Metabolic Synergy between Human Symbionts Bacteroides and Methanobrevibacter.

Trophic interactions between microbes are postulated to determine whether a host microbiome is healthy or causes predisposition to disease. Two abundant taxa, the Gram-negative heterotrophic bacterium Bacteroides thetaiotaomicron and the methanogenic archaeon Methanobrevibacter smithii, are proposed to have a synergistic metabolic relationship. Both organisms play vital roles in human gut health; B. thetaiotaomicron assists the host by fermenting dietary polysaccharides, whereas M. smithii consumes end-stage fermentation products and is hypothesized to relieve feedback inhibition of upstream microbes such as B. thetaiotaomicron. To study their metabolic interactions, we defined and optimized a coculture system and used software testing techniques to analyze growth under a range of conditions representing the nutrient environment of the host. We verify that B. thetaiotaomicron fermentation products are sufficient for M. smithii growth and that accumulation of fermentation products alters secretion of metabolites by B. thetaiotaomicron to benefit M. smithii. Studies suggest that B. thetaiotaomicron metabolic efficiency is greater in the absence of fermentation products or in the presence of M. smithii. Under certain conditions, B. thetaiotaomicron and M. smithii form interspecies granules consistent with behavior observed for syntrophic partnerships between microbes in soil or sediment enrichments and anaerobic digesters. Furthermore, when vitamin B12, hematin, and hydrogen gas are abundant, coculture growth is greater than the sum of growth observed for monocultures, suggesting that both organisms benefit from a synergistic mutual metabolic relationship. IMPORTANCE The human gut functions through a complex system of interactions between the host human tissue and the microbes which inhabit it. These diverse interactions are difficult to model or examine under controlled laboratory conditions. We studied the interactions between two dominant human gut microbes, B. thetaiotaomicron and M. smithii, using a seven-component culturing approach that allows the systematic examination of the metabolic complexity of this binary microbial system. By combining high-throughput methods with machine learning techniques, we were able to investigate the interactions between two dominant genera of the gut microbiome in a wide variety of environmental conditions. Our approach can be broadly applied to studying microbial interactions and may be extended to evaluate and curate computational metabolic models. The software tools developed for this study are available as user-friendly tutorials in the Department of Energy KBase.

INTRODUCTION

From birth, the human microbiome plays an important role in maintaining human health (1–4). Newborns are colonized in their first days of life (5–7), and the microbiome grows and develops with the child into adulthood (5, 8). Depending on the age, geographic location, diet, and health of its host, 10 to 100 trillion microbial cells reside in the intestines alone (1, 9). These organisms are part of a dynamic, closely interconnected ecosystem made up of bacteria, archaea, and eukarya (7, 8, 10). Interactions between host-associated microbes affect many aspects of human health. A well-balanced, healthy microbiota offers protection against infection (11, 12), metabolizes nutritional compounds (13), provides essential vitamins and nutrients, adds 15 to 30% of human caloric intake (14, 15), manages weight gain (16), and influences the human immune system and its development (17–19). However, an unbalanced microbiome is linked to obesity, infections, asthma, allergies, Crohn’s disease, irritable bowel disease, neurodegenerative disease, and cancer (1–4, 12, 17, 19–34).

It has been proposed that mutualistic relationships within the gut microbiome maintain a balance that is necessary for a healthy host digestive system (8, 29, 35). The dynamic interactions between microbes in a multispecies gut community are difficult to study both in the laboratory and using computational modeling and analysis due to the sheer complexity of the system. This complexity is due to many factors, including the combinatorial genetic space, sampling heterogeneity, unknown environmental conditions (e.g., local microvariation in temperature and pH and nutrient availability), and unknown relationships between interacting genetic and environmental variables that make it difficult to confidently ascribe major (or minor) organism functions in a mixed microbiome community. Fortunately, cutting-edge software system research approaches, such as statistical sampling and decision trees, can be used to develop tools for management and analysis of complex microbial systems (36). To benchmark new computational tools, we constrain the gut microbiome system to studying the relationship between two key human gut inhabitants: the fermenter Bacteroides thetaiotaomicron and the methanogenic archaeon Methanobrevibacter smithii.

B. thetaiotaomicron (B. theta) is one of the most prominent fermenters in the human gut, making up between 5 and 50% of the overall gut community (15). B. theta is a generalist fermenter: it consumes dietary plant polysaccharides such as starch as well as mucosal glycans such as heparin and produces fermentation products such as hydrogen, carbon dioxide, formate, acetate, and succinate (15). It plays a crucial digestive role by partially breaking down polysaccharides human cells are unable to degrade and produces short-chain fatty acids for human and microbial consumption (acetate, succinate, propionate, lactate). In the process, it protects intestines against infection by activating host immune defenses, through direct interactions, and by competition with other bacteria (12, 37).

M. smithii is the dominant methanogen in the human gut and makes up between 1% and 10% of the human gut community (38, 39). It colonizes the human gut in infancy, remains present in the majority of the population through adulthood (6, 38), and has been detected in up to 95.7% of individuals (39–42). Imbalance in M. smithii gut communities has been associated with obesity and malnutrition-related digestive diseases (2, 30, 41, 43). M. smithii is a hydrogenotrophic methanogen and can conserve energy through methanogenesis using carbon dioxide and hydrogen gas or formate as substrates (39, 44, 45). M. smithii has an incomplete reductive tricarboxylic acid (TCA) cycle and must assimilate acetate as a carbon substrate. It is hypothesized that growth of M. smithii in the gut could benefit B. theta and other fermenters by metabolizing fermentation products that would inhibit further fermentation by feedback inhibition as they accumulate. Moreover, the removal of hydrogen could allow B. theta to increase metabolic efficiency and contribute to effective metabolism of dietary substrates, since high hydrogen partial pressure inhibits bacterial NADH dehydrogenases, reducing the yield of ATP and causing fermentation overall to become endergonic (unfavorable) (35, 46). The proposed nature of the complementary metabolism between B. theta and M. smithii suggests a mutualistic relationship in which M. smithii relies on the metabolic products of B. theta fermentation to survive, while B. theta relies on M. smithii to remove products that would inhibit fermentation, which would be an example of metabolic syntrophy (46).

