Introduction

Recent improvements in sequencing technologies have made it possible to sequence entire communities and to use genome-enabled approaches for the study of microbial interactions that lead to community-level properties (Song et al., 2015). Because B vitamin-related enzyme cofactors are essential but expensive to produce, their precursors are commonly shared by community members. It is recognized that many microorganisms are unable to synthesize all required cofactors de novo and instead salvage select precursors to make them. The genetic basis for this salvage has been defined in many organisms, revealing a great diversity of transporters (Jaehme and Slotboom, 2015a), alternative enzymes and pathways for converting precursors to cofactors, and the tight regulation of precursor salvage and cofactor biosynthesis (Winkler et al., 2002a,b; Nahvi et al., 2004; Ames et al., 2010; Leyn et al., 2016; Suvorova and Rodionov, 2016). New genes involved in cofactor metabolism continue to be discovered, largely due to the increasing availability of whole genome sequences and the ability of comparative genomics and regulon analysis to discover new protein families that participate in this process (Sun et al., 2013; Rodionova et al., 2015).

At present, studies of cofactor precursor exchange in natural communities have been limited to surveys of biosynthetic and salvage potential in metagenomes (Magnusdottir et al., 2015) where the sources of precursors are not defined, or in assays of binary cultures where the number of precursors exchanged are minimal (Croft et al., 2005; Kazamia et al., 2012; Grant et al., 2014). Here we focused instead on two nearly identical consortia derived from benthic microbial mats that occur in a heliothermal saline (epsomite, MgSO4) lake in northern Washington State (Lindemann et al., 2013; Zachara et al., 2016). These unicyanobacterial consortia (UCC) consist of a single, but distinct, cyanobacterium (Phormidesmis priestleyi ANA in UCC-A versus Phormidium sp. OSCR in UCC-O) and its associated heterotrophic cohorts (Cole et al., 2014) (Supplementary Table S1). Although the abundances of the heterotrophs relative to the cyanobacterium differ between the consortia, they each harbor the same 16 heterotrophs. A single additional heterotroph, Porphyrobacter sp. HL-46, was detected in only UCC-A (Nelson et al., 2016). Complete or near-complete genome sequences are available for most members enabling prediction of producers of cofactors and/or consumers of cofactor precursors. Importantly, no vitamin supplements were added to the growth medium during enrichment or subsequent cultivation. Hence, these consortia are dependent entirely on endogenous production of cofactors.

Here we used the subsystems approach and comparative genomics analysis to reconstruct vitamin cofactor biosynthetic pathways and predict transport capabilities in the 19 organisms that comprise these consortia. This approach, supported by supplementary growth studies on heterotrophic isolates, revealed an extensive cofactor auxotrophy among members of these consortia and differentiation in the type of precursors salvaged and the complement of cofactor-dependent pathways present. In addition, we identified potential transporters and regulators involved in precursor salvage. Insights gained from this study support the concept of syntrophic partnership based on vitamin exchange and suggest mechanisms by which this is achieved.

Materials and methods

Growth studies on UCC heterotrophic isolates

Hot Lake strains were routinely cultured at 30°C on Hot Lake Heterotroph broth, pH 8.0 (Cole et al., 2014). For testing vitamin-dependent growth, the cells were diluted to an OD600 of ~0.2 and inoculated into HL base medium (Hot Lake Heterotroph without yeast extract) supplemented with 5 mm carbon source and 1 × Wolfe’s vitamin mix (Wolin et al., 1963) or selected vitamins (0.74 nm cyanocobalamin, 0.015 μm thiamine and/or 0.082 μm biotin). When cultures reached mid-log phase, the cells were washed three times with HL base medium to remove residual metabolites and growth supplements. Ten replicate samples were evaluated every 1–2 h for all assays. For Erythrobacter HL-111, which grows slowly, the OD600 of replicated 20 ml cultures was measured using a SmartSpec Plus Spectrophotometer (Bio-Rad, Hercules, CA, USA). For the remaining strains, a Norden Lab Professional-Bioscreen, C Edition was used to analyze growth (OD600) of 250 μl samples in the 100-well Bioscreen plate.

Consortial composition

A combination of the species-resolved metagenome bins or isolate genome sequences (Nelson et al., 2016) from two unicyanobacterial consortia (UCC-A and UCC-O whose cyanobacterial members are designated Phormidium priestleyi ANA and Phormidesmis OSCR, respectively) were evaluated for the presence or absence of pathways for de novo biosynthesis of B vitamin-related enzyme cofactors. In addition to the cyanobacterial members, the closed genome sequences derived from 10 axenic cultures (Algoriphagus marincola str. HL-49, Aliidiomarina sp. HL-53, Roseibaca calidilacus HL-91, Halomonas sp. HL-48 and HL-93, Marinobacter excellens str. HL-55, Marinobacter sp. HL-58, Erythrobacter sp. HL-111, Salinivirga fredricksonii HL-109 and Porphyrobacter sp. HL-46) and binned assemblies from nine additional members (Bacteroidetes bin01, Oceanicaulis bin04 and Rhodobacteraceae bins 7, 8, 9, 12 and 18) were examined. Based on occurrence of conserved single-copy genes, it was estimated that the binned assemblies have a genomic coverage of at least 98% with the exception of one member, Rhodobacteraceae bin09, whose estimated coverage is 87.6% (Nelson et al., 2016).

