1 The ER-localised Hrd1 ubiquitinates and inactivates Usp15 to promote

2 TLR4-induced inflammation during bacterial infection

3 Yao Lu1*, Ying Qiu1*, Peng Chen2, Haishuang Chang3, Luqiang Guo3, Fang Zhang4, Li 4 Ma1, Chi Zhang1, Xin Zheng1, Jun Xiao1, Ruiyue Zhong1, Lei Han1, Xiaoyan Xu1,5, 5 Yanbo Zhang1,6, Dangsheng Li1, Guisheng Zhong7, Rosemary Boyton8, Ying Huang3, 6 Yongning He3, Ronggui Hu2#, Bin Wei4,9#, Hongyan Wang1,10#

7 1 State Key Laboratory of Cell Biology, Key Laboratory of Systems Biology, CAS 8 Center for Excellence in Molecular Cell Science, Shanghai Institute of Biochemistry and 9 Cell Biology, Chinese Academy of Sciences, University of Chinese Academy of Sciences, 10 Innovation Center for Cell Signaling Network, Shanghai, 200031, China; 2 State Key 11 Laboratory of Molecule Biology, Key Laboratory of Systems Biology, CAS Center for 12 Excellence in Molecular Cell Science, Shanghai Institute of Biochemistry and Cell 13 Biology, Chinese Academy of Sciences, University of Chinese Academy of Sciences, 14 Shanghai, 200031, China; 3 State Key Laboratory of Molecular Biology, National Center 15 for Protein Science Shanghai, Shanghai Science Research Center, Shanghai Key 16 Laboratory of Molecular Andrology, CAS Center for Excellence in Molecular Cell 17 Science, Shanghai Institute of Biochemistry and Cell Biology, Chinese Academy of 18 Sciences, University of Chinese Academy of Sciences, Shanghai, 201210; 4 Wuhan 19 Institute of Virology, Chinese Academy of Sciences, Wuhan, China; 5 Experimental 20 Immunology Branch, National Cancer Institute, US National Institutes of Health, 21 Bethesda, Maryland, USA; 6 Division of Immunology, Department of Microbiology and 22 Immunobiology, Harvard Medical School, 77 Avenue Louis Pasteur, Boston, MA 02115, 23 USA; 7 iHuman Institute, School of Life Science and Technology, ShanghaiTech 24 University, Shanghai, China; 8 Lung Immunology Group, Department of Infectious 25 Diseases, Faculty of Medicine, Imperial College London, London W12 0NN, United 26 Kingdom; 9 School of Life Sciences, Shanghai University, Shanghai 200444; 10 Cancer 27 Center, Shanghai Tenth People's Hospital, Tongji University, School of Medicine, 28 Shanghai 200072, China

29 1

30 *Contributed equally; #Corresponding author: [email protected]; 31 [email protected]; [email protected]

32 Running title: Hrd1 enhances TLR4 pathway upon infection

33 Keywords: Hrd1, Ubiquitination, Usp15, TLR4 pathway, Inflammation and Infection

34

35 ABSTRACT

36 The special organelle-located MAVS, STING and TLR3 are important for clearing viral

37 infections. Although TLR4 triggers NF-κB activation to produce proinflammatory

38 cytokines for bacteria clearance, effectors with special organelle localisation have not

39 been identified. Here, we screened over 280 E3 ubiquitin ligases and discovered that the

40 endoplasmic reticulum-located Hrd1 regulated TLR4-induced inflammation during

41 bacterial infection. Hrd1 directly interacted with the deubiquitinating enzyme (DUB)

42 Usp15. Unlike the classical function of Hrd1 in ER-associated degradation, Usp15 was

43 not degraded but lost its DUB activity for IκBα deubiquitination, resulting in excessive

44 NF-κB activation. Importantly, Hrd1 deficiency in macrophages protected mice against

45 LPS-induced septic shock, and knock-down of Usp15 in Hrd1 KO macrophages restored

46 the reduced IL-6 production. This study has proposed the crosstalk between Hrd1 and

47 TLR4 linking the ER-plasma membrane function during bacterial infection.

48

2

49 Introduction

50 Macrophages express pattern recognition receptors (PRRs) such as Toll-like

51 receptors (TLRs) and RIG-like receptors (RLRs) to sense pathogen-associated molecular

52 patterns (PAMPs) and trigger the innate immune response, leading to inflammation and

53 microbe clearance.1 The mitochondria-located MAVS (Mitochondrial antiviral-signaling

54 protein, also named VISA (virus-induced signaling adapter), IPS-1 and Cardif)2, 3, 4, 5, the

55 endoplasmic reticulum (ER)-located STING (Stimulator of interferon genes)6, 7, 8 and the

56 endosome-located TLR3 (Toll-like receptor 3)9 are important for type I IFN production

57 and clearance of viral infections4. Although the surface receptor TLR4-triggered NF-κB

58 activation is well studied for proinflammatory cytokine production and bacteria clearance,

59 downstream effectors with special organelle localisation have not been identified in the

60 TLR4 pathway. In recent years, studies of membrane contact sites within cells and their

61 role have been rapidly advancing, in particular insights have recently demonstrated the

62 contact sites exist and function between the largest organelle endoplasmic reticulum (ER)

63 and other organelles for cell homeostasis or disease pathogenesis by sensing the intra- or

64 extracellular stimulation 2, 3, 6, 10, 11.

65 While a balanced inflammatory response is pivotal to protecting the host against

66 microbes and self-injury, excessive activation of the TLR or RLR signalling pathway

67 could lead to serious inflammatory diseases, including septic shock or autoimmune

68 diseases. Septic shock is the most common cause of death in hospitalized patients and the

69 proinflammatory cytokine IL-6 is crucial in the pathophysiology of severe sepsis, and

70 IL-6 levels most significantly correlate with mortality rates compared to other cytokines12.

71 Accumulating studies have shown that E3 ubiquitin ligases are involved in TLR

3

72 signalling. TRAF6 (TNF receptor associated factor 6), a typical RING-type E3, can be

73 autoubiquitinated at Lys124, which is then recognized by the TAB (TGF-β activated

74 kinase 1 binding protein) 2/3 complex to further activate TAK1 (TGF-β activated kinase

75 1) and NEMO (NF-κB essential modulator). Pellino-1 increases LPS-driven Lys63-linked

76 polyubiquitination of IRAK1, TBK1 (TANK binding kinase 1) and TAK1 in TLR

77 signalling. Since E3 ligases and their substrates can be targeted to attenuate excessive

78 inflammation and sepsis, we were interested in investigating whether any E3 ligases with

79 special organelle localisation could be identified in the TLR4 pathway.

80 We screened over 280 E3 ligases using Dharmacon RNAi Screening Libraries and

81 identified Hrd1 as a RING-type E3 ubiquitin ligase, which positively regulated IL-6

82 production in LPS-treated macrophages. Hrd1, a homologue of yeast Hrd1p/Der3p13,

83 contains a transmembrane domain and is specifically located in the endoplasmic

84 reticulum (ER). The best-defined function of Hrd1 is to ubiquitinate misfolded/unfolded

85 proteins with help from other proteins in the ER-associated degradation (ERAD)

86 complex14, 15, 16, 17, which protects cells from ER-stress-induced apoptosis18. In agreement

87 with this role, Hrd1 (also known as Synoviolin, Syvn1) expression is enhanced in

88 synovial fibroblasts from rheumatoid arthritis (RA) patients19, and Hrd1+/− mice are

89 resistant to collagen-induced arthritis due to increased synovial cell apoptosis20, 21.

90 Previous studies have demonstrated that TLR4 is highly expressed in RA synovial tissue

91 lining and sublining macrophages22, and excessive levels of TNF-α and IL-6 accelerate

92 RA development. However, no knowledge is available about how Hrd1 affects

93 TLR4-induced inflammation. More importantly, the ER has a broad localisation

94 throughout the cell and can form direct physically contacts with the cell membrane11, 23.

4

95 We found that macrophages enhance ER membranes upon bacterial infection. Therefore,

96 it was interesting to further elucidate how the ER-located Hrd1 participates in

97 TLR4-induced inflammation in macrophages during bacterial infection.

98 This study has identified an ERAD-independent function of Hrd1 to increase

99 TLR4-induced proinflammatory cytokine production. We fished out Usp15

100 (Ubiquitin-specific protease 15) as a binding partner for Hrd1. Usp15 is a member of the

101 largest subfamily of cysteine protease DUBs (deubiquitinating enzymes). Prior studies

102 have indicated that Usp15 promotes cell survival by stabilizing IκBα in TNF-α stimulated

103 HeLa cells24, and Usp15 promotes type I interferon responses and pathogenesis during

104 neuroinflammation25. Other studies have demonstrated Usp15 function in anti-tumor

105 response, including that Usp15 regulates p53 function to promote cancer-cell survival and

106 Usp15 inhibits T cell activation and immune surveillance26. However, it remains unclear

107 about how Usp15 affects TLR4-induced inflammation. This study has demonstrated that

108 Hrd1 promoted polyubiquitination of Lys21 in Usp15. Unlike other Hrd1 substrates,

109 ubiquitinated Usp15 was not degraded, but rather lost its DUB activity and failed to

110 deubiquitinate IκBα, which resulted in excessive TLR4-NF-κB activation.

111 Results

112 RNAi screening identifies the ER-localised Hrd1 to positively regulate inflammation

113 in LPS-stimulated macrophages

114 To identify E3 ligases involved in the regulation of TLR4-signalling, we transfected

115 over 280 siRNAs (Dharmacon RNAi Screening Libraries) into murine primary peritoneal

116 exudate macrophages (PEMs), followed by LPS stimulation for 6 hr. IL-6 concentrations

5

117 in supernatants were measured by ELISA (Figure S1a).We controlled macrophage

118 survival by checking the am-blue absorbance and excluded genes that significantly

119 induced macrophage death (indicated by blue colour in Figure 1a). Approximately 20

120 candidate genes were identified to affect IL-6 production, including Birc2 and Traf5 that

121 were previously identified in regulating inflammation (Figure 1a)27, 28. We next verified

122 the functions of several candidate genes, including Hrd1, using single siRNAs in smart

123 pools. In order to select our most interested genes for further investigation, we treated

124 PEMs with LPS-coated latex beads. Interestingly, electron microscopy analysis showed

125 that macrophages displayed enhanced ER membrane upon stimulated with LPS-coated

126 latex beads (Figures 1b and S1b). We next checked how ER might be affected in

127 response to bacterial infection in MEFs, which were transiently transfected with

128 KDEL-mCherry to label ER. After infected with Salmonella typhimurium (strain

129 SL1344), mCherry formed aggregation in response to bacterial infection (Figure 1c).

130 ER is the largest organelle in the cell which forms an interconnected network with

131 almost every membrane-bound organelles, including cell membrane and mitochondria29,

132 30. Advanced studies have suggested that ER is essential for cell homeostasis or disease

133 pathogenesis by sensing the intra- or extracellular stimulation through contact sites with

134 other organelles.31 Despite that Hepatitis B virus (HBV) might induce ER dysfunctions

135 that leads to liver injury32; and the ER-located STING participates in the innate immune

136 response against viruses7, little is known about how ER functions in bacterial infection

137 and no many effectors with ER localisation have not been identified in the TLR4 pathway.

138 We therefore took this advantage to further elucidate how the ER-located Hrd1

6

139 cooperates with the cell membrane receptor TLR4 to regulate inflammation in vivo and in

140 vitro.

141 Knock-down (KD) of Hrd1 using 4 different siRNAs reproducibly reduced the

142 mRNA levels of Il6 and Tnfa in LPS-stimulated PEMs (Figure 1d, KD efficiencies of the

143 4 different siHrd1 were shown in Figure S1c). Similarly, when infected with

144 Gram-negative bacteria, such as Salmonella typhimurium (strain SL1344) and

145 Escherichia coli (E. coli), Hrd1-silenced macrophages produced much lower levels of Il6

146 and Tnfa (Figure 1e). We also confirmed decreased concentrations of IL-6 and TNF-α by

147 ELISA in Hrd1-silenced PEMs after LPS stimulation (Figure 1f, siHrd1 KD efficiency is

148 shown in Figure S1d.). In agreement with these results, stable overexpression of Hrd1 in

149 human THP-1 cells or primary mouse embryonic fibroblasts (MEFs) significantly

150 increased LPS-induced Il6 production at the mRNA level (Figure 1g).