Syntrophy is a form of mutualism in which two organisms form a tightly coupled, mutually beneficial metabolic relationship (46–49). Syntrophic relationships allow bacteria to overcome energy barriers and to break down substrates more efficiently (35, 47, 49, 50). Methanogens have often been observed in syntrophic relationships with soil bacteria, removing hydrogen gas and other fermentation inhibitors to benefit the bacteria (46, 49, 51). B. theta and M. smithii have been previously proposed to form a syntrophic relationship based on genomic evidence (45). Gnotobiotic mouse studies suggest that the presence of M. smithii assists B. theta in the breakdown of polysaccharides and increases host digestive efficiency (35), and our previous study showed that feedback of B. theta is inhibited by acetate and formate (52), suggesting that a partner organism that consumes these metabolic by-products, such as M. smithii, may benefit B. theta. However, the relationship between B. theta and M. smithii is not fully characterized. One of the greatest obstacles is the difficulty in systematically identifying what conditions could lead to a synergistic metabolic relationship, such as syntrophy, that benefits both organisms.

To examine the metabolic relationship between B. theta and M. smithii, we established laboratory conditions in which both monocultures and cocultures of B. theta and M. smithii can be studied and characterized. We leverage an approach to test complex software, partitioning the input space and systematically manipulating program inputs and inferring interactions and relationships between the inputs. In the system we studied here, we observed organism and coculture growth behaviors as outputs dependent on culture nutrient inputs. We then devised an assay to compare monocultures and cocultures in 128 different nutrient conditions and evaluated growth using decision trees from machine learning to identify neutral, favorable, and unfavorable conditions for M. smithii and B. theta growing in coculture.

RESULTS

B. theta and M. smithii grow independently in both rich and defined media when supplemented with appropriate carbon sources.

Before we could examine the interactions between B. theta and M. smithii, we needed to understand how they grow separately in monocultures. The rich medium is based on a standard tryptone yeast extract recipe with high nutrient availability. In sealed anaerobic culture tubes supplemented with 0.05% glucose (Fig. 1a), B. theta doubled every 1.48 h (Table 1). Supplemented with 10 mM acetate and 10 mM formate under an atmosphere of 80% H2, M. smithii grew at a much lower rate (Fig. 1b), with a doubling time of 8.8 h (Table 1). The difference in growth rates means that if two cultures are started at the same time, B. theta has already reached the stationary stage of growth when M. smithii is entering early log growth (Fig. 1a and b). In defined medium, the difference between B. theta and M. smithii monoculture growth is more pronounced. B. theta doubled every 1.4 h, while M. smithii doubled every 15.4 h (Fig. 1a and b, Table 1).

FIG 1

Growth phenotypes on rich and defined culture media. (a) Growth of B.theta monocultures in anaerobic culture tubes in rich (black) or defined medium (gray) compared to B.theta plus M. smithii cocultures (red) grown on defined medium on 0.5% glucose (n = 5). (b) Growth of M. smithii monocultures in anaerobic culture tubes in rich (black) or defined medium (gray) supplemented with 10 mM acetate and 10 mM formate under an atmosphere of 80% H2 (n = 5). (c) Growth of B.theta (gray), M. smithii (blue), and B.theta plus M. smithii cocultures (red) on defined medium with 0.5% glucose as sole carbon and energy source in 96-well plates (n = 8). Error bars have been omitted for clarity. OD600, optical density at 600 nm. G, 0.5% glucose; B, vitamin B12 (cyanocobalamin); H, hematin; F, 10 mM formate; A, 10 mM acetate; 2, 80% H2 atmosphere.

TABLE 1

Culture doubling times (hours)

Treatment	B. theta 0.05% glucose			M. smithii 10 mM acetate + 10 mM formate			Coculture 0.05% glucose
	Maximum OD	Doubling time (SD)	P vs rich	Maximum OD	Doubling time (SD)	P vs rich	Maximum OD	Doubling time (SD)	P vs B. theta	P vs M. smithii
Anaerobic culture tube rich†	1.11 ± 0.013	1.48 ± 0.140	1	0.55 ± 0.051	8.84 ± 0.332	1	0.54 ± 0.020	1.46 ± 0.137	0.890	0.000
Anaerobic culture tube defined†	0.55 ± 0.017	1.40 ± 0.092	0.699	0.15 ± 0.007	15.46 ± 1.803	0.000	0.52 ± 0.016	1.44 ± 0.05	0.872	0.000
96-well plates rich‡	0.64 ± 0.033	1.62 ± 0.097	1	0.17 ± 0.024	7.93 ± 0.825	1	0.745 ± 0.016	2.02 ± 0.073	0.001	0.000
96-well plates defined‡	0.69 ± 0.015	2.04 ± 0.100	0.000	0.14 ± 0.007	318 ± 23.13	0.000	0.61 ± 0.018	2.18 ± 0.067	0.030	0.000

Data were collected from n = 5 (†) or n = 8 (‡) biological replicates. Significance P values were determined by unpaired Student’s t test.

The presence of B. theta is sufficient for M. smithii growth in cocultures.

In defined medium, optical density of cocultures was very similar to that of B. theta monocultures (Table 1) except culture density did not decrease after 50 h during stationary phase. A decrease in culture optical density may indicate a number of causes; among these are cell lysis or a change in cell shape or volume. The observation that cocultures did not show this decrease in optical density suggested that M. smithii protects B. theta from lysis, that it prevents a change in cell shape or volume, or that the optical density reflected growth of M. smithii even though M. smithii monocultures do not grow in medium lacking H2 and CO2 or formate and acetate (Fig. 1c). When cocultures were grown in rich medium in anaerobic culture tubes supplemented with 0.05% glucose, we observed aggregation of cells after 1 week of incubation (Fig. 2a). We did not observe aggregation in B. theta monocultures of the same age (Fig. 2b). Microscopy indicated that aggregates comprised intact B. theta with associated M. smithii (Fig. 2c and d) in addition to intact planktonic and extracellular matrix or dead cells (Fig. 2e and f). The accumulation of M. smithii cells in coculture lacking methanogenic substrates indicates that B. theta fermentation products are sufficient to support M. smithii growth. It should be noted that while lysed B. theta may provide metabolite precursors, M. smithii has an absolute requirement for acetate for acetyl-CoA biosynthesis and it is restricted to using H2 and CO2 and possibly formate as carbon and energy sources.

