The TRIM37 variants in Mulibrey nanism patients paralyze follicular helper T cell differentiation

The Mulibrey (Muscle–liver–brain–eye) nanism caused by loss-of-function variants in TRIM37 gene is an autosomal recessive disorder characterized by severe growth failure and constrictive pericarditis. These patients also suffer from severe respiratory infections, co-incident with an increased mortality rate. Here, we revealed that TRIM37 variants were associated with recurrent infection. Trim37 FINmajor (a representative variant of Mulibrey nanism patients) and Trim37 knockout mice were susceptible to influenza virus infection. These mice showed defects in follicular helper T (TFH) cell development and antibody production. The effects of Trim37 on TFH cell differentiation relied on its E3 ligase activity catalyzing the K27/29-linked polyubiquitination of Bcl6 and its MATH domain-mediated interactions with Bcl6, thereby protecting Bcl6 from proteasome-mediated degradation. Collectively, these findings highlight the importance of the Trim37-Bcl6 axis in controlling the development of TFH cells and the production of high-affinity antibodies, and further unveil the immunologic mechanism underlying recurrent respiratory infection in Mulibrey nanism.


Introduction
High-affinity antibodies derived from the germinal center (GC) are critical to protective immune responses against pathogen infection 1 . The generation of highaffinity antibodies is dependent on T FH cells, which are a subset of CD4 + T cells that supports germinal center B cell differentiation 2,3 . The transcription factor Bcl6 is the master regulator of T FH cell differentiation and function. Bcl6 deletion in CD4 + T cells completely abrogates T FH cell differentiation and subsequent GC formation [4][5][6] . Bcl6 can be upregulated by T-cell receptor (TCR) stimulation, ICOS costimulation, and CD28 costimulation, especially in a milieu with the cytokines IL-12, IL-6, and IL-21, which activate STAT1 or STAT3 2,3,7 . Many T FH cell key transcription factors, including TCF1, LEF1, and BATF, positively regulate Bcl6 transcription and T FH cell differentiation [8][9][10] , while BLIMP1 4 , KLF2 11,12 , FOXO1 13 , and STAT5 [14][15][16] are negative regulators of Bcl6 transcription and T FH cell differentiation. In addition, OPN-I promotes T FH differentiation by protecting Bcl6 against ubiquitindependent proteasome degradation 17 , indicating that post-transcriptional regulation of Bcl6 plays an important role in T FH differentiation. However, the detailed mechanisms for the post-transcriptional modification of the Bcl6 protein in T FH cells are not clear.
Many studies have shown that defects in T FH cell differentiation or function lead to severe infections in humans. For example, mutations in ICOSL and Sh2d1a, the key genes instructing T FH differentiation, result in combined immunodeficiency and mortality due to recurrent infections 18,19 . Mutations in CD40L and IL21, the key genes promoting high-affinity antibody production, manifest as severe immunodeficiency with decreased T FH cells, recurrent infections, and reduced pathogenspecific antibody titers 19,20 . Hence, these studies suggest that mutations in T FH cell-related key genes may be the cause of recurrent infection in patients.
To uncover the molecular mechanism of recurrent infection, we carried out whole-exome sequencing data of patients with recurrent infection and identified TRIM37 gene variants with a statistically significant association with recurrent infection. TRIM37, an E3 ligase, contains a RING, B-box, and coiled-coil domain in the N-terminus and a specific MATH domain in the C-terminus 21 . TRIM37 is involved in many biological processes, including autophagy 22 , tumorigenesis 23,24 , post-transcriptional modification 23 , peroxisome genesis 25,26 , signal transduction 22 , and centrosome dysfunction 24,[27][28][29] . TRIM37 mutations in humans cause an autosomal recessive disorder named Mulibrey (Muscle-liver-brain-eye) nanism, characterized by severe growth failure and constrictive pericarditis 30 . A total of 26 mutations within the gene loci of TRIM37 have been identified 31 , most of which occurred in patients from Finland. In addition to growth failure, a clinical review of 85 FIN major (the so-called "Finnish major mutation") patients' hospital records from birth to diagnosis at the age of 0.02-52 years old revealed that these patients also manifest recurrent respiratory tract infections, co-incident with an increased mortality rate 32,33 . However, the mechanisms that drive recurrent respiratory tract infections remain unclear.
Here, we found that Trim37 mutant mice were also susceptible to influenza virus infection. These mice showed T FH cell differentiation and antibody production defects following vaccine immunization or influenza virus infection. We revealed that Trim37 controlled the differentiation of T FH cells in a T-cell-intrinsic manner, which also relied on its E3 ligase activity and implicated a direct interaction between Bcl6 and the MATH domain of Trim37. Trim37 catalyzed the K27/29-linked polyubiquitination of Bcl6, thereby preventing Bcl6 from proteasome-mediated degradation and promoting T FH cell differentiation. Thus, our data demonstrate the essential role of the Trim37-Bcl6 axis in the differentiation of T FH cells that underlies the recurrent respiratory infections observed in patients with Mulibrey nanism.