151 We then performed the luciferase reporter assay to determine how Hrd1 cooperates

152 with MyD88-dependent and -independent pathways to promote TLR4-NF-κB activation.

153 The key TLR4 signalling effectors were transfected alone or together with Hrd1 into

154 HEK293T cells to measure luciferase readings and the ER location of overexpressed

155 Hrd1 was determined by Western blot (Figure S1e). Hrd1 co-expression cooperated with

156 MyD88, TRAF6, IKKα/β (Inhibitor of nuclear factor kappa B kinases alpha/beta) and

157 TRIF (Toll/IL-1 receptor domain-containing adaptor) to substantially further increase

158 luciferase readings (Figure 1h, middle and right panels) compared to the effect mediated

159 by transfecting the TLR4 signalling effectors alone. Furthermore, this enhancement was

160 shown to be Hrd1 dose-dependent when either TRAF6 or TRIF was co-transfected

161 (Figure 1i, Hrd1 levels were shown in Figure S1f). In contrast, Hrd1 did not further

7

162 increase luciferase readings when it was co-expressed with p65 in HEK293T cells

163 (Figure 1h), suggesting that Hrd1 was located upstream of p65 and downstream of

164 IKKα/β. We also noted that overexpression of Hrd1 alone slightly enhanced luciferase

165 readings (Figure 1h, left panel), but this effect was very minor compared to those

166 mediated by transfecting MyD88, TRAF6, IKKα, IKKβ, TRIF or p65 alone (Figure 1h,

167 white columns).

168 To elucidate whether the ER-localisation of Hrd1 was critical for NF-κB activation,

169 we truncated the transmembrane domain of Hrd1, termed ΔTM (Figure 1j, left and lower

170 panel). Using the NF-κB luciferase system, we indeed found that the Hrd1 ΔTM mutant

171 failed to increase luciferase readings compared to WT Hrd1 (Figure 1j, left and top panel).

172 To better explore the importance of the ER location of Hrd1 in the regulation of

173 inflammation, Hrd1 was targeted to mitochondrial membrane (termed mito-Hrd1) (Figure

174 S1g & Figure 1j, right and lower panel). Similar with the ΔTM truncation, mito-Hrd1

175 failed to increase NF-κB luciferase readings compared to WT Hrd1 (Figure 1j, right panel)

176 and could not promote pro-inflammatory cytokines expression in MEFs (Figure 1k, the

177 expression level of Hrd1 was shown in right panel). These results taken together, we have

178 identified that the ER-located E3 ligase Hrd1 enhances TLR4-induced inflammation in

179 macrophages.

180

181 Hrd1 KO macrophages reduce TLR4-induced inflammation and NF-κB activation

182 To better understand Hrd1 function, we generated Hrd1 conditional knock-out mice.

183 Exon 6 in the Hrd1 gene was floxed, resulting in a stop codon after 125 amino acids

8

184 (named Hrd1fl/fl, Figure S2a). Hrd1fl/fl mice were crossed with Lysozyme M (LysM)-Cre

185 mice to generate conditional knock-out (cKO) mice lacking Hrd1 expression in myeloid

186 cells. The efficiency of Hrd1 KO in macrophages was confirmed by western blot (Figure

187 2a). Hrd1 deficiency did not affect the development of bone marrow cell-derived

188 macrophage, which showed normal mRNA expression levels of Adgre1 (Adhesion G

189 Protein-Coupled Receptor E1, also known as F4/80) and MerTK (Mer tyrosine kinase) as

190 well as normal percentages of F4/80+CD11b+ macrophages (Figure S2b). In addition,

191 when bred in an SPF facility, Hrd1 cKO mice showed normal T cell development in the

192 thymus and normal percentages of B220+, CD4+, CD8+ lymphocytes and F4/80+, CD11b+

193 macrophages in the spleen (Figure S2c). Although Hrd1 is associated with

194 ER-stress-induced cell apoptosis20, Hrd1-deficient macrophages did not exhibit reduced

195 cell viability after being challenged by E. coli infection (Figure S2d). Also Hrd1

196 deficiency did not modulate LPS- and ATP-induced cleavage of Caspase1 (Figure S2e).

197 This result was in agreement with the unchanged cell survival we observed after

198 knock-down of Hrd1 in our siRNA library screening system (Figure 1a).

199 Consistent with our luciferase data, Hrd1-deficient PEMs exhibited no differences in

200 IKKα/β phosphorylation (Figure 2b, left panel). Upon IKKα/β phosphorylation and

201 activation, IκBα is phosphorylated then polyubiquitinated for subsequent degradation;

202 this results in nuclear entry of NF-κB to turn on target gene expression33. Indeed, we

203 observed that Hrd1-deficient macrophages degraded IκBα less efficiently (Figure 2b,

204 right panel) and less p65 translocation into the nucleus decreased (Figure 2c) upon LPS

205 stimulation. Knock-out or knock-down of Hrd1 in PEMs did not change phosphorylation

206 levels of JNK, ERK or p38 (Figures 2d and S2f). Similar to the data from our Hrd1

9

207 knock-down experiments, Hrd1-deficient PEMs and bone marrow derived macrophages

208 (BMMs) also exhibited reduced expression of Il6, Tnfa and Il1b at the mRNA levels

209 (Figures 2e, Figures S2g) or concentrations of IL-6 and TNF-α (Figure 2f) after treatment

210 with LPS, Salmonella typhimurium or E. coli. To determine whether NF-κB is the major

211 transcription factor downstream of Hrd1, we next blocked NF-κB activation by

212 knock-down of the NF-κB subunit p65 or by using the inhibitor PDTC (pyrrolidine

213 dithiocarbamate) (Figures 2g, S2i, sip65 KD efficiency is shown in Figure S2h).

214 Knock-down of p65 (Figure 2g) and PDTC treatment (Figure S2i) both profoundly

215 suppressed IL-6 expression to similar levels in WT and Hrd1 KO macrophages.

216 Interestingly, unlike the LPS/TLR4-induced response, Hrd1 did not modulate CpG

217 (TLR9 stimuli)-induced IL-6 production, Poly(I:C) (TLR3 stimuli)-induced IFN-β

218 production (Figure 2h) and PGN (TLR2 stimuli)-induced pro-inflammatory cytokine

219 production (Figure S2j). Collectively, we have demonstrated that Hrd1 specifically

220 regulates TLR4-induced IκBα degradation to increase NF-κB activation and

221 proinflammatory cytokine production in macrophages, without significantly affecting

222 TLR2/TLR3/TLR9 signalling.

223

224 The E3 ligase activity of Hrd1 is critical for TLR4-induced NF-κB activation

225 independent of ERAD

226 As an ER-located E3 ligase, the classical physiological function of Hrd1 is to

227 catalyse addition of ubiquitin molecules to lysine (Lys) residues in substrate proteins to

228 promote their degradation. After we confirmed that the ER location of Hrd1 is critical for

10

229 TLR4 signalling (Figure 1j), we next assessed whether the E3 ligase activity of Hrd1 was

230 also required for modulating TLR4 signalling. Hrd1 contains 8 conserved cysteine and

231 histidine residues in the core of its RING domain that maintain its three-dimensional

232 structure. Based on previous reports, the cysteine residue at position 329 is critical, and

233 its mutation to serine (C329S) abolishes Hrd1 E3 activity18. Notably, the catalytic

234 inactive mutant C329S failed to elevate NF-κB luciferase activity when co-expressed

235 with MyD88, TRAF6 or TRIF (Figure 3a, similar overexpression levels of Hrd1 and

236 C329S were shown in the right panel). In agreement with this result, stable

237 overexpression of the C329S mutant in THP-1 cells (Figure S3a) decreased Lys48-linked

238 polyubiquitination of IκBα when compared to that seen in THP1 cells stable

239 overexpressing Hrd1 (Figure 3b). Stable overexpression of the C329S mutant or the

240 mitochondrial located Hrd1 mutant (i.e. mito-Hrd1) in MEFs could not promote

241 LPS-triggered IκBα degradation as effectively as those in MEFs expressing WT Hrd1

242 (Figure 3c). The expression levels of Hrd1 and its mutant were shown in Figure S3b and

243 Figure 1k, right panel. We performed immunoprecipitation experiments and found that

244 Hrd1 or the C329S mutant did not bind IκBα (Figure S3c), which does not support the

245 possibility that Hrd1 might directly ubiquitinate IκBα for degradation. Also Hrd1 or the

246 C329S mutant could not interact with TLR4, Myd88, TRIF or TBK1 (Figure S3c, left

247 panel), which indeed bound Hrd3 (Figure S3c, right panel). We next purified the nucleus

248 and cytoplasm fractions, and the C329S mutant reduced p65 levels in the nuclei of THP-1

249 cells (Figure 3d). Using immunofluorescence strategies, we also confirmed that stable

250 overexpression of the C329S mutant decreased p65 translocation into the nucleus in

11

251 MEFs (Figure 3e, the relative amount of p65 in the nucleus was quantified in the right

252 panel).

253 Importantly, we reconstituted Hrd1-deficient BMMs to stably overexpress WT Hrd1,

254 C329S, ΔTM mutant or a GFP control using a retrovirus transduction system and similar

255 overexpression levels were confirmed by FACS analysis (Figure S3d, about 20% cells

256 were successfully transduced). Partially reconstituted WT Hrd1 expression, but not the

257 catalytic inactive mutant C329S or the transmembrane domain deletion mutant ΔTM,

258 could rescue IL-6 production in some degree in Hrd1-deficient BMMs (Figure 3f).

259 Hrd1, a core component in regulating ERAD14, 34, was recently elucidated to form a

260 dimer with Hrd3 (also named Sel1L) during ERAD35. To determine whether Hrd1- or

261 Hrd3-dependent ERAD was involved in TLR4-signalling, we used two siRNAs to knock

262 down Hrd3 expression in PEMs (Figure S3e). Knock-down of Hrd3 (Figure S3e) or Hrd1

263 KO PEMs (Figure 3g) significantly promoted tunicamycin-induced expression of

264 unfolded protein response (UPR)-target genes such as ERdj4 (Dnajb9), Bip (Hspa5),

265 CHOP (Ddit3) and the splicing of Xbp1.Unexpectedly, knock-down of Hrd3 for 36 hr in

266 PEMs (Hrd1 expression was not affected at this time point: Figure S3f, right panel) did

267 not significantly affect LPS-induced Il6 production at the mRNA level (Figure S3f). We

268 further investigated whether Hrd1 deficiency affects expression of ER stress-associated

269 proteins upon triggering TLR4 signalling. In agreement with other reports36, we observed

270 that comparing to tunicamycin treatment, both LPS stimulation and Gram-negative

271 bacteria infection minimally affected or even reduced the ER stress response in WT

272 control cells (Figure 3g and Figure S3g). In addtion, in response to LPS or bacteria

273 treatment, Hrd1-deficient (Figure 3g) or Hrd1 knock-down (Figure S3g) macrophages did

12

274 not exhibit significantly altered expression levels of ER stress-associated genes. Together,

275 we have elucidated that Hrd1 depends on its E3 ligase catalytic activity and its ER

276 location to accelerate TLR4-induced IκBα degradation and p65 translocation, which

277 might be independent of ERAD.

278

279 The ER-localised Hrd1 directly binds Usp15 and regulates TLR4-induced

280 inflammation

281 Because Hrd1’s E3 ligase activity is critical for TLR4-NF-κB activation, we next

282 needed to determine its key substrate. To answer this question, human THP-1 cells stably

283 expressing Flag-tagged Hrd1 were stimulated with LPS, after which immunoprecipitation

284 using anti-Flag antibodies was performed for further mass spectrometry (MS) analysis. In

285 addition to proteins previously reported to participate in ERAD (i.e. Hrd3), our MS

286 analysis also identified interesting candidates including kinases, helicases and adaptor

287 proteins. We confirmed by immunoprecipitation assay that some candidate such as

288 WDR1 indeed interacted with Hrd1 in HEK293T cells, which however did not affect

289 NF-κB activation in the luciferase reporter assay (Figure S4a). After initial screening and

290 validation, we identified Usp15 (Ubiquitin specific peptidase 15) as an interesting

291 candidate for three reasons: First, Usp15 is a classical deubiquitinating enzyme (DUB)

292 and only a few studies to date have reported a DUB itself to bind and be ubiquitinated by

293 an E3 ligase. Second, Usp15 showed inhibitory effect on NF-κB activation when

294 co-transfected with MyD88 in HEK293T cells (Figure S4a). Third, Usp15 was previously

295 suggested to deubiquitinate IκBα in TNF-α-stimulated HeLa cells24, and, interestingly,

296 we observed a link between Hrd1 E3 ligase activity and IκBα degradation upon LPS 13

297 treatment in macrophages (Figures 2b and 3c). We therefore investigated whether Usp15

298 is the downstream effector through which the ER-located Hrd1 regulates IκBα

299 degradation.