FIG 2

Microscopy of B.theta and M. smithii in cocultures grown in rich medium for 8 days. (a) Coculture of B.theta and M. smithii grown on rich medium for 1 week showing visible aggregates. (b) Monoculture of B.theta grown on rich medium for 1 week. (c) Live/dead staining of a coculture aggregate appears to show that aggregates are comprised of live B.theta and M. smithii cells as well as what appears to be extracellular matrix and/or lysed cell debris. In live/dead micrographs, intact B.theta cells are green, intact M. smithii cells are blue, and dead cells are stained red. Yellow color results from colocalization of red and green channels. Methanogens autofluorescence due to oxidized coenzyme F420 and appear blue when viewed under a DAPI LED light filter. (d) Transmission micrographs of panel c. (e) Live/dead staining of planktonic cells in coculture showing live B.theta (green), live M. smithii (blue), and dead cells (red). (f) Transmission micrograph of panel d. Black bars indicate scale.

Quantification of B. theta and M. smithii in coculture.

To determine if the interaction is solely a cross-feeding interaction or if B. theta also benefits from coculture with M. smithii during long-term cultivation with glucose as the sole carbon and energy source, the ratio of each organism in coculture populations was quantified using quantitative PCR (qPCR) and flow cytometry. We calculated cell ratios by qPCR by probing for the 16S rRNA-coding region of the genome. Between 4 and 9 days of coculture, there was no change in the coculture optical density, and the ratio of B. theta to M. smithii remained steady at 11.56 (±3.03) on day 4 to 14.25 (±4.72) on day 9 (P = 0.313). To confirm the qPCR results, we also developed a flow cytometry assay to quantify changes in the population ratio according to cell wall staining and to measure cell aggregation. We found that when grown in the presence M. smithii, B. theta (Fig. 3b to d and f to h) forms larger cells or aggregates, which mimics the phenotype observed when B. theta is grown in rich medium with 0.5× glucose (Fig. 3a and e). When 0.5× glucose is provided, cultures deplete glucose carbon source and enter stationary phase more rapidly. Consistent with the qPCR data, after 4 days the ratio of planktonic B. theta to M. smithii cells was 10.97 (±3.81). These results suggest that B. theta and M. smithii reach balanced growth, in contrast to a purely cross-feeding interaction in which we would expect the ratios to transition from B. theta-dominated to M. smithii-dominated cultures over time. Culturing, microscopy, and quantification experiments suggest that B. theta forms irregular aggregates under limiting nutrient conditions that either passively “trap” or actively recruit M. smithii.

FIG 3

Quantification of B.theta and M. smithii in cocultures using flow cytometry. (a to d) Cells were grown in defined medium for 9 days and counted by flow cytometry. The x axis reports forward scattering to quantify the size of the cell populations. Blue histogram indicates B.theta staining. Green histogram indicates M. smithii signal. (e to h) Cells from the same experiments as in panels a to d were visualized by transmission electron microscopy and UV fluorescence with a DAPI filter to detect M. smithii (arrow). Black bars indicate 5 μm scale. (a and e) B.theta grown alone in 0.5× glucose medium. (b and f) B. theta grown alone on glucose defined medium. (c and g) B.theta grown in defined medium supplemented with 20 mM acetate. (d and h) B.theta and M. smithii grown in coculture supplemented with 20 mM acetate.

Evidence for metabolic cross-feeding between B. theta and M. smithii.

M. smithii growth may be supported by B. theta fermentation products acetate, formate, and H2 plus CO2. We next assessed the effect of acetate on cocultures, as acetate is required for M. smithii growth. Under these conditions, growth of M. smithii still requires H2 and CO2 provided by B. theta, as M. smithii cannot grow on acetate as an energy source. When cocultures were supplemented with 20 mM acetate, B. theta cells were larger and formed aggregates, and the relative proportion of B. theta to M. smithii decreased to 4.52 (±1.56) (Fig. 3d and h). The decrease in the relative proportion of B. theta to M. smithii with acetate supplementation is interpreted to suggest that acetate is growth-limiting for M. smithii in cocultures.

We hypothesized that if B. theta and M. smithii have a metabolic relationship, whether it be cross-feeding or syntrophy, then one or both of them may secrete unknown small molecules such as amino acids, bacteriocins, toxins, or quorum-sensing factors to promote growth of the other organism. We tested whether preconditioned medium may have a positive effect on growth of monocultures. In these experiments, media were preconditioned by inoculating with either B. theta or M. smithii monocultures and allowing the monoculture to grow to stationary phase before filter sterilizing with a 0.2-μm filter to remove intact cells. When B. theta was grown on rich medium that had been preconditioned by B. theta monocultures, growth was slower and maximum optical density was decreased (Fig. 4a). When B. theta was grown on medium that had been sequentially preconditioned first by B. theta and then by M. smithii, cultures grew faster to a higher maximum optical density; however, the growth enhancement was not significant versus growth of B. theta on rich medium preconditioned by B. theta (Fig. 4a). B. theta and M. smithii cocultures grown on defined medium that had been preconditioned by M. smithii or B. theta monocultures grew similarly to B. theta cultures using medium that had been preconditioned by sequential culturing of B. theta and then M. smithii monocultures. These results suggest that when grown sequentially, B. theta depletes rich medium of one or more nutrients that are provided at a low level by M. smithii. In an attempt to identify nutrients that might improve growth of B. theta, we supplemented rich and defined medium with hematin, vitamin B12, vitamin K3, formate, and acetate. Growth experiments indicate that the lysis observed in rich medium was reduced by addition of hematin with vitamin K (Fig. 4b). In defined medium, vitamin B12, hematin, formate, and acetate had no effect, but vitamin K improved growth (Fig. 4b). We also tested whether M. smithii growth is enhanced by rich medium preconditioned by B. theta. M. smithii was able to grow solely on rich medium preconditioned by B. theta, suggesting that molecules secreted by B. theta were responsible for stimulating M. smithii growth (Fig. 4c). M. smithii growth was further enhanced by adding hematin, acetate, and formate. However, the growth stimulation observed using preconditioned media was not as dramatic as the growth observed in cocultures (Fig. 1c and Fig. 2). Cross-feeding data indicated that a comprehensive systematic approach was needed to characterize the complex metabolic relationship between B. theta and M. smithii in coculture.

