Transcriptional profiling and genetic analysis of a cystic fibrosis airway-relevant model shows asymmetric responses to growth in a polymicrobial community

ABSTRACT Bacterial infections in the lungs of persons with cystic fibrosis are typically composed of multispecies biofilm-like communities, which modulate clinically relevant phenotypes that cannot be explained in the context of a single species culture. Most analyses to date provide a picture of the transcriptional responses of individual pathogens; however, there is relatively little data describing the transcriptional landscape of clinically relevant multispecies communities. Harnessing a previously described cystic fibrosis-relevant, polymicrobial community model consisting of Pseudomonas aeruginosa, Staphylococcus aureus, Streptococcus sanguinis, and Prevotella melaninogenica, we performed an RNA-Seq analysis on the biofilm population to elucidate the transcriptional profiles of the community grown in artificial sputum medium (ASM) as compared to growth in monoculture, without mucin, and in fresh medium supplemented with tobramycin. We provide evidence that, although the transcriptional profile of P. aeruginosa is community agnostic, the transcriptomes of S. aureus and S. sanguinis are community aware. Furthermore, P. aeruginosa and P. melaninogenica are transcriptionally sensitive to the presence of mucin in ASM, whereas S. aureus and S. sanguinis largely do not alter their transcriptional profiles in the presence of mucin when grown in a community. Only P. aeruginosa shows a robust response to tobramycin. Genetic studies of mutants altered in community-specific growth provide complementary data regarding how these microbes adapt to a community context. IMPORTANCE Polymicrobial infections constitute the majority of infections in the cystic fibrosis (CF) airway, but their study has largely been neglected in a laboratory setting. Our lab previously reported a polymicrobial community that can help explain clinical outcomes in the lungs of persons with CF. Here, we obtained transcriptional profiles of the community versus monocultures to provide transcriptional information about how this model community responds to CF-related growth conditions and perturbations. Genetic studies provide complementary functional outputs to assess how the microbes adapt to life in a community.

pwCF provides an ideal environment for the growth of bacterial and fungal pathogens; these chronic infections increase inflammation, resulting in tissue damage and increased mortality rates (6,7).
Chronic infections originate early in the lives of pwCF (8).Once infected, many of the microbes that colonize the lungs of pwCF transition from an acute infection to a chronic biofilm mode of growth (9)(10)(11)(12)(13).The development of chronic infections in pwCF is associated with poor clinical outcomes, with 80%-95% of pwCF succumbing to respiratory failure associated with chronic bacterial infections and the associated airway inflammation (14).
Examination of human infections with microscopy revealed the polymicrobial nature of many infections (15)(16)(17).Despite early knowledge that many infections are associated with more than one microbe, most efforts have focused on the study and treatment of individual bacterial species (15).Research focusing on polymicrobial communities has found that when multiple bacterial genera/species are present in a community, there is a disconnect between culture-based antimicrobial susceptibility testing and clinical responses (18)(19)(20).
Recent work from our lab has attempted to reconcile our lack of knowledge regarding polymicrobial biofilms in the context of CF infections (21).Our work has focused on a new clinically informed model community comprised of a set of four CF-relevant microbes: Pseudomonas aeruginosa, Staphylococcus aureus, Streptococcus sanguinis, and Prevotella melaninogenica, grown in artificial sputum medium (ASM) under anoxic conditions, mimicking the environment found in the mucus associated with the lungs of pwCF (21)(22)(23)(24)(25)(26).
In this study, we harnessed biofilm-associated cells to develop transcriptional profiles of the four constituent members of a four-microbe community in a clinically relevant growth condition, in monoculture, in the four-microbe community, and in response to several perturbations.This data set will serve as a resource for the community.
A key finding of these analyses is the differential impact on the transcriptome for these microbes grown in monoculture versus co-culture.We also performed genetic screens showing that a microbe with little change in transcription in the community (P.aeruginosa) also yield few mutants with small effect sizes in the community context.In contrast, for two microbes showing marked changes in the community transcriptome, we identified multiple mutants, some with large effect sizes.The transcriptional and genetic studies support the conclusion that microbes display asymmetric responses to growth in a polymicrobial community.

