Shifting the balance of fermentation products between hydrogen and volatile fatty acids: microbial community structure and function

Abstract

Fermentation is a key process in many anaerobic environments. Varying the concentration of electron donor fed to a fermenting community is known to shift the distribution of products between hydrogen, fatty acids and alcohols. Work to date has focused mainly on the fermentation of glucose, and how the microbial community structure is affected has not been explored. We fed ethanol, lactate, glucose, sucrose or molasses at 100 me- eq. L⁻¹, 200 me- eq. L⁻¹ or 400 me- eq. L⁻¹ to batch-fed cultures with fermenting, methanogenic communities. In communities fed high concentrations of electron donor, the fraction of electrons channeled to methane decreased, from 34% to 6%, while the fraction of electrons channeled to short chain fatty acids increased, from 52% to 82%, averaged across all electron donors. Ethanol-fed cultures did not produce propionate, but did show an increase in electrons directed to acetate as initial ethanol concentration increased. In glucose, sucrose, molasses and lactate-fed cultures, propionate accumulation co-occurred with known propionate producing organisms. Overall, microbial communities were determined by the substrate provided, rather than its initial concentration, indicating that a change in community function, rather than community structure, is responsible for shifts in the fermentation products produced.

Variations in the starting concentrations of fermentable substrates shift the products of fermentation from hydrogen (and methane) to fatty acids, resulting in substrate-dependent shifts in the associated microbial communities.

INTRODUCTION

Fermentation is a process of great importance in many industrial and environmental biotechnologies as well as in the natural environment. It is used industrially to produce beverages, foods and chemical compounds as well as to clean wastewaters (Parkin and Owen 1986; Agler et al. 2011) and to enhance bioremediation processes in groundwater (Ziv-El et al. 2012a,b; Loffler et al. 2013; Delgado et al. 2014). Recent work indicates that fermentation is a key element for host–microbe interactions in the human intestine and may contribute to some diseases (Ley et al. 2005; Samuel and Gordon 2006; Turnbaugh et al. 2006; Zhang et al. 2009; Krajmalnik-Brown et al. 2012; Kang et al. 2013). Fermentation is a metabolic process, often performed by microorganisms, where energy is extracted from carbon compounds without the use of an external electron acceptor, such as oxygen or sulfate (Hoelzle, Virdis and Batstone 2014).

During many fermentation reactions, microorganisms consume an electron donor and produce H₂, volatile fatty acids (VFAs), and alcohols (Thauer, Jungermann and Decker 1977; Rodríguez et al. 2006). Figure 1 illustrates some of the microbial pathways involved in the fermentative breakdown of glucose. During glycolysis, NAD⁺ is reduced to NADH and pyruvate is formed (reviewed in Bodner 1986). The NADH must be re-oxidized in order to replenish the NAD⁺ pool necessary for further glycolysis. Under fermentative conditions, the electrons from oxidizing NADH can be sent to reduce protons for H₂ production (Rodríguez et al. 2006). Pyruvate can be channeled to the TCA cycle for acetate formation, which involves the reduction of the protein ferredoxin. Similar to NADH, ferredoxin can be re-oxidized with the evolution of H₂ (Mortenson, Valentine and Carnahan 1962, 1963; Tagawa and Arnon 1962). In both cases, the production of H₂ faces thermodynamic limitations, as reviewed by Schink (1997).

Figure 1.

Selected pathways involved with the fermentation of glucose. Production of acetate, propionate and H₂, as well as the subsequent use of acetate or H₂ for methanogenesis are shown highlighted.

Briefly, the pathways that produce H₂ during fermentation are favored only at low H₂ partial pressures due to the lower Gibbs free energy for the chemical reactions (Tanisho, Kamiya and Wakao 1989; Schink 1997; Angenent et al. 2004). According to Angenent et al. (2004), generating hydrogen (H₂) with electrons from NADH during fermentation is inhibited at H₂ partial pressures above 0.0006 atm, or 0.47 μM in the solution. Due to the more negative redox potential of ferredoxin ( = −400 mV) compared to NADH ( = −320 mV), the production of H₂ through ferredoxin can be sustained up to H₂ partial pressures of 0.3 atm, or 237 μM in solution (Angenent et al. 2004). When the concentration of H₂ increases, the change in free energy caused by the reaction also increases until it becomes positive and the reaction is no longer thermodynamically favorable.

