Dynamics of dark fermentation microbial communities in the light of lactate and butyrate production

Background This study focuses on the processes occurring during the acidogenic step of anaerobic digestion, especially resulting from nutritional interactions between dark fermentation (DF) bacteria and lactic acid bacteria (LAB). Previously, we have confirmed that DF microbial communities (MCs) that fed on molasses are able to convert lactate and acetate to butyrate. The aims of the study were to recognize the biodiversity of DF-MCs able and unable to convert lactate and acetate to butyrate and to define the conditions for the transformation. Results MCs sampled from a DF bioreactor were grown anaerobically in mesophilic conditions on different media containing molasses or sucrose and/or lactate and acetate in five independent static batch experiments. The taxonomic composition (based on 16S_rRNA profiling) of each experimental MC was analysed in reference to its metabolites and pH of the digestive liquids. In the samples where the fermented media contained carbohydrates, the two main tendencies were observed: (i) a low pH (pH ≤ 4), lactate and ethanol as the main fermentation products, MCs dominated with Lactobacillus, Bifidobacterium, Leuconostoc and Fructobacillus was characterized by low biodiversity; (ii) pH in the range 5.0–6.0, butyrate dominated among the fermentation products, the MCs composed mainly of Clostridium (especially Clostridium_sensu_stricto_12), Lactobacillus, Bifidobacterium and Prevotella. The biodiversity increased with the ability to convert acetate and lactate to butyrate. The MC processing exclusively lactate and acetate showed the highest biodiversity and was dominated by Clostridium (especially Clostridium_sensu_stricto_12). LAB were reduced; other genera such as Terrisporobacter, Lachnoclostridium, Paraclostridium or Sutterella were found. Butyrate was the main metabolite and pH was 7. Shotgun metagenomic analysis of the selected butyrate-producing MCs independently on the substrate revealed C.tyrobutyricum as the dominant Clostridium species. Functional analysis confirmed the presence of genes encoding key enzymes of the fermentation routes. Conclusions Batch tests revealed the dynamics of metabolic activity and composition of DF-MCs dependent on fermentation conditions. The balance between LAB and the butyrate producers and the pH values were shown to be the most relevant for the process of lactate and acetate conversion to butyrate. To close the knowledge gaps is to find signalling factors responsible for the metabolic shift of the DF-MCs towards lactate fermentation. Video Abstract Supplementary Information The online version contains supplementary material available at 10.1186/s40168-021-01105-x.

showed the highest biodiversity and was dominated by Clostridium (especially Clostridium sensu stricto 12). LAB were reduced, other genera such as Terrisporobacter, Lachnoclostridium, Paraclostridium or Sutterella were found. Butyrate was the main metabolite and pH was 7. WGS analysis of the selected butyrate-producing microbial communities independently on the substrate, revealed C. tyrobutyricum as a dominant Clostridium species.
Conclusions: The batch tests revealed dynamics of metabolic activity and composition of DF microbial communities dependent on fermentation conditions. The results expand our knowledge on lactate to butyrate conversion by DF microbial communities. The relevant factor for conversion of lactate and acetate to butyrate in the presence of carbohydrates is pH in the range 5-6 and the balance between LAB (especially Lactobacillus), lactate and acetate producers (Bi dobacterium) and butyrate producers (mainly Clostridium) as well Prevotella. The pH below 4 and ethanol concentration might be the signalling factors responsible for metabolic shift of the dark fermentation microbial communities towards lactate fermentation.

