The Interior Surfaces of Wooden Barrels Are an Additional Microbial Inoculation Source for Lambic Beer Production

Although the coolship step is generally regarded as the main contributor to the spontaneous inoculation by environmental air of fresh worts for lambic beer production, it is known that microorganisms often associate with specific surfaces present in a brewery. However, knowledge about the association of microorganisms with the interior surfaces of wooden lambic barrels is limited. To clarify the role of casks and foeders as additional microbial inoculation sources, it was important to determine the influence of the barrel characteristics and the cleaning procedures on the microbial communities of the interior barrel surfaces. Moreover, this helped to elucidate the complex spontaneous lambic beer fermentation and maturation process. It will allow further optimization of the lambic beer production process, as well as the wooden-barrel-cleaning procedures applied.

ABSTRACT

Traditional lambic beer production takes place through wort inoculation with environmental air and fermentation and maturation in wooden barrels. These wooden casks or foeders are possible additional inoculation sources of microorganisms for lambic worts. To date, however, these lambic barrels have been examined only with culture-dependent techniques, thereby missing a portion of the microorganisms present. Moreover, the effects of the cleaning procedures (involving high-pressure water and/or fumigation) and the barrel type on the microbial community structures of the interior surfaces of wooden lambic barrels were unclear. The culture-dependent plating and culture-independent amplicon sequencing of swab samples obtained from the interior surfaces of different wooden casks and foeders used for traditional lambic beer production in Belgium revealed that the microbial compositions of these surfaces differed statistically throughout the barrel-cleaning procedures applied. At the end of the cleaning procedures, amplicon sequencing still detected fermentation- and maturation-related microorganisms, although only a few colonies were still detectable using culture-dependent methods. It is possible that some of the surviving microorganisms were missed due to the presence of many of these cells in a viable but not culturable state and/or engrained deeper in the wood. These surviving microorganisms could act as an additional inoculation source, besides brewery air and brewery equipment, thereby helping to establish a stable microbial community in the wort to diminish batch-to-batch variations in fermentation profiles. Furthermore, the microbial compositions of the interior barrel surfaces differed statistically based on the barrel type, possibly reflecting different characteristics of the lambic barrels in terms of age, wood thickness, and wood porosity.

IMPORTANCE Although the coolship step is generally regarded as the main contributor to the spontaneous inoculation by environmental air of fresh worts for lambic beer production, it is known that microorganisms often associate with specific surfaces present in a brewery. However, knowledge about the association of microorganisms with the interior surfaces of wooden lambic barrels is limited. To clarify the role of casks and foeders as additional microbial inoculation sources, it was important to determine the influence of the barrel characteristics and the cleaning procedures on the microbial communities of the interior barrel surfaces. Moreover, this helped to elucidate the complex spontaneous lambic beer fermentation and maturation process. It will allow further optimization of the lambic beer production process, as well as the wooden-barrel-cleaning procedures applied.

INTRODUCTION

Despite the convenience of the use of starter cultures, a whole range of fermented foods and beverages are still produced by spontaneous fermentation (1–4). However, a combination of different factors, such as the indigenous microbiota, processing conditions, and raw materials, can hamper the outcome of these processes. Belgium is known for its traditional spontaneous mixed-fermentation beers, such as lambic beers, which are traditionally produced through air inoculation of the wort in an open coolship during the winter months (5–7). The air provides the microorganisms necessary for successful lambic beer production, while its low temperature prevents the growth of harmful microorganisms. The actual fermentation and maturation of lambic beers, however, are carried out in horizontal wooden casks of oak or chestnut, whose contribution to microbial inoculation has scarcely been studied (8–10).

In (port) wine and whiskey production, wooden barrels are predominantly used for flavor formation through the extraction of typical wood compounds, such as phenolics and tannins, during maturation. However, these compounds are mostly lost after extensive usage of these barrels. Moreover, due to their physical inertness and porosity, their wooden surfaces are difficult to sanitize and thus harbor resident microbiota that can negatively affect the organoleptic properties of wines (5, 8, 11). Since the extraction of wood compounds is of no interest during lambic beer production, old barrels (port wine casks and wine foeders) are reused in spontaneous lambic beer production processes, thereby further exploiting the resident microbiota present (5). Indeed, some lambic beer maturation-related microorganisms (such as the yeast Dekkera bruxellensis and the lactic acid bacterial species Pediococcus damnosus) have been isolated from lambic beer barrels (10). They probably originated from previous production processes, since these casks are used continuously for lambic beer production and are cleaned only superficially, with high-pressure water, between consecutive production batches. Moreover, D. bruxellensis is known to reside up to 8 mm deep in wooden wine barrels, requiring a necessary steaming time of 12 min for inactivation (12). Hence, the wooden barrels can be regarded as a possible inoculation source for the production of lambic beer. However, since some microorganisms possibly occur in a viable but not culturable (VBNC) state and/or occur in low relative abundances after cleaning procedures, culture-dependent techniques can easily result in inaccurate enumeration and identification of resident microorganisms on wooden cask surfaces (8). Therefore, amplicon sequencing has been applied in several studies to determine the microbial communities present on different surfaces in food fermentation facilities (13–16). It was applied, for instance, to determine the bacterial and fungal communities within the environment of a brewery that acts both as a conventional beer-producing facility and as an American coolship ale (ACA) production facility (17). ACA fermentation-related microorganisms, mostly lactic acid bacteria (LAB) and molds, were found on the exterior barrel surfaces present in that brewery. Moreover, contacts of brewery equipment surfaces with grains, hops, yeasts, and fermenting or finished beers act as the driving forces behind the association of microorganisms with these brewery surfaces and the formation of microbial communities on the different brewery apparatuses. This surface-driven community structuring is a common phenomenon in, for example, wineries and artisan cheesemaking plants (13–16).

Although the studies mentioned above have revealed the association of microorganisms with specific surfaces present in breweries, barrels used in lambic beer breweries have been examined only with culture-dependent techniques (10). Therefore, many microorganisms might have been missed, as the greater part of these microorganisms cannot be cultivated. Moreover, the effects of the cleaning procedures, particularly the use of high-pressure water and/or the application of fumigation, as well as the barrel type, on the microbial community structure of the interior of wooden barrels are unclear. Therefore, both culture-dependent and culture-independent examination of swab samples obtained from the interior surfaces of different wooden barrels present in a traditional lambic beer brewery, spanning all cleaning procedures applied, needed to be performed. This study aimed at investigating the possibility of wooden barrels as an additional source of specific microorganisms for lambic beer fermentation and maturation, by examining the impact of different barrel characteristics and cleaning procedures in a traditional lambic brewery in Belgium.

