2020.03.10.984922V1.Full.Pdf

bioRxiv preprint doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

1 Molecular basis of ubiquitination catalyzed by the bacterial transglutaminase MavC

1,2* 3* 1,2* 1,2* 3 1,2 3 Hongxin Guan , Jiaqi Fu , Ting Yu , Zhao-Xi Wang , Ninghai Gan , Yini Huang ,

1 1,2 3† 1,2† 4 Vanja Perčulija , Yu Li , Zhao-Qing Luo and Songying Ouyang

1 6 The Key Laboratory of Innate Immune Biology of Fujian Province, Provincial University

7 Key Laboratory of Cellular Stress Response and Metabolic Regulation, Biomedical

8 Research Center of South China, Key Laboratory of OptoElectronic Science and

9 Technology for Medicine of the Ministry of Education, College of Life Sciences, Fujian

10 Normal University, Fuzhou, China

2 11 Laboratory for Marine Biology and Biotechnology, Pilot National Laboratory for Marine

12 Science and Technology (Qingdao), Qingdao, China

3 13 Purdue Institute for Inflammation, Immunology and Infectious Disease and Department of

14 Biological Sciences, Purdue University, West Lafayette, IN, USA

16 *These authors contributed equally to this work.

17 †Corresponding authors: [email protected] (ZQL); [email protected] (SO)

1 bioRxiv preprint doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

19 Summary

20 The Legionella pneumophila effector MavC is a transglutaminase that carries out atypical

21 ubiquitination of the ubiquitin (Ub) E2 conjugation enzyme UBE2N by catalyzing the

22 formation of an isopeptide bond between Gln40 of Ub and Lys92 (or to a less extent, Lys94)

23 of UBE2N, which results in inhibition of UBE2N signaling in the NF-κB pathway. In the

24 absence of UBE2N, MavC deamidates Ub at Gln40 or catalyzes self-ubiquitination.

25 However, the mechanisms underlying these enzymatic activities of MavC are not fully

26 understood at molecular level. In this study, we obtained the structure of the

27 MavC-UBE2N-Ub ternary complex that represents a snapshot of covalent cross-linking of

28 UBE2N and Ub catalyzed by MavC. The structure reveals the unique way by which the

29 cross-linked catalytic product UBE2N-Ub binds mainly to the Insertion and the Tail

30 domains of MavC prior to its release. Based on our structural, biochemical and mutational

31 analyses, we proposed the catalytic mechanism for both the deamidase and the

32 transglutaminase activities of MavC. Finally, by comparing the structures of MavC and

33 MvcA, the homologous protein that reverses MavC-induced UBE2N ubiquitination, we

34 identified several key regions of the two proteins responsible for their opposite enzymatic

35 activity. Our results provide insights into the mechanisms for substrate recognition and

36 ubiquitination mediated by MavC as well as explanations for the opposite activity of MavC

37 and MvcA.

39 Keywords

40 Legionella pneumophila; effectors; transglutaminase; ubiquitination; deamidase

2 bioRxiv preprint doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

42 Introduction

43 Signal transduction in cells is often mediated by posttranslational modifications

44 (PTMs), which impact the activity of existing proteins to allow rapid responses to upstream

1,2 45 cues. Among more than 200 different types of PTMs identified so far , ubiquitination is

46 one of the most widely used. Canonical ubiquitination requires the activities of the E1, E2

47 and E3 enzymes that respectively activate, conjugate and ligate the 76-residue ubiquitin

3 48 (Ub) to modify proteins . Ubiquitination itself is further regulated by ubiquitination and other

49 types of PTMs such as phosphorylation, acetylation and ADP-ribosylation that target Ub,

4 50 components of the ubiquitination machinery, or both . This complex crosstalk among

51 various PTMs allows cells to achieve better fine-tuning of their response to various stimuli,

4,5 52 particularly under disease conditions .

53 Pathogens have evolved diverse mechanisms to co-opt host functions to promote their

54 fitness. One such mechanism is the acquisition of virulence factors capable of effective

6 55 modulation of cellular processes by various PTMs . Legionella pneumophila, the causative

56 agent of Legionnaires’ disease, is one such example. The intracellular life cycle of this

57 bacterium utilizes the Dot/Icm type IV secretion system that injects hundreds of virulence

7,8 58 factors known as effectors into host cells . These effectors extensively modulate cell

59 signaling hubs such as small GTPases and the Ub network to create a niche permissive for

9 60 intracellular replication of the L. pneumophila .

61 Co-option of the host Ub network by L. pneumophila appears to be of particular

62 importance for modulating host cellular immune process to facilitate its intracellular

63 replication. More than 10 effectors with E3 Ub ligase activity have been identified. Although

64 their target proteins remain elusive in most cases, theseeffectors cooperate with E1 and E2

10 65 enzymes in host cells to form active ubiquitination machineries (Fig. S1A). A paradigm

66 shift discovery was made by the study of the SidE effector family (SidEs) that includes

+ 67 effectors such as SdeA, which catalyze a NAD -dependent ubiquitination. This mechanism

68 involves Ub activation via ADP-ribosylation and phosphodiesterase (PDE)-mediated

69 ligation of phosphoribosylated ubiquitin (PR-Ub) onto serine residues of substrate proteins

3 bioRxiv preprint doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

11-13 70 (Fig. S1B). Interestingly, two research groups recently reported that DupA and DupB,

71 the two highly homologous PDE domain-containing deubiquitinases from L. pneumophila,

14 72 similarly reverse phosphoribosyl serine ubiquitination on their substrates (Fig. S1B).

73 Moreover, the activity of SidEs is regulated by SidJ, another effector which inhibits the

15-17 74 mono-ADP-ribosyltransferase activity by calmodulin-dependent glutamylation (Fig.

75 S1B).

76 The modification of the E2 enzyme UBE2N by MavC represents another atypical

77 ubiquitination mechanism. In this reaction, UBE2N, which exists in cell primarily as a

18,19 78 UBE2N~Ub conjugate linked by a thioester bond , is ligated to Ub via an isopeptide

79 bond formed between Gln40 of Ub and Lys92 (i.e. γ-glutamyl-ε-Lys bond between UbGln40

20 80 and UBE2NLys92) or, to a lesser extent, Lys94 of UBE2N . This ligation is mediated by

21 81 transglutamination, a reaction that does not require exogenous energy (Fig. S1C-D).

82 Analogously to other transglutaminases that function as deamidases in the absence of

22 83 their target substrates , MavC uses catalytic Cys74 that is crucial for both enzymatic

23 84 activities (Fig. S1E). Ubiquitination at Lys92 abolishes the activity of UBE2N, which in

85 turn curbs the formation of K63-type polyubiquitin chains through canonical ubiquitination

20 86 otherwise mediated by UBE2N, E1 and UVE1, thereby inhibiting NF-κB activation (Fig.

87 S1A).

23 88 MavC and its homolog MvcA are structurally similar to the canonical ubiquitin

89 deamidase cycle inhibiting factor (Cif) effectors from enteropathogenic Escherichia coli and

24,25 90 its homolog in Burkholderia pseudomallei (CHBP) (Fig. 1). Both MavC and MvcA have

91 ubiquitin deamidase activity but only MavC is able to induce monoubiquitination of UBE2N.

92 Furthermore, we recently found that MvcA counteracts the trangslutamination activity of

21 93 MavC by removing ubiquitin from UBE2N-Ub . However, although the structures of MavC

23 94 and its homolog MvcA have been solved (Fig. 1C), the mechanism underlying

95 transglutaminase-induced UBE2N ubiquitination by MavC and the molecular basis for their

96 opposite catalytic activities both remain elusive. Here, by solving the structure of the

97 MavC-UBE2N-Ub ternary complex and comparing it to other available structures of MavC

98 and MvcA, we illustrate the structural basis for substrate recognition by MavC and the 4 bioRxiv preprint doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

99 mechanism that mediates the formation of the isopeptide bond between Lys92 in UBE2N

100 and Gln40 in Ub. In addition, structural comparison of the MavC and MvcA in their apo form

101 and in ternary complex has allowed us to gain insights into the basis of the opposite

102 biochemical activity exhibited by these two highly similar proteins in terms of regulation of

103 UBE2N ubiquitination.

104

105 Results

106 The Insertion domain of MavC is essential for UBE2N ubiquitination but not for

107 self-ubiquitination and ubiquitin deamidation activities of MavC

108 Purified MavC from E. coli exists primarily as a mixture of monomers and dimers in

109 solution, both of which interact with UBE2N in size-exclusion chromatography (SEC) (Fig.

110 S2). Unlike Cif and CHBP, both apo MavC (PDB ID: 5TSC) and MvcA (PDB ID: 6K11)

111 possess a unique Insertion domain, which likely plays an important role in interaction with

23 112 UBE2N (Fig. 1). To test this hypothesis, we constructed a MavC truncation mutant

113 missing Insertion domain (MavC∆mid, lacks residues Gln131 to Asn223) and then examined

114 its activities (Fig. 1A). In contrast to wild-type MavC (MavCWT) that robustly induced

115 UBE2N ubiquitination, both MavCC74A and MavC∆mid lost the ability to carry out UBE2N

116 ubiquitination (Fig. 2A left panel). In line with the biochemical results, although MavC∆mid

117 displayed high expression levels in the L. pneumophila strain ∆mavC and was delivered

118 into host cells, it failed to ubiquitinate UBE2N (Fig. 2B). Intriguingly, MavC∆mid is still

119 capable of catalyzing self-ubiquitination (Fig. 2A right panel) and ubiquitin deamidation

120 (Fig. 2C). These results suggest that the Insertion domain is essential for UBE2N

121 ubiquitination but not for ubiquitin deamidation and self-ubiquitination activities of MavC.