300 To confirm the interaction between Hrd1 and Usp15, a yeast two-hybrid system was

301 used. With the positive and negative controls, yeast indeed formed clones on the yeast

302 SD-Leu-Trp-His-Ura (SD-4) selection media after transfection with pDEST32-Hrd1 and

303 pDEST22-Usp15 plasmids (Figure 4a). To examine their direct interaction, plasmid

304 which expressed GST-tagged WT Hrd1 or the transmembrane domain truncation mutant

305 (i.e., ΔTM) as well as His-tagged Usp15 was respectively transformed into the bacterial

306 strain BL21 followed by protein purification. Using an in vitro pull-down assay, we

307 confirmed that WT Hrd1, but not Hrd1ΔTM, directly bound Usp15 (Figure 4b).

308 We next examined endogenous interactions between Hrd1 and Usp15 and detected

309 endogenous Hrd1 protein in anti-Usp15 immunoprecipitates from both PEMs (Figure 4c)

310 and primary MEFs (Figure S4b). Importantly, this endogenous interaction appeared to be

311 enhanced by LPS stimulation (Figures 4c and S4b). Interestingly, we could detect the

312 formation of endogenous protein complex containing Hrd1, Hrd3 and Usp15 in

313 anti-Usp15 immunoprecipitation. However, anti-Usp15 immunoprecipitation failed to

314 pull down Hrd3 and Hrd1 in Hrd1 KO PEMs (Figure S4c). This suggests that Usp15

315 might not be able to bind Hrd3 directly. Next, MEFs were transfected with Flag-tagged

316 Hrd1 and immunostained with anti-Flag and anti-Usp15 antibodies. Although Usp15 was

317 indeed located in the nucleus, we also observed interaction between Hrd1 and Usp15 in

318 the cytoplasm (Figure S4d). To further confirm these results, a cellular fractionation

319 assay was performed in lysed PEMs using centrifugation to harvest microsomes from

14

320 fragmented ER. Calnexin was used as a positive control to identify ER sediment, while

321 Caspase3 and Aif, which do not appear in the ER, were used as negative controls. In

322 agreement with other reports, we detected both Hrd1 and Usp15 in ER sediments37.

323 Interestingly, more Usp15 was translocated into the ER after LPS stimulation (Figure 4d).

324 Immunofluorescence assays also showed that Usp15 colocalised with the ER marker

325 Calnexin, and this colocalisation was enhanced by LPS treatment (Figure S4e, Pearson’s

326 correlation value was increased from 0.58 to 0.68).

327 We next performed a domain mapping experiment and used different tags to

328 precipitate WT Hrd1 and the transmembrane domain truncation mutant (ΔTM) as well as

329 Usp15. HA-tagged Usp15 could co-precipitate Flag-tagged Hrd1 in HEK293T cells

330 (Figure S4f). We also used Myc-tagged Hrd1 and Flag-tagged Usp15 to repeat the

331 immunoprecipitation assay and confirmed their interaction (Figure S4g). Next,

332 Flag-tagged WT Hrd1 or ΔTM was co-expressed with HA-tagged Usp15. While

333 HA-tagged Usp15 co-precipitated Flag-tagged WT Hrd1, the ΔTM mutant lost

334 interaction with Usp15 (Figure 4e). With these different strategies, we have demonstrated

335 that Hrd1 directly interacts with Usp15 depending on Hrd1’s transmembrane domain, and

336 that this interaction is enhanced by TLR4 stimuli.

337 Because the crystal structure of Usp15 (PDB 4A3O)38 and the cryo-EM structure of

338 Hrd1 (PDB 5V6P)35 have been demonstrated, we further propose a model to describe the

339 interaction between Usp15 and Hrd1: a concave surface is formed by the transmembrane

340 helices of Hrd1, the DUSP and UBL domains of Usp15 might bind this concave surface

341 and lie on each side of the Hrd1 dimer (Figure S4i).

15

342 Because Hrd1 regulated TLR4-triggered IκBα degradation in macrophages, we next

343 investigated whether its binding partner Usp15 was also involved in the TLR4 pathway.

344 Knock-down of Usp15 promoted LPS-induced production of IL-6, TNF-α and IL-1β at

345 the mRNA level (Figure 4f) as well as LPS-induced IκBα degradation (Figure 4g, left

346 panel), without affecting TLR4 expression in PEMs (Figure 4g, right panel). Importantly,

347 Usp15 knock-down rescued the reduced IL-6 production in Hrd1-deficient macrophages

348 upon LPS stimulation (Figure 4h). Hrd1-deficient macrophages did not affect Usp15

349 mRNA levels (Figure S4j; the knock-down efficiency of Usp15 was shown). To further

350 ask whether Usp15 regulated TLR4-induced inflammation mainly via NF-κB, we next

351 knocked down the NF-κB subunit p65 using two different siRNAs in Usp15 KD

352 macrophages and observed that knock-down of p65 substantially blocked the enhanced

353 IL-6 production in Usp15 KD macrophages by ELISA (Figure 4i) or by RT-PCR (Figure

354 S4k) assays. The knock-down efficiencies of Usp15 or p65 were shown in Figure S4k

355 (lower panels). Furthermore, the NF-κB luciferase activity was gradually reduced when

356 an increasing amount of Usp15 was co-expressed (in a dose-dependent manner) with

357 Hrd1 in HEK293T cells in the presence of IKKβ (Figure S4l, Usp15 expression levels

358 were shown in the lower panel).

359 Usp15 is a DUB, and previous studies have suggested that the C298A/C812A

360 catalytically inactive Usp15 mutant (termed M2 in this study) loses its deubiquitination

361 ability39. We next asked whether Usp15 DUB activity is crucial for Hrd1-mediated

362 NF-κB activation. Overexpression of Usp15 reduced Hrd1-induced NF-κB luciferase

363 activity; in contrast, overexpression of the Usp15 M2 mutant failed to inhibit

364 Hrd1-mediated effects in the presence of IKKβ (Figure 4j). These data suggest that Hrd1

16

365 directly interacts with Usp15 in macrophages and that the DUB activity of Usp15 is

366 critical for Hrd1-induced IκBα degradation and proinflammatory cytokine production,

367 specifically via the TLR4 pathway.

368

369 Hrd1 promotes polyubiquitination of Lys21 in Usp15 and inactivates Usp15

370 Usp15 is a DUB and, interestingly, its DUB activity is critical for Hrd1 to modulate

371 inflammation (Figures 4j). We next asked whether Hrd1 ubiquitinates Usp15, thus

372 affecting its DUB activity or degradation. To elucidate this, we first co-expressed WT

373 Hrd1, the C329S or ΔTM mutant with HA-tagged Usp15 and His-tagged ubiquitin (Ub)

374 in HEK293T cells, followed by immunoprecipitation with anti-HA antibody. When

375 overexpressed with WT Hrd1 in HEK293T cells, HA-Usp15 ubiquitination levels were

376 significantly enhanced compared to the GFP control, the Hrd1 catalytic inactive mutant

377 C329S and ΔTM mutant (Figure 5a, left panel). To confirm that the signals came from

378 ubiquitinated Usp15, we incubated the immunoprecipitated HA-Usp15 with the catalytic

379 core of human Usp2 (Ubiquitin-specific protease 2) (termed Usp2cc). Hrd1-induced

380 ubiquitination of Usp15 was indeed eliminated after incubation with Usp2cc (Figure 5a,

381 middle panel). His-tagged Usp15 was purified with Ni-NTA beads with the buffer

382 containing high concentrations of salt to minimalize the possibility of pulling down its

383 binding partners, followed by the Usp15 ubiquitination assays to confirm that the

384 detected ubiquitin conjugates were bound to Usp15 itself but not Usp15-associated

385 proteins (Figure 5a, right panel).

17

386 To further confirm that the DUB Usp15 is a substrate of Hrd1, we constituted the E1

387 Ub-activating enzyme (Uba1), the E2 Ub-conjugating enzyme (Ube2g2), the E3 WT

388 Hrd1 or the C329S mutant, HA-tagged Ubiquitin and His-tagged catalytically inactive

389 Usp15 (i.e., C298A/C812A mutant, termed Usp15 M2) into E. coli40, 41. Using the

390 assembled ubiquitin system in E. coli, His-tagged Usp15 was purified with Ni-NTA

391 beads using the buffer containing high concentrations of salt, followed by the in vitro

392 Usp15 ubiquitination assays. We indeed observed that WT Hrd1, but not the C329S

393 mutant, successfully ubiquitinated Usp15 (Figure 5b). Furthermore, the E3 ligase HACE

394 was used as a negative control and constituted into E. coli. HACE could catalyze free

395 ubiquitin conjugation to ubiquitin chains (Figure S5a) but could not catalyze Usp15

396 ubiquitination, which showed no ubiquitination signals in Usp15 (Figure 5b, last lane).

397 Together, we have confirmed that Hrd1 induces Usp15 ubiquitination by using the

398 eukaryotic HEK293T cells and a reconstituted prokaryotic system. Next, we

399 immunoprecipitated Usp15 in PEMs and immunoblotted with anti-ubiquitin antibody.

400 LPS stimulation significantly increased Usp15 ubiquitination levels (Figure 5c).

401 Importantly, in comparison to WT PEMs, the Usp15 ubiquitination signal was strongly

402 reduced in Hrd1-deficient macrophages (Figure 5d). This result was consistent with

403 enhanced Usp15 translocation to the ER (Figures 4d and S4e) and increased Usp15

404 interaction with Hrd1 (Figures 4c and S4b) after LPS treatment. Together, these results

405 demonstrate that Hrd1 binds Usp15 to promote Usp15 ubiquitination in macrophages.The

406 next important question was which lysine residues on Usp15 are ubiquitinated by Hrd1.

407 Using the E. coli reconstituted ubiquitin system, Hrd1-induced ubiquitinated Usp15 was

408 enriched by two rounds of immunoprecipitation using Ni-NTA (Usp15) and anti-HA

18

409 (Ubiquitin) antibodies then subjected to trypsin digestion followed by MS analysis. The

410 MS results identified ubiquitination sites at Lys21, Lys154 and Lys228 in Usp15. For

411 verification of these results, we generated Usp15 mutants bearing the individual

412 Lys-to-Arg substitutions, including K21R, K154R and K228R. We first found that these

413 substitution mutants of Usp15 formed stable complexes with Hrd1 via

414 immunoprecipitation assay in HEK293T cells (Figure S5b). By contrast, only when

415 Lys21 was mutated to Arg (i.e., K21R), Usp15 ubiquitination was eliminated (Figure 5e).

416 We next investigated the biological consequences of Hrd1-mediated ubiquitination

417 of Lys21 in Usp15. Unexpectedly, unlike the Hrd1 targets for ERAD, Hrd1-induced

418 ubiquitination of Usp15 did not lead to Usp15 degradation (Figure S5c). The Lys21

419 residue is located in the DUSP domain of Usp15, and previous studies have reported that

420 Usp15 requires the DUSP-Ubl domain to achieve its full catalytic efficiency42. This

421 raised the key question of whether Lys21 polyubiquitination regulated Usp15 catalytic

422 activity. We co-expressed Myc-tagged Hrd1 or the C329S mutant with Flag-tagged

423 Usp15 or the K21R mutant in HEK293T cells then purified Flag-Usp15 by

424 immunoprecipitation. During the immunoprecipitation process, we added excess

425 polyubiquitin chains which competitively protected Usp15 polyubiquitination (Figure

426 S5d). Using DUb-7-amido-4-methylcoumarin (DUb-AMC) as the substrate, we found

427 that activated Usp15 cleaved DUb-AMC and released a C-terminal derivatization of

428 ubiquitin with 7-amino-4-methylcoumarin. By measuring fluorescence intensity, we

429 observed that ubiquitinated Usp15 had significantly reduced DUB activity levels

430 compared to the K21R mutant or the un-ubiquitinated Usp15 (Figure 5f). We next used

431 another strategy to examine how Usp15 ubiquitinatation affected its DUB activity. Indeed,

19

432 when a K48-linked polyubiquitin chain (Ub2-16) was used as a substrate, the K21R mutant

433 showed higher activity in deubiquitinating polyUb (i.e., Ub4, 5, or Ub 6-Ub16) with an

434 enhanced amount of mono-Ub (Figure 5g, samples with long exposure). Next, NF-κB

435 luciferase reporter assays were performed to determine the biological function of Lys21

436 ubiquitination in Usp15. Compared to WT Usp15, the non-ubiquitinated K21R mutant

437 displayed a better inhibitory effect on Hrd1-induced luciferase activity in a

438 dose-dependent manner (Figure5h, increasing expression levels of Usp15 and K21R were

439 shown in the lower panel). Indeed, when comparable levels of WT Usp15 or the K21R

440 mutant were stably overexpressed in MEFs (Figure5i, right lower panel), the K21R

441 mutant showed a further decrease of Il6 production at the mRNA and protein levels after

442 E. coli treatment (Figure 5i, left two panels). Compare to the GFP control sample, stably

443 overexpression of WT Usp15 could remove ubiquitin in IκBα, resulting in stabilization

444 IκBα in response to LPS treatment (Figure S5e). The K21R mutant could further stabilize

445 IκBα, comparing to that in WT Usp15, but the M2 mutant lost the DUB activity and

446 maintained LPS-induced IκBα degradation (Figure S5e). These results agree with the

447 enhanced DUB activity of the K21R mutant (Figures 5f and 5g). Together, we have

448 demonstrated that Hrd1 inhibits Usp15 DUB activity by adding Lys27-linked

449 polyubiquitin chains to the Lys21 residue in Usp15.

