Cobamide-based interactions between soil bacteria can be predicted based on monoculture growth

Abstract

Interactions between microbes shape the structure and function of microbial communities. While studying interactions is key to understanding microbial communities as a whole, gaining a detailed mechanistic view is challenging due to the scale of co-occurring interactions. The model nutrient approach enables the study of a subset of interactions involving a single nutrient class and can shed light on broader interaction mechanisms involving other nutrients. Here, we focus on cobamides, the cobalamin (vitamin B₁₂) family of enzyme cofactors, to study nutrient competition and nutrient-sharing interactions in co-cultures and tri-cultures. We examined bacteria that were previously isolated from a grassland soil and were characterized as “dependents” (require cobamides but cannot synthesize them) or “producers” (synthesize cobamides). The outcome of competition between a pair of dependents was predictable based on monoculture growth characteristics, with the dominant microbe determined by its adaptation to a specific cobamide concentration range. Moreover, cobamide producers could support the cobamide-dependent growth of dependents in co-culture and influenced the outcome of competition between dependents in tri-culture. We analyzed the metabolic capacity encoded in the genomes of producers and dependents and found that cobamides are likely the main shared nutrient in our co- and tri-cultures. These results highlight the utility of the model nutrient approach to characterize and predict interactions in bacterial consortia of increasing complexity.

Introduction

From soils to oceans to the human gut, microbial communities shape processes crucial for human health and the planet. There is growing interest in predicting how different biotic and abiotic factors impact microbial communities (1-5). Interactions between microbes are strong determinants of community structure and function, with single microbes or single metabolites having the ability to impact the community broadly (6-10). Thus, deciphering how microbe-microbe interactions occur and are regulated is crucial for understanding whole communities. However, microbial communities encompass complex networks with many metabolites being produced and consumed at a given time, making it difficult to observe and disentangle specific interactions (11-14).

Applying a model nutrient approach allows us to focus on interactions involving a single shared nutrient. This approach enables the mechanistic study of specific microbial interactions, while setting aside other interactions that could have confounding effects (15, 16). Cobamides, the cobalamin (vitamin B₁₂) family of cofactors (Fig. 1), are useful as model nutrients for several reasons. First, cobamide-sharing interactions have been observed in bacterial co-cultures (15, 17-20). Moreover, cobamides are predicted to be shared widely among bacteria, based on genomic data showing an estimated 86% of sequenced bacterial species use cobamides yet only 36% have the complete biosynthesis pathway (21). Second, they are required enzyme cofactors for diverse metabolic processes that are relevant in different environments, including metabolism of specific carbon and nitrogen sources, methionine synthesis, and reductive dehalogenation (21, 22). Third, bacteria have distinct preferences for different cobamides that impact their growth and metabolism (23). These characteristics, combined with prior foundational research on cobamide biosynthesis and use, give us the ability to study cobamide-based microbial interactions. The investigation of interactions through the lens of a model nutrient can enhance our understanding of microbial competition and cooperation in laboratory consortia and in the environment.

Figure 1. Cobamides used in this study.

Structures of cobalamin (B₁₂) (left) and the lower ligands of 2-methyladeninylcobamide ([2-MeAde]Cba) and 5-hydroxybenzimidazolylcobamide ([5-OHBza]Cba) (right) are shown. The abbreviation used for each cobamide is in parentheses next to each structure. The lower ligand of cobalamin, 5,6-dimethylbenzimidazole, is shown inside the rectangle.

We previously isolated bacteria from a grassland soil in California and characterized their cobamide metabolism, gathering information necessary for application of the model nutrient approach to soil microbial communities (24). We classified isolates capable of producing cobamides de novo as “producers” and those that required cobamides for growth as “dependents” and observed that dependents in our isolate collection preferred cobalamin over other cobamides but could use different cobamides to various extents (24). The soil isolates we evaluated required lower cobamide concentrations for growth when compared to bacteria from other environments, suggesting that soil microbes have adapted to relatively low concentrations of bioavailable cobamide (23, 25). Because cobamide structure and concentration can greatly influence the growth and metabolism of bacteria (23), it is likely that competition between dependents is shaped by the identity and amount of cobamide present. Thus, we hypothesized that cobamide-dependent growth in monoculture can be used to predict the outcome of competition in consortia where competition for cobamides is the predominant interaction.

All of the cobamide producers that we isolated from soil were found to synthesize cobalamin in monoculture (24). Of this set, all contained intracellular cobalamin but only a subset of producers had detectable amounts of cobalamin in culture supernatants. We labeled this subset as cobamide “providers”—producers that release cobamides into the environment (24, 26). Although the mechanism of cobamide release remains unknown, cobamides have been detected in culture supernatants of many bacteria (16, 24, 26). Some producers released cobalamin at levels exceeding the requirements of the dependents, leading us to postulate that these bacteria could support the growth of dependents in co-culture. Further, we hypothesized that the extent to which each provider could support dependents would be predictable based on the amount of extracellular cobamide detected in monoculture.

