Dedicated bacterial esterases reverse lipopolysaccharide ubiquitylation to block immune sensing

Summary Pathogenic bacteria have evolved diverse mechanisms to counteract cell-autonomous immunity, which otherwise guards both immune and non-immune cells from the onset of an infection1,2. The versatile immunity protein Ring finger protein 213 (RNF213)3–6 mediates the non-canonical ester-linked ubiquitylation of lipopolysaccharide (LPS), marking bacteria that sporadically enter the cytosol for destruction by antibacterial autophagy4. However, whether cytosol-adapted pathogens are ubiquitylated on their LPS and whether they escape RNF213-mediated immunity, remains unknown. Here we show that Burkholderia deubiquitylase (DUB), TssM7–9, is a potent esterase that directly reverses the ubiquitylation of LPS. Without TssM, cytosolic Burkholderia became coated in polyubiquitin and autophagy receptors in an RNF213-dependent fashion. Whereas the expression of TssM was sufficient to enable the replication of the non-cytosol adapted pathogen Salmonella, we demonstrate that Burkholderia has evolved a multi-layered defence system to proliferate in the host cell cytosol, including a block in antibacterial autophagy10–12. Structural analysis provided insight into the molecular basis of TssM esterase activity, allowing it to be uncoupled from isopeptidase function. TssM homologs conserved in another Gram-negative pathogen also reversed non-canonical LPS ubiquitylation, establishing esterase activity as a bacterial virulence mechanism to subvert host cell-autonomous immunity.


116
However, the intracellular bacterial burden between WT and tssM-mutant Burkholderia was 117 indistinguishable in both MEFs and RAW264.7 macrophages (Fig. 2a). This finding, which is 118 in line with previous reports for the tssM mutant of B. mallei 9 , suggests that Burkholderia have 119 additional virulence mechanisms that counteract ubiquitin-mediated cell-autonomous 120 immunity. This prompted us to test whether TssM could promote the replication of the non-121 cytosol adapted pathogen, Salmonella. Exogenous expression of WT TssM Bp , but not 122 catalytically inactive TssM BpC292G , was sufficient to promote Salmonella replication (Fig. 2b), 123 providing further evidence that Burkholderia has at least one additional mechanism to block 124 host-mediated restriction of cytosolic bacteria.

136
As these polyubiquitin signals initiate the recruitment of ubiquitin-binding autophagy receptors 137 Optineurin (OPTN) 22 , NDP52 23 and p62 24 , we next explored the recruitment of these proteins 138 to Burkholderia. Consistent with the accumulation of polyubiquitin signals, NDP52, p62 and 139 OPTN were recruited to approximately 60% of E264::tssM pknock bacteria in an RNF213-140 dependent manner, in contrast to fewer than 5% of WT bacteria ( Fig. 2e-g). In line with the 141 overall lower levels of ubiquitin coating on E555::tssM pknock bacteria (Fig. 1e), whereas the 142 percentage of marker-positive bacteria in this strain was elevated compared to WT E555 143 bacteria in an RNF213-dependent manner, it remained below 20% of total bacteria (Extended 144 Data Fig. 2b-d). This finding is also consistent with our previous observation that RNF213 145 was recruited to a lower percentage of WT E555 B. thailandensis (Fig. 1b). Overall, we find 146 that without TssM, B. thailandensis becomes coated with polyubiquitin and associated with 147 autophagy receptors.

157
The percentages of WIPI2B-positive E264::tssM pknock and WT E264 were low and 158 indistinguishable at 6 hours post-invasion (Fig. 2i). When marker-positive bacteria were 159 quantified among the fraction of ubiquitin-positive E264::tssM pknock bacteria, the failure to 160 recruit critical autophagy proteins WIPI2B and LC3B was evident (Fig. 2j). Similarly, whether 161 quantified among the total or the ubiquitin-positive population, it was evident that WIPI2B and 162 LC3B were associated with fewer E555::tssM pknock bacteria than the autophagy receptors 163 (Extended Data Fig. 2f-h). Together, this suggests that the failure in antibacterial autophagy 164 occurs due to a block in its initiation. As B. pseudomallei prevents LC3 lipidation and 165 association of LC3 with bacteria through the action of BopA 10,12 , we propose that this, or 166 indeed other additional mechanisms that also include the polysaccharide capsule when 167 present, cooperate with the anti-RNF213 activity of TssM to promote the intracellular 168 replication of Burkholderia spp.