Genomic reconstruction of metabolic pathways and regulons

Genome-based reconstruction of eight B vitamin-related cofactor biosynthesis pathways in 19 UCC genomes was performed using the subsystem-based comparative genomic approach (Osterman et al., 2010) implemented in SEED/RAST (Aziz et al., 2008; Overbeek et al., 2014; Brettin et al., 2015) combined with genomic reconstruction of vitamin-specific transcriptional regulons and identification of candidate vitamin transporters as previously described (Rodionov et al., 2009; Rodionova et al., 2015). Extensive manual pathway curation was assisted by KEGG orthology assignments from the BlastKoala annotation tool (Kanehisa et al., 2016). Identification of orthologs in closely related genomes and their genome neighborhood analysis were performed in the IMG workspace (Markowitz et al., 2014). Cofactor requirements were asserted by cataloging respective enzymes from annotated genomes and connecting them to metabolic pathways in KEGG (Kanehisa et al., 2016) and SEED (Overbeek et al., 2014). Computational searches of B12, FMN and thiamine pyrophosphate (TPP) riboswitches were performed using covariance models from the Rfam database as recently reviewed in Sun and Rodionov (2014). The BirA, NrtR, NadQ regulons were reconstructed by the consistency check approach using the positional weight matrices of DNA motifs from the RegPrecise database as previously described (Leyn et al., 2016). A full list of reconstructed metabolic pathways and regulons is available in the Supplementary Data 1.

Genome sequence accession numbers

All sequence data were generated by the Joint Genome Institute. Assemblies derived from metagenome bins and from isolate genome sequence can be accessed via IMGer (https://img.jgi.doe.gov/cgi-bin/mer/main.cgi). In addition, binned metagenome assemblies have been deposited in the European Nucleotide Archive (ENA; www.ebi.ac.uk/ena) and Genbank under accession numbers LIHN00000000 (Bacteroidetes Bin01), LJSH00000000 (Oceanicaulis Bin04), LJSU00000000 (Rhodobacterales Bin07), LJSF00000000 (Rhodobacterales Bin08), LJNT00000000 (Rhodobacterales Bin09), LJZR00000000 (P. priestleyi ANA), LJSV00000000 (Rhodobacterales Bin12), LJZT00000000 (Phormidum OSCR) and LJSY00000000 (Rhodobacterales Bin18). Isolate genomes have been deposited under accession numbers GCA_001458075.1 (Aliidiomarina sp. HL-53), JMLY00000000 (Marinobacter sp. HL-58), GCA_001517585.1 (R. calidilacus sp. HL-91), NZ_JAFX00000000 (A. marinocola sp. HL-49), JMMC00000000 (Halomonas sp. HL-48), JYNR00000000 (M. excellens sp. HL-55), GCA_000744895.1 (Porphyrobacter sp. HL-46), GCA_001314745.1 (Halomonas sp. HL-93), GCA_001314785.1 (Salinivirga fredricksonii HL-109, formerly Rhizobiales sp. HL-109) and GCA_001314765.1 (Erythrobacter sp. HL-111).

Results

Composition of UCC-A and UCC-O

The unicyanobacterial consortia examined in this study are low complexity enrichments resulting from attempts to develop axenic cultures of cyanobacteria from mats found in Hot Lake Washington. The firm attachment of heterotrophs to the cyanobacteria have thus far made it impossible to isolate the autotroph, but resulted in communities that are expected to be enriched in interacting microbes because the heterotrophs cannot fix CO2 that was provided as the sole carbon source during enrichment. Metagenomic sequence was derived from two unicyanobacterial consortia (UCC-A and UCC-O), assembled and binned into 19-member sequences (Nelson et al., 2016). Because the two consortia were nearly identical in taxonomic composition (16 heterotrophs were common to both consortia), co-assembly of sequencing reads lead to near-complete assembly of nearly all members of the consortia (the abundance of a 20th Alphaproteobacterial member was too low to produce sufficient genomic coverage to be included in this study). Further improvement of member assemblies resulted from sequencing the genomes of axenic cultures of heterotrophs derived from the consortia. In addition to the six genomic sequences from heterotrophic members of the consortia (Porphyrobacter sp. HL-46, Halomonas sp. HL-48, Algoriphagus marincola HL-49, Aliidiomarinacea HL-53, Marinobacter excellens HL-55 and Marinobacter sp. HL-58) described previously, this study includes genomic sequence from four new UCC isolates, Roseibaca calidilacus HL-91, Halomonas sp. HL-93, Erythrobacter sp. HL-111 and Salinivirga fredricksonii HL-109, bringing the total number of members with closed genomes to 10. In summary, each community is composed of a single cyanobacterium belonging to the Oscillatoriales order, two Bacteroidetes, five Gammaproteobacteria and ~10 Alphaproteobacteria (see Supplementary Table S1).

