Monogenic autoimmunity caused by TLR7 gain-of-function

While circumstantial evidence supports enhanced TLR7 signaling as a mechanism of human systemic autoimmune disease, we have lacked the proof afforded by lupus-causing TLR7 gene variants. Here we describe monogenic human systemic lupus erythematosus (SLE) caused by TLR7 gain-of-function. We identied a de novo, novel, missense TLR7 Y264H variant in a child with severe lupus and additional novel or rare variants in probands with interferonopathies or systemic autoimmunity (Aicardi Goutieres Sd, SLE, Sjogren’s Sd, and juvenile idiopathic arthritis). The variants increased NF-κB and IFN-β activity and the de novo TLR7 Y264H variant was sucient to cause lupus when introduced in mice. We show that constitutive TLR7 signaling drives aberrant survival of BCR-activated B cells that would otherwise die, and accumulation of CD11c + age-associated B cells and germinal center (GC) B cells in a B cell-intrinsic manner. Follicular and extrafollicular helper T-cells were also increased but these phenotypes were cell-extrinsic. MyD88-deciency rescued autoimmunity, aberrant B cell survival, and all cellular and serological phenotypes. Despite prominent spontaneous GC formation in mice carrying the TLR7 Y264H variant, we show that TLR7-driven lupus was not ameliorated when the TLR7 Y264H mice were made GC-decient suggesting extrafollicular origin of pathogenic B cells. We establish the importance of TLR7 for human SLE pathogenesis, which paves the way for therapeutic TLR7 or MyD88 inhibition. 2 (15). RNA viruses such as HIV, inuenza, vesicular stomatitis and more recently SARS-CoV2 suggested to activate TLR7 (16-19). Reports over the last few years have suggested that the endogenous ligand for TLR7 is smRNP (20, 21) which contains the TLR7 stimulatory U1 RNA molecule (22). Solving TLR7’s crystal structure revealed that TLR7 binds to endogenously-produced 2’3’GMP, a cyclic nucleotide derived from GTP degradation (15), at the site (site 1) recognized by the ligands commonly used to activate TLR7 (R837, R848). Binding of R848 to site 1 is enough to trigger signaling, without the need for poly-Uridine binding to site 2 (15). Here, we describe the action of a novel TLR7 gain-of-function (GoF) variant that occurs in a residue that binds 2’3’-GMP, causing constitutive TLR7 activation and monogenic childhood-onset SLE. We report additional novel and ultrarare GoF variants in another ve patients and show how this pathway contributes to autoimmunity. B cells the autoimmune phenotype. B cells. GoF our in a SLE patients (4), suggests that TLR7 is a key upstream driver of human SLE. Combinations of gene variants downstream of this receptor causing increased MyD88 signaling are likely to be important contributors to human SLE. Therapies blocking TLR7 itself and MyD88 may be more effective than therapies blocking germinal centers in SLE patients due to increased TLR7 signaling.


