Coexistence of Microbial Species in Structured Communities by Forming a Hawk-Dove Game Like Interactive Relationship

Microorganisms evolve kinds of elaborate interaction models that can form relatively stable communities in a wide range of ecosystems. It is recognized that the spatial genetic structure of microbes in surface-attached environments lays a good foundation for the persistence of polymicrobial communities in adverse conditions. However, the interacting dynamics of microbes in facilitating the formation and stabilization of community structure still remains elusive. In this study, we identify a hawk-dove game like interspecific relationship between the two Gram-negative opportunistic pathogens Pseudomonas aeruginosa and Klebsiella pneumoniae, which naturally coexist in insect gut and can cocolonize human tissues. Specifically, although P. aeruginosa had significant competitive advantage over cocultured K. pneumoniae on solid medium with rich nutrient factors, K. pneumoniae could resist the suppression of P. aeruginosa by enhancing the expression of membrane transporters induced by the extracellular metabolites of P. aeruginosa. By contrast, under the condition that K. pneumoniae had a growth advantage but P. aeruginosa met a metabolic burden in producing quorum-sensing-controlled extracellular products, the frequency of K. pneumoniae would be slightly higher than P. aeruginosa during the coexistence because K. pneumoniae was also capable of exploiting the extracellular metabolite from P. aeruginosa. In addition, P. aeruginosa quorum-sensing variant could reap benefits from K. pneumoniae in turn and reach a relatively stable two species equilibrium. These findings provide an explanation for the formation and maintenance of polymicrobial communities in different spatially structured environments, and thus may contribute to understanding the complex interspecific interactions of microbes in local communities and shed new light on the development of social microbiology.


INTRODUCTION
Microorganisms are ubiquitous inhabitants of natural and unnatural ecosystems, and growth of them in a sessile lifestyle inevitably form communities composed of homo-/heterogeneous species with complex ecological interacting networks (Griffin et al., 2004;Adnani et al., 2017;Tshikantwa et al., 2018). Microbial species in symbiotic communities evolve kinds of chemical or physical interaction patterns such as metabolitemediated cooperation/conflict and space competition, and these interactions can be beneficial, neutral or harmful to the fitness of individuals in the community. Accordingly, the outcomes of microbial interspecific interactions can be mutualistic, commensalistic or parasitic, by which may integrate the local species to adhere to material surfaces or persistently cocolonize host tissues (Xavier, 2011;Hussa and Goodrich-Blair, 2013;Short et al., 2014;Scherlach and Hertweck, 2018).
Surface-attached polymicrobial communities are frequently densely packed, and this leads to the formation of spatial genetic structure and promotes the evolution of cooperation in digesting macromolecular compound using extracellular enzymes (Chao and Levin, 1981;Greig and Travisano, 2003;Nadell et al., 2009;Elias and Banin, 2012;Frost et al., 2018;Solden et al., 2018). Nutrient limitation as a hallmark of dense communities can result in social competition for extracellular products that significantly benefit the producers and neighboring recipients. This creates selection for individuals with deficiency in producing the key factor but can reap the benefits of producers' investment (Hansen et al., 2007;Xavier et al., 2011;Zengler and Zaramela, 2018). Theoretically, such kind of exploitive competition may occur in many aspects involving extracellular digestive proteases, antibiotic resistance enzymes and iron chelate, biofilm production, virulence determinants and so on (Diggle et al., 2007a;Brockhurst et al., 2008Brockhurst et al., , 2010Gore et al., 2009;Diard et al., 2013;van Gestel et al., 2014;Kelsic et al., 2015;Luján et al., 2015;Frost et al., 2018).
The classic "hawk-dove" game describes a coexisting mechanism of higher animal species in sharing the same ecological niche. In the payoff matrix, the aggressive hawk can easily defeat the mild dove and gets the full resource V once encountered, while dove gets 0. When two hawks meet, each of them has a 50% chance to win or to loss the fight and will pay a cost C for fighting, thus the average outcome is V/2 -C/2. When two doves meet, they share the resource and get V/2. Because C is generally greater than V in the natural world, then V > V/2 > 0 > V/2 -C/2, thus coexistence of the two species is evolutionarily stable and the proportion of them depends on the probability of meeting (Maynard-Smith and Price, 1973;Auger et al., 1998). By contrast, accumulating theoretical and empirical evidence suggest that spatially structured environments can effectively promote the biodiversity of microbial communities by enhancing the local feedbacks to balance the proportions of individuals with different fitness (Turner and Chao, 1999;Kerr et al., 2002;Sachs et al., 2004;Ross-Gillespie et al., 2007;Santos et al., 2008;Pande et al., 2016). On the basis of these findings, it seems that the concept of hawk-dove game may also apply to microbial interspecific interaction, especially in spatially structured environments. Hence, we test this hypothesis by using the two Gram-negative pathogens Pseudomonas aeruginosa and Klebsiella pneumoniae, which naturally coexist in the gut of silk moth and can be coisolated from pneumonia and burn patients (Liu et al., 2004;Li et al., 2007;Anand et al., 2010;Kelvin Lee et al., 2016;Acosta et al., 2017), as experimental models.
The interspecific interaction and coexisting mechanisms of P. aeruginosa and K. pneumoniae in spatially structured communities were explored by coculturing them on solid media from different ratios. The distribution and proportion of the two species in mixed communities were determined by fluorescence imaging and CFU counting. A series of RNA-sequencing were further performed to pinpoint their transcriptional changes under different coculture conditions. The results showed that although P. aeruginosa generally had significant competitive advantage, the enhanced membranechannel-related substance circulation of K. pneumoniae induced by the extracellular metabolites of P. aeruginosa enabled the invasion of K. pneumoniae toward P. aeruginosa. On the other side, when the two species were cocultured under the condition that P. aeruginosa had a metabolic burden, they could still invade each other's population and coexist in the communities by forming a hawk-dove game like interactive relationship.

