City, CA) following manufacturer’s instructions. Normalized samples were pooled and then quantified using Qubit (Thermo Fisher Scientific) dsDNA high-sensitivity assay. Amplicon sequencing was carried out on a NextSeq 2000 (Illumina, San Diego, CA) with 2 × 300 bp P2 chemistry at the Center for Quantitative Life Sciences at Oregon State University.

DNA sequence processing

Demultiplexed raw reads were pre-processed using cut-adapt (35), and then raw sequence data were processed using QIIME-2 (36). Due to short amplicon lengths, some improper adapter trimming was noted, which caused failed taxonomic assignment for merged reads beyond the level of kingdom. Therefore, only forward reads were imported into QIIME-2 and processed according to the standard QIIME-2 pipeline. ITS UNITE data were used to train fungi-based classifier (Version 9.0) (37) used for taxonomic assignment. The exported ASV table (Table S4) containing 1,753 detected taxa was manually filtered using the following criteria described by Smirnova et al. (38) that were found to perform adequately compared to algorithmic approaches (38). First, any taxa with reads in fewer than five samples were excluded. Then, remaining taxa were retained that exceeded 0.001% relative abundance and were either present in one sample at >1% relative abundance, present in at least 2% of samples at >0.1% relative abundance, or present in at least 5% of samples (Table S5). The first rule removed 1,053 taxa, while the second rule removed a further 353 taxa.

While filtering the data it was noted that species H. lachancei was present in high numbers in the mock Hanseniaspora community despite not being a component added to this control. Through BLAST searches of representative sequences assigned to H. lachancei, all were re-assigned to H. opuntiae. Additionally, BLAST searches of representative sequences assigned to Hanseniaspora at genus level were predominantly reassigned to H. uvarum. During this process, we also noted that Hanseniaspora-assigned reads were evident in the ATCC MSA-1010 and MSA-2010 mycobiome control samples despite no Hanseniaspora species being included in these samples. Sequencing error and index hopping are well-known phenomena that must be considered when assessing the detection of relatively low-abundance taxa, particularly for low-DNA environmental samples (39) False-positive detection thresholds were established for each Hanseniaspora species on the basis of mean relative abundances in the ATCC extraction control samples that received similar sequencing effort (total read count) to the must, cold-soak, and fermentation samples (Table S6).

Statistical analyses

All statistical analyses were performed in the R v 4.3.2 statistical computing language (40) and R Studio v 2023.03.0+386 (41). Base R statistical analysis was used for ANOVA with post hoc Tukey tests to determine differences between sub-AVAs and vineyards for the following parameters: brix, TT10 (time to 10% fermentation), pH, YAN, and must log CFU.

The filtered metabarcoding ASV table was imported into phyloseq v1.46.0 (42), which was used to calculate alpha and beta diversity. Vegan v2.6–4 (43) was used for ADONIS permutation analysis of diversity data using distance matrices after calculating Bray-Curtis dissimilarities and sample metadata, SIMPER (44), and mantel tests. Ggplot2 v 3.4.4 (45) and ggstatsplot v 0.12.1 (46) were used for metadata analysis and visualization as well as linear regressions.

The following additional packages were used for the construction and plotting of vineyard location maps: sf v1.0–14 (47), spdata v 2.3.0 (48), tmap v3.3–4 (49), tigris v2.0.4 (50), and ggrepel v 0.9.4 (51). Latitude and longitude of each vineyard were removed from supplemental to protect the privacy of sampled vineyards.

The output of SIMPER (44) on pairwise vineyard comparisons was used to determine which taxa contributed significantly and meaningfully to differences in beta diversity within each sub-AVA at each sampling stage. For each vineyard-vineyard comparison, SIMPER calculated the “average” contribution of each taxa to all pairwise comparisons between samples and assigned significance through permutation. Only taxa with at least one vineyard-vineyard SIMPER average that was significant (P < 0.05) were retained, and then “cumulative SIMPER average” was calculated by summing SIMPER average values of each taxa for all significant vineyard-vineyard comparisons within each sub-AVA and stage. These cumulative SIMPER averages were ranked, and the top 5 contributing taxa in each comparison identified.