122 Considering our findings, we further hypothesized that the Insertion domain of MavC is

123 involved in substrate recognition.

124

125 Binding affinities between MavC and its substrates

5 bioRxiv preprint doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

126 We first, examined the binding affinity between Ub and MavC. Although Ub

127 participates in MavC-induced transglutamination in vitro, we could not observe the

128 formation of a stable complex between Ub and MavCC74A in solution using SEC and no

129 detectable binding was detected in MST assays (Fig. S3A-B). High concentrations of Ub

130 (>3 mM) were insufficient to generate saturation curve in MST assay. We therefore

131 assumed that binding between these two proteins is too weak to be detected under the

132 assay conditions. Similarly, we did not observe crystal growth when mixing MavC and Ub

133 at 1:3 ratio. These findings are inconsistent with those of a previous study that confirmed

134 binding between MavC and Ub by monitoring chemical shift perturbations (CSPs) in NMR

23 135 titration experiments . The differences are probably due to the fact that the catalytically

15 136 active MavCWT used in this previous study deamidated the N-labeled ubiquitin during

23 137 titration in solution nuclear magnetic resonance (NMR) . Thus, although Ub participates in

138 MavC-mediated transglutamination and deamidation reactions, its interactions varies when

21 139 assayed with enzymatically active or inactive MavC .

140 We also examined binding between MavCC74A and UBE2N with the MST assay and

141 found that these two molecules bind at a Kd value of 1.14 µM in solution (Fig. S3C).

142 Moreover, since the best resolution we could obtain for the MavCC74A-UBE2N binary

143 complex was only about 3.5 Å, we speculated that the intermolecular interactions mediated

144 by the Insertion domain of MavC are dynamic. Interestingly, although both MavC and MvcA

23 145 contain an Insertion domain, MvcA has been showed to have no interaction with UBE2N .

146 Taken together, these results indicate that, in spite of ~50% sequence identity, MavC and

147 MvcA differ in recognizing UBE2N.

148 To investigate the molecular basis for substrate binding and transglutaminase activity

149 of MavC, we aimed to solve the structure of the MavC-UBE2N-Ub ternary complex. We

150 attempted to improve the binding stability between MavCC74A and its substrates Ub and

151 UBE2N by directly supplying the product of MavC-catalyzed transglutamination on the

152 premise that the binding affinity between MavC and cross-linked UBE2N-Ub binary

153 complex (Kd=1.4 µM) is similar to the binding affinity between MavC and UBE2N and would

154 thus solve the problem of weak interactions between and MavC and free Ub (Fig. S3D). 6 bioRxiv preprint doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

155 The MavCC74A and UBE2NK94A mutants were used for crystallization experiments to ensure

21 156 that UBE2N is ubiquitinated only at Lys92 .

157

158 Overall structure of MavC in complex with its product UBE2N-Ub and catalytic site

159 interactions

160 By following the protocol described above, we successfully crystallized and solved the

161 structure of the MavCC74A-UBE2NK94A-Ub ternary complex at a 2.85 Å resolution (Table 1).

162 Only one copy of the ternary complex could be found in the asymmetric unit (ASU).

163 Interestingly, the architecture of the MavCC74A-UBE2NK94A-Ub complex is highly similar to

21 164 that of the MvcA-UBE2N-Ub complex reported in our previous study . The ternary

165 complex solved in this study represents the stage of MavC-mediated UBE2N ubiquitination

166 in which UBE2N and Ub have already been cross-linked by an isopeptide bond but are still

167 bound to MavC (Fig. 3A-B, Movie S1). MavC, which assumes a concave shape, consists

168 of a Core domain flanked by a helical Tail domain and an Insertion domain (Fig. 1 and

169 3A-B). Loops 1, 2 and 5 along with loops 3 and 4 serve as flexible hinges that respectively

170 link the Tail domain and the Insertion domain to the Core domain, thus conferring the

171 flexibility to the two subdomains required for binding UBE2N-Ub (Fig. 3A-B). The

172 covalently linked UBE2N-Ub sits in the groove formed by the Tail domain and the Insertion

173 domain of MavC with the UBE2N and Ub portion of the molecule hanging on either side of

174 the enzyme.

175 The catalytic site for the transglutaminase is situated at the bottom of the concavity of

176 MavC, with the Cys74-His231-Gln252 catalytic triad (Cys74 was mutated to Ala in our

177 structure) at the center of the concave line. Several residues adjacent to the catalytic triad

178 contribute to the stabilization of the UBE2N loop and the Ub loop that respectively contain

179 Lys92 and Gln40, the two reactive residues covalently bonded by transglutamination.

180 Ala74, Thr230 and Trp255 of MavC interact with Gln40 of Ub through hydrogen bonding

181 that involves both side chain and backbone atoms. Moreover, Arg126 and Thr230 of MavC

182 form three pairs of hydrogen bonds with Asp39 of Ub, thereby further stabilizing the

7 bioRxiv preprint doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

183 Gln40-containing loop. The UBE2N loop containing Lys92 is held in place by hydrogen

184 bonding between Asn318 of MavC and Leu91 of UBE2N (Fig. 3D).

185

186 Key residues mediating interactions between MavC and UBE2N and their

187 implications in transglutaminase of MavC

188 In the ternary complex, interactions between MavC and UBE2N include hydrogen

189 bonding, electrostatic and hydrophobic interactions. Three pairs of hydrogen bonds

190 (Lys132-Glu61, Glu207-Lys6 and Tyr198-Gln100) at the interface between the Insertion

191 domain of MavC and UBE2N contribute to UBE2N recognition (Fig. 4A). Electrostatic

192 interactions further stabilize the binding of UBE2N-Ub with MavC. These electrostatic

193 interactions involve the loop between β6 strand and α7 helix composed of seven negatively

194 charged residues (Asp196, Glu197, Asp200, Glu202, Glu203, Glu206 and Glu207) and

195 three Tyr residues (Tyr189, Tyr192 and Tyr198) that interacts with a positively charged

196 region formed by four residues (Arg6, Arg7, Lys10 and Arg14) of the α1 helix in UBE2N

197 (Fig. 4A). According to the sequence conservation analysis, the above residues of MavC

198 are conserved except Y189, E197 and Glu202 (Fig. S4). Hydrophobic interactions are

199 mediated by Met317 from the Core domain of MavC which inserts into a hydrophobic

200 pocket in UBE2N formed by residues Ile86, Leu88, Ile90, Leu99, Val104, Ile108 and

201 Leu111 (Fig. 4B-C).

202 To determine the impact of the residues that play role in interaction and recognition of

203 UBE2N during ubiquitination, we designed a set of MavC mutants and tested their ability to

204 produce UBE2N-Ub. Mutations of Tyr192, Tyr198 and Met317 to alanine severely impaired

205 ubiquitination of UBE2N, whereas the Y189A and N318A mutants were only partially

206 defective. Furthermore, mutation to alanine of E207 that hydrogen bonds with Lys6 of

207 UBE2N caused a slight defect in catalyzing UBE2N ubiquitination, suggesting that this

208 residue is required for optimal ubiquitination activity. In contrast, we did not detect defects

209 associated with MavCE197A (Fig. 4D-E). These observations are consistent with the binding

210 results, which showed that MavCY192A and MavCY198A failed to bind UBE2N (Fig. 4F-G). 8 bioRxiv preprint doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

211 When introduced into the L. pneumophila ∆mavC mutant on a plasmid, each of the

212 abovementioned mutants can be expressed and translocated into host cells at levels

213 comparable to that of wild-type MavC. Our in vivo results, reveal that only the M317A

214 mutant has lost the ability to induce UBE2N ubiquitination, thus implying an important role

215 of this residue in the transglutaminase activity of MavC (Fig. 4H-I). Therefore, electrostatic

216 interactions and hydrophobic interactions mediated by Met317 are the main force to

217 stabilize the binding of UBE2N-Ub with MavC.

218 To further validate the substrate recognition mechanism implied by the ternary

219 structure, we designed UBE2NR6A/R7A, UBE2NK10A/R14A and UBE2N∆N6-14 (deletion mutant

220 missing the residues from Arg6 to Arg14) mutants predicted to affect intermolecular

221 electrostatic interactions and tested them for ubiquitination using the method described

222 above. We found that neither UBE2NR6A/R7A nor UBE2N∆N6-14 demonstrated detectable

223 ubiquitination and that UBE2NK10A/R14A can still be ubiquitinated but at a markedly reduced

224 level (Fig. 5A-D).