Here, we characterized cobamide-based interactions among four bacterial isolates—two dependents with different growth profiles on cobamides and two producers that release different amounts of cobamide in monoculture (24). We grew these bacteria in co- and tri-cultures to test the hypothesis that growth and cobamide measurements in monoculture and co-culture can reliably predict community dynamics in co-culture and tri-culture, respectively. Indeed, we found that the outcome of competition between dependents can be predicted based on their different cobamide requirements in monoculture, which may reflect the distinct niches they occupy in the environment. Further, we determined that cobamides released by producers can support the growth of dependents in co-culture. Finally, we found that cobamide release by a producer can impact the outcome of competition between dependents in tri-cultures. The centrality of cobamide-based cooperative and competitive interactions among these soil bacteria is reinforced by a comparison of their genome sequences. Overall, these insights contribute to a better comprehension of pairwise bacterial interactions and enhance our understanding of how cobamide diversity may impact microbial community dynamics at higher levels of complexity.

Results

Dependents have advantages at distinct cobamide concentrations

We hypothesized that the outcome of competition between two dependent bacteria could be predicted based on their relative cobamide-dependent growth rates in monoculture (23). We tested this hypothesis on a pair of cobamide-dependent soil isolates, Ca19 and Ca24A, initially selected because they grew reproducibly with no aggregation in liquid culture. Through the genome sequencing effort described in a later section, we found that the isolates belong to two different species of Caulobacter, as reflected by a whole genome average nucleotide identity (ANI) of 89% (27). The species were not identified by GTDB-Tk (28, 29), which suggests that they could be novel species.

We measured the maximum growth rates of Ca19 and Ca24A in media containing a range of concentrations of [5-OHBza]Cba, cobalamin, and [2-MeAde]Cba to determine how these cobamides influence growth in monoculture (Fig. 2A, D, G). Ca19 has a higher growth rate than Ca24A at low concentrations of [5-OHBza]Cba, whereas Ca24A grows faster at high concentrations (Fig. 2A). We observed a similar trend on cobalamin, but with significant differences in growth rate (p < 0.05) at fewer tested concentrations (Fig. 2D). The cobamide concentration required to reach the half-maximal growth rate (EC₅₀) was lower for Ca19 on these two cobamides (Fig. 2J), suggesting this isolate is adapted to lower cobamide concentrations. In contrast, the two isolates grow at similar rates at all concentrations of [2-MeAde]Cba, with Ca19 growing faster (p < 0.05) at two concentrations (Fig. 2G). We expected that, if cobamide-based competition is the dominant interaction, the isolate with the higher growth rate at a given cobamide concentration should outcompete the other. We tested this hypothesis by establishing co-cultures at different cobamide concentrations and differentiating between the two isolates based on their distinct colony morphologies (Fig. 2B, inset).

Figure 2. Dependent isolate growth rates in monoculture predict the outcome of competition at different cobamide concentrations.

The maximum growth rate of triplicate cultures grown in a range of concentrations of (A) [5-OHBza]Cba, (D) cobalamin, and (G) [2-MeAde]Cba is shown. Curve fits are for a four-parameter least squares fit. Asterisks denote Welch’s t-test p < 0.05 and dotted vertical lines denote the concentrations selected for further experiments. (B, C, E, F, H, I) Graphs show the relative abundance of Ca24A in co-culture with Ca19 passaged daily for seven days, based on the number of colony forming units (CFU/ml) of each isolate in samples taken at the indicated time points. (B, C) [5-OHBza]Cba, (E, F) cobalamin, and (H, I) [2-MeAde]Cba were provided at (B, E, H) low and (C, F, I) high concentrations, as indicated. Black and gray lines show data for five to six biological replicates from two independent experiments performed under the same conditions, except that a higher concentration of [2-MeAde]Cba (I) was used in the experiment shown in gray. The dotted horizontal line indicates 50% abundance, and the color bars indicate the dominating isolate at the corresponding region of the graphs (purple for Ca24A and pink for Ca19). Inset in B shows different colony morphologies of the two dependents grown in co-culture: Ca24A (small colonies, purple arrowhead) and Ca19 (large colonies, pink arrowhead) after 72 hours of growth on R2A agar containing cobalamin. (J) The cobamide concentration that leads to half maximal growth rate (EC₅₀) for each isolate on [5-OHBza]Cba, cobalamin, and [2-MeAde]Cba highlights the differences in their cobamide preferences. Error bars represent 95% confidence interval, three biological replicates each.

