You are on page 1of 17

International Journal of Food Microbiology 394 (2023) 110163

Contents lists available at ScienceDirect

International Journal of Food Microbiology


journal homepage: www.elsevier.com/locate/ijfoodmicro

New insights into the role of key microorganisms and wooden barrels
during lambic beer fermentation and maturation
Louise Vermote a, 1, Jonas De Roos a, 1, Margo Cnockaert b, Peter Vandamme b, Stefan Weckx a,
Luc De Vuyst a, *
a
Research Group of Industrial Microbiology and Food Biotechnology (IMDO), Department of Bioengineering Sciences, Vrije Universiteit Brussel, Pleinlaan 2, B-1050
Brussels, Belgium
b
Laboratory for Microbiology, Department of Biochemistry and Microbiology, Ghent University, K.L. Ledganckstraat 35, B-9000 Ghent, Belgium

A R T I C L E I N F O A B S T R A C T

Keywords: Belgian lambic beers are still produced through traditional craftsmanship. They rely on a spontaneous fermen­
Lambic beer tation and maturation process that is entirely carried out in wooden barrels. The latter are used repetitively and
Spontaneous fermentation may introduce some batch-to-batch variability. The present systematic and multiphasic study dealt with two
Wooden barrels
parallel lambic beer productions carried out in nearly identical wooden barrels making use of the same cooled
Shotgun metagenomic sequencing
wort. It encompassed a microbiological and metabolomic approach. Further, a taxonomic classification and
Metagenome-assembled genomes
Functional analysis metagenome-assembled genome (MAG) investigation was based on shotgun metagenomics. These investigations
Acetobacter lambici provided new insights into the role of these wooden barrels and key microorganisms for this process. Indeed,
besides their role in traditionality, the wooden barrels likely helped in establishing the stable microbial
ecosystem of lambic beer fermentation and maturation by acting as an inoculation source of the necessary mi­
croorganisms, thereby minimizing batch-to-batch variations. They further provided a microaerobic environment,
which aided in achieving the desirable succession of the different microbial communities for a successful lambic
beer production process. Moreover, these conditions prevented excessive growth of acetic acid bacteria and,
therefore, uncontrolled production of acetic acid and acetoin, which may lead to flavor deviations in lambic beer.
Concerning the role of less studied key microorganisms for lambic beer production, it was shown that the
Acetobacter lambici MAG contained several acid tolerance mechanisms toward the harsh environment of maturing
lambic beer, whereas genes related to sucrose and maltose/maltooligosaccharide consumption and the glyox­
ylate shunt were absent. Further, a Pediococcus damnosus MAG possessed a gene encoding ferulic acid decar­
boxylase, possibly contributing to 4-vinyl compound production, as well as several genes, likely plasmid-based,
related to hop resistance and biogenic amine production. Finally, contigs related to Dekkera bruxellensis and
Brettanomyces custersianus did not possess genes involved in glycerol production, emphasizing the need for
alternative external electron acceptors for redox balancing.

1. Introduction 2021; Bouchez and De Vuyst, 2022; De Roos and De Vuyst, 2019;
Pothakos et al., 2016; Spitaels et al., 2017; Verachtert and Derdelinckx,
Beer productions carried out on an industrial scale generally make 2005). These Belgian acidic beers get increasingly popular worldwide as
use of yeast starter cultures to produce stable and consistent end- trademarks of traditional craftsmanship. They possess a unique flavor
products (Bokulich and Bamforth, 2013). However, certain beers are profile, which is composed of lactic acid, acetic acid, ethyl esters (in
still produced through a spontaneous fermentation and maturation particular ethyl lactate and ethyl acetate), acetoin, and phenolic com­
process, such as particular acidic ales. Belgium is particularly known for pounds (in particular 4-ethylphenol and 4-ethylguaiacol) that are
its traditional lambic and lambic-based beers, which are produced important for the fresh acidic, sharp acidic, fruity, buttery, and Brett-
without the intended inoculation of yeasts or bacteria (Bongaerts et al., flavor notes, respectively (Bongaerts et al., 2021; Bouchez and De

* Corresponding author.
E-mail address: luc.de.vuyst@vub.be (L. De Vuyst).
1
Equal contribution

https://doi.org/10.1016/j.ijfoodmicro.2023.110163
Received 6 October 2022; Received in revised form 24 February 2023; Accepted 26 February 2023
Available online 6 March 2023
0168-1605/© 2023 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-
nc-nd/4.0/).
L. Vermote et al. International Journal of Food Microbiology 394 (2023) 110163

Vuyst, 2022; De Keersmaecker, 1996; Dysvik et al., 2020; Pothakos and maturation process of lambic beer wort that was carried out in a
et al., 2016; Steensels et al., 2015). single cask (De Roos et al., 2018a). This study targeted both bacteria and
Traditional Belgian lambic beers are produced through the sponta­ fungi and allowed microbial identifications at genus level. In addition, a
neous fermentation of an aqueous mixture of barley malt, unmalted shotgun metagenomic analysis of the same fermentation and maturation
wheat, and aged dry hops in horizontal wooden barrels (Bongaerts et al., process has allowed mapping of the microbial diversity at species level.
2021; De Keersmaecker, 1996; De Roos and De Vuyst, 2019; Spitaels This further indicated the functional potential of the microorganisms
et al., 2017). This fermentation and maturation process can proceed up found, allowing to link substrate consumption and metabolite produc­
to three years and is characterized by a microbial succession, acting tion to specific microbial groups or species during lambic beer produc­
during four different phases and comprising different yeast and bacterial tion (De Roos et al., 2020). Whereas these two studies aimed to use state-
species, in particular Saccharomyces cerevisiae, Saccharomyces kudriav­ of-the-art DNA-based methodologies to investigate the microbiota
zevii, Acetobacter lambici, Pediococcus damnosus, Dekkera bruxellensis, involved in the fermentation and maturation of lambic beer wort in a
and/or Brettanomyces custersianus (Bongaerts et al., 2021; De Roos et al., single wooden barrel, it is unclear which variability the use of different
2018a, 2018b; Spitaels et al., 2017; Van Oevelen et al., 1977; Verachtert wooden barrels of the same type add to lambic beer production pro­
and Iserentant, 1995). The lambic beer wort that was first boiled is cesses originating from the same coolship batch.
cooled down in an open coolship overnight, which allows its inoculation The present study aimed at a further determination of the influence
with microorganisms present in the surrounding air. This is followed by of individual wooden barrels on lambic beer production processes.
fermentation and maturation of the cooled wort in wooden barrels Therefore, two parallel lambic beer productions using nearly identical
(casks or foeders), which further serve the traditionality of the process wooden barrels of wine origin were carried out in a Belgian traditional
(Bongaerts et al., 2021; De Roos et al., 2019). Wooden casks are involved lambic brewery and were started using wort from the same coolship
in the maturation of several alcoholic beverages, such as (port)wine, batch. These productions were monitored temporally by means of a
whisky and cider, to allow some oxygen ingress as well as flavor for­ systematic and multiphasic analysis approach. This consisted of a high-
mation by extracting typical wood compounds, such as polyphenols and throughput microbiological analysis, in casu molecular identification of
tannins (Canas et al., 2016; del Alamo-Sanza and Nevares, 2014; Del isolates picked from selective agar media, and shotgun metagenomics of
Campo et al., 2003; Guzzon et al., 2011; White et al., 2017). However, it whole-community DNA, in combination with an extended metabolite
remains largely unknown why wooden barrels are traditionally used target analysis. This study further examined the genetic potential of
during lambic beer production (Spitaels et al., 2017). Moreover, wooden A. lambici, P. damnosus, D. bruxellensis, and B. custersianus.
surfaces display a microbial risk, as they are difficult to sanitize due to
their physical inertness and porosity. Alternatively, it has been shown 2. Materials and methods
that the resident microbiota, present on the interior barrel surfaces,
plays an inoculation and functional role during spontaneous lambic beer 2.1. Lambic beer production and sampling
productions (Bongaerts et al., 2021; De Roos et al., 2019; Oelofse et al.,
2008; Spitaels et al., 2017). For instance, P. damnosus, D. bruxellensis, In April 2015, two lambic beer production processes were started in a
and Dekkera anomala, which are considered as key microorganisms for Belgian, traditional lambic brewery, located in the Senne river valley
successful lambic beer productions, have been isolated from the interior southwest of Brussels. Lambic beer wort of 12.5◦ P was prepared ac­
surfaces of wooden lambic beer barrels and studied intensively (Bon­ cording to the brewer’s recipe (consisting of water, barley malt,
gaerts et al., 2021; Bouchez and De Vuyst, 2022; Spitaels et al., 2015). unmalted wheat, and overaged hop), manually acidified with lactic acid
The role of air-borne acetic acid bacteria (AAB), such as Acetobacter after wort boiling, and cooled overnight in a coolship open to the sur­
lambici, is less studied. This AAB species was originally isolated from a rounding air. The wort was then chilled to 4 ◦ C and transferred to two
maturing lambic beer wort and has not been found in other niches up to nearly identical 620-L oak casks of the same type (further referred to as
now (Spitaels et al., 2014a, 2014b). Although the whole-genome casks 1 and 2) for fermentation and maturation, hence representing
sequence of A. lambici has been published (Sombolestani et al., 2020), biological duplicates. The oak casks originated from a wine producer
no in-depth investigation of its genetic potential was performed so far. and were already used several times to produce lambic beer in the
Further, high-throughput, partial 16S rRNA gene amplicon brewery. Before their use, the oak casks were cleaned and treated as
sequencing of whole-community DNA from swab samples of the inside described previously (De Roos et al., 2018b). The casks were not closed
wooden surfaces of lambic beer barrels after extensive barrel cleaning airtightly and were located next to each other in a cellar at ambient
shows the presence of fermentation- and maturation-related microor­ temperature (ranging from 12 ◦ C to 21 ◦ C due to seasonal effects). These
ganisms (De Roos et al., 2019). Since the same wooden barrels are used two parallel lambic beer productions were monitored as a function of
several times for lambic beer batch productions and are only cleaned time by withdrawing samples of 100 mL or 200 mL (metagenomic
superficially with high-pressurized water in between consecutive pro­ samples) from the cooled wort before its transfer to the wooden casks,
duction batches, these fermentation- and maturation-related microor­ further referred to as the time point 0 h, and from the fermenting wort
ganisms presumably originate from previous lambic beer productions and maturing beer after 1 h; 1 and 3 days; 1, 2, 6, 9, and 16 weeks; and 6,
(De Roos et al., 2019; Spitaels et al., 2015). Besides cleaning with high- 12 (time point with volumetric adjustment of the liquid levels in the
pressurized water, sulphur dioxide is commonly used as a sanitation casks), 18, 24, and 30 months. Aseptic sampling was performed through
technique for the barrels (Barata et al., 2013; De Roos et al., 2019; the cork-plugged bunghole of the wooden casks at the middle of the
Guzzon et al., 2011; Oelofse et al., 2008). As it is known that some yeast fermenting wort and maturing beer, as described previously (De Roos
species enter a reversible, viable but non-culturable (VBNC) state after et al., 2018b). Part of the samples were chilled; another part was frozen
sulphur dioxide treatments, it is likely that some of these microorgan­ at − 20 ◦ C. The chilled samples were further handled for microbiological
isms survive the entire cleaning procedure, whether or not in a VBNC plating and to obtain cell-free supernatants for metabolite target anal­
state, and act as an additional inoculation source, besides the house ysis, as described previously (De Roos et al., 2018b). The frozen samples
microbiota of the brewery air and brewery equipment (Bongaerts et al., served for metagenomic DNA extraction.
2021; Bokulich et al., 2012, 2015; De Roos et al., 2019; Spitaels et al.,
2015). 2.2. Determination of temperature, pH, and apparent attenuation
A straightforward procedure to assess all microorganisms present,
including VBNC ones, encompasses culture-independent, DNA-based The temperature, pH, and apparent attenuation were measured on
analyses. Until now, only one study applied amplicon-based high- site, immediately after sampling, as described previously (De Roos et al.,
throughput sequencing of whole-community DNA during a fermentation 2018b).

