Genetic variants in UNC93B1 predispose to childhood-onset systemic lupus erythematosus

Rare genetic variants in toll-like receptor 7 (TLR7) are known to cause lupus in humans and mice. UNC93B1 is a transmembrane protein that regulates TLR7 localization into endosomes. In the present study, we identify two new variants in UNC93B1 (T314A, located proximally to the TLR7 transmembrane domain, and V117L) in a cohort of east Asian patients with childhood-onset systemic lupus erythematosus. The V117L variant was associated with increased expression of type I interferons and NF-κB-dependent cytokines in patient plasma and immortalized B cells. THP-1 cells expressing the variant UNC93B1 alleles exhibited exaggerated responses to stimulation of TLR7/-8, but not TLR3 or TLR9, which could be inhibited by targeting the downstream signaling molecules, IRAK1/-4. Heterozygous mice expressing the orthologous Unc93b1V117L variant developed a spontaneous lupus-like disease that was more severe in homozygotes and again hyperresponsive to TLR7 stimulation. Together, this work formally identifies genetic variants in UNC93B1 that can predispose to childhood-onset systemic lupus erythematosus.


Introduction
Systemic lupus erythematosus (SLE) is a chronic autoimmune disease that typically develops in adults, but can affect around 1/100,000 children 1 .There is a significant genetic contribution to the condition, ranging from common variants with small effects, through to fully penetrant diseasecausing alleles 2 .One example of this is TLR7, which marks a genetic interval that is a risk factor to develop SLE 3 , but can also drive a monogenic form of the disease due to gain-of-function mutations 4 .TLR7 typically functions as a sensor of viral ssRNA in endosomes, to which it is trafficked by the transmembrane protein UNC93B1 5 .TLR7 then signals through Myd88 to IRAK1/4 leading to NF-kB and type I IFN expression programs that are typically associated with SLE 6 , but critically required to fight viral infection.Consequently, people with loss-of-function mutations in TLR7 or UNC93B1 are immunodeficient 7,8 .On the other hand, again similar to TLR7, UNC93B1 expression is increased in SLE patients with active disease 9 , and mutations in murine UNC93B1 can cause a lupus-like disease 10,11 .There is also a sporadic lupus-like disease in dogs that is due to a mutation in UNC93B1 12 .Despite all of this compelling data linking UNC93B1 to disease, there was still no genetic variant in the human population that formally validated its role in SLE pathogenesis.