Results
Loss of Trim37 fails to mount protective humoral immunity against influenza virus A major goal in human genetics is to uncover the associations between natural variants and phenotypic consequences. To understand the relationships between natural variants and recurrent infection, we analyzed the whole-exome sequencing data of patients from the Children's Hospital of Fudan University. We assessed the association of altering protein-coding variants (proteintruncating variants, PTVs, missense or nonsynonymous variants, MISs) with recurrent infection in 16,330 participants (case group with recurrent infections in the main diagnosis, n = 447; control group without recurrent infections in all diagnoses during the hospital period, n = 15,883) by performing genetic Burden Test (Fig. 1a). Our screen identified 11 genes (NT5E, STK4, TBXAS1, TRIM37, SOS1, HPSE2, AFP, SEN4A, RARA, CYP2D6, FKBP10) carrying 137 variants that have a statistically significant association with the incidence of recurrent infection (P < 0.05) (Fig. 1b). Interestingly, NT5E and STK4 deficiency in humans has been reported as associating with combined immunodeficiency and recurrent infections [34][35][36] .
We also noticed that TRIM37 mutations in humans have been known to be the genetic cause of the autosomal recessive disorder Mulibrey nanism 30 . Patients with Mulibrey nanism not only have severe growth failure but also suffer from respiratory tract infections in infancy 32,33 . Echoing with these records, our findings also suggested that TRIM37 is highly associated with an increased incidence of severe infection (Fig. 1b). Considering a case report about antibody deficiency in a girl with Mulibrey nanism 37 , we speculated that TRIM37 might be involved in regulating antibody responses, and recurrent infection in Mulibrey nanism patients might be due to antibody deficiency.
The largest group of Mulibrey nanism patients carry the FIN major (c.493-2A > G) mutation, resulting in aberrant splicing at the next AG site and leading to a 164-aa truncated protein 30 . To explore whether TRIM37 is involved in regulating antibody production and recurrent infection, we constructed mice carrying the FIN major mutation (Fig. 1c-e). In the FIN major mice, we did not find significant defects in the development and homeostasis of the adaptive immune system by assaying the percentage of CD4 + and CD8 + cells in thymuses ( Supplementary Fig.  S1a) or the percentages of CD4 + T cells, CD8 + T cells and B cells in peripheral lymph organs ( Supplementary Fig.  S1b, c), as well as the percentages of naïve CD4 + T cells and naïve CD8 + T cells in peripheral lymph organs ( Supplementary Fig. S1d, e). Hence, Trim37 deficiency did not seem to impair the development of the adaptive immune system. Consistent with infertility as shown in Fig. 1 Loss of Trim37 fails to mount protective humoral immunity against the influenza virus. a The pipeline of Burden Test for recurrent infections. b Volcano plot of the genetic burden test results for cases with recurrent infections. Points indicate a certain gene with the corresponding log 2 (OR) and -log 10 (P value) of the Fisher test results. The threshold of OR > 2 and the threshold of P < 0.05 are marked by two red dashed lines. Genes above the threshold are shown as green dots, while TRIM37 is shown as a red dot. c The schematic design of FIN major mice. d Sanger sequencing of the WT and FIN major mice. e Schematic presentation of WT Trim37 and FIN major mutant (up) and immunoblot analysis of Trim37 levels in spleen samples obtained from the WT and FIN major mice (down). f The weight loss of WT (n = 4) and FIN major (n = 4) mice following PR8 influenza (0.5 LD 50 ) infection. g Histopathological analysis was performed by H&E staining 13 days after PR8 influenza (0.5 LD 50 ) infection. h WT (n = 3) and FIN major (n = 3) mice were infected with PR8 influenza (0.5 LD 50 ), and the sera were obtained 13 days post-infection. Viral-specific anti-HA IgG in the sera was measured by ELISAs. i-l WT and FIN major mice were vaccinated with 10 µg of PR8 vaccine adjuvanted with alum intraperitoneally and boosted with the same agents 2 weeks later. Serum samples were collected 14 days after the second vaccination. These mice were intranasally challenged with a high dose of PR8 influenza virus (10 LD 50 ) 19 days after the second vaccination. i The weight loss of WT (n = 4) and FIN major (n = 4) mice following PR8 influenza (10 LD 50 ) infection. The gray line (WT unV, n = 2) shows the body weight loss of naive control mice without vaccination. j Virus titers of the lungs 4 days post-infection (n = 3). k Mice were sacrificed on day 9 after influenza virus infection, and lung tissue was collected. Histopathological analysis was performed by H&E staining. l Viral-specific anti-HA IgG in the sera was measured by ELISAs (WT, n = 4; FIN major , n = 5; WT unV, n = 3). Data are representative of one (b), two (i-l) or at least three (c-h) independent experiments. Data were analyzed by Fisher test (b).
Mulibrey nanism patients and Trim37-deficient mice 38,39 , our in-house developed strains carrying FIN major variants have shown the infertility phenotype at 8 weeks. Interestingly, after we further infected these mice with the Rico/8/34 (PR8, H1N1) influenza virus (0.5 LD 50 ), the FIN major mice showed more weight loss at 10 days postinfection than the wild-type (WT) mice (Fig. 1f) and had much more severe lung histopathology than the WT mice (Fig. 1g). Highly reminiscent of the respiratory infections in the patients with Mulibrey nanism, these data clearly indicated that FIN major mice are susceptible to influenza virus infection and of a major defect in the production of anti-influenza IgG (Fig. 1h).
To verify whether the susceptibility to influenza virus infection is due to antibody deficiency, we immunized FIN major mice with two doses of the PR8 vaccine and challenged these mice with a lethal dose of the PR8 influenza virus (10 LD 50 ) 19 days after the second vaccination ( Supplementary Fig. S1f). As expected, when compared to the WT mice, the FIN major mice had greater body weight loss (Fig. 1i), harbored a higher amount of virus (Fig. 1j), and showed more severe damage in the lung (Fig. 1k), indicating insufficient protection that vaccinations may usually confer. Unsurprisingly, the FIN major mice had a considerably lower level of anti-HA IgG in serum (Fig. 1l).
T H 1 and cytotoxic CD8 + T-cell-mediated cellular immune responses are also important for protective immunity against pathogen invasion. In our study, we did not find a significant difference between IFN-γ + CD4 + and IFN-γ + CD8 + cells in the FIN major mice (Supplementary Fig. S1g-j). Taken together, our data show that the FINmajor variant of Trim37 leads to poor protective antibody responses upon viral infection or vaccination, which may account for the severe respiratory tract infections often observed with Mulibrey nanism patients.

Trim37 is required for T FH cell differentiation
T FH cells specifically support GC formation and thus play an essential role in antibody responses. To obtain insight into the Trim37-mediated regulation of protective antibody responses upon viral infection or vaccination, we infected the FIN major mice and their counterpart WT mice with PR8 influenza viruses through an intranasal challenge and analyzed the T FH cells and the antibody responses induced by viral infection. We found that CXCR5 + PD-1 + T FH cells ( Fig. 2a; Supplementary Fig.  S2a) and CXCR5 + Bcl6 + T FH cells ( Fig. 2b; Supplementary  Fig. S2b) in the lung-draining lymph nodes and the spleen were largely compromised (~50% loss) in the FIN major mice as compared to those in the WT mice. Consistently, a considerable decrease in GC B cells ( Fig. 2c; Supplementary Fig. S2c), and B220 lo CD138 hi plasma cells ( Fig.  2d; Supplementary Fig. S2d) was observed in lung-draining lymph nodes and the spleen. In addition, immunofluorescence staining of B220 and GL7 provided visual images that indicated reduced GC reaction in the spleen of the FIN major mice (Fig. 2e). Besides, we also observed a spontaneous defect in T FH cell and GC B cell differentiation even in the uninfected FIN major mice ( Supplementary Fig. S2e, f). Thus, these data suggest that Trim37 plays a critical role in T FH cell differentiation and GC formation.
To further pinpoint the critical role of Trim37 in regulating T FH cell differentiation, we immunized WT mice and FIN major mice with the PR8 vaccine to analyze T FH cell and GC B cell responses. Ten days after immunization, we found that the percentage of CXCR5 + PD-1 + T FH cells was dramatically decreased in the FIN major mice (Fig.  2f). Moreover, we observed a considerable decrease in B220 + Fas + CD38 lo GC B cells (Fig. 2g). These data were highly consistent with our observation that B cells from the FIN major mice had a diminished ability to produce fluspecific IgG (Fig. 1h). Thus, Trim37 is critical for GC responses and high-affinity antibody production following vaccine immunization. Taken together, these data indicate that Trim37 displays a remarkable impact on the GC response and the production of high-affinity neutralizing antibodies against influenza virus infection and vaccine immunization.
To further confirm the functional effects of Trim37 on T FH cell differentiation, we constructed another Trim37 ko mouse carrying a 7 bp insertion in the exon 4 of Trim37 ( Supplementary Fig. S2a, b). Like in FIN major mice, we did not find any defects in the development and homeostasis of the adaptive immune system in the Trim37 ko mice (Supplementary Fig. S3c-g). The Trim37 ko mice were also susceptible to influenza virus infection and recovered slowly until 12 days after intranasal challenge (Supplementary Fig. S4a, b), with more severe lung histopathology than the WT mice ( Supplementary Fig. S4c). We found that the frequency of CXCR5 + PD-1 + and CXCR5 + Bcl6 + T FH cells in the lung-draining lymph nodes is decreased by over 50% in the Trim37 ko mice as compared to those in the WT mice ( Supplementary Fig.  S4d, e). Consequently, there was a considerable decrease in GC responses ( Supplementary Fig. S4f-h). The Trim37 ko mice also exhibited a significant reduction in the secretion of influenza-specific and neutralizing antibodies, as detected by ELISAs and microneutralizing assays, respectively ( Supplementary Fig. S4i, j). Hence, the T FH cell differentiation and GC responses are impaired in both FIN major and Trim37 ko mice.