Bacterial Strains and Media
Strains of mCherry-labeled wild type P. aeruginosa PAO1 (WT PAO1), mCherry-labeled P. aeruginosa PAO1 isogenic lasR mutant (in-frame knockout) and green fluorescent protein (GFP)-labeled K. pneumoniae were previously preserved in the laboratory (Zhao et al., 2018). All the strains were routinely cultured in Luria-Bertani (LB) broth from a single colony.

On-Plate Competition Experiments
Overnight cultured bacterial strains in LB broth were harvested and washed with a sterile saline solution (autoclaved 20 min at 127 • C) for three times, and then adjusted to optical density at 600 nm (OD 600 ) of 1.0. WT PAO1 or lasR mutant strains were well-mixed with K. pneumoniae at different ratios (1:99, 1:1, and 99:1) in 1.0 ml of sterile saline solution. Subsequently, 5 µl of inocula were separately spotted on disposable sterile petri dishes containing 20 ml of LB agar medium or modified M9 agar medium supplemented with 0.5% (w/v) of skim milk powder (Zhao et al., 2019). The plates were sealed by sealing films and then statically cultured at 37 • C for different time periods. These experiments were independently repeated for six times. The growth status and distribution of bacterial cells were assessed by checking fluorescence using an IVIS XRII imaging system (Calpier, Inc.). Moreover, 1:1 mixed P. aeruginosa and K. pneumoniae were also stabbed (not reach the bottom) into LB plates containing 0.3% of agar using sterile toothpick or inoculated (2 µl of inocula) on the surface of LB plates containing 0.6% of agar to determine the swimming and swarming motilities of mixed communities, respectively. Pure cultures of P. aeruginosa and K. pneumoniae under the same conditions were set as controls.