Identification of rare UNC93B1 variants in patients with childhood-onset SLE
Given the recent observation that rare genetic variants in TLR7 can cause childhood-onset SLE 4 , we searched in this patient population for variants which may influence the interaction between TLR7 and the transmembrane (TM) protein which regulates its cellular localization, UNC93B1 13 .
We first identified the novel variant UNC93B1 (c.A940G p.T314A), which is not present in the general population, and is highly conserved (Fig. 1a).Based on a published structure of UNC93B1 in complex with TLR7 13 , the affected amino acid is at the start of helix 2 (H2) between TM helices TM6 and TM7 of UNC93B1 (Fig. 1b).H2 is positioned on top of TM helix of TLR7 responsible for the UNC93B1-TLR7 interaction.Additionally, T314 is in close proximity to the C-terminus of the protein, a disordered region that has been shown to have two phosphorylation sites (S547 and S550) that regulate TLR7 activation 11 .Looking more broadly in a cohort of 272 patients with pediatric lupus recruited at the Guangzhou Women's and Children's Medical Centre, we found that 7 encode UNC93B1 (c.G349T p.V117L) which lies at the interface between UNC93B1 protomers in the reported structure (Fig. 1b).Although these protomers are close to one another, there are no noticeable interactions that would prevent dimerisation of TLR7, similar to what occurs for TLR3 13 .Instead, V117L may act in a similar way to K333R, which is not proximal, but is also located at an UNC93B1 protomer interface and results in increased TLR7 activation.However as K333 is ubiquitinated this could represent a different mechanism of activation 11 .The highly conserved UNC93B1 V117L variant (Fig. 1a) is present in the general East Asian population, and is most prevalent in South Coast Han (Fig. 1c) 14 .We calculate that it confers a 17.9 fold increased risk of developing childhood-onset SLE (Supp Table 1).For the seven SLE patients identified to carry this allele, there were no immediate reports of affected family members, although follow-up studies would be required to confirm this.
Further gene set and pathway analysis pointed towards programs related to innate immune response, phagosome activity, and antigen processing and presentation (Fig. 3e,f and Supp Fig. 3b,c,e,f).Collectively, these results indicate that the UNC93B1 variants identified can intrinsically promote inflammation and immune dysfunction associated with type I IFN and NF-kB signaling pathways.Some of the most attractive therapeutic targets downstream of UNC93B1 and the nucleic acid sensing TLRs (NAS-TLRs) are the signaling adaptor kinases IRAK1 and IRAK4 20 .A dual targeting inhibitor was tested to confirm that this is the pathway triggered by UNC93B1 V117L and T314A, and assess potential clinical utility.Transcriptionally, IRAK1/4 inhibition was efficacious, returning IL-8, IL-12a, TNF-α, and the interferon-related genes, IFN-β, ISG20L2, and ISG-15 close to baseline (Fig. 4a).The pathway was confirmed by western blot of cell lysates, where induced phosphorylation of IRF5, NFκB, and MAPK (JNK1,2,3 and P38) was also downregulated due to IRAK1/4 inhibition (Fig. 4b).In agreement, the downstream cytokines produced by UNC93B1 V117L and T314A Thp-1 cells (eg.IL-8, TNF-α, IFN-β, and IP-10) were significantly blunted (Fig

Lupus-like disease in mice with a mutation orthologous to UNC93B1 V117L
Given that UNC93B1 (V117L) is a highly significant risk factor for childhood-onset SLE, but also present in the general population, we sought to confirm pathogenicity in vivo and created mice with the orthologous mutation V138L.Heterozygous and homozygous UNC93B1 V138L mice were born at normal Mendelian ratios and initially appear healthy, however they lose weight (Fig. 5a), develop splenomegaly with increased spleen cellularity (Fig. 5b) and have a low kidney size (Fig. 5c).In addition, lupus-associated serum anti-dsDNA and anti-smith autoantibodies were increased in knock-in mice (Fig. 5d), but no change was noted for total immunoglobulin G (data not shown).
Functionally, the intracellular production of IFN-γ in the immune cells of bone marrow was also upregulated in the mutant mice (Fig. 5f).In the spleen of UNC93B1 V138L mice there were increased B cells and a decreased CD4 T cell /B cell ratio, along with elevated regulatory T cells, activated T cells, germinal center B cells and CD45 + /CD4 + /CD44 + cells (Fig. 5g).Serum cytokine analysis demonstrated that UNC93B1 V138L leads to upregulation of IL-12p70, IP-10, IL-6 and IFN-γ (Fig. 6a).Histologically, there was overt pathology in kidney, spleen, lung, and pancreas where disease scores of mesangial expansion, extramedullary hematopoietic cell hyperplasia and inflammatory cell infiltrate, respectively, were elevated in mice with the UNC93B1 V138L variant (Fig. 6b).
Bone marrow-derived macrophages (BMDM) isolated from UNC93B1 V138L mice experienced increased mRNA of the interferon-related genes, IRF-7 and IFIT1 when compared to BMDM of UNC93B1 WT mice (Fig. 7a).Meanwhile, phosphorylation of the UNC93B1/TLR7 signaling pathway, including IRF5, NFκB, and MAPK (JNK1,2,3 and P38) was also observed to be activated in UNC93B1 V138L BMDM (Fig. 7b).Additionaly, UNC93B1 V138L BMDM exhibit elevated intrinsic intracellular production of TNF-α (Supp Fig. 4a).Unbiased RNA sequencing indicates that all of the most highly upregulated genes are known to be inducible by type I or type II IFN 21 and includes the central TLR signaling molecule IRF7, as well as IRF3 and TBK1 22 (Supp Fig. 4b, c).
Consistently, this was associated with pathway analysis implicating TLR signaling, TNF signaling, antigen processing and presentation (Supp Fig. 3d-f).Overall, there is a very significant association with genes listed in KEGG disease: Systemic Lupus Erythematosus (Fig. 7c).To investigate the specificity of NAS-TLR signalling in this mouse model, BMDMs were stimulated by TLR ligands and tested for inflammatory markers.As in human samples, data from these experiments showed the involvement of TLR7 in mouse lupus-like phenotype.IFIT1, IRF7, ISG-15, and TNF-α genes were more highly expressed in BMDM of mice with the UNC93B1 V138L variant after stimulation with R848 but not poly I:C, or CPG-C (Fig. 7d).Also, secretion of CXCL1, IP-10, and CCL5 in the supernatant of UNC93B1 V138L BMDM were elevated after R848 stimulation but not after stimulation by poly I:C, or CPG-C (Fig. 7e).These findings implicate enhanced TLR7 signalling as the pathway driving inflammation and a lupus-like phenotype in mice with the equivalent gene variant to humans contributing to childhood-onset SLE.
Therefore, UNC93B1 V138L drives a lupus-like disease in mice at a cellular level, with relevant end organ damage.This is associated with activation of the TLR7 signaling pathway and inflammatory genes expression that is consistent with the clinical presentation of patients with the orthologous genetic change, who suffer from childhood-onset SLE.