Based on their relative growth rates in different concentrations of [5-OHBza]Cba and cobalamin, we predicted that Ca19 would outcompete Ca24A when co-cultured at concentrations less than 0.012 nM [5-OHBza]Cba or 0.010 nM cobalamin, whereas Ca24A should outcompete Ca19 at higher cobamide concentrations (Fig. 2A, D). Predictions of co-culture outcomes in [2-MeAde]Cba were less clear because the average growth rate measurements for the two isolates were very similar at all concentrations (Fig. 2G). Nevertheless, due to the small differences in growth rate, we predicted that Ca19 would outcompete Ca24A at all concentrations of [2-MeAde]Cba.

We co-cultured Ca19 and Ca24A in media containing a low (0.003 or 0.009 nM) or high (1 nM) concentration of each cobamide and passaged every 24 hours to maintain cultures in exponential phase. The viable cell number of each isolate was determined for the inocula and co-cultures on days 1, 3, 5, and 7 (Fig. 2B, C, E, F, H, I).

Predictions for the outcome of competition between the two dependents were accurate in the low and high concentrations of [5-OHBza]Cba and cobalamin, with Ca19 reaching a higher abundance at low cobamide concentrations and Ca24A becoming predominant at high concentrations (Fig. 2B, C, E, F). These results were consistent with our hypothesis that the outcome of competition could be predicted from maximum growth rates in monoculture, which suggests that competition for cobamides is the main interaction driving dynamics in these co-cultures. Overall, the competition for these two cobamides is determined by concentration, suggesting that Ca19 is adapted to lower concentrations of [5-OHBza]Cba and cobalamin and Ca24A to higher concentrations.

In media containing [2-MeAde]Cba, the outcome of competition was not accurately predicted from monoculture growth rates at the low or high concentrations (Fig. 2H, I). The isolates remained at equal proportions in the low concentration, while Ca24A outcompeted Ca19 in the high concentration condition. When the experiment was repeated with a higher concentration of [2-MeAde]Cba, the results of the two replicates diverged (Fig. 2I, gray lines). This is likely because the differences in growth rates or EC50 were not large enough to define a cobamide-dependent outcome of competition on [2-MeAde]Cba.

Producer supernatants support growth of dependents in monoculture

Next, we hypothesized that each dependent isolate could be supported by a producer that releases cobamides into the culture media, and that the amount of cobamide the producer supplies would determine the outcome of competition between the two dependents in tri-culture. We selected two isolates from the same collection—Me12, a strain of Mesorhizobium opportunistum, and Pr24B, a strain of Priestia megaterium—that were previously found to produce and release different amounts of cobalamin. Pr24B was originally isolated in the same culture as the dependent Ca24A, suggesting these bacteria may interact in their natural soil environment, including via cobamide sharing. These producers accumulate cobalamin in stationary phase supernatants at concentrations that exceed the minimal requirements of both dependents in monoculture (24). The colony morphologies of these two producers were easily distinguishable from those of the dependents, making it possible to determine their abundances in co-culture based on colony numbers.

To evaluate the timing of cobalamin release in producer cultures, we prepared cell-free supernatants from producers during exponential and stationary phases of growth (Fig. S1A-C) and measured growth of the two dependent isolates in medium containing these supernatants. Dependent growth was supported by supernatants from both producers (Fig. 3). Me12 supernatants collected during both growth phases supported the growth of both dependents (Fig. 3A). The Me12 stationary phase supernatants led to 3- and 6-fold higher growth of Ca24A and Ca19, respectively, than Me12 exponential phase supernatants ), suggesting it contains a higher cobamide concentration or additional nutrients (Fig. 3A). Growth of Ca19 when provided with the Me12 stationary phase supernatant exceeded that of Ca19 in medium containing saturating cobalamin (Fig. 3A), suggesting Me12 releases additional nutrients that benefit Ca19. In contrast, while the stationary phase supernatant of Pr24B supported growth of both dependents, its exponential phase supernatant did not enhance growth of Ca24A relative to the unamended control, and it only slightly enhanced the growth of Ca19 (Fig. 3B). This suggests that cobamide release is regulated differently in the two producers.

Figure 3. Growth of cobamide dependent isolates in medium amended with producer supernatants.

Graphs show CFU/ml increase of dependent cultures supplemented with exponential (exp) or stationary (stat) phase supernatants from (A) Me12 or (B) Pr24B. Cobalamin addition (1nM B12) and no addition (NA) controls are shown for each experiment. All cultures were grown for 24 hours. Bottom portion of each graph shows points below zero on a linear y-axis. Points represent individual replicates and horizontal lines show the median CFU/ml increase for each isolate.