Transcriptional profiling studies
We performed a series of transcriptional profiling studies to better understand genes differentially expressed in monoculture versus our previously reported, clinically relevant multispecies community (24).We also assessed the transcriptional profile of the community upon treatment with tobramycin in fresh medium (using fresh medium as a control), in medium lacking mucin, and for a community including a P. aeruginosa variant frequently detected in the CF airway, that is, a ΔlasR mutant.We previously reported that LasR-defective strains show an increased tolerance to tobramycin in the mixed community (24).
To this end, we obtained transcriptional profiles of each of the four constituent species growing as a biofilm, P. aeruginosa PA14, S. aureus Newman, S. sanguinis SK36, and P. melaninogenica ATCC 25845, across the different growth conditions outlined above at 24 hours (Table 1; Tables S1 to S4).Co-culturing these four microbes in ASM supplemented with mucin provides a means to mimic the nutritional, environmental, and community composition common in the CF airway (23)(24)(25)(26), providing an ideal in vitro model to examine the transcriptional landscape of these four microbes in CF-rele vant growth conditions.
The 24 hour time point was used to assess the established polymicrobial bio film community.These conditions reflect the assay conditions used in our previous report (24).Furthermore, we believe that by assessing the transcriptional profile of an established 24 hour biofilm, we are more closely reflecting the chronic communities found in the lungs of persons with cystic fibrosis, as these are thought to be very slow growing or stationary phase communities (27)(28)(29)(30).Additionally, we have previously shown that the relative abundance of microbes in our system falls within what is detected in patient samples (24).We anticipated that the tobramycin treatment (and the associated fresh medium control) would alter the transcriptional activity of the community, but we would also argue that the addition of fresh medium could potentially mimic the transcriptional changes the bacteria undergo upon colonization of a new lung environment.In future studies, as we better understand the system, we plan to perform a more comprehensive time course study.

Response of the microbes to growth in the community
To begin analyzing our data, each genus' RNA profiles in each of the indicated growth conditions were used to produce a series of principal component analyses (PCA) plots (Fig. 1A through E).Each point shown in the PCA plots represents the first two principal components for an individual biological replicate for the indicated condition, each with four replicates.We observed that P. aeruginosa and the ΔlasR mutant monocultures (Fig. 1A; Mono and ΔlasR, respectively) clustered with their community-grown counterparts (Fig. 1A; Mix and Mix_ΔlasR, respectively).In contrast, the S. aureus monoculture (Fig. 1B; Mono, pink) and S. sanguinis monoculture (Fig. 1C; Mono, pink) were distinct from their corresponding community-associated counterparts (Mix, red).As previously reported, Prevotella does not grow either planktonically or as a biofilm in the model in monocul ture (24), so there is no "Prevotella monoculture" condition (Fig. 1D).The key for Fig. 1A through D is shown in Fig. 1E.
We performed a permutational multivariate analysis of variance (PERMANOVA) on the data associated with each of the PCA plots to determine if the groupings were significantly different from each other.Pairwise tests demonstrated that only groupings within the P. aeruginosa PCA were classified as significantly different from each other.Specifically, while there was no significant difference for the wild type (WT) or the lasR mutant in monoculture or the mixed community, or the WT or lasR communities exposed to tobramycin, we did observe a significant difference when mucin was removed from the medium.For all other PCA plots shown in Fig. 1, while we do see some significant changes in gene expression in response to perturbations, the overall transcriptional profiles are not statistically significant by PERMANOVA.
As a first step, the transcriptional data were compared using the mixed community (Mix) growth condition as a baseline to compare to the monoculture condition.P. aeruginosa WT monocultures were found to have 19 downregulated genes and 135 upregulated genes when compared to the polymicrobial biofilm (Table S1).Of the 135 genes upregulated in the P. aeruginosa WT monospecies biofilms, many of these genes are in pathways and/or operons implicated in biofilm formation and quorum sensing (Table S1).S. aureus monocultures showed 50 downregulated genes and 159 upregulated genes when compared to S. aureus grown in the polymicrobial biofilm (Table S2).Of the 50 downregulated genes, a large portion are implicated in metabolism and amino acid synthesis, while of the 159 upregulated genes, there is an enrichment for genes involved in S. aureus quorum sensing and thymidine metabolism (Table S2).S. sanguinis monocultures showed 240 downregulated genes and 211 upregulated genes when compared to S. sanguinis grown in the polymicrobial biofilm (Table S3).The 240 downregulated genes were enriched for pathways involved in amino acid synthe sis and D-xylulose-5-P synthesis, while the 211 upregulated genes were enriched for pathways involved in catabolism and metabolism (Table S3).
Although the S. aureus, S. sanguinis, and P. aeruginosa cultures appear to have a similar number of differentially regulated genes, the proportion of differentially regulated genes, in relation to genome size and number of genes is less by a factor of two for P. aeruginosa compared to either S. aureus or S. sanguinis.Additionally, a closer examina tion of the magnitude of differential gene expression between the monocultures and co-cultures were lower for P. aeruginosa than for S. aureus or S. sanguinis.P. aeruginosa monocultures differed from the multispecies cultures by a log 2 FC of −2.87-5.32, the magnitude of differential gene expression of S. aureus monocultures differed from the multispecies cultures by a log 2 FC of −7.55-9.38 and S. sanguinis cultures differed by a log 2 FC of −11.10-7.56.
To further compare the transcriptional differences between the various conditions used in this study, we produced a distance matrix comparing the mixed community (Mix) to the mean point for each of the other conditions displayed in each PCA plot (Fig. 1F).This distance matrix reinforces the finding that P. aeruginosa (WT and the ΔlasR mutant) shows a modest shift in overall transcriptional profile when grown in monoculture versus mixed culture.Thus, it appears that P. aeruginosa is somewhat agnostic, in terms of gene expression, to monoculture growth conditions versus growth in the polymicrobial community.In contrast, the relative shift in transcriptional profiles for S. aureus and S. sanguinis in monoculture is much more striking when compared to growth in the mixed community (Fig. 1F).