Exchange of cofactor precursors is necessary to support essential metabolic processes

We used a subsystems-based genomic reconstruction approach to infer biosynthesis, recycling and salvage pathways for eight B vitamin-related enzyme cofactors in the 19 members of UCC-A and UCC-O. Our analysis revealed that all members are able to synthesize PLP (pyridoxal phosphate; Supplementary Figure S1) and flavin nucleotides (Supplementary Figure S2), but only the cyanobacteria and both Halomonas sp. can produce the remaining six cofactors (summarized in Table 1). Four of the Rhodobacteraceae (bins 7, 9, 12 and 18) belonging to the Roseobacter group are auxotrophic for six cofactors, and the fifth (bin08) is only able to synthesize CoA de novo, making this group the most dependent on cofactor exchange within the consortia. By contrast, the Gammaproteobacteria require precursors to produce no more than two cofactors, cobalamin and/or TPP, each.

Table 1 Overview of predicted cofactor biosynthesis capability
Full size table

The consortia consist of 10 TPP, 11 biotin and 13 cobalamin auxotrophs, making their precursors the most widely sought after. Analysis of cofactor-dependent processes confirmed that TPP is essential for central carbon metabolism of all species in the two consortia (Supplementary Data 1 and Figure 1). Critical TPP-dependent enzymes found in all members include transketolase, a pyruvate oxidase, a TCA enzyme that degrades 2-oxogluatrate and 1-deoxy-d-xylulose-5-phosphate synthase, which produces a key metabolite needed for thiamine, pyridoxal phosphate and terpenoid biosynthesis. A single biotin-dependent enzyme, acetyl-CoA carboxylase, which is essential for the initiation of fatty acid biosynthesis, is present in all isolates.

Figure 1
figure 1

Overview of TPP, biotin and B12-dependent processes. TPP-dependent enzymes: AceE, pyruvate dehydrogenase E1 component; AcoA, acetoin dehydrogenase complex E1 component alpha subunit; AlsS, acetolactate synthase; AruI, 2-ketoarginine decarboxylase; BkdA, 2-oxoisovalerate dehydrogenase E1 component alpha subuit; ComD, sulfopyruvate decarboxylase alpha subunit; Cyano, 2-oxoglutarate dehydrogenase; Dxs, 1-deoxy-d-xylulose-5-phosphate synthase; IlvB, biosynthetic acetolactate synthase I catalytic subunit; IlvG, biosynthetic acetolactate synthase II catalytic subunit; IolD, decyclizing 3D-(3,5/4)-trihydroxycyclohexane-1,2-dione acylhydrolase; IorA, ferredoxin oxidoreductase alpha subunit; Kgd, 2-oxoglutarate decarboxylase; KorB, 2-oxoglutarate ferredoxin oxidoreductase beta subunit; LpdC, indolepyruvate decarboxylase; MenD, 2-succinyl-5-enolpyruvyl-6-hydroxy-3-cyclohexene-1-carboxylic-acid synthase; SucA, 2-oxoglutarate dehydrogenase E1 component; Tkt, transketolase; Xfp, xylulose-5-phosphate/fructose-6-phosphate phosphoketolase; and Xsc, sulfoacetaldehyde acetyltransferase. Biotin-dependent enzymes: AccB, acetyl-CoA carboxylase biotin carboxyl carrier protein; AtuF, geranyl-CoA carboxylase alpha subunit; LiuD, 3-methylcrotonyl-CoA carboxylase alpha subunit; OadA, oxaloacetate decarboxylase alpha subunit; PccA, propionyl-CoA carboxylase alpha subunit; Pyc, pyruvate carboxylase; and Uca, urea carboxylase. B12-dependent enzymes: BchE, magnesium-protoporphyrin IX monomethyl ester cyclase; Ecm, (2R)-ethylmalonyl-CoA mutase; EutC, adenosylcobalamin-dependent ethanolamine ammonia-lyase small subunit; HpnP, hopanoid 2-methyltransferase; McmA, methylmalonyl-CoA mutase B12-binding subunit; MetH, methionine synthase; Mut, methylmalonyl-CoA mutase; and NrdJ, ribonucleoside diphosphate reductase. Enzymes and transporters that obviate the need for cofactors: BhmT, betaine–homocysteine methyltransferase; MetE, methionine synthase; MetNIQ, ABC-type methionine transporter; MetT, methionine:cation antiporter; NrdEF, ribonucleoside diphosphate reductase; UreABC, ABC-type urea transporter. Additional enzymes without precise functions predictions are listed in Supplementary Data 1.