Competitivity Identification
At each sampling point of the competition experiments above, the colonies were gently scraped out from the plates by using autoclaved bamboo sticks with flat-shaped shovel at one end and dispersed in 1.0 ml of sterile saline solution, and then the bacterial cells were washed for three times. A fraction of appropriately diluted bacterial suspension was spread on M9-milk plate and cultured overnight (17 h) for species discrimination. WT PAO1 would form large colony size with clear proteolytic halo around the colony, P. aeruginosa lasR mutant would form relatively small colony without proteolytic halo, and K. pneumonia would form large colony size with no or faint proteolytic halo. The strains were further confirmed by IVIS XRII imaging to check the fluorescent colors or by colony PCR using specific primer pair 5 -CTTCATCGTCGGCAACTAC-3 and 5 -GTCTGGTAGATGGACGGTTC-3 to amplify the partial lasR gene of P. aeruginosa. Positive PCR-amplicon could only be obtained from WT PAO1. After calculating the proportion of each species in the mixed colonies at each sampling points, the relative fitness (v) of one species was then calculated by comparing the initial (x 0 ) and final frequencies (x 1 ) using the modified equation v = log10[x 1 (1-x 0 )/x 0 (1-x 1 )] based on previous description (Diggle et al., 2007a).

Small Metabolites Mediated Interspecific Competition
Equal amounts (10 µl of bacterial suspension with OD 600 of 1.0) of P. aeruginosa and K. pneumoniae were cocultured in disposable sterile 12-well polystyrene plate equipped with an inner well-containing a 0.4 µm permeable polycarbonate membrane (Corning, NY, United States) in the bottom of the inner well (Supplementary Figure 1). The two species were separated by the wall of the inner well, but could exchange the small extracellular metabolites through the membrane layer. The outer space of the well was supplied with 3.0 ml of LB broth and the inner space was supplied with 0.75 ml of LB broth. Coculture of 1:1 well-mixed P. aeruginosa and K. pneumoniae and pure cultures of the two species in normal 12-well polystyrene plate without membrane containing 4.0 ml of LB broth were set as control. All the capped plates were sealed by sealing films and statically cultured at 37 • C for 3 days.

RNA-Sequencing and Transcriptomic Analysis
Eight groups of experiments were carried out for RNAsequencing (RNA-seq): (1) coculture (1:1) of P. aeruginosa and K. pneumoniae on LB plates; (2) pure culture of P. aeruginosa on LB plates; (3) pure culture of K. pneumoniae on LB plates; (4) coculture (1:1) of P. aeruginosa and K. pneumoniae in conventional 12-well plates containing LB broth; (5) coculture of P. aeruginosa (inoculated in the outer space) and K. pneumoniae (inoculated in the inner space) in LB broth in the 12-well plate equipped with an inner well (Supplementary Figure 1); (6) coculture of K. pneumoniae (inoculated in the outer space) and P. aeruginosa (inoculated in the inner space) in LB broth in the 12-well plate equipped with an inner well; (7) pure culture of P. aeruginosa in conventional 12-well plates containing LB broth; (8) pure culture of K. pneumoniae in conventional 12well plates containing LB broth. All the cultures were statically incubated at 37 • C for 3 days, and bacterial cells of each culture were then harvested for total RNA isolation using TRIzol reagents (Invitrogen). All the experiments were independently repeated for three times. RNA samples from three parallel cultures of each experiment above were well-mixed and conducted for library construction and RNA-seq using prokaryotic strand-specific Illumina-based RNA-seq technology (Novogene Bioinformatics Technology, Co., Ltd., China). The generated sequencing data are deposited in the NCBI database under the Accession No. PRJNA504726.
The software Bowtie 2 (Langmead and Salzberg, 2012) was used to map the clean reads from coculture of P. aeruginosa and K. pneumoniae on LB plate (Experiment 1), pure culture of P. aeruginosa on LB plate (Experiment 2), coculture of P. aeruginosa and K. pneumoniae in LB broth (Experiment 4), P. aeruginosa cultured in the outer well (Experiment 5), and pure culture of P. aeruginosa in LB broth (Experiment 7) to the reference genome of P. aeruginosa 1 , respectively. Similarly, the clean reads from coculture of P. aeruginosa and K. pneumoniae on LB plate (Experiment 1), pure culture of K. pneumoniae on LB plate (Experiment 3), coculture of P. aeruginosa and K. pneumoniae in LB broth (Experiment 4), K. pneumoniae cultured in the outer well (Experiment 6), and pure culture of K. pneumoniae in LB broth (Experiment 8) to the reference genome of K. pneumoniae 2 , respectively. The software package Cufflinks (Trapnell et al., 2012) and DEGseq (Wang et al., 2010) were conducted to get transcriptome assembly and calculate the values of gene expression using expected fragments per kilobase of transcript per million fragments (FPKM). Differential gene expression was determined by comparing the FPKM values to those of monocultured corresponding species under the same growth condition (on plate or in liquid), respectively. For example, comparison of Experiment 1 to Experiment 2 indicates the profiling of differentially expressed genes of P. aeruginosa when P. aeruginosa was cocultured with K. pneumoniae on LB plate. The significance (P-value) of differentially expressed gene was corrected by calculating the false discovery rate Q, and Q < 0.05 was thought to be significantly different. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway prediction and protein classification analysis were performed by using KOBAS v2.0 (Mao et al., 2005).