Producers can support dependents by providing a cobamide in co-culture

We then tested whether producers could support dependent growth in co-culture, with the hypothesis that the isolates would grow differently in co-cultures, where release of cobalamin by the producer could enable the dependent to compete more effectively for other resources. We found that both producers supported dependent growth in co-culture. When co-cultured with Me12, dependent isolates Ca19 and Ca24A reached abundances similar to monocultures supplemented with a saturating concentration of cobalamin (1 nM) (Fig. 4). Me12 provided cobalamin despite being at a competitive disadvantage in co-culture, reaching a density around 4-fold less than when grown alone (Fig. 4). When co-cultured with Pr24B, both dependents grew to higher levels than in monocultures with no cobalamin, but their final densities were over 10-fold lower than when grown in monoculture with saturating cobalamin (Fig. 4). Furthermore, unlike Me12, Pr24B grew to a similar level in co-culture and in monoculture, perhaps because it experienced less competition with the dependents due to the dependents’ restricted cobamide availability (Fig. 4).

Figure 4. Growth of cobamide dependent isolates in co-culture with a producer.

The increase in colony forming units per ml is shown for monoculture controls in medium lacking cobalamin or containing 1 nM cobalamin (B12) and co-cultures of each producer with each dependent. All cultures were grown for 48 hours. Bottom portion of the graph shows points below zero on a linear y-axis. Points represent individual replicates and horizontal lines show the median CFU/ml increase for each isolate.

Upon comparing the number of dependent cells supported per producer cell, we found that dependents generally benefited more from being supplemented with supernatants than from growing in co-culture (Table 1). Producer growth and metabolism are likely different in monoculture and co-culture, due to factors like competition for resources, which might alter cobalamin availability. Nonetheless, these two producers support dependents by providing cobamides when grown together. Pr24B supernatant supports fewer dependent cells than Me12 supernatant because its cell numbers are much lower (Figs. 3, S1). However, the calculated ratios show that both Pr24B and Me12 support similar numbers of dependent cells per producer cell (Table 1).

Table 1.

Number of dependent cells supported per producer cell when grown with supernatants from the exponential phase (supernatant exp) or stationary phase (supernatant stat) and in co-culture.

		Dependent cells per producer cell
Producer	Condition	Ca19	Ca24A
Pr24B	supernatant exp	41 ± 24	−2.2 ± 7.4
	supernatant stat	130 ± 57	81 ± 17
	coculture	1.7 ± 0.64	1.5 ± 0.17
Me12	supernatant exp	32 ± 4.2	14 ± 7.4
	supernatant stat	20 ± 4.0	4.0 ± 0.30
	coculture	3.9 ± 0.88	5.3 ± 1.0

Cobamide providing and competition occur in three-species consortia

Upon confirming that each producer could support both dependents, we sought to determine how producers influence the growth of the two dependents when competing for cobalamin. We hypothesized that the amount of cobalamin released by each producer would determine which dependent was more abundant at the end of the competition. Thus, we established tri-cultures containing the two dependents with each producer in a cobamide-free medium with daily passaging into fresh medium. On days 1, 3, 5, and 7, the cultures were plated to determine the abundances of each isolate based on their unique colony morphologies (Fig. 5A, C).

Figure 5. Outcome of competition between Ca19 and Ca24A in tri-culture with each producer.

(A, C) Colony morphologies of both dependent isolates with (A) Me12 and (C) Pr24B after 96 and 72 hours of growth on solid medium, respectively. Arrow colors correspond to the color for each isolate in the rest of the figure. (B, D) Relative abundances of producer and both dependents in tri-culture with (B) Me12 and (D) Pr24B are shown. Bars represent the average of three biological replicates and error bars represent standard deviation. The tables in E and F show average and standard deviation for total cobalamin measured in the cells and supernatants via an E. coli bioassay in the tri-cultures and in monocultures of (E) Me12 and (F) Pr24B (n=3). <0.27 indicates the measurement was below the bioassay detection limit of 0.27 nM.

In both tri-cultures, all three isolates persist over the seven passages and the producer had the lowest abundance at most timepoints. While Me12 had similar abundances at the beginning and end of the experiment, Pr24B was barely detectable at the initial timepoint and increased to over 10% by day 7. Ca19 was the more abundant of the two dependents by day 3 in both tri-cultures. Ca19 relative abundance further increased at later timepoints while Ca24A abundance decreased (Fig. 5B, D).

We measured total cobamide in cells and supernatants of the tri-cultures and producer monoculture controls via an E. coli-based bioassay to evaluate if the outcome of competition correlated with cobamide concentrations (30). In the Me12 tri-culture, cobalamin levels were below the bioassay detection limit (0.27 nM) at all timepoints (Fig. 5E). These low cobalamin concentrations are the result of low Me12 cell numbers (on the order of 10³ CFU/ml) likely due to growth suppression by the dependents (Fig. 4) and 24-hour passages preventing this producer from reaching stationary phase (Fig. S1A). The E. coli bioassay detection limit is higher than the “low” cobalamin concentration of 0.003 nM (Fig. 2), making it difficult to draw a conclusion from these data. However, Ca19 outcompeting Ca24A in this tri-culture is consistent with low available cobalamin in the tri-culture supernatant (Figs. 5B, 2E).