Background
Anaerobic digestion (AD) is a complex and multistep conversion of biomass to methane and carbon dioxide resulting from metabolic activity and nutritional interactions between many groups of microorganisms. It involves four main stages: hydrolysis of polymeric organic matter to monomers, acidogenesis, acetogenesis and methanogenesis. [1][2][3] This study focuses on the processes during acidogenesis when the products of hydrolysis are converted to non-gaseous short-chain fatty acids, alcohols, aldehydes and the gases, carbon dioxide and hydrogen [4]. The dominant end-products of the fermentation process determine the type of fermentation. A part of acidogenesis, hydrogen-yielding fermentations (dark fermentation) are considered to be one of the most attractive alternative biological methods of hydrogen (biohydrogen) production. The main types of hydrogen-yielding fermentation under mesophilic conditions, especially from carbohydrates degradation, are acetic/butyric acid fermentation (Clostridium-type fermentation) and mixed-acid fermentation (Enterobacteriaceae-type fermentation) [3,5,6]. Hydrogen can be also produced during transformation of products other fermentation types. Fermentative biohydrogen production offers additional advantage of potentially using various waste streams from different industries as feedstock such as sugar beet industry. Optimization of biohydrogen yield during acidogenesis is challenging and requires better understanding of the microbial community dynamics in bioreactors and their metabolic substrate conversion along hydrogen-promoting pathways.
In a multispecies microbial communites, nutrient utilization is a complex process and frequently involves competition and symbiotic cross-feeding (syntrophy) [7]. The former is when two or more groups of microorganisms compete for susbtarte that usually leads to a temporary increase in relative abundance of one interacting partner over the other. The latter is when the metabolic products yielded by one microbe constitute energy resource or nutrients supporting growth for another one. Therefore, the analysis of nutrient metabolism in fermentative processes should integrate the dynamics of microbial community composition with metabolic nutrient conversion.
Lactic acid bacteria (LAB) are ubiquitous in the environment, they accompany the plant biomass to anaerobic bioreactors, and constitute a relevant component of acidogenic microbial communities. It is commonly believed that development of LAB in bioreactors inhibit hydrogen production due to substrate competition and/or excretion of bacteriocins that inhibit growth of other bacteria. In the homolactic fermentation two molecules of pyruvate formed during glycolysis are converted to lactate; in heterolactic fermentation, one molecule of pyruvate is converted to lactate and the other to ethanol and carbon dioxide [8]. Substrate competition includes replacement of hydrogen fermentation by lactic acid or ethanol fermentation. Decrease in hydrogen production is observed with simultaneous increase of lactic acid and ethanol concentrations among nongaseous fermentation products [9][10][11][12][13].
On the other hand, cross-feeding of lactate involves the conversion of lactate and acetate to butyrate, hydrogen and carbon dioxide. It is a symbiotic nutritional interaction recognized between lactate-and acetate-producing bacteria and butyrate producers. This phenomena of metabolic interactions between different bacterial groups was described in the gut of many animals including in the human gut. The end product, butyrate, is a crucial molecule necessary in maintaining gut health, homeostasis, and serving as an energy source for the colonic epithelial cells [14][15][16][17]. Cross-feeding of lactate is also observed in dark fermentation bioreactors during fermentative conversion of organic substrates to biohydrogen both in mesophilic [18][19][20][21][22][23] and termophilic conditions [24,25].
The studies on fermentation of agave bagasse, tequila vinasse and wastewater from nixtamalization supplied data supporting the thesis that cross-feeding of lactate is signi cant in the microbial communities of dark fermentation bioreactors. The authors postulated that conversion of lactate and acetate to butyrate is the main pathway of biohydrogen production [19][20][21][22][23]. A speci c succession of bacteria was observed in batch experiments. In the rst stage, the substrate was processed to acetate and lactate, which were transformed to butyrate and hydrogen in the second stage. The pH was an important factor ensuring balance and syntrophy between lactate-and butyrate-producers [19][20][21][22][23]. Studies on thermophilic dark fermentation of sugarcane vinasse also showed lactate as the primary substrate for biohydrogen production and relevance of pH in this process [24,25]. Cross-feeding of lactate was also observed in reduced microbial communities composed of two components: butyrateproducing Clostridium beijerinckii and lactate-producer Yokenella regensburgei [26]. Furthermore, pure cultures of Clostridium acetobutylicum [27], Butyribacterium methylotrophicum [28], Clostridium diolis [29], Clostridium buttyricum [18] and Clostridium tyrobutyricum [30,31] anaerobically grown in media with acetate and lactate as exclusive carbon sources produced carbon dioxide, hydrogen and butyrate.
Our previous work demonstrated that dark fermentation microbial communities fed molasses under mesophilic conditions are able to convert lactate and acetate to butyrate in batch experiments [18]. Here we propose a logical continuation and extension of the previously published studies aimed at (i) recognition of biodiversity and dynamics of dark fermentation microbial communities able and unable to convert lactate and acetate to butyrate and (ii) de nition of the conditions for the process of transformation. We examined batch cultures of dark fermentation microbial communities grown in media containing molasses or sucrose supplemented with lactate and acetate, or a mixture of lactate and acetate without added carbohydrates. The balance between lactic acid bacteria and the butyrateproducing clostridia and the pH values were shown to be the most relevant for the process of lactate and acetate conversion to butyrate. The putative main lactate producers and lactate and acetate utilizers were identi ed.
Since fermentation processes are ubiquitous in anaerobic environments, butyrate and lactate producers are found in anaerobic digesters and among the gut microbiota, the results obtained in this study should interest both the researchers dealing with studies on (i) AD and production of gaseous biofuels as well as (ii) the butyrate production by the gut bacteria.