RESULTS

Microbial enumerations.

The interior surfaces of six wooden barrels used in the traditional production of lambic beer were sampled throughout their cleaning procedures (Fig. 1 and 2). These barrels represented three identical oak port wine casks (referred to as C1, C2, and C3) and three oak wine foeders (F1, F2, and F3). In general, the microbial counts of the interior surfaces of these barrels were high immediately after their emptying (precleaning [PC] stage) (Table 1). However, all barrels displayed large variabilities in terms of microbial enumerations. Despite the high initial microbial loads, the viable counts decreased throughout the cleaning procedures, to counts below log 2.0 CFU/cm². Regarding the foeders, viable counts below log 2.0 CFU/cm² were determined from most agar media examined after the high-pressure water cleaning step (after cleaning [AC] stage). Although this step also removed bacteria and yeasts substantially in the samples obtained from the casks, viable counts of approximately log 3.0 CFU/cm² could still be obtained. When sulfuring was applied (after sulfuring [AS] stage), only counts below log 2.0 CFU/cm² were detected until the casks were refilled with fresh wort (before filling [BF] stage). F2 yielded counts below log 2.0 CFU/cm² on yeast extract-peptone-glucose with cycloheximide (YPGc) agar medium, which could be due to sampling biases or the lower presence of cycloheximide-resistant yeast species in this foeder.

FIG 1.

(A) Wooden cask used for lambic beer production. (B) Bunghole used for sampling (now closed with a loose wooden panel). The specific features of the casks are given below the pictures. (C) Wooden foeder (F1 and F2) used for lambic beer production. (D) Wooden foeder (F3) used for lambic beer production. (Panels E) Manhole used for sampling. The specific features of the foeders are given below the pictures.

FIG 2.

(A) Cleaning procedure for the wooden casks used for lambic beer production, between their emptying (removal of matured lambic beer) and filling with fresh wort. The different stages that were sampled are indicated as PC (precleaning), AC (after cleaning), AS (after sulfuring), and BF (before filling). (B) Cleaning procedure for the wooden foeders used for lambic beer production, between their emptying (removal of matured lambic beer) and filling with fresh wort. The different stages that were sampled are indicated as PC (precleaning) and AC (after cleaning).

TABLE 1.

Colony counts of samples from wooden casks and foeders used for lambic beer production on different agar media

Barrel and cleaning stagea	Colony count (log CFU/cm2)b
	PCA	PCAc	YPG agar	YPGc agar
Cask 1
PC	5.1 ± 0.1	3.2 ± 0.1	5.3 ± 0.1	4.9 ± 0.1
AC	3.3 ± 0.1	2.6 ± 0.2	3.2 ± 0.1	3.1 ± 0.1
AS	ULQ	ULD	ULQ	ULD
BF	ULQ	ULQ	ULD	ULD
Cask 2
PC	4.6 ± 0.1	2.7 ± 0.1	4.8 ± 0.1	4.7 ± 0.4
AC	3.2 ± 0.1	3.0 ± 0.1	3.2 ± 0.1	3.1 ± 0.1
AS	ULQ	ULQ	ULD	ULD
BF	ULQ	ULQ	ULQ	ULD
Cask 3
PC	3.9 ± 0.1	3.4 ± 0.1	4.0 ± 0.1	3.2 ± 0.3
AC	3.0 ± 0.1	2.5 ± 0.3	2.9 ± 0.2	2.9 ± 0.1
AS	ULQ	ULQ	ULQ	ULQ
BF	ULQ	ULQ	ULQ	ULD
Foeder 1
PC	4.1 ± 0.1	4.5 ± 0.1	3.2 ± 0.2	3.9 ± 0.1
AC	ULD	ULQ	ULQ	ULD
Foeder 2
PC	4.7 ± 0.1	4.8 ± 0.1	2.6 ± 0.1	ULQ
AC	ULQ	ULQ	ULQ	ULQ
Foeder 3
PC	4.5 ± 0.1	4.8 ± 0.2	5.2 ± 0.1	5.2 ± 0.1
AC	ULQ	ULQ	ULQ	ULQ

^aThe different cleaning stages are indicated as follows: PC, precleaning; AC, after cleaning; AS, after sulfuring; BF, before filling.

^bPCA was used for the total microbial counts, PCAc for the total bacterial counts, YPG agar for the yeast counts, and YPGc agar for the cycloheximide-resistant yeast counts. Values represent mean counts ± standard deviations. Standard deviations represent technical replicates from single swab samples. ULQ, under the limit of quantification (a few colonies were detected on the agar but below log 2.0 CFU/cm²); ULD, under the limit of detection (no colonies were detected on the agar).

The PC stage displayed the highest microbial counts on plate count agar (PCA), with C1 showing the highest average microbial load of log 5.1 CFU/cm², followed by C2 and C3 with log 4.6 CFU/cm² and log 3.9 CFU/cm², respectively. The foeders showed average microbial counts of log 4.1 CFU/cm², log 4.7 CFU/cm², and log 4.5 CFU/cm² for F1, F2, and F3, respectively. The bacterial counts on PCA with cycloheximide (PCAc) were generally lower than the counts obtained on PCA, with averages of log 3.2 CFU/cm², log 2.7 CFU/cm², and log 3.4 CFU/cm² for C1, C2, and C3, respectively. For F1, F2, and F3, the counts on PCAc and PCA were comparable, with even slightly higher counts on PCAc (log 4.5 CFU/cm², log 4.8 CFU/cm², and log 4.8 CFU/cm², respectively).