225

226 Interactions between MavC and Ub

227 Although MavCC74A was not found to interact with Ub directly in SEC and does not

228 exhibit observable affinity for it in MST assays, the structure of our ternary complex

229 structure clearly shows that MavCC74A indeed interacts with Ub via five distinct contact

230 regions (termed contact regions A–D and the carboxyl terminus contact region CTC) on the

231 Tail domain of MavC. Several pairs of hydrogen bonds and hydrophobic interactions from

232 these regions contribute to positioning of the ubiquitin molecule optimal for

233 transglutamination (Fig. S5A-B). Contact region A is formed by a hydrogen bond between

234 Glu42 of MavC and Lys6 of ubiquitin (Fig. S5C). Contact region B involves the loop of the

235 N-terminal β-hairpin in Ub, particularly residues Leu8, Thr9 and Gly10, which engage in

236 hydrophobic interactions with a hydrophobic pocket of MavC formed by residues Ile31,

237 Pro32, Ile35, Leu36 (Fig. S5D). In contact region C, three pairs of hydrogen bonds are

238 formed, namely Gln31(Ub):Glu123(MavC), Glu34(Ub):Asn79(MavC) and

9 bioRxiv preprint doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

239 Gly35(Ub):Arg121(MavC) (Fig. S5E). Contact region D features interactions mediated by a

240 single pair of hydrogen bonds formed between Asn25 of ubiquitin and Ile163 of MavC (Fig.

241 S5F). Finally, the flexible tail at the carboxyl terminus of ubiquitin is held in place by

242 contacting an adjacent groove of MavC through hydrogen-bonding involving backbone

243 atoms of Glu66, Gln69, His258 and side chains of Arg72 and Arg74 of the ubiquitin tail (Fig.

244 S5G).

245

246 Rotation of the Insertion and Tail domains of MavC is essential for delivering

247 UBE2N and Ub to the trangslutaminase active site

248 In comparison to the apo form, the Insertion domain of MavC underwent a distinct

249 approximately 30º anticlockwise rotation in the ternary complex. This movement allowed

250 formation of electrostatic interactions and hydrogen bonds between the Insertion domain

251 and positively charged region of UBE2N (Fig. 6A-C). The loop 1 (Leu88-Trp95) of UBE2N

252 and the loop2 (His311-Lys320) of MavC interlock and thus strengthen binding between

253 MavC and UBE2N and firmly stabilize the Lys92 of UBE2N in the enzyme activation center

254 (Fig. 6D-E, Movie S2). Similarly to Insertion domain, the Tail domain of MavC also rotates

255 anticlockwise relative to its position in apo form MavC. The rotation of the Tail domain

256 instigates hydrogen bonding between Gln42 of MavC and Lys6 of Ub and electrostatic

257 interactions (Fig. 6F-G). In addition, the loop 3 (Thr71-Ser73) and loop 4 (Gln252-Ser257)

258 play an important role in maintaining the Lys92 of UBE2N and Gln40 of Ub in the precise

259 location of the enzyme activation center (Fig. 6H). Based on these findings, we propose

260 that a anticlockwise rotation of the Insertion and Tail domains in MavC during ternary

261 complex formation moves Lys92 of UBE2N and Gln40 of Ub closer to the catalytic center,

262 and the Core domain serves as a scaffold to stabilize UBE2NLys92 and UbGln40 in the

263 catalytic center, these three domains complete the transglutamination reaction

264 cooperatively.

265

10 bioRxiv preprint doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

266 The mechanism for MavC-mediated transglutamination and molecular basis of

267 opposite enzymatic activities of MavC and MvcA

268 MvcA is a deubiquitinase that counteracts the activity of MavC by removing ubiquitin

21 269 from UBE2N-Ub . However, in spite of their opposite enzymatic activities, these two

270 proteins share approximately 50% identity and their structures are highly similar (Fig. 1C).

271 Apo forms of MavC and MvcA represent the closed catalytically inactive form of the two

272 proteins characterized by high conformational stability, which can be seen when

23 273 superimposing three available structures of apo MvcA (PDB ID: 5SUJ , 5YM9, and also

21 274 our previous work with PDB ID 6K11 ) — the three structures and all monomers in an

275 asymmetrical unit align with maximal RMSD of 0.613 Å (Fig. 1C). Hence, we used the apo

276 form structures as reliable reference point in our attempt to determine functional

277 divergence of MavC and MvcA by comparing the structures of MavC and MvcA in their apo

278 form and in ternary complexes. Interestingly, the structural similarity between MavC and

279 MvcA from respective ternary complexes is markedly higher than between apo-MavC and

280 apo-MvcA (Fig. S6A-D). This suggests that the two proteins undergo significant

281 conformational changes during transition to ternary complexes. In both proteins, these

282 conformational changes can be mainly contributed to rotational movement of the Insertion

283 domain and Tail domain relative to the Core domain.

284 To investigate whether the opposite enzymatic activities are the consequence of

285 divergent catalytic reactions, we searched for potential structural differences in the catalytic

286 triads of MavC and MvcA. The two proteins utilize an identical Cys-His-Glu catalytic triad in

287 which the catalytic cysteine (Cys74 in MavC and Cys83 in MvcA) is essential both for the

288 formation and the cleavage of the isopeptide bond between Lys92 of UBE2N and Gln40 of

20,21 289 Ub . In a comparable scenario, SdeA and DupA utilize an identical set of catalytic

290 residues, but conformational differences between these residues give rise to opposite

26 291 activities in phosphoribyl ubiquitination . As inferred by comparison of the apo-MavC and

292 apo-MvcA structures, the catalytic triads (Cys, His, Glu) of MavC and MvcA assume

293 equivalent conformations in the inactive state (Fig. 7A). Interestingly, the comparison of

294 MavC-UBE2N-Ub and MvcA-UBE2N-Ub ternary complexes showed that all three residues 11 bioRxiv preprint doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

295 in the catalytic triads are in similarly close alignment, and the differences in positioning of

296 the isopeptide bond between Lys92 of UBE2N and Gln40 of Ub are minute (Fig. 7B). We

297 therefore reasoned that transglutamination and deubiquitination mediated by MavC/MvcA

298 are in fact forward and backward reactions of the same reversible catalytic reaction,

299 respectively and that the preference for the direction in which the reaction proceeds is

300 governed by other factors such as substrate binding affinity and stability of the

301 substrate/product bound state. These observations and the fact that the Insertion domain

302 and Tail domain undergo rotational movement during formation of the MavC/MvcA ternary

303 complex prompted us to divert our attention to these two domains in further analysis.

304 While analyzing the Insertion domains of MavC and MvcA, we noticed that their

305 negatively charged regions vary considerably in terms of negatively charged amino acid

306 composition (Fig. 7C). Specifically, the negatively charged region of MavC Insertion

307 domain contains seven negatively charged amino acids, whereas the corresponding region

308 in MvcA contains only five, indicating that the net electric charge of the negatively charged

309 region in MavC is more negative than the one in MvcA (Fig. 7D). In both ternary complexes,

310 the Insertion domain utilizes the negatively charged region to interact with the positively

311 charged region of UBE2N through electrostatic forces (Fig. 7E). Moreover, the

312 conformational changes observed between the two ternary complexes and their apo form

313 structures imply that the Insertion domains of both proteins bind UBE2N via anticlockwise

314 rotation (Fig. 7F-G). Thus, the interaction forces and mechanisms for UBE2N by MavC and

315 MvcA are highly similar, but the binding affinity between UBE2N and MvcA is likely to be

316 lower than between UBE2N and MavC because of lower net negative charge.

317 As one of the substrates, Ub mainly interacts with the Tail domain of the MavC and

318 MvcA. Intriguingly, comparison of the MavC and MvcA ternary complexes to their apo

319 forms shows that the Tail domain of MavC rotates anticlockwise and the Tail domain of

320 MvcA rotates clockwise (Fig. 8A-B). Since the surface of Ub that interacts with the Tail

321 domains of MavC and McvA is positively charged, the anticlockwise rotation of MavC Tail

322 domain increases local negative potential, which likely results in tighter binding with Ub

323 (Fig. 8C-E). On the contrary, the clockwise rotation of MvcA Tail domain creates an 12 bioRxiv preprint doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

324 obvious positive bulge, which probably facilitates uncoupling of Ub from UBE2N (Fig.

325 8F-H). Another difference between MavC and MvcA is a residue in proximity to the active

326 site, i.e. Trp255 of MavC and the corresponding residue Phe268 in MvcA (Fig. 1B, 8I-J,

327 S4). The distinct properties of these two residues alter substrate binding in the active site

328 as well as orientations of the Tail and Insertion domains that contact Ub and UBE2N.

329 In conclusion, we summarize all the structural analyses described above to propose

330 the molecular basis for the opposite enzymatic activities of MavC and MvcA in terms of

331 regulation of UBE2N ubiquitination (Fig. 9A-B). The initial substrate engagement of

332 MavC/MvcA involves binding of the UBE2N~Ub conjugate (the primary form in which

18,19 333 UBE2N exists in cell )/cross-linked UBE2N-Ub through interactions that induce rotation

334 of the Insertion domain and Tail domain, which in turn positions UBE2N~Ub/ UBE2N-Ub in

335 the active site of the respective protein. In MavC, the rotation of the Insertion domain and

336 the Tail domain results in more stable binding between the UBE2N~Ub conjugate and

337 MavC, which is favorable for transglutamination (Fig. 9A). Conversely, the rotation of the

338 Tail domain in MvcA and lower binding affinity of the Insertion domain for UBE2N promote

339 the dissociation of Ub and UBE2N from the enzyme, thus leading to deubiquitination (Fig.