Experimental set-up for the examination of lactate to butyrate transformation in batch experiments
Tests on the transformation of lactate and acetate to butyrate were conducted in static batch experiments, analogous to those described previously [18], in 250-ml Erlenmayer asks for 18 days in a Vinyl Anaerobic Chamber (Coy Laboratory Products, Inc.) without shaking at 30°C. Five-milliliter samples of microbial community taken from the dark fermentation hydrogen producing packed bed reactor described previously was used as inoculum. The liquid growth medium (200 mL) was M9 after 10-fold dilution, without glucose, supplemented with 1% sucrose (Chempur Poland) or molasses at the concentration corresponding to 1% sucrose; sodium lactate (VWR Chemicals) 7.41 g/L; sodium acetate (Chempur Poland) 3.5 g/L; and 0.2% yeast extract (BD Bioscences USA). The following combination of nutrients were tested: molasses (Experiment M); molasses plus sodium lactate (Experiment ML); molasses plus sodium lactate and sodium acetate (Experiment MLA); sodium lactate and sodium acetate (Experiment LA); sucrose plus sodium lactate and sodium acetate (Experiment SLA). All the variants were tested in three independent repetitions designated as A, B, and C. Molasses is a by-product of sugar processing from sugar beets. It contains 50% of sucrose. Other components are water, glucose, fructose, amino acids, mineral salts, betaine, B vitamins, glutamic acid, inositol, nitrogen compounds. In this study, molasses came from the Dobrzelin Sugar Factory, branch of the Polish Sugar Company "Polski Cukier".
Starting pH of all media was 7.0. No additional means of pH control were used. Before inoculation, 4-5 sterile slag pieces were placed in each ask to be covered by bacterial bio lm. Bacterial growth in batch cultures was determined by OD 600 nm measurements. After every 3 days of incubation, the Erlenmayer asks were shaken, the digestive liquids were removed and the respective fresh media for further growth were added. After every passage (on days 3, 6, 9, 12, 15 and 18) the digestive liquids were centrifuged (7,000 × g for 10 min), the supernatants analyzed and the pellets used for total microbial DNA isolation as described below. Composition of the selective media, lactate and acetate concentration were selected based on the data from previous studies [18,29,32,33].

Analytical methods
The pH of the media and the digestive liquids was measured using a standard pH meter (ELMETRON model CP-502, Poland). Samples were centrifuged (7,000 × g for 10 min, 10°C) to remove microbial cells and debris, and concentrations of carbohydrates (sucrose, glucose and fructose), short-chain fatty acids, ethanol were determined. The carbohydrates and ethanol were analyzed using high performance liquid chromatography (HPLC) with refractometric detection (Waters HPLC system: Waters 2695 -Separations Module, Waters 2414 -Refractive Index Detector, a thermostat for column, and 300×6.5 mm Sugar Pak I column with guard column). The determination of carbohydrates was carried at 90°C, and ethanol at 70°C. The sample (10 µL) were injected onto the column and eluted for 20 min with an isocratic ow of 0.1 mM calcium disodium salt EDTA (0.5 mL/min). Short-chain fatty acids were analyzed by HPLC with photometric detection (Waters HPLC system as above, Waters 2996 -Photodiode Array Detector, and 300×7.8 mm Aminex HPX-87 H column with guard column at 30°C). The samples were eluted for 45 min with an isocratic ow (0.6 mL/min) of 4 mM sulphuric acid.
For the statistical analysis of bacterial growth (OD 600nm ), pH of the digestive liquids and the non-gaseous fermentation products, the STATISTICA (version 10.0) computer software (StatSoft, Inc.) was used. All variables were examined for normality and homogeneity of variance. Tukey's HSD (honestly signi cant difference) test was applied after ANOVA analysis to compare statistical signi cance among the variables in experiments. Statistical significance was considered at p < 0.05.

Microbial DNA extraction
The total DNA was isolated from the pellets obtained after centrifugation (see above) of 2 ml-samples of the digestive liquids taken after 3, 6, 9, 12, 15 and 18 days of the experiment. From each culture two samples (duplicates) were taken. DNA was extracted and puri ed using a DNeasy PowerSoil Pro Kit (Qiagen, Cat No. 47014) according to the manufacturer's protocol. Cell lysis was done using Vortex-Genie 2 equipped with a Vortex Adapter for 1.5-2 ml tubes (cat. no. 13000-V1-24). DNA was stored at −20 °C. The nal samples of DNA extracted from the two replicates were pooled.

16S rRNA amplicon sequencing and data analysis
The hypervariable V4 region of the 16S rRNA gene was ampli ed from each sample using barcoded reverse primers (806R) and common forward primer (515F). Both reverse and forward primes were extended with the sequencing primer pads, linkers, and Illumina adapters [34], and with MyFi™ Mix 12500). Pooled amplicons were diluted and denatured with 0.1N NaOH. The library was sequenced at the Microbiome Core at the Steele Children's Research Center, University of Arizona, using MiSeq platform (Illumina) and custom primers [34]. Due to the limited sequence diversity among 16S rRNA amplicons, 5% of the PhiX Sequencing Control V3 (Illumina, Cat No. FC-110-3001) made from phiX174, was used to spike the library to increase diversity. The raw sequencing data were demultiplexed using the idemp script (https://github.com/yhwu/idemp). Filtering, dereplication, chimera identi cation, and merging of pairedend reads were performed with dada2 [35]. The amplicon sequence variants (ASVs) taxonomy was assigned using the Ribosomal Database Project (RDP) classi er [36] against SILVA database release 132 [37].
Taxonomic richness and evenness (Shannon and Simpson indices) were calculated and statistical signi cance within in each experiment was calculated using Kruskal-Wallis rank sum test followed by Dunn's multiple comparison test with Bonferroni correction (dunn.test R package).
Differences in microbial communities were evaluated using non-metric multidimensional scaling (NMDS) ordination analysis on Bray-Curtis distances followed by permutational multivariate analysis of variance (PERMANOVA) to analyze the contribution of different metadata variables to microbial communities composition dissimilarities. Also, to investigate and visualize the association between metadata variables and their effect on the species distribution pattern, redundancy analysis was used in vegan R package [38]. The obtained results were visualized with a ggplot2 (ver 3.3.2) package [39] and with heatplus (ver. 3.11) R package [40].
The raw sequences generated in this study have been deposited in NCBI databases with the accession number PRJNA645198.