Regarding the yeast counts on yeast extract-peptone-glucose (YPG) agar, the PC stage also displayed the highest loads, with log 5.3 CFU/cm², log 4.8 CFU/cm², and log 4.0 CFU/cm² for C1, C2, and C3, respectively. The foeders displayed more pronounced yeast count variations, namely, log 3.2 CFU/cm², log 2.6 CFU/cm², and log 5.2 CFU/cm² for F1, F2, and F3, respectively. The major difference in the counts obtained from YPG agar for F1 and F2 versus F3 presumably reflected the difference in the age of the lambic brews the foeders contained before they were emptied (young lambic beer was present in F3). The high yeast counts in F3 indicated the presence of more viable yeast sediments from the maturation phase in that foeder. The yeast counts on YPGc agar were log 4.9 CFU/cm², log 4.7 CFU/cm², and log 3.2 CFU/cm² for C1, C2, and C3, respectively, and log 3.9 CFU/cm², below log 2.0 CFU/cm², and log 5.2 CFU/cm² for F1, F2, and F3, respectively. Independent of these different initial values, general decreases of the microbial loads were found after the AC stage, to counts below log 2.0 CFU/cm² for the foeders and to a more constant level of approximately log 3.0 CFU/cm² for the casks. This finding indicated that cleaning with high-pressure water had only a limited capacity to remove residual bacteria and yeasts. Only a few colonies were found at the AS and BF stages, indicating that the bacteria and yeasts still present were mostly killed or entered a VBNC state when exposed to this fumigation step. Besides the few colonies detected by swabbing, there could be more viable microorganisms present deeper in the wood and inaccessible for the swabbing method applied.

The most abundant microbial group present on the inner cask surfaces represented yeast species, among which cycloheximide-resistant yeasts made up a large part. Indeed, similar yeast counts were obtained for PCA, YPG agar, and YPGc agar, with the latter selecting for cycloheximide-resistant yeast species. In contrast, the smaller differences between microbial counts on PCA and PCAc and the fact that the latter were generally higher than those on YPG agar indicated that there were more bacteria than yeasts present in the foeders, in comparison with the casks. The differences in colony counts between the different wooden barrels, even barrels of the same type and age, indicated heterogeneous microbial compositions, as expected.

Bacterial species composition.

Amplicon sequencing of the bacterial V4 region of the 16S rRNA gene was used to investigate the total bacterial community structure of the inner surfaces of the wooden casks and foeders examined throughout the different cleaning procedures. The overall error rate, based on a mock community analysis, was 0.03%. Small differences in relative abundances of taxonomic assignments at the genus level were present when the methods based on operational taxonomic units (OTUs) (mothur) and amplicon sequence variants (ASVs) (DADA2) were compared. The OTU-based analysis generally attributed more reads to Pediococcus, whereas the ASV-based analysis attributed more reads to Cellulosimicrobium (Fig. 3). The latter method also attributed more reads to the “others” group. Neither method could perform taxonomic assignments to the species level. In general, statistical differences (P < 0.05) in relative abundances of OTUs and ASVs were found between the inner surfaces of the casks and foeders (Fig. 3). Regarding the foeders, the microbial community structure was dominated by Pediococcus species (Table 2 and Fig. 3). For the casks, Pediococcus species were also prevalent but relative abundances were more distributed across different taxa, such as Acetobacter, Acinetobacter, Brevibacterium, Cellulosimicrobium, Lactobacillus, Sphingomonas, and Staphylococcus. This resulted in greater diversity and evenness for the casks (Table 2).

FIG 3.

Relative abundances of bacterial OTUs (A and C) and ASVs (B and D) obtained by amplicon sequencing of all inner wooden surfaces of casks (A and B) and foeders (C and D) used for lambic beer production, sampled during their respective cleaning procedures. OTUs with occurrences of <50 were discarded; all OTUs and ASVs with occurrences of <1,000 were grouped as “others.” In the case of the OTU-based analysis, the “others” group consisted of 66 different genera present at low abundances; in the case of the ASV-based analysis, the “others” group consisted of 64 different genera present at low abundances.

TABLE 2.

Alpha diversity metrics based on the relative abundances of bacterial and yeast genera obtained by both the OTU- and ASV-based bioinformatic analysis methods, applied to samples from the interior surfaces of wooden casks and foeders used for lambic beer production

Barrel and cleaning stagea	Alpha diversity indexb
	Bacteria				Yeast
	OTU		ASV		OTU		ASV
	Simpson	Pielou	Simpson	Pielou	Simpson	Pielou	Simpson	Pielou
Cask 1
PC	0.51	0.24	0.55	0.45	0.22	0.16	0.09	0.18
AC	0.68	0.37	0.77	0.58	0.74	0.52	0.75	0.61
AS	0.86	0.61	0.88	0.71	0.66	0.43	0.69	0.56
BF	0.73	0.47	0.68	0.52	0.70	0.57	0.72	0.64
Cask 2
PC	0.55	0.27	0.56	0.55	0.12	0.10	0.10	0.15
AC	0.65	0.43	0.81	0.59	0.79	0.59	0.65	0.54
AS	0.76	0.50	0.74	0.59	0.73	0.52	0.59	0.45
BF	0.92	0.68	0.92	0.78	0.83	0.64	0.84	0.76
Cask 3
PC	0.84	0.53	0.86	0.66	0.62	0.46	0.50	0.48
AC	0.79	0.52	0.82	0.63	0.81	0.65	0.33	0.30
AS	0.87	0.58	0.86	0.67	0.80	0.60	0.81	0.64
BF	0.76	0.49	0.70	0.57	0.75	0.61	0.71	0.56
Foeder 1
PC	0.01	0.02	0.01	0.03	0.58	0.38	0.55	0.41
AC	0.01	0.02	0.02	0.04	0.61	0.41	0.57	0.52
Foeder 2
PC	0.01	0.02	0.01	0.04	0.60	0.47	0.56	0.58
AC	0.17	0.12	0.21	0.18	0.49	0.34	0.64	0.77
Foeder 3
PC	0.00	0.00	0.00	0.00	0.25	0.23	0.17	0.32
AC	0.36	0.21	0.40	0.28	0.44	0.28	0.39	0.36

^aThe different cleaning stages are indicated as follows: PC, precleaning; AC, after cleaning; AS, after sulfuring; BF, before filling.

^bThe Simpson (D) and Pielou (Je) indexes were calculated for all samples, to measure their diversity and evenness, respectively.