Abstract
While circumstantial evidence supports enhanced TLR7 signaling as a mechanism of human systemic autoimmune disease, we have lacked the proof afforded by lupus-causing TLR7 gene variants. Here we describe monogenic human systemic lupus erythematosus (SLE) caused by TLR7 gain-of-function. We identi ed a de novo, novel, missense TLR7 Y264H variant in a child with severe lupus and additional novel or rare variants in probands with interferonopathies or systemic autoimmunity (Aicardi Goutieres Sd, SLE, Sjogren's Sd, and juvenile idiopathic arthritis). The variants increased NF-κB and IFNβ activity and the de novo TLR7 Y264H variant was su cient to cause lupus when introduced in mice. We show that constitutive TLR7 signaling drives aberrant survival of BCR-activated B cells that would otherwise die, and accumulation of CD11c + age-associated B cells and germinal center (GC) B cells in a B cell-intrinsic manner. Follicular and extrafollicular helper T-cells were also increased but these phenotypes were cell-extrinsic. MyD88-de ciency rescued autoimmunity, aberrant B cell survival, and all cellular and serological phenotypes. Despite prominent spontaneous GC formation in mice carrying the TLR7 Y264H variant, we show that TLR7-driven lupus was not ameliorated when the TLR7 Y264H mice were made GC-de cient suggesting extrafollicular origin of pathogenic B cells. We establish the importance of TLR7 for human SLE pathogenesis, which paves the way for therapeutic TLR7 or MyD88 inhibition.
Main Text SLE is generally considered to be a polygenic autoimmune disease, although in some cases, rare alleles have been identi ed that have provided important insights into disease mechanisms (1,2). The discovery of monogenic lupus cases has highlighted the importance of complement, type I interferons, TLR7 ligation, and B cell survival in disease pathogenesis (3). There is accumulating evidence that human SLE patients display phenotypes consistent with increased TLR7 signaling, with elevated IgD -CD27 -(double negative, DN) B cells, and more speci cally, the CXCR5 -CD11c + subset (also known as DN2 B cells or age-associated B cells (ABCs)) in peripheral blood (4) and excessive accumulation of extra-follicular helper T cells (5). Genome wide association studies (GWAS) have identi ed common polymorphisms in or near TLR7 that segregate with SLE (6)(7)(8). In mice, increased TLR7 signaling due to the Yaa duplication or to transgenic TLR7 expression has been shown to exacerbate autoimmunity to RNA-related self-antigens typical of lupus and Sjogren's syndrome (9,10). In the same line, deletion of TLR7 prevents or ameliorates disease in other lupus models, such as the 564Igi transgenic mice in which B cells encode a receptor that binds to RNA ligands (11). Despite this mounting link between TLR7 and the pathogenesis of lupus, no monogenic lupus cases due to TLR7 variants have been reported to date. There is also con icting evidence as to how TLR7 overexpression causes autoimmunity, particularly, the relative roles of TLR7-driven spontaneous germinal centers versus the role of TLR7-driven DN B cells; the latter have been proposed to originate extrafollicularly and be pathogenic in lupus (4). Most mouse lupus models in which TLR7 plays a role in pathogenicity display increased GC and Tfh formation (9, 10) and it has been proposed that TLR7 drives germinal centers enriched in self-reactive B cells (12). However, recent reports have demonstrated that lupus can develop independently of GCs in mouse models in which disease is dependent on MyD88 signaling (13,14).
TLR7 has been considered to be predominantly a receptor for single stranded RNA and it typically gets activated by guanosine-and uridine-rich RNA, with guanosine binding to site 1 and polyuridine to site 2 (15). RNA viruses such as HIV, in uenza, vesicular stomatitis and more recently SARS-CoV2 have been suggested to activate TLR7 (16)(17)(18)(19). Reports over the last few years have suggested that the endogenous ligand for TLR7 is smRNP (20,21) which contains the TLR7 stimulatory U1 RNA molecule (22). Solving TLR7's crystal structure revealed that TLR7 binds to endogenously-produced 2'3'GMP, a cyclic nucleotide derived from GTP degradation (15), at the site (site 1) recognized by the ligands commonly used to activate TLR7 (R837, R848). Binding of R848 to site 1 is enough to trigger signaling, without the need for poly-Uridine binding to site 2 (15). Here, we describe the action of a novel TLR7 gain-of-function (GoF) variant that occurs in a residue that binds 2'3'-GMP, causing constitutive TLR7 activation and monogenic childhood-onset SLE. We report additional novel and ultrarare GoF variants in another ve patients and show how this pathway contributes to autoimmunity.