450

451 Hrd1 deficiency protects against LPS- and CLP- induced septic shock

452 Since excessive TLR4-NF-κB activation plays a critical role in sepsis, we asked

453 whether Hrd1 deficiency in macrophages protects against LPS- and CLP-induced septic

454 shock in mice. Upon intra-peritoneal injection of LPS, more Hrd1 cKO mice survived 20

455 compared to the WT littermate controls (Figure 6a). In agreement with the notion that an

456 uncontrolled inflammatory response is critical for induction of septic shock, we observed

457 significantly lower concentrations of serum IL-6 in Hrd1 cKO mice (Figure 6b).

458 Similarly, Hrd1 cKO mice exhibited less robust Il6 production in multiple organs,

459 including the liver, lung and kidney (Figure 6c). Immunohistological studies showed that

460 the WT mice displayed more severe lung injury than Hrd1 cKO mice (Figure 6d).

461 To confirm the role of Hrd1 in pathogen-induced septic shock, the WT and cKO

462 littermates were subjected to caecal ligation and puncture (CLP) surgery. In agreement

463 with the results from LPS-induced septic shock, Hrd1 cKO mice died at later time points

464 and had increased survival rates (Figure 6e). IL-6 concentrations in serum (Figure 6f) as

465 well as mRNA levels in the liver, lung, kidney and spleen (Figure 6g) were significantly

466 reduced in Hrd1 cKO mice compared to their WT littermates 6 hr after CLP operation.

467 Immunohistological staining of the lungs showed enhanced lung damage in WT mice 24

468 hr post CLP surgery (Figure 6h).

469 We next asked a critical question: how Usp15 deficiency affects Hrd1 KO

470 macrophage function in vivo. Because Usp15 and Hrd1 double KO mice were not

471 available, we depleted endogenous peritoneal macrophages via intra-peritoneal (i.p.)

472 injection of clodronate-containing liposomes followed by adoptive transferring of

473 CFSE-labeled WT or Hrd1-deficient macrophages transfected with scrambled siRNA or

474 siUsp15. We confirmed that treatment with clodronate-containing liposomes resulted in

475 marked depletion of endogenous F4/80+CD11b+ macrophages in the peritoneal cavity

476 (Figure S6a). Usp15 KD efficiency was confirmed in WT and Hrd1 KO PEMs (Figure 6i,

477 right panel). After adoptively transferring Hrd1 KO/Usp15 KD PEMs for 24 hr, mice

21

478 were challenged with LPS for 6 hr followed by collection of CFSE+ PEMs. In agreement

479 with the in vitro data, Hrd1 KO macrophages exhibited lower IL-6 production (Figure 6i,

480 left panel). Importantly, knock-down of Usp15 rescued reduced Il6 production in Hrd1

481 KO macrophages upon LPS treatment, which reached levels similar to those in WT cells

482 (Figure 6i, left panel).

483 We previously excluded the role of Hrd1 in TLR3/TLR9-induced inflammation in

484 vitro (Figure 2h). To further examine whether Hrd1 regulated TLR3/9 signalling in vivo,

485 we injected CpG or Poly(I:C) respectively, into WT and Hrd1 cKO mice through the tail

486 vein. Consistent with the in vitro data, serum IL-6 or IFN-β concentrations were not

487 significantly changed between the paired WT and Hrd1 cKO mice (Figures 6j and 6k, left

488 panel), and mRNA levels of Il6 and Ifnb in multiple organs were also comparable

489 (Figures 6j and 6k, right panels). These results confirmed that the Hrd1/Usp15 module

490 also specifically affected TLR4 signalling in vivo.

491 Together, our data have elucidated a model linking the TLR4 innate signalling

492 from the cell membrane to the ER (Model in Figure 6l). As an ER-located E3 ligase,

493 Hrd1 is important for transducing signals from the cell membrane receptor TLR4 to the

494 DUB Usp15, which is recruited to the proximal ER membrane to induce IκBα

495 degradation and NF-κB activation.

496

497 Discussion

498 More and more E3 ubiquitin ligases are being identified to precisely regulate

499 TLR4-NF-κB activation, such as TRAF6 and Pellino-1. In this study, we have identified

22

500 that Hrd1, the ER-located RING-type E3 ligase, is critical for the plasma membrane

501 receptor TLR4-induced signalling and proinflammatory cytokine production in

502 macrophages. Deletion of the Hrd1 gene or knock-down of Hrd1 expression specifically

503 attenuated IL-6, TNF-α and IL-1β production in macrophages after LPS treatment,

504 Salmonella typhimurium or E. coli infection. Interestingly, Hrd1 deficiency did not

505 modulate IL-6 or IFN-β production after activation of the TLR2/TLR3/9 pathways. In

506 vivo Hrd1-specific deficiency in myeloid cells protected mice from LPS- or CLP-induced

507 septic shock, suggesting that targeting Hrd1 might be applied to protect the host from

508 severe bacterial infections.

509 Expression of Hrd1, also known as Syvn1, is increased in synoviocytes and

510 peripheral blood cells from RA patients19, and this increase is correlated with the onset of

511 RA. Previous work suggests that Hrd1+/− mice increase apoptosis of synovial cells to

512 protect against collagen-induced arthritis20, 21. In contrast to Hrd1 KO synovial cells, we

513 did not observe abnormal apoptosis in Hrd1 KO macrophages. Importantly, our study

514 suggests that Hrd1 acts as a positive regulator for IL-6 and TNF-α production via the

515 TLR4 pathway in macrophages. Previous immunohistological analyses of inflamed RA

516 joint tissue have demonstrated that TLR4 is highly expressed in infiltrated macrophages22,

517 43, and macrophages are one of the main sources of proinflammatory cytokines44. IL-6

518 and TNF-α play crucial roles in the pathology of RA, and the most widely used biologic

519 anti-rheumatic drugs are anti-TNF agents. It will be intriguing to examine whether Hrd1

520 affects IL-6 and TNF-α production in infiltrated macrophages from inflamed RA joints.

521 We therefore propose that inhibition of Hrd1 E3 ligase activity might be a promising

23

522 approach to treat RA by targeting two major cell types, including increasing apoptosis of

523 synovial cells and proinflammatory cytokine secretion in macrophages.

524 Our study has identified Hrd1 as an ER-located protein that regulates

525 TLR4-signalling during bacterial infection in vivo and in vitro. Cellular localisation is

526 important for certain proteins to perform their functions, including the

527 mitochondria-located MAVS4, the ER-located STING6, 7 and the endosome-located

528 TLR3 that participate in the innate immune response against viruses. In particular, several

529 key amino acids substitutions in STING were mapped from patients exhibiting

530 autoimmunity. Interestingly, these residues make up a small linker region connecting the

531 N-terminal transmembrane domain of STING to the C-terminal cyclic

532 dinucleotide-binding domain45. These residues are critical in retaining STING on the ER

533 in unstimulated cells, and disease mutations disrupt ER retention46 and causing abnormal

534 interferon response6, 47. The ER has a broad localisation throughout the cell and can form

535 direct physical contacts with all other membranous organelles2, 3, 6, 10, 11, which are

536 involved in lipid exchange between membranes, controlling Ca2+ homeostasis and other

537 key biological functions. In the past several years, advancing studies focus on the

538 structure and function of membrane contact sites within cells, including the contact of the

539 ER with mitochondria and endosome10, 11. However, it is largely unknown whether ER

540 functions during bacterial infection and which molecules with special ER localisation

541 could modulate the TLR4 pathway or anti-bacterial infection. Our study provides an

542 interesting example on how two membrane proteins (i.e. TLR4 and Hrd1) from the

543 contacts of the ER with the cell plasma membrane function in regulating bacterial

544 infection and inflammation. This provides an important hint linking ER with

24

545 anti-bacterial infection. Hopefully it will trigger more interest in understanding the

546 biology and function of the endomembrane system as well as the crosstalk between ER

547 and the plasma membrane.

548 As an ER-located E3 ligase, the classical role of Hrd1 is to ubiquitinate incorrectly

549 folded proteins in the ER to promote their degradation and to prevent ER-stress-induced

550 cell apoptosis. Our study has illuminated that unlike the role of Hrd1 in ERAD, Hrd1 in

551 LPS-stimulated macrophages catalyses Usp15 ubiquitination at Lys27 which does not

552 lead to Usp15 instability and degradation, but rather inactivates Usp15. This might open a

553 research angle for studying the ERAD-independent function of Hrd1.

554 Prior studies have implicated Usp15 in the regulation of a variety of cellular

555 signalling pathways, including the TGF-β, NF-κB, β-catenin, spliceosome and p53

556 signalling pathways39, 48. Usp15 dysregulation has been demonstrated in cancer and

557 immune responses26, 49. Despite its critical functions, the mechanism of Usp15 regulation

558 has not yet been fully elucidated. In this work, we demonstrated that Usp15 translocates

559 to the ER and is modified by Hrd1-mediated polyubiquitination after TLR4 activation in

560 macrophages. The ER-located E3 Hrd1 catalyses the Lys21 in Usp15, disrupting its DUB

561 activity and preventing it from deubiquitinating downstream effectors, including IκBα.

562 The Lys21 in Usp15 represents the first-identified post-translational modification of

563 Usp15. In addition, recent studies have implicated ER dysfunctions and ER stress are

564 involved in liver diseases, Alzheimer’s disease, cancer32, 50. This suggests that inhibit ER

565 dysfunctions or targeting the Hrd1/Usp15 module might be useful in treating related

566 pathological conditions.

25

567 In summary, we have provided compelling evidence that the ER-located E3

568 ubiquitin ligase Hrd1 ubiquitinates the DUB Usp15 to modulate TLR4-NF-κB signalling

569 in macrophages in response to Gram-negative bacterial infection. Given the role of Hrd1,

570 Usp15 and the proinflammatory cytokines IL-6 and TNF-α in pathological conditions

571 including rheumatoid arthritis, Parkinson’s disease, cancer and anti-pathogen infections,

572 this study might provide therapeutic hints for drug design.

573

574 Material and methods

575 Mice. The sixth exon of Hrd1 gene was flanked by two loxP sites to generate Hrd1

576 conditional knock-out mice (the F0 generation i.e. Hrd1fl/- mice) (Sidansai Biotechnology

577 Company, Nanjing University). Hrd1fl/- mice were crossbred with LyzM-Cre+/+ mice to

578 generate Hrd1fl/-LyzM-Cre+/- mice, which were then backcrossed with Hrd1fl/- mice. The

579 Hrd1fl/flLyzM-Cre+/- cKO mice specifically knocked out Hrd1 in myeloid cells, and the

580 littermate Hrd1fl/flLyzM-Cre-/- mice were used as the controls. Genotyping was performed

581 with the following primers: forward: 5′-GCATAGTCTCTACTCTGTTC-3′; reverse:

582 5′-CTTCCTGCTCCACGACAATC-3′. The mice were on C57BL/6 background. All

583 mice were maintained under specific pathogen-free conditions with approval of the

584 institutional animal facility of Shanghai Institute of Biochemistry and Cell Biology

585 (protocol IBCB0057) and the National Institute for Viral Disease Control and Prevention.

586 Animals were randomly allocated to experimental groups. The animal experiments

587 performed with a blinded manner were described below. Statistical methods and advice

588 from the related publications were considered to determine the number of mice used. All

589 animal experiments were approved by the Institutional Animal Care and Use Committee 26

590 (IACUC) of Institute of Biochemistry and Cell Biology, Shanghai Institutes for

591 Biological Sciences, Chinese Academy of Sciences.

592 Cells and reagents. Cells were maintained at 37 °C under 5% CO2. HEK293T cells were

593 kindly provided by Dr. Xiaohui Zhang (SIBCB, CAS). Primary MEF cells as well as

594 MEF cell lines were kind gifts from Dr. Anning Lin (SIBCB, CAS). Primary peritoneal

595 macrophages, HEK293T, primary MEF, MEF cell lines were tested and confirmed no

596 mycoplasma contamination, which were cultured in Dulbecco’s modified Eagle

597 medium (DMEM) supplemented with 10% (v/v) FBS, L-Glutamine (2m M), and

598 penicillin–streptomycin (100 U/ml). THP-1 cells were cultured in complete RPMI 1640

599 medium, induced by Phorbol-12-myristate-13-acetate (PMA) (final concentration 300

600 ng/ml) for 12 hr to differentiate to macrophages. Mouse peritoneal macrophages were

601 prepared from 12 weeks old mice through intraperitoneal injection with 3 ml 3% Brewer

602 thioglycollate medium. To generate bone-marrow-derived macrophages (BMM), bone

603 marrow cells were cultured with 30% L929-conditioned media for a week. All cell lines

604 used in this study were not contaminated by mycoplasma, which were examined by

605 PCR-based detection of mycoplasma. Lipopolysaccharide was from Escherichia coli

606 055:B5 and purified by phenol extraction. CpG-B (ODN1826) and Poly(I:C) (HMW)

607 Rhodamine and PGN (tlrl-pgnb3, InvivoGen) were from Invivogen. Clodronate

608 liposomes were from FormuMa. Antibodies used in this study were listed in Table S2.

609 Plasmids. cDNAs for human HRD1, USP15, NFKBIA were amplified from

610 reverse-transcribed cDNA from HEK293T cells. These genes were cloned into the

611 pcDNA3.1-IRES-GFP vector for transient expression in HEK293T cells. For stable

612 expression, human HRD1 and USP15 were subcloned into the pMSCV-IRES-GFP vector 27

613 with Flag tag and HA tag respectively. For recombinant expression in E.coli, USP15 was

614 inserted into the pET-Duet vector with 6xHis tag. Mutations in HRD1 or in USP15 were

615 constructed by standard PCR cloning strategy. Plasmids expressing ubiquitin

616 lysine-to-arginine mutations were kindly provided by Dr. Ronggui Hu (SIBCB, CAS).

617 Usp15 truncation mutations were kindly gifts from Dr. Shaocong Sun (MD Anderson

618 Cancer Center, USA). All plasmids were confirmed by DNA sequencing. Plasmids used

619 in this study were listed in Table S3.

620 RNAi. siRNA (Dharmacon) targeting specific genes were delivered into PEMs and

621 BMMs by Lipofectamine RNAiMAX Reagent (Invitrogen) according to the

622 manufacturer’s instructions. Sequences of all siRNA were listed in Table S4.

623 Quantitative RT-PCR analysis. Total RNA was isolated by RNAiso Plus regent

624 according to the manufacturer’s instructions (TAKARA). First-strand cDNA was

625 generated using M-MLV reverse transcriptase (RNase H-) (TAKARA). Samples were

626 amplified by CFX-96 machine (Bio-rad) using SYBR Green master mix (DBI Bioscience)

627 according to the manufacturer’s instructions, and data were normalized to an Actb control.

628 Sequences of all primers were listed in Table S5.

629 Stable cell line establishment and BMM transduction. Retroviral particles were

630 prepared by transfecting HEK293T cells with pCL-10A and MIGR or MSCV vectors

631 expressing target genes. MEFs were infected with retroviral supernatants in the presence

632 of polybrene (final concentration 1 μg/ml) (Santa Cruz). Cells overexpressing the

633 indicated genes were selected by sorting GFP positive cells or by puromycin (final

634 concentration 1.5 μg/ml) treatment. To generate stably transduced BMMs,

635 L929-conditioned media was supplemented with retroviral supernatants at 1:1 ratio to 28

636 culture BM cells at the first 3 days. After 7 days induction, BM cells were derived into

637 macrophages which stably overexpressed the indicated genes.

638 ELISA. Concentrations of mouse IL-6 or TNF-α in serum or cell supernatants were

639 measure by ELISA commercial kits (eBioScience) according to the manufacturer’s

640 instructions.

641 Dual-luciferase reporter assay. HEK293T cells were transfected with the NF-κB

642 luciferase reporter, TK-renilla and the indicated plasmids by Polyethylenimine

643 (Polysciences) as the manufacturer’s instructions. Cells were harvested after 24 hr to

644 measure luciferase readings with a dual-luciferase reporter assay system according to the

645 manufacturer’s instructions (Promega).

646 Isolation of cytoplasmic and nuclear fractions. PEMs or PMA-induced THP-1 cells

647 were stimulated with 1 μg/ml LPS for the indicated time points. Cells were harvested,

648 washed with PBS, then lysed with buffer A (10 mM HEPES, 1.5 mM MgCl2, 10 mM

649 KCl, 0.5 mM DTT, 1 mM PMSF, 0.1% (v/v) NP-40 and protease inhibitor cocktail, pH

650 7.9). Lysates were centrifuged at 6,000 rpm for 15 min at 4°C. Supernatants containing

651 cytoplasmic proteins were collected. Nuclear proteins were extracted from the pellet by

652 cold buffer C (20 mM HEPES, 1.5 mM MgCl2 , 0.42 M NaCl, 0.2 mM EDTA, 25% (v/v)

653 glycerol, 0.5 mM DTT, 1 mM PMSF and protease inhibitor cocktail, pH 7.9), and

654 insoluble material was removed by centrifugation at 12,000 rpm for 1 min at 4°C.

655 Immunostaining. MEF cells or PEMs were stimulated with 1 μg/ml LPS as the indicated

656 time points, mounted on coverslips, fixed, permeabilized, and then stained with the

657 indicated antibodies and Hoechst. Images were taken by Olympus IX81 microscopy or

29

658 Leica confocal microscopy. Fluorescent imaging analysis was conducted in a blinded

659 fashion, and densitometry quantification or Pearson’s correlation was analyzed by

660 Image-Pro 6.0 software.

661 Mass spectrometry analysis. To identify the substrates of Hrd1, THP-1 cells which

662 stably overexpressed Flag-dsRed or Flag-Hrd1 were treated with PMA (final

663 concentration 300 ng/ml) for 12 hr to derive into macrophages. After washing, adherent

664 THP-1 macrophages were stimulated with 1 μg/ml LPS for 1hr and then lysed with lysis

665 buffer (20 mM Tris-Cl, 100 mM KCl, 1 mM EDTA, 0.1% (v/v) NP-40, 10% (v/v)

666 glycerol, and protease inhibitor cocktail, pH 7.5). Whole cell extracts were

667 immunoprecipitated with anti-Flag beads. Proteins were eluted by the Flag peptides and

668 subjected to make freeze-dried powder.

669 To investigated which lysines in Usp15 were ubiquitinated by Hrd1, BL21

670 competent cells transferred with indicated plasmids were inoculated and induced by

671 IPTG (1 mM) at 16 °C for 16 hr. Bacteria harvested in RIPA buffer were lysed by sonic

672 and centrifuged at 4 °C for 15 min. Following incubation of cell lysates and Ni-NTA

673 beads at 4 °C for 2 hr, the beads were washed for five times with RIPA buffer and eluted

674 with 200 mM imidazole. The eluent was dialyzed in PBS containing 10% glycerol at

675 4 °C overnight and collected the liquid in dialysis bag. Next, the resulting proteins were

676 immunoprecipitated with anti-HA beads to enrich ubiquinatinated Usp15. The beads

677 were washed for five times with RIPA buffer, and a part of the beads were harvested with

678 SDS loading buffer and the sample were determined by western blot. Protein pellet was

679 dissolved in 10 μl of 8 M Urea, 100 mM Tris-HCl, pH 8.5, and diluted to 80 μl 1M Urea

680 with 100 mM Tris-HCl, pH 8.5. The pH was adjusted to pH 1.0 by adding 1 M HCl.

30

681 Digestion was performed in the presence of 100 mM Tris-HCl using sequencing-grade

682 soluble trypsin (Roche).

683 Yeast two-hybrid. To examine the interaction between Hrd1 and Usp15, GAL4-based

684 yeast two-hybrid system (ProQuest™ Two-Hybrid System, Invitrogen) were applied.

685 Full length human Hrd1 ORF (open reading frame) was cloned into the donor vector

686 pDONR221 and transfected into pDEST32 via Gateway cloning reaction (Invitrogen). In

687 pDEST32 plasmid, Hrd1 was in-frame fused with GAL4 DNA binding domain,

688 generating the bait plasmid. Human Usp15 was in-frame fused to the GAL4 activation

689 domain (Invitrogen) in prey vector pDEST22. To perform the Yeast two-hybrid assay,

690 the yeast strain Mav203 competent cells were transformed with the bait clone

691 pDEST32-Hrd1 and the prey clone pDEST22-Usp15. The empty vectors pDEST22 and

692 pDEST32 were used as the negative prey or bait controls. pEXP32-Krev1 and

693 pEXP22-RalGDS-WT were used as the positive controls. Mav203 competent cells that

694 were transformed with the bait and prey plasmids could grow and form clones on

695 SC-Leu-Trp (SD-2) plates, while only cells containing the interacting proteins could

696 grow on SC-Leu-Trp-His-Ura (SD-4) plates. When pDEST32- Hrd1 and

697 pDEST22-Usp15 were transformed into Mav203 competent cells, all reporter genes were

698 activated.

699 Expression and purification of recombinant proteins. The full length and the

700 transmembrane domain truncation mutant (ΔTM) of human Hrd1 were constructed to the

701 pGEX4T-1-GST vector (GE Healthcare) to generate GST-fusion proteins. Human Usp15

702 was cloned into pET-Duet-His vector. Recombinant proteins were expressed in E. coli

703 BL21 (DE3) Codon-Plus strain (Novagen). BL21 cells were transformed with the above

31

704 plasmids and grown in L-broth supplemented with ampicillin (50 μg/ml). Expression of

705 the recombinant proteins was induced with 0.1 mM IPTG at 16 °C for 20 hr. For

706 purification, 6×His-Usp15 was purified by Ni-NTA affinity chromatography (Qiagen,

707 Germany); GST-Hrd1, GST-ΔTM or GST were purified by Glutathione-agarose beads

708 according to the manufacturer’s instructions (GE). Purified Hrd1 protein was identified

709 by western blotting without boiling.

710 GST pull-down. GST-Hrd1 (3 μg), GST-Hrd1-ΔTM (2.4 μg) or GST proteins (0.92 μg)

711 at equimolar concentrations were incubated with His-Usp15 (4 μg) at 4 °C for 2 hr in 100

712 μl pull-down buffer (20 mM Tris-Cl, 100 mM NaCl, 5 mM MgCl2, 1 mM EDTA, 1 mM

713 DTT, 0.5% (v/v) NP-40 and 10 μg/mL BSA, pH 7.5) followed by washing three times.

714 Samples were combined with SDS loading buffer and subjected to SDS-PAGE without

715 boiling.

716 Generation of the interaction model. The interaction model between Usp15 and Hrd1

717 complex was generated with the Pymol program (The PyMOL Molecular Graphics

718 System, Version 1.6, Schrödinger, LLC).

719 Immunoprecipitation analysis. HEK293T cells overexpressing the indicated proteins

720 were washed with cold PBS before lysed with cold lysis buffer (25 mM Tris-Cl, 150 mM

721 NaCl, 1% (v/v) NP-40, 5 mM EDTA, 0.5% sodium deoxycholate and protease inhibitor

722 cocktail, pH 7.2). Cell lysates were then centrifuged at 13, 500 rpm for 15 min, 4 °C.

723 Following incubation of cell lysates with protein G Sepharose coated with indicated

724 antibodies and rotating at 4 °C for 2 hr, beads were then washed for five times with lysis

725 buffer and resuspended in SDS-PAGE loading buffer for western blot analysis.

32

726 Cellular fractionation. PEMs were harvested into 5 ml ice-cold PBS and centrifuged at

727 1,000 g for 5 min at 4 °C. Cell pellets were resuspended in 0.4 ml Buffer F (10 mM

728 HEPES-KOH at pH 7.2, 250 mM sorbitol, 10 mM KOAc, 1.5 mM Mg(OAc)2 and

729 protease inhibitor cocktail), passed through a 22-gauge needle 20 times, incubated on ice

730 for 15 min, and centrifuged at 1,000 g for 5 min at 4 °C. Supernatants were transferred to

731 microtubes, and centrifuged at 16,000 g for 3 min at 4°C. The pellets were resuspended

732 in 0.5 ml Buffer E (50 mM HEPES-KOH at pH 7.2, 250 mM sorbitol, 70 mM KOAc, 5

733 mM potassium EGTA, 2.5 mM Mg(OAc)2 and protease inhibitor cocktail), and

734 centrifuged at 16000 g for 3 min at 4 °C to obtain microsomes.

735 EM Data Collection. LPS (1 μg/ml) was used to coat latex beads (3 μm, Sigma) via

736 shaking at 4 °C overnight. PEMs were treated with LPS-coated or non-coated latex beads

737 for 60 min, and fixed with 2.5% glutaraldehyde in PBS for 1.5 hr at 4 °C. After washing

738 with PBS three times, the fixed samples were post-fixed with 1% osmium tetroxide for 1

739 hr at 4 °C. The samples were washed with PBS three times and dehydrated in an ethanol

740 series (25 %, 50 %, 75 %, 95 % ethanol in distilled water, each for 10 min). Then the

741 samples were dehydrated in pure ethanol twice (each for 30 min) and infiltrated with

742 ethanol:epoxy 812 (at 2:1, 1:1 and 1:2 ratio, each for 30 min), and then infiltrated with

743 pure epoxy 812 overnight. Then the samples were infiltrated with fresh epoxy 812 for 1

744 hr and embedded in epoxy 812. After polymerizing at 65 °C for 48 hr, the resin blocks

745 were trimmed and thin sectioned with a thickness of 70 nm on an ultramicrotome using a

746 diamond knife. The thin sections were mounted onto formvar-coated copper grids and

747 counterstained with 3 % uranyl acetate in 70 % methanol for 7 min and then by lead

33

748 citrate for 3 min. The stained sections were viewed under a FEI electron microscope

749 (Tecnai G2 Spirit 120kV) and images were recorded on a FEI Eagle CCD camera.

750 In vivo ubiqunitination assay. For in vivo ubiquitination of Usp15, HEK293T cells

751 transiently transfected with HA-tagged Usp15, His-tagged WT or mutant ubiquitin and

752 Flag-tagged Hrd1 or its E3 ligase-dead mutant C329S or PEMs lysed in RIPA buffer (50

753 mM Tris-Cl, 150 mM NaCl, 5 mM EDTA, 1% (v/v) Triton X-100, 0.5% sodium

754 pyrophosphate, 0.1% SDS, and protease inhibitor cocktail, pH 7.4) containing 10 mM

755 N-Ethylmaleimide. Cell lysates were immunoprecipitated with anti-HA beads or with

756 protein G Sepharose coated with 2 μg anti-Usp15 antibody/sample and then washed for

757 five times with RIPA buffer, followed by immunoblotting analysis to detect indicated

758 proteins. For Figure 5a left panel, after the immunoprecipitation, HA beads were

759 resuspended in PBS buffer and subjected to Usp2cc (catalytic core of Usp2) digestion

760 (final concentration 100 μg/ml), followed by immunoblot analysis. For Figure 5a right

761 panel and Figure 5b, HEK293T cells and E.coli overexpressing His-Usp15 and the

762 indicated proteins were suspended in denaturing lysis buffer (6 M Guanidine

763 Hydrochloride, 20 mM Sodium Phosphate, pH 7.8, 500 mM NaCl) and lysed by

764 sonication. Cell lysates were incubated with Ni-NTA beads at 4 °C for 2 hr, and washed

765 three times with binding buffer ( 8 M Urea, 20 mM Sodium Phosphate, pH 7.8, 500 mM

766 NaCl) and wash buffer (8 M Urea, 20 mM Sodium Phosphate, pH 6.0, 500 mM NaC),

767 respectively.

768 LPS shock and CLP model. To monitor survival conditions, 8 weeks old male WT and

769 cKO mice were injected intraperitoneally with LPS (35 mg/kg) to induce LPS shock. To

770 generate the CLP model, mice were anesthetized, and an abdominal incision was made

34

771 for identification of the cecum. The distal one third of the cecum was ligated with silk

772 suture and was punctured twice with a 21-gauge needle. A small amount of cecal content

773 was extruded through the perforation. The peritoneum and skin were closed with

774 autoclips (Becton Dickinson) after the cecum was returned to the abdomen. 1 ml saline

775 was injected intraperitoneally for resuscitation. For sham-treated mice, all of the same

776 steps were performed, except for ligation and puncture of the cecum.

777 Adoptive macrophage transfer. 8-week-old WT male mice were injected

778 intraperitoneally with 100 μl of Clodronate liposomes (FormuMax) to delete peritoneal

779 macrophages and used as the recipient. PEMs obtained from WT and cKO mice were

780 transfected with scramble or mouse siUsp15 in a 1.5 ml Eppendorf tube for 6 hr followed

781 by CFSE labeling. Equal numbers of the control or siUsp15-transfected CFSE+ PEMs (10

782 million) were i.p injected to the Clodronate liposome-treated recipient mice. 24 hr later,

783 mice were i.p challenged with 25 mg/ kg LPS for 6 hr, sacrificed and adoptively

784 transferred CFSE+ PEMs were harvested to extract mRNA for RT-PCR analysis.

785 In vivo Poly(I:C) and CpG treatment. Age- and sex-matched adult mice were i.v

786 injected with 100 μg Poly(I:C)/ mouse or 100 μg CpG-B/mouse. Mice were sacrificed on

787 the indicated time points, serum and organs were harvested to measure expression levels

788 of IL-6 and IFN-β.

789 Flow cytometry. All samples passed through a 70 µm nylon cell strainer to make

790 single-cell suspension. For surface staining, cells were labeled with the indicated

791 antibodies in FACS buffer (2.5 L PBS, 50 ml NBS and 0.5 g NaN3) for 30 min at 4 °C

792 avoiding from light. To analysis macrophages apoptosis, cells were incubated with

793 Annexin V (BioLegend) in Annexin V Binding Buffer (BD Bioscience) at room 35

794 temperature for 20 min in the dark, and then stained with PI (Propidium bromide) for a

795 maximum of three minutes. The samples were washed with cold FACS buffer and

796 analyzed with C6 (BD Bioscience).

797 Statistics. Adequate power was ensured when choosing the sample sizes. Data are

798 represented as the means ± SEM of at least three experiments. Statistical analyses are

799 performed using Graphpad Prism6 software, version 6. Statistical significance is

800 calculated using Student's two-tailed unpaired t-test for comparison between groups, and

801 using paired t-test for western blot and serum ELISA analysis. The log-rank test is used

802 for survival comparisons. NS, not significant (p>0.05); *P < 0.05, **P < 0.01.

803 Author Contributions: Y.L, and Y.Q. performed the majority of experiments and

804 statistical analysis with the help of H.C., L.G., F. Z., L. M., C. Z., X. Z., J. X., R.Z., L. H.,

805 X.X., Y. Z., P.C. participated in part of the ubiquitination experiments. Y.H. generated

806 the interaction model of Hrd1 and Usp15. H.W., B.W., D.L., Y.L., Y.Q. G. Z., R. B., Y.H.

807 and R. H. designed the study. Y.L. and H.W. drafted the manuscript.

808 Acknowledgments: We thank Drs. ShaoCong Sun, Chen Wang, Bing Sun for providing

809 plasmids, Dr. Anning Lin for MEF cell lines, Dr. WH Fang for Salmonella typhimurium

810 (strain SL1344), Dr. Yulong He for Lysozyme M (LysM)-Cre+/+ mice, Dr. Ping Wang for

811 DUb-AMC reagents. We would like to the Core Facility of Chemical Biology and Core

812 Facility of Molecular Biology for technique help; thank the National Center for Protein

813 Science Shanghai for MS analysis and EM Data Collection.

814 This work was supported by grants from the Strategic Priority Research Program of

815 the Chinese Academy of Sciences (XDB19000000), the Ministry of Science and

36

816 Technology of China (2016YFD0500207, 2016YFD0500407, and 2016YFC0905902),

817 National Natural Science Foundation of China (81825011, 81630043, 81571617,

818 81671572, 81571552, 81701569, 31700781, 8180060957), and the State Key Laboratory

819 of Cell Biology, SIBCB, CAS (SKL CBKF2013003). We thank the Genome Tagging

820 Project (GTP) Center, Shanghai Institute of Biochemistry and Cell Biology, CAS for

821 technical support. Dr. H. Wang is supported by the Hundred Talents Program of the

822 Chinese Academy of Sciences.

823 Competing interests: None declared.

824 Data availability: The original data that support the findings of this study are available

825 from the corresponding author upon request. Supplementary figures are available in the

826 Supplementary Information.

827

37

828 Figure 1. RNAi screening identifies the ER-localised Hrd1 to positively regulate

829 inflammation in LPS-stimulated macrophages

830 (a) The relative amount of IL-6 concentrations in the siRNA screening plate that contains

831 LPS-treated mouse peritoneal macrophages (PEMs). (b) Electron microscopic images

832 from macrophages after incubation with the untreated beads (left panel) or LPS-coated

833 beads (right panel). Triangles indicate ER membrane. (c) Confocal microscopy analysis

834 of the ER in MEFs transfected with KDEL-mCherry before or after SL1344 infection.

835 Nucleus were labelled with DAPI. (d, e, f) The relative mRNA levels was checked by

836 qRT-PCR or cytokine concentrations was quantified by ELISA in PEMs transfected with

837 the control siRNA (N.C.) or Hrd1 siRNA for 72 hr, followed by LPS (1 μg/ml)

838 stimulation (d, f), SL1344 or E. coli infection (e). (g) qRT-PCR analysis of Il6 mRNA

839 levels in PMA-induced THP-1 cells or primary MEFs with or without LPS (1 μg/ml)

840 stimulation. THP-1 cells or primary MEFs were generated with a retrovirus transduction

841 system to stably overexpress GFP or Hrd1. (h) Luciferase (Luc) activity was measured in

842 HEK293T cells after transfected with the plasmids expressing NF-κB luciferase reporter,

843 renilla, GFP or Hrd1 without (left panel) or with (right panel) MyD88, TRAF6, IKKα,

844 IKKβ, p65 or TRIF for 24 hr. (i) Luciferase activity in HEK293T cells transfected with

845 the plasmids expressing NF-κB luciferase reporter, renilla, TRAF6, TRIF together with

846 the increasing doses of Hrd1. (j) Luciferase activity in HEK293T cells transfected with

847 the plasmids expressing NF-κB luciferase reporter, renilla, IKKβ together with GFP, WT

848 Hrd1, the transmembrane domain-deletion (ΔTM) or mitochondrial located (mito-Hrd1)

849 mutant. Similar expression levels of WT Hrd1, Hrd1 ΔTM and mito-Hrd1 were

850 confirmed by immunoblotting. (k) qRT-PCR analysis of Il6 mRNA levels in MEFs with

38

851 or without LPS (1 μg/ml) stimulation. MEFs were generated with a retrovirus

852 transduction system to stably overexpress GFP, WT Hrd1 or the mitochondrial located

853 mutant (mito-Hrd1). Similar expression levels of Hrd1 and mito-Hrd1 were confirmed by