2
L. Vermote et al. International Journal of Food Microbiology 394 (2023) 110163

2.3. Analysis of the microbial community dynamics and species diversity (Vermote et al., 2018). The DNA fragment libraries were sequenced
using an Ion Torrent Personal Genome Machine (PGM) with an Ion PGM
2.3.1. Culture-dependent analysis Hi-Q View Sequencing Kit (Thermo Fisher Scientific). For each of the 15
samples sequenced, one Ion 316 Chip v2 BC (Thermo Fisher Scientific)
2.3.1.1. Selective plating and culturing for colony enumeration and iso­ was used, following the manufacturer’s instructions, which resulted in
lation. Four selective agar media supplemented with the appropriate 15 data sets. All 15 metagenomic data sets are accessible at the European
antibiotics were used for colony enumeration. This was based on dilu­ Nucleotide Archive of the European Bioinformatics Institute (ENA/EBI;
tion series of the original samples. The counts were expressed as log of Hixton, UK) under the study accession number PRJEB53155.
colony-forming units (CFU)/mL. The agar media also served for colony
isolation (for subsequent identification) of the presumptive lactic acid 2.3.2.3. Bioinformatics analysis
bacteria (LAB; modified de Man-Rogosa-Sharpe or mMRS agar medium), 2.3.2.3.1. Metagenomic sequence reads quality trimming. The quality
AAB (modified deoxycholate-mannitol-sorbitol or mDMS agar medium), of the metagenomic sequence reads obtained was assessed using FastQC
yeasts (yeast extract-peptone-glucose or YPG agar medium), and (version 0.10.1; Andrews, 2018), followed by a quality trimming step
cycloheximide-resistant yeasts (YPG agar medium supplemented with with prinseq-lite (version 0.20.2; Schmieder and Edwards, 2011), as
cycloheximide or YPGc agar medium), as described previously (Papal­ described previously (Vermote et al., 2018). This resulted in a set of
exandratou et al., 2013; De Roos et al., 2018a, 2018b). Additionally, 50,819,736 quality-trimmed metagenomic sequence reads (MSRs), ac­
100 μL of each dilution was plated on violet red-bile-glucose (VRBG) counting for 11.7 Gbp, ranging from 3,030,740 to 3,864,756 MSRs and
agar medium (Oxoid, Basingstoke, Hampshire, UK), supplemented with from 641.4 to 1016.9 Mbp per sample. All bioinformatic analysis steps
5 ppm of amphotericin B and 200 ppm of cycloheximide, for the described below were performed on all these MSR data sets using the
enumeration of presumptive enterobacteria during the first days of same databases.
fermentation after aerobic incubation of the plates at 30 ◦ C for 7 days. 2.3.2.3.2. Taxonomic classification at genus level. The taxonomic
composition of all samples was assessed in a software- and database-
2.3.1.2. Dereplication and identification of microbial isolates by MALDI- independent manner (Illeghems et al., 2012), using four classification
TOF MS. The microbial isolates were sub-cultivated twice on their tools, namely blastn (Altschul et al., 1990), DIAMOND (Buchfink et al.,
respective agar media for dereplication and identification purposes. The 2015), Kaiju (Menzel et al., 2016), and Kraken 2 (Wood et al., 2019), as
resulting third-generation colonies were cultivated in their respective described previously (Verce et al., 2021). The results of Kraken 2 will be
liquid media, stored at − 80 ◦ C, and further used for matrix-assisted laser further referred to as Kraken 2-nt for the classification at nucleotide
desorption ionization-time of flight mass spectrometry (MALDI-TOF MS) level and Kraken 2-aa for the classification at amino acid level. All an­
fingerprinting, as described previously (De Roos et al., 2018b). Minor alyses were performed with databases constructed or downloaded
modifications included that the final cell pellets were resuspended in 40 before the reclassification of the Lactobacillus genus (Zheng et al., 2020).
μL of 70 % (v/v) formic acid (Merck, Darmstadt, Germany) and vortexed 2.3.2.3.3. Taxonomic classification at species level based on fragment
thoroughly. Also, 40 μL of acetonitrile (Merck) was added and the recruitment plotting. Fragment recruitment plotting (FRP) was per­
contents were mixed. After centrifugation (21,000 ×g, 3 min, 4 ◦ C), 1 μL formed as described previously (Verce et al., 2019), using an in-house
of these solutions was spotted in duplicate on a MBT Biotarget 96 database containing a representative genome sequence (if available) of
stainless steel plate (Bruker, Billerica, MA, USA) and overlaid with 1 μL each species of the genera detected in at least one of the samples with at
of matrix solution, consisting of 5 mg/mL of α-cyano-4-hydroxycin­ least one taxonomic classification method and with a minimum relative
namic acid in water:acetonitrile:trifluoroacetic acid (48:50:2). The mass abundance of 0.1 %. This in-house database was extended with a
spectra were measured by means of a Microflex MALDI-TOF mass representative genome sequence (if available) of species of the genera
spectrometer (Bruker) using built-in software, and the spectra of indi­ found with the culture-dependent analysis and the genera detected in a
vidual isolates were dereplicated into clusters using BioNumerics previous metagenomics-based study on a lambic beer production from
(version 7.5; Applied Maths, Sint-Martens-Latem, Belgium). Represen­ the same brewery (De Roos et al., 2020).
tative isolates from each cluster were identified through comparative 2.3.2.3.4. Statistical analysis. Rstudio software (Rstudio Team,
gene sequence analysis, as described previously (De Roos et al., 2018a). 2020) was used for all statistical analyses, using the tidyverse (Wickham
et al., 2019), ggrepel (Slowikowski, 2020), vegan (Oksanen et al., 2019),
2.3.2. Culture-independent analysis and RVAideMemoire (Hervé, 2021) packages. A data set based on the
relative abundance of species obtained by FRP was used for the statis­
2.3.2.1. Metagenomic DNA extraction. Cell pellets obtained from the tical analyses. Principal component analysis (PCA) was performed to
samples taken at 0 h, 3 days; 1, 6, 9, and 16 weeks; and 18 and 30 identify patterns between the wort sample before transfer to the casks
months were used for metagenomic DNA extraction, which was based on and the time-dependent fermenting wort/maturing beer samples of cask
enzymatic lysis, chemical treatment and mechanical disruption, as 1 and cask 2. Intra-sample diversity (alpha-diversity) was assessed by
described previously (Vermote et al., 2018), with some modifications. calculating the Simpson (diversity) and Pielou (evenness) indices
After the lysis steps, a phenol/chloroform/isoamyl alcohol (Merck) pu­ (Mouillot and Leprêtre, 1999) and inter-sample diversity (beta-di­
rification was applied, followed by an RNase A treatment on the aqueous versity) was assessed to determine the differences between casks 1 and 2
phase (55 U; Thermo Fisher Scientific, Waltham, MA, USA) at 37 ◦ C for by permutational multivariate analysis of variance (PERMANOVA),
10 min. The DNA was further purified using a DNeasy Blood and Tissue based on Bray-Curtis dissimilarity scores (Anderson, 2017).
Kit (Qiagen, Hilden, Germany), following the manufacturer’s in­ 2.3.2.3.5. Co-assembly of metagenomes, contig binning, and contig
structions and applying an elution with 100 μL of nuclease-free water annotation. The 15 MSR data sets were co-assembled using MEGAHIT
(VWR International, Darmstadt, Germany). The DNA purity was (version 1.2.9; Li et al., 2015) with a minimum contig length set at 1000
assessed with a NanoDrop 2000 spectrophotometer and the concentra­ bp. The resulting set of contigs was imported into anvi’o (version 6.2;
tion was measured with a Qubit 2.0 fluorometer using a Qubit dsDNA HS Eren et al., 2015), only considering contigs of at least 2500 bp. Each
Assay Kit (all from Thermo Fisher Scientific). MSR data set was mapped to those contigs using Bowtie 2 to build up a
profile database (version 2.3.5.1; Langmead and Salzberg, 2012). The
2.3.2.2. Shotgun metagenomic sequencing. The metagenomic DNA taxonomic classification of the contigs was determined by classifying the
extracted was used for the construction of DNA fragment libraries, with predicted genes from anvi’o using Kaiju by means of the database
an intended fragment length of 350 bp, as described previously mentioned above used for taxonomic analysis. The contigs were binned

3
L. Vermote et al. International Journal of Food Microbiology 394 (2023) 110163

using CONCOCT (version 1.1.0; Alneberg et al., 2014). The bins ob­ methyl-1-butanol, as described previously (De Roos et al., 2018a,
tained were imported into anvi’o and manually refined based on the 2018b). Therefore, a Focus gas chromatograph (Interscience, Breda, The
sequence composition and differences in coverage, reported complete­ Netherlands), equipped with a Stabilwax-DA column (Restek, Belle­
ness, and contamination statistics. After refinement, only bins with a fonte, PA, USA) coupled to a FID-80 detector (Interscience), was used.
high completion (> 75 %) and low contamination level (< 10 %) were
considered and the taxonomy assigned was checked with minimap2 2.4.5. Determination of other organic acid concentrations
(version 2.11; Li, 2018), using the database used for FRP. Ultra-performance liquid chromatography with tandem mass spec­
The most complete metagenome-assembled genome (MAG), namely trometric detection (UPLC-MS/MS), applying external calibration, was
that of A. lambici, was compared with the genome of A. lambici LMG applied to determine the concentrations of citric acid, gluconic acid,
27439T [National Center for Biotechnology Information (NCBI) acces­ malic acid, and succinic acid, as described previously (De Roos et al.,
sion number GCF_011516885.1; Sombolestani et al., 2020] to calculate 2018a, 2018b). Therefore, an Acquity system chromatograph (Waters),
the average nucleotide identity (ANI) value using OrthoANI (Lee et al., equipped with a HSS T3 column (Waters) and coupled to a triple
2016) with the default settings. Contigs belonging to bins assigned to quadrupole (TQ) tandem mass spectrometer with a Zspray™ electro­
bacterial genera were annotated with Prokka (version 1.14.6; Seemann, spray ionization source in negative ionization mode (Waters), was used.
2014), using the options “metagenome” and “adding gene features”, and
with an e-value cut-off of 1 × 10− 20. Contigs belonging to bins assigned 2.4.6. Determination of biogenic amine concentrations
to fungal genera were annotated with Maker2 (version 2.31.10; Holt and UPLC-MS/MS, using multiple reaction monitoring and internal
Yandell, 2011), as described previously (Verce et al., 2021). A targeted standardization, was applied to determine the concentrations of agma­
search among the high completion bins and bins derived thereof during tine, cadaverine, histamine, 2-phenylethylamine, putrescine, trypt­
a manual refinement process was performed to find metabolic pathways amine, and tyramine, as described previously (De Roos et al., 2018a).
or genes relevant for lambic beer production. If a particular gene was not Therefore, an Acquity chromatograph, equipped with a HSS T3 column
found, a sequence alignment was performed manually, using an avail­ coupled to a TQ tandem mass spectrometer with a Zspray™ electrospray
able amino acid sequence for the particular gene, to possibly link hy­ ionization source used in positive ionization mode (Waters), was used.
pothetical proteins to a function.
2.4.7. Statistical analysis
2.4. Metabolite target analysis of the substrate and metabolite dynamics A Spearman correlation analysis (p < 0.05) was performed between
the metabolites measured and the main microbial species detected with
The cell-free supernatants were used to determine substrate and FRP, using the Hmisc package (version 4.6–0; Harrell, 2021). The
metabolite concentrations in the fermenting wort and maturing beer. All calculated correlation was visualized and hierarchically clustered using
samples were both prepared and analyzed in triplicate. the ComplexHeatmap package (version 2.6.2; Gu et al., 2016).