Novel and ultrarare TLR7 variants in autoimmune patients
We undertook whole genome sequencing (WES) of a Spanish girl who had presented at age 7 with refractory immune thrombocytopenia and hemichorea. She was found to have renal involvement after admission with a hypertensive crisis (140-150 / 80 mmHg > P99 for her height and sex) that required treatment with enalapril, amlodipine and spironolactone. She was also found to have mild mitral insu ciency and severe thrombocytopenia (lowest of 8.000/µL). She also suffered intermittent episodes of chorea. Bioinformatic analysis revealed a de novo, TLR7 p.Tyr264His (Y264H) novel missense variant ( Fig. 1a-b, Family A and Table S1), predicted to be damaging by SIFT and CADD (Table S2). This variant was not present in the databases of normal human sequence variation (gnomAD, ExAC, dbSNP). Examination of the BAM les ( Fig. 1c) together with paternity analysis (Fig. S1) con rmed the mutation occurred de novo. The mutated tyrosine residue lies in the eighth leucine-rich repeat of TLR7 (23), within the endosomal part of the receptor (Fig. 1d) and is highly conserved across species including zebra sh (Fig. 1e).
WES of additional cases of SLE or related autoimmune diseases and interferonopathies including Sjogren's Sd, Aicardi-Goutières Sd (AGS) and juvenile idiopathic arthritis (JIA) identi ed other novel (C.II.1 R28G), ultrarare (B.II.1 E906K; D.I.1 S724R) and rare (E.II.1, A448V) variants in TLR7 (Fig. 1a, d, e and Tables S1, S2). Of these, the most severe case was that of B.II.1, a North American girl diagnosed clinically with interferon-related encephalitis/AGS spectrum disease at 2.5 years of age. Within 6 months she developed progressive ascending hypertonicity, hyperre exia, and weakness leading to gross motor regression (loss of ability to walk or sit independently) and expressive language regression to a non-verbal state. Fine motor skills were less affected, with some sting and cortical thumbing. Her neurological disease was preceded 1-2 months prior by a month-long febrile illness. Lumbar puncture excluded viral encephalitis and she was found to have an interferon signature with an ISG expression score that was 2.5 to 6.6-fold above the controls (p<0.05) (data not shown). Sequencing identi ed a maternally inherited heterozygous TLR7 missense p.Glu906Lys (E906K) variant, predicted to be damaging by PolyPhen-2 and SIFT. She also carried a paternally-inherited novel variant in MYD88: c.16_34del, p.Ala6Profs*39 (deletion of 19 nucleotides from positions 16 to 34, predicted to result in an A6P substitution and a new reading frame with a stop codon 39 amino acids downstream).
TLR7 variants are gain-of-function Flow cytometric analysis of PBMCs from the probands with the de novo and ultrarare TLR7 variants revealed elevated plasmablasts with variable proportions of memory B cells (Fig. 1f) and elevated CD19 hi IgD -CD27 -CD21 lo B cells also known as double negative (DN) B cells (Fig. 1g); the latter have been associated with increased TLR7 signaling (4). The most signi cant increases in plasmablasts and DN B cells were noted in children A.II.1 with SLE and the TLR7 Y264H de novo variant and B.II.1 with AGS carrying the combination of TLR7 E906K and MYD88 A6Pfs39* variants.
We speculated pathogenic variants would enhance TLR7 signaling and therefore downstream NF-kB and IFN-b activation. To test this, we transfected RAW 264.7 macrophage-like cells with plasmids encoding the different TLR7 mutants together with MYD88. Compared with wild type, overexpression of TLR7 constructs showed enhanced NF-kB and IFN-b activity for most mutant variants suggesting priming of TLR7 for enhanced activation. Interestingly, the de novo Y264H variant did not show increased activity (Fig. 1h), suggesting that it was constitutively active and either refractory to exogenous or over-expressioninduced stimulation or was toxic (24, 25) (see below). Activation of transfected cells with the TLR7 agonists R837 and R848 showed similar results (Fig. S1c). We also expressed the truncated A6Pfs39* MYD88 found in proband B.II.1 and found it enhanced IFN-b signaling while it dampened NF-kB (Fig. 1i).

TLR7 Y264 causes autoimmunity in mice
Given the different luciferase assay results with the Y264H variant, we were intrigued by the TLR7 Y264H allele. The proband carrying this allele had a very severe and early onset SLE and the variant occurred de novo, suggesting a possible monogenic form of disease. Using the published TLR7 structure, we mapped the mutated residue to site 1 of TLR7 ligand binding site, where 2'3'-cGMP binds (15) (Fig. 1j). Indeed, the Y264 side-chain OH forms a hydrogen bond with 2'3'-cGMP (15). Upon binding to site 1, small molecules like R837 induce TLR7 dimerization and formation of the active signaling form (26) (Fig 1k). This suggests TLR7 Y264H substitution gives rise to a constitutive, maximally activated form, likely to be toxic to RAW264.7 cells: high albeit therapeutic doses of TLR7 agonists drive apoptosis of cancer cells (24,25).
To investigate whether the TLR7 Y264H allele could cause monogenic SLE we introduced the orthologous allele into mice using CRISPR/Cas9 editing. WES of affected and unaffected male littermates con rmed the Tlr7 Y264H was the only relevant CRISPR induced coding variant that segregated with the phenotype. The resulting strain was named "kika" by the girl carrying the TLR7 Y264H variant (A.II.1), with Tlr7 Y264H being abbreviated as the "kik" allele henceforth. A strain we generated lacking TLR7 protein due to a 1bp deletion was included as a control ( Fig. S2a-b). Twelve-week-old mice carrying one or both kika alleles displayed splenomegaly (Fig. 2a). Kika mice also had decreased survival (Fig. 2c).