Full size image

Cobalamin is required for essential processes such as methionine biosynthesis. However, it is not universally required due to mechanisms that bypass the need for it (dashed lines, Figure 1). Examples include the use of alternative enzymes (e.g., MetE to synthesize methionine) and pathways (e.g., methyl citrate pathway to degrade propionyl-CoA) that do not require cofactors or transporters that import metabolites that would otherwise require a cofactor for synthesis (e.g., MetT or MetNIQ for methionine). The Gammaproteobacterial members use the most cobalamin-independent pathways, which is surprising as the two Halomonas spp. are among five of the only members that can synthesize this metabolically expensive cofactor. Unlike S. fredricksonii HL-109, the Halomonas spp. do not control cobalamin synthesis with a riboswitch (Supplementary Data 2 and Supplementary Note 1), and therefore have the potential to make excess cofactor and be important providers of B12 for the community.

Exchange of canonical B vitamins can complement most, but not all, deficiencies in de novo biosynthesis of respective cofactors

Although, by definition, B vitamins include a variety of different cofactor precursors (Supplementary Table S2), the chemicals included in the conventional medium supplement, Wolfe’s vitamin solution (Wolin et al., 1963) are the standard growth supplements used to cultivate microbes and more widely assumed to be exchanged in microbial communities. Here we found, however, that pantothenate, nicotinic acid, folic acid, para-aminobenzoic acid (pABA), and cyanocobalamin are the only standard vitamin additives that can complement UCC member deficiencies in cofactor biosynthesis.

Pantothenate, a precursor to coenzyme A (CoA), is synthesized de novo through the actions of (i) PanB and PanE (or IlvC), which convert 3-methyl-2-oxobutanoate to pantoate and (ii) PanC, which ligates β-alanine to pantoate to form pantothenate (Figure 2). All UCC members can convert pantothenate to CoA, but four Rhodobacteraceae cannot synthesize pantothenate and thus are expected to be auxotrophs even though no pantothenate transporter could be identified. An aspartate carboxylase, PanD or PanP, that is needed for the synthesis of β-alanine from aspartate is present in all members except Phormidesmis ANA, R. calidilacus HL-91, Erythrobacter HL-111, Porphyobacter HL-46 and S. fredricksonii HL-109. As HL-91 and HL-111 can grow in defined medium lacking CoA precursors (Supplementary Figure S7), we hypothesize that an alternative aspartate carboxylase or novel pathway to synthesize β-alanine exists.

Figure 2
figure 2

Coenzyme A biosynthesis and salvage of precursors. Metabolites that occur in the pathway for coenzyme A biosynthesis appear in open gray rectangles: 1, D-4′-phosphopantothenate; 2, (R)-4-phosophopantothenyoyl-L-cysteine; 3, pantetheine 4-phosphate; 4, dephospho-CoA; 5, N-carbomoyl-β-alanine; and CoA, coenzyme A. Enzymes appear in open black rectangles: CoaA, type I pantothenate kinase; CoaB, phosphopantothenate-cysteine ligase; CoaC, phosphopantothenoylcysteine decarboxylase; CoaD, pantetheine-phosphate adenylyltransferase; CoaE, dephospho-CoA kinase; CoaX, type III pantothenate kinase; IlvC, ketol-acid reductoisomerase; PanB, 3-methyl-2-oxobutanoate hydroxymethyltransferase; PanC, pantoate-β-alanine ligase; PanD, pyruvoyl-dependent aspartate 1-decarboxylase; PanE, 2-dehydropantoate 2-reductase; PanG, ketopantoate reductase; PanP, PLP-dependent aspartate 1-decarboxylase; PreA, NAD-dependent dihydropyrimidine dehydrogenase subunit; PreT, NAD-dependent dihydropyrimidine dehydrogenase subunit; PydB, dihydropyriminidase; PydC, beta-ureidopropionase. PanF refers to the Na+/pantothenate symporter and the boxed question mark to unknown transporters. Non-orthologous replacements are indicated by filled gray circles. Dashed arrows indicate precursor salvage routes for coenzyme A precursor uptake.