Impact of fresh medium and tobramycin on transcriptional profiles
In a previous study, we found that the polymicrobial biofilm community conferred tolerance to tobramycin for biofilm-grown S. aureus and S. sanguinis compared to their planktonic counterparts (24).For S. aureus, growth in the mixed community conferred a modest (1-log) higher viability when exposed to tobramycin, while for S. sanguinis, growth in a mixed community resulted in 8-log greater viability than in the pure culture when both were exposed to tobramycin.In contrast, there was a statistical increase in the viability of the S. sanguinis, and no change to S. aureus when grown in mixed community when compared to the monoculture in the absence of tobramycin (24).Furthermore, community-grown P. melaninogenica biofilms were unaltered by the presence or absence of tobramycin in the mixed community (Table S4) (24).Unexpectedly, our previous work found a decrease in viability of biofilm-grown P. aeruginosa in the mixed community following the addition of tobramycin; this phenotype was reversed for the P. aeruginosa ΔlasR mutant strain (24).We explored the impact of the lasR mutation below.
To better understand the basis of these changes in tolerance for microbes grown in the absence versus presence of the other community members (24), we assessed the transcriptional profiles of each of the constituent members of the four-microbe community upon the addition of fresh medium with tobramycin or fresh medium alone as the control.To assess the acute effects of addition of tobramycin, we harvested RNA 15 minutes after antibiotic (or medium control) exposure.We hypothesized that a 15 minutes incubation would be sufficient to alter the transcriptional profiles without a significant loss in bacterial viability, which we confirmed experimentally (Fig. S1).The relatively short turnover time of mRNAs in bacteria (31,32) supports our ability to investigate the acute effects of exposure to tobramycin.
We analyzed the response to acute tobramycin exposure versus the medium control in two ways.First, we generated a series of heatmaps to visualize the 50 most variably expressed genes (Fig. 2) across all growth conditions in each tested species (Table 1).The list of these 50 genes is shown in Table S5.The differentially expressed genes for the tobramycin exposure (Mix_tob, in brown) and the medium control (Mix_swap, in yellow, because the medium was "swapped") formed a tight cluster for WT P. aeruginosa and for the ΔlasR mutant, discrete from all other conditions-all other conditions were grown for 24 hours without the addition of fresh medium (Fig. 2A).The tight association of the profiles of the top differentially expressed genes for the tobramycin exposure (Mix_tob) and the medium control (Mix_swap) are also reflected in the tight clustering of the respective samples in the P. aeruginosa PCA plot (Fig. 1A).
To further illustrate the influence of fresh medium +/− tobramycin on the mixed communities, a series of Venn diagrams were created to show the numbers of genes differentially expressed in these conditions compared to the mixed cultures at 24 hours (Fig. 3).To generate the gene lists used to produce the Venn diagrams (Tables S6 to S9), the RNA-Seq data were filtered to select only genes with a |log 2 FC| > 1.5 and an adjusted P-value < 0.05, then comparing samples exposed to only fresh medium to those exposed to fresh medium plus 100 µg/mL tobramycin.
Each of the Venn diagrams in Fig. 3 consist of 15 different regions.For P. aeruginosa, this analysis revealed a total of 47 upregulated and 93 downregulated genes associated with addition of fresh medium, and an additional 193 upregulated and 107 downregula ted genes in response to tobramycin treatment (Table S6).These differentially regulated genes were dominated by genes implicated in various metabolic pathways.Thus, P. aeruginosa showed a robust transcriptional response to the addition of tobramycin independent of the effect of adding fresh medium to the mixed community (Fig. 1F).
The three other constituent members of the polymicrobial community showed a minimal response to the presence of tobramycin in the heatmap analysis of the top 50 genes, but their response to the addition of fresh medium was robust (Fig. 1B through F and 2B through D; Tables S7 to S9).This observation is best illustrated in Fig. 2B through D by the co-clustering of samples exposed to fresh medium and fresh medium plus tobramycin for the WT and ΔlasR mutant communities.In contrast, for P. aeruginosa, as detailed above, the samples from cultures exposed to fresh medium formed a discrete cluster from those cultures exposed to fresh medium plus tobramycin (Fig. 2A).The lack of a transcriptional response for S. aureus and P. melaninogenica in response to the antibiotic is consistent with our previous report that these microbes do not show a robust change in the viability of biofilm-grown bacteria in monoculture versus a mixed community when exposed to tobramycin (24).
Interestingly, despite its lack of transcriptional response to tobramycin treatment, S. sanguinis biofilms treated with tobramycin do show a large and significant shift in viability when grown as a monoculture as compared to a mixed community (24).This change in behavior may be due to non-transcriptional mechanisms and/or response(s) to tobramycin that occur after the 15-minute exposure used in this study, or may be due to an unknown polymicrobial interaction protecting the S. sanguinis from tobramycin.The differential expression of genes in P. aeruginosa but not the other three species in the polymicrobial community upon the addition of tobramycin to the cultures is consistent with the modest efficacy of tobramycin in removal of Pseudomonas from the lungs of pwCF, however, does not provide insight into reducing the remaining community members.
The transcriptional response to the exposure to fresh medium, independent of the exposure to tobramycin, for S. aureus, S. sanguinis, and P. melaninogenica, is illustrated by genes in the 7 different P. melaninogenica growth conditions.Variation in gene expression was calculated with the rowVars function in the matrixStats package in R. The mean expression across all growth conditions was calculated, and the difference between the individual growth condition and the mean was utilized to produce a heatmap depicting the most differentially expressed genes across our various growth conditions through the use of the pheatmap package in R. The gene lists used to generate these diagrams can be found in Table S5. the separation of the data points for these three microbes in the PCA plots in Fig. 1, Venn diagrams in Fig. 3; Tables S7 through S9.As expected, the addition of fresh ASM, whether with or without tobramycin, drastically altered the transcriptional profiles of each species in the community (33)(34)(35).An analysis of the transcriptional responses of each of these species upon the addition of fresh medium failed to uncover any specific pathways.That is, many of the differentially regulated genes belonged to the "carbon metabolism" or "microbial metabolism in diverse environments" pathways as assessed by KEGG pathway analysis.

Impact of the lasR mutation on transcriptional profiles
As mentioned above, we previously showed a substantial decrease in viability of biofilm-grown P. aeruginosa in the mixed community following the addition of tobramy cin; this phenotype was reversed for the P. aeruginosa lasR mutant strain (24).Therefore, we included the ΔlasR mutant in several conditions: in monoculture, in the mixed community, and in the mixed community exposed to fresh medium +/− tobramycin.
For P. aeruginosa, as expected, the top differentially expressed genes for the WT and the ΔlasR mutant formed discreet clusters (Fig. 2A), whether grown in monoculture or in the mixed community for 24 hours or exposed to fresh medium +/− tobramycin.
In contrast, S. aureus mixed cultures or the cultures exposed to fresh medium +/− tobramycin were unaffected by the P. aeruginosa ΔlasR mutant, which can be observed in the PCA plot in Fig. 1B and by the overlap of the most variably expressed genes in Fig. 2B.Similarly, S. sanguinis appears to be agnostic to the presence of a P. aeruginosa ΔlasR mutant as the transcriptional profile of this microbe in the mixed community with WT P. aeruginosa and mixed community with the ΔlasR mutant are co-mingled in PCA plots (Fig. 1C) and by clustering (Fig. 2C).Finally, the ΔlasR mutant does not appear to impact the transcriptome of P. melaninogenica, as the profiles of this anaerobe grown with the WT and ΔlasR mutant are intermixed in the PCA plot (Fig. 1D) and by clustering (Fig. 2D).Overall, while loss of LasR function impacts the transcriptome of P. aeruginosa, as expected (36)(37)(38), mutating this quorum sensing system has a minimal impact on the other microbes in the community in the time frame used in our study.