The tri-culture containing Pr24B had a stable cobalamin concentration of approximately 2 nM over multiple passages (Fig. 5F). This concentration was also sustained in Pr24B monocultures passaged in the same medium, suggesting that the dependents do not trigger changes to cobamide production in Pr24B (Fig. 5F). From our dependent co-culture experiments we expected Ca24A to outcompete Ca19 in high cobalamin concentrations (Fig. 2F), but the opposite occurred, likely because the amount of cobalamin available to dependents during the exponential growth phase of Pr24B is low (Fig. 3B). Thus, cobamide concentrations in the tri-culture supernatant were likely low during most or all of the time between passages. This tri-culture result (Fig. 5D) aligns with our observation that exponential phase supernatants from Pr24B favored Ca19 (Fig. 3B, Table 1). In conclusion, Ca19 is likely favored in this tri-culture because Pr24B only released a fraction of the cobalamin it produced.

Genome analysis indicates cobamides are the main shared nutrient in these consortia

Because dependent growth in co-culture with producers did not exceed dependent growth when supplemented with excess cobalamin (Fig. 4), we hypothesized that cobalamin is the main shared nutrient in our co- and tri-cultures. However, the observation that Ca19 grows to a higher density when supplemented with Me12 stationary phase supernatant suggests that nutrients other than cobalamin could be provided by Me12 when the producer is not competing with another isolate (Fig. 3A).

To investigate metabolic capabilities that could impact growth of the isolates in co- and tri-cultures, we sequenced and analyzed the genomes of the four isolates to investigate whether they could produce nutrients that are commonly shared, such as amino acids and vitamins (Fig. 6). We also considered the possibility that certain isolates could degrade carbon sources that others could not, making byproducts available, so we investigated the catabolism pathways for the four carbon sources present in our medium.

Figure 6. Amino acid, vitamin, and carbon catabolism pathways present or absent in the genome of each isolate.

This analysis confirmed that Me12 and Pr24B are predicted to produce cobalamin while Ca24A and Ca19 lack most of the biosynthesis pathway, aligning with the phenotypic observations reported previously (Fig. 6, S2A-B) (21, 24). All isolates are congruent in their amino acid biosynthesis capabilities, except that Ca19 is missing one gene in the leucine biosynthesis pathway, likely rendering it unable to produce leucine (Fig. 6). This amino acid is a component of the culture medium, although it is possible that additional leucine was provided to Ca19 by another isolate. Methionine was the only amino acid missing in our medium as its biosynthesis is often cobamide-dependent (24). Only Pr24B encodes the cobamide-independent methionine synthase, while Ca19, Ca24A, and Me12 are predicted to require cobamides for methionine production (Fig. S2C). The four isolates also have the same pathways for the catabolism of carbon sources in the growth medium; they share the capability to degrade glucose, N-acetylglucosamine, and xylose, but lack xylan catabolism genes (Fig. 6). This result points to carbon as one of the possible resources being competed for in co- and tri-cultures. Vitamin biosynthesis pathway presence and absence are aligned for Ca19, Ca24A, and Me12. These three isolates are predicted to synthesize pantothenate and tetrahydrofolate and to lack complete pathways for the synthesis of thiamine, riboflavin, biotin, and niacin (Fig. 6). Pr24B is predicted to have additional vitamin biosynthesis capabilities, as it contains complete biosynthesis pathways for all the evaluated vitamins except thiamine. While this may result in Pr24B providing additional nutrients or having a competitive advantage in co-culture, all of these vitamins were present in the culture medium and likely did not largely impact the outcome of co- and tri-cultures. Thus, cobalamin is likely to be the main shared nutrient in our co-cultures and tri-cultures.

We further characterized the cobalamin biosynthesis and dependence genes in the four genomes to understand production and dependence of this key nutrient in our isolates (Fig. S2). Me12 and Pr24B are genomically classified as producers according to previously established thresholds (21). They both encode bluB, the gene required to synthesize the lower ligand for cobalamin, 5,6-dimethylbenzimidazole (Fig. 1), in accordance with our previous observation that both synthesize cobalamin (Fig. S2A-B) (24, 31, 32). Pr24B encodes several anaerobic corrin ring biosynthesis enzymes, suggesting it may produce cobamides via the anaerobic pathway. The two dependent genomes lack corrin ring and lower ligand biosynthesis genes, while they contain genes for tetrapyrrole precursor biosynthesis, which are not specific to cobamide biosynthesis (Fig. S2B). However, both dependent genomes contain adenosylation, nucleotide loop assembly, and lower ligand attachment genes, which enable salvaging of late cobamide precursors (21, 33, 34). The latter observation aligns with previous experimental results which showed these two dependent isolates can grow when provided with cobinamide, an incomplete cobamide lacking the lower ligand (24). However, as the two dependent genomes lack the known cobamide lower ligand biosynthesis genes (21), they must also salvage 5,6-dimethylbenzimidazole when they salvage cobamide precursors, or they salvage cobinamide to produce a cobamide other than cobalamin. Taken together, the analysis of the genomes of the dependents and producers in this study aligns with our experimental results demonstrating cobamide-based competition and cooperation.