Full size image

Six heterotrophs (S. fredricksonii HL-109, Rhodobacteraceae bins 7, 8, 9, 12 and 18) are unable to synthesize NAD(P) from aspartate or tryptophan (Supplementary Figure S3) and the same six cannot produce tetrahydrofolate (Supplementary Figure S4). The presence of the nicotinate phosphoribosyltransferase (PncB and PncB2) in all auxotrophs and six prototrophs suggest that nicotinic acid is salvaged, even though no candidate nicotinic acid transporters were identified. S. fredricksonii HL-109 is missing the entire pathway while five Rhodobacteraceae (bins 7, 8, 9, 12 and 18) lack only the ability to synthesize the precursor pABA. These five members encode neither the enzymes with pABA synthetase or 4-amino-4-deoxychorismate lyase activity, which are needed to convert chorismate to pABA nor the recently described alternative pABA-biosynthetic enzyme (Adams et al., 2014; Satoh et al., 2014). No candidate folate or pABA transporters were detected in respective auxotrophs, but four prototrophs (Phormidium OSCR, Oceanicaulis Bin04, Aliidiomarina HL-53 and Bacteroidetes Bin01) encode AbgT family transporters, which are proposed to export pABA (Delmar and Yu, 2016), suggesting that they share this resource.

A diverse set of precursors can support cobalamin auxotrophs

Although cyanocobalamin is expected to support the growth of all the UCC cobalamin auxotrophs, the presence of partial biosynthetic pathways in some members suggests that additional precursors might also be salvaged. The biosynthesis of B12 family cofactors is initiated by synthesis of the corrin ring from uroporphyrinogen III via either an oxygen-dependent (Warren et al., 2002; Heldt et al., 2005) or oxygen-sensitive pathway (Warren et al., 2002; Roessner and Scott, 2006; Supplementary Figure S5). Only five members (both cyanobacteria, both Halomonas spp. and S. fredricksonii HL-109) possess either of these pathways and hence synthesize cobalamin de novo. The second phase of B12 family cofactor biosynthesis involves attachment of the upper α-axial component, which is derived from aminopropanol and either 5′-deoxyadensine or a methyl group, and a lower β-axial component typically composed of a cobalt-coordinated nucleotide with DMB (5,6-dimethylbenzimidazole) as the base (Gray and Escalante-Semerena, 2007; Supplementary Figure S6). The attachment of different combinations of ligands in the upper and lower positions results in the formation of distinct cofactors (e.g., pseudo-B12, methylcobalamin, adenosylcobalamin and hydroxocobalamine; Giedyk et al., 2015).

Both cyanobacteria genomes encode corrin ring biosynthetic genes, but lack cobT that is necessary for attachment of DMB as the lower ligand and, therefore, likely produce alternative coenzymes (Supplementary Figure S6) such as adenylylcobamide (pseudo-B12), which has been reported as the predominant form of B12 coenzyme in other cyanobacteria (Watanabe, 2007; Tanioka et al., 2009). Six auxotrophs (Bacteroidetes bin01, A. marincola HL-49, M. excellens HL-55, Porphyrobacter HL-46, Oceanicaulis bin04 and Rhodobacteraceae bin08) lack all biosynthetic genes and thus require an external source of cobalamin that cannot originate in the cyanobacteria. With the exception of Bacterioidetes bin01, the remaining auxotrophs encode partial pathways for biosynthesis of adenosylcobalamin, the TonB-dependent receptor (BtuB), and the ABC-type inner membrane translocase (BtuCDF), which are required for the uptake of cobalamin and related precursor corrinoids (Woodson et al., 2005; Degnan et al., 2014). The missing ABC transporter genes in Bacteroidetes bin01 are presumably in a sequencing gap. Interestingly, all three of the heterotrophic cobalamin producers lack the high affinity BtuB outer membrane receptor. Based on the gene complement present in these members, we predict that they salvage one or more of the precursors hydrogenobyrinic acid a,c-diamide, cobyrinic acid, cobyrate acid and cobinamide-guanosine diphosphate and consequently, some community members may benefit from more precursors than others (Figure 3).

Figure 3
figure 3

Model of potential cobalamin precursor exchange among UCC community members. The two types of B12 producers (adenylylcobamide versus adenosylcobinamide) are shown at the top and the auxotrophs are shown at the bottom. Intermediates that are predicted to be salvaged are shown in the middle; HBA, hydrogenobyrinic acid; Cbr, cobyrinate a,c-diamide; Cby, cobyrate; Cbi-GDP, cobinamide-guanosine diphosphate. Arrows between intermediates depict the order (earliest to latest from left to right) in which they enter the cobalamin biosynthetic pathway. Pseudo-B12 (PsB12) has an adenine rather than a 5,6-dimethylbenzimidazole (DMB) at the alpha-axial ligand position. Note that auxotrophs at the far bottom right are unable to make use of precursors produced by the cyanobacteria.