In anoxic environments, hydrogenotrophic organisms keep H₂ concentrations low enough that the production of H₂ during fermentation remains thermodynamically favorable (Lee and Zinder 1988; Ragsdale and Pierce 2008). Hydrogenotrophs include hydrogenotrophic methanogens, which produce methane, and homoacetogens, which produce acetate. Maintenance of low H₂ concentrations by hydrogenotrophs leads to faster fermentation rates (Schink 1997; McInerney et al. 2008; Zhang et al. 2009; Krajmalnik-Brown et al. 2012).

When the thermodynamic threshold is reached, due to H₂ production outpacing H₂ consumption, the fermenting organisms must slow down H₂ production. Under these conditions, fermenters do not cease consuming available electron donors. Instead, there is a shift in the metabolites produced and the microorganisms produce compounds such as ethanol, propionate and lactate as discussed by Ruzicka (1996), Angenent et al. (2004) and Hoelzle, Virdis and Batstone (2014). During propionate formation, bacteria use electrons from NADH in either of two pathways for the production of propionyl-CoA (Swick and Wood 1960; Hetzel et al. 2003), followed by the release of propionate (highlighted by blue ovals in Fig. 1). These, and similar reactions, enable the continued fermentation of more complex, higher energy-carrying compounds, such as glucose, through the production of simpler compounds, such as propionate and butyrate, without necessarily producing more H₂. The electrons in ferredoxin can be transferred to either NAD⁺, to form products as described above, or the electrons can be transferred to NADP⁺ and used for biosynthesis (Jungermann et al. 1973).

Previously, work has focused on glucose fermentation as a representative compound, the process of methanogenesis is not usually included, and the associated microbial communities and their responses to these changes have not been examined. The human gut, natural environments and engineered systems do not only involve glucose; therefore, it is important to include other common fermentable electron donors when studying fermentation. Given the difficulty in preventing methanogenesis in systems that have a continuous influx of microbial biomass, it is important to include methanogens when studying the effect of electron donor loading on the distribution of fermentation products. Finally, whether or not microbial communities respond to changes in electron donor concentrations, as opposed to responding only to electron donor type, by altering their structure or function can have consequences in bioreactor stability and efficiency. For the first time, we report changes in the microbial community structure related to changes in the balance of fermentation products produced from various fermentable electron donors.

In order to study the shift in the balance between H₂ and VFA production during fermentation and its effect on microbial communities, we set up fermentation cultures in batch serum bottles fed ethanol, lactate, glucose, sucrose and molasses as electron donor. We varied the initial concentrations of electron donors and monitored VFAs, methane and H₂ as fermentation products as well as how the associated microbial community responded as a result of these electron donor variations.