FIG 4

Stimulation of B. theta and M. smithii growth by preconditioning or nutrient supplementation. (a) Growth of B. theta monocultures (Bth) in rich medium is enhanced by preconditioning with M. smithii (Msm; n = 8). Bth G, B. theta monocultures grown on control glucose medium; Bth sG, growth of B. theta monocultures on glucose medium conditioned by B. theta; Bth+Msm sG, B. theta plus M. smithii cocultures grown on glucose medium conditioned by B. theta monoculture; Bth ssG, growth of B. theta monocultures on glucose medium preconditioned first by B. theta monoculture, then by M. smithii monoculture; Msm GFA, growth of M. smithii on rich medium control with glucose, formate, and acetate supplementation. (b) Lysis of B. theta monocultures in defined medium is delayed by the addition of vitamin K3. (c) Growth of M. smithii monocultures on rich medium with glucose (Rich) is enhanced when medium is preconditioned by B. theta (sRich, n = 8). Error bars have been omitted for clarity. OD600, optical density at 600 nm; G, 0.5% glucose; B, vitamin B12; H, hematin; K, vitamin K3; F, 10 mM formate; A, 10 mM acetate.

B. theta growth is inhibited by the presence of hydrogen gas, formate, and acetate.

To tease apart the metabolic interactions between B. theta, M. smithii, and the culture environment, we used a machine learning technique to analyze large-scale growth culture phenotype data. We assessed the growth of each organism alone and in coculture in 128 different combinations of culture media (27). The combinatorial media recipes included or omitted the following nutrient components: 0.05% glucose, 10 mM acetate, 10 mM formate, 5.8 mM vitamin K3, 0.0037 μM vitamin B12, and a mixture of 0.2 mM histidine and 0.02 mM hematin in the presence or absence of 5% H2 headspace gas. The optical densities of B. theta, M. smithii, and cocultures were measured after 1 day to capture maximal B. theta optical density, 7 days to capture slow-growing B. theta and maximal M. smithii optical density, and 14 days to measure culture stability after prolonged incubation. The same time points were observed to allow comparisons between monocultures and cocultures for every growth condition. To analyze and visualize these data, we utilized C4.5 decision trees to identify the relationships between nutrient factors in the culture medium (36). Decision trees draw branch nodes depending on whether the presence or absence of the factors (in this case, the provided nutrients) affects growth of the culture.

In defined medium, glucose was the only carbon source available to B. theta and was the primary determinant of growth (Fig. 5a). The structure of the decision tree suggests that the presence of hydrogen gas alone inhibits growth (Fig. 5b). Hydrogen gas inhibited B. theta growth up to 37.6% (Fig. 6a). When combined with acetate and formate supplementation, growth is decreased by 55% (supplemental material). Acetate and formate alone do not have a significant effect on growth of B. theta (Fig. 6a), although the tree suggests that the presence of acetate and formate in combination may increase growth under certain conditions. Interestingly, the tree indicates that the inhibition caused by H2 is mitigated by the presence of vitamin K or a combination of vitamin B12 and hematin/histidine (Fig. 5b). Additional experiments are needed to understand how this could be occurring, as inhibition of growth by hydrogen gas is not well understood in B. theta. M. smithii is unable to grow under conditions that favor B. theta unless hydrogen, formate, and acetate are present (Fig. 6c).

FIG 5

Effect of medium composition on B. theta growth. Triplicate B. theta monocultures were grown in replicates of 8 on defined medium supplemented with 128 different combinations of glucose, vitamin B12, hematin, vitamin K3, acetate, formate, and a 5% H2 atmosphere. Data across 3 experiments were compared to a positive control (glucose, hematin, vitamin B12, acetate, formate, and 5% H2) and combined. (a) B. theta growth relative to a positive control. (b) Decision tree representation of growth data relative to glucose, hematin, and vitamin B12 control condition. The decision tree representation is read from left to right where each node represents one of the 7 nutrient conditions. The “−” path means the nutrient was not included, and the “+” path means it was included. As the path is traversed, the nutrients are additive. The values in parentheses represent how many total conditions are covered by that leaf, followed by how many are incorrectly classified by the model. Growth relative to the positive control is indicated as follows: open circles, no growth (NONE, <0.25); gray circles, low growth (LOW, ≥0.25, <0.75); blue circles, similar growth (SIMILAR, ≥0.75, <1.25); and pink circles, high growth (HIGH, ≥1.25).

FIG 6

Identification of defined medium that supports independent growth. (a and c) Growth of B. theta in dropout medium. (b and d) Growth of M. smithii in dropout medium. Data for each panel were obtained from triplicate biological and 8 technical replicates (n = 24). Error bars represent standard deviation. G, 0.5% glucose; B, vitamin B12; H, hematin; K, vitamin K3; F, 10 mM formate; A, 10 mM acetate; 2, 5% H2 atmosphere.

M. smithii grows best on a combination of glucose, vitamin B12, hematin, acetate, formate, and H2 gas but depends on acetate for biomass.

As expected, B. theta is incapable of growth (Fig. 6b) under conditions that M. smithii prefers (Fig. 6d). M. smithii grew well in defined medium containing formate, acetate, and hydrogen, even with the addition of glucose, vitamin B12, and hematin (Fig. 6c). Overall, M. smithii was able to grow under a wider variety of conditions than B. theta, albeit to a lower maximum optical density (Fig. 7a). The decision tree for M. smithii suggests that acetate is the primary determinant for the growth of M. smithii (Fig. 7b). The presence of hydrogen or formate is necessary but not sufficient for growth (Fig. 6d). Both vitamin K3 (menadione) and vitamin B12 can inhibit M. smithii growth, but under certain circumstances vitamin B12 rescues vitamin K inhibition and hematin can rescue vitamin B12 inhibition (Fig. 7b).