At the PC step for the casks, the bacterial species composition of C1 and C2 was relatively limited, consisting of mainly three genera, namely, Acetobacter, Cellulosimicrobium, and Pediococcus, although their relative abundances were distributed differently (Fig. 3). C2 and C3 exhibited a high relative abundance of Acetobacter at the expense of Cellulosimicrobium and Pediococcus, compared to C1. This could be linked to a possible greater influx of oxygen into C2 and C3 during the previous lambic beer production process. Compared to C1 and C2, C3 contained greater bacterial species diversity (Table 2), which was possibly due to its deformed front panel, likely resulting in a greater influx of oxygen. In addition to the three genera mentioned above, Acinetobacter, Gluconacetobacter, Gluconobacter, and Staphylococcus were abundantly found (Fig. 3), resulting in more diversity and evenness. At the PC stage, the bacterial community structure of the foeders was characterized by very limited species diversity (Table 2), also dominated by the genus Pediococcus. In contrast to the casks, no variability among the three different foeders sampled was found during the PC stage. Besides the overabundance of Pediococcus species, only Acetobacter was found in the three foeders (Fig. 3), albeit at very low relative abundances, possibly indicating a lower oxygen influx into the foeders, compared to the casks.

After cleaning of the casks and foeders, the overabundance of fermentation- and maturation-related bacteria was washed away and the relative abundances of the sequences were redistributed over the microbial community structure, resulting in increased diversity and evenness of the microbial community, except for C3 and F1, for which the redistribution of the relative abundances was very limited (Table 2). Despite the treatment with high-pressure water, Pediococcus remained prevalent in all barrels, indicating again the limited capacity of this cleaning step to remove residual bacteria from the cask and foeder inner surfaces. Species belonging to the genera Brevibacterium, Lactobacillus, Lactococcus, Leuconostoc, and Staphylococcus became more prevalent throughout the cleaning procedures. Although similar genera were detected in both the foeders and the casks after cleaning (except for Gluconacetobacter and Nocardiopsis, which were retrieved only from F2 and F3), the species diversity and evenness were much more restricted in the foeders than in the casks and Pediococcus species remained more abundant (Table 2 and Fig. 3).

As a consequence of the increased time of exposure to the environment and the low concentrations of DNA extracted from the interior cask surfaces, due to the removal of most fermentation and maturation remnants, further decreases in Pediococcus species and increases in the bacterial species diversity and evenness (Table 2) for the three casks were found after sulfuring (AS stage). The latter consisted of Arthrobacter, Brachybacterium, Brevibacterium, Brevundimonas, Chryseobacterium, Curvibacter, Enterococcus, Nocardiopsis, Pseudomonas, Sphingobacterium, Sphingomonas, Staphylococcus, and Streptomyces species (Fig. 3). Also, the relative abundances of the “others” group increased throughout the cleaning procedure for the casks.

Before the casks were filled, minor changes in the relative abundances were found but the bacterial species composition remained generally the same (Fig. 3). At the end of the entire cleaning procedure, with or without the application of a fumigation step (sulfuring), Pediococcus species and, to a lesser extent, Acetobacter species were still detected. The viability of these bacteria could not be assessed with the technique applied, but it was possible that some bacteria survived the entire cleaning procedure, in a VBNC state or not, as viable colonies could still be retrieved (albeit at counts below log 2.0 CFU/cm²) and hence could act as an additional inoculation source of fresh worts (Table 1).

Fungal species composition.

Amplicon sequencing of the fungal internal transcribed spacer 1 (ITS1) ribosomal DNA (rDNA) region was used to investigate the total fungal community structure of the inner surfaces of the wooden casks and foeders throughout the different cleaning procedures. Small differences in relative abundances of taxonomic assignments at the genus level were present when the OTU-based and ASV-based methods were compared. The OTU-based analysis generally attributed more reads to Candida, Debaryomyces, and Saccharomyces, whereas the ASV-based analysis attributed more reads to Dekkera and Pichia (Fig. 4). The latter method also attributed more reads to the “others” group. For some ASVs, taxonomic assignments to the species level were possible, although the results need to be considered with caution, due to the restricted length of the amplicon sequences (Fig. 5). In general, statistical differences (P < 0.05) in the relative abundances of OTUs and ASVs were found between the inner surfaces of the casks and foeders (Fig. 4). More specifically, the casks displayed a more diverse fungal community structure before refilling, compared to the foeders (Table 2). Dekkera and Pichia were generally encountered in samples from the casks and foeders. In the casks, mostly D. bruxellensis, Brettanomyces custersianus, and Pichia membranifaciens were encountered (Fig. 5). In contrast, in the foeders, mostly D. bruxellensis and Pichia fermentans were present. Also, the foeders displayed greater relative abundance of Saccharomyces species, consisting of both Saccharomyces cerevisiae and Saccharomyces kudriavzevii. The casks harbored greater relative abundances of Candida, Debaryomyces, and Kregervanrija species. The abundant presence of B. custersianus in the casks and not in the foeders could reflect the preference of the latter yeast species for more aerobic environments.

FIG 4.

Relative abundances of fungal OTUs (A and C) and ASVs (B and D) obtained by amplicon sequencing of all inner wooden surfaces of casks (A and B) and foeders (C and D) used for lambic beer production, sampled during their respective cleaning procedures. OTUs with occurrences of <50 were discarded; all OTUs and ASVs with occurrences of <1,000 were grouped as “others.” The OTU Dipodascaceae was also grouped under “others,” because taxonomical classification yielded only the family level (representing 2,465 sequences). In the case of the OTU-based analysis, the “others” group consisted of 13 different genera present at low abundances; in the case of the ASV-based analysis, the “others” group consisted of 23 different genera present at low abundances.

FIG 5.

Relative abundances of fungal ASVs, assigned to the species level, obtained by amplicon sequencing for all inner wooden surfaces of casks (A) and foeders (B) used for lambic beer production, sampled during their respective cleaning procedures. ASVs with occurrences of <1,000 were grouped as “others.”