Whole Metagenome Sequencing (WGS) and data analyses
The libraries for WGS were constructed for the selected samples from the static batch experiments using QIASeq FX DNA Library Kit (QIAGEN) according to the manufacturer's protocol. Brie y, 50 ng of DNA from each sample (or pooled samples) was randomly fragmented with FX Enzyme Mix followed by the adapter ligation step. Both i5 and i7 adapters contain unique 8 nucleotide barcodes. After removing free adapters from the reaction with AMPure XP magnetic beads, all individual libraries were ampli ed by PCR followed by the size selection with 2-step puri cation (the negative selection followed by the positive selection step) with AMPure XP magnetic beads. The quality and quantity of all libraries was determined with Agilent 4150 TapeStation DNA analyzer. The libraries were normalized and pooled, and the sequencing was performed on the Illumina NextSeq 500/550 platform using Illumina 400M HighOutput 300 cycles sequencing chemistry.
The raw sequences generated in this study have been deposited in NCBI databases with the accession number PRJNA640235.

General characteristics of the microbial communities in static batch experiments
To examine the capabilities of dark fermentation microbial communities to convert lactate and acetate to butyrate, ve independent static batch experiments in three replicates were performed. Each one was inoculated with the same community derived from hydrogen-producing packed-bed reactors described previously [18,46]. The experiments provided different carbon sources as shown in Table 1: molasses (Experiment M), molasses supplemented with lactate (Experiment ML), molasses supplemented with lactate and acetate (Experiment MLA), sucrose supplemented with lactate and acetate (Experiment SLA) and lactate and acetate (Experiment LA). The batch experiments were maintained for 18 days and passaged every 3 days. Bacterial growths measured by OD 600nm of the digestive liquids after every passage are presented in Table 1. The results clearly show that sucrose stimulates bacterial growth. The densities were higher (OD 600nm after every three days ≈ 2-3) when bacteria grew on the media containing sucrose (either from molasses or used a pure additive; Experiments M, ML, MLA and SLA) compared to Experiment LA when lactate and acetate were provided as an exclusive carbon source (OD 600nm after every three days ≈ 1), 0.001 < p < 0.005 between LA group and any other group (Tukey's HSD test; Table 1, Additional le 1). Interestingly, in comparison to molasses and lactate alone (Experiment ML), addition of acetate in Experiment MLA increased bacterial growth on days 6, 9, and 12 (p < 0.05, Tukey's HSD test). Differences in bacterial growth were also found on day 9 between Experiments M and ML as well as between Experiments M and SLA (p < 0.05, Tukey's HSD test).
Biodiversity and microbial changes in all the experiments were analysed by sequencing of the 16S V4 amplicon pro ling. Total of 119 samples were sequenced in one MiSeq run, and 7,431 ASVs were detected. After chimera identi cation and removal, 93.15% ASVs remained. 29 samples from an unrelated project were ltered out, and the remaining 90 samples were further analyzed. For detailed taxonomic assignments see Additional le 2. All negative controls for the V4 ampli cation by PCR (collection day 0 for each experiment) did not show any ampli cation and these controls were removed from analysis during the quality control steps due to insu cient number of reads. Alpha diversity analysis revealed that the microbial communities are moderately rich in taxa, and that communities grown in media supplemented with molasses only or molasses and lactate (Experiments M and ML) had the lowest diversity as compared to the inoculum alone or to other groups (  Table 2).

Analysis of metabolites and microbial community composition after the initial 3 days of fermentation
After the intial 3 days of fermentation, we found no statistically signi cant differences in the concentration of the analyzed non-gaseous fermentation products among the batch experiments where  Table 2 and Additional le 4.
The dynamics of the fermenation process were followed over four additional passages until 18 days post-inoculation and showed considerable differences between experimental groups. These are discussed in detail in the following sections.

Dynamics of fermentation processes with molasses only (Experiment M)
When molasses were fermented without exogenous SFCAs, pH of the digestive liquids after six days dropped below 4 and remained in the 3.6 -3.9 range (Table 2, Figure 4, Additional le 4). During the whole experiment, the main non-gaseous fermentation products were ethanol and lactate. Lactate was the only metabolite that signi cantly changed over time (ANOVA, p = 0.000015), with a gradual increase from day 3 to day 18 (2.6, 3.9, 5.2, 6.3, 5.2 and 5.6 g/L, respectively). The results of the detailed statistical comparisons are presented in Table 2 and Additional le 1. Between 6-18 days, the concentrations of ethanol remained relatively stable at 5.6, 3.8, 3.1, 4.1 and 4.9 g/L. The concentrations of butyrate and acetate were low (≤1 g/L) throughout the experiment (Table 2, Figure 4).