Discussion

Cooperative interactions such as cross-feeding and competitive interactions such as resource competition between pairs of bacteria can shape microbial community structure (35, 36). Understanding the contribution of particular interactions in detail, however, is challenging because many factors contribute to microbial community interdependence (14). Thus, studying simple consortia with a focus on a single model nutrient can allow for the elucidation of specific interactions in a controlled environment.

Here, we applied the model nutrient approach to test predictions of cobamide-based interactions in co- and tri-cultures of soil bacteria. We successfully used monoculture growth dynamics to predict the outcome of competition for cobamides between two dependent bacteria. This points to cobamides as key nutrients capable of influencing community composition. Further, we confirmed that cobamide producers can support dependents by sharing cobamides in co-culture and can coexist with dependents in tri-culture. These results point to the possibility that cobamide producers broadly influence microbial community dynamics. Finally, whole genome analysis suggests that cobamide-dependent interactions are the main nutrient sharing interaction in our co- and tri-cultures, supporting the large effect we observed from cobamides in culture experiments. Together, our results reveal that cobamide sharing and competition are central determinants of interaction dynamics among soil bacterial isolates and suggest that these interactions may occur in the native environment.

Using pairwise interactions to predict growth in consortia of increasing complexity is limited by the emergence of higher-order interactions (14). The cobamide sharing interactions we observed in co-cultures were conserved in both tri-cultures, suggesting that cobamide sharing is maintained as community complexity increases. Competition for cobamides between dependents was also conserved in tri-cultures. We attempted to measure supernatant cobamide concentrations from the tri-cultures, but these were below the detection limit of our bioassay, likely because the system is in flux and released cobamide is quickly taken up by dependent cells. The outcome of competition was correctly predicted based on total tri-culture cobamide concentrations in the case of Me12. The outcome of competition between dependents in the Pr24B tri-culture was not predictable based on total cobamide concentrations. This unpredictability was likely due to Pr24B providing only a small fraction of the cobamide it produces. The outcome of this tri-culture was aligned with supernatant growth results where Ca19 reached higher cell numbers than Ca24A when grown on exponential phase Pr24B supernatant (Fig. 3B). This result is also consistent with Ca19 thriving at lower cobalamin concentrations (Fig. 2E). It is also possible that other interactions occurred in the tri-culture that benefited Ca19 over Ca24A, such as providing of other nutrients or growth inhibition of Ca24A. It should also be considered that interactions may be the result of changes in gene expression that impact production or release of this enzyme cofactor, thus affecting dependent partners. The conservation of cobamide-based interactions in consortia of higher complexity would enable further predictions and shed light on the importance of cobamide sharing and competition in communities.

Interactions between Ca24A and Pr24B were of particular interest because their joint isolation suggests they could have a pre-established metabolic partnership. They were indeed able to grow together in co-culture when Pr24B provided cobalamin. Additionally, Ca24A was sustained in the tri-culture when Pr24B was the producer, suggesting it was somehow favored more by this producer than by Me12 (Fig. 5D). However, the advantage was not sufficient for Ca24A to outcompete Ca19.

Despite belonging to different species within the Caulobacter genus, the two dependents had the same sets of cobamide biosynthesis and dependence genes (Fig. S2B, C). Only the methionine and tRNA synthesis genes, metH and queG, are relevant in our growth conditions. The epoxyqueuosine reductase encoded by queG has no known effect in laboratory culture (37), thus their cobamide dependence in our culture medium may be solely due to metH. It is curious that although the two dependents completely overlap in the cobamide metabolism genes they encode, they have adapted to different cobamide concentrations. This could be due to experiencing distinct conditions in their native soil environment, such as different physical proximity to cobamide producers. Like preferences for cobamides, adaptations to specific cobamide concentration ranges could occur due to differences in regulation or efficiency of cobamide transporters, riboswitches, adenosylation enzymes or cobamide-dependent enzymes (16, 38). This contrast in preferred concentrations shows that even for genera where cobamide metabolism is conserved, different species or strains may occupy distinct niches. Niche specialization based on cobamide concentration may enable coexistence of closely related bacteria with similar metabolic capabilities in the soil environment (39). The observation that bacteria have preferred cobamide concentrations in addition to preferred structures highlights the key role of cobamides in microbial communities and adds a layer of complexity to predictions of cobamide metabolism, as these preferences cannot be predicted from genome data.

