Chaperone mediated coupling of subunit availability to activation of flagellar Type III Secretion

Bacterial flagellar subunits are exported across the cell membrane by the flagellar Type III Secretion System (fT3SS), powered by the proton motive force (pmf) and a specialized ATPase that enables the flagellar export gate to utilise the pmf electric potential (ΔΨ). Export gate activation is mediated by the ATPase stalk, FliJ, but how this process is regulated to prevent wasteful dissipation of pmf in the absence of subunit cargo is not known. Here, we show that FliJ activation of the export gate is regulated by flagellar export chaperones. FliJ binds unladen chaperones and, using novel chaperone variants specifically defective for FliJ binding, we show that disruption of this interaction attenuates motility and cognate subunit export. We demonstrate in vitro that chaperones and the FlhA export gate component compete for binding to FliJ, and show in vivo that unladen chaperones, which would be present in the cell when subunit levels are low, sequester FliJ to prevent activation of the export gate and attenuate subunit export. Our data indicate a mechanism whereby chaperones couple availability of subunit cargo to pmf-driven export by the fT3SS.


Introduction
Here, we identify point mutant variants of the chaperones FlgN and FliT that are 96 specifically defective for binding to the FliJ ATPase stalk yet retain the ability to bind 97 cognate subunits and the FlhA component of the fT3SS export gate. We show in 98 vitro that the FlgN and FliT chaperones compete with FlhA for binding to FliJ, and 99 that chaperone sequestration of FliJ in vivo blocks the activating interaction between 100 FliJ and the export gate. This suggests a mechanism whereby binding of FliJ by 101 unladen chaperones acts as a signal to the export machinery that subunit cargo is 102 unavailable and, accordingly, export activity is limited to prevent wasteful dissipation 103 of the pmf. This model is supported by ribosome profiling data, which show that the 104 cytoplasmic ratios of cognate subunits to chaperones are low for FlgN and FliT, 105 relative to the ratio of flagellin to its chaperone, FliS, suggesting that levels of 106 unladen FlgN and FliT would be a sensitive proxy measure of intracellular flagellar 107 subunit levels. Our data indicate a mechanism by which chaperones modulate the 108

Isolation of chaperone variants that are specifically defective for FliJ binding 113
To investigate the function of chaperone binding to the ATPase stalk, FliJ, we sought 114 to identify variants of the FlgN and FliT chaperones that did not bind FliJ yet could 115 interact with cognate subunits and the FlhA component of the fT3SS export gate. To 116 do this, we first constructed mutant alleles of flgN encoding variants with three-117 residue deletions in the FlgN α3 helix (residues 70-102), which contains the 118 overlapping binding sites for cognate subunits and FliJ ( Fig. 1; SI Appendix, Fig. S1) [8]. Screening of the FlgN deletion variants using affinity chromatography pull-down 120 assays with GST-FlgK subunit or GST-FliJ identified a single deletion variant, 121 FlgND76-78, that could bind its cognate subunit but not the ATPase stalk (Fig. 1B). 122 Protein sequence comparisons revealed FlgN W78 to be highly conserved and 123 replacement of this residue with alanine (FlgN-W78A) severely reduced FlgN binding 124 to FliJ but did not affect binding to FlgK subunit (  Table 3). 130

131
Having identified residues in the FlgN α3 helix that, when mutated, could decouple 132 chaperone binding of cognate subunits and FliJ, we targeted conserved residues in 133 the equivalent α3 helix of FliT for site directed mutagenesis to alanine ( Fig. 1D; SI 134 Appendix, Fig. S1). To facilitate screening for FliJ binding, mutations were introduced 135 into a fliT allele encoding a truncated variant, FliT94, which lacks the α4 helix and 136 shows enhanced binding to FliJ in vitro [20]. Affinity chromatography pull-down 137 assays identified five FliT94 variants (I68A, L72A, N74A, E75A and L78A) that showed 138 severely reduced binding to GST-FliJ (Fig. 1E). Three of these FliJ non-binders -139 FliT94-I68A, FliT94-L72A and FliT94-N74A -had previously been shown to bind cognate 140 subunit FliD [20]. For this reason, and after assessing the expression and stability of 141 these three FliT94 variants in Salmonella (not shown), we chose to further investigate 142 the function of truncated FliT94-L72A and full-length FliT-L72A. 143 Binding of FliT-L72A and truncated FliT94-L72A to cognate subunit FliD was confirmed 145 using pull-down assays (GST-FliD; SI Appendix, Fig. S2). To assess binding to the 146 export gate component FlhA, pull-down assays were carried out using GST-FlhAC 147 that had been incubated with cell lysates containing FliD cognate subunit, which is 148 required for high affinity binding of FliT to FlhA, and either wild type FliT or full-length 149 FliT-L72A (Fig. 1F) [9]. The FliT-Y106A variant, which cannot bind FlhA, was included 150 as a negative control (Fig. 1F)