TLR7 Y264H causes cell-intrinsic expansion of effector B cells
Flow cytometric analysis of kika spleens revealed a reduced T:B ratio (Fig. S4b): total B cells were increased by a median of 1.7-fold in both heterozygous and homozygous female kika mice, whereas total T cell numbers did not change (data not shown). Analysis of B cell subsets revealed marked spontaneous GC formation in spleens (Fig 3a) as well as increased ABCs in both spleen (Fig. 3b) and blood (Fig. S4c). Splenic marginal zone B cells were decreased in percentages but not in total numbers (Fig. S4d). Both Tfh cells and CXCR3 + extrafollicular helper CD4 + T (eTh) were elevated in kika mice (Fig. 1c, d). Strikingly, kika mice had circulating Tfh cells in the blood expressing high amounts of PD-1 and CXCR5, comparable to splenic GC-Tfh cells (Fig. S4e). Plasma cells (Fig. 3e) and effector/memory CD4 + CD44 hi T cells (Fig 3f) were also expanded in kika spleens. Plasmacytoid dendritic cells (pDCs) appeared to be activated with increased MHCII and reduced Siglec-H expression (27) (Fig S4f). By contrast, TLR7-de cient mice had opposite phenotypes, lacking spontaneous GC B cell, Tfh cells and ABCs, and showing reduced plasma cells (Fig. S4g). In general, the cellular phenotypes of female mice carrying either one or two kika alleles were largely comparable.
We then set out to establish which cellular phenotypes were cell-autonomous. Mixed bone marrow chimeras were generated by adoptively transferring 50:50 mixes of either WT CD45.1:kika CD45.2 or WT CD45.1:WT CD45.2 bone marrow into sub-lethally irradiated Rag1 -/mice. Examination of spleens 22 weeks post-reconstitution revealed that TLR7 acts B cell-intrinsically to promote increased formation of GCs, ABCs and plasma cells (Fig. 3g). By contrast, accumulation of effector CD4 + T cells, Tregs, Tfh cells and eTh and reduction in MZ B cells were largely extrinsic, with comparable changes in mutant and wild type cells (Fig. 3g and Fig. S4h).
Proband A.II.1 was also heterozygous for another rare variant in a gene associated with SLE: RNASEH2B, p.Ala177Thr, which, when homozygous, causes SLE and AGS (28). RNASHE2B activates cGAS-STING which can increase type 1 IFN production and TLR7 signaling (29,30). We therefore wondered whether a single Rnaseh2b loss-of-function allele might exacerbate the phenotype conferred by TLR7 Y264H. To test this, we generated mice hemizygous for a Rnaseh2b 1 base pair deletion leading to a premature stop codon at amino acid 175 (Fig. S5a). Heterozygous mice were viable and had no immunological phenotypes whereas as previously reported, mice homozygous for Rnaseh2b deletion were embryonically lethal (31) (Fig. S5b). Evaluation of Rnaseh2b +/-Tlr7 kik/+ double heterozygous female mice revealed that Rnaseh2b hemizygosity did not exacerbate the phenotype of Tlr7 kik/+ female mice (Fig. S5c) suggesting that TLR7 GoF alone can cause monogenic SLE. TLR7 GoF allele promotes survival of BCR-activated B cells Next, we investigated the mechanism by which the kik allele alters TLR7 function. First, we asked whether the TLR7 Y264H variant induces constitutive TLR7 activation, which can be assessed by the presence of spontaneous TLR7 cleavage in the absence of stimulation (26). Western blots of splenocytes from TLR7 from kika and wild-type littermates using two different antibodies against both C and N termini revealed the presence of the ~75kda kDa C terminal and 60kda kDa N terminal cleaved TLR7 product in splenocytes from unimmunized kika mice (Fig. 4a). Such cleaved fragment has been reported to be the active form of TLR7 (24).
We then explored the stage at which TLR7 Y264H breaks B cell tolerance. We hypothesized that constitutive TLR7 signaling may provide an aberrant signal 2 to self-reactive B cells that have bound self-antigen through their BCR (signal 1) and would otherwise die within 72 hours, as occurs in anergic B cells (31) and in immature CD93 + B cells stimulated with a-IgM (32). We could not separately activate immature/transitional B cells from spleen, because we found that agonistic TLR7 treatment of mature splenic B cells upregulates CD93 (Fig. S6a). Not surprisingly, CD93 + B cells were abundant in kika mice and it was not possible to determine what fraction of these were truly immature. Hence, we activated total MACS-puri ed splenic B cells from kika, WT littermates and Yaa mice carrying a TLR7 duplication (10,11), stimulated them with R837 or a-IgM and performed live cell counts 72 hours later. We observed that a-IgM, but not R837, enhanced survival of both total and mature kika B cells compared to control cells (Fig. 4b and S6b). This differs from reports using TLR7 transgenic B cells, which only displayed increased survival when activated with a TLR ligand and not with a-IgM (32), again suggesting constitutive activation of TLR7 Y264H. Given that B cell development and proportions of immature B cells in the bone marrow were largely comparable between kika and control littermates ( Fig. S6c) we looked at a-IgM-induced survival in this compartment. Increased survival was also observed in sorted bone marrow immature kika B cells (CD93 + B220 int ) when stimulated with a-IgM (Fig. 4c). These results suggest that constitutive TLR7 signaling allows the survival of B cells that are binding selfantigen through their surface BCR.