MATERIALS AND METHODS

Fermentation batch experiments

We performed fermentation batch experiments in triplicate in 150 mL serum bottles. The serum bottles contained 100 mL of enrichment media: 1.5 mM KH₂PO₄, 9.3 mM NH₄Cl, 17.1 mM NaCl, 4.2 mM MgCl₂, 6.7 mM KCl, 1.4 mM CaCl₂, 100 mM NaHCO₃, 0.1 mM Na₂S, 25 μM FeCl₂, and 2 mL L⁻¹ of trace minerals solution. The trace minerals solution contained 1.7 mM EDTA, 345 μM CoCl₂, 1 mM CaCl₂, 162 μM H₃BO₃, 97 μM Na₂MoO₄, 5.8 μM Na₂SeO₃, 39 μM Na₂WO₄, 154 μM NiCl₂, 12.2 mM MgCl₂, 4.7 mM MnCl₂, 367 μM ZnCl₂, 400 μM CuSO₄ and 23 μM AlK(SO₄)₂. We sparged 80% N₂:20% CO₂ gas through the media to make it anaerobic prior to and during the addition of the reagents, adjusted the initial pH of the media to ∼7.6 with 1 M HCl, and sterilized all culture bottles by autoclave. We performed fermentation experiments with ethanol, lactate, glucose, sucrose and molasses at starting concentrations of 100, 200 and 400 me⁻ eq. L⁻¹. The starting concentrations were normalized to electron equivalents for comparison across electron donors, obtained by multiplying the desired number of electron equivalents by the number of electron equivalents per mol for each electron donor. Ethanol and lactate have 12 e⁻ eq. per mol, glucose has 24 e⁻ eq. per mol and sucrose has 48 e⁻ eq. per mol. The corresponding ethanol and lactate concentrations are 8.3, 16.7 and 33.3 mM. For glucose, the concentrations are 4.2, 8.3 and 16.6 mM. The equivalent concentrations for sucrose are 2.1, 4.2 and 8.3 mM. For the molasses experiments, the chemical oxygen demand (COD, measured in mg COD/L) of the molasses used was measured with HACH high range COD digestion vials (HACH Catalog no. 2125915). This value was then converted to e⁻ eq. L^-1 using the ratio of 8 g COD/e⁻ eq. The composition of molasses was determined to be 1:1 glucose to sucrose and the concentration added was verified using high pressure liquid chromatography (HPLC) analysis, described below. We also performed lactate experiments at 0.8, 1.7 and 3.3 mM. We inoculated each culture with 1 mL of anaerobic digested sludge obtained from the Northwest Water Reclamation Plant, Mesa, AZ. The cultures were incubated at 37°C and 150 rpm. Both liquid and gas samples were taken daily for the first four days, and then periodically to quantify VFAs, H₂ and methane. Electron balances shown are the averages of measurements from biological triplicates.

Microbial community analysis

In order to perform community analysis, we collected biomass samples and extracted DNA. Each replicate was analyzed separately. At the beginning of each experiment, we took 1 mL of the anaerobic digested sludge used for all of the cultures, placed it in a clean sterile 1.5 mL falcon tube previously cross linked in a SpectroLinker for 5 min, and froze it for future DNA extractions. We also obtained 1 mL of each culture when the original electron donor was depleted (day 4 for ethanol, glucose, sucrose and molasses and day 3 for lactate). The samples were placed into −80°C storage. DNA was extracted from the samples using the MoBio Power Soil DNA kit (Mobio Laboratories, Carlsbad, CA, cat. 12888) and quality and quantity were checked on a Nanodrop ND-1000 spectrophotometer. DNA samples were sent to the ASU Microbiome Analysis Lab (http://krajmalnik.environmentalbiotechnology.org/microbiome-lab.html) for sequencing on an Illumina MiSeq sequencer using third generation chemistry. Primers targeting the V4 region (515F: GTGCCAGCMGCCGCGGTAA and 806R: GGACTACHVGGGTWTCTAAT) (Caporaso et al. 2012) were used for paired end sequencing and approximately 280 bp were sequenced from each end. Sequences were deposited in the Sequence Read Archive (www.ncbi.nlm.nih.gov/sra) under BioProject accession number PRJNA330962.

Paired ends were combined using Pandaseq (Masella et al. 2012), the assembly algorithm from the Ribosomal Database Project paper (Cole et al. 2013), elimination of all sequences with uncalled nucleotides, and a threshold of 0.95. Sequences were analyzed using the QIIME software package as described in Ruiz et al. (2014), except as noted. We used QIIME version 1.8 (Caporaso et al. 2010) and a sequence length cutoff of 293 bp. We assigned taxonomy using the 97% similarity set of operational taxonomic units (OTUs) from the Greengenes 13_5 database (McDonald et al. 2012). We rarefied the samples to 19289 OTUs, the number of OTUs present in the smallest sample as described before (Delgado et al. 2014). We used the QIIME beta diversity through plots script to generate a principal coordinate analysis of the communities based on the weighted UNIFRAC distance metric (Lozupone and Knight 2005). We used the QIIME make distance boxplots script to compare the distances, also based on the weighted UNIFRAC distance, between samples based on concentration and electron donor provided, including a two sample t-test to test the significance of the differences in the distances between sets of samples. Pearson's correlation coefficients were calculated between electron donor concentrations, measured fermentation products and OTUs observed along with accompanying P-values based on the t-distribution.