Dynamics of fermentation processes with molasses supplemented with lactate (ML)
After the addition of lactate, pH of the digestive liquids after the rst passage (days 6-18) remained in the range 4.4-4.6. The concentrations of ethanol decreased steadily (ANOVA, p = 0.004) from 6.7 g/L on day 3 to 3.4, 2.7 and 2.8 g/L on days 9, 15 and 18, respectively. Concentration of lactate varied over time (ANOVA, p = 0.006). It peaked on day 6 at 12 g/L and decreased to 6.3 and 6.8 g/L on days 15 and 18, respectively. The concentration of butyrate gradually increased (ANOVA, p = 0.006) from 0.03 g/L on day 3 to 3.2 g/L on days 15 and 18. The concentration of acetate remained low and steady (≤1 g/L) throughout the experiment (Table 2, Figure 2, Additional les 1 and 4).
Dynamics of fermentation processes with molasses or sucrose supplemented with lactate (MLA and SLA) In the MLA and SLA Experiments, the media contained molasses as a source or sucrose or pure sucrose, both supplemented with lactate and acetate. These two experiments are described together due to similar tendencies observed, which re ects the dominant efect of lactate/acetate supplementation over the source of sucrose (Figures 3 and 4, Table 2, Additional les 1 and 4). In the MLA Experiment, the pH of the digestive liquids changed from 4.6 on day 3 to 5.4 and 5.5 on days 6 and 9 (p = 0.0002, p = 0.0002, respectively; Tukey's HSD test). In the SLA Experiment, the pH changed from 4.6 on day 3 to 6.2 and 5.8 on days 6 and 9 (p=0.002, p = 0.02, respectively; Tukey's HSD test). In the MLA Experiment, the pH increase was associated with increased butyrate concentration, from 0.2 g/L on day 3 to 7.1 and 6.5 g/L on days 6 and9 (p = 0.002, p = 0.004, respectively; Tukey's HSD test). the SLA Experiment, the concentration of butyrate increased from 0.9 g/L on day 3 days to 7.0 g/L and 8.0 g/L on days 6 and 9 (p = 0.0004, p = 0.0003, respectively; Tukey's HSD test). During longer fermentation (days 12, 15 and 18), butyrate remained an abundant fermentation product and pH was maintained at ca. 5. On day 6 and onward, ethanol concentration decreased and it became a minor metabolite compared to samples collected on day 3 in either MLA or SLA Experiment (p = 0.0002, Tukey's HSD test) .
Clustering analysis of each experiment revealed that in some individual experimental replicates differed from the other counterparts and were more similar to those from other experiments. For example, the replicate B after the 3 rd passage (9 days) from the Experiment MLA grouped with the samples collected after the 3 rd passage (9 days) from the Experiment ML. Other examples are replicates from the Experiment SLA after the 4 th , 5 th and 6 th passages, respectively, 12, 15 and 18 days ( Figure 3). We had no logical explanation and thus no reason to exclude these replicates from analysis.

Dynamics of fermentation of lactate and acetate as the main carbon sources (LA)
A distinct scenario occured for Experiment LA when the source of carbon was limited to lactate and acetate ( Table 2 and Additional les 1 and 4). The pH of the digestive liquids was maintained at approximately 7, and the lowest reached value of 6.6 was observed after the rst three days. Since the second passage (after 6 days), lactate was e ciently utilized (90-100%) and the dominant fermentation product, butyrate, was maintained at similar level during the whole experiment at around 4 g/L ( Table 2). Acetate was detected as the second component (1 -2 g/L) of the digestive liquids, wheareas propionate (0.2 -0.4 g/L) and ethanol (0.05 -0.2 g/L) were detected as minor products.

Summary of the static batch experiments and redundancy analysis
Detailed statistical comparison of the pH and metabolite (ethanol, butyrate, propionate, and lactate) formation between Experiments M, ML, MLA, LA and SLA are depicted in Additional le 1. For simplicity, in this section we focus on the results from day 6 and 9. The pH values were signi cantly different among all experiments (0.001 > p > 0.0002, Tukey's HSD test; Additional le 1) with the lowest pH recorded in Experiment M (molasses only;  Table 2).
To integrate the targeted metabolomic data with the analyses of sample biodiversity, we performed redundancy analysis (RDA), a direct gradient analysis technique which summarises linear relationships between components of response variables that are "redundant" with (i.e. "explained" by) a set of explanatory variables. The results of RDA analysis and the correlation between the fermentation products, pH of the digestive liquids and the dominant bacterial genera in respective experiment are presented in Figure 5. This corresponds with a higher contribution of Clostridium in the microbial communities. In all scenarios, propionate remains a minor product during the experiments, a decreasing contribution of Fructobacillus is observed over time, and Lactobacillus remains to be an abundant genus. Butyrate formation is related to pH increase, higher contribution of Clostridia (e.g. Clostridium sensu stricto 12) in the microbial community and increase of biodiversity that is the especialy proiminent in Experiment LA.