FIG 7

Effect of medium composition on M. smithii growth. Triplicate M. smithii monocultures were grown in replicates of 8 on defined medium supplemented with 128 different combinations of glucose, vitamin B12, hematin, vitamin K3, acetate, formate, and a 5% H2 atmosphere. Data across 3 experiments were compared to a positive control of acetate, formate, and 5% H2 and combined. (a) M. smithii growth relative to a positive control. (b) Decision tree representation of growth data relative to H2, acetate control condition. Growth relative to the positive control is indicated as follows: open circles, no growth (NONE, <0.25); gray circles, low growth (LOW, ≥0.25, <0.75); blue circles, similar growth (SIMILAR, ≥0.75, <1.25); and pink circles, high growth (HIGH, ≥1.25).

Dynamics of coculture phenotypes.

To compare the growth of the cocultures to the monocultures, growth was normalized to the control treatment that contains the necessary additive requirements for both organisms: glucose, hematin, vitamin B12, formate, acetate, and H2. Relative to that of the control, coculture growth closely resembles the B. theta monoculture tree after 1 day (Fig. 8a) but increasingly resembles M. smithii by day 7 through day 14 (Fig. 8bc). On day 1, glucose is the primary growth factor (Fig. 8a). By day 7, biomass is primarily linked to glucose catabolism, but acetate and hydrogen gas also influence coculture growth (Fig. 8b). By day 14, growth is dependent less on glucose and more on acetate, indicating increased importance of M. smithii in the coculture (Fig. 8c). However, the decision trees for the coculture do not directly mimic the monoculture trees. These data indicate that the metabolic relationship between B. theta and M. smithii is dynamic over a 2-week period and that coculture growth is not a simple additive relationship that can easily be predicted from monoculture data. This is consistent with observations made by others in which metabolism of diverse microbes is shaped by other members in a consortium (53, 54).

FIG 8

Decision trees of coculture growth results. Triplicate B. theta plus M. smithii cocultures were grown in replicates of 8 on glucose defined medium supplemented with 64 different combinations of vitamin B12, hematin, vitamin K3, acetate, formate, and a 5% H2 atmosphere. (a) Cocultures after 1 day of growth. (b) Cocultures after 7 days of growth. (c) Cocultures after 14 days of growth. Data were obtained from triplicate biological and eight technical replicates (n = 24). Growth relative to the positive control is indicated as follows: open circles, no growth (NONE, <0.25); gray circles, low growth (LOW, ≥0.25, <0.75); blue circles, similar growth (SIMILAR, ≥0.75, <1.25); and pink circles, high growth (HIGH, ≥1.25).

The synergistic interaction index uncovers nutrient-dependent trophic interactions.

The complex growth patterns we observed suggested that under some conditions, B. theta and M. smithii could be growing independently, while in other culture media they may cross-feed each other and/or may compete for nutrients. We developed a synergistic interaction index to analyze the growth time course data (Fig. 9). Because B. theta will not grow without glucose, only the cases where glucose was supplemented were examined, and the synergistic interaction index (SI) was determined at 7 and 14 days (Fig. 10, supplemental material). The SI for the control condition was 1 on both days, indicating an additive mutualistic relationship between B. theta and M. smithii growth as expected (Table 2 and 3). After 7 days, two cocultures (glucose plus vitamin B12 plus heme plus H2 and glucose plus vitamin B12 plus H2) had an SI of ≥1.5, indicating that the relative growth of the coculture was at least 1.5× that of the two monocultures together (supplemental material). One of these conditions (glucose plus vitamin B12 plus H2) had an SI of 2.307, indicating coculture growth 2 times that of the monocultures together (Table 2 and 3). By 14 days, 7 cocultures had an SI of ≥1.5 with a maximum value of 2.36 for the glucose plus vitamin B12 plus H2 condition (Table 2 and 3). Cultures showing the highest indexes on day 14 were supplemented with vitamin B12, which is surprising because B12 inhibited M. smithii growth in monoculture (Fig. 6b). One of the seven conditions with an SI of ≥1.5 included two fermentation products, acetate and hydrogen, and four cultures supplemented with acetate showed an increase of at least 0.5 SI units between days 7 and 14. Of the 11 cultures with day 14 SI of >1.25, 10 showed an increase from day 7 to day 14, likely indicating the differences in growth rates between B. theta and M. smithii and suggesting that depletion of glucose (and starvation of B. theta) is required for synergistic growth under these culture conditions. Measurement of SI after 7 days correlated very well with SI measured at 14 days with an r2 value of 0.7922 (Fig. 10a). Twenty-four cultures had an SI of <0.75, suggesting the possibility of antagonistic inhibition or competition between B. theta and M. smithii. Of those 24 cultures, 20 saw a decrease in index from day 7 to day 14. Overall, synergistic growth was dependent on the availability of heme, vitamin B12, acetate, and hydrogen (Fig. 9).

FIG 9

Metabolic synergistic interaction index (SI). (a) Effect of nutrient supplementation on growth. (b) Effect of nutrient combinations on growth. (c and d) SI of conditions shown in panels a and b, respectively. An index of 1 (dotted gray line) indicates the null hypothesis where organisms grow independently in coculture. An index greater than 1 indicates that the relative growth of the coculture is greater than the sum of the monocultures, suggesting a syntrophic relationship. An index of less than one indicates a competitive or inhibitory effect. Error bars represent standard deviation from triplicate biological and eight technical replicates (n = 24). G, 0.5% glucose; B, vitamin B12; H, heme; K, vitamin K3; F, 10 mM formate; A, 10 mM acetate; 2, 5% H2 atmosphere.

FIG 10

Synergistic coculture growth is inversely related to B. theta growth. (a) Metabolic relationships are established and stable between 7 and 14 days in culture. (b) B. theta monoculture relative growth is inversely correlated with the 7-day synergistic interaction index (SI). (c) M. smithii monoculture relative growth is not correlated with the 7-day SI.