Trim37 promotes T FH cell differentiation in a T-cell-intrinsic manner
We examined the Trim37 expression level in T FH cells. We infected WT mice with influenza viruses and sorted non-T FH cells (CXCR5 -PD-1 -) and T FH cells (CXCR5 + PD-1 + ) 10 days after PR8 influenza virus infection (Fig. 3a). RT-PCR assays revealed that Trim37 mRNA was highly expressed in T FH cells (Fig. 3b). Then, we took advantage of Trim37 3×Flag knock-in mice with 3× Flag at the N-terminus of Trim37 to determine Trim37 protein levels in T FH cells (Fig. 3c). We infected the Trim37 3×Flag mice with influenza viruses and sorted non-T FH cells and T FH cells 10 days post-infection. Immunoblot analysis revealed that Trim37 protein levels were also higher in T FH cells than in non-T FH cells (Fig. 3d), suggesting that Trim37 might specifically regulate T FH cell differentiation and function.
We next determined whether Trim37 controls T FH cell differentiation in a T-cell-intrinsic manner. We reconstituted lethally irradiated Tcrb -/mice with mixed CD45.2 FIN major and CD45.1 WT bone marrow (BM) cells at a ratio of 1:1 to generate mixed chimeric mice. Then, these mixed bone marrow chimeric mice were infected with PR8 influenza viruses. T FH cell differentiation was analyzed 12 days post-infection ( Supplementary  Fig. S5a). Under the same immune environment, CD4 + T cells derived from CD45.2 + FIN major BM cells showed an impaired ability to differentiate into T FH cells compared with CD4 + T cells derived from CD45.1 + WT BM cells (Fig. 3e, f). Likewise, we reconstituted lethally irradiated Tcrb -/mice with mixed CD45.2 Trim37 ko and CD45.1 WT BM cells at a ratio of 1:1 to generate mixed chimeric mice ( Supplementary Fig. S5b). CD4 + T cells derived from CD45.2 + Trim37 ko BM cells also showed . These mice were sacrificed on day 12 after infection. a Representative flow cytometry plots illustrate the frequency of CXCR5 + PD1 + T FH cells as a percentage of CD4 + Foxp3 -T cells in the lung draining lymph nodes, and quantification of CXCR5 + PD1 + T FH cells. b Representative flow cytometry plots illustrate the frequency of CXCR5 + Bcl6 + T FH cells as a percentage of CD4 + Foxp3 -T cells in the lung-draining lymph nodes, and quantification of CXCR5 + Bcl6 + T FH cells. c Representative flow cytometry plots illustrate the frequency of Fas + PNA + GC B cells as a percentage of B220 + B cells in the lung-draining lymph nodes, and quantification of GC B cells. d Representative flow cytometry plots illustrate the frequency of B220 lo CD138 hi plasma cells as a percentage of live cells in the lung-draining lymph nodes, and quantification of plasma cells. e Confocal microscopy of the spleen's germinal center (B220 + GL7 + ). f, g WT (n = 5) and FIN major (n = 5) mice were vaccinated with 10 µg of PR8 vaccine adjuvanted with alum intraperitoneally. Spleens were collected 10 days after vaccination. f Representative flow cytometry plots illustrate the frequency of CXCR5 + PD1 + T FH cells as a percentage of CD4 + Foxp3 -T cells in the spleens and quantification of CXCR5 + PD1 + T FH cells. g Representative flow cytometry plots illustrate the frequency of Fas + CD38 lo GC B cells as a percentage of B220 + B cells in the spleens, and quantification of GC B cells. Data are representative of at least three independent experiments and were analyzed by two-tailed unpaired Student's t-test (a-d, f, g). Data are mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, ns, not significant. Fig. S5c, d). Thus, Trim37 controls T FH cell differentiation in a T-cell-intrinsic manner.

deficient T FH cell differentiation compared with CD4 + T cells derived from WT BM cells 12 days after influenza virus infection (Supplementary
To uncover the specific function of Trim37 in T cells, we generated conditional Trim37 knockout mice (referred to as Trim37 cko ) by crossing Trim37 flox mice (Supple- Similarly, conditional knockout of Trim37 in T cells led to significantly decreased T FH differentiation (Fig. 3g, h), GC formation (Fig. 3i), and plasma cell differentiation (Fig. 3j) in the lung draining lymph nodes 12 days after PR8 virus infection. Meanwhile, Trim37 deficiency seemed not to impair the activation and proliferation of CD4 + T cells ( Supplementary Fig. S6a, b). In addition, we found that Trim37 did not affect iT reg , T H 1, T H 2, or T H 17 cell differentiation in vitro and in vivo (Supplementary Fig.  S6c-j). Collectively, these data show that Trim37 governs the GC response by directly regulating T FH cell differentiation.

Trim37 promotes Bcl6 stability in T FH cell differentiation
We then asked how Trim37 regulates T FH cell differentiation. Bcl6 is the master regulator of T FH cell differentiation. We found that the protein levels of Bcl6 were significantly decreased in the Trim37 ko CXCR5 + PD1 + T FH cells (Fig. 4a). However, Trim37 deficiency did not affect Bcl6 mRNA levels in T FH cells (Fig. 4b). These results suggested that Bcl6 might be regulated by Trim37 Fig. 3 Trim37 promotes T FH cell differentiation in a T-cell-intrinsic manner. a Sorting of CXCR5 -PD1non-T FH cells and CXCR5 + PD1 + T FH cells from lung-draining lymph nodes and spleens in the C57BL/6 mice infected with PR8 influenza virus. b Cxcr5, Bcl6, and Trim37 mRNA expressions in non-T FH cells and T FH cells were measured by RT-PCR. c Schematic design of the Trim37 3×Flag mice. d CXCR5 -PD1non-T FH cells and CXCR5 + PD1 + T FH cells were sorted from lung-draining lymph nodes and spleens in the Trim37 3×Flag mice infected with PR8 influenza virus. Flag, Bcl6, and β-actin protein expression was measured by Western blotting. e, f FIN major bone marrow chimeric mice were infected intranasally with PR8 H1N1 influenza virus (0.5 LD 50 ) and sacrificed on day 12 after infection. Representative flow cytometry plots showing the frequency of CD45.1 + (WT, n = 4) or CD45.2 + (FIN major , n = 4) CXCR5 + PD1 + T FH cells (e) and CXCR5 + GL7 + GC-T FH cells (f) as a percentage of CD4 + T cells in the spleens. g-j Flow cytometry analysis of lymphocytes from WT (n = 3) and Trim37 cko (n = 4) mice at day 10 after infection of PR8 H1N1 influenza virus (0.5 LD 50 ). g, h The frequency and quantification of CXCR5 + PD1 + (g) and CXCR5 + Bcl6 + (h) T FH cells as a percentage of CD4 + T cells in the draining lymph nodes. i, j The frequency and quantification of Fas + PNA + GC B (i) cells and IgD lo CD138 hi plasma cells (j) as a percentage of B220 + B cells in the draining lymph nodes. Data are representative of at least three independent experiments, and were analyzed by two-tailed unpaired Student's t-test (e-g, j), or twoway ANOVA (b). Data are mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, ns, not significant.
in a post-translational manner. In T cells, anti-CD3 and anti-ICOS stimulation can upregulate Bcl6 expression 17 . After addition of CHX to cultures of CD4 + T cells followed by stimulation with anti-CD3 and anti-ICOS antibodies, we observed a significant decrease in the stability of Bcl6 protein of the Trim37 ko CD4 + T cells compared to the WT CD4 + T cells (Fig. 4c). Moreover, we obtained consistent results when we used flow cytometry to detect the expression of Bcl6 protein ( Supplementary Fig. S7a). Thus, Trim37 regulates T FH cell differentiation by promoting Bcl6 stability.
We next investigated how Trim37 regulates Bcl6 expression. Through co-immunoprecipitation assays and confocal experiments, we demonstrated that Trim37 and Bcl6 could directly interact with each other (Fig. 4d, e). Moreover, we investigated whether Trim37 interacted with the Bcl6 protein in T cells. We took advantage of Bcl6 TST knock-in mice with Twin-Strep-Tag (TST) at the C-terminus of Bcl6 (Fig. 4f). CD4 + T cells from WT and Bcl6 TST mice were cultured under T FH-like conditions and stimulated with anti-CD3 and anti-ICOS for 4 h, followed by affinity enrichment of TST with Strep-tactin. Immunoblot experiments indicated that Trim37 and Bcl6 could directly interact with each other in T cells (Fig. 4f). Together, these data suggest that Trim37 is associated with Bcl6.