Statistical Analyses
Data analysis and statistical tests were performed by using the software GraphPad Prism version 7.0 (San Diego, CA, United States). Mean values of standard deviation were compared by using two-tailed unpaired t-test.

P. aeruginosa and K. pneumoniae Can Coexist in Structured Communities
To gain a general insight on the interactive relationship of P. aeruginosa and K. pneumoniae during static growth, we first grew the two species in different combinations on LB plates to monitor their frequencies. LB is a rich medium containing abundant resources which enables the normal growth of P. aeruginosa and K. pneumoniae. Our prior work showed that no significant growth difference was detected between monocultured P. aeruginosa and K. pneumoniae on LB plates, while P. aeruginosa had a competitive advantage over K. pneumoniae under coculture condition (Zhao et al., 2018). In this study, we found that in comparison to the moderately increased proportion of P. aeruginosa in the coculture with K. pneumoniae from an initial ratio of 1:1, either P. aeruginosa or K. pneumoniae from low initial proportion (1%) in the mixture could rapidly invade each other's population (Figures 1A,B). In addition, the relative fitness of P. aeruginosa was always higher than K. pneumoniae under the same starting conditions after 2 days (Figure 1C), indicating that the invasion of P. aeruginosa toward K. pneumoniae was easier than the invasion of K. pneumoniae toward P. aeruginosa when they FIGURE 1 | Competition of Pseudomonas aeruginosa (PA) and Klebsiella pneumoniae (KP) on LB plates. (A) Growth status and fluorescence detection of singly cultured P. aeruginosa or K. pneumoniae (first row) or coculture of them from different initial ratios. Red color, P. aeruginosa. Green color, K. pneumoniae. (B) Frequency of P. aeruginosa when cocultured with K. pneumoniae for different time periods. (C) Relative fitness of P. aeruginosa or K. pneumoniae when they were cocultured from different ratios. Data shown are mean values of standard deviation (n = 6). * P < 0.05, * * P < 0.01, * * * P < 0.001. Two-tailed unpaired t-test.
Frontiers in Microbiology | www.frontiersin.org were cocultured on LB plate. Therefore, the unexpected mutual invasion relationship between P. aeruginosa and K. pneumoniae was similar with the outcome of hawk-dove game. This result was different from our previous finding that P. aeruginosa could overcome cocultured K. pneumoniae, while K. pneumoniae failed to invade P. aeruginosa in liquid culture with shaking (Zhao et al., 2018). We reasoned that this divergence could be due to the spatial distribution difference of bacterial cells between shaking and static culture conditions, since the structured population frequently led to coexistence of individuals with different relative fitness (Kerr et al., 2002;Ross-Gillespie et al., 2007;Santos et al., 2008;Pande et al., 2016).
Compared with the uniformly distributed P. aeruginosa in the communities, K. pneumoniae was excluded in the middle parts of colonies that grew from 99:1 or 1:1 mixture of P. aeruginosa and K. pneumoniae for 4 days (Supplementary Figure 2). As we had reported previously (Zhao et al., 2018), the peripheral distribution of K. pneumoniae was caused by the H1-T6SS (type VI secretion system) elicited cytotoxicity of P. aeruginosa, which could kill K. pneumoniae after 4 days of coculture. However, no inhibitory effect was observed when P. aeruginosa was cocultured with K. pneumoniae from an initial ratio of 1:99 (Supplementary Figure 2). This result indicated that in addition to the direct cell-cell contact with competitor cells as a necessary condition (Hood et al., 2010), high cell-density of P. aeruginosa might be a sufficient condition for the activation of P. aeruginosa T6SS during further interspecific competition. Moreover, we also noticed that P. aeruginosa had intrinsically stronger swimming ability than K. pneumoniae, and the mixed communities composed of both species showed remarkably enhanced swimming motility, instead of swarming, compared to the pure cultures (Supplementary Figure 3). Our previous work also found that the flagellum-encoding genes of P. aeruginosa were up-regulated during the coevolution with K. pneumoniae in liquid medium (Zhao et al., 2018). Therefore, these findings indicated that the presence of K. pneumoniae in the communities might improve the motility of P. aeruginosa.

