Bioactive exometabolites drive maintenance competition in simple bacterial communities

ABSTRACT During prolonged resource limitation, bacterial cells can persist in metabolically active states of non-growth. These maintenance periods, such as those experienced in stationary phase, can include upregulation of secondary metabolism and release of exometabolites into the local environment. As resource limitation is common in many environmental microbial habitats, we hypothesized that neighboring bacterial populations employ exometabolites to compete or cooperate during maintenance and that these exometabolite-facilitated interactions can drive community outcomes. Here, we evaluated the consequences of exometabolite interactions over the stationary phase among three environmental strains: Burkholderia thailandensis E264, Chromobacterium subtsugae ATCC 31532, and Pseudomonas syringae pv. tomato DC3000. We assembled them into synthetic communities that only permitted chemical interactions. We compared the responses (transcripts) and outputs (exometabolites) of each member with and without neighbors. We found that transcriptional dynamics were changed with different neighbors and that some of these changes were coordinated between members. The dominant competitor B. thailandensis consistently upregulated biosynthetic gene clusters to produce bioactive exometabolites for both exploitative and interference competition. These results demonstrate that competition strategies during maintenance can contribute to community-level outcomes. It also suggests that the traditional concept of defining competitiveness by growth outcomes may be narrow and that maintenance competition could be an additional or alternative measure. IMPORTANCE Free-living microbial populations often persist and engage in environments that offer few or inconsistently available resources. Thus, it is important to investigate microbial interactions in this common and ecologically relevant condition of non-growth. This work investigates the consequences of resource limitation for community metabolic output and for population interactions in simple synthetic bacterial communities. Despite non-growth, we observed active, exometabolite-mediated competition among the bacterial populations. Many of these interactions and produced exometabolites were dependent on the community composition but we also observed that one dominant competitor consistently produced interfering exometabolites regardless. These results are important for predicting and understanding microbial interactions in resource-limited environments.

alter community functioning (5,6).It is expected that interspecies interactions play an important role in shaping microbial community dynamics (7).However, multiple stimuli in the environment make it difficult to disentangle the separate influences of abiotic versus biotic stimuli on microbial community dynamics (8).Therefore, efforts to characterize and distinguish community responses to biotic stimuli, such as those that facilitate interspecies interactions, will provide insights into the specific roles that microbial interactions play in shaping their communities (9).
Interspecies interactions can be facilitated through small molecules (10).Extracellular small molecules are collectively referred to as exometabolites (11)(12)(13).Depending on the exometabolite produced, these molecules can mediate interspecies interactions that range from competitive to cooperative (14).Of these interaction types, competition has been shown to have a major influence in structuring microbial communities (15)(16)(17).Thus, competitive interactions that are mediated by exometabolites are also expected to influence microbial community dynamics.In addition, different types of exometabolites can be employed by bacteria to gain the advantage in both exploitative (e.g., nutrient scavenging) and interference (direct cell damage) categories of competition.
Traditionally, competition has been viewed through the lens of resource acquisition (18).In previous studies, competitiveness is modeled with respect to yield given resource consumption and growth (19,20).However, competition for survival or maintenance may be just as important as competition for yield, especially during periods of resource limitation (21,22).Competition during maintenance is likely common in environments that experience relatively long periods of nutrient famine punctuated by short periods of nutrient influx, for example, as in soils, sequencing batch reactors, and the gut (23)(24)(25)(26).The stationary phase of a bacterial growth curve falls within this context of growth cessation, and pulses of nutrients may be transiently available as cells die and lyse (necromass), while the total population size remains stagnant.The stationary phase is often coordinated with a metabolic shift to secondary metabolism (27,28).There fore, an effective "maintenance" competitor may produce bioactive exometabolites, like antibiotics, which are often produced because of secondary metabolism.Bacteria can activate biosynthetic gene clusters (BSGCs) to produce bioactive exometabolites (29).The activation of BSGCs is closely tied to stress responses, suggesting that bacteria can sense the stress of competition (30,31).While it is known that certain exometabolites can trigger BSGC upregulation and, more generally, alter transcription (32), there is much to understand about the outcomes of interspecies interactions for BSGCs in multi-mem ber microbial communities.
Here, we build on our previous research to understand how exometabolite-mediated interactions among bacterial neighbors contribute to community outcomes in a simple, three-member community (Table 1).These three members are commonly associated with terrestrial environments (soils or plants) and were chosen because of reported (33) and observed interspecies exometabolite interactions in the laboratory.We used a synthetic community (SynCom) approach (34) by applying our previously described transwell system (35), which allowed for the evaluation of "community goods" within a media reservoir that was shared among members.The members' populations were physically isolated by membrane filters at the bottom of each transwell but could interact chemically via the reservoir.In our prior work, we investigated each member's exometabolites and transcription over a stationary phase, and the objective was to understand monoculture responses (in minimal glucose media) before assembling the more complex two-and three-member communities.We previously found that each member in monoculture produced a variety of exometabolites in the stationary phase, including bioactive molecules involved in competition (36).In this work, we build twoand three-member communities to ask: How do members interact via exometabolites in simple communities during maintenance (stationary phase), and what are the competi tive strategies and outcomes of those interactions?What genetic pathways, molecules, and members drive the responses?We found that B. thailandensis had a major influence on the transcriptional respon ses of both C. subtsugae and P. syringae and that this influence could be attributed to an increase in both interference and exploitative competition strategies.These findings show that diverse competitive strategies can be deployed even when bacte rial neighbors are surviving rather than exponentially growing.Therefore, we suggest that contact-independent, exometabolite-mediated interference and exploitation are important competitive strategies in resource-limited environments and support the non-yield outcome of maintenance.

Bacterial strains and culture conditions
We selected three environmental bacterial strains for the SynCom experiments that were originally isolated from various plant/soil habitats and that had prior evidence of exometabolite interactions among them in the laboratory (Table 1; 33,[37][38][39][40].Freezer stocks of B. thailandensis, C. subtsugae, and P. syringae were plated on half-concentration Trypticase soy agar (TSA50) at 27°C for at least 24 h.Members were inoculated in 7 mL of M9-0.2% glucose medium and grown for 16 h at 27°C, 200 rpm.Cultures were then diluted into 50 mL M9-0.2% glucose medium such that the exponential growth phase was achieved after 10 h of incubation at 27°C, 200 rpm.Members were diluted in 50 mL M9 glucose medium to target ODs (B.thailandensis 0.3 OD, C. subtsugae: 0.035 OD, P. syringae 0.035 OD).The high initial OD for B. thailandensis was necessary such that the stationary phase would be achieved by all members within a 2-h window after 24-h incubation in the transwell plate.The glucose concentration in the final dilution varied upon community membership-0.067%for monocultures, 0.13% for pairwise cocultures, and 0.2% for the three-member community.For each member, 48 mL of diluted culture was transferred as 4 mL aliquots in 12, 5 mL Falcon tubes to more efficiently prepare replicate transwell plates.