340 9B).

341

342 Discussion

343 MavC is a transglutaminase that catalyzes covalent cross-linking of Gln40 of

344 ubiquitin to either Lys94 or Lys92 of UBE2N (Fig. S1C), and it also possess deamidase

23 345 activity that targets Ub at Gln40 in the absence of UBE2N (Fig. S1D). The effects of

346 MavC is counteracted by MvcA, which is a close homolog of MavC that functions as a

21 347 deubiquitase against UBE2N-Ub (Fig. S1D). Like MavC, MvcA also exhibits ubiquitin

348 deamidase activity in a manner similar to such bacterial deamidase such as Cif and

24,25 349 CHBP that deamidate Gln40 of Ub and Ub/NED88, respectively . Structural comparison

350 of these four enzymes reveals that the structures of CHBP and Cif align strikingly well with

351 the Core and Tail domains of MavC and MvcA (Fig. 1). However, the Insertion domain

13 bioRxiv preprint doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

352 appears to be characteristic of MavC and MvcA as its structural equivalent is absent in

353 CHBP or Cif (Fig. 1). Therefore, we reason that the Core and Tail domains of MavC and

354 MvcA are sufficient for their deamination activity. Indeed, deletion of the Insertion domain

355 abolishes the transglutamination activity of MavC while remains largely active in its

356 deamination activity (Fig. 2).

357 That being said, what is the functional significance of the Insertion domain? CHBP is

358 known to target multiple signaling pathways with its ubiquitin deamidation activity via

359 blocking free ubiquitin chain synthesis by different E3-E2 pairs, leading to the inhibition of

360 ubiquitination of RhoA mediated by a Cullin-based E3 complex, and subsequently cell

27,28 361 cycle arrest . In contrast, the scope of activity for MvcA and MavC appears narrower as

362 they specifically regulate ubiquitination in UBE2N-related pathways such as NF-κB 29 363 signaling . Previously reported structure of the MvcA-UBE2N-Ub ternary complex reveals

21 364 that the Insertion domain is involved in recognition of the UBE2N-Ub substrate . In line

365 with that, our study shows that the Insertion domain of MavC is also involved in the

20 366 recognition of UBE2N . Hence, the Insertion domain enables MvcA and MavC to act

367 specifically on UBE2N, thereby making the regulation of host ubiquitination signaling by

368 MavC and MvcA more specific and precise.

369 Catalytic domains that mediate chemically opposite reactions in highly homologous

370 proteins are not unprecedented: the PDE domains of DupA/B and SidE enzymes are

371 another example. PDE domains of SidE enzymes have moderate binding affinity for Ub

372 and catalyze PR ubiquitination, whereas PDE domains of DupA/B bind strongly to Ub and

373 mediate the removal of PR-Ub. The formation of stable enzyme-substrate complexes is

374 required in mediate cleavage reaction, while the ligation reaction requires moderate

26 375 binding affinity to substrates, probably allowing efficient release of products . MavC and

376 MvcA are highly similar proteins with approximately 50% sequence identity and identical

377 catalytic triad (Fig. 1B), yet MavC primarily catalyzes the formation of an isopeptide bond

378 between UBE2NLys92 and UbGln40, whereas MvcA is responsible for breaking of the same

379 isopeptide bond. The structure of MavC ternary complex shows that the negatively charged

380 surface of Tail domain attracts the positively charged surface of Ub, resulting in stable

381 binding between MavC and Ub. Concurrently, the anticlockwise rotation of the Ub-bound 14 bioRxiv preprint doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

382 Tail domain draws Ub closer to UBE2N and the active center to facilitate isopeptide bond

383 formation. In MvcA, however, the surface charge of Tail domain is repulsive to the Ub and

384 low binding affinity of UBE2N favors the dissociation of Ub from MvcA, which in turn

385 facilitates the isopeptide bond breaking and separation of Ub and UBE2N from MvcA. Thus,

386 given their overall structural similarity and equivalent chemical nature of the ubiquitin

387 ligation by transglutamination and deubiquitination activities, the fact that MavC and MvcA

388 are licensed with opposite enzymatic activities is due to differences in substrate binding

389 preference and stability of the substrate/product-bound intermediates. We proposed that

390 the formation of stable enzyme-substrate complexes is required for the synthesis reaction,

391 whereas weak interactions between enzyme and product favor cleavage reactions.

392 For the transglutaminase activity of MavC, the Insertion domain enables the specific

393 binding of MavC to UBE2N, whereas the Tail domain maintains the binding with Ub, and

394 the Core domain serves as scaffold to stabilizes UBE2NLys92 and UbGln40 in the catalytic

395 center. These three domains coordinate to complete the catalytic reaction. The shape

396 complementarity and architecture, rather than specific individual interactions, is crucial for

397 isopeptide formation of UBE2N and Ub. For the deamidase activity of MavC, the Insertion

398 domain is dispensible. In the catalyzing process of MavC, Cys74 first attacks Gln40 in

399 ubiquitin to form a thioester intermediate. When UBE2N is present, the acylated MavC

400 reacts with the amine donor from the ε -lysine in UBE2N to form an intermolecular

401 isopeptide bond. In the absence of UBE2N, the acylated MavC is further attacked by a

402 nucleophilic water to produce UbQ40E. MvcA can deubiquitinate the UBE2N-Ub product

20,21 403 from MavC transglutaminase activity (Fig. S7).

404 As a key checkpoint for activation of the NF-κB pathway, UBE2N appears to be a

405 common target of such cellular subversion as bacterial infection; it can be regulated by

30 4 406 diverse mechanisms such as deamidation and ISGylation . MavC and MvcA regulate

407 the activity of UBE2N in spatial and temporal manner via their opposite enzymatic activities.

408 Cross-linking of UBE2N and Ub catalyzed by MavC leads to the inhibition of the

409 UBE2N-mediated NF-κB activation, which is essential in the early infection, whereas the

410 deubiquitinase activity of MvcA restores the activity of UBE2N to allow the recovery of host

411 signaling pathways in the later phases of infection when the bacteria have already 15 bioRxiv preprint doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

412 successfully evaded host detection. Such mechanisms may allow less disruptions to the

413 pathogen or promote symbiotic coexistence between the pathogen and their hosts under

414 conditions of evolutionary pressure.

415

16 bioRxiv preprint doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

416 Data availability

417 Coordinates and structure factors have been deposited in the Protein Data Bank (PDB)

418 under accession numbers 6KL4.

419

420 Acknowledgements

421 This work was supported by the National Natural Science Foundation of China grants

422 31770948, 31570875 (SO) and 31900879 (HG), the National Institutes of Health grant

423 R01AI127465 (ZQL), Marine Economic Development Special Fund of Fujian Province

424 (FJHJF-L-2020-2) and the High-level personnel introduction grant of Fujian Normal

425 University (Z0210509). This work was also supported by the scientific research innovation

426 program “Xiyuanjiang River Scholarship” of the College of Life Sciences, Fujian Normal

427 University. The diffraction data were collected at the beamline BL-17U1 of Shanghai

428 Synchrotron Radiation Facility (SSRF).

429

430 Author contributions

431 SO and ZQL conceived the project. HG, TY and YH crystalized the complexes and

432 determined the structures; YL made the movie; JF and NG constructed and analyzed the

433 mutants; HG, ZQL, JF and SO analyzed the data. HG, JF, S.O., and ZQL wrote the

434 manuscript.

435

17 bioRxiv preprint doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

436 Figure legends

437 Fig. 1 Primary sequence and three-dimensional structure comparison of MavC and

438 its homologs MvcA, Cif and CHBP.

439 A. Domain organization of MavC. The Core domain, Insertion domain and Tail domain

440 (divided into Tail domain-I and -II) are colored green, yellow and blue, respectively.

441 B. Primary sequence alignment of MavC with MvcA, Cif and CHBP generated by ClusterW

442 (https://www.genome.jp/tools-bin/clustalw) and ESpript 3 (http://espript.ibcp.fr/ESPript/

443 ESPript/). Every tenth residue is indicated with a dot (.) above it. Strictly conserved

444 residues are indicated in white on a red background. The yellow triangles indicate the

445 residues of catalytic triad sites. Residues Trp255 of MavC and Phe268 of MvcA proximal to

446 the active site are marked by a red dotted rectangle box and a red hexagon above them.

447 The part of sequence corresponding to the Insertion domain is enclosed by yellow brackets,

448 whereas the parts of sequence corresponding to the Tail domain (-I and -II) are underlined

449 by a blue line.

450 C. Three-dimensional structure comparison of apo MavC (PDB ID:5TSC) with MvcA (PDB

451 ID:6K11 and 6JKY), Cif (PDB ID: 4F8C) and CHBP (PDB ID: 4HCN) in two different

452 orientations.