Full size image

S. fredricksonii HL-109 and Rhodobacteraceae bin09 are TPP auxotrophs that cannot salvage thiamine

The de novo biosynthesis of TPP involves the formation of phosphorylated hydroxyethylthiazole (HET) and hydroxymethylpyrimidine (HMP) intermediates, which are first joined by ThiE (thiamine-phosphate pyrophosphorylase) and then phosphorylated by ThiL (thiamine monophosphate kinase; Du et al., 2011; Figure 4). Only six members (each of the cyanobacteria, Halomonas spp. and Marinobacter spp.) have all the genes necessary for TPP biosynthesis. Erythrobacter HL-111, Porphyrobacter HL-46 and Oceanicaulis bin04 lack known enzymes for synthesis of the iminoglycine precursor. However, HL-111 can grow in defined medium lacking thiamine (Supplementary Figure S7), which suggests that an alternative iminoglycine biosynthesis pathway exists and that all three organisms are also prototrophs.

Figure 4
figure 4

Thiamine, hydroxyethylthiazole and 4-amino-t-aminomethyl-2-methylpyrimidine are salvaged for thiamine pyrophosphate biosynthesis. Metabolites that occur in the thiamine pyrophosphate biosynthetic and precursor salvage pathways: AIR, 5-amino-1-(5-phospho-D-ribosyl)imidazole; amino-HMP, 4-amino-5-aminomethyl-2-methylpyrimidine; DXP, 1-deoxy-d-xylulose-5-phosphate; G3P, glyceraldehyde-3-phosphate; HET, hydroxyethylthiazole; HMP, 4-amino-5-hydroxymethyl-2-methylpyrimidine; IG, iminoglycine; -P, phosphate; -PP, diphosphate; TMP, thiamine monophosphate; TPP, thiamine pyrophosphate. Enzymes involved in biosynthesis: CdsH, cysteine desulfhydrase; Dxs, 1-deoxy-d-xylulose-5-phosphate synthase; IscS, cysteine desulfurase; MoeB, adenylyltransferase/sulfurtransferase; ThiC, phosphomethylpyrimidine synthase; ThiD, bifunctional hydroxymethylpyrimidine/phosphomethylpyrimidine kinase; ThiE, thiamine-phosphate pyrophosphorylase; ThiF, sulfur carrier protein adenylyltransferase; ThiG, thiazole synthase; ThiL, thiamine monophosphate kinase; ThiO, glycine oxidase; and ThiS, sulfur carrier protein. Enzymes involved in salvage and recycling: TenA, thaiminase; ThiK, thiamine kinase; ThiM, 4-methyl-5-(beta-hydroxyethyl) thiazole kinase; ThiN and ThiN2, thiamine pyrophosphokinase. Thiamine transporters include the Omr1 TonB-dependent receptor, the PnuT ATP-dependent transporter, the ThiBPQ ABC-type transporter, The YkoCDE ECF-type transporter and the ThiV2 sodium-dependent transporter. Hydroxymethylpyrimidine transporters are the ThiV sodium-dependent transporter and the ThiXYZ ABC-type transporter. Trap refers to the single unnamed Trap-type hydroyethylthiazole transporter.

Full size image

Two main salvage strategies are used by our consortia to produce TPP (Figure 5). Nine members of the consortia use the most commonly recognized strategy, which involves uptake of thiamine followed by its phosphorylation to TPP. Five different thiamine transporters were detected including ThiV2, which was discovered through our analysis (Supplementary Note 2). Five auxotrophs encode a partial TPP biosynthetic pathway (ThiD, ThiM, ThiE and ThiL), which can be used to convert salvaged HET and HMP into TPP. Two transporters were detected in auxotroph genomes; ThiXYZ that had been previously associated with HMP transport and a TRAP-type transporter that we predict to be used for HET uptake (Supplementary Note 2). The ThiV and/or TRAP transporters were also found in several TPP prototrophs, suggesting that they can reduce the cost of biosynthesis by salvaging HET and/or HMP. Salvage of both precursors is the only option for generating TPP by S. fredricksonii HL-109 and Rhodobacteraceae bin09 and thus Wolfe’s vitamin solution should not support growth of these organisms.

Figure 5
figure 5

Overview of proposed interactions between thiamine producers and auxotrophs. TPP precursor transporters are depicted by rounded rectangles with gray edges. Precursors that are exchanged include thiamine, hydroxyethylthiazole (HET) and 4-amino-5-hydroxymethyl-2-methylpyrimidine (HMP). An exporter for HET has not been identified. Note that M. excellens HL-55 and both Halomonas spp. also have uptake transporters, but they are not shown here for clarity.

Full size image

Non-canonical vitamins can be salvaged for biotin, NAD(P) and PLP biosynthesis

Over half of the community members are biotin auxotrophs (Table 1 and Supplementary Figure S8), but all are predicted to salvage biotin or a precursor. Biotin is salvaged by either the the YigM transporter or the BioY ECF-type transporter, which can function even in the absence of the auxiliary components BioMN (Finkenwirth et al., 2013). Three auxotrophs lack YigM and BioY but encode the biotin synthase (BioB), which catalyzes the last step in biosynthesis. Thus, we predict that dethiobiotin is a precursor that is available for exchange in communities and that transporters other than BioY or YigM mediate its uptake.

