Condensins are essential for Pseudomonas aeruginosa corneal virulence through their control of phenotypic programs

Pseudomonas aeruginosa is a significant opportunistic pathogen responsible for a variety of human infections. Its high pathogenicity resides in a diverse array of virulence factors and an ability to adapt to hostile environments. We report that these factors are tied to the activity of condensins, SMC and MksBEF, which primarily function in structural chromosome maintenance. This study revealed that both proteins are required for P. aeruginosa virulence during corneal infection. The reduction in virulence was traced to broad changes in gene expression. Transcriptional signatures of smc and mksB mutants were largely dissimilar and non-additive, with the double mutant displaying a distinct gene expression profile. Affected regulons included those responsible for lifestyle control, primary metabolism, surface adhesion and biofilm growth, iron and sulfur assimilation, and denitrification. Additionally, numerous virulence factors were affected, including type 3 and type 6 secretion systems, hemagglutinin, pyocin and macroglobulin production, and a host of virulence regulators. in vitro properties of condensin mutants mirrored their transcriptional profiles. MksB-deficient cells were impaired in pyocyanin, c-di-GMP production, and sessile growth whereas smc mutants mildly upregulated c-di-GMP, secreted fewer proteases and were growth deficient under nutrient-limiting conditions. Moreover, condensin mutants displayed an abnormal regulation upon transition to stationary phase. These data reveal that condensins are integrated into the control of multiple genetic programs related to epigenetic and virulent behavior, establishing condensins as an essential factor in P. aeruginosa ocular infections. Author Summary Bacterial pathogenicity is a complex phenomenon dependent on the ability of a bacterium to thrive in a hostile environment while combating the host using an array of virulence factors. This study reports that pathogenicity is also tied to structural chromosome maintenance through condensins, proteins that are responsible for the global organization of the chromosome. We show that the two Pseudomonas aeruginosa condensins, SMC and MksB, act as global regulators of gene expression. The inactivation of SMC and MksB induces opposite regulatory programs in the cell that resemble those observed during the acute and chronic phases of infection. A substantial portion of this regulation is mediated by the intracellular signaling network of P. aeruginosa. Accordingly, virulence regulation is altered in condensin mutants. The results were validated by genetic, phenotypic and virulence studies of condensin mutants. Overall, these data establish condensins as an essential factor during ocular P. aeruginosa infections revealing their involvement in the regulatory virulence network and the control of the bacterial lifestyle.

Introduction genes being oppositely affected in two of the mutants. Interestingly, most of differentially 159 upregulated genes were unique to each mutant ( Fig 2B). Among downregulated genes, a large 160 overlap was observed between mksB and mksB smc mutants but not with the smc strain.

161
Principle component analysis (PCA) further highlighted this relationship revealing a virtually 162 equidistant arrangement of the three strains in PCA coordinates (Fig. 2C). Together, this reveals 163 predominately unique transcriptional signatures for single and double condensin mutants with a 164 distinct overlap between mksB and mksB smc cells for downregulated genes.

166
Differentially regulated pathways are dominated by lifestyle and virulence 167 We next analyzed gene expression using hierarchical clustering. Linkage maps were 168 constructed based on expression changes across all three mutants followed by ranking of the 169 genes according to the separation distance (Fig. S1A). The number of significant clusters in the 170 data set was estimated using the elbow method ( Fig S1B). Most of the genes in significantly 171 affected clusters were downregulated in mksB and smc mksB cells but were mixed in smc 172 mutants (Fig. S1A).

173
Ten out of the 29 statistically identified clusters consisted of artificially separated outliers 174 and were merged with functionally similar groups of genes. Overall, we recognized 18 distinct 175 groups of differentially affected genes (Fig. 3A). One third of the genes (31%) in the clusters 176 were of unknown function or associated with various aspects of primary metabolism (8%; Fig.   177 3B). For the remaining genes, the response was dominated by lifestyle and virulence pathways.

178
Sessile growth. Consistent with the previous physiological analysis (10), genes involved 179 in biofilm formation were prominently present among affected pathways (18 genes, mostly in 180 cluster C11; Table S1). Of these, two pel and nine psl exo-polysaccharide genes, a major 181 component of biofilms, were identified. Interestingly, a gene encoding hemagglutinin (PA0041; 182 cluster C16) was greatly affected for two condensin mutants. Hemagglutinin is a large 183 filamentous adhesion protein known for its ability to cause red blood cell agglutination and also 184 directly contributes to biofilm formation (25,26).