TABLE 2

Synergistic interaction index statistics

Index range (Ii)	No. of treatments on day:
	7	14
>1.5	2	7
>1.25	8	11
0.75≥ x ≥1.25	79	5
<0.75	24	29
<0.5	2	1

TABLE 3

Synergistic interaction index statistics

Extrema	Index value (Ii) on day:
	7	14
Maximum	2.307	2.366
Minimum	0.376	0.397
Positive controla	0.928	0.950

Defined medium supplemented with glucose, histidine and hematin, vitamin B12, formate, acetate, and H2. In this treatment, B. theta and M. smithii are provided nutrients for both to grow independently.

Synergistic growth is correlated with nutrient limitation and feedback inhibition.

We noted that conditions indicating coculture synergy occurred when B. theta monocultures grew poorly. When the day 7 SI is compared to the growth of B. theta in monocultures under the same conditions, we observed a strong inverse correlation with an r2 value of 0.6118 (Fig. 10b). In contrast, the day 7 SI was correlated only very weakly with M. smithii growth with an r2 value of 0.0657 (Fig. 10c). The decision tree analysis (Fig. 8) shows that highest growth occurs when the coculture is supplemented with acetate, formate, and hematin. When cocultured, the data suggest that both organisms compete for hematin (Fig. 9c and d), especially when M. smithii growth is stimulated by added formate, acetate, and/or hydrogen (supplemental material).

DISCUSSION

Our study highlights the benefit of using an interdisciplinary approach to discern patterns of microbial growth as environmental conditions are varied. We have been able to adapt the use of decision trees from machine learning to parse through large-scale phenotype data to identify critical nutrient factors that contribute to growth. Machine learning decision trees are an unbiased tool to search for patterns in large-scale data to uncover hierarchy and interactions between variables. Here, we used optical density to measure culture growth as an output-dependent variable due to the ease of obtaining high-throughput data. However, decision trees can be adapted to use for any combination of dependent and independent variables in biological systems. This approach allowed us to ascertain patterns in growth phenotypes as well as emergent behaviors, in this case competition and synergistic growth, when two species are cocultured. We then developed a method to quantify these emergent behaviors in the form of a synergistic interaction index. Our approaches apply well to studying the interactions between microbes as diverse as the Gram-negative bacterium B. theta and the archaeal methanogen M. smithii. While we used endpoint optical density as an output parameter, which may mask phenotypes such as changes in cell size, shape, lysis, and aggregation, other complementary signals, such as 16S sequence or metatranscriptomic abundance, biomass, metabolites produced or consumed, or the rates of change of these signals, could theoretically be used instead with minimal adaptation. Likewise, while we constrained our study here to two organisms, with minor adjustment to account for number of species the approach could be retooled for more complex microbial communities.

A major consideration when studying complex microbial communities would be to use statistical sampling techniques such as BioSIMP (55) to develop technically and logistically feasible experiments that still retain a suitable confidence interval for observed responses. BioSIMP involves using statistical sampling techniques to reduce the number of experiments required to uncover unexpected behavior/interactions in a biological system. For instance, instead of testing a full-factorial array of environmental conditions, statistical subsampling can be used to reduce the experiments by 50% or more to find which culture conditions can be expected to produce a growth effect of a certain magnitude. Precious resources in time, money, materials, and personnel can be devoted to more detailed mechanistic experiments that require full-factorial experimentation once “interesting” or “unexpected” results are identified for follow-up study.

The results of this study show that fermentation products secreted by B. theta are sufficient to support growth of M. smithii (Fig. 1c) and that M. smithii enhances growth of B. theta (as determined by growth curve data [Fig. 1c], microscopy [Fig. 2e], and qPCR experiments). One explanation for this result could be that B. theta becomes feedback inhibited by secreted metabolic by-products but in the presence of M. smithii the products are consumed, thus reducing growth inhibition. We showed in a previous study that the major fermentation products secreted by B. theta during growth on glucose suppress B. theta growth rates and alter metabolism, resulting in increased production of carbon dioxide and amino acids (52). M. smithii is capable of growing solely on the metabolites secreted by B. theta (Fig. 1c), and medium that has been sequentially conditioned first by B. theta and then by M. smithii supports higher B. theta biomass than medium that was preconditioned by B. theta alone (Fig. 4a). It is unknown what M. smithii might secrete besides methane, biomass, or intracellular metabolites such as heme or corrinoids released by lysed cells. Heme iron is a well-studied nutrient requirement, and bacteria have many heme acquisition mechanisms, such as siderophores and pathogenesis factors, that they synthesize to acquire iron for enzyme cofactor synthesis from the environment, from neighboring microbes, or from the host (56, 57). Corrinoids have also been shown to play a key role in several interspecies systems, and competition for B12 is thought to be a significant factor that shapes human gut microbial community ecology (58). Our results show that B. theta is capable of growth in medium without supplemented heme or B12 when grown in coculture with M. smithii presumably because B. theta can scavenge heme and corrinoids synthesized by M. smithii.

It is formally possible that B. theta and M. smithii secrete one or more unknown small molecules that are sensed as interspecies signals in addition to the central metabolic interactions we tested in our treatments. Synergistic growth could be mediated by soluble and insoluble signals that may be expressed when B. theta is undergoing the starvation response. It is tempting to speculate that B. theta forms multicellular aggregates under starvation conditions to attract M. smithii, which would benefit from physical association with B. theta by allowing rapid mass transfer of hydrogen and acetate (Fig. 2). Intriguingly, these aggregates and the association of M. smithii with them statistically decreased when acetate was supplemented into the culture medium (Fig. 3), suggesting either that acetate induces signaling molecules by B. theta or that M. smithii, which requires exogenous acetate for growth, secretes molecule(s) that promote aggregate formation with B. theta. Whether the increase in B. theta growth is a result of secreted product from M. smithii or by the protective effect of removing inhibitory compounds is still to be decided. Additional experiments are needed to identify any attractants secreted by B. theta and/or M. smithii under conditions that favor and disfavor aggregate formation.