2.4.1. Determination of simple carbohydrate and maltooligosaccharide 3. Results


concentrations.
High-performance anion exchange chromatography coupled to 3.1. Course of temperature, pH, and apparent attenuation
pulsed amperometric detection (HPAEC-PAD) with internal standardi­
zation was applied, making use of an ICS3000 chromatograph (Dionex, Two nearly identical oak casks of the same type were filled with
Sunnyvale, CA, USA), equipped with a Carbopac™ PA10 column (Dio­ lambic beer wort coming from the same brew with a temperature, pH,
nex) to determine the concentrations of fructose, glucose, maltose, and and apparent attenuation of 13.7 ◦ C, 4.3, and 12.5◦ P, respectively. The
sucrose, or a Carbopac™ PA100 column (Dionex) to determine the fermenting wort and maturing beer displayed comparable temperature,
concentrations of maltotriose, maltotetraose, maltopentaose, malto­ pH, and apparent attenuation profiles in the two casks (Fig. 1A). The
hexaose, and maltoheptaose, as described previously (De Roos et al., temperature inside the casks followed the seasonal changes of the
2018a, 2018b). temperature in the cellar. During the first week of fermentation, the pH
of the fermenting wort in both casks decreased from 4.3 to 3.9, after
2.4.2. Determination of ethanol and short-chain fatty acid concentrations which the pH remained constant. From week 6 onwards, the pH
High-performance liquid chromatography with refractive index continued to drop to values below 3.6. The apparent attenuation of the
detection (HPLC-RI), applying external calibration, was applied to fermenting wort rapidly decreased during the first six weeks of
determine the concentrations of ethanol, acetic acid, propionic acid, and fermentation, followed by a slow decrease later on in the production
butyric acid, as described previously (De Roos et al., 2018a, 2018b). process.
Therefore, a Waters Alliance 2695 HPLC chromatograph (Waters, Mil­
ford, MA, USA), equipped with an ICSep ICE ORH-801 column (Trans­ 3.2. Culture-dependent microbial community dynamics and species
genomic North America, Omaha, NE, USA) and coupled to a W410 RI diversity
detector (Waters), was used.
The lambic beer production processes in both nearly identical oak
2.4.3. Determination of lactic acid stereoisomer concentrations casks examined displayed four phases, namely an initial or enterobac­
HPLC with ultraviolet detection (HPLC-UV) and external calibration terial fermentation phase, a main or alcoholic fermentation phase, an
was applied to determine the concentrations of D-lactic acid and L-lactic acidification phase, and a maturation phase (Figs. 1B, C, and 2). The
acid, as described previously (De Roos et al., 2018a). Therefore, a Wa­ presence of colony counts of presumptive LAB (mMRS), enterobacteria
ters Alliance 2695 HPLC chromatograph (Waters), equipped with a (VRBG), AAB (mDMS), yeasts (YPG), and cycloheximide-resistant yeasts
Shodex Orpak CRX-853 column (Showa Denko, Tokyo, Japan) and (YPGc) before the transfer of the cooled wort to the oak casks indicated
coupled to a UV detector operating at 253 nm (Waters), was used. that the starting lambic beer wort was spontaneously inoculated over­
night during the coolship step and by the house microbiota of the
2.4.4. Determination of volatile organic compound concentrations brewery equipment used.
Gas chromatography with flame ionization detection (GC-FID), During the short enterobacterial and AAB phase (first two weeks),
applying internal standardization, was applied to determine the con­ Enterobacter species were the most prevalent bacteria, exceeding log 5.0
centrations of 2,3-butanediol, 2,3-butanedione, acetoin, ethyl acetate, (CFU/mL) on VRBG agar medium, in the fermenting wort of both casks.
ethyl lactate, isoamyl acetate, (iso)butyric acid, (iso)valeric acid, and The same bacterial species were isolated from both VRBG and mMRS

4
L. Vermote et al. International Journal of Food Microbiology 394 (2023) 110163

Fig. 1. (A) Course of the physicochemical parameters temperature, pH, and apparent attenuation during 30-month lambic beer production processes carried out in
two nearly identical oak casks [cask 1 (light coloured) and cask 2 (dark coloured)]. (B, C) Counts of presumptive yeasts, cycloheximide-resistant yeasts, acetic acid
bacteria, lactic acid bacteria and enterobacteria, expressed as log (CFU/mL), as plated on yeast extract-peptone-glucose (YPG) agar medium, YPG agar medium
supplemented with cycloheximide (YPGc), modified deoxycholate-mannitol-sorbitol (mDMS) agar medium, modified de Man-Rogosa-Sharpe (mMRS) agar medium
and violet red-bile-glucose (VRBG) agar medium, respectively, during 30-month lambic beer production processes carried out in two nearly identical oak casks [cask
1 (B), cask 2 (C)]. The four characteristic phases of a lambic beer production process are indicated at the top of the graphs. (For interpretation of the references to
colour in this figure legend, the reader is referred to the web version of this article.)

5
L. Vermote et al. International Journal of Food Microbiology 394 (2023) 110163

Fig. 2. Community dynamics of the presumptive acetic acid bacteria (AAB) species, as determined on modified deoxycholate-mannitol-sorbitol (mDMS) agar
medium, lactic acid bacteria (LAB) species, as determined on modified de Man-Rogosa-Sharpe (mMRS) agar medium, yeast species, as determined on yeast extract-
peptone-glucose (YPG) agar medium, and cycloheximide-resistant yeast species, as determined on YPG agar medium supplemented with cycloheximide (YPGc agar
medium), during 30-month lambic beer production processes carried out in two nearly identical oak casks [cask 1 (top graphs), cask 2 (bottom graphs)]. The number
of isolates picked and identified is given between brackets at the bottom line of the graphs. If no isolates (indicated as 0) were picked, the bacterial counts were < log
2.0 (CFU/mL). The following species were identified on mDMS: Acetobacter cerevisiae (Genbank accession number KF537430.1), Gluconobacter japonicus
(NR_041445.1), Acetobacter indonesiensis (NR_028616.1), Yarrowia lipolytica (NR_111212.1), Acetobacter fabarum (HG329536.1), Kosakonia sacchari (NR_118333.1),
Gluconobacter cerevisiae (HG329585.1), Gluconobacter oxydans (FN391653.1), Acetobacter lambici (HG329531.1), and Dekkera bruxellensis (KY103321.1). The
following species were identified on mMRS: Enterobacter species (CP017184.1, CP017183.1, and NR_117547.1), Carnobacterium maltaromaticum (LC065032.1),
A. lambici (HG329531.1), Pediococcus damnosus (NR_042087.1), Lacticaseibacillus paracasei (NR_025880.1), Gluconobacter liquefaciens (FN391626.1), Gluconobacter
nephelii (HG329579.1), Gluconobacter cerinus (FN391644.1), Serratia liquefaciens (CP006253.1), A. cerevisiae (KF537430.1), A. fabarum (HG329536.1), Acetobacter
persici (KF537423.1), A. indonesiensis (NR_028616.1), Acetobacter lovaniensis (NR_113551.1), G. oxydans (FN391653.1), G. cerevisiae (HG329585.1), and Debar­
yomyces hansenii (KY103230.1). The following species were identified on YPG: Hanseniaspora uvarum (KY103558.1), Saccharomyces cerevisiae (KC881067.1), Pichia
kudriavzevii (AY939808.1), D. hansenii (KY103230.1), Saccharomyces kudriavzevii (NR_111355.1), Pichia fermentans (NR_130688.1), A. cerevisiae (KF537430.1),
D. bruxellensis (KY103321.1), Millerozyma farinosa (NR_111254.1), Pichia membranifaciens (NR_111195.1), and Brettanomyces custersianus (NR_138184.1). The
following species were identified on YPGc: H. uvarum (KY103558.1), A. lambici (HG329531.1), Ogataea pini (NR_138175.1), D. bruxellensis (KY103321.1), Dekkera
anomala (KY103306.1), Candida lundiana (NR_137645.1), and B. custersianus (NR_138184.1).

agar media. A first peak of AAB was found during this initial fermen­ fermenting wort, respectively.
tation phase in both casks as well, up to almost log 5.0 (CFU/mL), for Oxidative yeasts, such as Brettanomyces and Pichia species, co-
which Acetobacter cerevisiae and Acetobacter fabarum were most preva­ occurred in the maturing beer of both oak casks from month 6 on­
lent. During the initial stages of the fermentation processes (after 24 h wards, which marked the start of the maturation phase. Acetobacter
till 9 weeks), high counts of both cycloheximide-resistant yeasts, up to lambici was the prevalent AAB species that could be isolated from month
almost log 6.0 (CFU/mL), among which Hanseniaspora uvarum was most 13 onwards, when Saccharomyces species were absent. This indicated an
prevalent, and cycloheximide-sensitive yeasts, above log 6.0 (CFU/mL), increase in oxygen availability due to lower carbon dioxide production
among which S. cerevisiae was most prevalent, were found in the fer­ by the yeasts. However, the AAB counts fluctuated as a function of time
menting wort of both casks. Both H. uvarum and S. cerevisiae were pre­ and were relatively low, which indicated still limited oxygen availability
sent in nearly equal counts, up to almost log 6.0 (CFU/mL), in the in the oak casks. The counts of the AAB decreased when the yeast counts
fermenting wort of both oak casks during the first day of fermentation. were high, from month 13 and month 18 onwards in the maturing beer
The numbers of cycloheximide-resistant yeasts reached a maximum of of casks 1 and 2, respectively.
log 5.8 (CFU/mL) after 24 h of fermentation, before they gradually
decreased to counts below log 2.0 (CFU/mL) after 6 weeks of fermen­
3.3. Shotgun metagenomic analysis
tation. The ratio of Saccharomyces species to H. uvarum increased to a
value higher than 1:1 after 24 h in the fermenting wort of both oak casks.
3.3.1. Taxonomic classification at genus level using all metagenomic
The most prevalent yeast species shifted from S. cerevisiae to
sequence reads (MSRs)
S. kudriavzevii after 9 weeks of fermentation. Low counts of between log
The MSRs obtained were used for taxonomic classification at genus
2.0 (CFU/mL) and log 4.0 (CFU/mL) of these Saccharomyces species
level using blastn, DIAMOND, Kaiju, and Kraken 2, to obtain software-
were found until month 6 of the production processes.
and database-independent results, and showed similar trends for the
The LAB counts increased when the counts of the yeasts decreased
fermenting wort of both casks (Fig. 3A). The 0-h sample included the
and the temperature exceeded 15 ◦ C, which marked the start of the
genera Acinetobacter, Arcobacter, Chryseobacterium, Flavobacterium,
acidification phase (Fig. 2). LAB, all identified as P. damnosus, were
Janthinobacterium, Lactococcus, and Pseudomonas, which were not found
isolated from week 6 (cask 1) and 9 (cask 2) until month 6, which gave
in subsequent time points during these fermentation and maturation
counts up to almost log 7.0 (CFU/mL) and log 5.0 (CFU/mL) in the
processes. Fungi (such as Mucor, Penicillium, Rhizopus, and Trichococcus)

6
L. Vermote et al. International Journal of Food Microbiology 394 (2023) 110163

Fig. 3. Taxonomic classification of the metagenomic sequence reads during 30-month lambic beer production processes carried out in two nearly identical oak casks.
The 0-h sample on the top of the figure represents the cooled lambic beer wort before its transfer to the wooden casks. (A) Genus level identification using blastn with
the NCBI nt database (nt), DIAMOND with the NCBI nr database (Diam), Kaiju with a customized database of bacterial, archaeal, and lower eukaryotic protein
sequences (Kaiju), and Kraken2 with a customized database of bacterial, archaeal, and lower eukaryotic nucleotide (K2nt) or protein (K2aa) sequences. The category
“Minorities” represents all genera present with a relative abundance below 1.0 % for all methods used. The category “Higher than genus” represents all taxonomic
levels assigned above genus level. The category “Unassigned” represents reads that were not assigned to any taxonomic level. The category “No hits” includes all
reads that could not be classified at all. (B) Species level identification, using fragment recruitment plotting with a custom-made database containing genera pre­
viously assigned through taxonomic classification at genus level [cask 1 (left) and cask 2 (right)].

were found, depending on the taxonomic tool used. The genera Gluco­ 3.3.2. Taxonomic classification at species level based on fragment
nobacter, Hanseniaspora, and Saccharomyces were the only ones detected recruitment plotting
in the 0-h sample that were also present upon fermentation and matu­ MSRs were aligned to an in-house database containing the genome
ration. For the fermenting wort and maturing beer in both casks, the sequence of 3662 microbial species, corresponding to 175 genera, also
yeast genera Saccharomyces, Hanseniaspora, Brettanomyces, Ogataea, and taking into account the new taxonomy for Lactobacillus (Zheng et al.,
Pichia, and the bacterial genera Acetobacter, Gluconobacter, Pediococcus, 2020). Overall, 95 % of the MSRs could be classified with FRP for all
and Lactobacillus were found as the main genera, i.e., having a relative time points.
abundance of at least 1 %, although many other genera were present in For the 0-h sample, many species from various genera were found,
lower relative abundances. Overall, the percentage of MSRs that could even when applying a MSR ANI cut-off of at least 95 % (Table 1). The
be assigned at genus level ranged from 75 % for the samples taken at the largest number of reads, namely 30.33 %, aligned to Mucor circinelloides,
start of the fermentation processes until 15 % for the 18- and 30-month with a MSR ANI value of 94.34 %. This suggests that a closely related
samples. Mucor species was present in the sample, for which there was no genome
sequence available in the database used. In decreasing order of relative
abundance, H. uvarum, S. cerevisiae, Penicillium roqueforti, Acinetobacter

7
L. Vermote et al. International Journal of Food Microbiology 394 (2023) 110163

Table 1 and P. damnosus were found, albeit at lower relative abundances.


List of microbial species identified in the metagenome from the 0-h sample, this In the fermenting wort and maturing beer of cask 2, similar trends
is the sample before the wort was transferred to the barrels, based on fragment were seen (Fig. 3B), although at day 3, A. cerevisiae (0.07 % of all reads)
recruitment plotting. Only (sub)species with a metagenomic sequence read and Acetobacter malorum (0.16 %) were found. Saccharomyces cerevisiae
average nucleotide identity (MSR ANI) of >95 % are listed. (*) Species with a recruited most of the MSRs (67.54–54.93 %) from day 3 until week 9,
high percentage of reads assigned but with an MSR ANI score lower than 95 %.
and P. damnosus was only found starting from week 9. At month 18, next
Species % of reads Species % of reads to D. bruxellensis, also A. lambici was found. At month 30, >90 % of the
assigned assigned
MSRs aligned to B. custersianus.
Mucor circinelloides 30.33* Lelliottia amnigena 0.12
Hanseniaspora uvarum 3.50 Buttiauxella gaviniae 0.11 3.3.3. Statistical analysis
Saccharomyces 3.46 Carnobacterium inhibens 0.10
Considering all fermenting wort samples, a PCA of the relative spe­
cerevisiae subsp. inhibens
Penicillium roqueforti 2.28 Carnobacterium inhibens 0.10 cies abundance based on FRP resulted in two principal components
subsp. gilichinskyi (PCs) that together covered 86.8 % of the total variance (Fig. 4A). PC1
Acinetobacter guillouiae 1.63 Saccharomyces bayanus 0.09 was characterized by negative loadings for most species found in the 0-h
Lactococcus raffinolactis 1.53 Penicillium biforme 0.09
sample and positive loadings for S. kudriavzevii, S. cerevisiae,
Geotrichum candidum 0.67 Penicillium carneum 0.09
Chryseobacterium bovis 0.58 Penicillium camemberti 0.09 P. damnosus, S. uvarum, D. bruxellensis, B. custersianus, A. lambici, and
Flavobacterium 0.55 Carnobacterium viridans 0.09 D. anomala. PC2 was characterized by negative loadings for H. uvarum,
hibernum Gluconobacter oxydans, Gluconobacter japonicus, S. cerevisiae, Saccharo­
Yarrowia lipolytica 0.51 Buttiauxella agrestis 0.09 myces bayanus, A. fabarum, A. cerevisiae, Pichia fermentans, S. uvarum,
Trichococcus 0.50 Candida sake 0.08
Clavispora lusitaniae, and S. kudriavzevii and positive loadings for
flocculiformis
Pseudomonas saxonica 0.46 Rahnella variigena 0.08 D. bruxellensis, B. custersianus, P. damnosus, and A. lambici. The 0-h
Penicillium solitum 0.46 Acetobacter cerevisiae 0.08 sample was separated from the other samples and associated with a
Pseudomonas veronii 0.45 Buttiauxella noackiae 0.08 negative loading of PC1 and an almost neutral loading of PC2. The
Enterobacter soli 0.44 Acetobacter malorum 0.08
fermenting wort samples from the casks were associated with slightly
Wickerhamomyces 0.42 Chryseobacterium 0.08
anomalus balustinum
positive loadings of PC1 and varying loadings of PC2, forming a vertical
Pseudomonas rhodesiae 0.42 Carnobacterium 0.07 line, and those from months 18 and 30 fell together. For this reason, a
maltaromaticum second PCA was performed with only the fermenting wort samples from
Saccharomyces 0.39 Cellulosimicrobium 0.07 casks 1 and 2. The first two PCs obtained covered together 54.7 % of the
kudriavzevii aquatile
total variance (Fig. 4B). PC1 was characterized by positive loadings of
Clavispora lusitaniae 0.34 Saccharomyces uvarum 0.07
Pichia membranifaciens 0.34 Kregervanrija fluxuum 0.07 C. lusitaniae, H. uvarum, Pi. fermentans, S. cerevisiae, G. oxydans,
Pseudomonas helleri 0.27 Pseudomonas simiae 0.06 G. japonicus, Enterobacter soli, A. cerevisiae, S. uvarum, and S. bayanus.
Gluconacetobacter 0.27 Pseudomonas putida 0.06 PC2 was characterized by positive loadings of A. fabarum, S. bayanus,
liquefaciens
G. japonicus, A. cerevisiae, and G. oxydans and negative loadings of
Gluconobacter japonicus 0.22 Yersinia enterocolitica 0.05
subsp. enterocolitica
S. kudriavzevii, Wickerhamomyces anomalus, S. uvarum, Pe. roqueforti, and
Pichia fermentans 0.22 Cellulosimicrobium 0.05 S. cerevisiae. Except for the 3-day sample from cask 1, which was asso­
terreum ciated with positive loadings of PC1 and positive loadings of PC2, all
Torulaspora delbrueckii 0.21 Cellulosimicrobium funkei 0.05 other fermenting wort samples were associated with either positive
Duganella zoogloeoides 0.21 Pediococcus damnosus 0.05
loadings of PC1 and negative loadings of PC2 or negative loadings of
Penicillium paneum 0.20 Acetobacter fabarum 0.05
Paraburkholderia 0.20 Dekkera bruxellensis 0.05 PC1 and positive loadings of PC2. All fermenting wort samples of the
tropica two casks did not differ much regarding the PC combination, except for
Flavobacterium 0.19 Yersinia enterocolitica 0.04 the 3-day and 9-week samples.
sinopsychrotolerans subsp. paleartica The intra-sample (alpha) diversity showed that the highest diversity
Sphingobacterium 0.18 Acetobacter indonesiensis 0.04
siyangense
and highest evenness was associated with the 0-h sample (Table 2). The
Acinetobacter albensis 0.18 Carnobacterium jeotgali 0.04 1-week and 6-week fermenting wort samples of cask 1 had slightly lower
Cellulosimicrobium 0.15 Kazachstania unispora 0.03 values for diversity and evenness, followed by the 6-week and 9-week
cellulans samples of cask 2. The 18-month sample of cask 1 and the 30-month
Penicillium 0.15 Enterobacter ludwigii 0.03
sample of cask 2 had a very low diversity and low evenness. The inter-
fuscoglaucum
Pseudomonas carnis 0.14 Sphingomonas 0.03 sample diversity showed no significant differences (p < 0.05) between
paucimobilis the two casks.

3.3.4. Co-assembly of metagenomes, contig binning, and contig annotation


guillouiae, Lactococcus raffinolactis, Geotrichum candidum, Chrys­ Assembly and subsequent binning with CONCOCT of the 15 MSR
eobacterium bovis, Flavobacterium hibernum, Yarrowia lipolytica, and Tri­ data sets resulted in 24 bins; after manual refinement, 42 bins were
chococcus flocculiformis were found in the 0-h sample, ranging from 3.50 obtained, of which 7 bins had a completion of minimum 75 % and a
% to 0.50 % of the MSRs. contamination of <10 % (Table 3). Two of these bins were related to
In the fermenting wort and maturing beer of cask 1, most MSRs bacteria and five to yeasts. Two bins, Bin_23_A and Bin_18_A_1, could be
aligned to S. cerevisiae at day 3, week 1, and week 6 (Fig. 3B), followed considered as high-quality MAGs (> 90 % completion and < 5 %
by S. kudriavzevii and H. uvarum, both displaying lower relative abun­ contamination). Using minimap2, Bin_23_A could be identified as
dances. At day 3, A. cerevisiae (0.15 % of all reads) and A. fabarum (0.39 A. lambici and Bin_18_A_1 as P. damnosus. Because of the high comple­
%) were found. At week 1, MSRs aligned to Saccharomyces uvarum, albeit tion and low contamination of Bin_23_A, the contigs were compared to
at a low relative abundance. From week 6 onward, P. damnosus the genome of A. lambici LMG 27439T, resulting in an ANI value of
occurred, having the highest relative abundance at weeks 9 and 16. 99.92 %. Minimap2 was also used to check the identity of the other bins,
Hanseniaspora uvarum, S. cerevisiae, and S. kudriavzevii were found until as represented in Table 3.
week 16. At month 18, >90 % of the MSRs aligned to D. bruxellensis,
after which this shifted to B. custersianus, with >80 % of the MSRs
3.3.4.1. Acetobacter lambici MAG (Bin_23_A). In the A. lambici MAG, no
aligned at month 30. At that time point, also A. lambici, D. bruxellensis,

8
L. Vermote et al. International Journal of Food Microbiology 394 (2023) 110163

Fig. 4. Statistical analysis of the metagenomic sequence data of the lambic beer fermentation and maturation process samples analyzed, obtained during 30-month
lambic beer production processes carried out in two nearly identical oak casks. (A, B) Principal component analysis of the taxonomic composition based on the
relative abundances of all species detected by fragment recruitment plotting (FRP). Samples are colour- and shape-coded according to the time point and sample type,
respectively. (A) All wort samples, including the 0-h sample (cooled lambic beer wort before its transfer to the casks). (B) Only the fermenting wort samples. (C)
Hierarchical clustering analysis of the Spearman correlation values (p < 0.05) obtained by comparing the main microbial species, determined by FRP, and the
concentrations of the substrates consumed and metabolites produced. Positive correlations are represented in blue and negative correlations in red; the intensity of
the colour represents the degree of correlation. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of
this article.)

genes for the conversion of sucrose or maltose/maltooligosaccharides involved in the production of (R)-acetoin and (S)-acetoin from pyruvate
were found, neither with Prokka nor with a manual search using blastp via α-acetolactate and the further oxidation of acetoin into acetaldehyde
(Fig. 5). Yet, some genes encoding maltooligosyl trehalose synthase and and acetyl-CoA were found. Genes responsible for pyruvate decarbox­
maltooligosyl trehalase were found in other, not complete bins, on ylase and the pyruvate dehydrogenase complex, converting pyruvate
contigs identified with minimap2 as A. malorum and A. cerevisiae, into acetaldehyde or acetyl-CoA, respectively, were present. Genes
respectively. All genes coding for the enzymes involved in the Embden- encoding acetyl-CoA synthetase, phosphotransacetylase, and acetate
Meyerhof-Parnas (EMP) pathway, except for phosphoglucose isomerase kinase were found, catalyzing the reversible conversion of acetate into
and 6-phosphofructokinase, and for the enzymes involved in the pentose acetyl-CoA, acetyl-CoA into acetyl phosphate, and acetyl phosphate into
phosphate pathway (PPP) were found. Also, a gene coding for pyruvate acetate, respectively. All genes involved in a modified tricarboxylic acid
phosphate dikinase was found. Genes coding for pyrroloquinoline (TCA) cycle were found, with genes coding for succinyl-CoA:acetate
quinone-dependent membrane-bound alcohol and aldehyde de­ CoA-transferase, instead of succinyl-CoA synthetase, converting
hydrogenases (all subunits) as well as for cytoplasmic NADH-dependent succinyl-CoA and acetate into succinate and acetyl-CoA, and for malate:
dehydrogenases were found. Genes coding for all subunits of NAD(P) quinone oxidoreductase, instead of malate dehydrogenase, converting
transhydrogenase, both terminal ubiquinol oxidases (cytochrome bd malate into oxaloacetate. All genes coding for the 2-methylcitrate cycle
quinol oxidase and cytochrome bo3 ubiquinol oxidase) and ATP syn­ were found; however, no genes of the glyoxylate shunt were found. A
thase, were found. For NADH:quinone oxidoreductase, not all genes gene coding for the NAD-dependent malic enzyme was found, respon­
encoding its subunits were found. In addition, genes encoding enzymes sible for the conversion of malate into pyruvate. A gene coding for a

9
L. Vermote et al. International Journal of Food Microbiology 394 (2023) 110163

Table 2 acetyl esterase Aes (two copies; HAMAP annotation MF_01958), which
Alpha diversity indexes based on the relative abundances of species detected produces short-chain fatty acid esters (acyl chains containing up to eight
using fragment recruitment plotting in the different lambic beer fermentation carbon atoms), and an esterase YbfF (UniProt accession number
and maturation process samples obtained during 30-month lambic beer pro­ P75736), displaying an esterase activity toward palmitoyl-CoA and
duction processes carried out in two nearly identical oak casks. The Simpson (D) malonyl-CoA.
and Pielou (Je) indices were calculated for all samples to determine their di­
versity and evenness, respectively. C1, cask 1; C2, cask 2.
3.3.4.2. Pediococcus damnosus MAG (Bin_18_A_1). The P. damnosus
Sample Simpson (D) Pielou (Je)
MAG had genes coding for sucrose 6-phosphate hydrolase and maltose
0h 0.93 0.79 phosphorylase, allowing to use sucrose and maltose, respectively
C1 3 d 0.43 0.37 (Fig. 6). In this MAG, genes encoding all enzymes involved in both the
C1 1 w 0.55 0.47
homo- and heterolactic fermentation pathways were found, as well as
C1 6 w 0.64 0.62
C1 9 w 0.30 0.36 genes coding for D-lactate dehydrogenase and L-lactate dehydrogenase,
C1 16 w 0.16 0.20 allowing to convert pyruvate into D-lactate and L-lactate, respectively.
C1 18 m 0.02 0.04 Genes involved in the conversion of pyruvate into acetyl phosphate and
C1 30 m 0.20 0.28
its subsequent conversion into either acetate or ethanol were found. Also
C2 3 d 0.37 0.31
C2 1 w 0.37 0.33 a gene coding for the malolactic enzyme was found, allowing to convert
C2 6 w 0.51 0.42 S-malate into L-lactate. A ferulic acid decarboxylase gene was found,
C2 9 w 0.53 0.43 enabling the conversion of ferulic acid and p-coumaric acid into 4-vinyl
C2 16 w 0.23 0.32 guaiacol and 4-vinyl phenol, respectively. Further, in a bin constructed
C2 18 m 0.26 0.41
during manual refinement of Bin_18, also genes encoding enzymes
C2 30 m 0.02 0.04
involved in the production of (R)-acetoin and (S)-acetoin from pyruvate
via α-acetolactate were found in the constituting contigs identified as
P. damnosus. This bin also contained genes coding for tyrosine decar­
Table 3
boxylase [98.55 % identity and 99.20 % similarity to the Tetragenococcus
Bins obtained after contig binning with CONCOCT and two rounds of manual
halophilus protein (NCBI accession number WP_031944088.1); contig
refinement using anvi’o. The taxonomy shown here was determined with Kaiju
within the anvi’o pipeline used. Only bins with at least 75 % completion and identified as Pediococcus cellicola] and histidine decarboxylase [100 %
<10 % contamination were considered. identity to the Lentilactobacillus hilgardii protein (UniProt accession
number Q5DLT9); contig identified as Pediococcus parvulus], catalyzing
Metagenomic bin Size Completion Contamination Taxonomy
(Kaiju) (Mbp) (%) (%) (minimap2)
the conversion of tyrosine into tyramine and histidine into histamine,
respectively. Also, genes coding for the hop resistance proteins HorA
Bin_23_A 2.61 97.18 1.41 Acetobacter
[99.66 % identity and 99.83 % similarity to the Pediococcus claussenii
Acetobacter lambici
Bin_18_A_1 1.87 94.37 1.41 P. damnosus protein (UniProt accession number G8PFG1)], HorB [100.00 % identity
Pediococcus to the Loigolactobacillus backii protein (UniProt accession number
damnosus A4UX85)], HorC [98.83 % identity and 99.30 % similarity to the Levi­
Bin_5_A_1 11.37 87.95 2.41 D. bruxellensis
lactobacillus brevis protein (UniProt accession number Q6I7K2)], and
Brettanomyces
bruxellensis
HitA [99.78 % identity and 100.00 % similarity to the L. brevis protein
Bin_7_A 9.99 85.54 2.41 Brettanomyces (UniProt accession number Q93V04)] were found in this bin, all on the
Brettanomyces custersianus constituting contigs identified as P. damnosus, as well as a gene coding
bruxellensis for a putative glucan synthase [97.88 % identity and 98.94 % similarity
Bin_13_A_1 9.75 81.93 9.64 S. cerevisiae,
to the P. damnosus protein (UniProt accession number Q336K4); contig
Saccharomyces S. kudriavzevii
cerevisiae x identified as P. claussenii].
Saccharomyces
kudriavzevii 3.3.4.3. Yeast bins (Bin_2_A, Bin_4_A_1, Bin_5_A_1, Bin_7_A, and Bin
Bin_4_A_1 10.25 77.11 9.64 Saccharomyces
13_A_1). The yeast bins differed in predicted carbohydrate usage
Unknown cerevisiae
Saccharomyces (Fig. 7). In Bin_5_A_1, identified as D. bruxellensis, a gene encoding
Bin_2_A 8.52 78.31 2.41 H. uvarum maltodextrin phosphorylase was found, allowing to convert maltooli­
Hanseniaspora gosaccharides into glucose 6-phosphate via glucose 1-phosphate. In Bin
uvarum 13_A_1, identified as S. kudriavzevii, and in a bin constructed during the
refinement of Bin 4_A (contigs identified as S. cerevisiae), genes coding
glycerol uptake transporter was found as well as all genes to convert for a maltase that can hydrolyze maltose, maltotriose and sucrose, and
glycerol into dihydroxyacetone phosphate, or vice versa. No genes cod­ an invertase, converting sucrose into glucose and fructose, were found.
ing for lactate dehydrogenase were found, although genes coding for a In the S. cerevisiae bin (Bin_4_A_1), a gene encoding an α-glucosidase,
lactate permease and a lactate utilization protein were found. which can hydrolyze maltose, was found. Both Saccharomyces bins also
The A. lambici MAG was also searched for several genes known to be possessed a gene coding for a sporulation-specific glucoamylase, capable
responsible for acid tolerance in AAB. A gene coding for an acetic acid of degrading glycogen, starch, maltotriose and maltose into glucose, as
resistance ABC transporter ATP-binding protein Uup was found and this well as a gene coding for a polygalacturonase, a pectinolytic enzyme. In
gene had also 82.09 % identity and 92.04 % similarity to the acetic acid the D. bruxellensis bin (Bin_5_A_1), a gene coding for a putative glycosyl
resistance ABC transporter AatA of Acetobacter pasteurianus NBRC 3188 transferase [97.14 % identity and 98.45 % similarity to the
(UniProt accession number A0A401WXJ5). Also, the genes coding for D. bruxellensis AWRI1499 protein (NCBI accession number EIF48743)]
urea carboxylase [97.66 % identity and 99.07 % similarity to the was found, indicating the possible breakdown of β-glycosides, such as
A. fabarum protein (UniProt accession number A0A269XV37)] and the β-1,3-glucan laminarin. Another β-glucosidase that can assimilate
allophanate hydrolase responsible for the conversion of urea into cellobiose, gentiobiose, and arbutin showed a sequence identity of only
ammonia were found. Finally, a gene coding for N5-carbox­ 66.91 % and a similarity of 81.84 % to the D. bruxellensis AWRI1499
yaminoimidazole ribonucleotide (N5-CAIR) mutase, involved in purine protein (NCBI accession number EIF45415). One gene in Bin_2_A,
biosynthesis, was found. The MAG also possessed genes coding for an identified as H. uvarum, showed 51.24 % and 56.50 % identity and

10
L. Vermote et al. International Journal of Food Microbiology 394 (2023) 110163

Fig. 5. Overview of the main catabolic pathways reconstructed with genes found in Bin_23_A, identified as Acetobacter lambici, through functional analysis of
metagenomic contigs derived from whole-community DNA of samples from lambic beer fermentation and maturation processes carried out in two nearly identical
oak casks. Catabolic reactions are indicated with black arrows; dashed arrows depict multiple reactions. Chemical conversions are indicated with red arrows. TCA,
tricarboxylic acid. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

68.88 % and 71.75 % similarity with these putative glycosyl transferase phenol reductase [98.70 % identity and similarity to the D. bruxellensis
and β-glucosidase genes, respectively, hinting to the presence of another AWRI1499 protein (UniProt accession number I2JWC1)] was found,
β-glucosidase gene in this bin. Almost all yeast bins (Bin_2_A, Bin_5_A_1, converting the latter metabolites into 4-ethyl guaiacol and 4-ethyl
and Bin_13_A_1) contained all genes involved in the EMP pathway and phenol, respectively, but also known to have a superoxide dismutase
PPP. Although the S. cerevisiae bin (Bin_4_A_1) was missing a gene activity. In the other yeast bins (Bin_2_A, Bin_4_A_1, Bin_7_A, and
coding for fructose bisphosphate aldolase in the EMP pathway, this gene Bin_13_A_1), the gene coding for superoxide dismutase was found with
was found in a bin obtained during refinement (contig identified as identities of 64.10 %, 73.86 %, 88.96 %, and 73.86 % and similarities of
Saccharomyces paradoxus). In Bin_7_A, identified as B. custersianus, the 77.56 %, 83.66 %, 92.86 %, and 84.79 %, respectively. All yeast bins
gene encoding ribose 5-phosphate epimerase in the PPP was missing. All possessed a gene coding for an NAD-dependent glycerol 3-phosphate
yeast bins contained all genes allowing to convert pyruvate into acet­ dehydrogenase; however, in the Brettanomyces/Dekkera bins
aldehyde, α-acetolactate, and (R)-acetoin by multifunctional pyruvate (Bin_5_A_1 and Bin_7_A), no gene coding for glycerol 3-phosphate
decarboxylase; however, the H. uvarum bin was the only bin that did not phosphatase was found. The S. kudriavzevii (Bin_13_A_1) and Brettano­
possess genes encoding diacetyl reductase, converting diacetyl into (R)- myces/Dekkera bins (Bin_5_A_1 and Bin_7_A) possessed a gene coding for
acetoin, and its subsequent conversion into (R,R)-butanediol by (R,R)- glycerol kinase and a quinol-dependent glycerol 3-phosphate dehydro­
butanediol reductase. All yeast bins possessed genes to further convert genase, which is involved in the degradation of glycerol. The
acetaldehyde into ethanol or acetate and subsequently acetate into S. kudriavzevii bin (Bin_13_A_1) harbored also a gene coding for a glyc­
acetyl-CoA to enter the TCA cycle. Also, a gene coding for malate de­ erol uptake/efflux facilitator protein.
hydrogenase, converting malate into pyruvate, was found in all yeast
bins. The H. uvarum bin was the only bin missing genes involved in the 3.4. Substrate consumption and metabolite production dynamics
glyoxylate cycle. The H. uvarum (Bin_2_A) and the S. kudriavzevii
(Bin_13_A_1) bins were the only bins containing all genes encoding all Unless stated otherwise, the concentrations and profiles of the sub­
subunits of the pyruvate dehydrogenase complex, converting pyruvate strates and metabolites measured were comparable for the fermenting
into acetyl-CoA. In the Saccharomyces bins (Bin_4_A_1 and Bin_13_A_1), a wort and maturing beer of the two oak casks of the lambic beer pro­
gene coding for ferulic acid decarboxylase, converting ferulic acid and p- duction processes examined.
coumaric acid into 4-vinyl guaiacol and 4-vinyl phenol, respectively, Glucose (7.3 g/L), fructose (2.0 g/L), sucrose (7.0 g/L), maltose
was found. In the D. bruxellensis bin (Bin_5_A_1), a gene coding for vinyl (55.0 g/L), maltotriose (12.0 g/L), and higher maltooligosaccharides

11
L. Vermote et al. International Journal of Food Microbiology 394 (2023) 110163

Fig. 6. Overview of the main catabolic pathways reconstructed with genes found in Bin_18_A_1, identified as Pediococcus damnosus, through functional analysis of
metagenomic contigs derived from whole-community DNA of samples from lambic beer fermentation and maturation processes carried out in two nearly identical
oak casks. Catabolic reactions are indicated with black arrows; dashed arrows depict multiple reactions. Chemical conversions are indicated with red arrows.
Catabolic reactions in grey were found in contigs of related bins that were identified as P. damnosus. (For interpretation of the references to colour in this figure
legend, the reader is referred to the web version of this article.)

(6.3 g/L) were the most abundant carbohydrates present in the initial continuous production of lactic acid in the maturing beer of casks 1 and
lambic beer wort (Fig. 8A and B). Also organic acids were present, 2, even when no LAB were found, reflected the presence of LAB at low
namely lactic acid (1.2 g/L), malic acid (180 mg/L), citric acid (170 mg/ counts, below log 2.0 (CFU/mL), throughout the entire maturation
L), and gluconic acid (53 mg/L) (Fig. 8A and C). No short-chain or phase. D-lactic acid and L-lactic acid were both produced in a nearly
branched-chain fatty acids were found during the short enterobacterial equimolar ratio. Malic acid was rapidly depleted after the occurrence of
and AAB phase, indicating limited growth of this background micro­ LAB at week 6 in the fermenting wort of cask 1 and week 9 in cask 2,
biota. A sequential consumption of the mono- and disaccharides took which coincided with the production of lactic acid and thus indicated
place (Fig. 8A), with a nearly complete depletion of the mono-, di-, and malolactic fermentation (Fig. 8A and C). During the final year of
trisaccharides as follows. The depletion of sucrose in the fermenting maturation, when all other substrates were exhausted or present in very
wort after three days of fermentation was followed by the depletion of low concentrations, citric acid was metabolized, probably by the yeasts
glucose and fructose after two weeks, which corresponded with the high present (Fig. 8C). The concentration of gluconic acid that was present
prevalence of S. cerevisiae. A continuous degradation of maltose from the start of the fermentation remained stable throughout the
occurred until it was nearly completely exhausted after week 9. The lambic beer production processes performed.
same degradation profile was followed by maltotriose, which indicated The most abundant esters of lambic beer, ethyl lactate and ethyl
the ability of the strains of the Saccharomyces species present to take up acetate, were produced from week 2 onwards, when ethanol was present
and consume simultaneously maltose and maltotriose, although small abundantly (Fig. 8A and C). In the fermenting wort of both oak casks,
concentrations of maltotriose remained present until month 13. Coin­ maximal concentrations of over 50 mg/L of ethyl lactate were reached
ciding with the low concentrations of both maltose and maltotriose, the after 6 weeks. Maximal concentrations of over 300 mg/L and around
counts belonging to Saccharomyces species decreased. From week 2 200 mg/L of ethyl acetate were found in the maturing beer of casks 1 and
onwards, a simultaneous degradation of the higher maltooligo­ 2 after 30 months, respectively. Isoamyl acetate was not found. Acetoin
saccharides (from maltotetraose to maltoheptaose) took place. production was negligible in the fermenting wort of both casks, indi­
During the first seven weeks of the lambic beer production processes cating minimal AAB growth or oxidation.
examined, the yeast-associated metabolites ethanol, methyl-1-butanol, Some biogenic amines were present at low concentrations in the
and succinic acid were produced (Fig. 8A, C, and D). Acetic acid was initial lambic beer wort (Fig. 8D). It concerned agmatine (12 mg/L),
barely produced in the fermenting wort and maturing beer of both oak putrescine (6 mg/L), and cadaverine (2 mg/L). The concentrations of
casks, which reflected the low counts of the AAB present. Only during those biogenic amines remained stable upon fermentation. Histamine
the final year of the maturation process, which coincided with the and tyramine were produced mainly during the acidification phase in
growth of AAB and Brettanomyces species, acetic acid was produced up the fermenting wort of both oak casks, which coincided with the pres­
to 1000 mg/L and 400 mg/L in the maturing beer of casks 1 and 2, ence of LAB. The concentrations of histamine were around 20 mg/L in
respectively. The differences in acetic acid concentrations indicated the fermenting wort of casks 1 and 2. The most prevalent biogenic
differences in the AAB counts because of oxygen availability, likely amine, tyramine, was still present in the fermenting wort of both casks at
depending on barrel porosity. Lactic acid was produced from weeks 2 the end of the maturation phase. It reached concentrations between 40
and 9 onwards in the fermenting wort of casks 1 and 2, respectively. The and 50 mg/L in the maturing beer. Phenylethylamine and tryptamine

12
L. Vermote et al. International Journal of Food Microbiology 394 (2023) 110163

Fig. 7. Overview of the main catabolic pathways reconstructed with genes found in fungal bins through functional analysis of metagenomic contigs derived from
whole-community DNA samples from lambic beer fermentation and maturation processes carried out in two nearly identical oak casks. Pathways found in Bin_5_A_1,
Bin_7_A, Bin_2_A, Bin_4_A_1, and Bin_13_A_1, identified as Dekkera bruxellensis, Brettanomyces custersianus, Hanseniaspora uvarum, Saccharomyces cerevisiae, and
Saccharomyces kudriavzevii, respectively, are displayed. Catabolic reactions present in all bins are indicated with black arrows; dashed arrows depict multiple re­
actions. Catabolic reactions that were not present in all bins are indicated with purple arrows. Reactions represented by genes present at low identities are indicated
with brown arrows. Chemical conversions are indicated with red arrows. TCA, tricarboxylic acid. (For interpretation of the references to colour in this figure legend,
the reader is referred to the web version of this article.)

were not found. represent species without a sequenced genome and/or novel species. In
A Spearman correlation analysis showed two clusters, one consisting earlier work, FRP has proven to be successful to identify MSRs at species
of the yeast species present during the enterobacterial and main level and even to unravel the presence of novel species, as was the case
fermentation phases and another consisting of the species present during for Acetobacter sicerae in water kefir samples (Verce et al., 2019, 2020).
the acidification and maturation phases (Fig. 4C). The carbohydrates Second, functional analysis of two high-quality bacterial MAGs and five
and malic acid correlated positively with the first cluster (correlation yeast bins obtained after manual refinement showed the presence of
values ranging from 0.56 to 0.99), whereas acetoin and ethyl acetate genes that have not been shown before, for instance those coding for
displayed a positive correlation with P. damnosus (correlation values of ferulic acid decarboxylase (P. damnosus) and maltodextrin phosphory­
0.73 and 0.76, respectively). Methyl 1-butanol, ethyl lactate, acetic acid, lase (D. bruxellensis) that are of importance in the maturation of lambic
histamine, tyramine, lactic acid, ethanol, and succinic acid were posi­ beer. Finally, microbial growth and gene prediction could be backed up
tively correlated with the second cluster (correlation values ranging by metabolite target analysis.
from 0.54 to 0.89). The present study further contributed to the unravelling of the role of
wooden casks in the spontaneous inoculation of the wort and maturing
4. Discussion beer during lambic beer production. Although shotgun metagenomics
revealed that the cooled wort had the highest diversity and evenness, its
Lambic beer productions have been examined thoroughly, in harboring species did not contribute to the wort fermentation and beer
particular regarding their microbiology, using culture-dependent tech­ maturation in the casks, as many of these species were solely present at
niques (De Roos et al., 2018a, 2018b; Spitaels et al., 2014b, 2015), and the coolship stage. The present multiphasic study of two parallel lambic
to a lesser extent culture-independently, either through amplicon-based beer production processes, starting from the same wort and carried out
high-throughput sequencing (De Roos et al., 2018a, 2019) or shotgun in nearly identical oak casks of wine production origin, showed that the
metagenomics (De Roos et al., 2020). Some of those studies have com­ wooden barrels did have an impact on the production process from a
bined the microbiological assessment with a thorough metabolomic microbiological point of view. The four characteristic phases of a lambic
analysis (De Roos et al., 2018a, 2018b, 2019, 2020). Although the beer production process occurred as described before (De Roos et al.,
shotgun metagenomic sequencing approach of the present study showed 2018a), with similar microbial community, substrate consumption, and
similar trends as the culture-dependent investigations, it unraveled metabolite production dynamics, and typical characteristics, including
additional features. First, methodologically, taxonomic analysis of the malolactic fermentation performed by P. damnosus and simultaneous
MSRs using FRP, allowing MSR-based taxonomic assignments at species degradation of maltooligosaccharides by yeasts. Besides these generally
level, showed that more reads could be assigned compared with a series encountered characteristics, some specific differences with previously
of software tools and databases used that only allowed taxonomic as­ studied lambic beer production processes in the same brewery were
signments at genus level. The remaining unassigned MSRs could found, possibly highlighting that a combination of different coolship

13
L. Vermote et al. International Journal of Food Microbiology 394 (2023) 110163

Fig. 8. Concentrations of monosaccharides, disaccharides, lactic acid, and ethanol (A); maltooligosaccharides (maltotriose, M3; maltotetraose, M4; maltopentaose,
M5; maltohexaose, M6; and maltoheptaose, M7) and acetic acid (B); other organic acids, ethyl acetate, and ethyl lactate (C); and acetoin, higher alcohols, and
biogenic amines (D) during 30-month lambic beer production processes carried out in two nearly identical oak casks (cask 1 and cask 2). The four characteristic
phases of a lambic beer production process are indicated at the top of the graphs.

batches and the casks used may influence lambic beer production. AAB oxygen to be flushed out, is mostly dependent on the presence of
generally grow during both the enterobacterial and main fermentation Saccharomyces species, whereas pellicle formation is mostly determined
phases, depending on the oxygen availability (De Roos et al., 2018b). In by Brettanomyces and Pichia species. Besides regulating the oxygen
the fermenting wort of both oak casks of the present study, culture- availability, wooden barrels harbor resident microorganisms and can
dependent analysis showed that the AAB that grew during the entero­ thus act as an additional microbial inoculation source, in particular of
bacterial phase mostly belonged to the species A. fabarum and LAB and yeasts (del Alamo-Sanza and Nevares, 2014; De Roos et al.,
A. cerevisiae, as opposed to A. orientalis as the most prevalent AAB spe­ 2019). This additional inoculation helps to diminish batch-to-batch
cies during the first AAB peak of the fermenting wort in a previous study variations concerning fermentation profile, thereby outcompeting un­
(De Roos et al., 2018b). Although the type strains of A. fabarum, desirable microorganisms (De Roos et al., 2018b). The fact that barrels
A. cerevisiae, and A. orientalis were isolated from different niches, strains act as an additional inoculation source was further strengthened by the
of these three species have been isolated from lambic beer productions presence of highly similar microorganisms in the present study and a
before (Cleenwerck et al., 2002, 2008; Lisdiyanti et al., 2001; Spitaels previous one carried out in the same brewery (De Roos et al., 2018b).
et al., 2015). As these three AAB species are relatively distinct, the risk The additional inoculation caused by wooden surfaces has also been
for misidentifications is minimal. The shotgun metagenomic analysis of reported for spontaneous cider fermentations, for which the use of
the present study identified A. cerevisiae and A. fabarum during the main traditional wooden presses versus pneumatic stainless steel presses and
fermentation phase of cask 1 and A. cerevisiae and A. malorum (phylo­ stainless steel fermentors shows a profound influence on the production
genetically closely related to A. cerevisiae) in that of cask 2. Hence, the process (Del Campo et al., 2003; Morrissey et al., 2004). In addition, it is
identity of the AAB species prevailing during the enterobacterial and known, for example, that the heterogenous microbial communities that
AAB phase of a lambic beer production process may vary between inhabit the internal wine barrel surfaces include Brettanomyces species
production batches. Moreover, their presence seemed to be mostly (Barata et al., 2013; Fugelsang and Edwards, 2007). These yeast species
dependent on the environmental inoculation during the overnight are of importance during lambic beer maturation (Bongaerts et al., 2021;
cooling of the wort and the use of non-sterile brewery equipment instead Bouchez and De Vuyst, 2022; De Roos and De Vuyst, 2019; Spitaels
of potential inoculation of AAB species residing in the wood of the et al., 2017). Moreover, due to their physical inertness and porosity,
barrels. The maturing lambic beer in the oak casks used in the present wooden surfaces are not easy to sanitize (De Roos et al., 2019; Guzzon
study displayed a delayed and limited second booming of AAB, mainly of et al., 2011; Oelofse et al., 2008; Spitaels et al., 2017). The higher the
the species A. lambici, likely due to a limited oxygen availability as a porosity of the barrels, the better the conditions for survival and biofilm
result of the low porosity of the casks. The oxygen availability and, formation by microorganisms are. Therefore, wooden barrels, with their
hence, the growth of AAB during lambic beer production processes highly abundant pores and cracks, permit a good substrate for biofilm
seems to be a combined effect of barrel characteristics (porosity, head­ formation and a protective barrier for microorganisms (Guzzon et al.,
space, and surface to volume ratio), and the presence of specific yeast 2011; Swaffield and Scott, 1995; Swaffield et al., 1997), which are likely
species that cause carbon dioxide production and pellicle formation (De of importance for later inoculation of the contents of the barrel (De Roos
Roos et al., 2018b). The production of carbon dioxide, which causes et al., 2019).

14
L. Vermote et al. International Journal of Food Microbiology 394 (2023) 110163

Several new insights into the genetic potential of typical members of enterobacterial growth, MAG-originating genes related to amino acid
lambic beer fermentation and maturation processes were obtained based decarboxylation linked P. damnosus with biogenic amine formation,
on the metagenomic bin analysis. Several of the genes related to acetic which supported previous findings (De Roos et al., 2018a). Nevertheless,
acid tolerance, identified in the A. lambici MAG, were common for the biogenic amine concentrations were well below the health-
Acetobacter species to allow their survival in an acid environment, such acceptable concentrations that have been issued by the panel on bio­
as the conversion of acetate into acetyl-CoA by different enzymatic re­ logical hazards of the European food safety authority for fermented
actions, an acetic acid resistance ABC transporter, the conversion of urea foods (Biohaz, 2011).
into ammonia, the 2-methylcitrate cycle, and a N5-CAIR mutase (Ill­ In conclusion, the present study showed that a systematic and
eghems et al., 2013; Lynch et al., 2019; Nakano et al., 2006; Yang et al., multiphasic analysis approach, encompassing classical microbiology,
2019). The A. lambici MAG further confirmed the uptake and conversion metabolite target analysis, and shotgun metagenomics could further
of lactate and glycerol as carbon sources (Pelicaen et al., 2019) as well as unravel the secrets of lambic beer production. The casks used likely
a pyruvate phosphate dikinase as the driving force of gluconeogenesis in helped in establishing a stable microbiota of yeasts, LAB, and AAB, by
AAB (Adler et al., 2014). Although predicted before in an Acetobacter bin acting as an additional inoculation source of the necessary microor­
that was composed of different Acetobacter contigs, among which some ganisms, as the cooled wort transferred to the casks was still lacking
belonging to A. pasteurianus (De Roos et al., 2020), maltose/maltooli­ most fermentation and maturation microorganisms, thereby minimizing
gosaccharide consumption by A. lambici was probably not possible based batch-to-batch variations. Further, given the stable microbiota obtained,
on the current metagenomic analysis. Yet, a maltooligosyl trehalose it is obvious that the oak casks used provided a desired microaerobic
synthase and maltooligosyl trehalase could be attributed to A. malorum environment to achieve appropriate microbial community dynamics for
and A. cerevisiae, respectively, which may reflect their presence in a the fermentation and maturation of lambic beer wort. In particular, they
larger Acetobacter bin that was found in a previous study (De Roos et al., likely helped in preventing excessive growth of AAB and, therefore,
2020). The former two AAB species were present in the initial lambic excessive production of acetic acid and acetoin, of which high concen­
beer wort as well as during the enterobacterial fermentation phase, trations might cause flavor deviations in lambic beer. A functional
shown both culture-dependently and -independently. Furthermore, su­ analysis showed that the A. lambici MAG did not possess genes related to
crose consumption by A. lambici did not seem to be possible, as the sucrose and maltose/maltooligosaccharide consumption and the
concomitant gene was absent in the MAG. The gene encoding a putative glyoxylate shunt, but harbored genes involved in several acid tolerance
glucan synthase and hop resistance proteins in a refined bin related to mechanisms. Several typical traits, such as hop resistance, ropy pheno­
the P. damnosus MAG was in accordance with the strain’s ropy pheno­ type, and biogenic amine production, of P. damnosus were probably
type (Walling et al., 2005) and resistance to hop (Bergsveinson et al., plasmid-based and the P. damnosus MAG also possessed a ferulic acid
2015; Garcia-Garcia et al., 2017). These genes are plasmid-based decarboxylase, indirectly contributing to lambic beer flavor formation.
(Bergsveinson et al., 2015; Garcia-Garcia et al., 2017; Walling et al., The lack of a gene coding for glycerol 3-phosphate phosphatase in the
2005). Therefore, they were possibly removed during the manual D. bruxellensis and B. custersianus bins indicated that glycerol could not
curation of the P. damnosus MAG. Indeed, binning tools can have be produced, emphasizing the need for alternative external electron
problems with plasmids, as plasmids replicate autonomously, resulting acceptors for redox balancing.
in deviating relative abundance profiles (Beaulaurier et al., 2017). The
sequence composition profiles and GC contents can also differ (Beau­
laurier et al., 2017; Shintani et al., 2015). As the genes coding for his­ Declaration of competing interest
tidine decarboxylase, tyrosine decarboxylase, and enzymes involved in
the production of (R)-acetoin and (S)-acetoin from pyruvate were The authors declare that they have no known competing financial
located on the same bin, on contigs identified as P. damnosus or other interests or personal relationships that could have appeared to influence
Pediococcus species, they could be plasmid-based as well (Shintani et al., the work reported in this paper.
2015). All these genes were also detected during a previous meta­
genomic study of lambic beer, but they were not associated with a Data availability
plasmid location because another binning strategy was used (De Roos
et al., 2020). The presence of genes coding for ferulic acid decarboxylase Data will be made available on request.
indicated potential 4-vinyl compound production by P. damnosus, next
to their production by Saccharomyces species. However, previously, this Acknowledgements
gene has also been shown in D. bruxellensis but not in S. kudriavzevii (De
Roos et al., 2020). The degradation of starch, maltotriose, and maltose This work was financially supported by the Research Council of the
could be attributed to a sporulation-specific glucoamylase of the Vrije Universiteit Brussel (SRP7, IOF2442, and IOF3017 projects), the
Saccharomyces species, whereas maltooligosaccharide phosphorylase Hercules Foundation (projects UABR09004 and UAB13002), and the
could be responsible for maltooligosaccharide usage by D. bruxellensis, KMO-Portefeuille (projects 2014KMO084991, 2015KMO091056,
as all the concomitant genes were present in the respective bins. The 2016KMO149170, and 2017KMO112091, in collaboration with the
absence of a gene encoding glycerol 3-phosphate phosphatase in the brewery Oud Beersel). Part of the computational resources and services
D. bruxellensis and B. custersianus bins indicated that glycerol could not were provided by the Vrije Universiteit Brussel, the Flemish Super­
be produced, proving the need to use alternative external electron ac­ computer Centre (VSC), and the Research Foundation – Flanders (FWO-
ceptors, such as 4-vinyl compounds and acetoin to recuperate NAD+ Vlaanderen). LV and JDR were the recipients of a PhD fellowship of the
(Bongaerts et al., 2021; Steensels et al., 2015). The Acetobacter lambici Vrije Universiteit Brussel. We thank ing. Wim Borremans for providing
MAG and H. uvarum bin were not equipped with a glyoxylate cycle, technical assistance.
which indicated its species specificity within these genera (Illeghems
et al., 2013; Pelicaen et al., 2022; Sakurai et al., 2013; Seixas et al.,
References
2019). As has been shown for D. bruxellensis strains from wine, the
presence of two β-glucosidase genes could be responsible for the hy­ Adler, P., Frey, L.J., Berger, A., Bolten, C.J., Hansen, C.E., Wittmann, C., 2014. The key to
drolysis of β-1,3 (laminarin) and likely β-1,4 (cellobiose) and β-1,6 acetate: metabolic fluxes of acetic acid bacteria under cocoa pulp fermentation-
bonds (gentiobiose), respectively. Furthermore, this may have an in­ simulating conditions. Appl. Environ. Microbiol. 80, 4702–4716.
Alneberg, J., Bjarnason, B.S., de Bruijn, I., Schirmer, M., Quick, J., Ijaz, U.Z., Lahti, L.,
fluence on the flavoring capability of D. bruxellensis in lambic beer Loman, N.J., Andersson, A.F., Quince, C., 2014. Binning metagenomic contigs by
(Crauwels et al., 2014, 2015). Although mainly ascribed to coverage and composition. Nat. Methods 11, 1144–1146.

15
L. Vermote et al. International Journal of Food Microbiology 394 (2023) 110163

Altschul, S.F., Gish, W., Miller, W., Myers, E.W., Lipman, D.J., 1990. Basic local Fugelsang, K.C., Edwards, G.C., 2007. Wine Microbiology: Practical Applications and
alignment search tool. J. Mol. Biol. 215, 403–410. Procedures, Second edition. Springer, Boston, USA. 394 pp.
Anderson, M.J., 2017. Permutational multivariate analysis of variance (PERMANOVA). Garcia-Garcia, J.H., Damas-Buenrostro, L.C., Cabada-Amaya, J.C., Elias-Santos, M.,
Available online:. In: Balakrishnan, N., Colton, T., Everitt, B., Piegorsch, W., Pereyra-Alférez, B., 2017. Pediococcus damnosus strains isolated from a brewery
Ruggeri, F., Teugels, J.L. (Eds.), Wiley StatsRef: Statistics Reference Online, environment carry the horA gene. J. Inst. Brew. 123, 77–80.
pp. 1–15. https://doi.org/10.1002/9781118445112.stat07841. Gu, Z., Eils, R., Schlesner, M., 2016. Complex heatmaps reveal patterns and correlations
Andrews, S., 2018. FastQC: a quality control tool for high throughput sequence data. in multidimensional genomic data. Bioinformatics 32, 2847–2849.
Available online: http://www.bioinformatics.babraham.ac.uk/projects/fastqc/. Guzzon, R., Widmann, G., Malacarne, M., Nardin, T., Nicolini, G., Larcher, R., 2011.
Barata, A., Laureano, P., D’Antuono, I., Martorell, P., Stender, H., Malfeito-Ferreira, M., Survey of the yeast population inside wine barrels and the effects of certain
Querol, A., Loureiro, V., 2013. Enumeration and identification of 4-ethylphenol techniques in preventing microbial spoilage. Eur. Food Res. Technol. 233, 285–291.
producing yeasts recovered from the wood of wine aging barriques after different Harrell, F.E., 2021. Hmisc: Harrell miscellaneous. R package version 4.6-0. Available
sanitation treatments. J. Food Res. 2, 140–149. online: https://CRAN.R-project.org/package=Hmisc.
Beaulaurier, J., Zhu, S., Diekus, G., Mogno, I., Zhang, X.-S., Davis-Richardson, A., Hervé, M., 2021. RVAideMemoire: testing and plotting procedures for biostatistics. R
Canepa, R., Triplett, E.W., Faith, J.J., Sebra, R., Schadt, E.E., Fang, G., 2017. package version 0.9.79. Available online. https://CRAN.R-project.org/package
Metagenomic binning and association of plasmids with bacterial host genomes using =RVAideMemoire.
DNA methylation. Nat. Biotechnol. 36, 61–69. Holt, C., Yandell, M., 2011. MAKER2: an annotation pipeline and genome-database
Bergsveinson, J., Baecker, N., Pittet, V., Ziola, B., 2015. Role of plasmids in lactobacillus management tool for second-generation genome projects. BMC Bioinform. 12, 491.
brevis BSO 464 hop tolerance and beer spoilage. Appl. Environ. Microbiol. 81, Illeghems, K., De Vuyst, L., Papalexandratou, Z., Weckx, S., 2012. Phylogenetic analysis
1234–1241. of a spontaneous cocoa bean fermentation metagenome reveals new insights into its
Biohaz, 2011. European Food Safety Authority (EFSA) Panel on Biological Hazards. bacterial and fungal community diversity. PLoS One 7, e38040.
Scientific opinion on risk based control of biogenic amine formation in fermented Illeghems, K., De Vuyst, L., Weckx, S., 2013. Complete sequence and comparative
foods. Available online: EFSA J. 9, 2393. https://doi.org/10.2903/j.efsa.2011.2393 analysis of acetobacter pasteurianus 368B, a strain well-adapted to the cocoa bean
www.efsa.europa.eu/efsajournal. fermentation ecosystem. BMC Gen. 14, 526.
Bokulich, N.A., Bamforth, C.W., 2013. The microbiology of malting and brewing. Langmead, B., Salzberg, S.L., 2012. Fast gapped-read alignment with bowtie 2. Nat.
Microbiol. Mol. Biol. Rev. 77, 157–172. Methods 9, 357–359.
Bokulich, N.A., Bamforth, C.W., Mills, D.A., 2012. Brewhouse-resident microbiota are Lee, I., Kim, Y.O., Park, S.-C., Chun, J., 2016. OrthoANI: an improved algorithm and
responsible for multi-stage fermentation of american coolship ale. PLoS One 7, software for calculating average nucleotide identity. Int. J. Syst. Evol. Microbiol. 66,
e35507. 1100–1103.
Bokulich, N.A., Bergsveinson, J., Ziola, B., Mills, D.A., 2015. Mapping microbial Li, H., 2018. Minimap2: pairwise alignment for nucleotide sequences. Bioinformatics 34,
ecosystems and spoilage-gene flow in breweries highlights patterns of contamination 3094–3100.
and resistance. elife 4, e04634. Li, D., Liu, C.M., Luo, R., Sadakane, K., Lam, T.W., 2015. MEGAHIT: an ultra-fast single-
Bongaerts, D., De Roos, J., De Vuyst, L., 2021. Technological and environmental features node solution for large and complex metagenomics assembly via succinct de bruijn
determine the uniqueness of the lambic beer microbiota and production process. graph. Bioinformatics 31, 1674–1676.
Appl. Environ. Microbiol. 87 e00612-21. Lisdiyanti, P., Kawasaki, H., Seki, T., Yamada, Y., Uchimura, T., Komagata, K., 2001.
Bouchez, A., De Vuyst, L., 2022. Acetic acid bacteria in sour beer production: friend or Identification of acetobacter strains isolated from indonesian sources, and proposals
foe? Front. Microbiol. 13, 957167. of acetobacter syzygii sp. nov., acetobacter cibinongensis sp. Nov., and acetobacter
Buchfink, B., Xie, C., Huson, D.H., 2015. Fast and sensitive protein alignment using orientalis sp. nov. J. Gen. Appl. Microbiol. 47, 119–131.
DIAMOND. Nat. Methods 12, 59–60. Lynch, K.M., Zannini, E., Wilkinson, S., Daenen, L., Arendt, E.K., 2019. Physiology of
Canas, S., Caldeira, I., Anjos, O., Lino, J., Soares, A., Belchior, A.P., 2016. acetic acid bacteria and their role in vinegar and fermented beverages. Compr. Rev.
Physicochemical and sensory evaluation of wine brandies aged using oak and Food Sci. Food Saf. 18, 587–625.
chestnut wood simultaneously in wooden barrels and in stainless steel tanks with Menzel, P., Ng, K.L., Krogh, A., 2016. Fast and sensitive taxonomic classification for
staves. Int. J. Food Sci. Technol. 51, 2537–2545. metagenomics with kaiju. Nat. Commun. 7, 1–9.
Cleenwerck, I., González, Á., Camu, N., Engelbeen, K., De Vos, P., De Vuyst, L., 2008. Morrissey, W.F., Davenport, B., Querol, A., Dobson, A.D., 2004. The role of indigenous
Acetobacter fabarum sp. nov., an acetic acid bacterium from a ghanaian cocoa bean yeasts in traditional irish cider fermentations. J. Appl. Microbiol. 97, 647–655.
heap fermentation. Int. J. Syst. Evol. Microbiol. 58, 2180–2185. Mouillot, D., Leprêtre, A., 1999. A comparison of species diversity estimators. Res. Popul.
Cleenwerck, I., Vandemeulebroecke, K., Janssens, D., Swings, J., 2002. Re-examination Ecol. 41, 203–215.
of the genus acetobacter, with descriptions of acetobacter cerevisiae sp. nov. and Nakano, S., Fukaya, M., Horinouchi, S., 2006. Putative ABC transporter responsible for
Acetobacter malorum sp. nov. Int. J. Syst. Evol. Microbiol. 52, 1551–1558. acetic acid resistance in acetobacter aceti. Appl. Environ. Microbiol. 72, 497–505.
Crauwels, S., Van Assche, A., de Jonge, R., Borneman, A.R., Verreth, C., Troels, P., De Oelofse, A., Pretorius, I.S., du Toit, M., 2008. Significance of brettanomyces and dekkera
Samblanx, G., Marchal, K., Van de Peer, Y., Willems, K.A., Verstrepen, K.J., during winemaking: a synoptic review. S. Afr. J. Enol. Vitic. 29, 128–144.
Curtin, C.D., Lievens, B., 2015. Comparative phenomics and targeted use of Oksanen, J., Blanchet, F.G., Friendly, M., Kindt, R., Legendre, P., McGlinn, D.,
genomics reveals variation in carbon and nitrogen assimilation among different Minchin, P.R., O’Hara, R.B., Simpson, G.L., Solymos, P., Stevens, M.H.H., Szoecs, E.,
brettanomyces bruxellensis strains. Appl. Microbiol. Biotechnol. 99, 9123–9134. Wagner, H., 2019. Vegan community ecology package. R package version 2.5-4.
Crauwels, S., Zhu, B., Busschaert, P., De Samblanx, G., Marchal, K., Willems, K.A., Available online: https://CRAN.R-project.org/package=vegan.
Verstrepen, K.J., Lievens, B., 2014. Assessing genetic diversity among brettanomyces Papalexandratou, Z., Lefeber, T., Bahrim, B., Lee, O.S., Daniel, H.-M., De Vuyst, L., 2013.
yeasts by DNA fingerprinting and whole-genome sequencing. Appl. Environ. Hanseniaspora opuntiae, Saccharomyces cerevisiae, lactobacillus fermentum, and
Microbiol. 80, 4398–4413. acetobacter pasteurianus predominate during well-performed malaysian cocoa bean
De Keersmaecker, J., 1996. The mystery of lambic beer. Sci. Am. 275, 375–385. box fermentations, underlining the importance of these microbial species for a
del Alamo-Sanza, M., Nevares, I., 2014. Recent advances in the evaluation of the oxygen successful cocoa bean fermentation process. Food Microbiol. 35, 73–85.
transfer rate in oak barrels. J. Agric. Food Chem. 62, 8892–8899. Pelicaen, R., Gonze, D., Teusink, B., De Vuyst, L., Weckx, S., 2019. Reconstruction of
Del Campo, G., Santos, J., Berregi, I., Velasco, S., Ibarburu, I., Duenas, M., Irastorza, A., acetobacter pasteurianus 386B, a candidate functional starter culture for cocoa bean
2003. Ciders produced by two types of presses and fermented in stainless steel and fermentation. Front. Microbiol. 10, 2801.
wooden vats. J. Inst. Brew. 109, 342–348. Pelicaen, R., Weckx, S., Gonze, D., De Vuyst, L., 2022. Application of comparative
De Roos, J., De Vuyst, L., 2019. Microbial acidification, alcoholization, and aroma genomics of acetobacter species facilitates genome-scale metabolic reconstruction of
production during spontaneous lambic beer production. J. Sci. Food Agric. 99, the acetobacter ghanensis LMG 23848T and acetobacter senegalensis 108B cocoa
25–38. strains. Front. Microbiol. 13, 1060160.
De Roos, J., Vandamme, P., De Vuyst, L., 2018a. Wort substrate consumption and Pothakos, V., Illeghems, K., Laureys, D., Spitaels, F., Vandamme, P., De Vuyst, L., 2016.
metabolite production during lambic beer fermentation and maturation explain the Acetic acid bacteria in fermented food and beverage ecosystems. In: Matsushita, K.,
successive growth of specific bacterial and yeast species. Front. Microbiol. 9, 2763. Toyama, H., Tonouchi, N., Okamoto-Kainuma, A. (Eds.), Acetic Acid Bacteria:
De Roos, J., Van Der Veken, D., De Vuyst, L., 2019. The interior surfaces of wooden Ecology and Physiology. Springer, Tokyo, Japan, pp. 73–99.
barrels are an additional microbial inoculation source for lambic beer production. RStudio Team, 2020. RStudio: integrated development for R. Available online. RStudio,
Appl. Environ. Microbiol. 85 e02226-18. PBC, Boston, MA, USA. http://www.rstudio.com.
De Roos, J., Verce, M., Aerts, M., Vandamme, P., De Vuyst, L., 2018b. Temporal and Sakurai, K., Yamazaki, S., Ishii, M., Igarashi, Y., Arai, H., 2013. Role of the glyoxylate
spatial distribution of the acetic acid bacterium communities throughout the wooden pathway in acetic acid production by acetobacter aceti. J. Biosci. Bioeng. 115,
casks used for the fermentation and maturation of lambic beer underlines their 32–36.
functional role. Appl. Environ. Microbiol. 84 e02846-17. Schmieder, R., Edwards, R., 2011. Quality control and preprocessing of metagenomic
De Roos, J., Verce, M., Weckx, S., De Vuyst, L., 2020. Temporal shotgun metagenomics datasets. Bioinformatics 27, 863–864.
revealed the potential metabolic capabilities of specific microorganisms during Seemann, T., 2014. Prokka: rapid prokaryotic genome annotation. Bioinformatics 30,
lambic beer production. Front. Microbiol. 11, 1692. 2068–2069.
Dysvik, A., Leanti La Rosa, S., De Rouck, G., Rukke, E.-O., Westereng, B., Wicklund, T., Seixas, I., Barbosa, C., Mendes-Faia, A., Güldener, U., Tenreiro, R., Mendes-Ferreira, A.,
2020. Microbial dynamics in traditional and modern sour beer production. Appl. Mira, N.P., 2019. Genome sequence of the non-conventional wine yeast
Environ. Microbiol. 86 e00566-20. Hanseniaspora guilliermondii UTAD222 unveils relevant traits of this species and of
Eren, A.M., Esen, Ö.C., Quince, C., Vineis, J.H., Morrison, H.G., Sogin, M.L., Delmont, T. the hanseniaspora genus in the context of wine fermentation. DNA Res. 26, 67–83.
O., 2015. Anvi’o : an advanced analysis and visualization platform for ‘omics data. Shintani, M., Sanchez, Z.K., Kimbara, K., 2015. Genomics of microbial plasmids:
PeerJ 3, e1319. classification and identification based on replication and transfer systems and host
taxonomy. Front. Microbiol. 6, 242.

16
L. Vermote et al. International Journal of Food Microbiology 394 (2023) 110163

Slowikowski, K., 2020. ggrepel: automatically position non-overlapping text labels with Verce, M., De Vuyst, L., Weckx, S., 2019. Shotgun metagenomics of a water kefir
‘ggplot2’. R package version 0.8.2. Available online: https://CRAN.R-project. fermentation ecosystem reveals a novel oenococcus species. Front. Microbiol. 10,
org/package=ggrepel. 479.
Sombolestani, A.S., Cleenwerck, I., Cnockaert, M., Borremans, W., Wieme, A.D., De Verce, M., De Vuyst, L., Weckx, S., 2020. The metagenome-assembled genome of
Vuyst, L., Vandamme, P., 2020. Novel acetic acid bacteria from cider fermentations: candidatus oenococcus aquikefiri from water kefir represents the species oenococcus
acetobacter conturbans sp. nov. and acetobacter fallax sp. nov. Int. J. Syst. Evol. sicerae. Food Microbiol. 88, 103402.
Microbiol. 70, 6163–6171. Verce, M., Schoonejans, J., Hernandez Aguirre, C., Molina-Bravo, R., De Vuyst, L.,
Spitaels, F., Li, L., Wieme, A., Balzarini, T., Cleenwerck, I., Van Landschoot, A., De Weckx, S., 2021. A combined metagenomics and metatranscriptomics approach to
Vuyst, L., Vandamme, P., 2014a. Acetobacter lambici sp. Nov., isolated from unravel costa rican cocoa box fermentation processes reveals yet unreported
fermenting lambic beer. Int. J. Syst. Evol. Microbiol. 64, 1083–1089. microbial species and functionalities. Front. Microbiol. 12, 641185.
Spitaels, F., Wieme, A.D., Janssens, M., Aerts, M., Daniel, H.-M., Van Landschoot, A., De Vermote, L., Verce, M., De Vuyst, L., Weckx, S., 2018. Amplicon and shotgun
Vuyst, L., Vandamme, P., 2014b. The microbial diversity of traditional metagenomic sequencing indicates that microbial ecosystems present in cheese
spontaneously fermented lambic beer. PLoS One 9, e95384. brines reflect environmental inoculation during the cheese production process. Int.
Spitaels, F., Wieme, A.D., Janssens, M., Aerts, M., Van Landschoot, A., De Vuyst, L., Dairy J. 87, 44–53.
Vandamme, P., 2015. The microbial diversity of an industrially produced lambic Walling, E., Gindreau, E., Lonvaud-Funel, A., 2005. A putative glucan synthase gene dps
beer shares members of a traditionally produced one and reveals a core microbiota detected in exopolysaccharide-producing pediococcus damnosus and oenococcus
for lambic beer fermentation. Food Microbiol. 49, 23–32. oeni strains isolated from wine and cider. Int. J. Food Microbiol. 98, 53–62.
Spitaels, F., Wieme, A.D., Snauwaert, I., De Vuyst, L., Vandamme, P., 2017. Microbial White, B., Smyth, M.R., Lunte, C.E., 2017. Determination of phenolic acids in a range of
ecology of traditional beer fermentations. In: Bokulich, N., Bamforth, C. (Eds.), irish whiskies, including single pot stills and aged single malts, using capillary
Brewing Microbiology: Current Research, Omics and Microbial Ecology. Caister electrophoresis with field amplified sample stacking. Anal. Methods 9, 1248–1252.
Academic Press, Poole, UK, pp. 179–196. Wickham, H., Averick, M., Bryan, J., Chang, W., McGowan, L.D., François, R.,
Steensels, J., Daenen, L., Malcorps, P., Derdelinckx, G., Verachtert, H., Verstrepen, K.J., Grolemund, G., Hayes, A., Henry, L., Hester, J., Kuhn, M., Pedersen, T.L., Miller, E.,
2015. Brettanomyces yeasts – from spoilage organisms to valuable contributors to Bache, S.M., Müller, K., Ooms, J., Robinson, D., Seidel, D.P., Spinu, V., Takahashi, K.,
industrial fermentations. Int. J. Food Microbiol. 206, 24–38. Vaughan, D., Wilke, C., Woo, K., Yutani, H., 2019. Welcome to the tidyverse. J. Open
Swaffield, C.H., Scott, J.A., 1995. Existence and development of natural microbial Source Softw. 4, 1686.
populations in wooden storage vats used for alcoholic cider maturation. J. Am. Soc. Wood, D.E., Lu, J., Langmead, B., 2019. Improved metagenomic analysis with kraken 2.
Brew. Chem. 53, 117–120. Genome Biol. 20, 257.
Swaffield, C.H., Scott, J.A., Jarvis, B., 1997. Observations on the microbial ecology of Yang, H., Yu, Y., Fu, C., Chen, F., 2019. Bacterial acid resistance toward organic weak
traditional alcoholic cider storage vats. Food Microbiol. 14, 353–361. acid revealed by RNA-seq transcriptomic analysis in acetobacter pasteurianus. Front.
Van Oevelen, D., Spaepen, M., Timmermans, P., Verachtert, H., 1977. Microbiological Microbiol. 10, 1616.
aspects of spontaneous wort fermentation in the production of lambic and gueuze. Zheng, J., Wittouck, S., Salvetti, E., Franz, C.M.A.P., Harris, H.M.B., Mattarelli, P.,
J. Inst. Brew. 83, 356–360. O’Toole, P.W., Pot, B., Vandamme, P., Walter, J., Watanabe, K., Wuyts, S., Felis, G.
Verachtert, H., Derdelinckx, G., 2005. Acidic beers: enjoyable reminiscences of the past. E., Gänzle, M.G., Lebeer, S., 2020. A taxonomic note on the genus lactobacillus:
Cerevisia 30, 38–47. description of 23 novel genera, emended description of the genus lactobacillus
Verachtert, H., Iserentant, D., 1995. Properties of belgian acidic beers and their beijerinck 1901, and union of lactobacillaceae and leuconostocaceae. Int. J. Syst.
microflora. I. The production of gueuze and related refreshing acid beers. Cerevisia Evol. Microbiol. 70, 2782–2858.
20, 37–41.

17

You might also like