Autoimmunity in TLR7 Y264H mice is MyD88-dependent and GC-independent
To con rm that the observed aberrant B cell survival upon IgM stimulation was due to constitutive TLR7 signaling, we crossed kika mice to Myd88 knockout mice to generate kika.Myd88 -/mice. MyD88 de ciency completely rescued each and every kika phenotype including splenomegaly (Fig. 4d), accumulation of ABC (CD23 lo , CXCR5 lo , CD11c + , CD19 hi ), GC B, PC, eTh (Fig. 4e-i and Fig. S7) and autoantibody formation (Fig. 4j) con rming GoF TLR7 signaling via MyD88 causes the autoimmune phenotype. Consistent with constitutive TLR7 signaling being responsible for the aberrant survival of B cells receiving only signal 1 (BCR), the enhanced survival was completely abrogated in a-IgM-activated kika B cells lacking MyD88 (Fig. 4k). Concomitant cell-trace violet dye dilution analysis excluded increased proliferation as the cause of increased B cell numbers in the kika cultures (Fig. 4l).
It remains controversial whether the spontaneous GCs of lupus-prone mice contribute to the autoimmune phenotype. This is particularly the case for TLR7driven autoimmunity, where GCs are abundant. There are opposing views with some suggesting that TLR7 promotes the appearance of self-reactive GC B cells that give rise to the autoantibodies (12) and others proposing that the pathogenic B cells are ABC cells of extrafollicular origin (4). To resolve this question, we crossed kika mice to Bcl6 ox/ ox .CD23 Cre mice that cannot form germinal centers.
In the F2-intercross offspring, we enumerated GC B cells and con rmed that kika Bcl6 ox/ ox .CD23 Cre mice had a profound reduction in GC B cells (Fig. 4m).
Despite the paucity of GC B cells, kika Bcl6 ox/ ox .CD23 Cre mice developed the autoimmune phenotype observed in GC-forming littermate kika mice. This included expansion of ABCs, PC (Fig. 4n), eTh and Tfh (Fig. 4o) and formation of autoantibodies to DNA, RNA and SmRNP (Fig. 4p). These results indicate that TLR7-driven autoimmunity can be GC-independent.
We conclude that TLR7 GoF can cause monogenic SLE as well as contribute to polygenic human autoimmunity. TLR7 Y264 directly interacts with the recently described endogenous ligand 2'3'-cGMP (15). The Y264H substitution may thus create a structural change at site 1 or increase the binding a nity for 2'3'-cGMP, leading to constitutive TLR7 activation -R848 binding at this site (site 1) has been shown to be enough to trigger signaling without the need of poly-Uridine binding to site 2. Analysis of the mouse model revealed that TLR7 plays a crucial role in promoting the survival of BCR-activated immature B cells. This is likely to be an important tolerance checkpoint given the immature B cells are enriched in self-reactivity (33). Activation of self-reactive B cells by selfantigen in the presence of a constitutive signal 2 is likely to break B cell anergy and lead to differentiation and autoantibody production of B cells that would otherwise be destined to die in the absence of T cell help.
Intriguingly, despite our demonstration that enhanced TLR7 signaling causes cell-autonomous accumulation of both ABCs and GC B cells, we show that GC B cells are dispensable for the autoimmune phenotype. It is therefore likely that extrafollicular ABCs are the main source of pathogenic B cells. It will be important to determine if this is true for all forms of human SLE, or it is restricted to patients in which excessive TLR7 signaling is the dominant pathogenic pathway. Although highly damaging TLR7 GoF mutations are rare, our data, together with evidence of increased TLR7 signaling in a large fraction of SLE patients (4), suggests that TLR7 is a key upstream driver of human SLE. Combinations of gene variants downstream of this receptor causing increased MyD88 signaling are likely to be important contributors to human SLE. Therapies blocking TLR7 itself and MyD88 may be more effective than therapies blocking germinal centers in SLE patients due to increased TLR7 signaling.

