Blood samples from the patient and her parents were obtained for genetic analysis in accordance with the Declaration of Helsinki. Following informed consent, genomic DNA from the proband was used for whole-exome sequencing. Exome capture was performed by in-solution hybridization using SureSelect Human All Exon V6 bait (Agilent Technologies, Santa Clara, CA, USA). Massively parallel sequencing was performed with the Illumina HiSeq2500 platform with 150-bp paired end-reads (Illumina, San Diego, CA). The reads produced were aligned to the hg19 reference human genome using Burrows-Wheeler Aligner software with default parameters and a –mem option (8). PCR duplicates were removed using MarkDuplicates in Picard tools (https://broadinstitute.github.io/picard/). Candidate variants were called using VarScan2 (http://massgenomics.org/varscan) and were annotated using ANNOVAR (http://annovar.openbioinformatics.org/). Common variants defined by >1% minor allele frequency in ExAC (http://exac.broadinstitute.org/), 1000 genomes (http://www.1000genomes.org/), or ESP6500 (http://evs.gs.washington.edu/EVS/) were excluded from analysis.

C57BL/6J mice was purchased from Japan SLC (Hamamatsu, Japan). All mice were fed a commercial CE-2 diet (CREA Japan, Tokyo) and had ad libitum access to water. The mice were bred in a pathogen-free facility at the Institute for Laboratory Animal Research, Graduate School of Medicine, Nagoya University, and maintained under a controlled temperature of 23 ± 1°C, a humidity of 55 ± 10%, and a light cycle of 12-hour light (from 09:00 to 21:00)/12-hour dark (from 21:00 to 09:00). Animal care and all experimental procedures were approved by the Animal Experiment Committee, Graduate School of Medicine, Nagoya University, and were conducted according to the Regulations on Animal Experiments of Nagoya University.

Targeted disruption of the Jak1 gene on a C57BL/6J background was carried out using the CRISPR/Cas9 method as previously described (9). CRISPR RNA (crRNA, 5’- CAA GAA CAC ATA TCT ATT CT-3’) targeting exon 13 was designed using the CRISPOR website (10). The designed crRNA and trans-activating crRNA (tracrRNA) (Genome CraftType CT, FASMAC, Kanagawa, Japan) and Cas9 protein (New England Biolabs, Tokyo, Japan) were mixed and incubated at 37 °C for 20 min to form a ribonucleoprotein complex (RNP). The ssODN (5’- cct tcc tca gGG TGA GCA CCT TGG CAG AGG CAC AAG AAC AGA TAT ATA CTC TGG GAC CCT GCT GGA CTA CAA GGA TGA GGA AGG AAT TG-3’) was designed to include silent (synonymous) c.1788C>A and c.1791T>C mutations to avoid re-cleavage by Cas9, and the target c.1783C>G mutation was obtained from FASMAC. The final concentrations of components in RNP preparation with ssODN were 8 μM guide RNA (crRNA + tracrRNA), 200 ng/μl Cas9 protein, and 250 ng/μl ssODN. The mixture was electroporated into zygotes using a NEPA 21 electroporator (NEPA GENE Co. Ltd., Chiba, Japan) and the embryos were transferred into the oviductal ampulla of pseudo-pregnant ICR mice.

For sequencing and genotyping, genomic DNA was extracted using KAPA Express Extract (Kapa Biosystems, Woburn, MA) from the pinna and tail of the offspring and were used for PCR amplification. The region targeted by the Cas9 nuclease was amplified by using a GoTaq Green Master mix (Promega, Madison, WI, USA) and a primer pair (5’- CAG GTT TGT GAT AGA CTG CAG CTG-3’ and 5’- CAT CAT TCT CCC CTC ACT ACT CCC-3’). Mutations in the Jak1 gene in offspring were confirmed by Sanger sequencing of the PCR products using Eurofins DNA sequence service (Eurofins Genomics, Tokyo, Japan). Potential off-target cleavage sites predicted by the CRISPOR website ( Supplemental Table 1 , the five regions with the highest Mit off-target scores) were sequenced and no mutations were detected in these sites.