Finally, we developed a synergistic interaction index to score the likelihood that two or more organisms are growing synergistically or independently or are inhibiting each other. The index scores suggest that in some situations, the coculture is metabolically more efficient than expected from the biomass observed from independent monocultures. While the index does not identify the molecular mechanism of synergy or inhibition, by using a multicomponent culturing strategy and decision trees, we are able to discern patterns of coculture behavior that can lead to testable hypotheses. Our observations suggest that B. theta and M. smithii have a mutually beneficial syntrophic relationship when vitamin B12 and hydrogen gas are provided and that they compete for hematin when sufficient acetate, formate, and/or hydrogen gas are available for M. smithii to grow. Future experiments are needed to determine how B. theta and M. smithii compete for heme. We speculate that competition could proceed via passive mechanisms (autolysis, diffusion, and ATP-dependent transport) or by active processes using small molecules such as bacteriocins, toxins, quorum-sensing factors, or other mechanisms to obtain iron that have been discovered in other microbes (59–61).

MATERIALS AND METHODS

Strains and culturing conditions.

B. theta vpi-5482 (ATCC 29148, NB203) (62, 63) and M. smithii DSM0861 (ATCC 35061, NB215) (44) were used for the described studies. Strains were grown at 37°C in 18 mm by 150 mm Balch culture tubes under strict anaerobic conditions, in either a rich tryptone and yeast extract growth medium (TYG) or a defined medium. Previous work with B. theta utilized tryptone, yeast extract, and glucose medium (TYG) as a rich growth medium (15, 35, 64). Rich medium for M. smithii was also dependent on yeast extract (65, 66). Using TYG as a base, the recipes were combined to ensure growth for both organisms. M. smithii was initially grown in DSMZ Methanobacterium medium 119 (66) and then passaged into the modified rich medium supplemented with 10 mM formate and 10 mM acetate with 138 kPa 20% CO2 80% H2 headspace atmosphere. The defined medium was designed through a comparison of recipes (44, 45, 67–70). Rich medium contained tryptone peptone (10 g), Bacto yeast extract (5 g), 100 mM KPO4 (pH 7.2), 40 mL TYG salts (0.5 g MgSO4·7H2O, 10 g NAHCO3, 2 g NaCl per L), 54.4 μM CaCl2, 1.4 μM FeSO4, 4 μM resazurin, 2.8 mM cysteine·HCl, and 25 μM Na2S per L. Defined medium contained 7.5 mM NH4SO4, 11.9 mM Na2CO3, 100 mM KPO4 (pH 7.2), 14 μM FeSO4, 50 mL mineral salts (18 g NaCl, 0.53 g CaCl2·2H2O, 0.40 g MgCl2·6H2O, 0.20 g MnCl2·4H2O, 0.20 g CoCl2·6H2O per L), 4 μM resazurin, 2.8 mM cysteine·HCl, and 25 μM Na2S per L. When indicated, cultures were supplemented with 10 mM sodium acetate, 10 mM formate, 20 μM histidine and 2 μM hematin, 3.7 nM vitamin B12 (cyanocobalamin), or 5.8 mM vitamin K3 (menadione). B. theta was supplemented with glucose to 0.05% (2.78 mM). Culture headspace was either 20% CO2, 80% N2 or 5% H2, 20% CO2, 75% N2 atmosphere at 138 kPa. M. smithii was grown on a rotary shaker operating at 45 rpm. M. smithii culture tubes were pressurized to 138 kPa with a 20% CO2 80% H2 gas mixture twice daily during growth curves or every 3 days for culture maintenance. Preliminary data suggested that vitamin K3 may hinder M. smithii growth. B. theta/M. smithii cocultures were grown on or off the shaker under a 5% H2, 20% CO2, 75% N2 atmosphere in rich or defined medium supplemented with glucose to 0.05% (2.78 mM).

Growth curves.

B. theta and M. smithii were grown in 10 mL TYG and then transferred to media containing appropriate carbon sources and compounds. Growth in 18 mm by 150 mm anaerobic culture tubes was assessed by measuring changes in optical density at 600 nm using a Spec20D spectrophotometer modified with an 18 mm tube adapter. Growth in 96-well plates was assessed by measuring change in optical density at 600 nm using a Tecan Sunrise plate reader under a 5% H2, 20% CO2, 75% N2 atmosphere.

Microscopy.

Microscopy was performed using an EVOS FL auto cell imaging system with DAPI (4′,6-diamidino-2-phenylindole), Texas Red, and green florescent protein LED light cubes in the University of Nebraska Morrison Microscopy Core Facility. B. theta and M. smithii were grown in medium containing an appropriate carbon source as described above. Cocultures were grown in defined medium with only glucose as a carbon source. Five hundred-microliter samples were taken in an anaerobic atmosphere and stained with a combination of 5 μL propidium iodide and 1 μL SYTO 9 green per 500 μL cells while remaining in anaerobic conditions. M. smithii coenzyme F420 autofluorescence was viewed using a DAPI light cube (6, 38, 44, 71), propidium iodide with a Texas Red light cube, and Syto 9 green with a green florescent protein light cube.

Flow cytometry.

Flow cytometry was performed using a BD Biosciences FACS Aria II flow cytometer (BD Biosciences) in the Nebraska Center for Biotechnology Flow Cytometry Core Facility. B. theta and M. smithii were grown anaerobically in rich medium supplemented with variable concentrations of glucose and sodium acetate over 4 and 9 days. Cell concentrations were recorded via optical density at 600 nm (OD600) and cultures were concentrated anaerobically via centrifugation at 500 × g for 10 min in a ThermoScientific Sorvall Legend Micro21 rotor, washed with phosphate-buffered saline (PBS; 137 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, 1.8 mM KH2PO4), and resuspended to a concentration of ~1.0 × 106 cells per mL. Cells were dyed using LIVE/DEAD fixable yellow dead cell stain kit (Invitrogen) according to the protocol provided. Flow cytometry was performed exciting with a 405 nm laser until 10,000 events were recorded. Flow cytometry size standards were obtained from Fisher (1.0 to 15 μm diameter; catalog number F13838). The gating control is shown in the supplemental material in Fig. S1.

qPCR quantification.