Enhanced Substance Circulation May
Contribute to the Resistance of K. pneumoniae Against P. aeruginosa Suppression We were interested to explore why P. aeruginosa and K. pneumoniae could invade each other by detecting the transcriptional changes of the two species under coculture condition using RNA-seq. To prevent the potential interference of T6SS-related cell toxicity, the colonies of mixed (1:1) and single species cultured on LB plates for 3 days were harvested for RNA-seq ( Figure 1A). The transcriptional patterns of pure P. aeruginosa and K. pneumoniae were set as controls to determine the transcriptional changes of the two species under coculture condition.
The result of comparative-transcriptomic analysis revealed that only eight significantly up-regulated genes and 147 downregulated genes with no specific enrichment in any pathways were detected in P. aeruginosa when cocultured with K. pneumoniae (Supplementary Dataset 1). By contrast, a large scale of transcriptional change including 1734 significantly up-regulated genes enriched in ABC transporters (137 out of 274 genes, P = 9.10E-08) and 1116 down-regulated genes enriched in carbon metabolism (64 out of 126 genes, P = 0.034) and oxidative phosphorylation (31 out of 47 genes, P = 0.034) were detected in K. pneumoniae (Figure 2 and Supplementary Dataset 2). Additionally, when P. aeruginosa and K. pneumoniae were statically cocultured in LB broth, 55 significantly up-regulated genes enriched in aminobenzoate degradation (anthranilate biosynthesis, 3 out of 7 genes, P = 0.048) and 6 downregulated genes with no specific enriched pathways were detected in P. aeruginosa (Supplementary Dataset 3). Similarly, 1334 significantly up-regulated genes enriched in ABC transporters (119 out of 274 genes, P = 3.01E-08) and benzoate degradation (20 out of 33 genes, P = 0.021) and 1220 down-regulated genes enriched in ribosome (46 out of 72 genes, P = 0.012), oxidative phosphorylation (33 out of 47 genes, P = 0.013), and carbon metabolism (66 out of 126 genes, P = 0.013) were detected in K. pneumoniae (Supplementary Dataset 4). These data suggested that K. pneumoniae was more susceptible to the presence of P. aeruginosa, and the enhanced membrane-channelrelated substance circulation might contribute to the proportion increase of K. pneumoniae in the coculture with P. aeruginosa.

