key: cord-0714846-lschbtfs
authors: Gruber, Conor N.; Calis, Jorg J.A.; Buta, Sofija; Evrony, Gilad; Martin, Jerome C.; Uhl, Skyler A.; Caron, Rachel; Jarchin, Lauren; Dunkin, David; Phelps, Robert; Webb, Bryn D.; Saland, Jeffrey M.; Merad, Miriam; Orange, Jordan S.; Mace, Emily M.; Rosenberg, Brad R.; Gelb, Bruce D.; Bogunovic, Dusan
title: Complex Autoinflammatory Syndrome Unveils Fundamental Principles of JAK1 Kinase Transcriptional and Biochemical Function
date: 2020-08-03
journal: Immunity
DOI: 10.1016/j.immuni.2020.07.006
sha: 12297a7de079df11d3233414acd4560e22c721f7
doc_id: 714846
cord_uid: lschbtfs

Autoinflammatory disease can result from monogenic errors of immunity. We describe a patient with early-onset multi-organ immune dysregulation resulting from a mosaic, gain-of-function mutation (S703I) in JAK1, encoding a kinase essential for signaling downstream of >25 cytokines. By custom single-cell RNA sequencing, we examine mosaicism with single-cell resolution. We find that JAK1 transcription was predominantly restricted to a single allele across different cells, introducing the concept of a mutational “transcriptotype” that differs from the genotype. Functionally, the mutation increases JAK1 activity and transactivates partnering JAKs, independent of its catalytic domain. S703I JAK1 is not only hypermorphic for cytokine signaling but also neomorphic, as it enables signaling cascades not canonically mediated by JAK1. Given these results, the patient was treated with tofacitinib, a JAK inhibitor, leading to the rapid resolution of clinical disease. These findings offer a platform for personalized medicine with the concurrent discovery of fundamental biological principles.

Monogenic errors of the Janus kinase (JAK) family, essential signal transduction hubs of the immune system, have dire consequences in immune function. Gruber et al. describe a JAK1 gain-offunction mutation with mosaicism and monoallelic expression that underlies a multi-system autoinflammatory disease, which is rescued by JAK inhibitor therapy.

Monogenic disease mutations afford the opportunity to study the bona fide function of human genes in vivo, which have guided our understanding of biology and medicine for decades. Undiagnosed disease programs, by means of next-generation sequencing, have recently provided a platform to identify, diagnose, and study these rare patients with unusual clinical presentations (Lee et al., 2014; Splinter et al., 2018; Yang et al., 2014) . In turn, clinical management can, in some cases, be highly personalized.

To date, studies of rare immunologic diseases have identified germline gain-of-function (GoF) and loss-of-function (LoF) mutations throughout the Janus kinase-signal transducer and activator of transcription (JAK-STAT) signaling axis (Dupuis et al., 2001; Etheridge et al., 2014; Flanagan et al., 2014; Hambleton et al., 2013; Holland et al., 2007; Kofoed et al., 2003; Macchi et al., 1995; Mead et al., 2012; Minegishi et al., 2006 Minegishi et al., , 2007 Russell et al., 1995; van de Veerdonk et al., 2011) , the primary signal transduction pathway for cytokines. The Janus kinase (JAK) family contains four tyrosine kinases (JAK1, JAK2, JAK3, TYK2) constitutively associated with cytokine receptors. Upon cytokine binding, JAKs act in partnership to phosphorylate themselves, the receptors, and then STATs, which can then act directly as transcription factors or activate other signaling pathways further downstream (O'Shea et al., 2015) . JAK1 is activated by a broad range of cytokines (gc, gp130, interferon [IFN] , and interleukin-10 [IL-10] family cytokines). It can phosphorylate any signal transducer and activator of transcription (STAT) protein (STAT1-6) and is universally expressed in all tissues (O'Shea et al., 2015) . Through the formation of specific combinations of cytokine receptors, JAK partners, and STAT dimers, JAK1 orchestrates unique downstream signals for each cytokine.

The need to better understand JAK regulation has deepened with the expanding clinical use of JAK inhibitors (O'Shea and Gadina, 2019) . The breadth of successfully treated inflammatory conditions signifies the central pathophysiological role of JAK hyperactivity across immune diseases. However, the complete list of disorders resulting from JAK-STAT dysregulation remains unknown. Furthermore, it is unclear which specific JAK-mediated pathways drive disease, a key issue for the design of inhibitors with greater selectivity for individual members of the JAK family.