185
Iron and sulfur are critical microelements required for bacterial growth, especially during 186 competitive growth or infection. Besides being key components of the respiratory chain, iron 187 also serves as a signal for biofilm development (27) whereas sulfur is required for cysteine and 188 methionine biosynthesis. P. aeruginosa harbors multiple iron acquisition pathways, many of 8 189 which altered their expression, either upward or downward, in condensin mutants. 38 genes 190 from iron metabolism were found predominantly in clusters C7, C11, and C16. In particular, all 191 pyochelin biosynthesis genes were significantly expressed in cluster 7 (11 genes). Fifteen 192 pyoverdine genes were found significantly expressed across several clusters. Seven of these 193 genes were grouped together in C11 and four in C16. Pyocyanin, a major virulence factor in P.
194 aeruginosa which also acts in iron acquisition (28)

228
DNA metabolism. Only 2 genes involved in DNA maintenance and repair, ligD and 229 PA2150, have been found within significantly affected clusters (C7). These are both related to 230 non-homologous end joining and showed downregulation in both the smc and mksB mutants.

231
The opposite would be expected were this a part of a DNA damage response.

235
suggestive of a dysregulation in gene expression rather than a concerted response. In 236 particular, the 11 genes involved in amino acid metabolism were associated with biosynthesis of 237 arginine, histidine, lysine, tyrosine, valine, leucine and isoleucine.

238
Transporters. 39 genes representing 22 ABC (ATP binding cassette) transporters were 239 scattered across several clusters (Table S1). These protein machines are typically dedicated to 240 acquisition or efflux of specific substrates. The affected transporters were associated with 241 diverse functions including import of metals, sulfur and various nutrients, osmoregulation, and 242 export of exopolysaccharides. In addition, three RND transporters, mexH, mexV, and mexW, 243 associated with pyocyanin and multi-drug efflux (40, 41) changed their expression in at least 244 one mutant greater than 2 fold (FDR < 0.1).

245
Regulators. We identified 28 transcriptional regulators in significant clusters (Table S2) 246 and 56 more were found among genes whose expression changed more than two-fold in a 247 significant manner (FDR less than 0.1) in at least one of the mutants (Worksheet Regulators in 248 Table S3). Most of them belonged to pathways that were significantly affected by the mutations.

249
In particular, 8 and 4 genes, respectively, were involved in the regulation of T3SS and T6SS.

250
BexR and a hemolysin repressor PA2463 (42) were also on this list, whereas four other genes  Table S3). Only one pathway, bacteriophage Pf1 genes, was similarly 258 affected in all three mutants. The others gave rise to three unique transcriptional profiles.

259
The smc strain was characterized by an almost complete downregulation of T3SS 260 accompanied by a broad if mild upregulation of T6SS, BexR regulon, pyocin, hemolysin and 2-261 macroglobulin production, denitrification, and sulfur assimilation. Iron assimilation was mostly 262 upregulated with a notable exception of pyochelin biosynthesis. Biofilm genes were barely 263 affected except for the mildly induced hemagglutinin (Table S1) and the psl regulon, which 264 showed a small but broad upregulation of all genes including the polysaccharide transporter 265 pslD ( Fig 4A). Taken together, these factors might explain the high surface adhesion of SMC-266 deficient cells.

272
The smc mksB cells displayed similar regulation patterns to mksB for many growth and PppA (46) and RsmA (47) whereas the T3SS expression mirrored the levels of the activator

297
Two pathways could explain the sessile/planktonic behavior of the condensin mutants.

298
The switch between these two lifestyles is often controlled by the c-di-GMP signaling network 299 (50). c-di-GMP biosynthesis occurs through the siaA/D protein system which is in turn, 300 negatively regulated by free RsmA, that is the fraction of RsmA that is not associated with its 301 inhibitor RNAs, rsmY and rsmZ (47). We found large changes in the expression of SiaA, which 302 correlated with the dominance of the sessile and planktonic phenotypes observed in condensin 303 mutants (Fig. 5). In contrast, the expression of the RsmA regulon, which includes a number of 304 lifestyle genes, showed little correlation with the propensity of condensin mutants for biofilm 305 formation.

306
In summary, expression changes in the affected regulons and their immediate regulators 307 often paralleled each other, suggesting that at least some of the changes were not caused by a 308 direct transcriptional control from condensins but rather propagated through the signaling 309 network of P. aeruginosa.

312
Transcriptomic analysis revealed many physiological deficiencies that could explain the reduced 313 pathogenicity of condensin mutants. We first examined the mutants for fitness defects. We did 314 not find notable growth abnormalities in a rich medium (Fig. 6A). In M9 medium, however, 315 growth of smc but not mksB or smc mksB cells was markedly delayed (Fig. 6B). This lag varied were added into the medium, suggesting that it is caused by multiple factors. The lag was 318 indeed caused by an absence of SMC since it was largely abolished by an extrachromosomal 319 expression of the gene (Fig. 6E).