453

454 Fig. 2 The Insertion domain of MavC is essential for UBE2N ubiquitination but not

455 for ubiquitin deamidation and self-ubiquitination

456 A. Deletion of the Insertion domain of MavC (MavCmid) abolished UBE2N ubiquitination

457 but still had no effect on ubiquitin deamidation and self-ubiquitination. Ubiquitination

458 reactions containing MavC, MavCC74A or MavC∆mid were resolved by SDS-PAGE and

459 visualized by Coomassie staining (left panel) or immunoblotting with MavC-specific

460 antibodies (right panel).

461 B. The MavC∆mid mutant did not catalyze UBE2N ubiquitination in cells infected by L.

18 bioRxiv preprint doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

462 pneumophila. U937 cells were infected with the indicated L. pneumophila strains for 1 h at

463 a MOI of 10, after which the cell lysates separated by SDS-PAGE were probed by

464 immunoblotting with the indicated antibodies. Note that the MavC∆mid mutant was

465 translocated into host cells at levels higher than wild-type bacteria.

466 C. MavC mutant lacking the Insertion domain is capable of ubiquitin deamidation. Proteins

467 in reactions containing ubiquitin and MavC or its mutants were separated by native PAGE

468 and visualized by Coomassie staining.

469

470 Fig. 3 The overall structure of the MavC-UBE2N-Ub complex

471 A-B. Ribbon representation of the MavC-UBE2N-Ub ternary complex. In MavC, the Tail

472 domain (helices α2, α3 and α14, blue) is linked to the Core domain (green) by three loops

473 (loop 1, 2 and 5, red), whereas the Insertion domain (yellow) is linked to the Core domain

474 by two loops (loop 3 and 4, red). Ub and UBE2N are shown in magenta and cyan,

o 475 respectively. The view in panel B is generated by rotating the image in panel A by 180

476 around the indicated axis.

477 C. MavC-induced linkage between Lys92 of UBE2N (cyan) and Gln40 of Ub (magenta).

478 The catalytic triad (yellow) of MavC (Cys74 was mutated to Ala in our structure) and other

479 residues participating in the reaction are shown as sticks. Hydrogen bonds are shown as

480 dashed lines.

481 D. The 2Fo-Fc map of (C) contoured at 1.2 σ . The catalytic triad

482 (Ala(Cys)74-His231-Gln252, yellow) and Ser73 of MavC, Ile90, Leu91, Lys92, Asp93, and

483 Ala94 of UBE2N, and Ile36, Asp39, Gln40 and Gln41 of Ub are shown as sticks.

484

485 Fig. 4 The effects of mutations in MavC on MavC-induced ubiquitination.

486 A-B. Detailed views of the interactions between MavC and UBE2N in the ternary complex.

487 Panel A: Three pairs of hydrogen bonds formed between the Insertion domain of MavC 19 bioRxiv preprint doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

488 and UBE2N are shown as dashed lines, and the hydrogen-bonded residues are labeled

489 with red numbers. Residues involved in electrostatic interactions between the negative

490 surface of MavC and positive surface of UBE2N are labeled with black (MavC) and blue

491 (UBE2N) numbers. The interacting residues are colored yellow (Insertion domain) and

492 cyan (UBE2N). Panel B: The hydrophobic residues forming pocket 2 of UBE2N are shown

493 as cyan sticks, whereas Met317 of MavC that inserts into the pocket is shown as a green

494 stick and labeled by red number (B).

495 C. The hydrophobic pocket of UBE2N (designated pocket 2) into which Met317 of MavC

496 inserts in the ternary complex. The hydrophobic residues lining the pocket (magenta sticks)

497 are labeled with numbers in black and UBE2N is shown as cartoon with partially

498 transparent surface (gray).

499 D-E. Mutational analysis of MavC residues involved in the MavC-UBE2N interaction

500 interface and their importance for UBE2N ubiquitination. MavC or its mutant derivatives

501 were added to reactions containing UBE2N and ubiquitin for 8 min or 30 min at 37°C.

502 Samples resolved by SDS-PAGE and visualized by Coomassie staining were quantified

503 using Image Studio. The ratios are from three independent experiments. Error bars

504 indicate standard error of the mean (SEM).

505 F-G. SDS-PAGE images of the ubiquitination assay shown in panels D-E. MavC and its

506 mutant derivatives were added to reactions containing UBE2N and Ub, the reactions were

507 allowed to proceed for 8 or 30 min at 37°C before SDS-PAGE and visualization with

508 Coomassie staining.

509 H-I. The ability of MavC mutant derivatives to catalyze UBE2N ubiquitination during L.

510 pneumophila infection. Plasmids carrying mavC mutants were introduced into the ∆mavC

511 mutant and the bacteria were used to infect U937 cells. Saponin-soluble lysates of infected

512 cells resolved by SDS-PAGE were probed with antibodies specific for MavC or UBE2N.

513

514 Fig. 5 The effects of mutations in UBE2N on MavC-induced ubiquitination.

20 bioRxiv preprint doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

515 A. Detailed views of the interactions between MavC and UBE2N in the ternary complex,

516 similar to Fig. 4A.

517 B. The effects of mutations in UBE2N on MavC-induced ubiquitination. MavC was added to

518 reactions containing ubiquitin, UBE2N or its mutants. The reactions were allowed to

o 519 proceed for for 8 min or 30 min at 37 C. Samples resolved by SDS-PAGE and visualized

520 by Coomassie staining were quantified using Image Studio. The ratios are from three

521 independent experiments. Error bars indicate standard error of the mean (SEM).

522 C-D. The effects of mutations in UBE2N on MavC-induced ubiquitination. MavC was added

523 to reactions containing ubiquitin, UBE2N or its mutants. The reactions were allowed to

o 524 proceed for for 8 min or 30 min at 37 C before SDS-PAGE and detection by Coomassie

525 staining.

526

527 Fig. 6 Structural details of UBE2N and Ub recognition by MavC.

528 A. Superimposition of Insertion domains of UBE2N-Ub bound and apo MavC. The rotation

529 of the Insertion domain from its position in apo MavC to its position in the ternary complex

530 is indicated by an arrow.

531 B-C. The Insertion domain of MavC in the ternary complex undergoes a rotation during the

532 catalysis that links UBE2N and Ub. Positively charged residues of UBE2N (cyan cartoon)

533 and the negatively charged region of MavC are indicated by a circle (B). Negatively

534 charged residues of the Insertion domain of MavC-UBE2N-Ub (yellow cartoon) and

535 positively charged region of UBE2N (cyan cartoon) are indicated by a circle (C).

536 D-E. Two loops stabilize the binding of UBE2N to MavC. Loop1 (Leu88-Trp95, red cartoon)

537 of MavC is indicated by a circle. Met317 of MavC is shown as red stick which inserts into

538 the pocket of UBE2N electrostatic surface (D). Loop2 (His311-Lys320) of UBE2N is

539 indicated by a circle. Lys92 of UBE2N is shown as cyan stick which inserts into the pocket

540 of ternary complex electrostatic surface (E).

21 bioRxiv preprint doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

541 F. Superimposition of Tail domains of UBE2N-Ub bound and apo MavC. The rotation of the

542 Tail domain from its position in apo MavC to its position in the ternary complex is indicated

543 by an arrow.

544 G. The interacting interface of MavC-UBE2N-Ub Tail domain and Ub are shown as

545 electrostatic surface. Negatively charged Glu42 and positively charged Lys6 are shown as

546 sticks.

547 H. Two loops stabilize the binding of isopetide bond between UBE2N and Ub to MavC. The

548 isopetide bond between UBE2N (cyan) and Ub (magenta) is shown as sticks. Two

549 hydrophobic loops (loop3 and loop4) are shown as red cartoon and labeled on the outer

550 side of the electrostatic surface.

551

552 Fig. 7 Structural comparison and analysis of Insertion domains and catalytic triads

553 of MavC and MvcA.

554 A-B. Superimposition of the catalytic triads (panel A) of Apo-MavC (purple) and Apo-MvcA

555 (gray). Superimposition of the catalytic triads (panel B) of MavC-UBE2N-Ub (colorized) and

556 MavC-UBE2N-Ub (gray) ternary complexes. The catalytic triads of MavC and MvcA are

557 shown as sticks.

558 C. Structure-based sequence alignment of the loop between β6 and α7 that contains

559 negatively charged amino acids. The negatively charged region is boxed by black

560 rectangle, and the key residues involved in the interactions between MavC and UBE2N are

561 labeled by stars.

562 D-E. Structural comparison of Insertion domains of MavC and MvcA. The residues

563 involved in the formation of the negatively charged region are shown as sticks,

564 corresponding residues in MavC and MvcA are labeled by black boxes, and the key

565 residues are highlighted by red boxes (panel D). Superimposition of the Insertion domains

566 of MavC-UBE2N-Ub and MavC-UBE2N-Ub ternary complexes. The Insertion domains of

567 MavC and MvcA are shown as electrostatic surface (panel E). 22 bioRxiv preprint doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

568 F-G. Superimposition of Insertion domains of MavC-UBE2N-Ub ternary complex (colorized)

569 and apo MavC (gray) (panel F), and MvcA -UBE2N-Ub ternary complex (colorized) and

570 apo MvcA (gray) (panel G). The rotation of the Insertion domains from their position in the

571 apo protein to their position in ternary complex is indicated by an arrow.