Carbon balance in the selected static batch experiments
We have previously described an approximate balance of carbon during the fermentation of lactate and acetate to butyrate by Clostridium butyricum and proposed a model of lactate/acetate conversion to butyrate [18]. To illustrate metabolic transformations in the batch experiments performed in this study, the approximate millimolar balance of carbon for the selected data from Experiments LA and SLA (as shown in Figure 4) was calculated and presented in Table 3. sequencing. The goal of this analysis was to identify species potentially responsible for sucrose, acetate and lactate utilization. However, due to limitations of the approach we chose, we limited the data interpretation to two aspects. Since the MLA community produces initially (on day 3) large quantity of lactate, the rst goal was to identify the putative main lactate producers from sucrose (molasses) fermentation. The species more highly represented in MLA3 vs LA3 communities (>2-fold higher in MLA3, > 0.02% abundance in MLA3) were selected and 72 species that may be involved in fermentation of sucrose to lactate were identi ed ( Figure 6, Additional le 7). They were the Lactobacillus, Leuconostoc, Bi dobacterium, Weissella, Enterococcus, Gardnerella, Pediococcus, Oenococcus, Peptoaerobacter species. The top species were Lactobacillus uvarum (7-fold higher in MLA3, =11.3%), L. brevis The species more highly represented in LA18 vs LA3 communities (>2-fold higher in LA18, > 0.02% abundance in LA18) were selected and 48 species were identi ed ( Figure 6, Additional le 7), They were mostly the Clostridium, Terrisporobacter as well as Romboutsia, Shigella, Aerocolum, Gottschalkia, Klebsiella and Lactococcus species. The top species were Clostridium tyrobutyricum (106-fold higher in LA18, =28.6%) and Terrisporobacter glycolicus (7.2-fold higher in LA18, =4.1%) suggesting that these species contributed to butyrate synthesis. Interestingly, In the LA-3-BC sample (low butyrate formation) the top species were Clostridium sul digenes (7%) and Clostridium beijerinckii (4%). The former maintained at the level 4.8% whereas the latter dropped to 0.3% in the LA-18-AB sample.
Finally, the common species between MLA9 and LA18 communities were found. All of them were the Clostridium species (the most abundant C. tyrobutyricum and minor C. coskatii, C. kluyveri, C. ljungdahlii, C. ragsdalei, C. arbusti, C. estertheticum, Clostridium sp. DMHC 10, C. pasteurianum, C. carboxidivorans and C. acetobutylicum) ( Figure 6, Additional le 7). There are putatively the most involved in butyrate formation independent on the growth medium (MLA or LA).