854 immunoblotting. Scale bar, 1 μm (b), 10 μm (c); ns, not significant (P> 0.05); *P <0.05;

855 **P< 0.01 (two-tailed Student’s t-test). Data are from 3 independent experiments (mean ±

856 SEM) (d-g, k) or representative of 3 independent experiments (b-c, h-j). n≥3.

857

858 Figure 2. Hrd1 KO macrophages specifically reduce TLR4-induced inflammation

859 and NF-κB activation

860 (a) Hrd1 KO efficiency was analyzed by immunoblotting with anti-Hrd1 antibody in

861 BMMs obtained from WT and cKO mice. (b-d) Immunoblot analysis of phosphorylated

862 (p-) IKKα/β and degradation of IκBα (b) or p65 in cytoplasmic or nuclear fraction (c) or

863 p-ERk, p-p38, p-Jnk (d) in WT or Hrd1 KO PEMs upon LPS stimulation at the indicated

864 time points. Tubulin was used as cytoplasmic protein control. Lamin-B was served as

865 nuclear protein control (c). (e, f) The relative mRNA levels of Il6, Tnfa, and Il1b (e) or

866 concentrations of IL-6 and TNF-α (f) in WT or Hrd1 KO PEMs or BMMs after treatment

867 with LPS, SL1344 or E. coli for 6 hr. (g) WT or Hrd1 KO PEMs were transfected with

868 the control siRNA/N.C. or two different p65 siRNA followed by LPS stimulation to

869 measure the mRNA levels of Il6 by qRT-PCR. (h) WT or Hrd1 KO BMMs or PEMs

870 were stimulated with CpG (5 μg/ml) for 6 hr, Poly(I:C) (10 μg/ml) for 3 hr to measure the

871 relative mRNA levels of Il6 and Ifnb. ns, not significant (P> 0.05); *P <0.05; **P< 0.01

872 (two-tailed Student’s t-test). Data are from 3 independent experiments (mean ± SEM)

873 (e-h) or representative of 3 independent experiments (a, b-d). n≥3. 39

874

875 Figure 3. The E3 ligase activity of Hrd1 is critical for TLR4-induced NF-κB

876 activation independent of ERAD

877 (a) The NF-κB Luciferase readings were measured in HEK293T cells upon transfection

878 with plasmids expressing NF-κB luciferase reporter, renilla and MyD88, TRAF6 or TRIF,

879 and the GFP control, WT Hrd1 or C329S for 24 hr. Similar expression levels of WT Hrd1

880 or C329S were confirmed by immunoblotting. (b) PMA-induced THP-1 cells stably

881 overexpressing GFP control, WT Hrd1 or C329S were stimulated with LPS for 60 min,

882 immunoprecipitated with anti-IκBα antibody or rabbit IgG, followed by immunoblotting

883 with anti-IκBα antibody and anti-K48-Ubiquitin antibody to measure levels of

884 Lys48-linked polyubiquitinatin of IκBα. (c) IκBα degradation was checked by

885 immunoblotting in MEFs stably overexpressing WT Hrd1, the C329S or mito-Hrd1

886 mutant upon LPS stimulation. (d) Immunoblot analysis of p65 in the cytoplasmic or

887 nuclear fraction from PMA-induced THP-1 cells upon LPS stimulation for 30 min, which

888 stably overexpressed GFP, WT Hrd1 or C329S. (e) Confocal microscopy analysis of p65

889 nuclear translocation in MEFs before and after LPS stimulation, which stably

890 overexpressed GFP, WT Hrd1 or C329S. More than 300 cells in each sample were

891 measured to quantify p65 nuclear translocation rates. Scale bar, 20 μm. (f) WT or Hrd1

892 KO BMMs stably overexpressing with GFP, WT Hrd1 or C329S (left panel), or the

893 transmembrane domain deletion mutant ΔTM (right panel) were generated by retroviral

894 transduction system, which were treated with LPS to measure Il6 mRNA level. (g) WT or

895 Hrd1 KO PEMs were treated with LPS or tunicamycin for 6 hr, followed by qRT-PCR

896 analysis of the mRNA levels of Hspa5, Ddit3, Dnajb9 or Xbp1s. ns, not significant (P>

40

897 0.05); *P <0.05; **P< 0.01. Data are from 3 independent experiments (mean ± SEM) (f-g)

898 or representative of 3 independent experiments (a-e). n≥3. Statistical significance is

899 calculated using Student's two-tailed unpaired t-test for comparison between groups (e).

900

901 Figure 4. The ER-localised Hrd1 directly binds Usp15 and regulates TLR4-induced

902 inflammation

903 (a) The Yeast two-hybrid data to measure Hrd1 interaction with Usp15. Repeated twice.

904 (b) GST pull-down assay showed that the recombinant GST-tagged full-length Hrd1 (not

905 the transmembrane domain-deletion (ΔTM) mutant) directly bound the recombinant

906 His-tagged Usp15 (upper lane). Lower panel showed the equimolar loading

907 concentrations of GST-Hrd1 (3 μg), GST-ΔTM (2.4 μg) or GST proteins (0.92 μg). (c)

908 PEMs were stimulated with LPS at the indicated time points to detect the endogenous

909 interaction between Hrd1 and Usp15 by immunoprecipitation with anti-Usp15 antibody,

910 followed by immunoblotting with anti-Hrd1 or anti-Usp15 antibodies. (d) The levels of

911 Usp15 were measured in ER fractions from resting or LPS-stimulated PEMs by

912 immunoblotting. Calnexin, Caspase3 and Aif (apoptotic inducing factor) were

913 respectively as the control proteins of ER, cytoplasmic, mitochondrial fractions. (e)

914 HEK293T cells were transfected with Flag-tagged WT Hrd1 or ΔTM and HA-tagged

915 Usp15, followed by immunoprecipitation with anti-HA antibody, then immunoblotting

916 with the indicated antibodies. (f) The relative mRNA levels of Il6, Tnfa and Ilb by

917 qRT-PCR (f), or levels of IκBα degradation by immunoblotting (g) were checked in

918 LPS-treated PEMs after transfected with the control siRNA or two different Usp15

919 siRNAs. The knocking down efficiencies of Usp15 and the protein levels of TLR4 were 41

920 checked by immunoblotting. (h) WT or Hrd1 KO PEMs and BMMs were transfected

921 with the control siRNA or two different Usp15 siRNAs, followed by LPS stimulation to

922 check Il6 mRNA level by qRT-PCR. (i) PEMs were transfected with N.C. or two

923 different Usp15 siRNAs, followed by transfection with N.C. or two different p65 siRNAs.

924 After LPS stimulation, IL-6 concentrations were quantified by ELISA. (j) The NF-κB

925 luciferase activity was measured in HEK293T cells after transfection with plasmids

926 expressing NF-κB luciferase reporter, renilla, IKKβ, WT Usp15 or the C298A/C812A

927 mutant (M2), WT Hrd1 or C329S. Expression levels of Hrd1 or WT Usp15 and M2 were

928 shown by immunoblotting. ns, not significant (P> 0.05); *P <0.05; **P< 0.01 (two-tailed

929 Student’s t-test). Data are representative of 3 independent experiments (mean ± SEM) (f,

930 h, i) or representative of 3 independent experiments (a-e, g, j). n≥3.

931

932 Figure 5. Hrd1 promotes polyubiquitination of Lys21 in Usp15 and inactivates

933 Usp15

934 (a) HEK293T cells were transfected with His-tagged Ubiquitin (Ub), HA-tagged Usp15

935 (left and middle panels) or His-tagged Usp15 (right panel) together with GFP,

936 Flag-tagged WT Hrd1, the C329S or ΔTM mutants. Immunoprecipitation was performed

937 with anti-HA antibody or Ni-NTA to enrich Usp15, followed by immunoblotting with

938 anti-His or anti-Ub antibodies to measure Usp15 ubiquitination levels. Usp2cc (the

939 catalytic core of Usp2) was used to remove ubiquitination as a control. (b) Expression of

940 the E1, E2, HA-tagged Ub, the E3 WT or C329S Hrd1 and the control E3 HACE,

941 together with His-tagged Usp15 M2 were induced by IPTG (1 μM) in E. coli. His-tagged

942 Usp15 M2 was enriched by Ni-NTA followed by immunoblotting with anti-HA (Ub) 42

943 antibody to measure Usp15 ubiquitination levels. (c, d) WT PEMs (c), or WT and Hrd1

944 KO PEMs (d) were stimulated with LPS and prepared for immunoprecipitation with

945 anti-Usp15 antibody followed by immunoblotting with anti-Ub antibody to check Usp15

946 ubiquitination levels. (e) HEK293T cells were transfected with Hrd1 together with

947 HA-tagged Usp15 or its mutants. Immunoprecipitation was performed with anti-HA

948 (Usp15) , followed by immunoblotting with anti-His (Ub) to check Usp15 ubiquitination

949 levels. (f, g) HEK293T cells were transfected with Myc-tagged WT or C329S Hrd1

950 together with Flag-tagged Usp15 or the K21R mutant. Enriched Usp15 and K21R were

951 obtained by immunoprecipitation with anti-Flag antibody to measure deubiquitinase

952 activity in an ubiquitin-AMC assay (f). Alternatively, enriched Usp15 and K21R was

953 incubated with K48-linked poly-ubiquitin chains (Ub2-16), followed by immunoblotting

954 with anti-Ub antibody to measure mono/poly-ubiquitin levels (g). (h) The NF-κB

955 luciferase activity was checked in HEK293T cells, which were transfected with NF-κB

956 luciferase reporter, renilla, IKKβ, Flag-tagged Hrd1, different doses of HA-tagged WT

957 Usp15 or K21R for 24 hr. (i) MEFs stably overexpressing GFP, WT Usp15 or K21R,

958 were infected with E.coli for 3 hr. IL-6 production at mRNA and protein levels was

959 analyzed by qRT-PCR and ELISA. ns, not significant (P> 0.05); *P <0.05; **P< 0.01

960 (two-tailed Student’s t-test). Data are from 3 independent experiments (mean ± SEM) (i)

961 or representative of 3 independent experiments (a-h). n≥3.

962

963 Figure 6. Hrd1 deficiency protects against LPS- and CLP- induced septic shock

964 (a, e) WT and Hrd1 cKO mice were i.p. injected with LPS (35 mg/kg) or subjected with

965 CLP operations to record survival rates. Each dot represents one mouse. (b-c, f-g) Serum

43

966 IL-6 concentrations (b, f), Il6 mRNA expression in different organs (c, g) were measured

967 from WT and Hrd1 cKO mice 6hr after LPS (25 mg/kg) injection or CLP surgery. (d, h)

968 H&E staining of lungs was performed in WT and Hrd1 cKO mice 24hr after LPS (25

969 mg/kg) injection or CLP surgery. (i) WT or Hrd1 cKO PEMs were transfected with N.C.

970 or Usp15 siRNA, labeled with CFSE, and adoptively transferred into the recipient mice

971 that were previously injected with clodronate-containing liposomes to deplete

972 endogenous peritoneal macrophages. After the recipient mice were challenged in vivo

973 with LPS, CFSE+ PEMs were harvested to measure Il6 mRNA expression by qRT-PCR

974 (left panel). The KD efficiency of Usp15 was shown in the right panel. (j, k) WT and

975 Hrd1 cKO mice were i.v. injected with Poly(I:C) (100 μg/mouse) (j) or CpG (100

976 μg/mouse)(k), and expression of IFN-β and IL-6 in serum or at mRNA levels in organs

977 were measured. (l) The model. The ER-localised Hrd1 ubiquitinates and inactivates the

978 DUB Usp15 to promote TLR4-NF-κB induced inflammation independent of ERAD.

979 Scale bars in (d, h) 50 µm. ns, not significant (P> 0.05); *P< 0.05, **P< 0.01 and ***P<

980 0.001 (a-k, two-tailed Student’s t-test). Data are from 3 independent experiments (mean ±

981 SEM) of triplicate assays (b, c, f, g, i-k) or three independent experiments (d, h). n≥3.

982

983 REFERENCES

984 1. Liu J, Cao X. Cellular and molecular regulation of innate inflammatory responses. 985 Cellular & molecular immunology 2016, 13(6): 711-721. 986 987 2. Hornung V, Ellegast J, Kim S, Brzozka K, Jung A, Kato H, et al. 5'-Triphosphate RNA is 988 the ligand for RIG-I. Science 2006, 314(5801): 994-997. 989

44

990 3. Kawai T, Takahashi K, Sato S, Coban C, Kumar H, Kato H, et al. IPS-1, an adaptor 991 triggering RIG-I- and Mda5-mediated type I interferon induction. Nature immunology 992 2005, 6(10): 981-988. 993 994 4. Seth RB, Sun L, Ea CK, Chen ZJ. Identification and characterization of MAVS, a 995 mitochondrial antiviral signaling protein that activates NF-kappaB and IRF 3. Cell 2005, 996 122(5): 669-682. 997 998 5. Xu LG, Wang YY, Han KJ, Li LY, Zhai Z, Shu HB. VISA is an adapter protein required 999 for virus-triggered IFN-beta signaling. Molecular cell 2005, 19(6): 727-740. 1000 1001 6. Jeremiah N, Neven B, Gentili M, Callebaut I, Maschalidi S, Stolzenberg MC, et al. 1002 Inherited STING-activating mutation underlies a familial inflammatory syndrome with 1003 lupus-like manifestations. The Journal of clinical investigation 2014, 124(12): 1004 5516-5520. 1005 1006 7. Ishikawa H, Barber GN. STING is an endoplasmic reticulum adaptor that facilitates 1007 innate immune signalling. Nature 2008, 455(7213): 674-678. 1008 1009 8. Sun W, Li Y, Chen L, Chen H, You F, Zhou X, et al. ERIS, an endoplasmic reticulum 1010 IFN stimulator, activates innate immune signaling through dimerization. Proceedings of 1011 the National Academy of Sciences of the United States of America 2009, 106(21): 1012 8653-8658. 1013 1014 9. Alexopoulou L, Holt AC, Medzhitov R, Flavell RA. Recognition of double-stranded 1015 RNA and activation of NF-kappaB by Toll-like receptor 3. Nature 2001, 413(6857): 1016 732-738. 1017 1018 10. Phillips MJ, Voeltz GK. Structure and function of ER membrane contact sites with other 1019 organelles. Nature reviews Molecular cell biology 2016, 17(2): 69-82. 1020 1021 11. Saheki Y, De Camilli P. Endoplasmic Reticulum-Plasma Membrane Contact Sites. 1022 Annual review of biochemistry 2017, 86: 659-684. 1023 1024 12. Sullivan JS, Kilpatrick L, Costarino AT, Jr., Lee SC, Harris MC. Correlation of plasma 1025 cytokine elevations with mortality rate in children with sepsis. The Journal of pediatrics 1026 1992, 120(4 Pt 1): 510-515. 1027 1028 13. Nadav E, Shmueli A, Barr H, Gonen H, Ciechanover A, Reiss Y. A novel mammalian 1029 endoplasmic reticulum ubiquitin ligase homologous to the yeast Hrd1. Biochemical and 1030 biophysical research communications 2003, 303(1): 91-97. 1031

45

1032 14. Bordallo J, Plemper RK, Finger A, Wolf DH. Der3p/Hrd1p is required for endoplasmic 1033 reticulum-associated degradation of misfolded lumenal and integral membrane proteins. 1034 Molecular biology of the cell 1998, 9(1): 209-222. 1035 1036 15. Carvalho P, Goder V, Rapoport TA. Distinct ubiquitin-ligase complexes define 1037 convergent pathways for the degradation of ER proteins. Cell 2006, 126(2): 361-373. 1038 1039 16. Baldridge RD, Rapoport TA. Autoubiquitination of the Hrd1 Ligase Triggers Protein 1040 Retrotranslocation in ERAD. Cell 2016, 166(2): 394-407. 1041 1042 17. Xie W, Kanehara K, Sayeed A, Ng DT. Intrinsic conformational determinants signal 1043 protein misfolding to the Hrd1/Htm1 endoplasmic reticulum-associated degradation 1044 system. Molecular biology of the cell 2009, 20(14): 3317-3329. 1045 1046 18. Kaneko M, Ishiguro M, Niinuma Y, Uesugi M, Nomura Y. Human HRD1 protects 1047 against ER stress-induced apoptosis through ER-associated degradation. FEBS letters 1048 2002, 532(1-2): 147-152. 1049 1050 19. Gao B, Calhoun K, Fang D. The proinflammatory cytokines IL-1beta and TNF-alpha 1051 induce the expression of Synoviolin, an E3 ubiquitin ligase, in mouse synovial fibroblasts 1052 via the Erk1/2-ETS1 pathway. Arthritis research & therapy 2006, 8(6): R172. 1053 1054 20. Amano T, Yamasaki S, Yagishita N, Tsuchimochi K, Shin H, Kawahara K, et al. 1055 Synoviolin/Hrd1, an E3 ubiquitin ligase, as a novel pathogenic factor for arthropathy. 1056 Genes & development 2003, 17(19): 2436-2449. 1057 1058 21. Yamasaki S, Yagishita N, Sasaki T, Nakazawa M, Kato Y, Yamadera T, et al. 1059 Cytoplasmic destruction of p53 by the endoplasmic reticulum-resident ubiquitin ligase 1060 'Synoviolin'. The EMBO journal 2007, 26(1): 113-122. 1061 1062 22. Udalova IA, Mantovani A, Feldmann M. Macrophage heterogeneity in the context of 1063 rheumatoid arthritis. Nature reviews Rheumatology 2016, 12(8): 472-485. 1064 1065 23. Kopito RR. ER quality control: the cytoplasmic connection. Cell 1997, 88(4): 427-430. 1066 1067 24. Schweitzer K, Bozko PM, Dubiel W, Naumann M. CSN controls NF-kappaB by 1068 deubiquitinylation of IkappaBalpha. The EMBO journal 2007, 26(6): 1532-1541. 1069 1070 25. Torre S, Polyak MJ, Langlais D, Fodil N, Kennedy JM, Radovanovic I, et al. USP15 1071 regulates type I interferon response and is required for pathogenesis of 1072 neuroinflammation. Nature immunology 2017, 18(1): 54-63. 1073

46

1074 26. Zou Q, Jin J, Hu H, Li HS, Romano S, Xiao Y, et al. USP15 stabilizes MDM2 to mediate 1075 cancer-cell survival and inhibit antitumor T cell responses. Nature immunology 2014, 1076 15(6): 562-570. 1077 1078 27. Buchta CM, Bishop GA. TRAF5 negatively regulates TLR signaling in B lymphocytes. 1079 Journal of immunology 2014, 192(1): 145-150. 1080 1081 28. Leulier F, Lhocine N, Lemaitre B, Meier P. The Drosophila inhibitor of apoptosis protein 1082 DIAP2 functions in innate immunity and is essential to resist gram-negative bacterial 1083 infection. Molecular and cellular biology 2006, 26(21): 7821-7831. 1084 1085 29. Rowland AA, Voeltz GK. Endoplasmic reticulum-mitochondria contacts: function of the 1086 junction. Nature reviews Molecular cell biology 2012, 13(10): 607-625. 1087 1088 30. Friedman JR, Voeltz GK. The ER in 3D: a multifunctional dynamic membrane network. 1089 Trends in cell biology 2011, 21(12): 709-717. 1090 1091 31. Cockcroft S, Raghu P. Phospholipid transport protein function at organelle contact sites. 1092 Current opinion in cell biology 2018, 53: 52-60. 1093 1094 32. Kim SY, Kyaw YY, Cheong J. Functional interaction of endoplasmic reticulum stress 1095 and hepatitis B virus in the pathogenesis of liver diseases. World journal of 1096 gastroenterology 2017, 23(43): 7657-7665. 1097 1098 33. Liu S, Chen ZJ. Expanding role of ubiquitination in NF-kappaB signaling. Cell research 1099 2011, 21(1): 6-21. 1100 1101 34. Hampton RY, Gardner RG, Rine J. Role of 26S proteasome and HRD genes in the 1102 degradation of 3-hydroxy-3-methylglutaryl-CoA reductase, an integral endoplasmic 1103 reticulum membrane protein. Molecular biology of the cell 1996, 7(12): 2029-2044. 1104 1105 35. Schoebel S, Mi W, Stein A, Ovchinnikov S, Pavlovicz R, DiMaio F, et al. Cryo-EM 1106 structure of the protein-conducting ERAD channel Hrd1 in complex with Hrd3. Nature 1107 2017, 548(7667): 352-355. 1108 1109 36. Martinon F, Chen X, Lee AH, Glimcher LH. TLR activation of the transcription factor 1110 XBP1 regulates innate immune responses in macrophages. Nature immunology 2010, 1111 11(5): 411-418. 1112 1113 37. Jongsma ML, Berlin I, Wijdeven RH, Janssen L, Janssen GM, Garstka MA, et al. An 1114 ER-Associated Pathway Defines Endosomal Architecture for Controlled Cargo Transport. 1115 Cell 2016, 166(1): 152-166. 1116 47

1117 38. Elliott PR, Liu H, Pastok MW, Grossmann GJ, Rigden DJ, Clague MJ, et al. Structural 1118 variability of the ubiquitin specific protease DUSP-UBL double domains. FEBS letters 1119 2011, 585(21): 3385-3390. 1120 1121 39. Eichhorn PJ, Rodon L, Gonzalez-Junca A, Dirac A, Gili M, Martinez-Saez E, et al. 1122 USP15 stabilizes TGF-beta receptor I and promotes oncogenesis through the activation of 1123 TGF-beta signaling in glioblastoma. Nature medicine 2012, 18(3): 429-435. 1124 1125 40. Keren-Kaplan T, Attali I, Motamedchaboki K, Davis BA, Tanner N, Reshef Y, et al. 1126 Synthetic biology approach to reconstituting the ubiquitylation cascade in bacteria. The 1127 EMBO journal 2012, 31(2): 378-390. 1128 1129 41. Xu X, Li C, Gao X, Xia K, Guo H, Li Y, et al. Excessive UBE3A dosage impairs retinoic 1130 acid signaling and synaptic plasticity in autism spectrum disorders. Cell research 2018, 1131 28(1): 48-68. 1132 1133 42. Clerici M, Luna-Vargas MP, Faesen AC, Sixma TK. The DUSP-Ubl domain of USP4 1134 enhances its catalytic efficiency by promoting ubiquitin exchange. Nature 1135 communications 2014, 5: 5399. 1136 1137 43. Joosten LA, Koenders MI, Smeets RL, Heuvelmans-Jacobs M, Helsen MM, Takeda K, et 1138 al. Toll-like receptor 2 pathway drives streptococcal cell wall-induced joint inflammation: 1139 critical role of myeloid differentiation factor 88. Journal of immunology 2003, 171(11): 1140 6145-6153. 1141 1142 44. Maini RN, Taylor PC. Anti-cytokine therapy for rheumatoid arthritis. Annual review of 1143 medicine 2000, 51: 207-229. 1144 1145 45. Zhang X, Shi H, Wu J, Zhang X, Sun L, Chen C, et al. Cyclic GMP-AMP containing 1146 mixed phosphodiester linkages is an endogenous high-affinity ligand for STING. 1147 Molecular cell 2013, 51(2): 226-235. 1148 1149 46. Dobbs N, Burnaevskiy N, Chen D, Gonugunta VK, Alto NM, Yan N. STING Activation 1150 by Translocation from the ER Is Associated with Infection and Autoinflammatory 1151 Disease. Cell host & microbe 2015, 18(2): 157-168. 1152 1153 47. Liu Y, Jesus AA, Marrero B, Yang D, Ramsey SE, Sanchez GAM, et al. Activated 1154 STING in a vascular and pulmonary syndrome. The New England journal of medicine 1155 2014, 371(6): 507-518. 1156 1157 48. Inui M, Manfrin A, Mamidi A, Martello G, Morsut L, Soligo S, et al. USP15 is a 1158 deubiquitylating enzyme for receptor-activated SMADs. Nature cell biology 2011, 13(11): 1159 1368-1375. 1160 48

1161 49. Cornelissen T, Haddad D, Wauters F, Van Humbeeck C, Mandemakers W, Koentjoro B, 1162 et al. The deubiquitinase USP15 antagonizes Parkin-mediated mitochondrial 1163 ubiquitination and mitophagy. Human molecular genetics 2014, 23(19): 5227-5242. 1164 1165 50. Gerakis Y, Hetz C. Emerging roles of ER stress in the etiology and pathogenesis of 1166 Alzheimer's disease. The FEBS journal 2017. 1167

49