Analysis of fermentation products

We analyzed liquid samples from the cultures using HPLC (Model LC-20AT, Shimadzu, Columbia, MD) and gas samples from the headspace of the cultures using gas chromatography as previously described (Miceli et al. 2014). For sugars, we analyzed the liquid samples on the same HPLC system, at 30°C with 18 MOhm reverse osmosis water as eluent and a refractive index detector (Waters, RID 2414). Student's two tailed t-tests were used to determine the likelihood of observing the differences in the ratio of VFAs to methane seen among experimental cultures.

RESULTS AND DISCUSSION

Increasing electron donor concentration decreases the fraction of electrons channeled to methane production

Figure 2 shows the distribution of electrons from the initial electron donors to acetate, propionate, butyrate and methane when the original electron donor was depleted (day 4 for ethanol, glucose, sucrose and molasses and day 3 for lactate). The days shown were selected based on the initial electron donor having been consumed and the fact that acetate concentrations had not begun to decline yet, evidence that processes that consume acetate, like acetoclastic methanogenesis in this case, were not active at this time in the process. Table 1 shows the sum of electrons channeled to VFAs (acetate, propionate and butyrate), the number of electrons channeled to methane and the ratio of VFAs to methane. Asterisks indicate differences in the ratio of VFAs to methane between different initial substrate concentrations that are considered significant according to a student's t-test (P-values < 0.05). Across all fermentation cultures, as the starting electron donor concentration increased, the fraction of electrons going to methane decreased (averaged across electron donors, from 37% to 8%) while the fraction of electrons going to VFAs (acetate, propionate and butyrate) increased (from 43% to 80%). This agrees with our expectations that by producing sufficient H₂, and thus overloading the thermodynamics of H₂ production and the rate of consumption, fermentation cultures are forced to route electrons to more reduced carbon compounds. P-values student's t-tests of the changes in the ratio of VFAs to methane are provided in the Tables S1 and S2 (Supporting Information). P-values for student's t-tests of changes in methane, acetate, propionate and butyrate between different initial substrate concentrations are provided in Table S3 (Supporting Information).

Figure 2.

Distribution of electrons from initial electron donor to methane, acetate, propionate and butyrate. The data shown are from day 4 of the ethanol, glucose, sucrose and molasses experiments and from day 3 of the lactate experiment. Error bars represent one standard deviation from biological triplicates.

Table 1.

The ratio of VFAs to methane, measured as a fraction of the electrons from initial substrate, organized by substrate type and concentration. For all substrates, low, medium and high concentrations correlate to 100 or 10 me⁻ eq. L⁻¹, 200 or 20 me⁻ eq. L⁻¹ and 400 or 40 me⁻ eq. L⁻¹, due to the difference in concentrations tested for lactate cultures. Concentrations which have significantly different ratios of VFAs to methane are marked with asterisks, as measured by a student's t-test with P-value < 0.05.

Experiments performed with lactate at 100, 200 and 400 me⁻ eq. L⁻¹ resulted in the majority of electrons being diverted to propionate, showing that this concentration range was too high to see any changes in the distribution of fermentation products. Lactate fermentations had to be performed at 10, 20 and 40 me⁻ eq. L⁻¹, corresponding to 0.83, 1.66 and 3.33 mM, to observe shifts in the products, between methane and VFAs, similar to the other electron donors. For further explanation on the lactate-fed cultures and breakdowns of the products over time for individual experimental conditions, see Figs S1–S6 (Supporting Information).