Trim37-mediated ubiquitination promotes the stability of Bcl6 and differentiation of T FH cells
Given that Trim37 is an E3 Ub ligase, we assessed whether Trim37 could directly ubiquitinate Bcl6 to regulate T FH cell differentiation. In a reconstituted Escherichia coli (E. coli) ubiquitination system 40 , we observed a strong ubiquitination signal of Bcl6 when Trim37 and Bcl6 were co-transformed into competent E. coli BL21 cells. Bcl6 could be ubiquitinated by WT Trim37 but not the enzymatically inactive mutant (Trim37 C18R ) 23 (Fig. 5a). Consistently, we also found that Bcl6 was ubiquitinated by WT Trim37 but not Trim37 C18R in mammalian cells (Fig. 5b). These findings suggest that Trim37 targets Bcl6 for ubiquitination.
To investigate whether Trim37-mediated K27/29-linked polyubiquitination leads to enhanced Bcl6 expression, we Trim37 binds directly to Bcl6. a, b CXCR5 -PD1non-T FH cells CXCR5 + PD1 + T FH cells were sorted from lung-draining lymph nodes and spleens in the WT and Trim37 ko mice infected with PR8 influenza virus. a Bcl6 protein expression was measured by Western blot analysis. b Bcl6 mRNA expression was measured by RT-PCR. c The WT and Trim37 ko CD4 + T cells were cultured under T FH-like conditions and stimulated with anti-CD3 and anti-ICOS for 4 h. Then, CHX was added to the medium for the indicated hours. Immunoblot analysis of Bcl6 in lysates of CD4 + T cells was performed. Residual Bcl6 protein was normalized to actin and presented relative to that before the addition of CHX (right). d Immunoblot analysis of lysates from the HEK293T cells transfected with HA-tagged Trim37 and Flag-tagged Bcl6 plasmids, followed by immunoprecipitation with anti-Flag affinity gel and immunoblot analysis with anti-HA or anti-Flag antibodies (Abs). e Confocal microscopy of the HEK293T cells transfected with HA-tagged Trim37 and Flag-tagged Bcl6 plasmids. Immunofluorescence was performed with anti-HA (green) and anti-Flag (red) Abs. Nuclei were stained with DAPI (blue). f The schematic design of Bcl6 TST mice (up). The WT and Bcl6 TST CD4 + T cells were cultured under T FH-like conditions and stimulated with anti-CD3 and anti-ICOS for 4 h, followed by affinity enrichment of TST with Strep-tactin and immunoblot analysis with anti-Bcl6, anti-Trim37 and anti-actin Abs (down). Data are representative of at least three independent experiments, and were analyzed by two-way ANOVA (b). Data are mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, ns, not significant.
co-expressed Bcl6 with Trim37 or E3 ligase dead Trim37 C18R mutant in HEK293T cells. The overexpression of WT Trim37, but not Trim37 C18R mutant up-regulated the level of Bcl6 protein in HEK293T cells (Fig. 5d). We also introduced the protein synthesis inhibitor cycloheximide (CHX) to treat HEK293T cells that overexpressed WT Trim37 or Trim37 C18R . Immunoblot experiments indicated that WT Trim37 but not Trim37 C18R prolonged the stability of Bcl6 (Fig. 5e). These data suggest that Trim37 could promote the stability of Bcl6 via its E3 ligase activity.
To determine whether Trim37 controls T FH cell differentiation through its ubiquitin E3 ligase activity in vivo, we generated Trim37 C18R knock-in mice ( Supplementary  Fig. S7b, c). Then, we reconstituted lethally irradiated Tcrb -/mice with Trim37 C18R/C18R or WT BM cells to generate bone marrow chimeric mice and challenged these chimeric mice with PR8 influenza virus  S7d). On day 10 post-infection, the Trim37 C18R/C18R bone marrow chimeric mice showed fewer T FH cells than the WT bone marrow chimeric mice (Fig. 5f, g). Consistent with this observation, the Trim37 C18R/C18R bone marrow chimeric mice had dramatically decreased GC B cells (Fig. 5h) and plasma cells (Fig. 5i). We also found that the Trim37 C18R/C18R chimeric mice had a diminished ability to produce HAspecific IgG and IgG1, as shown by ELISA (Fig. 5j, k). All these data demonstrate that the E3 ligase activity of Trim37 is essential to control T FH differentiation, by promoting the stability of Bcl6.
Trim37 ubiquitinates Bcl6 at the K227, K302, K327, K535, and K689 residues A previous study demonstrated that canonical ubiquitination sites in proteins bore Gly-Gly adducts to the side chain of lysine (K) residues 40 . Gly-Gly adducts were found on five lysines (K227, K302, K327, K535, K689) in Bcl6, as revealed by mass spectrum analysis (Fig. 6a). Then, we simultaneously replaced all 5 Bcl6 lysine residues with arginine (5KR) and co-expressed ubiquitin, Trim37, and the WT (Bcl6 WT ) or 5KR mutant (Bcl6 5KR ) of Bcl6 in HEK293T cells. The in vivo ubiquitination assay showed that the ubiquitination of the Bcl6 5KR was significantly attenuated compared to that of the Bcl6 WT (Fig. 6b). In addition, the Bcl6 5KR displayed decreased stability in the presence of Trim37 (Fig. 6c).
Moreover, we investigated whether this mutant Bcl6 can influence the function of Bcl6 in vivo. To test this hypothesis, we transduced Bcl6-deficient OT-II CD4 + T cells with retroviruses expressing GFP alone (RV-GFP), WT Bcl6 (Bcl6 WT -GFP), or 5KR mutant Bcl6 (Bcl6 5KR -GFP), transferred these cells into B6 WT mice and analyzed T FH cell differentiation 7 days after NP-OVA plus alum immunization ( Supplementary Fig. S8a). Consistently, we observed a sharp decrease in T FH cell differentiation in the OT-II cells with Bcl6 5KR -GFP compared to the OT-II cells with Bcl6 WT -GFP (Fig. 6d). Compared to the Bcl6 mRNA level, we observed a slightly low expression of Bcl6 protein in Bcl6 5KR -GFP + T cells before adoptive transfer ( Supplementary Fig. S8b, c). We investigated whether this mutant Bcl6 can influence the function of Bcl6. Compared to the overexpression of Bcl6 WT in CD4 + T cells, overexpression of Bcl6 5KR barely induces the expression of CXCR5 and PD-1 (Fig. 6e, f). In contrast, overexpression of Bcl6 WT in CD4 + T cells represses Prdm1, Il7r, Ifngr1, Runx3, Gata3, Klf2, Ccr7 and S1pr1, while overexpression of Bcl6 5KR in CD4 + T cells barely represses expression of these genes ( Fig. 6g; Supplementary Fig. S8d). All these data indicated that Trim37-mediated non-proteolytic ubiquitination of Bcl6 at the 5 lysine residues, is critical for the stability of Bcl6 and significantly impacts the transcription of its target genes in T cells.