During the PC stage, Dekkera and Pichia were by far the most abundant genera found in the casks and foeders (Fig. 4). Greater fungal diversity was found in C3, compared to the other two casks, which could be indicative of its status (deformed front panel, resulting in a possibly greater influx of oxygen), as this cask also showed the highest bacterial diversity before cleaning (Table 2). After cleaning of the casks, the overabundance of fermentation- and maturation-related yeasts, which consisted mostly of Dekkera species, was removed. The relative abundances of the sequences were then redistributed over the fungal community structure and more diverse fungal profiles were obtained, generally resulting in increased diversity and evenness of the microbial community (Table 2). Candida, Debaryomyces, Guehomyces, Meyerozyma, Trichosporon, and Wickerhamomyces were found, in addition to Dekkera, Kregervanrija, and Pichia (Fig. 4). The relative abundances of the “others” group increased when the overabundance of fermentation and maturation remnants was removed. Also, the filamentous fungi Penicillium and, to a lesser extent, Aspergillus were found. Their presence on the inner surfaces of the casks could be indicative of the availability of oxygen during fermentation. The OTUs and ASVs assigned to Penicillium and Aspergillus were still detected after the AS stage but, since no mold growth was found on YPG and YPGc agar, it could be assumed that these were mostly killed. The fungal taxonomic diversity remained relatively unaltered after cleaning of the foeders (Table 2). Only the relative abundances of the different genera changed slightly, with a general increase of the Saccharomyces taxon (Fig. 4). No Aspergillus or Penicillium was found in the fungal profiles, although no fumigation with sulfur dioxide was performed during the cleaning procedure for the foeders.

Apart from minor changes in relative abundances, the fungal taxonomical diversity after sulfuring (AS stage) and before filling of the casks (BF stage) remained comparable to that found after cleaning (AC stage), as shown by the alpha diversity metrics (Table 2 and Fig. 4). Again, fermentation- and maturation-relevant yeasts (namely Dekkera, Pichia, and Saccharomyces) remained present at the end of the cleaning procedure. Although only low viable yeast counts (i.e., below log 2.0 CFU/cm²) were found at the end of the cleaning procedure, it was possible that these yeasts survived the entire cleaning procedure, in a VBNC state or not, and started to grow again when the conditions became favorable (when fresh wort was added), as well as acting concurrently as an additional inoculation source.

Beta diversity.

The foeder and cask samples were statistically different (P < 0.05), based on a permutational multivariate analysis of variance (PERMANOVA), indicating differences in their microbial community structures. Also, the fungal and bacterial compositions differed statistically (P < 0.05) throughout the cleaning procedures applied. Both cleaning with high-pressure water and fumigation with sulfur dioxide caused major shifts in the microbial compositions. Pairwise PERMANOVA tests revealed that fumigation with sulfur dioxide resulted in a major shift in beta diversity of the bacterial profiles, whereas cleaning with high-pressure water caused a major shift in beta diversity of the fungal profiles. Analysis of the composition of microbiomes (ANCOM) of the different samples showed no statistically relevant data at a significance level of 0.05. At a significance level of 0.1, however, Pediococcus and Sphingomonas were the major genera responsible for the shift in the bacterial profiles after fumigation. Candida, Clavispora, Debaryomyces, Dekkera, Kregervanrija, Penicillium, Pichia, and Wickerhamomyces were the major genera responsible for the shift in the fungal profiles of the casks after cleaning (P < 0.1). Dekkera and Saccharomyces were revealed as the major genera responsible for the shift in the fungal profiles of the foeders after cleaning (P < 0.1). The scattering of samples obtained from the casks (Fig. 6) indicated a more pronounced barrel-to-barrel variation for the casks than for the foeders.

FIG 6.

PCoA biplots based on Bray-Curtis dissimilarity scores of the bacterial (A) and fungal (B) community structures of the inner surfaces of the wooden casks and foeders used for lambic beer production. Samples are categorized according to the different steps in the cleaning procedures. Foeders are circled in black.

DISCUSSION

Although the coolship step (environmental air) is considered to be the main contributor to the spontaneous inoculation of fresh wort for traditional lambic beer production, the present study showed that the wooden (port) wine barrels used for the fermentation and maturation process for lambic beer were very likely additional sources of the microbiota. First, it is well known that internal wine barrel surfaces harbor, if the wood structure allows, a heterogeneous community of microorganisms and that, for example, Dekkera species can reside and survive, even down to 8 mm deep, inside wooden wine and cider barrels (18–20). Moreover, it is well known that wooden surfaces are difficult to sanitize, due to their physical inertness and porosity (5, 8, 11). This may cause direct access of the wood inhabitants to the fresh wort when the barrels are filled during lambic beer production. Second, Dekkera anomala, D. bruxellensis, and P. damnosus are among the species that have been isolated previously from the inner surfaces of wooden casks, and they are part of the core microbiota of lambic beer production processes (10). Also, in the present study, D. anomala and D. bruxellensis, among other fermentation- and maturation-related microorganisms, were shown to cover the inner surfaces of the wooden barrels used.

The amplicon sequencing methodology used in the present study mapped the evolution of the microbial communities in different barrel types and throughout the different cleaning procedures applied in a traditional Belgian lambic brewery. It became evident that the microorganisms detected on the interior surfaces of the wooden casks and foeders mostly reflected the microbiota present in the lambic beer they contained previously. For instance, the two foeders that contained the youngest lambic beer showed the highest relative abundances of Saccharomyces species, which mostly dominate the first months of the lambic beer production process (6, 9, 10, 21). Moreover, because all of the casks and foeders sampled contained lambic beer 1 year of age or older, mostly microorganisms related to the acidification and maturation phases of the lambic beer production process were retrieved from the interior surfaces of the wooden barrels, namely, species of Dekkera, Pediococcus, and, to a lesser extent, Acetobacter (6, 9, 10, 22, 23). The presence of Acetobacter may be linked to casks of greater age and porosity, allowing greater influx of oxygen and hence growth of aerobic acetic acid bacteria (AAB) during the lambic beer production process (23). Although Pediococcus species (in particular, P. damnosus) are considered the main LAB species present during the acidification and maturation phases of lambic beer production, the detection of Lactobacillus species on the inner surfaces of the wooden barrels should be no surprise (6, 9, 10, 23).