Discussion
This study describes dynamics of the metabolic activity and the structure of microbial communities sampled from hydrogen-producing dark fermentation bioreactor and tested in static batch experiments.
The results contribute to better understanding of the dynamics and plasticity of dark fermentation microbial communities that are key factors responsible for the stability and instability of hydrogen production process. This study continues and expands our previous research [18] on lactate and acetate conversion to butyrate by the bacteria of dark fermentation. Here we provide new data on (i) the conditions that favour and unfavour transformation of lactate and acetate to butyrate by dark fermentation microbial communities and (ii) the key players of the microbial communities determining ability to the conversion process. The study involved ve independent static batch experiments performed under anaerobic conditions differing with the media subjected to fermentation. The media contained molasses as a source of sucrose or pure sucrose with addition of lactate and acetate, or exclusively a mixture of lactate and acetate. Molasses has been used in our studies for hydrogen and methane production in a two-stage process [46][47][48]. Sucrose is an attractive substrate for glycolytic fermentations. Lactate and acetate are substrates for butyrate, hydrogen and carbon dioxide formation. Previously we have shown that dark fermentation microbial communities are unable to grow on lactate only-containing media [18].
The experiments in this study focused on the analysis of non-gaseous fermentation products and the examination of the microbial communities by 16S rDNA pro ling complemented by metagenomics analysis of the selected samples from Experiments MLA and LA (Additional les 5 and 6). Changes in the microbial communities, selection of the speci c groups of bacteria during the batch experiments point to the signi cance of substrates and metabolic activity of bacteria for the community structure and plasticity. It is noteworthy that the differences were also observed between replicates of the same experiments. This variability revealed in the batch experiments may explain unstable operation of dark fermentation bioreactors observed in many studies [9]. It may also illustrate metabolic microniches that can be formed in the bioreactors.
In  [20]. In their studies the majority of fermentable carbohydrates were metabolized to lactate by Lactobacillus, Sporolactobacillus, Streptococcus and acetate by Acetobacter that dominated in the microbial communities. Lactate and acetate were further used for production of butyrate and hydrogen by hydrogen producing bacteria [19,20]. In our Experiments SLA and LA (the latter with no carbohydrates in the medium), after the rst three days the most abundant genus was Clostridium sensu stricto 1, unlikely to convert lactate to butyrate.
Our observations concerning microbial communities selected in the Experiment M con rm commonly recognized fact about replacement of clostridial type fermentation by lactic acid fermentation in hydrogen-producing bioreactors and support the thesis about negative role of lactic acid bacteria in the dark fermentation microbial communities [9,49]. Analysis of the non-gaseous fermentation products in digestive liquids from the Experiment M revealed that besides lactic acid, the dominant product was also ethanol, both products of heterolactic fermentation. Butyric acids was a minor product and a drop in pH to < 4.0 was observed. Noteworthy was a high concentration of ethanol that together with a low pH (< 4.0) might be a relevant factor responsible for metabolic shift of microbial community towards lactic acid fermentation and maintenance of its stable metabolic activity. Previously, we have shown butyrate formation by the dark fermentation microbial communities grown on the medium containing only molasses [18]. However, compared to this study, the previous media contained a higher concentration of Na 2 HPO 4 and KH 2 PO 4 , that increased their buffering capacity and helped to maintain pH at 5.
Presence of external lactate or lactate and acetate in the medium was a stimulating factor for the growth of butyrate producers (Experiments ML, MLA and SLA). Lactate and acetate are also fermentation products. Our results clearly show that the balance between lactic acid bacteria and butyrate producers is key for the conversion of lactate and acetate to butyrate. Dark fermentation microbial communities that the most effectively converted lactate and acetate to butyrate (Experiments SLA and MLA) were composed of Clostridium sensu stricto 12, Lactobacillus, Fructobacillus, Bi dobacterium and Prevotella. Summary of the batch experiments where lactate and acetate were transformed to butyrate clearly shows a signi cant consumption of lactate (or its low concentration) among the non-gaseous products despite the presence of lactic acid bacteria in the microbial communities. It is in accordance with the previous studies where the e uents from hydrogen producing bioreactors and the microbial communities were examined [46,47,50]. It can also explain a lack or a weak correlation between Lactobacillus and lactate in Experiments ML, MLA and SLA. Interestingly, Esquivel-Elizondo and co-workers [51] considered Lactobacillaceae as a putative butyrate-producers.
It is noteworthy that Clostridium sensu stricto 12 was an abundant taxon in all butyrate producing microbial communities. Metagenomic analysis of the selected samples from Experiments MLA (day 9) and LA (day 18) revealed a signi cant contribution of Clostridium tyrobutyricum in the microbial communities. C. tyrobutyricum is a recognized hydrogen-and butyrate-producing bacterium via conversion of lactate and acetate [30,31].
Previously, we have postulated that pH may be a critical factor responsible for a balance of dark fermentation microbial communities. In our experiments, pH was established and maintained intrinsically in the asks with no additional pH adjustments. Lactate and acetate were transformed to butyrate at pH ≈ 7 when the substrate did not contain carbohydrates and or 5-6 when the substrate contained molasses or pure sucrose. Other studies reported pH in the range of 5.5-6.5 as optimal for hydrogen production and butyrate formation from lactate and acetate [13,21,24,[29][30][31]50].
Interestingly, Garcia-Depraect et al. [21] reported that increase of pH above 6.5 caused domination of Blautia and Propionicum genera in the microbial community processing tequila vinase and nixtamalization wastewater, and a metabolic shift leading to propionate production. However, in our study, low concentrations of propionate were detected within the non-gaseous fermentation products in all the samples. Propionate-type fermentation characteristic of e.g. Clostridium propionicum [32] was thus seemed irrelevant.
The results of our research are generally consistent with those of other groups. On one hand they con rm that lactic acid bacteria compete with dark fermentation bacteria and inhibit their growth [9][10][11][12][13]. However, the role of ethanol as the promoting factor is novel. Ethanol and a low pH are thought to provide unfavourable conditions for butyrate producers and conversion of lactate and acetate to butyrate. On the other hand our results strongly support thesis that conversion of lactate and acetate to butyrate is a common process in dark fermentation bioreactors. Furthermore, it is belived that this metabolic pathway is the main route of hydrogen production during acidic fermentation of carbohydrates-rich substrates [19][20][21][22][23][24][25]. Further investigation should concentrate on search for quorum-sensing mechanisms regulating hydrogen producing microbial communities with regard to pH and ethanol contribution. The regulation seems to be more complex than maintaining lactate and acetate transformation and likely includes mutual metabolic stimulation of bacteria.
Although during the batch test in this study the hydrogen production was not measured, we calculated the balances of carbon for selected time points. The balances of carbon performed for Experiment SLA differ dependently on the contribution of butyrate formation. The X value (meaning bacterial biomass and other fermentation products including fermentation gases) was higher when transformation of lactate and acetate to butyrate was observed. Since the bacterial biomass was similar in every experiment the differences included fermentation gases and eventually other non-analysed products.
Balance of carbon for the transformation of lactate and acetate to butyrate in the Experiment LA was similar to that for pure culture of Clostridium butyricum [18] with regard to milimoles of butyrate. Formation of butyrate was not limited by propionate synthesis. A lower concentration of ethanol within the non-gaseous fermentation products in comparison to the previous study [18] may have resulted from a lower concentration of acetate in the medium and/or activity of other bacteria within the community. It should be noted that acetate is a substrate and an intermediate on the pathway of lactate to butyrate transformation [15,27,28].
It is noteworthy that the biodiversity of microbial communities measured by the taxonomic richness and evenness increased with the capability to transform lactate into butyrate. This capacity was the lowest when the microbial communities were dominated by lactic acid bacteria and the most abundant fermentation products were lactate and ethanol. Additionally, the highest biodiversity was observed in microbial communities grown in the presence of lactate and acetate (Experiment LA) and was related to the presence of taxa not found in other Experiments. These include Terrisporobacter, Lachnoclostridium, Paraclostridium or Sutterella. Paraclostridium strain CR4 was isolated from sugar cane bagasse involved in hydrogen and butyric acid production [52]. Other genera were found in and were isolated from human intestine microbiome [53][54][55][56].
Since lactic acid and butyric acids fermentations are ubiquitous and some analogies can be found between anaerobic digesters and the mammalian gut, our results are also relewant in the context of human microbiomes analyses. Interestingly, studies on the microbiomes from the patients with autism spectrum disorder revealed an increased number of Terrisporobacter, Lachnoclostridium [55]. Lachnoclostridium was identi ed as a novel bacterial marker for the non-invasive diagnosis of colorectal adenoma [56]. It was also found that most of the gut butyrate-producing bacteria were signi cantly decreased in patients with non-small-cell lung cancer compared to healthy adults [57]. Also diarrhoea is often associated with the accumulation of lactate in the hindgut in the case of intestinal disorders such as short-bowel syndrome, in ammatory bowel disease, ulcerative colitis, dyspepsia, antibiotic-associated diarrhoea [58].

Conclusions
The batch tests revealed dynamics of metabolic activity and relevant differences in composition of dark fermentation microbial communities dependent on fermentation conditions. These results expand our knowledge on lactate to butyrate conversion by dark fermentation microbial communities and are relevant for understanding the processes inside hydrogen-producing bioreactors. The microbial communities unable to butyrate formation are dominated by lactic acid bacteria. The main fermentation products are lactate and ethanol, the drop of pH to < 4.0 is observed. Further investigations should concentrate on the role of pH and ethanol in the changes of microbial communities structure and metabolic shifts towards lactate fermentation.
With the ability to convert lactate and acetate to butyrate, biodiversity of microbial communities increase. The process of conversion proceeds at pH ≈ 5-6 when the media contain carbohydrates. The most relevant for lactate to butyrate conversion is the balance between lactic acid bacteria (mainly Lactobacillus) and butyrate producers (especially Clostridium sensu stricto 12, also Prevotella). Mutual metabolic stimulation of bacteria is likely and should not be discounted. In the absence of carbohydrates, the process of conversion lactate and acetate to butyrate proceeds at pH ~ 7 and the most abundant bacteria belong to the clostridial species. The increased contribution of Terrisporobacter, Lachnoclostridium, Paraclostridium or Sutterella is also observed. Metagenomic analysis revealed C. tyrobutyricum as the most abundant Clostridium species in the butyrate producing microbial communities fermenting media with and without carbohydrates. the sequence data and contributed to writing the manuscript. ASa performed analyses of short chain fatty acids and ethanol. ASi, AD, DL, PK analyzed the results. AD contributed to writing the manuscript.

List Of Abbreviations
AB, YC, FY analyzed the metagenomics data. PK revised the manuscript and contributed to writing the manuscript. ASi wrote the paper. MKB revised the paper. All authors read and approved the nal manuscript.     The main fermentation product of the medium containing exclusively lactate and acetate was butyrate; lactate as a substrate was e ciently utilized  Taxonomic composition (genus level) of the microbial communities selected in batch experiment based on hypervariable V4 region of the 16S rRNA gene, sequenced on MiSeq platform (Illumina). The taxonomy was assigned using RDP classi er against SILVA database. All taxa with Relative Abundance lower than 0.1% were removed.

Figure 3
Heatmap showing the relative abundance of genera in the individual Experiments for all timepoints and annotation with measured metabolites and pH. The heatmap was generated in R (Heatplus package, annHeatmap2 function) using the relative abundance of the observed genera. For clarity, the "Inoculum" sample and all genera with summarized relative abundance lower than 0.1% were removed. Rows were clustered using average linkage hierarchical clustering based on the Bray-Curtis dissimilarity matrix of the dataset ('vegdist' from the vegan package).

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.