Trim37 promotes the stability of Bcl6 and differentiation of T FH cells dependent on the MATH domain
Trim37 contains the MATH domain on its C-terminal following the RBCC (RING-B-box-coiled-coil) domain, which is predicted to mediate protein-protein interactions 41 . To explore whether the MATH domain governed the interaction between Trim37 and Bcl6, we performed coimmunoprecipitation experiments by overexpressing full-length Trim37 or various Trim37 truncations with Bcl6 (Fig. 7a). The results showed that full-length Trim37 and truncated Trim37 mutants, except the MATH domain-deleted mutant, were able to immunoprecipitate with Bcl6 (Fig. 7b). These data indicated that Trim37 interacts with Bcl6 via its MATH domain.
(see figure on previous page) Fig. 5 Trim37 targets Bcl6 for K27/29-linked ubiquitination. a Reconstituting the E. coli ubiquitination system by transforming the pACYC vector that expresses HA-tagged Ub, E1, and E2, with or without Trim37 WT or Trim37 C18R , along with the pET22b-His-Bcl6 vector into BL21 cells. This process was followed by affinity enrichment (Ni-NTA pulldown) of ubiquitinated Bcl6 and immunoblot analysis of Bcl6 expression with anti-His, anti-HA, and anti-Trim37 Abs. b Immunoblot analysis of lysates obtained from the HEK293T cells transfected with Flag-tagged Bcl6, His-tagged ubiquitin, and HAtagged empty vector, Trim37 WT , or Trim37 C18R . This process was followed by immunoprecipitation with an anti-Flag affinity gel and analysis with anti-HA, anti-Flag, and anti-His Abs. c Immunoblot analysis of lysates obtained from the HEK293T cells transfected with various combinations of plasmids such as HA-tagged Trim37 and Flag-tagged Bcl6 with His-tagged WT-Ub, K6-Ub, K11-Ub, K27-Ub, K29-Ub, K33-Ub, K48-Ub, K63-Ub, and all K-R mutated Ub. Then, assays were performed as in b. d Immunoblot analysis of Bcl6 in lysates of the HEK293T cells transfected with HA-tagged Bcl6 and increasing doses of His-tagged Trim37 WT and Trim37 C18R for 24 h. Relative Bcl6 protein, normalized to actin (down). e Immunoblot analysis of Bcl6 in lysates obtained from the HEK293T cells transfected with HA-tagged Bcl6 and empty vector, Trim37 WT , Trim37 C18R for 24 h and then treated with CHX for the indicated hours. Residual Bcl6 protein was normalized to actin and presented relative to that before the addition of CHX (down). f-k WT (n = 4) and Trim37 C18R/C18R (n = 4) bone marrow chimeric mice were infected with influenza virus for 10 days, and the spleen and serum were collected for the following assays. f, g Flow cytometry analysis of lymphocytes from WT and Trim37 C18R/C18R bone marrow chimeric mice. The frequency and quantification of CXCR5 + PD1 + (f) and CXCR5 + Bcl6 + (g) T FH cells as a percentage of CD4 + T cells in the draining lymph nodes. h, i The frequency and quantification of Fas + PNA + GC B (h) cells and IgD lo CD138 hi plasma cells (i) as a percentage of B220 + B cells in the draining lymph nodes. j, k Viral-specific anti-HA IgG (j) and IgG1 (k) in the sera were measured by ELISAs. Data are representative of two (f-k) or at least three (a-e) independent experiments, and were analyzed by two-tailed unpaired Student's t-test (f-i). Data are mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, ns, not significant.
To investigate whether the Trim37-Bcl6 interaction is required for Bcl6 ubiquitination, we co-transfected HA-Trim37 △MATH , Flag-Bcl6, and His-tagged ubiquitin and harvested cells 28 h after transfection. Expression of the MATH domain-deleted Trim37 with Bcl6 did not result in its ubiquitination (Fig. 7c). A c.965G > T (p.G322V) mutation located in the MATH domain of TRIM37 was described in Mulibrey nanism patients 42 . Interestingly, this G322V mutation (Trim37 G322V ) abolished the ability of Trim37 to ubiquitinate Bcl6 (Fig. 7c). These data indicated that Trim37 ubiquitinates Bcl6 in a MATH domain-dependent manner.
To further confirm whether the Trim37-Bcl6 interaction is required for the stability of Bcl6, we also introduced CHX to treat HEK293T cells that overexpressed WT Trim37, Trim37 △MATH or Trim37 G322V . Immunoblot experiments indicated that expression of MATH domain-deleted and G322V mutant Trim37 led to the instability of Bcl6 (Fig. 7d). These data suggest that Trim37-Bcl6 interaction is essential for promoting the stability of Bcl6.
To explore the functional consequence of the G322V mutation in vivo, we constructed mice carrying the Trim37 G322V mutation ( Supplementary Fig. S9a, b) and Fig. 6 Trim37 ubiquitinates Bcl6 at the K227, K302, K327, K535, and K689 residues. a Reconstituting the E. coli ubiquitination system by transforming the pACYC vector that expresses HA-tagged Ub, E1, and E2, with or without Trim37 WT , along with the pET22b-His-Bcl6 vector into BL21 cells. This step was followed by two-step affinity enrichment. The supernatant containing ubiquitinated Bcl6 was analyzed by mass spectrometry. Site mapping for ubiquitination of Bcl6 is shown (right). b Immunoblot analysis of lysates obtained from the HEK293T cells transfected with Flag-tagged Bcl6 WT or Bcl6 5KR and plasmids as indicated. This step was followed by immunoprecipitation with an anti-Flag affinity gel and analysis with anti-HA, anti-Flag, and anti-His Abs. c HEK293T cells were transfected with HA-tagged Bcl6 WT or Bcl6 5KR and Trim37 WT . Then, the cells were treated with CHX for the indicated hours. Immunoblot analysis of Bcl6 in lysates was performed. Residual Bcl6 protein was normalized to actin and presented relative to that before the addition of CHX (right). d OT-II Bcl6-deficient CD4 + T cells transduced with RV, Bcl6 WT, or Bcl6 5KR retroviruses, and then transfer to WT mice (RV, n = 4, Bcl6 WT , n = 3, Bcl6 5KR , n = 4). Representative flow cytometry plots showing the frequency of CXCR5 + PD1 + T FH cells as a percentage of CD4 + GFP + T cells in the spleen on day 7 after NP-OVA/alum immunization. e The median fluorescence index (MFI) of CXCR5 and PD1 in GFP + OT-II T cells transduced with RV (n = 4), Bcl6 WT (n = 3), or Bcl6 5KR (n = 4) retroviruses. f The mRNA expression of Cxcr5 and Pdcd1 was detected in T H 0 cells transduced with RV, Bcl6 WT , or Bcl6 5KR retroviruses. g The mRNA expression of Prdm1, Il7r, Ifngr1 and Runx3 was detected in T H 0 cells transduced with RV, Bcl6 WT, or Bcl6 5KR retroviruses. Data are representative of one (a) or at least three (b-g) independent experiments, and were analyzed by One-way ANOVA (d-g). Data are mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, ns, not significant. b Immunoblot analysis of lysates obtained from the HEK293T cells transfected with Flag-tagged Bcl6 and HA-tagged full-length Trim37 or Trim37 mutants as indicated, followed by immunoprecipitation with anti-Flag affinity gel and immunoblot analysis with anti-HA or anti-Flag Abs. c Immunoblot analysis of lysates obtained from the HEK293T cells transfected with Flag-tagged Bcl6, His-tagged ubiquitin, and the HA-tagged vectors Trim37 WT , Trim37 △MATH , Trim37 G322V . This process was followed by immunoprecipitation with an anti-Flag affinity gel and analysis with anti-HA, anti-Flag, and anti-His Abs. d Immunoblot analysis of Bcl6 in lysates of the HEK293T cells transfected with HA-tagged Bcl6 and empty vector, Trim37 WT , Trim37 △MATH , and Trim37 G322V . Then, the cells were treated with CHX for the indicated hours. Residual Bcl6 protein was normalized to actin and presented relative to that before the addition of CHX (right). e, f WT and Trim37 G322V/G322V bone marrow chimeric mice were infected with the influenza virus for 10 days, and the spleens were collected for FACS analysis. e Representative flow cytometry plots showing the frequency of CD45.1 + (WT, n = 4) or CD45.2 + (Trim37 G322V/G322V , n = 4) CXCR5 + PD1 + T FH cells as a percentage of CD4 + T cells in the spleen. f Representative flow cytometry plots showing the frequency of CD45.1 + (WT, n = 4) or CD45.2 + (Trim37 G322V/G322V , n = 4) CXCR5 + Bcl6 + T FH cells as a percentage of CD4 + Foxp3 -T cells in the spleen. g Schematic model for the Trim37-mediated ubiquitination of Bcl6 and regulation of T FH cell differentiation. Data are representative of two (e, f) or at least three (a-d) independent experiments, and were analyzed by two-tailed paired Student's t-test (e, f). Data are mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, ns, not significant. generated Trim37 G322V/G322V mixed bone marrow chimeric mice (Supplementary Fig. S9c). Ten weeks after bone marrow reconstitution, we infected these chimeric mice with the PR8 influenza virus. As shown in (Fig. 7e, f), the Trim37 G322V/G322V CD4 + T cells had impaired T FH cell differentiation, compared to the WT CD4 + T cells. Altogether, these data demonstrated that Trim37 promotes the stability of Bcl6 dependently on its MATH domain, and the G322V mutation disrupts the function of Trim37 in promoting T FH cell differentiation.