The HEK293 cell line (JCRB9068, Graham, F.L. established) was obtained from the Japanese Collection of Research Bioresources (Osaka, Japan). JAK1 complementary DNA carried on the pFN21A vector (Halotag ORF Clone FHC01306) was purchased from the Kazusa DNA Research Institute (Chiba, Japan), and the JAK1 mutation c.1786C>G;p.H596D was introduced to FHC01306 by Promega Japan (Tokyo, Japan). The HEK293 cells were cultured in DMEM containing 1.8 mM calcium supplemented with 10% fetal bovine serum at 37°C with 5% CO₂. For the transfections, the HEK293 cells were cultured in 12-well dishes and then transfected with wild-type JAK1, mutant JAK1 (H596D), or HaloTag Control Vector plasmids using Screen Fect A plus transfection reagent (FUJIFILM Wako Pure Chemical Corporation, Tokyo, Japan) according to the manufacturer’s protocol. Cells were cultured for 24 or 48 hours after being transfected with the indicated plasmids and were collected for Western blotting analysis.

Immunohistochemical analysis of skin samples from the participants and mice was performed as described previously (11), with slight modifications. Thin sections (3 μm) were cut from samples embedded in paraffin blocks. The sections were soaked for 20 min at room temperature in 0.3% H₂O₂/methanol to block endogenous peroxidase activity. After washing in PBS with 0.01% Triton X-100, the sections were incubated for 30 min in PBS with 4% BSA followed by an overnight incubation with the primary antibodies in PBS containing 1% BSA according to the manufacturer’s instructions. After washing in PBS, the thin sections were stained with the corresponding secondary antibodies for 1 hour at room temperature and washed in PBS. The Vectastain Elite ABC-PO kit (Vector Laboratories, Burlingame, CA) was used for staining. The following polyclonal antibodies were purchased from commercial sources: anti-p-JAK1 [phospho-Tyr1022, anti-p-Jak1 (Tyr1021), Sigma Aldrich, St Louis, MO), anti-p-STAT1 (phospho-Tyr701, anti-p-Stat1 (Tyr701)] (ab30645; Abcam, Cambridge, UK), anti-p-STAT3 (phospho-Tyr705, anti-p-Stat3 (Tyr705) (#11045; Signalway Antibody, College Park, MD). The following monoclonal antibodies were purchased from commercial sources: anti-p-STAT5 [phospho-Tyr694, anti-p-Stat5 (Tyr694)] (ab32364; Abcam), anti-p-STAT6 [phospho-Tyr641, anti-p-Stat6 (Tyr641)] (ab263947; Abcam). n=3. Each experiment was performed twice.

Zirconia balls were added to the proteins extracted from the liver of P0 newborns. Then, the proteins were dissolved in 1 ml sample buffer (NuPAGE LDS sample buffer 250μL, sample reducing agent 100μL, 25×protease inhibiter 40μL, and water 610μL) and crushed. After centrifugation at 10,000 rpm for 10 min at 4°C, the supernatant of each sample was subjected to SDS-PAGE. Strips of membrane were incubated with anti-p-JAK1 (Tyr1021), anti-JAK1 (ab125051; Abcam), anti-p-STAT1 (Tyr701), anti-STAT1 (ab99415; Abcam), anti-p-STAT3 (Tyr705), anti-STAT3 (SAB4300708, Sigma Aldrich), anti-p-STAT5 (Tyr694), anti-STAT5 (ab16276; Abcam), anti-p-STAT6 (Tyr641), anti-STAT6 (ab32520; Abcam) or anti-GAPDH (ab9485; Abcam) antibodies. The antibody–antigen complexes were detected with horseradish peroxidase–conjugated goat anti-rabbit IgG (Dako, Glostrup, Denmark) at a dilution of 1:1,000, followed by detection with enhanced chemiluminescence Western blotting substrate (GE Healthcare BioSciences, Little Chalfont, UK), as described by the manufacturer. n=3. Each experiment was performed three times. For HEK293 cell lysates, additional anti-HaloTag monoclonal antibody (G9211; Promega Corporation, WI) and anti-JAK1 rabbit monoclonal antibody (#3344; Cell Signaling Technology, MA) were used as the primary antibodies.

Total RNA extracted from the brain, liver, and skin of P0 newborns was purified using the RNeasy Mini Kit (QIAGEN, Hiden Germany). Two Jak1 ^{H595D/+;I596I/+;Y597Y/+} mice and five WT mice were analyzed. The quality of RNA was assessed with a 2100 Bioanalyzer (Agilent Technologies). The skin samples (n=7) had an average RNA integrity number (RIN) values of 6.48 (5.8 - 7.1). The blood and liver samples (n=7 each) had an average RIN value of 9.63 (9.2 - 10). RNA sequencing was performed by the Macrogen Japan Corp. using the TruSeq RNA Library Prep Kit v2 (Tokyo, Japan). Next-generation sequencing was performed using the Illumina Novaseq 6000 platform to obtain 101-bp paired-end reads. The reads were adapter-trimmed using our in-house script and were aligned to the mm9 reference genome using TopHat2 (https://ccb.jhu.edu/software/tophat/). Reads Per Kilobase of transcript, per Million mapped reads (RPKM) were calculated using GFOLD (12). Changes in expression levels between groups were estimated using GFOLD, and the resulting GFOLD (0.1) values (conservative estimation of fold changes at the confidence level of q = 0.1) were used for gene set enrichment analyses (GSEA) (13, 14). Volcano plots were generated using log2-fold change values and adjusted p values calculated by DESeq2 (15).

Our proband was a 22-year-old Japanese female who was the older of two siblings born to non-consanguineous parents with no significant family history ( Figure 1A ). She was delivered at 38 weeks of gestation by spontaneous vaginal delivery with a birth weight of 2718g. From birth, she was noted to have dry skin with mild erythema and was diagnosed with atopic dermatitis (AD) at the age of 2 months. At around the same time, she was found to have hepatosplenomegaly. Other clinical manifestations noted included low height and body weight, moderate motor impairment and learning disability, hyperlipidemia, and autism. She was treated with a growth hormone, although it was ineffective. She was provisionally diagnosed with glycogen storage disease type IV, and at 3 years of age, she received a living donor liver transplant due to severe liver failure. At 8 years of age, she developed erythematous cheeks and ichthyotic erythema on her trunk ( Figures 1B, C and S1 ) and extremities, with a SCORing Atopic Dermatitis (SCORAD) score of 70.8%. She was treated with narrow-band UVB therapy, but it led to only minimal improvement. She had eosinophilia and elevated serum thymus and activation-regulated chemokine (TARC; 4,092 pg/mL (normal range, <450 pg/mL)) and IgE with high titers for various antigens ( Supplemental Table 2 ). She died of unknown cause at the age of 22 years.

At 7 months of age, a skin biopsy specimen from the affected skin showed compact hyperkeratosis with normal-appearing granular layers, acanthosis, spongiosis, and lymphocytic infiltration from the upper dermis to the granular layers ( Figure 1D ). Additionally, excised liver tissue demonstrated liver cirrhosis with remodeling of the nodular structure due to fibrosis ( Figure 1E ). Most hepatocytes were brightened or vacuolar due to presumed lipid or glycogen deposition. Although the inflammatory cell infiltrates mainly consisted of lymphocytes (with a minority of plasma cells) and were located principally within fibrotic lesions, they were also observed in hepatocyte nodules and sinusoids, and they caused a mild necrotic inflammatory reaction accompanied by hepatocyte shedding ( Figure 1F ).

Following institutional ethical approval, informed written consent was obtained in compliance with the Declaration of Helsinki guidelines. We started by searching for mutations in ABHD5 (αβ-hydrolase domain-containing 5), the gene implicated in Chanarin–Dorfman syndrome (MIM 275630). Sanger sequencing of ABHD5 revealed no mutations in the genomic DNA of the patient. Moreover, intra-cytoplasmic lipid droplets within venous neutrophils were not observed. Whole-exome capture was then performed (using peripheral blood genomic DNA isolated from the patient and both parents) by in-solution hybridization using SureSelect All Exon 50 Mb Version 5.0 (Agilent, Santa Clara, CA, U.S.A.) followed by massively parallel sequencing (HiSeq2500; Illumina, San Diego, CA, U.S.A.) with 150-bp paired-end reads. After filtering for low frequency variants (frequency of less than 0.1% in the ExAC, 1000 genomes and ESP6500 databases) and against the parental exome dataset, 53 de novo heterozygous variants were identified ( Supplemental Table 3 ). Amongst these variants, a nonsynonymous heterozygous mutation was identified in JAK1 (c.1786C>G; H596D), which was validated by Sanger sequencing ( Figure 1G ). This mutation had not been described in our in-house database of 777 Japanese exomes, nor in the gnomAD Database (16), which includes data for 125,748 whole exomes and 15,708 whole- genomes. In silico analysis with MutationTaster (17) and SIFT (http://sift.jcvi.org/) predicted this variant to be ‘damaging’. Results from other predictive tools included a Combined Annotation Dependent Depletion (CADD) score of 15.27 and Genomic Evolutionary Rate Profiling (GERP) score of 4.09 (highly conserved); thus, the mutation is thought to be functionally relevant. The histidine residue at codon 596 of JAK1 is conserved among three diverse species: P. troglodytes, M. mulatta and M. musculus ( Figure 1H ). The variant allele frequency of H596D mutation in the patient’s peripheral blood mononuclear cells was 0.421053. As such, we cannot determine if this mutation is a germline or somatic mutation. We did not identify potentially pathogenic mutations in other genes implicated in autoinflammatory disease, ichthyosis, or glycogen storage disorders.

We conducted immunohistochemical analyses of phosphorylated-JAK1 (p-JAK1), -STAT1 (p-STAT1), -STAT3 (p-STAT3), -STAT5 (p-STAT5) and -STAT6 (p-STAT6) in lesional skin from the patient. The patient’s epidermis showed strong nucleocytoplasmic JAK1 expression, in contrast to the mostly cytoplasmic staining of normal skin ( Figure 1I ). Epidermal cytoplasmic p-STAT1 expression with focal nucleocytoplasmic localization was also seen in the skin from the patient ( Figures S2A, B ). In addition, p-STAT3, p-STAT5 and p-STAT6 were strongly expressed in the nuclei of keratinocytes in the patient ( Figures S2D, F, H ) compared with predominantly cytoplasmic staining in normal control skin samples ( Figures S2C, E, G ).

To gain deeper insights into the role of JAK1 hyperactivity in the autoinflammatory pathogenesis in vivo, we used a CRISPR-Cas9 gene-targeting approach to generate KI mice. We initially attempted to generate mice harboring only the c.1783C>G (H595D) substitution (Jak1 ^H595D/+), which is identical to that found in the patient. However, we could not design a high-quality guide RNA for CRISPR to target the conserved histidine residue. To avoid re-cutting by Cas9, we introduced two additional synonymous variants at the following amino acids: c.1788C>A (I596I) and c.1791T>C (Y597Y) ( Figure 2G ). Furthermore, we utilized the Jak1KI-mosaic male mouse born by chance because heterozygous Jak1 ^{H595D/+;I596I/+;Y597Y/+} mice (H595D, I596I, Y597Y) could not survive longer than 4 weeks. Mosaic-Jak1 ^{H595D/+;I596I/+;Y597Y/+} male mouse manifested a milder phenotype than Jak1 ^{H595D/+;I596I/+;Y597Y/+} mice. The majority of somatic cells in the mosaic-Jak1 ^{H595D/+;I596I/+;Y597Y/+} mice harbored only wild-type (WT) alleles. The mosaic-Jak1 ^{H595D/+;I596I/+;Y597Y/+} male mouse survived for more than 1 year and were able to produce offspring including heterozygous Jak1 ^{H595D/+;I596I/+;Y597Y/+} mice ( Supplemental Table 4 ).

Jak1 ^{H595D/+;I596I/+;Y597Y/+} mice seemed to have small body sizes and low weights at birth. From a few days after birth, these mice showed hyperkeratosis and scales on the ears, feet, and tail ( Figures 2A, B ). KI mice with only the two introduced synonymous heterozygous mutations (Jak1 ^{I596I/+;Y597Y/+}) did not show any abnormal phenotypes and survived as long as WT mice. Of note, the survival rate was significantly lower for Jak1 ^{H595D/+;I596I/+;Y597Y/+} mice than for WT and Jak1 ^{I596I/+;Y597Y/+} mice ( Figure 2H ). We incidentally generated two types of heterozygous Jak1 mice carrying null alleles caused by frame-shift mutations (c.1792_1793insT and c.1790_1794del5; Jak1 ^+/-). Both types of heterozygous Jak1 ^+/- mice exhibited no cutaneous phenotypes.

As expected, Jak1 ^{H595D/+;I596I/+;Y597Y/+} mice exhibited moderate thickening of the stratum corneum, a reduced number of keratohyalin granules in the uppermost stratum granulosum, and epidermal hyperplasia ( Figures 2C, D and S3B ). Additionally, lymphocytic infiltration was seen in the liver of Jak1 ^{H595D/+;I596I/+;Y597Y/+} mice ( Figures 2F and S3D ), but not in the liver of WT mice ( Figures 2E and S3C ). These findings are consistent with the histological features of the skin and liver samples from our AiKD patient with the JAK1 mutation ( Figures 1E, F ).

Stronger nucleocytoplasmic staining of p-JAK1 was observed in the palmar skin of Jak1 ^{H595D/+;I596I/+;Y597Y/+} mice than in the palmar skin of WT mice ( Figure 2I ). p-STAT3 and p-STAT6 were also detected in the nucleus ( Figures S3E-H ). We also performed Western blot analyses to determine whether the phosphorylation of JAK1 and STAT were increased in the liver tissue of Jak1 ^{H595D/+;I596I/+;Y597Y/+} mice. JAK1, STAT1, STAT3, STAT5, and STAT6 were more highly phosphorylated in the liver cells of Jak1 ^{H595D/+;I596I/+;Y597Y/+} mice compared to WT mice ( Figure S3I ).

To compare global gene expression profiles, RNA sequencing was performed using extracted RNA from the brain, liver, and skin of newborn Jak1 ^{H595D/+;I596I/+;Y597Y/+} mice and WT mice ( Figure 3 and Tables S5-S7 ). We performed a gene set enrichment analysis using the hallmark gene set database. In all three samples (brain, liver, and skin) of Jak1 ^{H595D/+;I596I/+;Y597Y/+} mice, genes associated with the hyperactivation of tyrosine kinases ( Figure 3A and Figure S4A ) and with the activation of NF-κB signaling (downstream of tyrosine kinases) were upregulated ( Figure 3B and Figure S4B ). Brain samples also showed the upregulation of genes associated with the extracellular matrix and the downregulation of genes associated with oxidative phosphorylation, metabolism of lipid-associated genes, and haptotaxis-associated genes ( Figure S4C ). Intriguingly, the genes downregulated in Jak1 ^{H595D/+;I596I/+;Y597Y/+} mice showed a very strong correlation with those downregulated in the brain of the model mice with 22q11.2 deletion syndrome (q<0.001) ( Figures 3C, D and Table S5 ) (18). Liver tissue from the Jak1 ^{H595D/+;I596I/+;Y597Y/+} mice also showed the upregulation of genes associated with inflammation, including IL-6 ( Figures 3E, F and Table S6 ). In terms of STAT family genes, Stat3 was upregulated in differentially expressed genes in the skin ( Table S7 ), and Stat4 was upregulated in differentially expressed genes in the liver ( Table S6 ).

The enrichment of genes associated with interferon-gamma and tumor necrosis factor-alpha was similar among the three types of tissues. The enrichment of genes associated with 22q11 deletion was specific to the brain. The enrichment of genes associated with inflammation was prominent in the liver, while in the other two tissues the enrichment did not reach the statistical significance (q = 0.59 and 0.187 for the brain and the skin, respectively).

Finally, to investigate whether JAK1-STAT pathways were upregulated due to H596D mutation in JAK1, we transfected the JAK1 mutant (JAK1-H596D) into HEK293 cells and performed Western blotting analyses to evaluate the phosphorylation of JAK1 and STAT proteins. As shown in Figure S5 , JAK1 was more highly phosphorylated in the JAK1-H596D-transfected HEK293 cells than in the wild-type JAK1-transfected cells. In addition, the JAK1-H596D-transfected cells showed higher phosphorylation for STAT1 (48h), STAT5 (24h and 48h), and STAT6 (48h) than wild-type JAK1-transfected cells did. We observed no significant difference in the STAT3 phosphorylation levels between wild type and mutant in the present experiments. The results for phosphorylation levels of STAT1 were similar to those obtained in a transfection study of JAK1 A634D in the previous report (6).