Discussion
Our discovery of rare genetic changes in UNC93B1 that predispose to childhood-onset SLE was facilitated by a large body of work culminating in the discovery of lupus-causing variants in TLR7 4 .Interestingly, all patients with UNC93B1 variants identified in this study were female, which could relate to a lack of TLR7 X-chromosome inactivation 23 however could also reflect the strong gender predisposition of this disease in general.We found that the spontaneous inflammatory signaling via IRAK1/4, due to the variants in UNC93B1 identified here, is associated with increased responses to stimulation of TLR7/8.This seems logical based on the current literature, however further work is required for formally delete TLR7 and resolve pathology from the UNC93B1 V138L mouse model we generated.Theoretically, there should also be an endogenous ligand to stimulate TLR7 in this context, which could be guanosine or other nucleic acid 24 .
Furthermore, although the structural location of UNC93B1 T314A presents a logical mechanism to impact TLR7, the molecular effect of UNC93B1 V117L is not yet clear.Mechanistic insight into this process will be extremely useful, and potentially clinically actionable in the future as TLR7 inhibitors are being developed 25 .
Not only are TLR7 inhibitors in clinical trials, but also IRAK1/4 inhibitors could have therapeutic benefit for the patients identified, based on our results.So far, none of the patients characterised in this study have gone on to receive JAK inhibitors or biological therapeutics that would also target type I IFN signaling.Those approaches should be beneficial, and this could be particularly important given that UNC93B1 V117L is present in the general East Asian population, where it could be a significant cause of SLE 26 .Within this population there is a strong geographical bias (Fig. 1c) for which the underlying basis is unknown.Given that the family members of the affected individuals in this study were all apparently healthy, we currently consider that UNC93B1 V117L is not a monogenic disease-causing allele with incomplete penetrance, but rather a very strong risk factor to develop childhood-onset SLE.
Our mouse model represents an important confirmation of the patient findings, and shows that a gene dosage effect of the gain-of-function variant is present.Although no homozygous humans have been identified so far, we can speculate that they may have more severe, or greater likelihood to develop, disease.As UNC93B1 V117L is present in the general population as a rare but highly significant risk factor for disease, the preclinical efficacy of new therapeutic modalities can be accurately modelled for the resulting patient population using our mice avatars.The mouse model should also be useful to determine the cellular contribution of UNC93B1 to lupus-like disease, either intrinsically in B-cells, or with activation of innate immune signaling from antigen presenting cells, and the distinct contribution of type I IFNs compared to other inflammatory programs.We would expect this to be similar to other gain-of-function UNC93B1 mouse models of lupus that are published 27 .
Overall, these findings are not just the first patients with childhood-onset SLE due to variants in UNC93B1, but they bridge the gap from rare monogenic diseases to show that this signaling pathway is relevant for the incidence of disease more generally.Moreover, therapeutics targeting this specific pathway are being developed and can be tested first in a mouse model for which there is a corresponding patient population.n=3 biological replicates (CBA).Indicated p values were determined by unpaired t-test.(e,f) Gene Set Enrichment Analysis (GESA) with WT as control groups and UNC93B1 V117L or UNC93B1 T314A as treatment groups which shows a significant association for genes relating to antigen processing and cross presentation.mRNA was extracted from THP-1 of UNC93B1 WT , UNC93B1 V117L , and UNC93B1 T314A , n=3 biological replicates.   in spleen of UNC93B1 WT/WT , UNC93B1 WT/V138L , and UNC93B1 V138L/V138L mice (n=indicated).Mice were age matched 8 months (for B cells, CD4 T cell /B cell ratio, regulatory T cells, and activated T cells) or 14 months old (for germinal center B cells and CD45 + /CD4 + /CD44 + cells).
4c, d).These findings point towards IRAK1/4 as the dominant signaling pathway triggered by variants in UNC93B1 found in patients with childhood-onset SLE.As IRAK1/4 are adaptors for NAS-TLR signaling, we next investigated which could be involved in this phenotype by stimulating patient PBMCs using Poly I:C, R848, and CPG-C, the ligands of TLR3, TLR7/8, and TLR9 respectively.The results revealed that patient PBMCs stimulated by R848 showed induced genes expression of IFIT1 and TNF-α more than healthy controls, but not when stimulated by Poly I:C and CPG-C (Fig 4e).In addition, secretion of TNF-α, IL-1β, and IL-6 in the supernatant of PBMCs of P2 were increased specifically after stimulation by R848, but not Poly I:C or CPG-C (Fig 4f).These results were independently confirmed using PBMCs of P3 (Fig 4g).Taken together, genetic variation in UNC93B1 regulates inflammation in patients with childhood-onset SLE through a TLR7 -IRAK1/4 pathway.