Attenuation of cell motility and cognate subunit export by specific disruption 156 of the chaperone-FliJ interaction 157
Having identified chaperone variants (FlgND76-78, FlgN-W78A, FliT-L72A) that were 158 specifically defective for FliJ binding yet retained the ability to bind cognate subunits 159 and the fT3SS export gate, we went on to investigate the importance of the 160 chaperone-FliJ interaction for cell motility and subunit export. To enable expression 161 of chaperones at physiological levels, Salmonella strains were engineered to carry 162 variant chaperone genes (flgND76-78, flgN-W78A or fliT-L72A) at the natural genetic 163 loci, replacing the wild type chaperone genes. As negative controls, isogenic 164 Salmonella strains in which the chaperone genes were either deleted (DflgN or DfliT) 165 or replaced with variant genes encoding export defective chaperones unable to bind 166 either cognate subunit (flgND90-100) or the FlhA export gate (fliT-Y106A) were also 167 constructed. These strains were assessed for swimming and swarming motility, and 168 for subunit export efficiency relative to wild type Salmonella (Fig. 2). 169 Cell populations producing the FlgND76-78 or FlgN-W78A variants, which cannot 171 bind FliJ, showed slightly decreased swimming (70-90%) and swarming (75-90%) 172 motility relative to wild type Salmonella, though their motility was significantly better 173 than that of the flgN null strain or cells producing FlgND90-100, which binds neither 174 cognate subunits nor FliJ ( Fig. 2A and SI Appendix, Fig. S2). Secretion of cognate 175 subunit FlgK was also reduced to c.50% of wild type in cells producing either 176 cognate subunit export and cell motility. That motility and subunit export are only 221 partially attenuated by disruption of the chaperone-FliJ interaction indicates that the 222 variant chaperones are still able to pilot cognate subunits to the fT3SS machinery to 223 promote export. The data suggest that chaperone binding by FliJ increases the 224 efficiency of, but is not essential for, cognate subunit export (Fig. 2). This raises a 225 question as to whether the chaperone-FliJ interaction might have an additional 226 function related to the primary role of FliJ in activating the export machinery to 227 couple efficient use of the pmf to protein export [18]. 228 229

Unladen chaperones disrupt the FliJ-FlhAC interaction 230
The identification and characterisation of FlgN and FliT chaperone variants that 231 retained their subunit piloting function but were specifically unable to bind the FliJ 232 As anticipated, we found that FlhAC could bind (His)6-FliJ in the presence of 247 chaperone variants that could not bind FliJ (Fig. 3B). However, when assays were 248 performed with chaperones that could bind FliJ, the amount of FlhAC pulled down by 249 FliJ was severely reduced (Fig. 3B). These data show that FlgN and FliT can 250 compete with FlhAC for binding to FliJ, i.e. unladen chaperones bound to FliJ prevent 251 the FliJ-FlhAC interaction -an interaction known to be essential for activation of the 252 export gate to efficiently couple pmf use to protein export [18]. inhibiting pmf-driven export, until a cognate subunit is available to capture the 290 chaperone from FliJ, allowing export activity to resume. 291

Ribosome profiling reveals low cytoplasmic ratios of minor filament subunits 293 to their cognate chaperones, FlgN and FliT 294
Our accumulating data suggested that, in addition to piloting subunit cargo to the

Mutational analysis of the FlgN and FliT chaperones reveals key residues 329 specifically required for binding to the FliJ ATPase stalk 330
Our mutational analysis identified FlgN chaperone variants, FlgN W78A and 331 FlgNΔ76-78, which were defective in binding to FliJ but could still pilot their cognate 332 subunits FlgK and FlgL to dock at the export gate component FlhAC (Fig.1). Motility 333 of strains producing FlgN W78A or FlgNΔ76-78 was marginally reduced compared to 334 wild type, as was export of FlgK and FlgL, indicating that loss of the FlgN-FliJ 335 interaction reduced the efficiency of, but did not abolish, cognate subunit export (Fig.  336 2). 337 338 Mutational analysis of the FliT chaperone showed that several highly conserved, 339 surface exposed residues on the helix required for cognate subunit binding were also 340 critical for binding to FliJ (Fig.1C). Replacement of FliT leucine-72 with alanine was found to disrupt binding to FliJ, but did not affect FliT chaperone interactions with 342 FliD cognate subunit or FlhAC (Fig. 1). Cells producing FliT L72A did not display a 343 swimming motility defect compared to wild type Salmonella (Fig. 2C). However, a 344 marginal export defect was observed for cognate subunit FliD (Fig. 2D). Cells 345 producing FliT L72A displayed reduced population swarming, indicating that the FliT-346 FliJ interaction may be required for the hyperflagellation associated with this form of 347 surface motility. 348

Unladen chaperones modulate the interaction of the FliJ ATPase stalk with the 349 cytoplasmic domain of the FlhA export gate 350
A previous study showed that FliJ appeared to enhance the binding of a FliT/FliD subunits. In the situation where cellular levels of cognate subunit increase, 368 chaperones would be captured from FliJ, restoring FliJ-dependent activation of the 369 export gate and subsequent subunit export (Fig. 6). including proteins that inhibit efficient coupling between the F1 and F0 components 384 [37][38][39]. Our data demonstrate that unladen export chaperones disrupt FliJ-385 dependent activation of the flagellar export gate, which would prevent wasteful pmf 386 use in the absence of cognate subunits, by effectively preventing a FliJ-FlhA 387 interaction until the chaperone is captured from FliJ by a cognate subunit, relieving 388 the inactive state. 389

A mechanism coupling export gate activation to availability of subunits for 391 export 392
Our data support a model whereby, when subunits are at low levels in the cell, 393 unladen FlgN and FliT bind to FliJ, preventing the FliJ-FlhA interaction and 394 subsequent activation of the export gate. Ribosome profiling revealed the amounts of 395 chaperone and cognate subunit(s) produced by the cell and while all three flagellar 396 chaperones are produced at similar levels (Fig. 6), the intracellular ratio of 397 of the pmf when subunit cargo availability is low (Fig. 6).

Motility assays 473
For swimming motility, Salmonella strains were grown in LB broth to A600 1. Two 474 microliters of culture were inoculated into soft tryptone agar (0.25% agar, 10 g/L 475 tryptone, 5g/L NaCl). Plates were incubated at 37 °C for between 4 and 6 hours. For 476 swarming motility, one microliter of overnight cultures grown in LB broth was 477 inoculated onto dried tryptone agar plates (0.6% agar, 10g/L tryptone, 5g/L NaCl) 478 supplemented with 0.3% glucose and incubated at 30 °C for 16 hours. Libraries were prepared essentially as described previously [54, 55] with the 504 following modifications: Cells were grown in LB at 37°C with 180 RPM shaking to 505 A600 1.0. Chloramphenicol was added to the culture to a final concentration of 1500 506 µg/ml, followed by rapid cooling of the culture and harvesting of cells by 507 centrifugation (6,000 g, 1 min, 4°C). The cell pellet was quickly resuspended in 1 ml 508 ice cold bacterial profiling buffer (20 mM Tris-Cl pH 7.5, 140 mM KCl, 5 mM MgCl2, 509 1500 µg/ml chloramphenicol, 0.5 mM dithiothreitol (DTT), 0.5% NP40, 1% Triton X-510 100, 2.5% Sucrose) and flash-frozen for cryo-grinding in liquid nitrogen. The frozen 511 powder was thawed and clarified by centrifugation (13,000 g, 2 mins, 4°C) followed 512 by adjustment of A254 nm10. Lysates were either snap frozen in liquid nitrogen for 513 storage at -80°C or nuclease treated with RNase 1 (700U, Ambion) followed by 514 ribosomal RNA depletion using a bacterial ribozero kit (Illumina) prior to library preparation for ribosome profiling. For parallel RNA-Seq, total RNA was extracted 516 from corresponding lysates followed ribosomal RNA depletion using the bacterial 517 ribozero kit (Illumina) prior to library preparation. For Ribosome profiling, rRNA was 518 further depleted with duplex-specific nuclease as described previously [54]. 519 Ribosome profiling and RNA-Seq libraries were pooled and sequenced using the 520 NextSeq® 500/550 platform and data were trimmed and mapped to the 521 transcriptome assembly of Salmonella enterica subsp. enterica serovar Typhimurium 522 strain ST4/74 (GenBank accession CP002487.1). Paired ribosome profiling and 523 RNA-Seq analysis was performed with riboSeqR as previously described [55,56]. 524 525

Quantification and statistical analysis 526
Experiments were performed at least three times. Immunoblot data were quantified 527 using Image Studio Lite. The unpaired two-tailed Student's t-test was used to 528 determine p-values, and significance was defined as *p < 0.05. Data are represented 529 as mean ± standard error of the mean (SEM), unless otherwise specified and 530 reported as biological replicates. 531 532 Acknowledgements 533 We thank Colin Hughes and Lewis Evans for useful discussions, and Sangita Ahmed 534 for technical assistance. This work was funded by grants from the Biotechnology and 535