RESULTS

This study comprised grape berry samples from the Yamhill-Carlton and Eola-Amity Hills sub-AVAs within the Willamette Valley in Oregon (Fig. 1). To explore geospatial effects on fungal communities, sampling was designed to span three orders of magnitude of physical distance: vineyard block (nine replicates within ~300 m), sub-AVA (six vineyards within ~3 km), and AVA (two sub-AVAs within ~30 km). Must samples prepared aseptically in the laboratory were analyzed for basic parameters relevant to wine fermentation (pH, Brix, YAN, and yeast CFU). Two-way ANOVA results (Table S8) show that comparisons between sub-AVAs and between vineyards were significantly different for each of these parameters, though, in each case, vineyard explained substantially more variance between samples in two-way ANOVA models. Differences between vineyards remained evident for comparisons by one-way ANOVA within each sub-AVA (Table S8). As expected, due to the variation in grape ripeness across the sample set, °Brix was significantly positively correlated with log must yeast CFU and pH (Table S9; Fig. S1), other parameters known to be affected by ripeness.

The crushed must samples were subjected to independent cold soak and allowed to undergo uninoculated fermentations until ~10% of initial must sugar content had been consumed. Time taken for each sample to reach this point (TT10) differed significantly between vineyards but not sub-AVAs (Table S8). TT10 had a significant negative correlation with log must yeast CFU and YAN and a significant positive correlation with brix (Fig. S1; Table S9).

Overall, a total of 308 samples across must, cold soak, and early fermentation were then analyzed for fungal community structure by metabarcoding the ITS2 region alongside positive and negative controls. High-throughput sequencing of the indexed amplicon library was performed with paired reads, but due to short amplicon lengths and sequencing adapter readthrough, reverse reads were not utilized. Consequently, 133,883,561 forward reads were retained, with an average read-count per sample of 381,959. Sequencing effort was significantly greater for must, cold-soak, and fermentation samples compared to no template and negative extraction controls (Fig. S2).

Following quality filtering and rule-based removal of low-frequency and low-abundance ASVs, 347 taxa were detected across the data set.

Willamette Valley Pinot Noir must mycobiome

Individual must samples varied substantially in terms of detected taxa, ranging from 53 to 255 with an average of 135. Across the must samples, 83% of sequencing reads were assigned to filamentous fungi, while 15.5% corresponded to yeast genera. The top 5 most abundant yeast and filamentous fungi taxa (Table 2) collectively explained 81% of must sample sequencing reads.

TABLE 2.

Most abundant yeast and filamentous fungi in grape must samples

Type	Taxa	Read count	% of total
Yeast	Hanseniaspora uvarum	3,510,533	11.6
	Vishniacozyma victoriae	476,430	1.6
	Filobasidium chernovii	165,389	0.5
	Vishniacozyma carnescens	150,550	0.5
	Kazachstania sp.	56,439	0.2
Total reads		4,208,791	14.4
Filamentous fungi	Cladosporium basi-inflatum	6,958,956	23.1
	Aureobasidium pullulans	5,873,222	19.5
	Botrytis caroliniana	4,887,008	16.2
	Cladosporium herbarum	1,262,088	4.2
	Cladosporium sp.	1,095,819	3.6
Total reads		20,077,093	66.6

Geospatial and processing stage effects upon Willamette Valley Pinot Noir mycobiome diversity

To determine the variation in taxa diversity between samples and geospatial locations, the alpha and beta diversity metrics were used. Simply put, alpha diversity is the measure of the number of taxa present within a sample with various statistics determining taxa richness (Chao1) and evenness (Shannon). Beta diversity was assessed using a Bray-Curtis dissimilarity matrix that captures differences in taxonomic presence and relative abundance between samples and geospatial locations.