Mice
Mice were bred and maintained in speci c-pathogen-free conditions at the Australian National University (ANU), Canberra, Australia. Experimentation was performed according to the regulations approved by the local institution ethics committee, including the Australian National University's Animal and human Experimentation Ethics Committee.
Primers for human TLR7 DNA sequencing were used at 10 µM (primer sequences available on request). PCR ampli cation was carried out using Phusion Hot Start II DNA Polymerase II (Thermo Fisher Scienti c) and under conditions recommended by the manufacturer. PCR amplicons were electrophoresed and excised bands puri ed using the QIAquick Gel Extraction Kit (Qiagen). Sanger sequencing was completed using Big Dye Terminator Cycle sequencing kit v3.1 (Applied Biosystems, Carlsbad) using the same primers used for PCR amplication. Sequencing reactions were run on 3730 DNA Analyze (Applied Biosystems) at the ACRF Biomolecular Resource Facility, Australian National University.

Immunohistochemistry
Liver, pancreas and kidneys were xed in 10% Neutral buffer formalin (NBF) solution, embedded in para n and stained by hematoxylin and eosin (H&E).

Bone Marrow Chimera experimentation
For competitive bone marrow chimeras, Rag1 -/mice were irradiated and injected intravenously with equal numbers of bone marrow cells from either WT or Kika CD45.2 and WT CD45.1 mice. Mice were given Bactrim in their drinking water for 48hrs before injection and for 6 weeks following injection, and housed in sterile cages. Following 22 weeks of reconstitution mice were taken down for phenotyping by ow cytometry.

ADVIA blood analysis
Orbital bleeds were performed on mice and blood samples were run on the ADVIA (Siemens Advia 1200).

Statistics
Statistical analysis was carried out using R software version 3.6.1 (The R Foundation for Statistical Computing) and the Emmeans package. Mouse spleen mass data was analysed using two experiments as a blocking factor, followed by a pairwise estimated marginal means comparison of genotypes. Mouse cellular phenotyping, ELISAs, white blood cell and platelet count analysis were performed using a logged linear regression model, followed by a pairwise estimated marginal means comparison of genotypes. Puri ed B cell cultures were analysed using linear regression model followed by a pairwise estimated marginal means comparison of genotypes and stimulatory effect. Luciferase assays statistics were analysed using a one-way ANOVA with Tukey's multiple comparison (PRISM, GraphPad Software LLC). All data was graphed using PRISM.
Plates were washed and goat anti-mouse IgG-AP antibody (Alkaline Phosphatase, Southern Biotech) added for 1 hour at 37°C. Phosphatase substrate (Sigma