Extracellular Metabolites Dominate the Competition Between P. aeruginosa and K. pneumoniae
We further tested the possibility of extracellular-factormediated interspecific interaction between P. aeruginosa and K. pneumoniae by coculturing them in 12-well polystyrene plate containing a 0.4 µm permeable polycarbonate membrane  Figure 1). In this way, growth of bacterial cells in outer well might be influenced by the secreted small metabolites from the species of inner well. When P. aeruginosa was cultured in outer well, the result of RNA-seq showed that 118 significantly up-regulated genes enriched in flagellar assembly (6 out of 35 genes, P = 0.005) and glycine, serine and threonine metabolism (6 out of 49 genes, P = 0.014), and 29 downregulated genes with no specific enriched pathways were detected compared to monocultured P. aeruginosa (Figures 3A,B  and Supplementary Dataset 5). The increased expression of flagellum-encoding genes under this condition further demonstrated that the swimming motility of P. aeruginosa might be enhanced by the stimulation of unknown extracellular metabolites from K. pneumoniae. Additionally, the expression of the pyoverdine biosynthesis operon, anthranilate biosynthesis operon and stress response related genes (such as htpG, sigX, ibpA, groES, and hslV) of P. aeruginosa were also significantly increased. These transcriptional changes were accordant to previous findings that, iron-competition was more frequent among interspecific bacterial species, and the expression of ironchelators pyoverdine and Pseudomonas quinolone signal (PQS, the product of anthranilate metabolic pathway) were sensitive to the level of environmental iron. Moreover, PQS was capable of modulating various phenotypes including cell motility, biofilm production, and virulence, and could also influence the growth of other bacterial competitors such as Staphylococcus aureus and Escherichia coli (Mashburn et al., 2005;Diggle et al., 2007b;Tashiro et al., 2013;Khare and Tavazoie, 2015).

(Supplementary
By contrast, when K. pneumoniae was cultured in outer well, 1379 significantly up-regulated genes enriched in ABC transporters (124 out of 274 genes, P = 1.60E-09) and phenylalanine metabolism (20 out of 35 genes, P = 0.038), and 1032 down-regulated genes enriched in oxidative phosphorylation (35 out of 47 genes, P = 0.001), ribosome (43 out of 72 genes, P = 0.003), carbon metabolism (59 out of 126 genes, P = 0.016), RNA degradation (15 out of 16 genes, P = 0.018), and citrate cycle (21 out of 33 genes, P = 0.047) were detected compared to monocultured K. pneumoniae (Figures 3C,D and Supplementary Dataset 6). These results combined with the up-regulated anthranilate biosynthesis operon of P. aeruginosa (Supplementary Dataset 5), indicating that the PQS of P. aeruginosa might suppress the growth and oxygen consumption rate of cultured K. pneumoniae by inhibiting the respiratory chain, like those happened during the interaction of P. aeruginosa with E. coli or P. putida (Toyofuku et al., 2010).
Finally, as summarized in Table 1, the small metabolitesdirected transcriptional changes of P. aeruginosa and K. pneumoniae generally covered those cocultured on solid medium and in liquid condition. This result revealed that the interspecific competition between P. aeruginosa and K. pneumoniae were mainly dominated by extracellular products, and also confirmed the competitive advantage of P. aeruginosa and the role of K. pneumoniae membranetransport systems in resisting the survival pressure imposed by P. aeruginosa (Khare and Tavazoie, 2015). Taken together, although the mutual invasion relationship of P. aeruginosa and K. pneumoniae identified in the current study was similar to the hawk-dove game, their interior interacting dynamics might be different. Unlike the intense intraspecific competition of hawk and its absolute advantage over dove, the invasion of P. aeruginosa and K. pneumoniae toward each other should pay additional cost in producing specific factors. Moreover, the enhanced substance circulation of K. pneumoniae identified in our current study might also provide a plausible explanation for the increased tolerance of mixed population against sodium dodecylsulfate stress (Kelvin Lee et al., 2016). Further more detailed study concerning the roles of toxic factors and membrane transporters would significantly contribute to understanding the formation and stabilization mechanisms of polymicrobial communities in diverse environments.