Synthetic community experiments
Transwell plate preparation was performed as previously described (35).Briefly, we used sterile filter plates with 0.22-μm-pore polyvinylidene difluoride (PVDF) filter bottoms (Millipore MAGVS2210).Prior to use, filter plates were washed three times with sterile water using a vacuum apparatus (NucleoVac 96 vacuum manifold; Clontech Laborato ries).The filter of well H12 was removed with a sterile pipette tip and tweezer, and 31 mL of M9 glucose medium was added to the reservoir through well H12.The glucose concentration in the reservoir varied upon community membership-0.067%for monocultures, 0.13% for pairwise cocultures, and 0.2% for the three-member com munity.Glucose concentration was adjusted to plate occupancy (e.g., three-member communities had a higher number of wells occupied than two-or one-member).Our aim was for each member to achieve the stationary phase at similar times across all conditions to compare transcripts and exometabolites under similar growth trajectories.In other words, available resources were standardized while keeping the well occupancy for each member constant.With this design, transcripts and exometabolites in cocultures that deviated from those in monocultures could be attributed to interspecies interac tions and not complicated by offset in member growth trajectories across the experi mental conditions.
Each well was filled with 130 µL of culture or medium (prepared as described, above; see Materials and Methods section: Bacterial strains and culture conditions).For each plate, a custom R script [RandomArray.R (see the script at https://github.com/ShadeLab/PAPER_Chodkowski_mSystems_2017/blob/master/R_analysis/RandomArray.R)] was used to randomize community member placement in the wells so that each mem ber occupied a total of 31 wells per plate.In total, there were seven community conditions-three monocultures, three pairwise cocultures, and the three-member community.Each member occupied 31 wells per plate regardless of experimental condition.Thus, "baseline" exometabolites could be determined in the monocultures, and then deviations in exometabolite abundance or detection in the cocultures could be attributed to interspecies interactions.A time course was performed for each replicate.The time course included an exponential phase time point (12.5 h) and five time points assessed every 5 h over a stationary phase (25-45 h).Four biological replicates were performed for each community condition for a total of 28 experiments.For each experiment, six replicate filter plates were prepared for destructive sampling for a total of 168 transwell plates.
Filter plates were incubated at 27°C with gentle shaking (~0.32 rcf ).For each plate, a custom R script [RandomArray.R (see the script at https://github.com/ShadeLab/PAPER_Chodkowski_mSystems_2017/blob/master/R_analysis/RandomArray.R)] was used to randomize wells for each organism assigned to RNA extraction (16 wells) and flow cytometry (five wells).The following procedure was performed for each organism when a transwell plate was destructively sampled: (i) wells containing spent culture assigned to RNA extraction were pooled (~100 µL/well) into a 1.7 mL microcentrifuge tube and flash frozen in liquid nitrogen and stored at −80 until further processing.(ii) 20 µL from wells assigned for flow cytometry were diluted into 180 µL Tris-buffered saline [TBS; 20 mM Tris, 0.8% NaCl (pH 7.4)].In community memberships where P. syringae was arrayed with B. thailandensis, P. syringae had a final dilution of 70-fold in TBS.In community memberships where P. syringae was arrayed in monoculture or coculture with C. subtsugae, P. syringae had a final dilution of 900-fold in TBS.Final dilutions for B. thailandensis and C. subtsugae were one 300-fold and one 540-fold, respectively.Each member was diluted differently to achieve a suitable events/second range on the flow cytometer for accurate cell counting.Populations were then stained and analyzed on the flow cytometer for live/dead counting (see Supplementary Methods).(iii) Spent medium (~31 mL) from the shared reservoir was transferred to 50 mL conical tubes, flash-frozen in liquid nitrogen, and stored at −80°C prior to metabolite extraction.

Quality filtering and differential gene expression analysis
RNA extraction, sequencing, quality control, and count matrix generation were performed as previously published [ (36), see Supplementary Methods].Count matrices for each member were quality filtered in two steps: genes containing 0 counts in all samples were removed, and genes with a transcript count of ≤10 in more than 90% of samples were removed.DESeq2 (41) was used to extract size factor and dispersion estimates.These estimates were used as external input into ImpulseDE2 for the analysis of differentially regulated genes (42).ImpulseDE2 determines differential expression by comparing longitudinal count data sets.Case-control (Cocultures-monoculture control) analyses were analyzed to identify genes with differences in temporal regulation at an FDR-corrected threshold of 0.01.Genes that passed the FDR threshold were further filtered for genes that had at least one time point with a log2 fold-change (LFC) >= 1 or <= −1.Thus, we defined differentially expressed genes (DEGs) as genes that met both the FDR-corrected and LFC thresholds.For each member, differences in gene regulation between the three coculture conditions were visualized with Venn diagrams using the VennDiagram package (43).
Differentially expressed genes were first determined by comparing each coculture condition to the monoculture control and applying an LFC threshold (see above).We then determined a second set of DEGs by comparing pairwise cocultures to each other.ImpulseDE2 case-control analyses were performed as follows: B. thailandensis coculture with C. subtsugae (case) compared to B. thailandensis coculture with P. syringae (control), C. subtsugae coculture with B. thailandensis (case) compared to C. subtsugae coculture with P. syringae (control), and P. syringae coculture with B. thailandensis (case) compared to P. syringae coculture with C, subtsugae (control).Genes that passed the FDR-corrected threshold of 0.01 based on ImpulseDE2 analysis and had at least one time point with an LFC of >= 1 or <= −1 represented coculture-specific DEGs.The DEGs determined from monoculture comparisons and coculture comparisons were then categorically grouped using Clusters of Orthologous Groups (COG).

COG analysis
Protein fasta files were downloaded from NCBI and uploaded to eggNOG-mapper v2 (http://eggnog-mapper.embl.de/) to obtain COGs.The DEGs determined from Impul seDE2 and LFC thresholds were categorized as upregulated or downregulated based on temporal expression patterns.DEGs with consistent positive LFC throughout all stationary phase time points were categorized as upregulated.DEGs with consistent negative LFC throughout all stationary phase time points were categorized as down regulated.These DEGs were then assigned to COGs, grouped based on temporal up/ downregulation patterns, and plotted using ggplot2 (44).

Principal coordinates analysis and statistics
Normalized gene matrices were extracted from DESeq2 and filtered to only contain DEGs (coculture to monoculture comparisons) based on our previously described definition.A variance-stabilizing transformation was performed on normalized gene matrices using the rlog function in DESeq2.A distance matrix based on the Bray-Curtis dissimilarity metric was then calculated on the variance-stabilized gene matrices and principal coordinates analysis was performed using the R package vegan (45).Principal coordi nates were plotted using ggplot2.Coordinates of the first two PCoA axes were used to perform PROTEST analysis using the PROTEST function in vegan.Dissimilarity matrices were used to perform PERMANOVA and variation partitioning using the adonis and varpart functions in vegan, respectively.The RVAideMemoire package (46) was used to perform post hoc pairwise PERMANOVAs.Lastly, distances were extracted from the Bray-Curtis dissimilarity matrix that compared each coculture condition to the monocul ture condition at each time point within each member.These distances were used to produce time series distance plots.