170
LPS represents the ubiquitylated substrate

172
Current data strongly suggest that RNF213 ubiquitylates Salmonella LPS, rather than a 173 proteinaceous substrate 4 , but whether this is the case for a host cytosol-adapted bacteria has 174 not been tested. We therefore hypothesised that the RNF213-dependent ubiquitylation of 175 tssM pknock bacteria also represented modification of bacterial LPS. To test this, we created an 176 E555::ΔwbiI mutant or an E555::ΔtssM,ΔwbiI double mutant, in which the lack of WbiI prevents 177 formation of long O-antigen, leaving only the lipid A and inner core of the LPS moiety ( Fig.   178   3a). Surprisingly, a defined, lower molecular weight ubiquitin-positive banding pattern was 179 detected in LPS-enriched lysates from the E555::ΔwbiI-infected cells (Fig. 3b). This   to the E555::ΔwbiI mutant (Fig. 3b).

186
The strict correlation between the molecular weight of the detected ubiquitin signal and O-187 antigen size strongly suggests the direct ubiquitylation of B. thailandensis LPS. As the LPS 188 core lacks amino groups suitable for amide-linked ubiquitylation 4 , it is likely that available 189 hydroxyl groups are modified 27 , creating an ester-linked ubiquitin. To test this, we treated 190 E555::ΔtssM,ΔwbiI bacteria isolated from infected cells with sodium hydroxide, which 191 selectively hydrolyses ester-linked conjugates. The addition of increasing concentrations of 192 sodium hydroxide clearly resulted in progressive loss of the ubiquitin signature (Fig. 3c).

196
TssM is a highly potent ubiquitin esterase

198
TssM has reported isopeptidase activity 9 , cleaving amide bonds within a polyubiquitin chain, 199 yet our data imply that TssM reverses ester-linked ubiquitylation. To test whether TssM

233
We determined a 2.5 Å crystal structure of TssM BpΔN191 covalently bound to Ub at its active site 234 to visualise the molecular basis for ubiquitin esterase activity (Fig. 4a, Extended Data Fig.   235 4a-c, Extended Data Table 1). The structure of TssM revealed a Big5 (bacterial Ig-like domain 236 5) fold N-terminal to a catalytic USP-type DUB module (Fig. 4a). Ig-like domains facilitate 237 protein or ligand interactions 32 , and we detected unexplained electron density in the b1 groove,

242
The TssM USP domain is very minimal and contains no sequence insertions. Boxes 1, 2, 5, 243 and 6 are well-conserved and form the core USP module, including the catalytic triad (Fig. 4b,

244
Extended Data Fig. 4g-i). Ubiquitin is typically bound at the S1 substrate-binding site by a 245 set of "fingers" encoded within Boxes 3 and 4, as well as a Box 4 "blocking loop" that guides 246 the ubiquitin C-terminus into the active site. In contrast, TssM has no recognizable finger 247 structure and compensates for an extremely short Box 4 region by encoding an analogous 248 blocking loop within Box 6 instead (Fig. 4b, Extended Data Fig. 4g-i).

250
Within the TssM active site lies an aligned Cys-His-Asp catalytic triad, as well as a conserved

251
Asn that forms the oxyanion hole (Fig. 4c). Mutation of these sites ablated or diminished 252 isopeptidase activity toward a Lys-Ub substrate, but interestingly the acidic Asp or oxyanion 253 hole Asn were not required for ubiquitin esterase activity (Fig. 4d,e, Extended Data Fig. 5a,b).