572

573 Fig. 8 Structural comparison and analysis of Tail domains and catalytic triads of

574 MavC and MvcA

575 A. Superimposition of MavC-UBE2N-Ub (colorized) and apo MavC (gray). The rotations of

576 the Insertion domain and Tail domain from their positions in the apo MavC to their position

577 in the ternary complex are indicated by red arrows.

578 C-E. Electrostatic surfaces of apo MavC Tail domain, MavC-UBE2N-Ub Tail domain and

579 Ub.

580 F-H. Electrostatic surfaces of apo MvcA Tail domain, MvcA -UBE2N-Ub Tail domain and

581 Ub.

582 I-J. MavC and MvcA from their ternary complex are shown as electrostatic surfaces. The

583 isopeptide bond (cyan and magenta) in MavC-UBE2N-Ub and isopeptide bond (light blue

584 and pink) in MavC-UBE2N-Ub are shown as sticks. Trp255 of MavC and Phe268 of MvcA

585 in the active site are shown as ticks and labeled on the outer side of the electrostatic

586 surface.

587

588 Fig. 9 Models for mechanisms underlying MavC trangslutaminase and MvcA

589 deubiquitination activities

590 A-B. UBE2N~Ub binds to MavC in cells (A) and UBE2N-Ub binds to MvcA (B) by rotation

591 of their Insertion domains and Tail domains in directions indicated by red arrows (1-2).

592 Products are released and the catalytic reactions are completed (3).

23 bioRxiv preprint doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

593

594 Fig. S1 Mechanisms of ubiquitination and deubiquitination.

595 A. General scheme of canonical ubiquitination. The product UBE2N-Ub suppresses the

596 activation of NF-κB.

597 B. SidE-catalyzed ubiquitination that is negatively regulated by SidJ/CaM and

598 DupA/DupB-catalyzed deubiquitination.

599 C. Diagrams displaying different chemical structure of the bond between UBE2N and Ub in

600 UBE2N~Ub and UBE2N-Ub.

601 D. MavC-catalyzed transglutamination. The product UBE2N-Ub suppresses the activation

602 of NF-κB.

603 E. MavC-catalyzed deamidation of Ub in solution.

604

605 Fig. S2 Two variants of UBE2N-Ub complexes used for crystallization.

606 A. MavC eluted from peak 2 and peak 3 in gel filtration was incubated with UBE2N at a

607 ratio of 1:2 and purified by size-exclusion chromatography using a Superdex200 increase

608 column. The figure compares retention volumes of MavC (black), MavC (peak2) -UBE2N

609 complex (green) and MavC (peak3) -UBE2N (red) complex.

610 B-D. The elution peaks of MavC and MavC-UBE2N complexes were examined by

611 SDS-PAGE. The word “complex” in the third lane 3 of gel images (C) and (D) is used to

612 denote complex samples before gel filtration.

613

614 Fig. S3 Interactions of MavC and Ub, MavC and UBE2N, and MavC and UBE2N-Ub in

615 vitro.

616 A. Purified MavC was incubated with Ub at a ratio of 2:1 and purified by size-exclusion

24 bioRxiv preprint doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

617 chromatography using a Superdex200 Increase column. Peak 1 (dimeric MavC), peak 2

618 (monomeric MavC) and peak3 (Ub) were examined by SDS-PAGE. The word “complex” in

619 the third lane 3 of gel images is used to denote complex samples before gel filtration.

620 B. The interactions between MavC and UBE2N (panel B), and MavC and C74A C74A 621 UBE2N-Ub (panel C) were measured by microscale thermophoresis (MST).

622

623 Fig. S4 Multiple sequence alignment of MavC and MvcA sequences.

624 MavC and MvcA sequences alignments carried out using MUSCLE v3.8.31 and visualized

625 with Jalview v2.10.3. Residues involved in electrostatic interactions between the negative

626 surface of MavC and positive surface of UBE2N are labeled with cyan. Trp255 of MavC in

627 the active site is labeled with red.

628

629 Fig. S5 Interactions between MavC and Ub in the MavC-UBE2N-Ub complex.

630 A-B. The five contacting interfaces between MavC (green-yellow-blue surface) and Ub

631 (magenta cartoon) are indicated by circles.

632 C-G. Detailed views of contact A (panel C), contact B (panel D), contact C (panel E),

633 contact D (panel F), and the C-terminal contact (CTC) (panel G). Key residues involved in

634 interactions between MavC (green-yellow-blue) and Ub (magenta cartoon) are shown as

635 sticks. Hydrogen bonds are indicated by black dashed lines.

636

637 Fig. S6 Structural alignments of MavC and MvcA in apo form and ternary complex.

638 A. Structural alignment of the ternary complex of MavC-UBE2N-Ub (cyan) and

639 MvcA-UBE2N-Ub (green) with an RMSD of 2.5 Å.

640 B. Structural alignment of the MavC-UBE2N-Ub ternary complex (cyan) and apo MavC

25 bioRxiv preprint doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

641 (yellow) with an RMSD of 3.0 Å.

642 C. Structural alignment of apo MavC (yellow) and apo MvcA (magenta) with an RMSD of

643 3.9 Å.

644 D. Structural alignment of the MvcA-UBE2N-Ub ternary complex (green) and apo MvcA

645 (magenta) with an RMSD of 3.0 Å.

646 E. Structural alignment of three available apo MvcA structures (PDB ID: 5SUJ, 5YM9,

647 6K11) with a maximal RMSD of 0.613 Å.

648

649 Fig. S7 Proposed mechanism for transglutamination and deamidation reaction

650 catalyzed by MavC.

651 In the first step, Cys74 of MavC attacks Gln40 of Ub to generate a thioester intermediate.

652 This intermediate reacts with Lys92 on UBE2N to form UBE2N-Ub or with H2O to form

653 Glu40 in the absence of UBE2N. The ispoeptide bond in UBE2N-Ub complex catalyzed by

654 MavC can be cleaved by MvcA.

655

656

657

658

659

660

26 bioRxiv preprint doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

661 Methods

662 Media, bacteria strains, plasmid constructions and cell lines

663 Legionella strains used in this papers were derivatives of Philadelphia 1 strain Lp02

31 664 and were grown and maintained on CYE (charcoal-yeast extract) plates or in AYE . For

32 665 complementation experiments, the genes were cloned into pZL507 . E.coli strains

666 XL1-Blue grown in Luria broth (LB) were used for expression and purification of all the

667 recombinant proteins. Genes for protein purification were cloned into pQE30 (QIAGEN).

668 Raw 264.7 and U937 cells were cultured in RPIM1640 medium in the presence of 10%

669 FBS. When necessary, U937 cells were differentiated into macrophages by phorbol

33 670 12-myristate 13-acetate as described earlier . All cell lines were directly purchased from

671 ATCC.

672

673 Purification of proteins for biochemical experiments

674 For protein production, 10 mL overnight cultures were transferred to 200 mL LB

675 medium in the presence of 100 µg ampicillin and grown to OD600 nm of 0.6–0.8. The cultures

676 were then incubated at 18°C for 16–18 h after adding isopropyl

677 β-D-1-thiogalactopyranoside (IPTG) to a final concentration of 0.2 mM. Bacterial cells were

678 harvested at 12,000xg by spinning and lysed by sonication. The soluble lysates were

679 cleared by spinning at 12,000xg twice at 4°C for 20 min. His-tagged proteins were purified

2+ 680 with Ni -NTA beads (Qiagen) and eluted with 300 mM imidazole in PBS buffer. Purified

681 proteins were dialyzed with buffer containing 50 mM Tris-HCl (pH7.5), 150 mM NaCl, 5%

682 glycerol, and 1 mM DTT.

683

684 Purification of proteins for structural study

685 The gene of MavC (full length) was PCR amplified from L. pneumophila genomic DNA

686 and inserted into pGEX-6p-1. The gene sequences of MavCC74A (7-384 AA, Cys74 residue

687 mutated to Ala) and UBE2NK94A (full length) were also inserted into pGEX-6p-1. The gene

27 bioRxiv preprint doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

688 sequence of Ub (1-76 AA) was inserted into pQE30. These plasmids were transformed into

689 E. coli BL21(DE3) cells. The cells were grown in LB medium at 37°C with constant shaking

690 at 220 rpm until the cell concentration reached OD600 of 0.8, after which the recombinant

691 protein expression was induced by the addition of IPTG to a final concentration of 0.3 mM.

692 The recombinant proteins were expressed at 18°C for 16 h. The cells were pelleted by

693 centrifugation (5,000xg, 15 min) and subsequently resuspended in the cold lysis buffer (50

694 mM Tris-HCl pH 7.5, 150 mM NaCl). Following lysis by ultrasonication, the lysates were

695 centrifuged at 17,000 rpm for 30 min at 4°C. The proteins with glutathione S-transferase

696 (GST) tag or His tag were purified by affinity chromatography (glutathione agarose resin

2+ 697 and Ni resin, respectively). The GST tag was removed by the PreScission protease

698 (PPase). The tag-less protein was then purified by size-exclusion chromatography (SEC)

699 using a Superdex 200 increase column (GE Healthcare) equilibrated with a buffer

700 containing 25 mM HEPES, pH 7.5, 150 mM NaCl and 2 mM DTT.