BSGC analysis
NCBI accession numbers were uploaded to the antiSMASH 6 beta bacterial version (47) to identify genes involved in BSGCs using default parameters.Where possible, literature-based evidence and BSGCs uploaded to MIBiG (48) were used to better inform antiSMASH predictions.Log2 fold-changes (LFCs) were calculated for all predicted biosynthetic genes within each predicted cluster by comparing coculture expression to monoculture expression at each time point.Average LFCs were calculated from all predicted biosynthetic genes within a predicted BSGC at each time point.Temporal LFC trends were plotted using ggplot2.An upregulated BSGC was defined as a BSGC that had at least two consecutive time points in the stationary phase with an LFC > 1.

Network analysis
Unweighted co-expression networks were created from quality-filtered and normal ized expression data.Networks were generated for pairwise cocultures containing B. thailandensis.First, data were quality filtered as previously described (see methods section: Quality filtering and differential gene expression analysis).Then, normalized expression data were extracted from DESeq2.Twenty-three and twenty-four RNA-seq samples from each member were used for network analysis in the B. thailandensis-C.subtsugae and B. thailandensis-P.syringae cocultures, respectively (23/24 samples/mem ber; six time points, four biological replicates).Only 23 samples were used in the B. thailandensis-C.subtsugae network analysis because RNA-seq failed for C. subtsugae at 45 h, biological replicate 2. Interspecies networks were then inferred from the expression data using the context likelihood of relatedness (49) algorithm within the R package Minet (50).Gene matrices for each coculture pair were concatenated to perform the following analysis.Briefly, the mutual information coefficient was determined for each gene pair.To ensure robust detection of co-expressed genes, a resampling approach was used as previously described (51).Then, a Z-score was computed on the mutual information matrix.A Z-score threshold of 4.5 was used to determine an edge in the interspecies network.Interspecies networks were uploaded into Cytoscape version 3.7.1.for visualization, topological analysis, and enrichment analysis (52).
Gene annotation and gene ontology (GO) files were obtained for B. thailandensis, P. syringae, and C. subtsugae for enrichment analyses.For B. thailandensis, annotation and ontology files were downloaded from the Burkholderia Genome Database (https:// www.burkholderia.com).For P. syringae, annotation and ontology files were downloa ded from the Pseudomonas Genome Database (http://www.pseudomonas.com/strain/download).Annotation and ontology files for C. subtsugae were generated using Blast2GO version 5.2.5 (53).InterProScan (54) with default parameters was used to complement gene annotations from C. subtsugae.GO terms were assigned using Blast2GO with default parameters.In addition, genes involved in secondary metabolism were manually curated and added to these files as individual GO terms.These genes were also used to update the GO term GO:0017000 (antibiotic biosynthetic process), composed of a collection of all the biosynthetic genes.(see Materials and Methods section: Biosynthetic gene cluster (BSGC) analysis).
Topological analysis was performed as follows: Nodes were filtered from each coculture network to only select genes from one member at a time.The GLay community cluster function in Cytoscape was used to determine intra-member modules.Functional enrichment analysis was then performed on the modules using the BiNGO package (55) in Cytoscape.
To determine interspecies co-regulation patterns, we filtered network nodes that contained an interspecies edge.Functional enrichment analysis was performed on the collection of genes containing interspecies edges for each member using the BiNGO package in Cytoscape.Then, we selected all genes contained within modules of interest (e.g., B. thailandensis modules containing either thailandamide or malleilactone genes in the B. thailandensis-C.subtsugae coculture network or B. thailandensis-P.syringae coculture network, respectively) in Cytoscape.Node selection was extended by selecting the first neighbors of the selected nodes.This resulted in interspecies edges.The resultant nodes were transformed into a circular layout and exported for manual edits in InkScape.The biosynthetic gene cluster organization of thailandamide and malleilactone was obtained from MIBig and drawn in InkScape.
Protein sequences from an interspecies gene of interest (CLV_2968) within a network module that also contained thailandamide genes from the B. thailandensis-C.subtsugae network and an interspecies gene of interest (PSPTO_1206) within a network module that also contained malleilactone genes from the B. thailandensis-P.syringae network was obtained.A protein blast for each protein was run against B. thailandensis protein sequences.B. thailandensis locus tags were extracted from the top blast hit from each run.Normalized transcript counts for these four genes of interest were plotted in R. Time course gene trajectories were determined using a loess smoothing function.

LCMS, feature detection, and quality control
Standard operating protocols were performed at the Department of Energy Joint Genome Institute as previously described (36).MZmine 2 (56) was used for feature detection and peak area integration as previously described (36).Select exometabolites were identified in MZmine 2 by manual observation of both MS and MS/MS data.We extracted quantities of these identified exometabolites for ANOVA and Tukey HSD post-hoc analysis in R. We filtered features in three steps to identify coculture-accumula ted exometabolites.The feature-filtering steps were performed as follows on a per-mem ber basis: (i) retain features where the maximum peak area abundance occurred in any of the coculture communities ; (ii) a noise filter, the minimum peak area of a feature from a replicate at any time point needed to be three times the maximum peak area of the same feature in one of the external control replicates, was applied; and (iii) coefficient of variation (CV) values for each feature calculated between replicates at each time point needed to be less than 20% across the time series.
Four final feature data sets from polar and nonpolar analyses in both ionization modes were analyzed in MetaboAnalyst 5.0 (57), as reported in our prior work (36, see Supplementary Methods).In addition, exometabolites categorized as primary metabo lites were identified according to Metabolomics Standards Initiative (MSI) level 1 criteria (58), as reported in our prior work (36, see Materials and Methods).

Principal coordinates analysis and statistics
A distance matrix based on the Bray-Curtis dissimilarity metric was used to calcu late dissimilarities between exometabolite profiles.Principal coordinates analysis was performed using the R package vegan.Principal coordinates were plotted using ggplot2.Coordinates of the first two PCoA axes were used to perform Protest analysis using the protest function in vegan.Dissimilarity matrices were used to perform PERMANOVA and variation partitioning using the adonis and varpart functions in vegan, respectively.The RVAideMemoire package was used to perform post hoc pairwise PERMANOVAs.Monoculture controls were removed to focus on coculture trends.

Overview
Our major data types included both transcriptomics and metabolomics, and we integrate these to interpret SynCom dynamics and interactions.Our longitudinal design resulted in 288 RNAseq samples across the three members, and 168 com munity metabolomics samples analyzed in each of four mass spectral modes (polar/ nonpolar, positive/negative modes = 672 total mass spectral profiles).After qual ity control, we were left with 281 RNAseq and 605 total mass spectral profiles for the integrated analyses (https://github.com/ShadeLab/Paper_Chodkowski_3member_SynCom_2021/tree/master/SummaryOfSamples).First, we present a summary of experiments and cell viability (section "SynCom design/sampling scheme and member ship cell viability").Then, we present the results of general responses of transcription (section "Stationary phase transcript dynamics of microbial community members") and exometabolomics (section "Stationary phase exometabolite dynamics of microbial communities"), separately.Then, we integrate transcriptomic and metabolomic efforts to determine the upregulation of BSGCs and identify exometabolites of interest from mass spectrometry (section "B.thailandensis increases competition strategies in the presence of neighbors").Lastly, we then present a transcriptomics co-expression network to ask how the upregulation of BSGCs influenced interspecies interactions through coordinated longitudinal gene expression (section "Interspecies co-transcriptional networks reveal coordinated gene expression related to competition").

SynCom design/sampling scheme and membership cell viability
We had four replicate, independent time series for each of the seven community memberships (three of each monoculture, three of each pair in coculture, and the three-member community).We define membership as the specific strains present in a given condition.Here, we focus on the multi-member analyses (two-and three-mem ber combinations) to gain insights into community outcomes (Fig. 1A).The SynCom transwell system isolated member populations among separate transwells but permitted the exchange of their collective exometabolites via the plate's shared media reservoir (Fig. 1B).We collected data (transcripts, metabolites, etc) over a time series that included one exponential phase time point (12.5 h) followed by five stationary phase time points (25-45 h sampled at 5-h intervals; Fig. 1C).
We observed relatively unchanged viability in B. thailandensis across all conditions (Fig. S1; panels A through D).On the contrary, we observed a slight reduction (~2.1 log2 fold change) in C. subtsugae live cell counts, and a drastic reduction (~4.7 log2 fold change) in P. syringae live cell counts, when either member was cocultured with B. thailandensis (Fig. 2; panels A vs C and panels D vs F, respectively).Reductions in cell viability of C. subtsugae and P. syringae were also present in the 3-member community (Fig. S1; panels E and F). C. subtsugae and P. syringae had minimal effects on each other (Fig. 2; panels B and E).Dead cell accumulation of P. syringae plateaued in coculture conditions compared to monoculture, suggesting cell lysis (Fig. 2, panels D through F).We note that one doubling occurred in B. thailandensis and P. syringae monocultures, and in C. subtsugae in pairwise coculture with P. syringae.We elaborated on this finding as the possibility of a reductive cell division in our previous manuscript (36).

Stationary phase transcript dynamics of microbial community members
Differentially expressed genes were determined by comparing time series transcript trajectories applying an FDR and LFC threshold (see Materials and Methods: Quality filtering and differential gene expression analysis).First, we compared each coculture to the monoculture control.A range of 153 to 276 genes were differentially expressed by each member in the coculture, irrespective of the identity of neighbors (Fig. S2).In addition, each member also had differential gene expression that was unique to a particular neighbor(s).Summarizing across all cocultures, 1,089/5,639 (19.3%) coding sequences (CDSs), 1,991/4,393 CDSs (45.3%), and 3,274/5,576 CDSs (58.7%).DEGs were determined for B. thailandensis, C. subtsugae, and P. syringae, respectively.The primary drivers of transcriptional response patterns for each member were community membership (PCoA axis 1) and time (PCoA axis 2) (Fig. 3; Table S1).Together, these data suggest that there are both general and specific consequences of neighbors for the transcriptional responses of these bacterial community members.
Temporal trajectories in member transcript profiles were generally reproducible across replicates (PROTEST analyses, Table S2).Each member had a distinct transcript profile (0.480 ≤ r2 ≤0.778 by Adonis; P value, 0.001; all pairwise false discovery rate [FDR]-adjusted P values, ≤0.01 except for two community memberships, Table S3).For all ordinations, community membership had the most explanatory value (Axis 1), followed by time (Axis 2), with the most variation explained by the interaction between time and membership (Table S1).Membership alone accounted for 60.6% and 77.0% of the variation explained in C. subtsugae and P. syringae analyses, respectively, and 46.3% in the B. thailandensis analysis (Table S1).
When included in the community, B. thailandensis strongly determined the transcript profiles of the other two members.For example, the inclusion of B. thailandensis in a coculture differentiated transcript profiles for both C. subtsugae and P. syringae (Fig. 3B and C; Fig. S3 to S5).The transcript profile differences between monoculture and coculture conditions are largest for C. subtsugae (Fig. S4) and P. syringae (Fig. S5) when B. thailandensis is included in the coculture.Thus, B. thailandensis appears to have had a dominating influence on the transcriptional response of neighbors, and these responses were dynamic with respect to time.We analyzed clusters of orthologous groups of proteins (COGs) to infer the responses of members to their neighbors.DEGs were categorized as upregulated or downregu lated based on temporal patterns and representation in COGs (Fig. S6).We focused on the largest differences between total DEGs upregulated and total DEGs downregu lated within a COG, which provides insights into broad biological processes affected by community membership.COGs with large differences toward upregulation in B. thailandensis included cell motility [N], secondary metabolites biosynthesis, transport, and catabolism [Q], and signal transduction mechanisms [T] while COGs with large differences toward downregulation included defense mechanisms [V], energy produc tion and conversion [C], translation, and ribosomal structure and biogenesis [J].These results suggest that B. thailandensis responds to neighbors via downregulation of growth and reproduction and upregulation of secondary metabolism.We, therefore, hypothe sized that B. thailandensis was producing bioactive exometabolites against C. subtsugae and P. syringae to competitively inhibit their growth.
Because of the strong transcript response of C. subtsugae and P. syringae when neighbored with B. thailandensis (Fig. 3B and C), we focused on COGs within community memberships with B. thailandensis (Fig. S6B and S6C, rows 2 and 3).The COG with large differences toward upregulation in both C. subtsugae and P. syringae were translation, ribosomal structure, and biogenesis [J].COG groups tending toward downregulation in C. subtsugae and P. syringae were signal transduction mechanisms [T] and secon dary metabolites biosynthesis, transport, and catabolism [Q], respectively.These results suggest that the presence of B. thailandensis alters its neighbor's ability to respond to the environment and inhibits secondary metabolism.The effects of B. thailandensis on C. subtsugae and P. syringae were also evident by mapping timeseries LFCs onto KEGG pathways.Various enzymes involved in central metabolism, fatty acid degradation, growth, transport, and response systems were upregulated when B. thailandensis was grown with either or both members (https://figshare.com/s/b7f5e559a32cc5c8a61f).
The above analyses focused on DEGs determined by comparing each coculture to the monoculture control.However, we also wanted to understand differences between pairs to determine whether the alterations in transcripts were attributed to specific memberships (aka interspecies interactions).A total of 436, 1,762, and 2,962 DEGs were determined when comparing the pairs including B. thailandensis, the pairs including C. subtsugae, and the pairs including P. syringae, respectively.We detected member-specific effects on the COGs that were differentially expressed (Fig. S7).These data suggest that there were transcriptional changes driven by particular members and given their partners.Due to the physical separation of members in our SynCom plate system, these member-specific interspecies interactions were very likely exometabolite mediated.