254
Among USPs the length of the so-called Cys-loop that precedes that catalytic Cys is almost 255 perfectly conserved. TssM encodes a one amino acid insertion within the Cys-loop (Fig. 4f), 256 but its truncation had a minimal effect on esterase activity and instead DL287 specifically 257 reduced isopeptidase function (Fig. 4g,h, Extended Data Fig. 5a,c). As the ubiquitin C-258 terminus threads into the active site, TssM coordinates the basic R42, R72, and R74 residues 259 of ubiquitin with acidic residues E362 and E469 in Boxes 2 and 6, respectively (Fig. 4i).

260
Mutation of either residue severely diminished isopeptidase activity, but esterase activity was 261 more significantly impacted by an inability to coordinate R42 and R72 with residue E469 (Fig.   262 4j,k, Extended Data Fig. 5a,d). Lastly, the remainder of the S1 site is composed of

268
As we noted that several of our structure-guided mutations in the TssM active site or S1 site 269 affected isopeptidase function more so than esterase, we sought to determine if any had 270 completely lost isopeptidase activity. We selected N286A, E362R, F459A, V466R, and E469R 271 as TssM mutants that more significantly impacted isopeptidase activity ( Fig. 4c-  Ser-Ub and Thr-Ub substrates, the V466R mutant exhibited a ~21,000-fold reduction in kcat/KM 277 for Lys-Ub, amounting to an ~5,000-fold specificity for ester-over isopeptide-linked substrates 278 ( Fig. 4o, Extended Data Fig. 5g). We suggest that TssM esterase activity is more compatible 279 with weak or transient substrate encounters, with the differential dependency on certain active 280 site features allowing us to uncouple esterase and isopeptidase activity.

282
Conservation of ubiquitin esterase activity 283 284 Next, we asked how common esterase activity is among an array of bacterial DUBs. As 285 observed with human DUBs 28 , many bacterial DUBs were capable of ubiquitin esterase 286 activity at high (0.5 µM) enzyme concentration (Fig. 5a). Despite this, only TssM protected 287 Salmonella from LPS ubiquitylation when each DUB was expressed in trans (Fig. 5b).

288
Furthermore, recombinant TssM was the most potent DUB tested, removing all anti-ubiquitin 289 detected bands from E555::ΔtssM,ΔwbiI bacteria isolated from infected cells (Fig. 5c). This 290 revealed a specificity within TssM, compared to several other bacterial DUBs, for the removal 291 of ubiquitin from LPS.

293
To extend our analysis of substrate specificity, we specifically examined a subset of bacterial 294 peptidases from the C19 family, for which TssM is a member. Using MEROPS, we selected  306 6b), and exhibited over 100-fold greater activity toward the ester-linked substrates than the 307 isopeptide substrate (Fig. 5f,g). Finally, as recombinant TssM Cs hydrolysed ubiquitin from the 308 LPS of E555::ΔtssM,ΔwbiI (Fig. 5h), we compared the AlphaFold model 40,41 of TssM Cs to our 309 TssM structure to assess structural homology (Fig. 5i, Extended Data Fig. 6c,d). Consistent 310 with similar activities toward ubiquitylated LPS, there was considerable alignment between the 311 active site regions, where the AlphaFold model confidence was highest. Together, these 312 findings suggest that ubiquitin esterase activity is encoded by at least two cytosolic bacteria 313 as a mechanism of countering RNF213 defences.

315
Our finding that some conserved bacterial DUBs are esterases that selectively reverse the 316 ubiquitylation of a non-canonical substrate reveals a previously undescribed molecular 317 mechanism for the evasion of RNF213-mediated cell-autonomous immunity. For cytosolic 318 Burkholderia spp., we propose that at least three mechanisms cooperate to counteract 319 RNF213. First, the capsule provides a physical barrier as evidenced by lower RNF213 320 association with E555 bacteria compared to E264 bacteria, consistent with acapsular B.       Ub (grey) shown in the S1 site. Representative 2|Fo|-|Fc| electron density is shown at 1s. c)

617
The following primers were used: