Testing of D451G and K454fs11* ACT1 variants
Neither c.1325A>G or c.1335delA TRAF3IP2 variants were found in any public databases (GnomAd or ExAc). Moreover, c.1325A>G (p.D451G) was classified as pathogenic (Sift 0.01, Polyphen 0.998, CADD 24.8). The c.1335delA variant causes a frame shift resulting in a premature stop codon (p.K454fs11*). Familial segregation showed that the c.1325A>G mutant allele was inherited from the mother and the c.1335delA allele from the father(Figure 2B) . Both mutations are located within the SEFIR domain (Figure 2C) .
We then assessed WT and mutant (D451G or K454fs11*) ACT1 protein expression by overexpression in HEK-293T cells, together with the previously characterized T536I ACT1 variant [12]. We found similar levels of protein expression for all variants, comparable to WT, although the frameshift (K454fs11*) variant resulted in a truncated ACT1 protein (Figure 4A) . We next tested the interaction of WT and mutant ACT1 with IL-17RA by co-immunoprecipitation assay in HEK-293T cells (Figure 4B) . As compared to input fraction(Figure 4B, left) , we observed a reduced interaction of IL-17RA with K454fs11* ACT1, whereas, D451G did not apparently affect the binding of ACT1 to IL-17RA (Figure 4B, right).
We then tested the functional activity of the variants by using an NF-κB reporter assay. ACT1 over-expression is known to constitutively activate NF-κB, a feature that is dependent on the SEFIR domain of ACT1 [10]. We found a strong induction of NF-κB transcriptional activity when we over-expressed wild type ACT1, as compared to the empty vector (EV), which was further increased after IL-17A stimulation (Figure 4C) . In contrast, all three variants showed impaired constitutive NF-κB activity, with barely any further induction after IL-17A stimulation, suggesting that these variants are loss-of-function in this overexpression system.
We then tested the response of the patient’s SV40-fibroblasts upon stimulation with IL-17A, TNF-α, alone or combination, or IL-1β, as compared with fibroblasts from a healthy control, patients with AR IL-17RA or IL-17RC deficiency, or from the previously reported ACT1 deficient patient (Figure 4D) . We observed an induction of GRO-α (also known as CXCL1), after 24 hours of stimulation of control’s fibroblasts upon IL-17A, further increased by TNF-α. In contrast, patient’s fibroblasts, similarly those of patients with AR IL-17RA or IL-17RC deficiency, or ACT1T536I/T536I fibroblasts, showed no GRO-α induction upon IL-17A stimulation, and no further induction as compared to TNF-α alone, when stimulated with a combination of IL17A and TNF-α. All fibroblasts showed strong responses when stimulated with IL-1β.
Finally, proportions of ex vivo IL-17A-, IL-22-, IL-4- producing memory CD4+ cells were raised in the patient, whereas IFN-γ-producing memory CD4+ cells were within the low ranges of the controls (Figure 5).