Stationary phase exometabolite dynamics of microbial communities
Because member populations are physically separated in the SynCom transwell system but allowed to interact chemically, observed transcript responses in different community memberships are inferred to result from exometabolite interactions.Spent medium from the shared medium reservoir was collected from each transwell plate and analyzed using mass spectrometry to detect exometabolites.Our previous manuscript focused on exometabolite dynamics in monocultures (36).Here, we focused our analysis on those exometabolites that had maximum accumulation in a coculture (either in pairs or in a three-member community).Consistent with the transcript analysis, we found that both community membership and time explained the exometabolite dynamics and that the explanatory value of membership and time was maintained across all polarities and ionization modes (Fig. 4; Table S4).
Temporal trajectories in exometabolite profiles were generally reproducible across replicates with some exceptions (PROTEST analyses, Table S5; Supplementary File 1).Exometabolite profiles were distinct by community membership (0.475 ≤ r2 ≤0.662 by Adonis; P value, 0.001; all pairwise false discovery rate [FDR]-adjusted P values, ≤0.01 except for two comparisons, Table S6) and also dynamic over time.As observed for the member transcript profiles, the interaction between membership and time had the highest explanatory value for the exometabolite data (Table S4).
We found that the C. subtsugae-P.syringae coculture exometabolite profiles were consistently the most distinct from the other coculture memberships (Fig. 4), support ing, again, that the inclusion of B. thailandensis was a major driver of exometabolite dynamics, possibly because it provided the largest or most distinctive contributions to the community exometabolite pool.Indeed, we observed that a majority of the most abundant exometabolites were either detected uniquely in the B. thailandensis monoculture or accumulated substantially in its included community memberships (Fig. S8).Some exometabolites detected in B. thailandensis-inclusive communities were not detected in its monocultures (Fig. S8D), suggesting that the inclusion of neighbors contributed to the accumulation of these particular exometabolites (e.g., upregulation of biosynthetic gene clusters or lysis products).C. subtsugae and P. syringae contributed less to the three-member community exometabolite profile, as exometabolites detected in the C. subtsugae-P.syringae coculture were less abundant and had lower accumulation over time in the three-member community (Fig. S8A).Together, these results suggest that B. thailandensis can suppress or overwhelm expected outputs from neighbors.
Exometabolites categorized as primary metabolites were identified according to Metabolomics Standards Initiative (MSI) level 1 criteria (58).We identified primary metabolites accumulated in the shared medium reservoir over time in each monoculture (Fig. 5; (36)) to compare their dynamics in cocultures.These primary metabolites were detected to decrease in concentration across coculture conditions, suggesting meta bolic inhibition or interspecies uptake.In addition, we also found a subset of primary metabolites that accumulate substantially in the exponential phase in monocultures (Fig. S9).Taken together, each member contributed a unique set of primary metabolites to the community exometabolite pool.The uptake and metabolism of these primary metabolites by the non-producing members may directly affect the available pool of exometabolites in cocultures, particularly with respect to exometabolites contributed from secondary metabolism.
In summary, we observed both increased accumulation and unique production of exometabolites in pairs and in the three-member community, with B. thailandensis contributing the most to the shared exometabolite pool as determined by comparisons with its monoculture exometabolite profile.Related, the transcriptional responses of C. subtsugae and P. syringae in the three-member community are most like their respective transcriptional response when neighbored with B. thailandensis alone, despite the presence of the third neighbor.

B. thailandensis increases competition strategies in the presence of neighbors
Given the observed reduction in cell viability (Fig. 2) and that there have been competi tive interactions between B. thailandensis and C. subtsugae previously reported (33), we hypothesized that B thailandensis was using competition strategies to influence its neighbors via production of bioactive exometabolites.If true, we would expect transcrip tional upregulation in B. thailandensis biosynthetic gene clusters (BSGC) that encode bioactive exometabolites.Indeed, when compared to the monoculture control, we found evidence of upregulated BSGCs across various time points in stationary phase in B. thailandensis cocultures (Fig. 6; Table S7).Some of these upregulation patterns were associated with particular pairs of members and some upregulation patterns were strongest in the full community (e.g., thailandamide).For example, B. thailandensis upregulated an unidentified non-ribosomal peptide synthetase (NRPS) when paired with P. syringae, but when paired with C. subtsugae, upregulated a different BSGC encoding an unidentified beta-lactone.This suggests that B. thailandensis responded to neighbors by upregulating genes involved in the production of bioactive compounds, likely to gain a competitive advantage.However, not all BSGCs in B. thailandensis were upregulated.Some BSGCs were unaltered or downregulated (Fig. S10).C. subtsugae upregulated only 1 BSGC, an uncharacterized hybrid nonribosomal peptide synthetase-type I polyketide synthase, in coculture with B. thailandensis, while P. syringae did not upregulate any BSGC in any coculture (Fig. S11 and S12).Interspecies interactions led to the upregulation of BSGCs in both B. thailandensis and C. subtsugae and three of these BSGCs encode potentially novel bioactive exometabolites.
Because B. thailandensis upregulated the transcription of various BSGCs when grown in cocultures, we asked if this led to the unique production of or increased accumulation of secondary metabolites as compared to when it was grown in monoculture.We identified 6 of the 11 exometabolites from the BSGCs in B. thailandensis that were upregulated and quantified their abundances from mass spectrometry data (Fig. 7; Supplementary File 2).We found that each identified exometabolite differentially accumulated between community memberships containing B. thailandensis (Table S8), particularly when comparing the B. thailandensis monoculture to each coculture (Table S9).As expected, these identified exometabolites were not detected in communities that did not include B. thailandensis (data not shown).Bactobolin was the only identified exometabolite that accumulated in monoculture to equivalent levels of accumulation in all coculture conditions.The other identified secondary metabolites were not detected or did not accumulate in monoculture, suggesting interspecies induction of secondary metabolism.Thus, in response to an exometabolite from either C. subtsugae or P. syringae, B. thailandensis increased its competitive strategies through the upregulation and production of many bioactive exometabolites.Of these bioactive exometabolites, three are documented antimicrobials (59)(60)(61), two are siderophores (62,63), and one is a biosurfactant (64).We conclude that B. thailandensis produced bioactive exometabolites to competitively interact using both interference and exploitative competition strategies (65).Given that B. thailandensis upregulated competitive strategies, and responded more broadly in producing competition-supportive exometabolites when grown with neighbors, we hypothesized that these bioactive exometabolites are responsible for the altered transcriptional responses in C. subtsugae and P. syringae.
In our experimental design, we adjusted glucose concentration depending on plate occupancy.Glucose concentration increased as plate occupancy increased (31 wells vs 62 wells vs 93 wells) but a member consistently occupied 31 wells across all experimental conditions.One complication of this design is that population density and resource concentration could contribute to differences in transcripts and exometabolites in a member-agnostic manner.To address this, we performed additional SynCom experi ments to affirm confidence that some changes in transcripts and exometabolites are attributable to exometabolite-mediated interspecies interactions.In these experiments, we increased the plate occupancy of B. thailandensis in monoculture while subsequently increasing resource concentration.Pairwise cocultures and the three-member commun ity SynCom experiments were repeated as well (see Supplementary methods).We calculated the relative gene expression of three genes in the thailandamide operon (thaF, thaK, and thaQ) through RT-qPCR by comparing each experimental condition to the monoculture control (B.thailandensis, 31 wells in M9-0.067% glucose).Decreased gene expression was observed across all three genes as both plate occupancy and resource concentration increased in B. thailandensis monocultures.In fact, thaF, thaK, and thaQ gene expression was further reduced in the 93-well B. thailandensis monoculture compared to the 62-well B. thailandensis monoculture, suggesting that the thailanda mide operon trended toward reduced expression as a function of B. thailandensis plate occupancy in monoculture conditions.On the contrary, thaF, thaK, and thaQ had increased expression in all coculture memberships, suggesting that exometabolite interspecies interactions were responsible for the increased expression of a BSGC in B. thailandensis (Table S10).