701 To prepare the MavCC74A-UBE2NK94A complex, UBE2NK94A was incubated with

702 MavCC74A at 2:1 molar ratio at 4°C for 1 h. To cross-link UBE2NK94A and Ub, wild-type

703 MavC was incubated with UBE2NK94A and Ub at a molar ratio of 1:120:180 in a reaction

2+ 704 buffer containing 25 mM HEPES, pH 7.5, 150 mM NaCl, 2 mM DTT and 10 mM Mg at

705 25°C for 5 min. The UBE2NK94A-Ub was separated from MavC and excess UBE2N by

2+ 21 706 Ni -NTA beads at 4°C and then purified by gel filtration to remove the excess Ub . The

707 UBE2NK94A-Ub was collected and incubated with MavCC74A at 1:2 molar ratio at 4°C for 1 h.

708 Finally, the MavCC74A-UBE2N-Ub complex was separated by gel filtration, pooled and

709 concentrated to A280=14 for the use in crystallization screen.

710

711 Crystallization, data collection and structural determination

712 Crystallization screens were performed using the sitting drop vapor diffusion method at

713 16°C, with drops containing 0.5 µl of the protein solution mixed with 0.5 µl of reservoir

714 solution. Diffraction-quality MavCC74A-UBE2NK94A-Ub crystals were obtained in 0.2 M

715 sodium malonate pH 6.0, 20% w/v PEG 3,350, whereas MavCC74A-UBE2NK94A crystals

716 were obtained in 0.2 M Sodium malonate pH 7.0, 20% w/v PEG 3,350. The crystals were

717 harvested and flash-frozen in liquid nitrogen with 20% glycerol as cryoprotectant. Complete 28 bioRxiv preprint doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

718 X-ray diffraction datasets were collected at BL-17U1 beamline of Shanghai Synchrotron

719 Radiation Facility (SSRF). Diffraction images were processed with the HKL-2000 program

34 720 . Structures of the binary and ternary complex were solved by molecular replacement

23,35-37 721 (MR) using Phaser-MR (MavC: 5TSC, UBE2N: 1JAT, Ub: 4HCN) . Model building

38,39 722 and crystallographic refinement were carried out in Coot and PHENIX . The

723 interactions were analyzed with PyMol (http://www.pymol.org/) and PDBsum. The figures

724 were generated in PyMol. Detailed data collection and refinement statistics are listed in

725 Table 1.

726

727 MavC-induced ubiquitination and deamidation assays

728 For MavC-mediated ubiquitination of UBE2N, 5 µg Ub, 0.5 µg UBE2N or its mutant

729 derivatives and 0.05 µg MavC or its mutant derivatives were incubated for 8 min or 30 min

2+ 730 at 37°C in 25 µl reactions systems containing 50 mM Tris-HCl (pH7.5), 5 mM Mn and 1

731 mM DTT. Reaction products were resolved by SDS-PAGE, protein bands were visualized

732 by Coomassie brilliant blue staining and the amount of protein in bands was quantified by

733 densitometry.

734 For MavC-mediated deamidation of Ub, 10 µg Ub and 1 µg MavC or its mutant

735 derivatives were incubated for 1 h at 37°C in 25 µl reactions containing 50 mM Tris-HCl

736 (pH8.8). Reaction mixtures were then mixed with 5x native gel loading buffer and resolved

737 by native PAGE followed by Coomassie brilliant blue staining.

738

739 MavC-mediated ubiquitination of UBE2N during L. pneumophila infection

740 For L. pneumophila infection experiments, all Legionella strains including

741 complementation strains were grown overnight in AYE medium to post-exponential phase

742 (OD600nm=3.2-3.8) and were induced with 0.2 mM IPTG for 3 h at 37°C before infection.

743 Raw264.7 or U937 cells were infected with L. pneumophila strains at a MOI of 10 for 2 h.

744 Cells washed with PBS three times were collected and lysed with 0.2% saponin on ice for

29 bioRxiv preprint doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

745 30 min. The cell lysates were resolved by SDS-PAGE and probed with MavC-specific

746 antibody to check the translocation of MavC and its mutants, and modified UBE2N was

747 probed with UBE2N-specific antibody.

748

30 bioRxiv preprint doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

749 References

750 1 Mann, M. & Jensen, O. N. Proteomic analysis of post-translational modifications. Nat Biotechnol 21, 751 255-261, doi:10.1038/nbt0303-255 (2003). 752 2 Olsen, J. V. & Mann, M. Status of Large-scale Analysis of Post-translational Modifications by Mass 753 Spectrometry. Mol Cell Proteomics 12, 3444-3452, doi:10.1074/mcp.O113.034181 (2013). 754 3 Hershko, A. & Ciechanover, A. The ubiquitin system. Annual review of biochemistry 67, 425-479, 755 doi:10.1146/annurev.biochem.67.1.425 (1998). 756 4 Song, L. & Luo, Z. Q. Post-translational regulation of ubiquitin signaling. J Cell Biol 218, 1776-1786, 757 doi:10.1083/jcb.201902074 (2019). 758 5 Herhaus, L. & Dikic, I. Expanding the ubiquitin code through post-translational modification. EMBO 759 Rep 16, 1071-1083, doi:10.15252/embr.201540891 (2015). 760 6 Salomon, D. & Orth, K. What pathogens have taught us about posttranslational modifications. Cell 761 Host Microbe 14, 269-279 (2013). 762 7 Burstein, D. et al. Genomic analysis of 38 Legionella species identifies large and diverse effector 763 repertoires. Nat Genet 48, 167-175, doi:10.1038/ng.3481 (2016). 764 8 Ensminger, A. W. Legionella pneumophila, armed to the hilt: justifying the largest arsenal of effectors 765 in the bacterial world. Curr Opin Microbiol 29, 74-80, doi:10.1016/j.mib.2015.11.002 (2016). 766 9 Qiu, J. Z. & Luo, Z. Q. Legionella and Coxiella effectors: strength in diversity and activity. Nat Rev 767 Microbiol 15, 591-605, doi:10.1038/nrmicro.2017.67 (2017). 768 10 Qiu, J. & Luo, Z. Q. Hijacking of the Host Ubiquitin Network by Legionella pneumophila. Front Cell 769 Infect Microbiol 7, 487, doi:10.3389/fcimb.2017.00487 (2017). 770 11 Bhogaraju, S. et al. Phosphoribosylation of Ubiquitin Promotes Serine Ubiquitination and Impairs 771 Conventional Ubiquitination. Cell 167, 1636-1649 e1613, doi:10.1016/j.cell.2016.11.019 (2016). 772 12 Qiu, J. Z. et al. Ubiquitination independent of E1 and E2 enzymes by bacterial effectors. Nature 533, 773 120-124, doi:10.1038/nature17657 (2016). 774 13 Kotewicz, K. M. et al. A Single Legionella Effector Catalyzes a Multistep Ubiquitination Pathway to 775 Rearrange Tubular Endoplasmic Reticulum for Replication. Cell Host Microbe 21, 169-181, 776 doi:10.1016/j.chom.2016.12.007 (2017). 777 14 Clague, M. J., Urbé, S. & Komander, D. Breaking the chains: deubiquitylating enzyme specificity 778 begets function. Nature Reviews Molecular Cell Biology 20, 338-352, 779 doi:10.1038/s41580-019-0099-1 (2019). 780 15 Bhogaraju, S. et al. Inhibition of bacterial ubiquitin ligases by SidJ-calmodulin catalysed glutamylation. 781 Nature 572, 382-386, doi:10.1038/s41586-019-1440-8 (2019). 782 16 Black, M. H. et al. Bacterial pseudokinase catalyzes protein polyglutamylation to inhibit the 783 SidE-family ubiquitin ligases. Science 364, 787-792, doi:10.1126/science.aaw7446 (2019). 784 17 Gan, N. et al. Regulation of phosphoribosyl ubiquitination by a calmodulin-dependent glutamylase. 785 Nature 572, 387-391, doi:10.1038/s41586-019-1439-1 (2019). 786 18 Siepmann, T. J., Bohnsack, R. N., Tokgöz, Z., Baboshina, O. V. & Haas, A. L. Protein Interactions 787 within the N-end Rule Ubiquitin Ligation Pathway. Journal of Biological Chemistry 278, 9448-9457, 788 doi:10.1074/jbc.M211240200 (2003). 789 19 Stewart, M. D., Ritterhoff, T., Klevit, R. E. & Brzovic, P. S. E2 enzymes: more than just middle men. 790 Cell Research 26, 423-440, doi:10.1038/cr.2016.35 (2016). 791 20 Gan, N., Nakayasu, E. S., Hollenbeck, P. J. & Luo, Z. Q. Legionella pneumophila inhibits immune