Considering the whole data set, there was a significant difference between vineyards when using the Chao1 statistic (Chao1, P = 3.2e−5) but not the Shannon statistic (Shannon, P = 0.0645). In addition, we found a significant difference in alpha diversity between sample stages (Shannon P ≤ 2e−16, Chao1 ≤2e−16, Fig. 2), but there was no difference in alpha diversity between sub-AVAs (Chao1 P = 0.314, Shannon P = 0.912), (Table S10). Alpha diversity of must samples showed no significant differences between sub-AVAs for both Shannon and Chao1 metrics; however, there was a significant difference between vineyards for both (Table S10). Notably, the decrease in values for both indices during processing indicates a collapse of taxonomic richness during cold soak and ferment. Regardless, the significant difference between vineyard must samples persisted in cold-soak and ferment samples (Table S10).

Fig 2.

Fungal diversity varies according to stage of grape processing and location of sampling. Alpha diversity richness plot of Chao1 index for richness and Shannon index for richness and evenness for all samples, grouped by grape-processing sample stage and colored by location of sampling.

Prior to assessing whether sub-AVA, vineyard, or processing stage contributed to mycobiome structure, mantel tests (Table S11) were performed on the must sample Bray-Curtis dissimilarity matrix to determine whether physical distance and must metrics correlate to differences in must sample mycobiomes. Physical distance significantly correlated with beta diversity of all the must samples (Mantel, P = 5.00e−04) as well as within Eola (Mantel, P = 1e−04) but not within Yamhill (Mantel, P > 0.05). For Eola, TT10, YAN, and log must yeast CFU were significantly correlated with beta diversity, while in Yamhill Brix, TT10 and log must yeast CFU were significantly correlated (Table S11).

Non-metric multidimensional scaling (NMDS) plots of Bray-Curtis distances (Fig. 3) and PERMANOVA revealed a significant difference in beta diversities between sub-AVAs, vineyards, and sample stages (Table S12). Significant differences remained evident for vineyard and sub-AVA when the data set was stratified by sample stage and, for sub-AVAs and stages, when stratified by vineyard (Table S12). However, the difference in beta diversity between sub-AVAs noted for must was not evident at cold-soak or ferment processing stages (Table S12). PERMANOVA on separate sub-AVA data sets reinforced the existence of vineyard-level differences in beta diversity (Table S12).

Fig 3.

Fungal diversity varies according to stage of grape processing and location of sampling. (A) Beta diversity represented by two-dimensional NMDS plots of pairwise Bray-Curtis dissimilarity of sample ASV counts. Data organized by the location of sampling and colored by grape-processing sample stage, with 95% confidence intervals for sample stage indicated by colored ellipses. (B) Bray-Curtis dissimilarity NMDS plots of vineyards. Data are organized by sub-AVA and processing stage and colored by vineyard with 95% confidence intervals for vineyard indicated by colored ellipses.

Taxa that made the largest contributions to the differences in beta diversity between vineyards and processing stages were identified using the SIMPER method of partitioning Bray-Curtis dissimilarities (Fig. 4). The top 5 taxa at each stage in each sub-AVA accounted for at least 70% of the proportion of contribution to total SIMPER cumulative averages (Table S13) and exceeded 80% for Eola cold-soak and Yamhill ferment (Table S13). Despite greater relative abundance of filamentous fungi in must samples, taxa contributing the most to differences between vineyards were evenly split between yeast and filamentous fungi, and it was evident that filamentous fungi decrease in relative contribution to beta diversity through cold soak and fermentation, likely due to the enrichment of yeasts during these processes. Among yeast taxa, H. uvarum had the highest cumulative SIMPER average in Yamhill must and second highest in Eola, and was one of the top 5 contributors to vineyard differences at every stage in both sub-AVAs, except fermentation in Eola.

Fig 4.

Variation in fungal beta diversity and taxa contributing to differences between vineyards at different processing stages. SIMPER averages for all taxa contributing significantly to differences in beta diversity for at least one pairwise vineyard-vineyard comparison were summed (cumulative SIMPER average) and ranked within sub-AVA and processing stage. The top 5 taxa within each sub-AVA and processing stage were retained and plotted.

Hanseniaspora uvarum is the most important Willamette Valley Pinot Noir Hanseniaspora species

For this study, we constructed a positive control comprising six Hanseniaspora species previously associated with grapes/wine. All six species included in the Hanseniaspora control were assigned reads in samples and passed manual filtering criteria. As shown in Fig. S3B, species H. osmophila and H. vineae were over-represented in terms of average read count vs CFUs, while species H. uvarum, H. guilliermondii, and H. valbyensis were under-represented. Across the entire grape must, cold-soak, and fermentation sample set, the most abundant Hanseniaspora species were H. uvarum and H. valbyensis, comprising 90.3% and 3.1% of Hanseniaspora reads, respectively. The relative abundance of both species in the positive Hanseniaspora controls was lower than expected based on input cell numbers, thus it is possible their true contribution to Hanseniaspora in our samples was underestimated. Additional filtering criteria based upon 99% CI derived from baseline Hanseniaspora reads in the defined ATCC-2010 Mycobiome sample (Table S6; Fig. S4) were applied to account for potential sequencing/index-hopping errors when estimating prevalence. Results post-filtering showed that all six Hanseniaspora species were detected in must samples (Table 3) and allowed the prevalence of each species at each stage of processing to be more reliably assessed (Table 3). However, H. uvarum was the most prevalent, occurring in 59% of must samples above threshold. As samples progressed into cold soak and ferment, there was a noticeable decline in species prevalence for every species except H. uvarum, which increased in prevalence and maximum relative abundance. H. lachancei/opuntiae was the second most prevalent in must samples; however, the maximum relative abundance in an individual sample was only 0.004 compared to H. uvarum at 0.73. Indeed, low maximum relative abundances were observed for all Hanseniaspora species other than H. uvarum. Differences in Hanseniaspora species abundance between sample controls were compared. There was no significant difference for randomized samples processed in batches with or without the Hanseniaspora mock community standard for all species other than H. lachancei/opuntiae (Table S14).

TABLE 3.

Prevalence of detected Hanseniaspora species at each processing stage and the maximum RA for an individual sample^a

Taxa	Must		Cold soak		Ferment
	Prevalencea	Max RAc	Prevalence	Max RA	Prevalence	Max RA
H. uvarum	0.59	0.73	0.96	0.93	1.0	0.96
H. valbyensis	0.09	0.06	0.02	0.002	0.04	0.01
H. lachancei/opuntiae	0.05	0.004	0.00	–b	0.00	–
H. osmophila	0.11	0.002	0.06	0.0008	0.07	0.006
H. vineae	0.08	0.002	0.04	0.0008	0.03	0.0008
H. guilliermondii	0.03	0.001	0.00	–	0.01	0.0007

^a Samples scored as positive if measured RA exceeded 95% CI for mean plus two standard deviations of RAs for that species within replicate (n = 7) ATCC-2010 Mycobiome control samples.

^b –, maximum RA below 95% CI cutoff.

^c RA, relative abundance.

Two-way ANOVA of H. uvarum relative abundances across the whole data set showed significant differences between stages and vineyards (P = 2.00e−16, P = 2.41e−6) but not between sub-AVAs (P = 0.492). When the data were split into separate sub-AVAs, Yamhill was found to have a significant difference in H. uvarum abundance between vineyards (ANOVA, P = 2.10e−7), but Eola did not (ANOVA, P = 0.0603). ANOVAs for each stage confirmed there was no significant difference in H. uvarum abundance between sub-AVAs (Table S15). Sub-AVA data were then analyzed separately. Within both Eola and Yamhill, there was a significant difference in H. uvarum abundance between stages (ANOVA, P = 2e−16, P = 2e−16). Furthermore, within Eola and Yamhill, there were significant differences in H. uvarum relative abundance between vineyards at the must stage (Table S15), while at cold soak, this difference persisted only for Yamhill (Table S15). At the ferment stage, there were no differences between vineyards within either sub-AVA for H. uvarum relative abundance (Table S15). These observations are consistent with H. uvarum’s relative contributions to between vineyard differences at each stage based on SIMPER output (Fig. 4).

H. uvarum relative abundances for samples at each stage were sorted from largest to smallest and given a rank (1 − n), with 1 representing the sample with highest relative abundance. Spearman rank correlations (Fig. 5; Table S16) show that samples with high rank for H. uvarum abundance in must were more likely to have high rank at cold soak (P = 1.8e−16, r = 0.71) and fermentation (P = 0.02, r = 0.23), and high rank for H. uvarum abundance at cold soak corresponded to higher ranking in early fermentation (P = 2.23e−9, r = 0.55). Of the measured must compositional parameters, only YAN and log-CFU were positively correlated with H. uvarum relative abundance (Table S16).

Fig 5.

Relative abundance (RA) of H. uvarum during cold soak and early fermentation is dependent upon RA in previous stage. (A) Distribution of H. uvarum sample RA colored by rank (1 = highest abundance) and organized by sub-AVA. (B) Regression of RA rankings for cold soak vs must with 95% CI. (C) Regression of RA rankings for ferment vs cold soak with 95% CI.

DISCUSSION

Mycobiome diversity of must indicates potential for microbial terroir in the Willamette Valley, OR

While the mycobiome does not represent all microbial diversity in the vineyard, it has proven to be a good indicator of regionality of microorganisms, or microbial terroir (8, 52). Only a limited number of studies have evaluated grape berry/must mycobiomes on the West Coast of Northern America (8, 53, 54), and none from Oregon. Consistent with these other studies, we observed the majority of grape must metabarcoding reads were associated with filamentous fungi. The most abundant taxa belonged to Cladosporium (30.9% of all sequencing reads), a similar finding to Bokulich et al. (8) from CA winery-prepared musts. We also observed high relative abundances of Aureobasidium (19.5%) and Botrytis (16.2%), both detected with relative abundances of at least 20% of grape-wash/crushed-must samples from Washington State, USA (WA) and British Columbia, Canada (BC) (53, 54). Yeast taxa in our study were primarily made up of H. uvarum, which contributed 11.6% of total must mycobiome reads. This is higher than reported by Bokulich et al. (8), who found 5% of sequencing reads from CA must samples corresponded to H. uvarum. The BC study only identified to the genus level but did find Hanseniaspora in low levels during early fermentations, and some Hanseniaspora ASVs were identified in late-stage fermentation (54). The WA study identified some Hanseniaspora species, specifically H. uvarum and H. osmophila ASVs in veraison (ripening grapes) and harvest samples (55). These studies found higher proportions of Vishniacozyma and Filobasidium, though it is worth noting both genera were represented among the five most abundant yeast taxa in our study.

Observed differences in taxa relative abundance between different studies reflect differences in geography, vineyard climate, and vineyard practices but are also likely contributed to variation in sample collection method, among other known artifactual causes of variation between metabarcoding studies (56). The studied grape-growing regions vary between hot and dry in California to wet and humid in Oregon (57), and as shown in Bokulich et al. (8), variations in climate can result in differences in mycobiome and microbiome structure. However, Bokulich et al. (8) collected must samples processed commercially within wineries, meaning community structures were a composite of those in the vineyard and the winery and its equipment (58). In other cases where grape berry samples were taken directly from the vineyard, studies that used the washing method of grapes (32), such as reference (54), are less likely to detect fermentative yeast than those who sample from winery-prepared must/juice (8, 59) or laboratory-prepared must, as we chose to do in this study. Nevertheless, taking this into consideration, the major taxa we observed in our sample set were comparable to other studies from the west coast of North America.