Impact of mucin on transcriptional profiles
The polymicrobial biofilms of both P. aeruginosa and P. melaninogenica are impacted by the presence or absence of mucin, while the S. aureus and S. sanguinis cultures show a minimal response to the presence/absence of mucin.Previous work demonstrated that bacteria like P. melaninogenica play a large role in mucin degradation (39,40).Even though S. sanguinis has previously been shown to play a role in mucin degradation (39,41,42), we did not observe a transcriptional response indicating differential gene regulation when comparing Mix_base (no mucin) growth condition to the Mix growth condition (i.e., with mucin added) at 24 hrs.
As mentioned above, mucin degradation is associated with fermentative anaerobes such as P. melaninogenica (39).The fermentation of mucin by P. melaninogenica generates acetate and propionate, on which P. aeruginosa is partially dependent for growth in bacterial communities that mimic the lower airway of pwCF (39).In this work, the downregulation of prpR, which codes for the propionate catabolism regulator, in P. aeruginosa grown in the absence of mucin points to the bioavailability of propionate in the plot, genes had to be statistically differentially expressed (P < 0.05) and have a log 2 fold change greater than or equal to the absolute value of 1.5.To provide a more informative figure, genes that were upregulated and downregulated in comparison to the Mix biofilm condition were separated into separate Venn diagrams, left and right, respectively.Further analysis focused on the intersection of all data sets to elucidate the influence of fresh medium, and the intersection of solely the tobramycin conditions to elucidate the influence of tobramycin on the biofilm population.The gene lists used to generate these diagrams can be found in Tables S6 through S9. produced by P. melaninogenica in the presence of mucin.Likewise, a decrease in P. melaninogenica HMRPREF0695_A6215 transcription, coding for acetate kinase, which is responsible for the production of acetate and propionate, in the Mix_base (−mucin) community when compared to the Mix community (+mucin) points to a change in the metabolic profile of P. melaninogenica when grown in the mixed community in the absence of mucin.

Summary of transcriptional responses
To further explore the transcriptional variability in each of the strains in the community, and across each growth condition, an upset plot was created for each bacterial genus to compare the number of co-regulated genes across each of the growth conditions (Table 1) using the mixed community (Mix) as the comparison condition (Fig. 4).The large number of co-regulated genes that were differentially regulated in similar conditions across each of the upset plots demonstrates the robustness of the data set.To visualize the changes in how the transcriptional profiles changed across the various growth conditions, upregulated and downregulated genes were separated to better understand how the transcriptional profiles of each organism was affected by the various growth conditions tested herein.
Figure 4 demonstrates the roles that mucin and fresh medium have on the communi ties.With the exception of S. sanguinis, the lack of mucin produced the most differentially regulated genes in each of the other species across all of the treatments tested.The complexity of the mucin may play a role in secondary metabolism in each of these other microbial species.Since the human host produces 22 different mucins, it is also possible that the composition of the available mucin could play a role in the metabolism and interspecies interactions between each of the bacteria in this polymicrobial model (43,44).Fresh medium, as described above, also plays a large role in the transcriptional profiles of each member of the tested polymicrobial communities.The large number of differentially regulated genes in the samples containing fresh medium support the ability of these bacteria to rapidly alter their transcription in response to the presence of fresh nutrients.

How genetic studies correlate with the transcriptional analyses
To complement the transcriptional analysis presented here, we performed genetic screens to ascertain the genes that, when mutated, would reduce survival in the community compared to monoculture, and at what magnitude (Fig. 5).Genetic screens were performed with P. aeruginosa, S. sanguinis, and S. aureus mutant libraries (45)(46)(47), and we compared viability of each mutant (via CFUs) when grown in a polymicrobial community versus in monoculture.
For the full P. aeruginosa transposon library (~4,500 mutants in non-essential genes), only five mutants were isolated; all of these strains resulted in a modest ~1 log loss in CFU in the polymicrobial community as compared to when the mutant was grown in monoculture.
Following the same procedures, screening the whole S. sanguinis transposon library (~2,100 mutants) yielded 34 genes with reduced CFU when mutants were grown in a mixed community compared to a monoculture.When compared with the five genes that restricted the growth of P. aeruginosa in the community, these 34 genes produced a significantly larger log-fold change (Fig. 5A; right side), varying from 1-log to 3.5-log fold change in viability.Interestingly, screening half of the S. aureus Newman transposon library (~1,500 mutants) did not yield mutants with reduced viability when grown in the community.
Due to the lack of genetic tools in P. melaninogenica, and its inability to grow as a monoculture in ASM, we focused on P. aeruginosa mutants associated with reduced growth of P. melaninogenica in co-culture compared to co-culture with WT P. aeruginosa.The genetic screen of~4,500 mutants from the P. aeruginosa mutant library identified 15 P. aeruginosa mutants that impacted the growth of P. melaninogenica (Fig. 5B).Thirteen of the P. aeruginosa mutations provided ~1-log fold change in the growth of P. melaninogenica, while two mutants abolished the ability of P. melaninogenica to grow in co-culture conditions.All of the transposon mutants identified here grew as well as their respective WT strain in monoculture.
Together, the genetic analysis largely mirrors the transcriptional study, that is, P. aeruginosa, which is agnostic to community growth yielded few mutants with small effects on viability in the community.In contrast, the screens involving S. sanguinis and P. melaninogenica, which showed a more robust transcriptional response to growth in the community, also yielded more mutants with a larger range of phenotypes.The one exception was S. aureus, which despite its robust transcriptional responses across the various conditions analyzed here, failed to yield any mutants that showed a defect specific to the community, perhaps suggesting that S. aureus is well suited to compete sanguinis (right) when grown in the four-microbe community compared to monoculture.(B) Mutants of P. aeruginosa that altered the growth of P. melaninogenica when grown as a co-culture.To be included, mutants had to decrease the colony forming units (CFUs) of the strain compared to growth in the monoculture(A) or when compared to P. melaninogenica grown in the presence of WT P. aeruginosa by ≥1 log.All mutants here showed no defect in monoculture growth (not shown).Statistics were performed using a t-test though the ggsignif package in R. ****P < 0.0001.against other microbes in a community via redundant systems not accessible to genetic screens.

Conclusions
Our findings indicate that in the model polymicrobial community, the transcriptional profile of P. aeruginosa remains relatively unchanged by the presence of the other members of this community, a finding reflected by the genetic analysis.That is, both experimental approaches used here reiterate the capacity of P. aeruginosa to thrive in the presence of other microbes across a range of environments.In contrast, the transcrip tional profiles of the other members of the community are influenced by each other and environmental perturbations.Again, the transcriptional analysis is largely reflected by our genetic studies, with the exception of S. aureus.
While a great deal of work has focused on P. aeruginosa monospecies biofilms, and the role of the transcriptional regulator LasR impacting the transcription of hundreds of genes in this microbe (48)(49)(50)(51)(52)(53), our findings indicate that a ΔlasR mutant of P. aeruginosa had a negligible influence on the transcriptional profiles of the other members of the community.
As expected, the addition of fresh medium, both with and without tobramycin, drastically altered the transcriptional responses of each of the members of the polymi crobial community.However, surprisingly, the addition of tobramycin only altered the acute transcriptional response of P. aeruginosa.The lack of a transcriptional response to tobramycin by the other members of this community confirms previous work performed in our lab demonstrating that upon tobramycin treatment, the viability of S. sanguinis, S. aureus, and P. melaninogenica largely remain unchanged when grown in the polymicro bial community (24).
Here, we have generated and performed an initial transcriptional analysis of a model polymicrobial community relevant to CF.We hope that providing the complete data set to the broader research community will stimulate further hypotheses and help to advance the treatment of polymicrobial infections in CF-relevant infections.Due to the complexity of many polymicrobial infections, we believe that we have established a framework to further investigate polymicrobial infections using an approach that enables hypothesis generation, while simultaneously implicating specific genes and pathways for their roles in complex microbe-microbe interactions.

Microbial growth assays
Culture experiments were performed as previously described by (24) with modifications.Briefly, cells from overnight liquid cultures of P. aeruginosa, S. aureus, S. sanguinis, and P. melaninogenica were individually collected and washed twice (for P. aeruginosa and S. aureus) or once (for S. sanguinis and P. melaninogenica) in sterile phosphate-buffered saline (PBS) by centrifuging at 10,000 × g for 2 minutes.After the final wash, cells were resuspended in the ASM base.The optical density (OD 600 ) was then measured for each bacterial suspension and diluted to an OD 600 of 0.2 in either ASM or ASM base, depending on the subsequent growth condition.Monocultures and co-cultures were prepared from the OD 600 = 0.2 suspensions and further diluted to a final OD 600 of 0.01 for each microbe in ASM or ASM base, depending on the subsequent growth condition.For each monoculture and co-culture condition, a full polystyrene flat-bottom 96-well plate was inoculated with 100 µL of bacterial suspension.Plates were incubated using an AnaeroPak-Anaerobic container with a GasPak sachet (ThermoFisher) at 37°C for 24 hours.After incubation, plates were processed and split into three separate groups.For group 1, which included monocultures and co-cultures communities grown in ASM or ASM base, unattached cells were aspirated with a multichannel pipette.Pre-formed biofilms were then washed once with sterile PBS and resuspended in 50 µL of RNAp rotect (Qiagen) using a sterile 96-pin replicator.For the remaining groups (co-culture communities with WT or ΔlasR P. aeruginosa), plates were anaerobically transferred from an AnaeroPak-Anaerobic container with a GasPak sachet to an anoxic environmental chamber (Whitley A55; Don Whitley Scientific, Victoria Works, UK) with 10% CO 2 , 10% H 2 , and 80% N 2 mixed gas at 37°C.Unattached cells of co-culture communities grown in ASM were aspirated with a multichannel pipette and 100 µL of either fresh ASM (group 2) or ASM supplemented with 100 µg/mL tobramycin (group 3) was added to the pre-formed biofilms for 15 minutes.After incubation, the supernatants were removed and the biofilms were washed once with sterile PBS.For all groups, a volume of 50 µL of RNAprotect was added to the biofilms and collected using a sterile 96-pin replicator.For all conditions, a control plate of the monocultures and co-cultures was 10-fold serially diluted and plated on selective media, as previously reported (24), as a control to confirm the expected viable counts of bacteria grown under each condition.We confirmed that CFU/mL of the mono-and co-cultures used in the transcriptional assays were consistent with the values typically associated with this experiment (data not shown).

RNA purification, sequencing, and analysis
Following growth and collection of biofilms in RNAprotect, bacterial lysis and RNA extraction were performed following the protocols outlined in the RNeasy Mini Kit (Qiagen).Because of S. aureus in the mixed communities, cells lysis conditions included a combination of lysozyme (800 µg/mL) and lysostaphin (12.5 µg/mL).The concentra tion and quality of the RNA was checked on a Nanodrop 2000 (ThermoFisher) prior to sending the samples for library preparation and sequencing (SEQCENTER).Briefly, samples were DNase treated (Invitrogen) prior to library preparation using the Illu mina Stranded Total RNA Prep Ligation with Ribo-Zero Plus Kit and 10 bp IDT for Illumina indices.Sequencing was performed on a NextSeq 2000 (2 × 51 bp reads).SEQCENTER performed demultiplexing, quality control, and adapter trimming with bcl-convert (v3.9.3) (54).Upon obtaining the data, samples were subsequently analyzed with Cutadapt (55) to a phred score of 20, a minimum length of 10 bp and using Cutadapt's nextseq-trim function set at 10. Sequencing reads were mapped to the P. aeruginosa PA14 genome (NC_008463.1), S. aureus Newman genome (NC_009641.1), S. sanguinis SK36 genome (NC_009009.1), or the P. melaninogenica ATCC 25845 genome (NC_014370.1 and NC_014371.1)using bowtie2 (56) in paired end read using default parameters.The sam files were converted to sorted.samfiles using samtools (57) prior to using featureCounts (58) with default parameters to calculate reads per gene.The differential expression of genes in each condition was determined through the use of DEseq2 (59).Read counts were normalized using the estimateSizeFactors function.Genes with read counts under 10 were dropped from our data set.The biological coefficient of variance was estimated, and a linear model was fit to determine the differential expression of genes between samples.Our false discovery rate was adjusted from P-values using the Bonferroni correction (60).To compare the transcriptional profiles between samples, our data sets were further filtered to ensure that only genes with a differential expression by a log 2 fold change (logFC) threshold greater than the absolute value of 1.5 and an adjusted P-value less than 0.05 were used in our downstream analyses.PCA plots were produced in base R, while heatmaps utilized the ComplexHeat map (61) and pheatmap (62) R packages.Venn diagrams and upset plots were produced with the VennDiagram (63) and ggupset (64) R packages, respectively.

Genetic screens
For the genetic screens, we utilized non-redundant transposon mutant libraries for each of P. aeruginosa PA14, S. sanguinis SK36, or S. aureus Newman (45)(46)(47).Overnight liquid cultures of mutant libraries were grown in polystyrene flat-bottom 96-well plates with each well containing a different mutant growing in 100 µL of TSB for P. aeruginosa, THY for S. sanguinis, and TSB + YE for S. aureus Newman.The 96-well plates containing monoculture conditions and co-culture conditions in ASM with S. aureus, S. sanguinis, and P. melananogenica were prepared using methods described above and inoculated with 1-2 μL of the P. aeruginosa PA14 (45), S. sanguinis (46), or S. aureus Newman (47) mutant libraries using a 96-pin replicator.
For all screens, plates were incubated using an AnaeroPak-Anaerobic container with a GasPak sachet (ThermoFisher) at 37°C for 24 hours.After incubation, unattached cells were aspirated with a multichannel pipette.Pre-formed biofilms were then washed once with sterile PBS.Cells were then resuspended in 50 µL PBS and spotted onto the appropriate selective medium for each microbe using a sterile 96-pin replicator, as previously reported (24).Mutants were assayed for growth in both monoculture and mixed cultures.Mutants unable to grow to WT levels in the monoculture growth condition were discarded from the genetic screen to focus on mutations that impacted growth specifically in the context of the polymicrobial community.In the screen, we used a 1-log cutoff as the minimum cutoff for selecting candidate mutants in the original screen as we did not feel the assay could reliably detect smaller effect sizes.
For confirmation, each candidate was grown in triplicate in the monoculture and mixed conditions as described above.After incubation, all cultures were 10-fold serially diluted and plated on PIA for P. aeruginosa, SSA from S. sanguinis, TSB for S. aureus, or TSB + YE for P. melaninogenica to quantify cell counts by CFUs, as reported.

Statistical analyses
All statistical analyses performed here were done in R. PCA were performed on normalized and log transformed count data obtained from a DESeq2 analysis performed on the read counts for each of the four RNA-Seq replicates analyzed in this paper.
The adonis2 function from the package Vegan (65) was utilized to compare the overall data set, and found a significant difference between each strain (P < 0.001 for all four species tested).To examine pairwise groupings between the various strain groupings in the PCA plots (Fig. 1A through D), we utilized the pairwise.adonisfunction from the package Vegan (65,66).Analyses were performed using the PC1 and PC2 data for each respective species, and comparisons made between each of the respective growth conditions.Both the adonis 2 and pairwise.adonisanalyses utilized Euclidean distances as the "method, " or "sim.method"variable, respectively, to remain consistent with the Euclidean distances that were calculated to compare groupings in Fig. 1F.
To compare the P. aeruginosa and S. sanguinis screen data, P-values were calculated through the use of the wilcox.testfunction and the ggsignif package in R (67).Due to the difference in how the screen data were generated for P. aeruginosa, S. aureus, S. sanguinis, and the P. melaninogenica samples, statistical analysis was not done to compare the P. melaninogenica data to the P. aeruginosa and S. sanguinis data.

FIG 1
FIG 1 Variation in transcriptional profiles of monoculture versus the four-microbe community (Mix) in different culture conditions.(A) PCA plot of P. aeruginosa transcriptional profiles across all culture conditions containing P. aeruginosa.(B) PCA plot of S. aureus transcriptional profiles across all culture conditions containing S. aureus.(C) PCA plot of S. sanguinis transcriptional profiles across all culture conditions containing S. sanguinis.(D) PCA plot of P. melaninogenica transcriptional profiles across all culture conditions containing P. melaninogenica.The gene lists used to generate the diagrams in panels A-D can be found in Tables S1 to S4. (E) Legend for the PCA plots in panels A-D.(F) A distance matrix comparing the mixed community (Mixed) to the mean point for each of the other conditions (see Table1) displayed in each PCA plot.Values were calculated via Euclidean distance between means and are included in the respective value ) displayed in each PCA plot.Values were calculated via Euclidean distance between means and are included in the respective value marker present with each distance heatmap.

FIG 2 (
FIG 2 (Continued) differentially expressed genes in the 8 different S. aureus growth conditions.(C) The top 50 differentially regulated genes in the 8 different S. sanguinis growth conditions.(D) The top 50 differentially regulated

7 FIG 3
FIG3 Genes differentially expressed from the mixed community (Mix) growth condition.Venn Diagrams were created to depict the number of genes differentially regulated in each of the growth conditions when compared to the mixed community grown in ASM (Mix) for P. aeruginosa (A), S. aureus (B), S. sanguinis (C), and P. melaninogenica (D).To be included (Continued on next page)

FIG 4
FIG 4Upset plots to depict the number of differentially regulated genes.Upset plots were created to depict the number of genes differentially regulated in each of the growth conditions for both upregulated and downregulated genes when compared to the mixed community grown in ASM (Mix).All genes used in the production of this plot were statistically differentially expressed (P < 0.05) with a log 2 fold change greater than or equal to the absolute value of 1.5.

FIG 5
FIG5 Fold change of mutants identified in genetic screens.Genetic screens using non-redundant transposon libraries were used to screen for mutations that reduced the growth of microbes in the polymicrobial community versus monoculture.(A) Mutants that impacted the growth of P. aeruginosa (left) and S.

TABLE 1
Labels for the strains and growth conditions used in the transcriptional studies