Interspecies co-transcriptional networks reveal coordinated gene expression related to competition
We performed interspecies co-expression network analysis to infer interspecies interac tions.We used temporal profiles from transcriptional responses to generate co-expres sion networks for B. thailandensis-C.subtsugae and B. thailandensis-P.syringae cocultures, respectively (Table S11).As expected, the majority of nodes in the network had intraspe cies edges only, with interspecies edges comprising 1.85% and 1.90% of the total edges in the B. thailandensis-C.subtsugae and B. thailandensis-P.syringae networks, respectively.We explored interspecies edges for evidence of interspecies transcriptional co-regula tion.We performed two analyses, module analysis and Gene Ontology (GO) enrichment, to validate networks and infer interspecies interactions (Fig. S13).Module analysis validated networks as intraspecies modules enriched for biological processes (Supplementary File 3).To infer interspecies interactions, we filtered genes with interspecies edges and performed enrichment analysis (Supplementary File 4).The top enriched GO term for B. thailandensis when paired with C. subtsugae was an antibiotic synthesis of thailandamide, supporting the interpretation of interference competition.Though the top enriched GO term in B. thailandensis when paired with P. syringae was bacterial-type flagellumdependent cell motility, antibiotic synthesis of malleilactone was also enriched.Both thailandamide genes from the B. thailandensis-C.subtsugae network (Fig. 8) and mallei lactone genes from the B. thailandensis-P.syringae network (Fig. S14) formed nearcomplete modules within their respective BSGCs.In addition, genes that were part of the BSGC modules contained interspecies edges with both C. subtsugae and P. syringae.
At least one gene from each of B. thailandensis's upregulated BSGCs (Fig. 6) had an interspecies edge, except for rhamnolipid.Our interpretation of this result is that, generally, B. thailandensis's upregulated BSGCs had co-expression patterns with genes from the other members.In the thailandamide and malleilactone modules, some of these interspecies genes were related to stress, transport, and iron-scavenging (Supple mentary File 5).The top GO term for both C. subtsugae and P. syringae genes that had edges shared with B. thailandensis was bacterial-type flagellum-dependent motility.Other notable enriched GO processes were efflux activity for C. subtsugae and signal transduction for P. syringae.Specifically, a DNA starvation/stationary phase gene (CLV04_2968, Fig. 8), dspA, was within the network module that also contained thailanda mide genes from the B. thailandensis-C.subtsugae network and a TonB-dependent siderophore receptor gene (PSPTO_1206, Fig. S14) was within the network module that also contained malleilactone genes from the B. thailandensis-P.syringae network.Interestingly, both CLV04_2968 and PSPTO_1206 were DEGs and downregulated when cocultured with B. thailandensis (Fig. S15A and S16A, respectively).In addition, the closest homolog for dspA in B. thailandensis was unaltered (BTH_I1284, Supplementary File 6) when cocultured with C. subtsugae (Fig. S15B) and the closest homolog to the TonBdependent receptor in B. thailandensis (BTH_I2415, Supplementary File 7) was a DEG and upregulated when cocultured with P. syrinage (Fig. S16B).Taken together, these coexpression networks revealed interspecies coordinated expression patterns.Specifically, we detected interspecies co-expression patterns related to antibiotic upregulation in B. thailandensis, suggesting C. subtsugae and P. syringae were sensing and responding directly to these competition strategies of B. thailandensis.