31 bioRxiv preprint doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

792 signalling via MavC-mediated transglutaminase-induced ubiquitination of UBE2N. Nat Microbiol 4, 793 134-143, doi:10.1038/s41564-018-0282-8 (2019). 794 21 Gan, N. et al. Legionella pneumophila regulates the activity of UBE2N by deamidase-mediated 795 deubiquitination. EMBO J, e102806, doi:10.15252/embj.2019102806 (2019). 796 22 Eckert, R. L. et al. Transglutaminase regulation of cell function. Physiol Rev 94, 383-417, 797 doi:10.1152/physrev.00019.2013 (2014). 798 23 Valleau, D. et al. Discovery of Ubiquitin Deamidases in the Pathogenic Arsenal of Legionella 799 pneumophila. Cell reports 23, 568-583, doi:10.1016/j.celrep.2018.03.060 (2018). 800 24 Cui, J. et al. Glutamine deamidation and dysfunction of ubiquitin/NEDD8 induced by a bacterial 801 effector family. Science 329, 1215-1218, doi:10.1126/science.1193844 (2010). 802 25 Yao, Q. et al. Structural mechanism of ubiquitin and NEDD8 deamidation catalyzed by bacterial 803 effectors that induce macrophage-specific apoptosis. P Natl Acad Sci USA 109, 20395-20400, 804 doi:10.1073/pnas.1210831109 (2012). 805 26 Shin, D. et al. Regulation of Phosphoribosyl-Linked Serine Ubiquitination by Deubiquitinases DupA 806 and DupB. Molecular Cell 77, 164-179.e166, doi:https://doi.org/10.1016/j.molcel.2019.10.019 (2020). 807 27 Li, W., Tu, D., Brunger, A. T. & Ye, Y. A ubiquitin ligase transfers preformed polyubiquitin chains from 808 a conjugating enzyme to a substrate. Nature 446, 333-337, doi:10.1038/nature05542 (2007). 809 28 Chen, Y. et al. Cullin Mediates Degradation of RhoA through Evolutionarily Conserved BTB Adaptors 810 to Control Actin Cytoskeleton Structure and Cell Movement. Molecular Cell 35, 841-855, 811 doi:10.1016/j.molcel.2009.09.004 (2009). 812 29 Losick, V. P., Haenssler, E., Moy, M.-Y. & Isberg, R. R. LnaB: a Legionella pneumophila activator of 813 NF-kappaB. Cell Microbiol 12, 1083-1097, doi:10.1111/j.1462-5822.2010.01452.x (2010). 814 30 Losick, V. P. & Isberg, R. R. NF-kappaB translocation prevents host cell death after low-dose 815 challenge by Legionella pneumophila. The Journal of experimental medicine 203, 2177-2189, 816 doi:10.1084/jem.20060766 (2006). 817 31 Berger, K. H. & Isberg, R. R. Two distinct defects in intracellular growth complemented by a single 818 genetic locus in Legionella pneumophila. Mol Microbiol 7, 7-19, 819 doi:10.1111/j.1365-2958.1993.tb01092.x (1993). 820 32 Xu, L. et al. Inhibition of host vacuolar H+-ATPase activity by a Legionella pneumophila effector. 821 PLoS Pathog 6, e1000822, doi:10.1371/journal.ppat.1000822 (2010). 822 33 Tilney, L. G., Harb, O. S., Connelly, P. S., Robinson, C. G. & Roy, C. R. How the parasitic bacterium 823 Legionella pneumophila modifies its phagosome and transforms it into rough ER: implications for 824 conversion of plasma membrane to the ER membrane. J Cell Sci 114, 4637-4650 (2001). 825 34 Otwinowski, Z. & Minor, W. Processing of X-ray diffraction data collected in oscillation mode. 826 Methods Enzymol 276, 307-326 (1997). 827 35 VanDemark, A. P., Hofmann, R. M., Tsui, C., Pickart, C. M. & Wolberger, C. Molecular insights into 828 polyubiquitin chain assembly: Crystal structure of the Mms2/Ubc13 heterodimer. Cell 105, 711-720, 829 doi:Doi 10.1016/S0092-8674(01)00387-7 (2001). 830 36 McCoy, A. J. Solving structures of protein complexes by molecular replacement with Phaser. Acta 831 Crystallogr D 63, 32-41, doi:10.1107/S0907444906045975 (2007). 832 37 Yao, Q. et al. A bacterial type III effector family uses the papain-like hydrolytic activity to arrest the 833 host cell cycle. Proc Natl Acad Sci U S A 106, 3716-3721, doi:10.1073/pnas.0900212106 (2009). 834 38 Adams, P. D. et al. PHENIX: a comprehensive Python-based system for macromolecular structure 835 solution. Acta Crystallogr D Biol Crystallogr 66, 213-221, doi:10.1107/S0907444909052925 (2010).

32 bioRxiv preprint doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

836 39 Emsley, P., Lohkamp, B., Scott, W. G. & Cowtan, K. Features and development of Coot. Acta 837 Crystallogr D Biol Crystallogr 66, 486-501, doi:10.1107/S0907444910007493 (2010). 838 839 840

33 Fig. 1

A Core domain

1 Tail domain-I Insertion domain Tail domain-II 482

bioRxiv preprintN doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint C (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Construct for crystallization B Fig. 2

A Tail domain-I

75 MavC*Ub 50 75 MavC/ 37 MavC *Ub 50 ∆mid

UBE2N-Ub MavC∆mid 25 37 20 IB: MavC UBE2N Insertion domain 10 B C L. Pneumophila

2 UBE2N-Ub 25 Ub 0 IB: UBE2N UBE2N 1 MavC∆mid 5 MavC

5 UbQ40E IB: MavC 0 3 75 IB: Tubulin 0 Translocation Tail domain-II

Insertion domain C 5TSC: apo MavC C-ter

90°

6K11: apo MvcA 4F8C: Cif-NEDD8 6JKY: MvcA-UBE2N-Ub 4HCN: CHBP-Ub N-ter Fig. 3

C D Fig. 4

A C D E UBE2N

F G

MavC MavC

UBE2N-Ub UBE2N-Ub

UBE2N UBE2N Ub Ub 8 min 30 min B H I Fig. 5

R6 E207 K10 Y189 Y192 R14 Y198 R7 K132 E197 Q100

B C D

MavC MavC

UBE2N-Ub UBE2N-Ub

UBE2N UBE2N Ub Ub

8 min 30 min Fig. 6

bioRxiv preprint doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint (which was not certified by peer Areview) is the author/funder, who has granted bioRxiv a license to displayB the preprint in perpetuity. It is C made available under aCC-BY 4.0 International license. negative negative region positive positive region

D E

M317 loop1 loop2

MavC-UBE2N-Ub (ternary complex) K92 Apo-MavC F G H

loop3 loop4 E42 K92 K6 Q40 Fig. 7 A B H231 H244

Apo-MavC Q40 Apo-MvcA K92

MavC-UBE2N-Ub (ternary complex) MvcA-UBE2N-Ub (ternary complex) C

D E

MavC-Insertion domain:UBE2N Apo-MavC Apo-MvcA

MvcA-Insertion domain:UBE2N

F G

MavC-UBE2N-Ub (ternary complex) Apo-MavC

MvcA-UBE2N-Ub (ternary complex) Apo-MvcA Fig. 8

MavC-UBE2N-Ub (ternary complex) Apo-MavC:TD MavC-UBE2N-Ub:TD MavC-UBE2N-Ub:Ub Apo-MavC B F G H

MvcA-UBE2N-Ub (ternary complex) Apo-MvcA:TD MvcA-UBE2N-Ub:TD MvcA-UBE2N-Ub:Ub Apo-MvcA I J

Q40 Q40

K92 K92 W255 F268 Fig. 9

Core Tail domain Insertion UBE2N Core Ub domain domain domain

Ub Ub

MavC UBE2N~Ub UBE2N-Ub (In vivo)

B 1 2 3

Core domain Iinsertion Ub Core Tail domain domain domain UBE2N + Ub Ub

MvcA UBE2N-Ub UBE2N Ub Fig. S1

bioRxiv preprint doi: https://doi.org/10.1101/2020.03.10.984922; this version posted March 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is A made available under aCC-BY 4.0 International license. NF-κB UBE2N E1 E2 E3 UBE2N + Ub activation ATP Canonical ubiquitination UBE2N-Ub

CaM B SidJ

substrate SidEs + Ub substrate NAD+ Ub SidE-catalyzed ubiquitination PR-Ub-sustrate

+H O substrate 2 + Ub PR-Ub C MavC UBE2N Ub UBE2N + Transglutamination Ub

UBE2N-Ub NF-κB Ub D UBE2N UBE2N~Ub activation

E Ub MavC + H O Ub 2 Deamidation + NH3 Fig. S2

A peak 1 peak 2 peak 2

MavC

peak 1 peak 3 Ub

B C

MavCC74A: UBE2N MavCC74A: UBE2N-Ub

Kd: 1.14 μM Kd: 1.4 μM Fig. S4

MavC

MvcA Fig. S5

MavC of MavC-UBE2N-Ub MavC of MavC-UBE2N-Ub MvcA of MvcA-UBE2N-Ub Apo-MavC RMSD: 2.5 Å RMSD: 3.0 Å C D

Apo-MavC MvcA of MvcA-UBE2N-Ub Apo-MvcA RMSD: 3.9 Å Apo-MvcA RMSD: 4.5 Å E

5SUJ 5YM9 6K11 Maximal RMSD of 0.613 Å Fig. S7

UBE2N-Ub UBE2N

MavC

MavC MvcA Thioester intermediate

H20