Extracellular Products Mediated
Coexistence of P. aeruginosa and K. pneumoniae To investigate the extracellular product-based interspecific interaction between P. aeruginosa and K. pneumoniae, another batch of on-plate competition assay was performed by using M9 medium containing sodium caseinate or skim milk powder as the sole carbon source. Previous studies reported that P. aeruginosa can grow in M9-casein (or milk) by synthesizing quorum-sensing (QS) system-controlled costly extracellular product elastase (Wilder et al., 2011;Darch et al., 2012). K. pneumoniae is also capable of producing this proteolytic enzyme, albeit its encoding gene still remained unknown (Trishin et al., 2004(Trishin et al., , 2005. Indeed, both P. aeruginosa and K. pneumoniae can grow on M9-milk plates, and the growth rate of P. aeruginosa was significantly slower than K. pneumoniae (Supplementary Figure 4). However, compared with the apparent proteolytic ring formed around P. aeruginosa colony, there was only a faint halo around K. pneumoniae colony. These results combined with the equal growth rates of P. aeruginosa and K. pneumoniae on rich medium (Zhao et al., 2018), indicating that the production of elastase through energy-intensive QS-regulation might burden the growth of P. aeruginosa on M9-milk plate.
In contrast to the competitive advantage of P. aeruginosa over K. pneumoniae on LB plates (Figure 1), the proportion of P. aeruginosa decreased gradually when cocultured with K. pneumoniae on M9-milk plate from an initial ratio of 1:1, while either P. aeruginosa or K. pneumoniae from low proportion (1%) in the mixture could still invade each other (Figures 4A,B).
The results of relative fitness analyses further revealed that K. pneumoniae always grew faster than P. aeruginosa when they were cocultured on M9-milk plates from any initial ratios ( Figure 4C). These findings suggested that K. pneumoniae had a competitive advantage over P. aeruginosa when cocultured on M9-milk plates, and the invasion of K. pneumoniae toward P. aeruginosa was easier than the invasion of P. aeruginosa toward K. pneumoniae. Additionally, the proportion increase of P. aeruginosa cultured from low proportion (1%) on M9-milk plates was significantly delayed compared with that on LB plates, and also significantly slower than K. pneumoniae under the same given condition (Supplementary Figure 5). Intriguingly, K. pneumoniae also emerged in the region of proteolytic ring formed by P. aeruginosa (Supplementary Figure 6), indicating FIGURE 4 | Competition of P. aeruginosa and K. pneumoniae on M9-milk plates. (A) Growth status and fluorescence detection of singly cultured P. aeruginosa or K. pneumoniae (first row) or coculture of them from different initial ratios. Red color, P. aeruginosa. Green color, K. pneumoniae. (B) Frequency of P. aeruginosa when cocultured with K. pneumoniae for different time periods. (C) Relative fitness of P. aeruginosa or K. pneumoniae when they were cocultured from different ratios. Data shown are mean values of standard deviation (n = 6). * P < 0.05, * * P < 0.01, * * * P < 0.001. Two-tailed unpaired t-test. that K. pneumoniae might benefit from the extracellular products of P. aeruginosa. Therefore, our finding here revealed that although the development of QS system endowed P. aeruginosa with an advantage in resource competition (Zhao et al., 2018), K. pneumoniae was capable of overcoming P. aeruginosa depending on the low metabolic cost or by stealing the extracellular elastase produced by P. aeruginosa. Furthermore, no T6SS-related inhibitory effect of P. aeruginosa on K. pneumoniae like that on LB plate was observed even when they were cocultured for 4 days (Supplementary  Figure 7). This could be explained by the finding that the activation of H1-T6SS was negatively regulated by the central QS regulator LasR (Lesic et al., 2009), and thus the H1-T6SSmediated killing of P. aeruginosa on K. pneumoniae would