DISCUSSION
Here, we used a synthetic community system to understand how exometabolomic interactions determine members' transcriptional responses and exometabolite outputs.Our experiment used a systems approach to compare the seven possible community memberships of three members, and their dynamics in member transcripts and com munity exometabolites over stationary phase.Differential gene expression across community memberships and over time show that the exometabolites released by a member were sensed and responded to by their neighbors.Furthermore, members' outputs in monocultures changed because of coculturing, as evidenced by differential exometabolite production.The largest transcriptional alterations in C. subtsugae and P. syringae occurred when cocultured with B. thailandensis.Global expression patterns in C. subtsugae and P. syringae when in the three-member community still resembled expression patterns in pairwise cocultures with B. thailandensis.These transcriptional alterations in C. subtsugae and P. syringae were coordinated with increases in B. thailan densis competitive strategies (evaluated by BSGC transcript upregulation and exometa bolite abundance).That interactions within a relatively simple community altered the transcriptional responses and exometabolite outputs of each member are important because these kinds of alterations could, in turn, drive changes in community structure FIG 8 B. thailandensis genes involved in thailandamide production are co-expressed with C. subtsugae genes.A network module containing the thailandamide BSGC is shown (A).The network module nodes are color-coded by according to the following criteria: thailandamide biosynthetic genes that had interspecies edges (magenta), thailandamide biosynthetic genes that did not have interspecies edges (orange), other B. thailandensis genes that were not part of the BSGC (yellow), and genes that were from C. subtsugae (blue).The chromosomal organization of the thailandamide BSGC is shown below the network module (B).The same colors are applied to the BSGC operon.The operon also depicts genes that were not detected within the interspecies network, shown in gray.Asterisks indicate core biosynthetic genes in the BSGCs, as predicted from antiSMASH.Table (C) shows upregulated B. thailandensis BSGCs (Fig. 6) and whether interspecies edges were detected (check is yes, x is no).and/or function in an environmental setting.For example, it was shown that interspecies interactions more strongly influenced the assembly of C. elegans gut communities than host-associated factors (66).Therefore, mechanistic and ecological characterization of interspecies interactions will inform as to the principles that govern emergent properties of microbial communities.
Overall, competitive interactions predominated in this synthetic community.This was first evidenced by reductions in viable cell counts in both C. subtsugae and P. syringae when cocultured with B. thailandensis.Interestingly, P. syringae was the only member to have an exponential increase in dead cell counts in monoculture.P. syringae dead cell count accumulation ceased in coculture conditions.We attribute this finding to the overall reduction of cell viability and/or lysis of dead cells when cocultured.
Our previous study found that, over the stationary phase in monocultures, each member released and accumulated at least one exometabolite documented to be involved in either interference or exploitative competition (36).This suggests that entry into stationary phase primed members for competitive interactions, regardless of heterospecific neighbors present.We interpret this strategy of preemptive aggression to be especially advantageous to B. thailandensis, as it successfully used competitive strategies against both C. subtsugae and P. syringae.B. thailandensis's success was supported by decreased viable P. syringae cells when cocultured with B. thailandensis.Though C. subtsugae viable cell counts were not as affected directly by the coculture with B. thailandensis, B. thailandensis-produced bactobolin (67) was detected in the shared medium reservoir.Bactobolin is a bacteriostatic antibiotic previously shown to be bioactive against C. subtsugae (33) through ribosome binding (59).However, C. subtsugae can resist bactobolin through upregulation of an RND-type efflux pump (68).This finding also is supported by our data, as all genes coding for the CdeAB-OprM RND-type efflux system were DEGs and upregulated in C. subtsugae cocultures with B. thailandensis (CLV04_2413-CLV04_2415).
When cocultured with B. thailandensis, we observed COG groups such as translation, ribosomal structure, and biogenesis [J] had large differences toward upregulation in both C. subtsugae and P. syringae.At first glance, this seems at odds with our inter pretation of B. thailandensis competitiveness toward C. subtsugae and P. syringae.In other words, how is B. thailandensis effectively competing via interference competition whether both C. subtsugae and P. syringae are upregulating machinery for growth?There are both theoretical (69) and experimental (70) evidence that show how cells treated with antibiotics stimulate ribosomal production to maintain a sufficient number of active ribosomes.As previously mentioned, B. thailandensis-produced bactobolin binds to the ribosome and can inhibit C. subtsugae (33,59).We also have evidence that bactobolin inhibits P. syringae (data not shown).It could be that bactobolin is stimulating riboso mal production in C. subtsugae and P. syringae as a survival mechanism to maintain protein production by maintaining enough active ribosomes.There was also evidence of B. thailandensis antibiotic efficacy against C. subtsugae and P. syringae, including general loss of cell viability and upregulation of various enzymes involved in central metabolism by both members when they were cocultured with B. thailandensis (https:// figshare.com/s/b7f5e559a32cc5c8a61f).These patterns are consistent with antibiotic treatments in Escherichia coli and Staphylococcus aureus where the upregulation oxidative phosphorylation due to drug treatment contributes to antibiotic efficacy (71,72).A barrage of B. thailandensis-produced antibiotics (Fig. 6 and 7) likely drove the transcriptional patterns in C. subtsugae and P. syringae.
We acknowledge that this study is limited in its ability to pinpoint the underly ing mechanisms driving the activation of secondary metabolism, particularly in B. thailandensis.Aside from self-activating mechanisms documented in B. thailandensis (e.g., quorum-sensing-driven bactobolin production) and/or sensing antibiotics and competitively responding (82), we note two major patterns in exometabolite production in the monocultures that may have contributed to the activation of secondary metabo lism in the cocultures.First, each member released and accumulated a unique set of primary metabolites over their time series.These exometabolites had relatively reduced concentrations in their coculture conditions.Second, because our experimental design included a comparative time point taken during exponential growth, we also identified a unique set of primary metabolites that had substantially accumulated by 12.5 h.Indeed, primary metabolites (83) have been documented to induce secondary metabolism in B. thailandensis.Thus, it is possible that the dynamics observed over the stationary phase could be attributed also to the uptake of exometabolites that were produced earlier in the exponential phase, or to the uptake of accumulated primary metabolites.Instead of pinpointing single molecule elicitors of secondary metabolism, our systems-level approach is better used to improve understanding of the environmental and ecological factors that contribute to member or community success.
C. subtsugae can inhibit B. thailandensis (33) but we did not observe B. thailanden sis inhibition based on cell counts.However, we did find that in stationary phase C. subtsugae-B.thailandensis cocultures, C. subtsugae upregulated an uncharacterized hybrid nonribosomal peptide synthetase-type I polyketide synthase.P. syringae was the least competitive of the three neighbors, as evidenced by a reduction in live cell counts when cocultured with B. thailandensis.Also, P. syringae did not increase compet itive strategies when cocultured, as no BSGCs were upregulated across all coculture conditions.In summary, though all three neighbors had a potential to use competitive strategies and maintained competitive strategies in monoculture (36), B. thailandensis was most successful in cocultures over the stationary phase through increased produc tion of exometabolites involved in interference and exploitative competition strategies.
Given the upregulation of BSGCs in B. thailandensis and the strong transcriptional responses of C. subtsugae and P. syringae to the presence of B. thailandensis, we hypothesized that competitive exometabolites were contributing to their commun ity dynamics.Thus, we used a co-expression network analysis with our longitudinal transcriptome series to infer interspecies interactions (84).The use of this approach was first demonstrated to infer coregulation between a phototroph-heterotroph commen sal pair (85).Our network confirmed that B. thailandensis BSGCs had coordinated gene expression patterns with both C. subtsugae and P. syringae.Interspecies nodes in both networks contained various genes involved in the upregulated B. thailanden sis BSGCs.We focused on interspecies edges within thailandamide nodes for the B. thailandensis-C.subtsugae network and interspecies edges within malleilactone nodes for the B. thailandensis-P.syringae network because these were significantly enriched as interspecies nodes.A C. subtsugae gene of interest, CLV04_2968, was contained within the thailandamide cluster of interspecies nodes.This gene codes for a DNA starvation/stationary phase protection protein and had the highest homology to the Dps protein in Escherichia coli across all C. subtsugae protein-coding genes.Dps mediates tolerance to multiple stressors and dps knockouts are more susceptible to thermal, oxidative, antibiotic, iron toxicity, osmotic, and starvation stressors (86).Interestingly, CLV04_2968 was downregulated when cocultured with B. thailandensis, suggesting that B. thailandensis attenuates C. subtsugae stress tolerance over the stationary phase.While we observed a slight decrease in viable C. subtsugae cells when cocultured with B. thailandensis, one may expect C. subtsugae to have increased sensitivity to subsequent stress (e.g., pH stress; 87) resulting from CLV04_2968 downregulation in the presence of B. thailandensis.
In the B. thailandensis-P.syringae co-expression network, a P. syringae gene of interest, PSPTO_1206, was contained within the malleilactone cluster of interspecies nodes.PSPTO_1206 is annotated as a TonB-dependent siderophore receptor.A P. syringae iron-acquisition receptor had coordinated expression with malleilactone, which has been characterized as a siderophore with antimicrobial properties (62).Interestingly, this gene was downregulated when in coculture with B. thailandensis.By contrast, the closest TonB-dependent siderophore receptor homolog to PSPTO_1206 in B. thailan densis, BTH_I2415, was upregulated coculture conditions with P. syringae.To summa rize, co-expression network analysis revealed interspecies coordinated gene expression patterns.Though determining directionality was beyond the scope of this analysis, we observed B. thailandensis-increased competition strategies were coordinated with a potential decrease in competition strategies in C. subtsugae via reduced stress tolerance and in P. syringae with reduced iron acquisition ability.
One feature of our study is that we adjusted glucose concentration depending on plate occupancy.Glucose concentration increased as membership increased but a member consistently occupied 31 wells across all experimental conditions.One could argue that resource concentration contributed to differences in transcripts and exometabolites and not interspecies interactions.However, DEGs were present when comparing pairwise coculture conditions and these were attributed to differences in temporal regulation of COG categories (Fig. S7).More specifically, regarding BSGCs, an unidentified NRPS was upregulated in B. thailandensis when cocultured with P. syringae but not when cocultured with C. subtsugae (Fig. 6) and, an unidentified NRPS-Type I polyketide synthase was upregulated in C. subtsugae when cocultured with B. thailanden sis but not when cocultured with P. syringae (Fig. S11).These differences occurred in experimental conditions where the glucose concentration was the same.Furthermore, we performed additional SynCom experiments where we increased the plate occupancy of B. thailandensis in monoculture while subsequently increasing resource concentration.Decreased gene expression was observed across all three RT-qPCR tested thailandamide genes as both plate occupancy and resource concentration increased in B. thailan densis monocultures.These same three genes had increased gene expression across all cocultures.These findings show that some undefined exometabolite interspecies interactions were responsible for the increased expression of a BSGC in B. thailandensis.Overall, we acknowledge that resource concentration and exometabolite output are intertwined, and subsequent work could test how initial resource availability determines SynCom outcomes.
A major goal in microbial ecology is to predict community dynamics for purposes of modulating and/or maintaining ecosystem function (88,89).At its core, microbial functional properties emerge, in part, from the concerted interactions of multi-spe cies assemblages.The SynCom system provides a tractable experimental system to understand the relationships between exometabolite interactions and environmental stimuli to inform higher-order community interactions.Higher-order interactions are those that are unexpected based on interactions observed in simpler situations (e.g., of member pairs) (90)(91)(92).Therefore, integrating different system variables, like transcrip tome and metabolome dynamics, within controlled microbial communities will inform how unexpected phenomena arise and how they contribute to deviations in predictive models of community outcomes.
Our results indicated that competition strategies were maintained despite stag nant population growth.B. thailandensis upregulated various bioactive exometabolites involved in both interference and exploitative competition when with neighbors.An effective competitor is often defined as its ability to outcompete neighbors via growth advantage that stems from efficient nutrient uptake and/or biomass conversion rates (93,94).We add to this that a competitor can also have a fitness advantage through effective maintenance, which can similarly employ interference or exploitative com petitive strategies despite no net growth.Maintenance may ensure survival in some environments that impose a stationary phase lifestyle, where long periods of nutrient depletion are punctuated with short periods of nutrient flux.In these scenarios, it warrants to understand how competitive strategies are deployed in the interim of growth and the extent to which these interactions contribute to long-term community outcomes.Though population levels remain constant, sub-populations of growing cells have been observed in the stationary phase (95), and continued production of compet itive exometabolites may serve as an advantageous strategy to hinder the growth of competitors.In addition, some antibiotics remain effective in non-replicating bacteria (96).The ability for continued maintenance via effective competition strategies during the stationary phase may provide spatiotemporal maintenance of population levels before growth resumption (97).Alternatively, both growth and non-growth strategies may be occurring simultaneously (e.g., as can occur in biofilms).The heterogeneity of biofilms may provide an environment where a bacterial population contains both stationary cells in the center of the colony with growing cells at the periphery of the colony that compete and alter the developmental patterns of neighboring populations (98,99).Thus, we expect that insights into the long-term consequences of competition for microbial community outcomes will be gained by considering competition in both active growth and maintenance scenarios.