The extent to which diversity metrics varied geospatially within our sample set was also in-line with the comparable west coast North American studies. In WA and BC, differences were observed between vineyards for Chao1 richness but not the Shannon index of richness and evenness (53, 54). However, a study across a larger distance (~300 km) did find significant differences in the Shannon index between regions (52) which we did not observe, possibly due to the smaller distance (~31 km) between the Eola and Yamhill sub-AVAs we sampled from. Nevertheless, we found that beta diversity varied between vineyards within sub-AVAs and between sub-AVAs themselves, similar to the CA and BC studies (8, 54).

Within winegrowing regions it has been established that vineyards (or vineyard blocks), as defined commercially, represent discrete cohesive units based on the grape varietal and viticultural management practices (60). We chose to sample from biodynamic or organically managed vineyards based upon reduced impact on grape berry fungal communities compared to conventional practices involving synthetic fungicides (61, 62) but could still observe clear vineyard-vineyard differentiation based upon grape must composition and mycobiome diversity. Vineyard as an operational unit can be confounding when seeking to address questions of microbial terroir as a function of geospatial location, particularly if insufficient within-vineyard replicate samples are collected. With nine replicates within vineyards, and between vineyard geospatial distance within sub-AVA constrained to an order of magnitude less than between sub-AVA distance, we were able to observe a difference in mycobiome beta diversity that was not solely a function of vineyard. Furthermore, within the Eola sub-AVA, we observed distance-decay relationships of mycobiome beta diversity across ~7.6 km, in-line with other grape and vineyard soil microbiome studies that observed distance-decay relationships across ~2 km (12, 54).

H. uvarum is the key Hanseniaspora species associated with Willamette Valley Pinot Noir must and is a major contributor to between vineyard differences in mycobiome beta diversity

Previously mentioned grape fungal diversity studies of the west coast of the United States report the detection of Hanseniaspora spp., including H. uvarum (8, 32, 54). Studies outside of North America find other Hanseniaspora species along with H. uvarum, such as H. osmophila, H. guilliermondii, and H. nectarophila (63–65). In reference (65), H. guilliermondii and H. nectarophila were the most abundant Hanseniaspora species, while H. uvarum was not detected. In our must sample set, we detected six Hanseniaspora species above thresholds set using a quality-controlled mycobiome standard from ATCC that did not contain any Hanseniaspora as DNA-equivalent negative control. That said, the six we found were the same species included in our mock Hanseniaspora community, and it cannot be excluded that detection thresholds used allowed for some false positives when scoring prevalence among must, cold-soak, and ferment samples. Furthermore, other than H. uvarum, the other Hanseniaspora species were only detected at low relative abundance numbers, consistent with the possibility this signal was due to index hopping (39) or cross-contamination during extraction.

The fact that H. uvarum is so ubiquitous in grape must samples across many studies (8, 14, 30, 66) may contribute to it being overlooked as a component of microbial terroir. Using SIMPER analysis, we found H. uvarum to be a major contributor to differences in beta diversity between vineyards. This method has been used in previous vineyard studies to determine the importance of taxa based on physical location, vineyard management, stages of fermentation, and soil microbiology (67–69). While all of these cited studies use SIMPER to compare Bray-Curtis dissimilarity as we did in this study, the approach has also been used to identify the similarity between samples as done in Pinto et al. (67).

While H. uvarum has shown to be highly abundant on wine grapes in this study and those previously mentioned, it has been found in abundance in other fruit-related niches such as berries, apples, and citrus (70–72). In addition to its classification as an American Viticultural Area, Oregon’s Willamette Valley produces a range of other crops, including small fruits (e.g., raspberries, blueberries, cherries, and strawberries), in close proximity to vineyards (73). These fruits can harbor H. uvarum in their own mycobiomes (72, 74), and it has been shown that the volatile aroma compounds produced by H. uvarum attract the pest Drosophila suzukii, also known as spotted wing drosophila (SWD) (75). SWD is a known pest affecting small fruits in Oregon (76) that has also shown affinity for wine grapes (77), thus it is possible the close proximity of vineyards to other fruit operations contributes to the presence of H. uvarum on wine grapes through insect transfer.