B. theta and M. smithii were grown in rich media supplemented with glucose. Cultures were incubated at 37°C, and 1-mL samples were collected anaerobically on days 4 and 9 of growth. DNA was isolated from the cells using a phenol-chloroform extraction (72), and qPCR was performed on a Mastercycler RealPlex (2) instrument (Eppendorf) detecting SYBR green using probes for 16S regions of the B. theta (B theta 16S fwd: 5′GGGATGCGTTCCATTAGGC; B theta 16S rev: 5′GGGACCTTCCTCTCAGAACC) and M. smithii (M smithii 16S fwd: 5′CGGCCGATTAGGTAGTTGGT; M smithii 16S rev: 5′GTTCCATCTCCGGGCTCTT) genomes. Threshold cycle (C) values were normalized to input DNA, amplification efficiency, and the apparent number of genomes per cell. To account for variability in chromosome copy number between the two organisms, an apparent genome count was calculated by counting cells using a hemocytometer and carefully extracting the DNA from 1 mL of cells. The DNA quantity was measured via NanoDrop spectrophotometer (Thermo Scientific) and was divided by the genome size to calculate average number of genomes per cell, which would then be divided by the number of cells harvested in 1 mL.

Dropout media preparation and growth assays.

For each biological replicate of the 7-component growth assay, dropout media were prepared that contained all combinations of vitamin K3, vitamin B12, formate, acetate, histidine-hematin, and glucose. Media were dispensed into 16 sterile 96-well culture plates, with two plates for each layout. One plate of each layout was kept under a 5% H2, 0.1% H2S, 20% CO2, 74.9% N2 atmosphere, and the other was kept under 0.1% H2S, 20% CO2, 79.9% N2, resulting in 128 media combinations total.

B. theta and M. smithii were grown in 10 mL rich medium and used to inoculate prepared 96-well culture plates. For monocultures, 1:20 inocula (5 μL) of B. theta, 1:10 inocula (10 μL) of M. smithii, or a coculture of both were added to sample wells. The inoculum volumes for each strain were empirically determined to yield measurable culture turbidity within the 2-week experiment. One plate of each layout was placed in a 35°C anaerobic incubator with either 5% H2, 0.1% H2S, 20% CO2, 74.9% N2 or 0.1% H2S, 20% CO2, 79.9% N2 atmosphere. A Tecan Sunrise plate reader measured optical density at 600 nm after 7 days or 14 days under strict anaerobic conditions. The process was repeated for a total of three biological replicates.

Growth data analysis.

The plate reader data were expected to contain errors due to splashing during handling, evaporation, and the formation of bubbles or cell clumps. Because each set of 8 samples or blanks contained similar contents, cells, and medium components, a normal distribution was assumed and Chauvenet’s criterion was applied to eliminate statistical outliers in preparation for statistical and algorithmic analysis (73). Chauvenet’s criterion specifies a probability band around the mean with a probability of . Data within the band are retained, while data outside the band are considered outliers. For 8 samples, Chauvenet’s criterion specifies a probability band that encompasses 93.8% of the population (equation 1):

This corresponds to 1.863 standard deviations from the mean. Samples or blanks outside the band with a standard deviation greater than 1.863 were eliminated as outliers. After applying Chauvenet’s criterion, differences in medium color were minimized by the subtraction of the mean value of the medium blanks from each sample. To allow data comparison across experiments, all samples were divided by the mean of a universal positive control grown in a H2 atmosphere and containing glucose, vitamin B12, hematin, formate, and acetate. These samples were joined. To compare error across biological replicates, we assumed that the standard deviations should be approximately equal across data sets and calculated pooled variance according to equation 2: where n is the sample size of population i and is the sample variance for population i or square of the standard deviation. Pooled standard error can then be calculated with equation 3:

Trees.

Decision trees are a divide-and-conquer machine learning technique that sorts data according to the data attributes that best divide the data. C4.5 decision trees were generated by running the data sets into a J28 classifier as previously described and drawn using Adobe Illustrator (36). BioSIMP software and tutorials can be found on the U.S. Department of Energy’s Knowledge Base (KBase) (74). For analysis of monoculture data and coculture time course, sample averages were sorted into 4 buckets relative to a positive control containing glucose, hematin, vitamin B12, acetate, and formate and grown under a 5% H2 atmosphere. Buckets related the growth to the positive control as follows: no growth (NONE, <0.25), low growth (LOW, ≥0.25, <0.75), similar growth (SIMILAR, ≥0.75, <1.25), and high growth (HIGH, ≥1.25).

Synergistic interaction index calculations.

If all the cells in a coculture or consortium grow entirely independently of each other, with no interactions, the number of cells in the coculture (ncoculture) would be equivalent to the sum of the number of monoculture cells grown under the same conditions, or in our case (equation 4):

Because the optical density of the culture is directly proportional to the number of cells, we can compare the relative growth, R,, of each culture of taxon s under culture condition i (equation 5):

As long as each organism can grow independently in a positive control (ctrl) condition, an x component coculture of B. theta and M. smithii should have an overall growth that is an average of the monocultures (equation 6). Therefore, assuming independent growth within a coculture of B. theta and M. smithii, the relative growth, R, of a coculture can be calculated with equation 7.

This relationship will be true only if the growth of B. theta and that of M. smithii are independent of each other (null hypothesis). If the cells are interacting in a positive, syntrophic way, the coculture will contain more cells than the sum of the monocultures. If the cells are competitive, total growth will be inhibited and there will be fewer cells in the cocultures than expected. To express coculture interactions as a numerical value, we divided the observed coculture growth by the sum of the independent cultures, creating a synergistic interaction index for each culture condition, I (equation 8).

A few caveats to this method should be considered depending on the organisms being studied. Under very high culture densities, the linear relationship between optical density and cell number is no longer valid, as the culture becomes increasingly opaque. In these situations, the index would be an underestimate of the coculture productivity. Another issue to keep in mind is that flocculation or aggregate formation would complicate experimental reproducibility. In spectrophotometric microplate readers, aggregates may be indicated when biological replicates produce signal variability depending on whether the beam hits an aggregate by chance. The result is high biological and technical variability that confounds statistical analysis. In either case, the index may be adapted by substituting biomass or another proxy growth measurement in place of OD variables in the above equations.

Data availability.

Data from this study are available in the figures, tables, and supplemental material. The software tools developed for this study are available as user-friendly tutorials in the Department of Energy KBase at https://narrative.kbase.us/narrative/19635.