Electron donor determines overall microbial community structure

Figure 3 shows the microbial community structures plotted on principal components 1 and 2 from a principal coordinate analysis of the distances, measured by weighted UNIFRAC, between the microbial communities. Weighted PCoA, in comparison to unweighted, considers changes in relative abundance that causes dominant microbial species to have a more substantial impact on the analysis. Principal component 1 accounts for 35.0% of the variation in the community structure data and principal component 2 accounts for 25.4%, which indicates that the majority of variation in the data (60.4%) is captured in this transformation. The inoculum communities were grouped very tightly, indicating that the starting inocula, although collected at different time points, were not likely to affect the outcome of the experiments. The experimental communities form clusters by type of electron donor fed to them rather than by electron donor concentration, indicating that the type of electron donor is more important than the concentration in determining the major taxa community composition, regardless of the functional changes we observed. A student's t-test of the UNIFRAC distances between subsets of the communities further supports that electron donor type has a stronger effect on community changes than the concentration and is provided in Fig. S7 (Supporting Information). PCoA based on the unweighted UNIFRAC distances only captured ∼21% of the variability in the first to principal components, indicating that presence/absence of different OTUs did not represent the major changes in the community data (data not shown).

Figure 3.

PCoA of microbial communities from fermentation cultures and inoculum. Based on weighted UNIFRAC distances, the circle sizes represent the inoculum (smallest) and increasing starting substrate concentration: 100 me- eq. L⁻¹, 200 me- eq. L⁻¹ and 400 me- eq. L⁻¹. (10, 20 and 40 me- eq. L⁻¹ for lactate). Inoculum communities are highlighted with a gray circle.

In Fig. 3, while the microbial communities do not form tight clusters based on varying the initial concentration of electron donor, a trend with increasing electron donor concentration does appear. For the glucose-fed cultures, the communities spread out across principal component 1 in the order of low to high concentration. Molasses-fed communities show this trend to a small degree, while this trend does not appear in ethanol, lactate and sucrose-fed communities. Lactate-fed communities, out of all the experimental communities and especially those fed 20 me⁻ eq. L⁻¹, cluster more closely to the inoculum communities, indicating the high similarity between these two groups.

Propionate producing bacteria correlate with electron donor concentration

Examining the microbial communities in more detail, Fig. 4 shows the abundance of those OTUs that accounted for at least 10%, averaged across replicates, of at least one of the experimental microbial communities, identified at the family level. Classification of OTUs down to the genera level, Pearson's correlation coefficients and associated P-values are available in the supplementary file CorrelationStats.xlsx (Supporting Information). The families Bacteroidaceae (Wallnöfer and Baldwin 1967; Macy, Ljungdahl and Gottschalk 1978) and Clostridiaceae (Leaver, Wood and Stjernholm 1955; Grupe and Gottschalk 1992; Vasconcelos, Girbal and Soucaille 1994) include known propionate producing bacteria. OTUs from the family Bacteroidaceae, (accounting for <1%–20% of communities) showed a positive correlation with increasing substrate concentration when all substrates were considered together, as well as when ethanol, glucose and molasses cultures were considered individually; however, only the correlations between Bacteroidaceae and substrate concentration overall, glucose concentration and molasses concentration were considered significant (P-value < 0.05). Clostridiaceae negatively correlated with substrate concentration when all substrates were considered (P-value < 0.05). When considered individually, however, the lactate-fed cultures showed a positive correlation between lactate concentration and the family Clostridiaceae (P-value < 0.05). In the inoculum and ethanol cultures, OTUs related to known propionate producing organisms represented only a small fraction of the sequences. The other OTUs that positively correlated with increasing initial electron donor concentration may contribute to propionate production or may play some alternative role such as biomass degradation: Porphyromonadaceae, Enterobacteriaceae and Streptococcaceae (P-values < 0.05).

Figure 4.

Relative abundance of OTUs, identified at the family level, averages of replicates (Inoculum, N = 5; Experimental, N = 3). Only OTUs which accounted for at least 10% (on average) of the community for at least one of the experimental conditions tested are included in this figure.