not happen under the condition requiring an active LasR regulon for growth. P. aeruginosa QS-Deficiency Contributes to the Coexistence With K. pneumoniae Evolution of P. aeruginosa under QS-required condition in vitro or in cystic fibrosis lungs frequently selects lasR mutants with impaired QS-regulation to reduce the cost in synthesizing extracellular factors (Sandoz et al., 2007;Köhler et al., 2009;Smith and Chapman, 2010;Zhao et al., 2016). To further investigate the role of P. aeruginosa QS system in extracellular productmediated competition with K. pneumoniae, the lasR mutant strain of P. aeruginosa with deficient elastase production ability FIGURE 5 | Competition of P. aeruginosa lasR mutant (lasR) and K. pneumoniae (KP) on M9-milk plates. (A) Growth status and fluorescence detection of singly cultured P. aeruginosa lasR mutant or K. pneumoniae (first row) or coculture of them from different initial ratios. Red color, P. aeruginosa lasR mutant. Green color, K. pneumoniae. (B) Frequency of P. aeruginosa lasR mutant when cocultured with K. pneumoniae for different time periods. (C) Relative fitness of P. aeruginosa lasR mutant or K. pneumoniae when they were cocultured from different ratios. Data shown are mean values of standard deviation (n = 6). * P < 0.05, * * P < 0.01, * * * P < 0.001. Two-tailed unpaired t-test.
was used to repeat the competition assay with K. pneumoniae on M9-milk plates. P. aeruginosa lasR mutant had been defined as a typical cheater which could invade QS-intact cooperators by stealing the extracellular products in the population (Sandoz et al., 2007;Wilder et al., 2011). We found that although the growth of P. aeruginosa lasR mutants at high cell-densities was poor, they could also produce less elastase during prolonged culture through an uncharacterized RhlR-dependent signaling ( Figure 5A, Dandekar and Greenberg, 2013). The results of coculture assays showed that either P. aeruginosa lasR mutant or K. pneumoniae could readily invade each other's population, and no significant proportion changes and relative fitness of P. aeruginosa lasR mutant and K. pneumoniae were detected when they were cocultured from an initial ration of 1:1 (Figure 5 and Supplementary  Figure 8). Moreover, the proportion increase of lasR mutant was significantly faster than WT PAO1 when competing with K. pneumoniae from a low initial proportion of 1% for 3 days (Supplementary Figure 9). Therefore, our data here suggested that when the QS-mediated metabolic burden or competitive advantage of P. aeruginosa was removed, K. pneumoniae could invade P. aeruginosa by the innate growth advantage on M9-milk medium, and P. aeruginosa lasR mutant had the ability to invade K. pneumoniae in turn by reaping the limited elastase produced by K. pneumoniae. Collectively, although the competitive advantages of P. aeruginosa and K. pneumonias were alterable in dependence of population structure and nutrient conditions, mutual cheating and metabolic advantage might jointly contribute to the hawk-dove game like coexisting of P. aeruginosa and K. pneumoniae in spatially structured communities.

CONCLUSION
This study provides evidence that P. aeruginosa and K. pneumoniae have the capacities to invade each other and can coexist in structured communities by forming a hawk-dove game like interactive relationship. These findings highlight the importance of spatially structured environments in the formation and stabilization of polymicrobial communities, open an avenue for further understanding the complex interspecific interactions of microbes during coevolution, and may also have general implications on the development of social microbiology and novel clinical therapies.

AUTHOR CONTRIBUTIONS
KZ designed the study and wrote the manuscript. KZ, JL, TH, YY, and JfL performed the experiments. KZ and BY coordinated the bioinformatics analysis. XW and YC coordinated the laboratory management and materials collection.