320
To determine which amino acids contributed to the lag, we grouped them into five sets 321 according to their biosynthetic mechanisms, and omitted one of each group from the growth 322 medium. To facilitate the comparison across multiple conditions, we measured the lag and 323 growth rate for each condition as illustrated in Figure 6E. The omission of any of the amino acid 324 groups delayed the growth of smc cells; however, the absence of hydrophobic amino acids 325 (alanine, valine, isoleucine and leucine) had a particularly strong effect (Fig. 6F). In contrast, the 326 growth rate of the smc cells, but not the other mutants, was equally reduced in the absence of 327 any set of amino acids (Fig. S1A). The addition of individual sets of amino acids into the minimal 328 medium reduced the lag of all tested bacteria but did not eliminate the distinctive delay in growth 329 of smc cells (Fig. 6G, S1B). The only exception observed was for the mixture of glycine, 330 cysteine, serine and histidine, which increased the lag instead of decreasing it. The increase 331 was due to cysteine (Fig. S1C, D), presumably caused by its iron chelating activity (51). Even in 332 this respect, the fitness of SMC-deficient cells was notably reduced.

333
We did not find any increase in susceptibility of condensin deficient P. aeruginosa to 334 lysozyme or defensins -defensin and lactoferrin. However, smc but not mksB or

341
The main pigments of P. aeruginosa, pyocyanin and pyoverdine, are among the more abundant 342 of its virulence factors and frequently contribute to initial host colonization by P. aeruginosa (52-343 55). We found that production of the two pigments mirrors the other phenotypes of condensin 344 mutants (Fig. 7A). Namely, production of pyocyanin was only marginally increased, if at all, 345 upon inactivation of SMC but was virtually undetectable in mksB and smc mksB cells (Fig 7B).

346
In contrast, pyoverdine production was not significantly affected in any of the condensin mutants 347 (Fig 7C). This phenotype was found to be indeed associated with mksB after deleting the gene 348 using a degron system (Fig. 7D). To this end, the endogenous mksB was replaced with its DAS4-tagged version in a strain that lacks a ClpXP adaptor protein SspB. SspB was then 350 expressed from an externally delivered plasmid. Degradation of MksB-DAS4 using the degron 351 system abolished pyocyanin production by P. aeruginosa (Fig. 7E).

352
Unlike with the pigments, protease secretion was reduced in SMC-but not MksB-353 deficient cells (Fig. 7F). This deficiency was completely restored by extrachromosomal 354 production of SMC (Fig. 7G). This result mirrors the downregulation of T3SS observed for smc 355 mutants (Fig.4).

357
Condensins are integrated into P. aeruginosa regulatory network 358 We next determined whether condensins act independently or as part of a regulatory network.

359
To this end, we explored genetic interactions between condensins and three regulators, rsmZ,

361
whereas the third one showed only borderline changes. All three regulators were implicated in 362 the control of planktonic and sessile behavior, each via its own mechanism (29, 56, 57). As a 363 proxy for these phenotypes, we followed competitive growth of the mutants in the presence of 364 their parental PAO1 strain as well as their propensity to stick to a surface.

365
Under our assay conditions, cells devoid of the quorum sensing regulator RhlI were 366 indistinguishable from the parental strain (Fig. 8A, B). The deletion of rsmZ, a negative regulator 367 of RsmA, had no effect on the surface adhesion phenotype of any of the condensin variants 368 (Fig. 8B) but modestly impaired their competitive growth (Fig. 8A). No clear evidence of genetic 369 interaction between rsmZ and condensins was detected. Curiously, however, an interaction was 370 found for VreI, a transcriptional factor that is primarily involved in iron acquisition (29). The 371 deletion of vreI increased the biofilm producing ability of wild type cells to the levels observed for 372 SMC deficient cells but had no effect on the smc mutants (Fig. 8B). Not dissimilarly, the 373 inactivation of VreI did not change the competitive growth of the wild type cells but diminished 374 the defects caused by the smc deletion (Fig. 8A).

375
Given the starkly different functions of condensins and the VreI regulon, the genetic 376 interaction found between the two proteins suggests that they affect the lifestyle of the 377 bacterium indirectly, by exploiting the existing signaling network. To evaluate this notion, we 378 measured the levels of c-di-GMP in condensin variant cells, which is a secondary messenger 379 responsible for lifestyle switching in P. aeruginosa (50). We indeed found that the level of c-di-deficient cells but was virtually undetected in mksB and smc mksB mutants (Fig. 8C). These 382 intracellular messenger levels are fully consistent with the observed phenotypes suggesting that 383 condensin mutations induce lifestyle switching by triggering the intracellular signaling network.