We previously reported that the bulk soil from which these bacteria were isolated contains levels of cobalamin that exceed microbial requirements for the vitamin (25, 40). However, because a large portion of the cobamide is thought to be adhered to the soil matrix, only a fraction may be available to the soil microbiota (25). Additionally, nutrient availability is known to vary greatly in different microenvironments within soil (41). The EC₅₀ values observed for cobamide dependent bacteria isolated from this soil were in the picomolar range (24). EC₅₀ values in this range are common for microorganisms in ocean water, which contains cobamide concentrations in the picomolar range, further suggesting that these bacteria have adapted to low cobamide bioavailability in their native environment (23, 24, 42, 43). We hypothesize that despite the high concentration of cobalamin detected in this soil, cobamides are not largely bioavailable and represent a growth-limiting nutrient for which bacteria must compete. This further highlights the importance of the adaptations in our two Caulobacter sp. isolates to picomolar and low nanomolar concentrations and suggests that the cobamide concentrations at which they thrive are relevant in the soil habitat.

A limitation of our study is that the soil environment from which these bacteria were isolated contains 95% cobalamin, making it one of the least diverse environments in terms of cobamide diversity (40). Hypotheses about competition for cobamides beyond concentration-dependent growth and the niche diversity enabled by different cobamides should be tested in bacteria from environments with a higher cobamide diversity, such as the human gut (40, 44).

Together, we were able to predict the outcome of dependent-dependent and producer-dependent interactions based on growth in monoculture. This represents a significant advance in the applicability of cobamides as model nutrients. Future studies should test the predictability of cobamide interactions between additional pairs of bacteria and evaluate the factors that make predictions possible. Further, elucidation of the mechanisms of cobamide release may enable genome-based predictions of producer-dependent interactions. This research adds to a body of literature that has shown that predictions about nutrient sharing are possible from a combined analysis of metagenomes, genomes, and phenotypic data (18, 36, 45, 46), and that nutrient availability impacts microbial competition and coexistence (47).

Materials and Methods

Isolate characterization and selection

The isolates used in this study were previously published as 5OH_024A (Ca24A), CRE_019 (Ca19), NOC_012 (Me12), and 5OH_024B (Pr24B) (24).

For all experiments, bacteria were cultured at 28°C with aeration in liquid VL60 medium, as described previously (24), or at 28°C on Reasoner's 2 (R2) agar supplemented with 1 nM cobalamin. Dependents were pre-cultured by inoculating from a colony and growing to saturation in liquid medium containing 1 nM cyanocobalamin and transferring to medium lacking cobamide at a 1:100 dilution after 24 hours. The latter cultures were used as the inocula for experiments. Cobamide producer isolates were pre-cultured by growing to saturation in medium lacking cobamide.

Growth measurements were performed in 200 μl cultures in 96-well plates in a Biotek Synergy 2 plate reader with OD₆₀₀ measurements taken every 15 minutes. Cobamide dose-response curves were acquired by culturing in 12 different concentrations of each cobamide with OD₆₀₀ recorded every 15 minutes, as described previously (24). To determine maximum growth rates, the OD₆₀₀ values were normalized to a 1 cm pathlength, blank subtracted in a replicate specific manner by subtracting the adjusted OD₆₀₀ of the inoculum from the mean of the first two timepoints, and smoothed with 0^th order smoothing with 3 neighbors on either side in GraphPad Prism (v10.4.0). Values below 0.009 were eliminated and maximum growth rates were obtained using AMiGA (48). Dose response curves were plotted and fitted in GraphPad Prism.

Co-cultures of dependent isolates

Experiments were performed on three independent cultures of each isolate. Each isolate was diluted to an OD₆₀₀ of 0.0015 for monoculture and co-culture experiments. 1 ml cultures were grown in 96-well plates with 2 ml wells covered with an AeraSeal (Excel Scientific) at 28°C with shaking at 800 rpm in a heated plate shaker (Southwest Science). 10 μl of culture were transferred into 990 μl of fresh medium containing the same cobamide conditions every 24 hours for 7 days. Serial dilutions and plating on R2A medium supplemented with 1 nM cyanocobalamin were done on days 0, 1, 3, 5, and 7 and colonies were counted after 48 hours of growth at 28°C.