Discussion
Here we establish the causality between the E3 Ub ligase TRIM37 and severe respiratory infection (Fig. 7g). Through a genetic burden test, we identified TRIM37 variants significantly associated with recurrent infection. By constructing multiple Trim37 mutant mice (FIN major , Trim37 ko , Trim37 cko , Trim37 C18R/C18R , Trim37 G322V/G322V ), we found that respiratory tract infections and antibody deficiency in Mulibrey nanism patients could be ascribed to TRIM37mediated regulation of protective antibody responses. Trim37-mediated ubiquitination prolonged the stability of Bcl6 and subsequently promoted differentiation of T FH cells. Thus, our findings suggest a previously unknown Trim37-Bcl6 axis that regulates T FH cell differentiation and high-affinity antibody production.
TRIM37 mutations found in humans cause a rare autosomal recessive disorder characterized by severe prenatal-onset growth failure, infertility, cardiomyopathy, fatty liver, type 2 diabetes, and tumorigenesis named Mulibrey nanism 30 . A clinical review of 85 FIN major patients' hospital records revealed that these patients suffer from respiratory tract infections, co-incident with an increased mortality rate 32,33 . All this information indicated the importance of TRIM37 in protective immunity against infection. In light of a previous case report about antibody deficiency in a girl with Mulibrey nanism 37 , TRIM37 might play a critical role in the control of T FH cell differentiation, germinal center formation, and antibody production. Our study then took advantage of genetic mouse models to address the exact cause of susceptibility to infection in patients with Mulibrey nanism. We found that Trim37 exerts a remarkable impact on the T FH cell response and the production of neutralizing antibodies against influenza infection in the Trim37 mutant mice. Moreover, the FIN major mice have poor protective humoral immunity followed by influenza vaccine immunization, resulting in susceptibility to influenza virus infection. Our results show that respiratory tract infections and antibody deficiency in Mulibrey nanism patients could be ascribed to deficient T FH cell and germinal center B cell responses.
We found that Trim37 may regulate Bcl6 levels in a post-transcriptional mechanism. Indeed, Trim37 directly interacts with Bcl6 and targets Bcl6 predominantly for K27/29-linked ubiquitination. Trim37 prolongs the stability of Bcl6 dependent on its E3 Ub ligase activity, while Trim37 mutant mice carrying the enzymatically inactive Trim37 C18R display a diminished ability to differentiate into T FH cells and an insufficient production of antiinfluenza high-affinity antibodies. All these findings demonstrated that Trim37 is the direct E3 Ub ligase of Bcl6 and that Trim37 promotes T FH cell differentiation by ubiquitinating and stabilizing Bcl6. Moreover, Trim37 mediates non-proteolytic ubiquitination of Bcl6 at 5 key lysines (K227, K302, K327, K535, and K689) residues. This 5KR mutant Bcl6 showed an abrogated ubiquitination level, resulting in decreased stability of Bcl6 and diminished T FH cell differentiation. Trim37 contains the MATH domain following the RBCC domain that is predicted to mediate protein-protein interactions 41 . As expected, Trim37 interacts with and ubiquitinates Bcl6 in a manner dependent on its MATH domain. Interestingly, the c.965 G > T (p.G322V) mutation located in the MATH domain of TRIM37 was described in Mulibrey nanism patients 42 . This G322V mutation can disrupt the Bcl6 ubiquitination by Trim37 and consequently dampen T FH cell differentiation. Taken together, these data established that the Trim37-Bcl6 interaction was essential for promoting the stability of Bcl6 and the differentiation of T FH cells. Furthermore, we generated Trim37 bko mice by crossing Trim37 fl/fl mice with Cd19-cre knock-in mice, and the preliminary study suggests that Trim37 can also directly regulate the differentiation of germinal center B cell differentiation (Data not shown). How Trim37 regulates the function of germinal center B cells and whether it regulates germinal center B cell differentiation by regulating Bcl6 ubiquitination remain for further research.
Moreover, post-transcriptional regulation also affects BCL6 protein levels in B lymphocytes 43 . For example, phosphorylation of BCL6 by MAPK results in BCL6 degradation 44 , and FBXO11 mediates proteasomemediated degradation of BCL6 in B lymphocytes 45 . In contrast, PELI1 induces K63-linked BCL6 polyubiquitination and promotes BCL6 stabilization 46 , and AIP inhibits BCL6 degradation by regulating the deubiquitinase UCHL1 47 . In addition, small molecule BI-3802 has been investigated to induce the degradation of BCL6 protein through E3 ubiquitin ligase SIAH1 48,49 . All these researches indicate the complex ubiquitin modification of the BCL6 protein. The linkage of E3 ligase-mediated ubiquitination of substrates determines the biological consequence. K11 and K48-linked polyubiquitination targets the substrate to the proteasome for degradation, while other atypical ubiquitin modifications (M1, K6, K27, K29, K33, or K63) may lead to nonprotelytic consequences 50 . In this work, we identified that Trim37 targets Bcl6 predominantly for K27/29 linked ubiquitination in T FH cells, thereby promoting the stability of the Bcl6 protein. However, whether other E3 ubiquitin ligases are involved in the post-transcriptional modification of the Bcl6 protein, especially the E3 ubiquitin ligase responsible for inducing the degradation of Bcl6 in T FH cells, needs further study.
A previous study has reported that specific TRIM37 mutations were associated with a selective impairment in both frequency and proliferative ability of the CD4 + T cell subset, along with a terminally differentiated memory phenotype in both CD4 + T cells and CD8 + T cells 51 . However, Trim37 mutant mice in our study did not exhibit these phenotypes. It is likely that the specific TRIM37 variants present in this case encoded mutated TRIM37 protein, altering its functions in T cells. Another possibility is that our inbred mice are maintained in specific pathogen-free barrier animal facilities, while the diverse human living environment, especially the presence of pathogens, is also important to modulate gene effects 52 .
In summary, we identified Trim37 as the key E3 Ub ligase for Bcl6 in T cells. We have demonstrated that the Trim37-Bcl6 axis is critical for T FH cell differentiation and antibody production. The defective T FH cell responses may actually account for the defect in protective humoral immunity observed with Trim37 mutant mice and patients who harbor the disease-causing TRIM37 variants. Therapeutic restoring the functionality of the TRIM37 mutant may provide a novel avenue to battle against recurrent infection and treat Mulibrey nanism.