In addition to some correlations appearing between substrate concentration and OTU abundances, the OTUs related to known propionate producers accounted for large portions of the cultures where propionate was observed. For example, Bacteroidaceae were abundant in glucose, sucrose and molasses cultures (1%–20%). OTUs classified to the order Bacteroidales, of which Bacteroidaceae is a member, were abundant in glucose, sucrose, molasses and lactate-fed cultures (1%–20%). Clostridaceae were abundant in sucrose, molasses and lactate-fed cultures (1%–25%).

Across all experimental cultures, several OTUs correlated with shifts in the spectrum of fermentation products. OTUs assigned to the families Pelobacteraceae, Eubacteriacae, Anaerolinaceae, Campylobacteraceae, and an OTU assigned to the order Bacteroidales were all positively correlated with more complete oxidation of starting substrates (IE positive correlations with acetate or methane production, or negative correlations with propionate or butyrate production). Moraxellaceae, Bacteroidaceae and Streptococcaceae were positively correlated with less complete oxidation (IE the production of propionate or butyrate). Cloacamonaceae showed mixed responses, correlating positively with the production of both propionate and methane, while showing a negative correlation with acetate production. This may indicate a role for Cloacamonaceae in processing intermediates of fermentation. These correlations were all deemed significant (P-values < 0.05). Here, we consider only OTUs that accounted for greater than 10% of at least one experimental community, averaged across replicates. The correlations and associated P-values for these OTUs as well as less abundant OTUs are presented in the chart provided in the supplementary file, CorrelationStats.xlsx (Supporting Information).

The community analysis was performed using extracted genomic DNA, hence, all analyses performed indicate relative abundance but not activity of the observed OTUs. Given that the communities were essentially enriched, through the feeding of substrate, the observation of increases in abundance of OTUs, from inoculum abundances, is strong evidence that the associated organisms were active in the experimental communities. Further, the observation of OTUs related to known propionate producing organisms accounting for large portions of propionate producing cultures strongly suggests that these organisms were responsible for propionate production. Thus, we are measuring function by determining both enrichment in the community and changes in metabolic products.

The initial concentration of electron donor drives microbial communities and determines what fermentation products can be produced. In cultures with higher electron donor concentrations, H₂ production leads to thermodynamically unfavorable conditions for further production of H₂ and acetate. Instead of producing acetate and hydrogen, the communities shifted towards organisms capable of excreting propionate. The decrease in thermodynamic energy available from H₂ production was a selective pressure on the community and the capacity to produce propionate thus became advantageous under these conditions by allowing additional electron donor to be consumed.

CONCLUSIONS

Regardless of electron donor provided, when initial electron donor concentration was higher, the fermentation cultures produced VFAs instead of acetate and methane. This shift occurred for most of the electron donors tested between 100 and 200 me⁻ eq. L⁻¹. Lactate required even lower concentrations of electron donor for this shift to occur (between 10 and 20 me⁻ eq. L⁻¹). The microbial community analysis indicated that the type of electron donor was the major driver in the development of the microbial community structures, but that the initial concentration of electron donor fed was also an important factor in crafting microbial communities. Known families of propionate producers were detected in cultures where propionate was produced: Bacteroidaceae in glucose, sucrose and molasses-fed cultures; members of the order Bacteroidales in glucose, sucrose, molasses and lactate-fed cultures; and Clostridiaceae in sucrose, molasses and lactate-fed cultures. Using electron donor loading to shift the balance of the products of fermentation not only deserves greater exploitation in industrial settings such as biohydrogen production, dechlorination and microbial electrochemical cells, it also offers insights into managing microbial communities in the human gut, where the distribution of fermentation products has been linked to obesity.

SUPPLEMENTARY DATA

Supplementary Data.

FUNDING

This work was supported by the National Institute of Health funded Arizona State University Institute for Maximizing Student Development [Grant Number R25GM099650], by the National Science Foundation CAREER Program [Grant Number 1053939], and by the United States Office of Naval Research [Grant Number N00014-10-M-0231].

Conflict of interest. None declared.