Lambic beer barrels are often secondhand (port) wine casks, thus harboring the (port) wine microbiota, such as highly ethanol-resistant lactobacilli (e.g., Lactobacillus fructivorans and Lactobacillus hilgardii) that are considered wine spoilers (24, 25). Moreover, the genus Lactobacillus is often linked to the production of lambic beers (in particular, Lactobacillus brevis [6]) and red-brown acidic ales (26), besides its contribution to beer spoilage (17, 27). For instance, it has been shown that the LAB species Lactobacillus brevis, Lactobacillus lindneri (also linked to wine grapes), and Lactobacillus paracollinoides harbor the hop resistance genes horA and horC (25, 27). Also, Pediococcus damnosus, Pediococcus dextrinicus, Pediococcus inopinatus, and Pediococcus cellicola harbor the hop-resistant genes horA and ORF5 (28–30). Although Lactococcus and Leuconostoc are less hop resistant, they are also encountered in breweries (24). Besides cycloheximide-resistant Dekkera species, which dominate the maturation phase of the lambic beer production process (6, 9, 23), the presence of other yeast genera indicates contamination from the wooden barrels, because of the multiple uses of these barrels during wine production (8). Candida is typically associated with the wine fermentation microbiota, as is the case for Debaryomyces and Wickerhamomyces; these yeasts are typically associated with grape surfaces (16, 31, 32).

The relatively high abundances of the genera Cellulosimicrobium and Acinetobacter were rather surprising, because those species have never been shown to grow during lambic beer or ACA production processes (6, 9, 10, 33). These bacteria probably survived the harsh lambic beer fermentation and maturation conditions and resided within the wood, rather than in the fermentation medium itself. Cellulosimicrobium has a respiratory metabolism and has the capacity to hydrolyze cellulose, making porous wood an ideal environment for its growth (34). However, the possibility that the lyticase and Zymolyase enzyme preparations that were used in the DNA extraction protocol for fungal cell lysis were possible sources of these DNA sequences cannot be ruled out, as lyticase and Zymolyase are isolated typically from Cellulosimicrobium cellulans and have been found frequently upon DNA extraction and high-throughput sequencing (35; M. Verce, L. De Vuyst, and S. Weckx, unpublished results). Moreover, it is known that reagent and laboratory contamination in sequencing data occurs widely and can affect sequence-based microbiome analyses (36).

The cleaning procedures for the barrels encompassed the use of high-pressure water to remove fermentation remnants. This high-pressure water indeed partly removed the fermentation- and maturation-related bacteria and yeasts, thereby revealing more diverse microbial compositions on the interior wooden surfaces. These diverse microbial compositions included microorganisms that are normally not associated with lambic beer fermentation and maturation. The occurrence of these microorganisms on the interior barrel surfaces could be the result of years of extensive barrel usage, with the barrels going through continuous cleaning cycles. These cleaning cycles always include cleaning with high-pressure water and opening of the barrels to the environmental air. Both the water and the environmental air could be possible sources of these microorganisms and, although these microorganisms most likely do not actively participate in fermentation and maturation, they can possibly survive the lambic beer production process and be carried over from one production process to the next production process. The microorganisms that were found that are normally not associated with lambic beer fermentation and maturation were mainly aerobic bacteria and molds. Their relative abundances seemed to be linked to the barrel type, which was presumably caused by differences in wood thickness and wood porosity. The casks, which were made of thinner, more porous wood than the foeders, probably allowed substantial oxygen influx, thereby creating favorable survival conditions for aerobic microorganisms such as AAB and molds. It is indeed known that AAB prevail during the maturation phase in red-brown acidic ales produced in vertical wooden barrels (with headspace); this is minimized during lambic beer production by the use of horizontal wooden barrels that are completely filled and undergo volume adjustments during production (23). Concerning the molds found, Aspergillus and Penicillium are filamentous fungi associated with grapes and barrel surfaces (16, 17, 31). Moreover, greater porosity of the barrels (due to aging, for instance) creates better conditions for survival and biofilm formation, as highly abundant pores and cracks provide a good substrate for the formation of biofilm, which acts as a protective barrier for microorganisms present inside the wood pores (11, 37, 38). This is further reflected by the presence of fermentation-unrelated bacteria, by the retrieval of still quantifiable viable counts from the casks but not from the foeders after cleaning, and by the finding that the most damaged and deformed cask also harbored the greatest microbial diversities. This finding emphasizes that the type and status of the casks could have an influence on the distribution of the relative abundances of the microbial community structures on their inner surfaces, although this needs additional research. Moreover, the possibility that, due to the low microbial biomass of the samples after cleaning, reagent and laboratory contamination could have a substantial impact on the microbial compositions determined cannot be ruled out (36).

Because the time between emptying and filling of the casks spans several weeks, a sulfuring step is part of the cleaning procedure, mainly to prevent the outgrowth of molds (39). In contrast, during wine production, fumigation is performed to prevent spoilage by Dekkera yeasts. Although sulfuring killed major parts of the microbiota present, dead cells and remaining DNA were not removed, explaining their detection with amplicon sequencing. Moreover, it has been shown that Dekkera yeasts, for example, can be very resistant to sulfur dioxide treatments, entering a VBNC state (8, 40–42). This state is reversible, which enables these yeasts to grow again when conditions become favorable, such as when fresh wort is added (10, 40, 41, 43). Furthermore, some Dekkera species display biofilm formation, which enhances their resistance to chemical cleaning agents and sanitizers (44–46). Additionally, relative enrichment of membrane-related genes occurs in D. bruxellensis, compared to the genomes of Pichia angusta, Pichia pastoris, and S. cerevisiae, making its adhesion to the wood in the interior of the lambic barrels possible and protecting the cells from being washed out during the high-pressure cleaning procedures applied (42, 47). Based on these findings and the detection of fermentation- and maturation-related microorganisms at the end of the cleaning procedures applied, it is hypothesized that some of the microorganisms survived the entire cleaning procedure, in a VBNC state or not for many of these cells and/or engrained deeper in the wood and protected from the full dose of sulfur dioxide. Therefore, these cells could act as an additional inoculation source, besides brewery air and brewery equipment. Further research should be conducted to confirm the presence of viable microorganisms, as the methodology applied in this study did not allow assessment of the presence of microbial cells engrained deeper in the wood and/or occurring in a VBNC state.

In conclusion, it became evident that, at the end of the cleaning procedure applied for lambic beer barrels in a traditional lambic brewery in Belgium, the microbial compositions were more diverse, although fermentation- and maturation-related microorganisms could still be detected. Microorganisms surviving the cleaning procedures could help to establish a stable microbial community in the wort, to diminish batch-to-batch variations in fermentation profiles, while outcompeting undesirable microorganisms. The barrel type (age, wood thickness, and wood porosity) seemed to be a major factor that determined the composition of the microbial communities present on the inner wooden surfaces, although this needs additional research. The condition of the barrels also seemed to have a strong influence on the compositions of the microbial communities found on the inner surfaces, thus most likely affecting the fermentation and final quality of the matured lambic beer. Sulfuring also helped to control mold growth, as no viable molds could be detected after this cleaning step.