Fig. 1 Fig. 2 Fig. 3
Fig. 1 Conservation, structural location, and geographical distribution of UNC93B1 SLE variants.(a) Amino acid conservation across 5 species, as indicated, for residues surrounding the variants of interest in UNC93B1.(b) UNC93B1 (PDB:7cyn) shown in light green and green (surface and cartoon representation) with T314 shown in blue sticks on H2.TLR7 protomers are shown in purple and pink (surface and cartoon representation).V117 is shown in red sticks between the interface of two UNC93B1 protomers, L117 is shown in magenta.(c) UNC93B1 V117L variant local geographical distribution.no 1. in wn 7L
in a, c, and d were determined by unpaired t-test.(e) Quantitative RT-PCR analysis of IFIT1 and TNF-α mRNA expression in the PBMCs of P3 compared to healthy controls after stimulation by 10μg HMW Poly I:C, 1μg R848, and 4μg CPG-C for 12hrs.Indicated p values were determined by two-way ANOVA.(f) Production of TNF-α, IL-1β, and IL-6 in the supernatant of PBMCs of P2 compared to healthy controls after stimulation by 10μg/mL HMW Poly I:C, 1μg/mL R848, and 4μg/mL CPG-C for 8hrs.Indicated p values were determined by two-way ANOVA.(g) Production of TNF-α, IL-6, IL-1β and IP-10 in the supernatant of PBMCs of P3 compared to healthy controls after stimulation by 10μg/mL HMW Poly I:C, 1μg/mL R848, and 4μg/mL CPG-C for 12hrs.Indicated p values were determined by two-way ANOVA.