All six NAD(P) auxotrophs (S. fredricksonii HL-109, R. calidilacus HL-91 and Rhodobacteraceae bins 7, 8, 9, 12 and 18) and five of the prototrophs (both cynobacteria, A. marincola HL-49 and both Halomonas spp.) encode nicotinamidase (PncA) and, therefore, are expected to salvage nicotinamide in addition to nicotinic acid. A. marincola HL-49 encodes the PnuC transporter and NadR kinase, which is required for the salvage of nicotinamide riboside (Supplementary Figure S3). It is recognized that pyridoxine kinase (PdxK) can convert pyridoxine, pyridoxal or pyridoxamine into PLP and, therefore, the three prototrophs that have PdxK can potentially salvage any of these three precursors (Supplementary Figure S1).

Biased distribution of one-component transporters among prototrophs and auxotrophs

We observed a consistent trend in the presence of one-component transporters in cofactor prototroph genomes and not in auxotrophs, which suggests that prototrophs might conditionally export substrates (Figure 6, Supplementary Data 4). YigM, which was found in all six of the heterotrophic biotin prototrophs, has been demonstrated to take up biotin (Ringlstetter, 2010; Finkenwirth et al., 2013) yet is classified as a DME family transporter (TC: 2.A.7.3; Saier et al., 2014). Most members of this transporter family are implicated in export functions, supporting the notion that this transporter operates in both directions.

Figure 6
figure 6

Reconstruction of syntrophic metabolism of B vitamin-related cofactors. Cofactor prototrophy and auxotrophy is represented by white and black ovals, respectively. Diamonds represent proposed bidirectional transporters (T=PnuT in HL-53 and ThiV in the rest, B: YigM, F: Fbt, P: PanF, N: PnuC) of cofactors or their precursors. Text within ovals and diamonds refers to precursor groups (vitamins) that contribute to the community metabolite pool: thiamine: Thi or T; niacin: Nia or N; pantothenate: Pan or P; biotin: Bio or B; folate: Fol or F; cobalamin: Cob. No cobalamin information is shown for Aliidiomarina HL-53 because it encodes no cobalamin-dependent enzymes. The only required precursor with no candidate exporter is cobalamin, but we speculate that one exists.

Full size image

Four transporters are classified as Na+ or H+:solute symporters and thus translocation directionality is likely controlled by pH and ionic strength. Three transporters (ThiV, ThiV2 and PanF) belong to the Na+:solute symporter family (TC: 2.A.21) and the fourth, Fbt, belongs to the folate-biopterin transporter family (TC: 2.A.71). Among these four, ThiV2 is the only one not found exclusively in prototrophs. Two additional one-component transporters, PnuC and PnuT, catalyze transport by facilitated diffusion. Their substrates are phosphorylated by a kinase after uptake, thereby trapping them inside the cell (Jaehme and Slotboom, 2015b). Transport directionality control would, therefore, be impacted by the energy status of the host. PnuC is found in the NAD prototroph, A. marincola HL-49, and PnuT is found in a single auxotroph, Aliidiomarina sp. HL-53. Taking into account other microorganisms that encode PnuC and PnuT, it is evident that their distribution is uneven, but most occur in prototrophs.

Discussion

Our results demonstrate, for the first time, how a self-sustaining defined microbial community can retain members that have numerous essential requirements for B vitamin-related enzyme cofactor precursors (Figure 6). The analysis of member biosynthetic capabilities revealed that many of the auxotrophs encode partial biosynthetic pathways that initiate with metabolites that are precursors of canonical vitamins rather than the vitamins themselves. Diversification and specialization of precursor salvage provides a mechanism for division of labor in microbial communities and a selective advantage for partners uniquely suited for optimal syntrophic biosynthesis of cofactors (see Fredrickson, 2015). These findings also have important implications for the cultivation of microbes from environmental samples, where the naive assumption is that the metabolites in Wolfe’s supplement are sufficient to support cofactor biosynthesis deficiency or to enable a desired community member to outcompete its neighbors. Although alternate salvage systems have been described previously, (Taga and Walker, 2008; Galeazzi et al., 2011), this knowledge has yet to be broadly applied to microbial physiology and cultivation. When cultivating uncharacterized microbes, we advise supplementing the standard vitamin supplements with additional precursors such as those described herein.

Precursor specialization may be driven by environmental conditions affecting precursor availability. Our consortia are grown under constant light conditions, and two cofactors, TPP and adenosylcobalamin, are susceptible to photolysis. Photolysis of thiamine yields HMP and HET (Ansari et al., 2004). Six of the community members have salvage systems for HMP and HET, with four members restricted to salvage of these precursors rather than thiamine itself (Figure 4 and Supplementary Data 3). Exclusive use of these metabolites likely reflects their ready availability. Photolysis of adenosylcobalamin causes loss of its adenosyl group (Garabato et al., 2016). Although it is predicted that all community members can salvage cob(II)alamin, it is likely that they do not all perform this task with the same efficiency. We predict that, in situ, concentrations of TPP and cobalamin precursors vary over the diel cycle and across mat depth, and that this variation may drive shifts in community function or composition.