MATERIALS AND METHODS

Lambic beer barrels and their cleaning procedures.

The interior surfaces of six wooden barrels used in the production of lambic beer were sampled in an artisan lambic brewery located in the Senne river valley, southwest of Brussels, Belgium. The barrels consisted of three identical oak casks (C1, C2, and C3, the latter with a deformed front panel) (Fig. 1) and three foeders (F1, F2, and F3) (Fig. 1). All barrels were used continuously in the brewery and contained maturing lambic beer prior to their emptying. Both the casks (Fig. 2A) and the foeders (Fig. 2B) were cleaned before being filled with a new batch of wort. When the casks were emptied, they were first cleaned with hot water under high pressure. The casks were allowed to dry and then fumigated with sulfur sticks to inhibit mold growth. Finally, the casks were stored at ambient temperature for several weeks, until they were filled with fresh wort. The foeders underwent a simpler cleaning procedure. After emptying, they were cleaned with hot water under high pressure and immediately refilled the next day with fresh wort to start a new fermentation batch. Samples for the PC and AC stages were taken on the same day (for both the casks and the foeders). Samples for the AS stage were taken a few days later, followed by samples for the BF stage, which were taken several weeks later (relative to the time point for the PC and AC stages).

Sampling of the inner wooden surfaces of lambic beer barrels.

The interior wooden surfaces of the three casks and three foeders were sampled in the middle of the barrels after emptying (PC stage), after cleaning (AC stage), after sulfuring (AS stage), and just before filling with fresh wort (BF). The casks C1, C2, and C3 contained lambic beer, the production of which was started in November 2014; they were sampled in October and November 2016 (representing matured 2-year-old lambic beer). The foeders F1, F2, and F3 contained lambic beer, the production of which was started in November 2014, March 2015, and January 2016, respectively; they were also sampled in October and November 2016 (representing matured 2-year-old, 1.5-year-old, and 1-year-old lambic beers). Samples from the casks were taken through the bunghole (closed with a loose panel) (Fig. 1). In the case of the foeders, a manhole located at the bottom in the front panel was used for sampling (Fig. 1).

For sampling, sterile swabs (Sterilux ES3 compress; Hartmann, Heidenheim an der Brenz, Germany) were used under aseptic conditions by applying flame sterilization. The sterile swabs were mounted onto a circular flame-sterilized surface of fixed size (approximately 20 cm²), allowing the swabbing, by rotation, of a fixed area (20 cm²; different areas per swabbing time) on the side panel of the barrels at a middle height. After swabbing, the sterile swab samples were immediately transferred, in the proximity of a flame, into sterile plastic tubes containing 20 ml of phosphate-buffered saline (PBS) (137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, 2 mM KH₂PO₄ [pH 7.4]). All samples were put on ice for transportation to the laboratory and were processed the same day. Immediately after transportation, samples were vortex-mixed for 3 min before the swabs were aseptically removed from the PBS solution. Portions of the resulting PBS solutions were then analyzed through culture-dependent plating, while the remaining portions of the PBS solutions were centrifuged (4,700 × g for 20 min at 4°C); the cell pellets obtained were stored at −32°C for culture-independent microbiological composition analysis.

Culture-dependent microbiological analyses. (i) Selective plating and culturing.

The chilled samples were serially diluted in PBS solution, and 100 μl of each dilution was plated onto multiple agar media, as follows: (i) PCA (Oxoid, Basingstoke, Hampshire, UK) for the enumeration of the total microbial counts after aerobic incubation at 30°C for 7 days; (ii) PCAc, i.e., PCA supplemented with 200 ppm of cycloheximide (Sigma-Aldrich, St. Louis, MO) and 5 ppm of amphotericin B (Sigma-Aldrich) to inhibit fungal growth, for the enumeration of the total bacterial counts; (iii) YPG agar (2.0% glucose [Merck, Darmstadt, Germany], 0.5% yeast extract [Merck], 1.0% peptone [Oxoid], and 1.5% agar [Oxoid] [wt/vol]), supplemented with 200 ppm of chloramphenicol (Sigma-Aldrich) to inhibit bacterial growth, for the enumeration of presumptive yeasts after aerobic incubation at 30°C for 7 days; and (iv) YPGc agar, i.e., YPG agar supplemented with 50 ppm of cycloheximide, for selection of cycloheximide-resistant yeasts after aerobic incubation at 30°C for 7 days.

(ii) Enumeration of colonies.

Agar media containing 30 to 300 CFU were counted along the cleaning procedures. The viable counts were expressed as log CFU per square centimeter of swabbed surface (1 cm² corresponded to 1 ml of sample solution).

Culture-independent microbiological analyses. (i) Total DNA extraction.

To enable culture-independent microbial species diversity analysis, total DNA extracts were prepared from the cell pellets obtained above, following a method combining enzymatic, chemical, and mechanical treatments for cell lysis and phenol-chloroform/isoamyl alcohol treatment for DNA extraction, and column chromatography using the DNeasy Blood and Tissue kit (Qiagen, Venlo, The Netherlands) was applied to purify the DNA obtained, as described previously (48).

(ii) Amplification of bacterial and fungal group-specific loci.

Amplification of group-specific loci was performed as described previously (48). Specific loci of bacterial and fungal marker genes were amplified by PCR with primers purchased from Integrated DNA Technologies (Leuven, Belgium). The hypervariable V4 region of the bacterial 16S rRNA gene was amplified with the primers F515 (5′- TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGGTGTGCCAGCMGCCGCGGTAA-3′) and R806 (5′- GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGGGACTACHVGGGTWTCTAAT-3′) (49), both with a specific Illumina platform 5′ tag (underlined). The thermal PCR program was run under the following conditions: denaturation at 94°C for 3 min, 35 amplification cycles of 94°C for 45 s (denaturation), 50°C for 60 s (annealing), and 72°C for 90 s (extension), and final extension at 72°C for 10 min. The fungal ITS1 region of the rDNA was amplified with the primers BITS (5′- TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGACCTGCGGARGGATCA-3′) and B58S3 (5′- GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGGAGATCCRTTGYTRAAAGTT-3′) (50), both with a specific Illumina platform 5′ tag (underlined). The thermal PCR program was run under the following conditions: denaturation at 95°C for 2 min, 40 amplification cycles of 95°C for 30 s (denaturation), 50°C for 30 s (annealing), and 72°C for 60 s (extension), and final extension at 72°C for 5 min. A commercially available bacterial mock community (HM-782D; BEI Resources, Manassas, VA) was amplified simultaneously with all samples, for subsequent detection of sequencing errors. PCR sample mixtures were prepared and checked for the appropriate lengths as described previously (48).