Genetic burden test and participants
Clinical exome sequencing data of patients from the Children's Hospital of Fudan University with potential underlying genetic disorders between October 1st, 2019, and September 30th, 2021, were collected. The patients were divided into the RI (recurrent infections) and non-RI groups based on the presence or absence of recurrent infections in the primary diagnosis. The genetic burden test was performed as follows: (1) Each variant was annotated to obtain its mutation type, which was further grouped into four types (protein-truncating variants (PTVs), missense or nonsynonymous variants (MISs), synonymous variants, and noncoding variants) according to the transfer table shown in Supplementary Table S1. (2) For each gene in each sample, the number of variants with each mutation type was calculated and summarized. (3) We applied Fisher's exact test to determine whether the number of variants of each mutation type in each gene was significantly higher in the RI group than in the non-RI group (Supplementary Table S2). In particular, tests of PTV and MIS variants were used to identify RI-related burden genes. Synonymous variants and noncoding variants were treated as the near-neutral background, and genes with significant differences found at the synonymous or noncoding level were filtered out. The threshold of significance was P < 0.05 and OR > 2. The criteria for genetic testing were approved by the ethics committees of Children's Hospital, Fudan University (2015-130). Written informed consent was signed by at least one of the patient's parents.

Mice
Mice were bred and maintained in specific pathogenfree barrier animal facilities. All mice were used according to protocols approved by the Institutional Animal Care and Use Committee of Shanghai Institute of Biochemistry and Cell Biology. C57BL/6N mice were purchased from Shanghai Laboratory Animal Company. Tcrb -/mice were purchased from the Model Animal Research Center of Nanjing University. CD45.1 congenic mice and Cd4-cre transgenic mice have been described previously 53 . OT-II and Bcl6 fl/fl mice were purchased from the Jackson Laboratories. Bcl6 fl/fl mice were bred with Cd4-cre and OT-II mice to generate OT-II Cd4-cre +/-Bcl6 fl/fl mice. Mice at 6-8 weeks were used for cell culture, immunization, and infection analyses. All mice were age-and sexmatched.
Generation of Trim37 mutant mouse models through zygote microinjection FIN major , Trim37 ko , and Trim37 C18R mutant mice were generated by zygote microinjection as described in a previous report 54 . In brief, the mixture of Cas9 mRNA (100 ng/μL), sgRNA (100 ng/μL), and oligo donor (50 ng/ μL) (without donor for the Trim37 ko model) was diluted in RNase-free water, centrifuged at 4°C and 13,200 rpm for 10 min and then injected into the cytoplasm of zygotes harvested from C57BL/6N females (mated with C57BL/ 6N males) using a micromanipulator and a FemtoJet microinjector (Eppendorf). The embryos were cultured in KSOM medium until the two-cell stage and then transplanted into the oviducts of 0.5-day post-coitum (dpc) pseudopregnant ICR females. F0 mosaic mice carrying expected genotypes were selected by Sanger sequencing of PCR products and then backcrossed with WT mice for 3-4 generations to obtain heterozygous mice. All sequence information for sgRNAs and primers is shown in Supplementary Table S3.

Infection with influenza virus and immunization
The WT A/Puerto Rico/8/34 H1N1 (PR8) influenza virus was obtained by cotransformation of 8 plasmids through a reverse genetic system 57 . The virus was expanded using Madin-Darby canine kidney (MDCK) cells, and the titer was measured. The expanded virus fraction was used after loading, freezing, and storage at -80°C. Before use, the viruses were melted on ice and diluted with 1× PBS filtered through a 0.22-μm filter. Eight-week-old mice were infected with the influenza virus at a dose of 0.5 LD 50 (15 PFU) or 10 LD 50 (300 PFU) per 30 μL. PR8 influenza virus vaccine was provided by Dr. Ze Chen. Mice were vaccinated with 10 μg of PR8 vaccine with an intraperitoneal alum adjuvant (Thermo Fisher Scientific) and boosted with the same agents 2 weeks later. Serum samples were collected 2 weeks after the second vaccination. These mice were intranasally challenged with a high dose of PR8 influenza virus (10 LD 50 ) 19 days after the second vaccination.

Flow cytometry
The procedures used in this study were previously described 53

Retroviral transduction
The ORFs of Bcl6 was cloned into the retroviral vector MSCV-IRES-GFP. The Bcl6 5KR mutation plasmids were constructed using the ClonExpress Ultra One Step Cloning Kit (Vazyme, C115-01). Retroviral plasmids containing sequences encoding Bcl6 were produced in Plat-E cells (cultured in DMEM with 10% heat-inactivated FCS and 1% penicillin-streptomycin). Plat-E supernatant containing retroviruses was collected 48 h after transfection. For transduction of retrovirus, purified naive CD4 + T cells were activated for approximately 28 h with anti-CD3 (5 μg/mL) and anti-CD28 (5 μg/mL) in plates precoated with a hamster IgG antibody (20 μg/mL) in complete T medium (described above). Cells were transduced with retrovirus-containing supernatant with polybrene (8 μg/mL) plus hIL2 (100 U/mL) and centrifuged for 1.5 h at 1800 rpm. After 20 h of culture, the transduced CD4 + T cells were washed from the plates and cultured with a complete T-cell medium with hIL2 (100 U/mL) for another 2 d before sorting.