Two mechanisms proposed for the release of vitamins to consumers are passive exchange, in which death and lysis of the producers benefits the consumers (Karl, 2002), and direct symbiosis, where producers synthesize excess vitamin and export it for use by consumers. Although passive exchange is assumed to occur in all systems, mathematical modeling of the growth dynamics of an obligate syntrophic partnership between a bacterial B12 producer and an algal consumer has provided strong evidence for direct symbiosis (Grant et al., 2014). Our data suggest that direct symbiosis occurs in this community. The predominant member in each of our consortia is the cyanobacterium (P. priestleyi ANA or Phormidium OSCAR), which is prototrophic for all vitamins (Cole et al., 2014). However, the cyanobacteria are unable to produce adenosylcobalamin, which is required by six members of our community (Figure 3). Consequently, these six members must acquire cobalamin from the low-abundance producers, Halomonas HL-48 and HL-93 and S. fredricksonii HL-109. For these members to provide the needed precursors through lysis, a high turnover of cells would be necessary. In addition, we observed a preferential occurrence of energy-independent vitamin transporters of biotin, precursors to TPP, CoA, folate and NAD in respective vitamin producers in our community. We propose that these transporters are conditionally bidirectional, a hypothesis that is supported by the identification of the RibM energy-independent facilitator (transporter family 4.B.1) as a likely bidirectional translocator of riboflavin in Streptomyces davawensis (Hemberger et al., 2011). Orthologs to RibM (designated PnuX) are found in nearly 400 genomes in the SEED database (see ‘Riboflavin biosynthesis and transport’ subsystem in http://pubseed.theseed.org/SubsysEditor.cgi), with only 10 predicted to lack riboflavin biosynthetic genes. The occurrence of energy-independent vitamin transporters is widespread among microbes. If many are indeed bidirectional, it suggests that syntrophic partnerships that rely on them are important in shaping community behavior and structure.

Although B12 is assumed to be exchanged between prototrophs and auxotrophs, no B12 exporters were identified. We speculate that one exists because previous studies suggest that Halomonas forms a syntrophic partnership with algae, with the former providing B12 and the latter a source of carbon (Croft et al., 2005). The Halomonas spp. in our community are the only members besides S. fredricksonii HL-109 that can provide suitable forms of cobalamin for all the members. Furthermore, unlike HL-109, their biosynthetic genes are not under control of the B12 riboswitch, and they encode enzymes or pathways that bypass essential requirements for this cofactor. Therefore, biosynthesis of cobalamin destined for export has reasonable potential to be achievable.

Metabolic reconstruction of cofactor biosynthesis and salvage pathways predicted the presence of precursor transporters (importers) that we were unable to identify. The most notable absence is a nicotinamide or nicotinate transporter required by 13 organisms in our system. A gene with this function has also been elusive in Escherichia coli; no candidate nicotinate transporter has been identified in Escherichia coli even though one has been known to exist for over 30 years (Rowe et al., 1985). One possibility is that a dedicated transporter does not exist and that available levels of precursor are sufficiently high to allow for passive transport. Other systems for which we could not identify transporters (folate, pantothenate and biotin) lack recognizable vitamin-responsive regulators, hampering efforts to use regulon analysis to identify these and other system components.

Conclusion

Our genomics-based metabolic reconstruction revealed that most members of the consortia have an essential requirement for cofactor precursors and thus must obtain them from other members of the community. We also found that some of the microbes have the potential to salvage a broader variety of precursors than are currently routinely used as growth supplements. Besides impacting strategies for the cultivation of microbes, this knowledge suggests that an organizational principle of metabolite exchange in communities is to diversify utilization of precursors. Last, our analyses revealed that cofactor producers typically encode different classes of transporters than do cofactor auxotrophs. We hypothesize that transporters translocate cofactor precursors inward or outward depending on the metabolic status of the cell, which suggests that the realized demand for precursors is dynamic, and that by balancing metabolite production and usage, a community can reap the benefits of division of labor. Therefore, while cell lysis could contribute to the shared pool of nutrients, the availability of cofactor precursors in the pool can also be controlled by the metabolic status and export capabilities of respective prototrophs. Furthermore, because certain processes that depend on cofactors are essential, they can serve as control points for coordination of community member abundance and function, suggesting that division of labor in cofactor production is closely tied to community emergent properties. Important knowledge gaps that remain to be explored include the characterization of mechanisms by which cofactor precursors are made available to the community and the means by which levels of precursor pools are controlled.