Producer supernatant assay

Three independent cultures of each producer were inoculated from glycerol stocks and grown to saturation in VL60 medium lacking cobamide. The cultures were sampled at exponential phase and stationary phase (16 and 36 hours for Me12 and 10 and 36 hours for Pr24B) and a portion was diluted and plated for viable cell counts (Fig. S1). The culture samples were centrifuged, and the supernatant was sterilized by filtration with a 0.22 μm syringe filter. Dependent pre-cultures were prepared as described above and inoculated into medium containing 10% producer supernatant in a 96-well plate. The Me12 exponential phase supernatant was further diluted five-fold based on preliminary experiments that showed it contained saturating amounts of cobamide, this dilution is accounted for in the CFU/ml shown in Figure 3A. Serial dilutions and plating on R2A medium supplemented with 1 nM cyanocobalamin were performed for the inoculum after 24 hours of growth of each culture to determine the viable cell count. Colonies were counted after 48 hours of growth and the number of dependents supported by each producer was calculated by the following: (Dependent CFU/ml gained when supplemented with supernatant – Dependent CFU/ml gained in no addition medium) / Producer CFU/ml_final * dilution factor, where CFU/ml gained was calculated by subtracting the initial CFU/ml from the final CFU/ml.

Producer-dependent co-cultures

Dependent cultures were pre-cultured as described above. Triplicate producer cultures were inoculated from colonies and cultured in VL60 medium lacking cobamide for 24 or 48 hours for Pr24B and Me12, respectively. Producer cultures were washed by centrifugation and resuspension of cell pellets fresh VL60 medium to remove spent medium containing cobamides. Both producer and dependent cultures were diluted to an OD₆₀₀ of 0.1 in fresh VL60 medium and mixed 1:1 in triplicate. These mixtures were then inoculated 1:100 into 1 ml fresh VL60 medium in a 96-well plate with 2 ml wells. Cultures were serially diluted and plated to determine the viable cell count in the inocula and after 48 hours of growth at 28°C. The number of dependent cells supported by each producer was determined by the following calculation: (Dependent CFU/ml gained in co-culture – Dependent CFU/ml gained in no addition medium) / Producer CFU/ml_final, where CFU/ml gained was calculated by subtracting the initial CFU/ml from the final CFU/ml.

Tri-culture growth

Me12 and Pr24B cultures were inoculated from colonies and grown for 48 and 24 hours, respectively, in liquid VL60 medium lacking cobamide. Dependents were pre-cultured as described above. Cultures were adjusted to an OD₆₀₀ of 0.1 and mixed in equal volumes to prepare the inoculum. 10 μl of inoculum was added to 990 μl of VL60 medium with no cobamide in a 96-well plate with 2 ml wells. Cultures were transferred and plated as described for the co-cultures of cobamide dependent isolates.

150 μl of culture were collected on days 1, 3, 5, and 7 and boiled to lyse the cells. This lysate was used to determine the total cobamide concentration via an E. coli bioassay as described previously (24, 30).

Genome sequencing and analysis

Genomic DNA was extracted from saturated cultures in VL60 medium using the Qiagen DNEasy Blood and Tissue kit adapted for Gram positive bacteria. Illumina libraries were prepared using 500 ng of input isolate DNA with the Illumina DNA prep kit with IDT^® for Illumina^® DNA/RNA index plates. Libraries were sequenced (2x150 paired end) on an Illumina NovaSeq X Plus instrument at Novogene (Novogene Corporation Inc., Sacramento, CA). Genomes were subsequently trimmed, assembled, and annotated using KBase as described previously (49, 50), and filtered for >95% completeness and <5% contamination. Amino acid sequences predicted by DRAM were used for downstream analyses (51).

The genomes were functionally annotated using BlastKOALA and cross-referenced with a list of HMMs for cobamide biosynthesis, cobamide-dependent functions, and independent alternatives to cobamide dependent pathways (21, 25, 52). Transport genes and cobalamin-binding domains were left out of the analysis due to poor annotations, and rSAM-B₁₂ proteins were omitted because this category included the Radical SAM superfamily, which includes but is not restricted to cobalamin-dependent rSAM enzymes. Additionally, genes that are not essential for cobamide biosynthesis or that are only found in archaea were not included.

Amino acid biosynthesis and small carbon source catabolism capability were assessed using GapMind (53). Pathway genes were considered to be present if they had at least a medium confidence hit as described previously (54). Synthesis capacity of B vitamins other than cobalamin was assessed by evaluating the biosynthesis pathways with previously established thresholds (45, 46). Genes of interest identified by previous work were cross-referenced against the DRAM annotations.

Importance.

Interactions between microbes shape their communities and the world around them. Because microbial communities are physically and chemically complex, predicting microbial interactions is often difficult. Using a set of four bacteria isolated from the same soil environment, we generated and tested predictions about the interactions involving a family of model nutrients related to vitamin B₁₂. Understanding and predicting interactions can lead to greater insights into microbiome function.

Data Availability

The sequencing data generated during the current study are available in the NCBI GenBank repository under Biosample Accession IDs SAMN51232832 - SAMN51232835.