Immunoprecipitation and immunoblot analysis
The ORFs of Trim37 was cloned into the pcDNA3.0 plasmids. The Trim37 mutation and truncation plasmids were constructed using the ClonExpress Ultra One Step Cloning Kit. Immunoprecipitation and immunoblot analysis were performed using standard protocols. Briefly, HEK293T cells were transfected (HighGene Transfection Reagent, ABclonal) with various combinations of pcDNA3.0 plasmids. At 24 h after transfection, the cells were washed with cold PBS twice, and lysates of the cells were prepared in RIPA lysis buffer (50 mM Tris base, 150 mM NaCl, 1% NP-40, 1 mM EDTA, 0.25% sodium deoxycholate, 0.1% SDS) containing 1× Protease Inhibitor Cocktail (Roche). For the in vitro ubiquitination assay, 20 mM MG132 (Sigma-Aldrich) was added to the culture medium 2 h before harvesting the cells. Ten percent of the lysates were mixed with 2× SDS loading buffer referred to as input. Ninety percent of the lysates were incubated with the anti-Flag Protein A/G Plus-Agarose gel (M2, Sigma-Aldrich) overnight at 4°C. The complexes were washed three times with RIPA lysis buffer and diluted with 2× SDS loading buffer. The immunoprecipitated samples were analyzed by immunoblotting. β-actin (I-19, Santa Cruz Biotechnology) was used as an internal control throughout. The antibodies against HRP-conjugated 6× His were obtained from Proteintech. The mouse anti-Bcl6 (K112-91) antibody was obtained from BD Biosciences. The rabbit anti-HA antibody and anti-Flag antibody were obtained from Sigma-Aldrich. The rabbit anti-Trim37 (13037-1-AP) antibody was obtained from Proteintech. The quantitative analysis of western blot data was performed using ImageJ software (NIH).
Ubiquitination assay in the reconstituted E. coli system The procedures were performed as previously described 40 . Briefly, the pACYC-Ub-HA-E1-E2, pACYC-Ub-HA-E1-E2-Trim37 WT , and pACYC-Ub-HA-E1-E2-Trim37 C18R plasmids were cotransformed with the pET22b-Bcl6-His plasmid into competent E. coli BL21 cells. Monoclonal cells were picked after sequencing, cultured in LB medium, and induced with 0.25 mM IPTG at 16°C for 16 h. The cells were pelleted by centrifugation and resuspended in 8 M urea lysis buffer (50 mM Tris-HCl, 50 mM Na 2 HPO 4 , 0.5% NP-40, 300 mM NaCl, 8 M urea, 20 mM imidazole, pH 8.0) for sonication for cell lysis. The supernatant was subjected to incubation with Ni-NTA affinity gel (Qiagen) for 4 h and washed 3 times with 8 M urea lysis buffer. The eluted proteins were analyzed by immunoblot analysis. For twostep enrichment of the proteins, Ni-NTA affinity gels were eluted with RIPA lysis buffer with 1 M imidazole before 9 volumes of RIPA lysis buffer were added. Then, the second eluted proteins were immunoprecipitated with anti-HA affinity gel (Sigma-Aldrich) and washed three times with RIPA lysis buffer. The eluted proteins were subjected to mass spectrum analysis.

Confocal microscopy and histology
The procedures were performed as previously described 58 . Briefly, HEK293T cells were transfected with plasmids encoding HA-tagged Trim37 and Flag-tagged Bcl6 for 24 h. Then, these cells were fixed with 4% PFA in PBS and permeabilized with Triton X-100. After blocking with 10% FBS in PBS, these cells were stained with mouse anti-HA and rabbit anti-Flag antibodies, followed by Alexa Fluor 647 goat anti-rabbit IgG and Alexa Fluor 488 rat anti-mouse IgG. Nuclei were stained with DAPI. The fluorescent images were captured with a Leica TCS SP8 laser confocal microscope. On day 12 after influenza virus infection, lymph nodes and spleens were fixed with 4% PFA and 10% sucrose in PBS for 1 h at 4°C. Then, fixed tissues were incubated overnight in 30% sucrose and embedded in OCT compound (Thermo Fisher Scientific). Cryosectioned tissues were blocked with 10% FBS and 1% Fc blocker (Invitrogen) and then stained with Alexa Fluor 647 anti-B220 (RA3-6B2, BD Biosciences) and Alexa Fluor 488 anti-GL7 (GL7, BioLegend) overnight at 4°C. Mounted sections were imaged on an Olympus FV3000 confocal microscope. Lungs were fixed with 4% PFA in PBS, embedded in paraffin, and stained with hematoxylin and eosin using standard protocols. Mounted sections were imaged on a Zeiss Scan.Z1 system.

Enzyme-linked immunosorbent and viral plaque/ microneutralization assay
The procedures used to measure HA-specific antibodies were previously described 53 . Briefly, 96-well plates (Nunc) were coated with 1 μg/mL HA protein (Influenza A H1N1 (A/Puerto Rico/8/1934) Haemagglutinin, SinoBiological) in Coating Buffer (0.1 M Na 2 CO 3 , 0.1 M NaHCO 3 , pH 9.5) at 4°C overnight. HRP-conjugated goat anti-mouse IgG1 (Southern Biotechnology Associates), and HRPconjugated anti-mouse IgG (R&D Systems) were used at 1:2000 to detect antigen-specific antibodies in serum. The procedures used in the microneutralization assay were previously described 53 . Briefly, MDCK cells were seeded into 96-well plates on day -1. Then, these cells were washed twice with PBS and incubated in DMEM with 2 μg/mL trypsin (T1426, Sigma-Aldrich) at Day 0. Serum samples were serially diluted 2-fold in 50 μL of DMEM and then mixed with 100 TCID50 of PR8 influenza virus in 50 μL of DMEM for 1 h at 37°C. 1 h later, the virusserum mixture was transferred to MDCK cells and incubated for 24 h. After 24 h of incubation, the supernatant was removed, the cells were washed twice with PBS and fixed in 80% acetone for 30 min, and viral antigen was detected by ELISAs with a polyclonal antibody against NP protein. The OD 450 was recorded. Viral plaque assays were performed as previously described 57 .

Quantitative real-time PCR analysis
The procedures used in this study were previously described 53 . Total RNA was prepared from cells using TRIzol reagent (Invitrogen). The purified RNA was quantified, and reverse transcribed by using HiScript III-RT SuperMix for the qPCR kit from Vazyme. The expression of mRNA was normalized to Hprt expression. qPCR primers are shown in Supplementary Table S3.

Statistical analysis
No statistical methods were used to predetermine sample size. GraphPad Prism software was used for all statistical analysis except the genetic burden test. Statistical significance was determined by Fisher test, two-tailed paired or unpaired Student's t-test and ANOVA as described in the figure legends. P values were considered significant when less than 0.05. ns, not significant, *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001. Data are mean ± SEM.