384
To further evaluate this conclusion, we examined the effects of condensins on quorum 385 sensing regulation. To this end, we compared condensin-induced gene expression changes in 386 exponential and stationary cells for the BexR regulon and T3SS, which have been implicated in 387 the transcriptomic analysis (Fig. 4A). Should condensins be directly involved in the control of 388 these genes, we expected similar effects on their expression in growing and stationary cells.

389
Using qRT-PCR, we measured the gene expression levels for spcS, exoT and pcrV from T3SS with smc cells initially showed a higher bacterial load than those with the parental strain, but that 407 was significantly reduced by Day 2 and continued to decline. The higher initial pathogenicity 408 was likely caused by an upregulation of T6SS, hemagglutinin and exopolysaccharide production 409 in these cells resulting in a higher propensity for biofilm formation (10

424
Transcriptional programs imposed by condensins were largely dissimilar. A substantial 425 overlap was only found for genes downregulated in mksB and smc mksB cells, whereas 426 upregulated gene sets were unique to each of the three mutants (Fig. 2B). In all cases, the 427 transcriptional response was dominated by lifestyle and virulence programs. MksB deficient 428 cells downregulated multiple systems that promote surface adhesion, making them fit for 429 planktonic growth, while upregulating T3SS (Fig. 4). This regulatory pattern is reminiscent of

437
Many changes could not be attributed to a lifestyle switch. In particular, the altered 438 expression of metabolic genes (Fig. 3) resulted in amino acid dependency and hydrogen 439 peroxide susceptibility of smc mutants (Fig. 6). The presence of amino acids in cystic fibrosis 440 sputum was shown to increase the ability of P. aeruginosa to establish the disease (62),

441
indicating that such metabolic deficiencies are not conducive to pathogenesis. In this respect, 442 transcriptional changes caused by the deletions of condensins bear elements of dysregulation.

443
The observed interference between condensin mutations and the normal signal transduction 444 (Fig. 8)  independent experiments were performed. The limit of detection for this assay was 10 CFU.

490
To analyze gross corneal pathology, mice were anesthetized with isoflurane and eyes 491 were imaged using an operating microscope (Zeiss OPMI Lumera and Medilive MindStream;  otherwise, all growth media were supplemented with a mixture of trace ions (Table S4; (67)).

516
Deletion mutants (Table 1) were constructed using allele replacement. Approximately 517 500 bp of chromosomal fragments flanking rhlI (PA3476), rsmZ (PA3621.1) and vreI (PA0675) 518 genes were generated by PCR and then spliced together as appropriate using overlap PCR with 519 a linker 5-ATGGCGGCCGCTTAA to generate complete deletions of the genes. The resulting 520 PCR products were inserted between the XhoI and KpnI sites of pEXG2 (68) to generate 521 plasmids pEXG2-rhlI, pEXG2-rsmZ and pEXG2-vreI. These plasmids were conjugationally 522 transferred into recipient P. aeruginosa strains to create the desired in-frame deletions (17).

525
Overnight cultures were inoculated into LB, grown with aeration to an OD 600 of 0.6, washed with 526 appropriate minimal medium and then inoculated into microplates containing M9, ATM or LB 527 supplemented with 100 μg/ml of the indicated amino acids at an OD 600 of 0.1. Amino acids were 528 combined in the following groups: Group 1 (Phe, Tyr and Trp), Group 2 (Asp, Met, Thr and Lys),

529
Group 3 (His, Gly, Cys and Ser), Group 4 (Ala, Val, Ile and Leu) and Group 5 (Glu, Gln, Pro and 530 Arg). The cell growth was followed using a Tecan Spark 10M microplate reader. The lag time, 531 t lag , and growth rate, r, were determined by fitting the growth curves to the equation 532 log 10 ( 600 ) = + • ( -

533
where a is the logarithm of the initial optical density of the culture. The doubling time was then 534 determined for the best-fit growth rate as t 1/2 = log 2 (10)/r.

535
Minimal inhibitory concentrations (MIC) were measured using the two-fold serial dilution 536 method. 10 4 exponential phase cells in 100 l were dispensed into microplates and incubated at 537 37 o C for up to 24 hours with shaking at 180 rpm.

538
Production of pyocyanin and pyoverdine was quantified as previously described (69, 539 70). Cells were grown overnight in LB with shaking at 37 °C, diluted 100-fold into fresh LB 540 medium in a 24-well plate, and grown as a standing culture at 37 °C for the indicated time.

541
The cells were then pelleted, and the concentration of pyoverdine in the supernatant 542 determined from its absorbance at 405 nm (  Million reads mapped), excluding the residual rRNA gene counts (less than 1.1% of total reads).

597
Genes having zero reads in any of the samples (89 total) were removed from the analysis as