(iii) High-throughput sequencing of the V4 and ITS1 amplicons.

High-throughput sequencing of the V4 and ITS1 amplicons was performed as described previously (48). The PCR amplicons obtained were purified using the Wizard SV gel and PCR cleanup system (Promega, Madison, WI), following the manufacturer’s instructions, after which they were eluted in 30 μl of nuclease-free water. Subsequently, they were subjected to size selection using Agencourt AMPure XP PCR purification magnetic beads (Beckman Coulter, Brea, CA), according to the manufacturer’s instructions except that the relative amount of bead solution was changed to 1.1× for the BITS1/B58S3 primer set, Eppendorf tubes were left to dry in the open air for a maximum of 3 min, and elution was performed with 30 μl of nuclease-free water. The quality and concentrations of the amplicons were validated by means of a 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA) and a fluorometric Qubit 2.0 assay (Thermo Fisher, Waltham, MA), as described previously (48). The bacterial mock community also was purified and size selected, following the protocol used for the bacterial V4 sequences mentioned above. Next, the bacterial V4 and fungal ITS1 amplicons originating from the same sample were pooled in a 2:1 molar ratio and barcoded with the same index prior to sequencing. All samples were paired-end sequenced, together with the amplified bacterial mock community, using the Illumina MiSeq platform (Illumina, San Diego, CA) at the Vrije Universiteit Brussel-Université Libre de Bruxelles sequencing facility (BRIGHTcore, Jette, Belgium). Two Fastq files were obtained for each sample, containing all forward and reverse reads from both bacterial and fungal amplicons.

(iv) Bioinformatic analysis.

The two Fastq files for each sample (forward and reverse sequences), containing both bacterial V4 and fungal ITS1 sequences, were split into an additional two files, containing only forward and reverse sequences of the V4 and ITS1 regions, as described previously (48). To circumvent software-based biases, analysis pipelines based on both OTUs (mothur) and ASVs (DADA2) were used.

For OTU-based analysis, different workflows were followed to process the bacterial and fungal sequences. For the bacterial V4 sequences (291 bp) for all samples, including the mock community, primers were removed using an in-house script (reverse primer for the forward sequences and forward primer for the reverse sequences) before mothur v1.36.1 software was used (51). After generation of the contigs, the unique sequences were clustered into groups based on a maximum of two mismatches. Chimeric sequences were removed with the UCHIME algorithm (52). The most abundant sequence of each group was chosen as the representative one and was taxonomically assigned by alignment against the bacterial 16S rRNA SILVA database (release 132) (53), to remove nonbacterial reads. The representative unique sequences that were assigned the same “order” of taxon were then clustered together. Within each cluster, multiple alignment analysis was performed to group the sequences into OTUs at a level of 97% identity. The sequence error rate was estimated in mothur by comparing the sequenced mock community against the in silico mock community (54). For the fungal ITS1 sequences, which could vary in length, the forward and reverse sequences were trimmed using Cutadapt software (55), to avoid possible adapter read-throughs of ITS1 sequences shorter than the 300-bp chemistry used by the Illumina platform, by trimming of the sequenced adapters and the overhangs at the 3′ end (56). The trimmed files were then quality screened and further processed using mothur software to generate contigs, groups of which were compared to the fungal UNITE_IT1 database (v6_sh_99) (57), as described previously (48).

The same trimmed files used for the OTU-based analysis were processed using the DADA2 package (58) in RStudio (59), following the DADA2 standard operation protocol. Briefly, sequences with at least one ambiguous base, reads containing the lowest possible quality score of 2, and reads with more than two “expected errors” were discarded. The reads were truncated before sample inference (the DADA2 core algorithm) was performed. Finally, the reads were merged, chimeras were removed, taxonomy was assigned using RDP v14 (60), and the resulting sequence table containing all ASVs was constructed.

Statistics.

Intrasample diversity (alpha diversity) was assessed by calculating the Simpson (diversity) and Pielou (evenness) indexes, for both bioinformatic analysis methods applied (OTUs and ASVs). Intersample diversity (beta diversity) to determine differences in bacterial and fungal compositions among the barrel types and cleaning steps was assessed by conducting PERMANOVA based on Bray-Curtis dissimilarity scores, again for both bioinformatic analysis methods applied. This analysis was followed by a series of pairwise PERMANOVA comparisons to assess differences among the PC, AC, AS, and BF cleaning stages. A significance level of 0.05 was considered for all statistical procedures. All statistical analyses and tests performed were executed through the SPSS v20 package (IBM, Chicago, IL). Differences in relative abundances of individual taxa throughout the cleaning procedures applied for the different barrels were tested with the software package ANCOM (61) in RStudio (59). ANCOM was run with the default settings, at a significance level of 0.05.

Finally, a principal-coordinate analysis (PCoA) was performed on the OTU-based data, enabling a visual interpretation of the beta diversity in a low-dimensional Euclidian space (62). The underlying trends in OTU diversity and community structure at different time points throughout the cleaning procedures, as well as according to the barrel type, were revealed. Before the calculation of the principal coordinates, the bacterial 16S rRNA gene data were rarefied at a level of 20,000 reads per sample (after which sample C3-BF was removed, because it contained fewer reads) and the fungal ITS1 data were rarefied at a level of 10,000 reads per sample (after which sample C3-AC was removed, because it contained fewer reads). PCoA plots were then calculated based on Bray-Curtis dissimilarity scores, whereby the magnitude of each axis displayed how much variation was summarized within that axis, using the statistical software package Vegan in RStudio (59, 63).

Accession number(s).

Sequencing data were deposited in the European Nucleotide Archive (ENA) of the European Bioinformatics Institute (EBI) under the accession number PRJEB28362 and is available at http://www.ebi.ac.uk/ena/data/view/PRJEB28362.