FIG 1
FIG 1 Experimental design and destructive sampling procedure of transwell plates.There were seven conditions, six time points/conditions, and four independent replicates/conditions (168 total transwell plates).Each member occupied 31 wells/condition to maintain member-specific population density across all conditions (A).The SynCom transwell plate maintains the physical separation of members in individual wells while permitting exometabolite exchange through a 0.22-μm-pore filter bottom.Exometabolite exchange occurs via a bottom-fitted shared medium reservoir (B; (35)).Six replicate transwell plates were prepared for a time-course experiment.The time-course experiment included one exponential phase time point and five stationary phase time points.At specified time points, a transwell plate was destructively sampled (C).Note that all members were diluted to different starting ODs to allow for all members to achieve stationary phase within a two-hour window of each other.This figure was created with BioRender.com.

FIG 2
FIG 2 Loss of cell viability in B. thailandensis cocultures.Live (green) and dead (blue) flow cytometry cell counts for C. subtsugae (Top row, panels A-C) and P. syringae (Bottom row, panels D-F) from Syto9-and propidium iodide-stained cells (n = 4 to 5 technical replicates/time point/community membership/transwell plate and n = 4 independent replicates/time point/community membership).Cell counts are from monocultures (panels A and D), cocultures with P. syringae (panel B) or C. subtsugae (panel E), and cocultures with B. thailandensis (panels C and F).The bottom and top of the box are the first (Q1) and third (Q3) quartiles, respectively, and the line inside the box is the median.The whiskers extend from their respective hinges to the largest value (top), and the smallest value (bottom) was no further away than 1.5 × the interquartile range.Points represent outliers that are less than 1.5 x the interquartile range of Q1 or greater than 1.5 x the interquartile range of Q3.

FIG 3
FIG 3 Transcriptional responses are driven by community membership and time.Shown are principal coordinate analysis (PCoA) plots for B. thailandensis (A), C. subtsugae (B), and P. syringae (C).Each PCoA sub-panel presents the time series of transcriptional patterns of the focal member given each of its four growth conditions (one monoculture condition, two pairs, and one three-member).Each point represents a mean transcript profile for a community member given a particular condition (indicated by symbol color) and sampled at a given time point over exponential and stationary phases (in hours since inoculation, h, indicated by symbol size, n = 3 to 4 replicates per time point/community membership).The Bray-Curtis distance metric was used to calculate dissimilarities between transcript profiles.Error bars are one standard deviation around the mean axis scores.Note that transcriptional responses are driven by community membership on PCoA axis 1 and time on PCoA axis 2 across all plots.

FIG 4
FIG 4 Bacterial community exometabolite profiles differ by community membership and time.Shown are PCoA plots for exometabolite profiles from the following mass spectrometry modes: polar positive (A), polar negative (B), nonpolar positive (C), and nonpolar negative (D).Each point represents the mean exometabolite profile (relative contributions by peak area) given a particular community membership (indicated by symbol color) at a particular time point (indicated by symbol size).The Bray-Curtis distance metric was used to calculate dissimilarities between exometabolite profiles.Error bars are 1 standard deviation around the mean axis scores (n = 2 to 4 replicates).Bt is B. thailandensis, Cs is C. subtsugae, and Ps is P. syringae.

FIG 5 FIG 6 B
FIG5 Primary metabolites accumulated in monocultures have altered dynamics in cocultures.A heat map of identified, primary metabolites is shown for C. subtsugae monoculture (Cs), P. syringae monoculture (Ps), B. thailandensis monoculture (Bt), C. subtsugae-P.syringae coculture (CsPs), B. thailandensis-P.syringae coculture (BtPs), B. thailandensis-C.subtsugae coculture (BtCs), and the three-member community (BtCsPs), where samples are in columns and exometabolites are in rows.These exometabolites were filtered based on their time series accumulation in monocultures (See supplementary methods for details).Data for each sample are the averages from independent time point replicates (n = 3 to 4).Euclidean distance was calculated from Z-scored mass spectral profiles.Features with similar dynamics were clustered by Ward's method.

FIG 7
FIG 7 Coculture upregulation of BSGCs from B. thailandensis translates to temporally accumulated secondary metabolites.Columns represent community membership and rows represent identified secondary metabolites in B. thailandensis.Known bioactive secondary metabolites produced by B. thailandensis were identified in MZmine 2 through the observation of MS and MS/MS data.The accumulation of each exometabolite was quantified through time (n = 2 to 4 integrated peak areas per time point).The bottom and top of the box are the first (Q1) and third (Q3) quartiles, respectively, and the line inside the box is the median.The whiskers extend from their respective hinges to the largest value (top), and the smallest value (bottom) was no further away than 1.5× the interquartile range.Points represent outliers that are less than 1.5× the interquartile range of Q1 or greater than 1.5× the interquartile range of Q3.

TABLE 1
Bacterial members used in the synthetic community (SynCom) system