Mycobiome diversity collapse during cold soak and early fermentation is driven by must H. uvarum populations

The process of cold soak, or pre-fermentation cold maceration, is a technique used in red wines to enhance color quality and aroma by increasing the concentration of anthocyanins through prolonged juice contact with the grape skins (78). This is a popular method in the production of Pinot Noir wine to increase the concentration of important quality markers, such as anthocyanins and tannins (79). However, cold soak when followed by uninoculated fermentation can cause unwanted changes in aroma composition due to growth of cryotolerant yeast species present, such as H. uvarum in the must (17). The extent to which presence/absence and relative abundances of cryotolerant yeasts in musts of different compositions, or from different geospatial locations, are predictive of potential impacts on wine quality has not been evaluated.

When comparing initial must mycobiome composition to sample composition after undergoing cold soak and fermentation, we observed significant differences in mycobiome alpha diversity between sample stages, which showed there was a larger diversity of fungal taxa in must samples than cold soak or ferment. This collapse in diversity seen through processing is consistent with other studies that observe such a collapse during various stages of fermentation (14). In our study, the collapse in alpha diversity coincided with loss of distinction between sub-AVAs in terms of beta diversity, though differences between vineyards persisted.

In terms of Hanseniaspora spp., studies of mycobiome diversity throughout processing find the abundance of H. uvarum is highest in must and decreases during fermentation as more alcohol is produced by predominant fermentation yeast, Saccharomyces cerevisiae (31, 80). This is a consequence of H. uvarum’s modest ethanol tolerance (16) and is the reason why fermentation samples in this study were taken after only 10% of sugar consumption to facilitate observation of Hanseniaspora communities prior to loss of viability. During cold soak, H. uvarum can be tolerant to the cold temperatures of the technique and produce unwanted aroma compounds during this process (17, 22), though it is not always detected potentially due to varying starting amounts of H. uvarum or differences in must composition. Growth capabilities of other Hanseniaspora species have not been noted in literature, so it is possible they also exhibit this cryotolerance; however, due to low abundance of other species in this study, correlations between abundance of other species and cold soak were not tested. When comparing H. uvarum abundance and sample stages, the largest increase in relative abundance for most samples in this study is between must and the end of cold soak, meaning that cold soak is a likely enrichment step for the species. These findings agree with previous reports of increases in H. uvarum populations during cold soaking (22). Importantly, by generating a large number of must samples that each continued independently through processing, we were able to show correspondence between initial abundances and those resulting from cold soak despite variation in must mycobiome and chemical composition.

This study provides the first biogeographical analysis of grape berry mycobiome composition in the state of Oregon, finding evidence of microbial terroir between vineyards and two sub-AVAs in the state. Hanseniaspora uvarum was not only one of the strongest contributing species to differences in vineyard beta diversity but was also the most prevalent and abundant Hanseniaspora species at each wine processing stage. We found processing had a significant impact on H. uvarum abundance, and cold soak is a likely enrichment step for the yeast. While further testing is required to determine functional impact, we found a significant positive correlation between H. uvarum abundance in must and the resulting abundance in cold soak and ferment. These results provide guidance to winemakers considering allowing uninoculated fermentations to proceed if H. uvarum is a spoilage concern.

DATA AVAILABILITY

Raw sequencing data are available at BioProject PRJNA1080583. Code for sequence processing and statistical analyses is available at https://github.com/curtinlab/Oregon_Pinot_Noir_Mycobiome.git.

SUPPLEMENTAL MATERIAL

The following material is available online at https://doi.org/10.1128/aem.00810-24.

Supplemental figures. aem.00810-24-s0001.tiff.

Figures S1 to S4.

Supplemental references. aem.00810-24-s0002.docx.

List of references cited in supplemental material.

Supplemental tables. aem.00810-24-s0003.xlsx.

Tables S1 to S16.

ASM does not own the copyrights to Supplemental Material that may be linked to, or accessed through, an article. The authors have granted ASM a non-exclusive, world-wide license to publish the Supplemental Material files. Please contact the corresponding author directly for reuse.