Fig. 5 UNC93B1
Fig. 5 UNC93B1 V138L mice develop lupus-like disease.(a) Appearance and weight of UNC93B1 WT/WT , UNC93B1 WT/V138L , and UNC93B1 V138L/V138L mice.(b) Spleen weight and splenocyte count of indicated mice.Mice were age matched 8-14 months old, the data pooled from three independent experiments (c) Kidney weight of indicated mice.Mice were age matched 8-14 months old, the data pooled from two independent experiments.(d) Serum autoantibodies to dsDNA or Smith for indicated mice (n=15).Mice were age matched 8-14 months old, the data pooled from three independent experiments.(e) Flow cytometric analysis of indicated immune cell populations in bone marrow (BM) of UNC93B1 WT/WT , UNC93B1 WT/V138L , and UNC93B1 V138L/V138L mice (n=indicated).(f) Intracellular cytokine staining of IFN-γ in BM immune cells of indicated mice (n=10).(g) Flow cytometric analysis of indicated immune cell populations

Fig. 6 UNC93B1
Fig.6UNC93B1 V138L mice develop systemic inflammation and organ damage.(a) Serum IL-12p70, IP-10, IL-6, and IFN-γ of indicated mice (n=5 for IL-12p70 and IP-10; n=4 for IL-6, and IFN-γ).For IP-10 mice were 8 months old and for IL-12p70, IL-6, and IFN-γ 10 months old.(b) Haematoxylin and eosin (H&E) staining of the kidneys from indicated mice (n=10), the green arrows indicated mesangial stroma hyperplasia, and the red arrows indicated endothelial and mesangial proliferation.Mice were age matched 8-14 months, the data pooled from three independent experiments.H&E staining of the spleens from indicated mice (n=5), the green arrows indicated the extramedullary hematopoietic cells and the red arrows indicated the inflammatory cells.Mice were age matched 8 months.H&E staining of the lungs from indicated mice (n=3), the yellow arrows (granulocytes) and red arrow (lymphocytes) indicated inflammatory cells infiltration, mice were 14 months old.H&E staining of the pancreases from indicated mice (n=3), the red arrows indicate inflammatory cells infiltration (granulocytes or lymphocytes), mice were 14 months old.The graphs show the pathological disease scores.(c) Quantitative RT-PCR analysis of MX1, ISG20L2, IFIT1 and IRF7 mRNA in kidney tissues from indicated mice (n=3).Mice were age matched 8 months old.Statistical analysis was done using unpaired ttests.The exact p values are shown.

Fig. 7 UNC93B1
Fig. 7 UNC93B1 V138L drives increased inflammation and TLR7 responses in mice.(a) Quantitative RT-PCR analysis of IRF7 and IFIT1 mRNA in bone marrow-derived macrophage (BMDM)from indicated mice (n=3).Mice were age matched 8 months.Statistical analysis was done using unpaired t-tests.The exact p values are shown.(b) Levels of phosphorylated NF-κB, JNK, P38, and IRF5 as measured by immunoblot, in lysates of BMDM of indicated mice (n=3).Mice were age matched 8 months.(c) RNA sequencing was performed for UNC93B1 V138L BMDM compared to control, n=3 biological replicates.Mice were age matched 8 months.Gene Set Enrichment Analysis significantly associated with Systemic Lupus Erythematosus.(d) Quantitative RT-PCR analysis of IFIT1, IRF-7, ISG-15 and TNFα mRNA expression in the BMDM from indicated mice (n=3) after stimulation by 40μg/mL HMW Poly