Herein, we identify a mutation (S703I) of JAK1 in a patient with a severe, early-onset immunodysregulatory syndrome identified in our undiagnosed disease program. Using extensive next-generation genomic, molecular, and multi-parametric immunological tools, we probe the effects of S703I JAK1 in vitro and ex vivo to investigate clinical dysfunction in vivo.

We studied an 18-year-old female who was referred to our undiagnosed disease program with a complex primary autoinflammatory and atopic syndrome ( Figure 1A ). The patient was born to a non-consanguineous family with no history of immunologic disease. At birth, the patient was noted to have a widespread pustular rash in a linear pattern that predominantly affected the left side of the body ( Figure 1B) . The lesions continually progressed, and a biopsy later revealed inflammatory linear verrucous epidermal nevus ( Figure S1A ). At~1 year of age, she developed recurrent emesis and diarrhea. Repeat endoscopic biopsies demonstrated chronic, unspecified inflammation at various sites (most frequently colonic, but also gastric, duodenal, ileal, and esophageal regions). Eosinophilic infiltration of the colon was consistently noted (Figures 1C and S1B) . Likewise, peripheral eosinophilia with a fluctuating pattern was frequently observed ( Figure S1C ). At 3 years of age, she developed rapid weight gain, edema, and proteinuria. Renal biopsy demonstrated membranous nephropathy (MN) ( Figures 1D and S1D) , which was refractory to treatment with corticosteroids, and later, cyclosporine and tacrolimus. According to her history, the nephrotic syndrome was ameliorated via an elemental diet, but this was not able to be consistently maintained. Of note, serology was negative for known MN autoantibodies or other autoreactive antibodies ( Figure S1E ). Kidney transplantation was performed at age 11, but MN recurred within 1 year, followed by antibody-mediated rejection, requiring subsequent hemodialysis. Over this time, she also experienced asthma, food and environmental allergies, severely stunted growth with leg length discrepancy, and poor weight gain (for an extended clinical report , see STAR Methods).

Mutation Given the overall healthy state of the parents and the early onset of disease in the patient, we hypothesized that either a recessive or de novo genetic mutation was the cause of the clinical syndrome ( Figure 1E ). We performed whole-exome sequencing on peripheral blood cells obtained from the patient and her parents. Subsequent variant analysis failed to produce any likely variants by a recessive model of inheritance (Table S1 ). Because of the asymmetric manifestations of disease, including limb length discrepancy and irregularly distributed dermatitis, we then considered the possibility of lower-read-frequency de novo mosaic mutations, which are typically excluded from common analysis pipelines. One candidate de novo variant, JAK1 c.2108G > T, which constituted 27% of the reads mapping to the region, was identified ( Figures 1E and 1F) . The presence of the c.2108G > T variant was confirmed by Sanger sequencing ( Figure 1G ), and this variant was absent from all of the publicly available genome sequences from healthy individuals. This mutation results in the substitution of serine to isoleucine at position 703 (S703I) in a highly conserved region ( Figure S1F ) and is predicted to be highly damaging (combined annotation-dependent depletion [CADD] score of 27.6). We then investigated the presence of c.2108G > T in non-hematopoietic tissues. We performed digital droplet PCR (ddPCR) with mutation-specific probes to estimate the fraction of cells carrying the mutation in different tissues. We identified the mutation at various frequencies in DNA from buccal swabs, granulocytes, peripheral blood mononuclear cells (PBMCs), and endoscopic biopsy samples fractionated into epithelia and associated immune cells (Figures 1H and S1G) . These tissues represent all three germ layers, signifying that the mutation must have arisen in the first~12 cell divisions between fertilization and gastrulation ( Figure 1I ) (Moore et al., 2015) .

The S703I mutation localizes to the pseudokinase domain of JAK1, a putative regulatory domain ( Figure 2A ). Although S703I is located between the germline JAK1 mutations identified to date, these other mutations diverged in their downstream consequences (LoF and GoF), making functional predictions for S703I difficult (Del Bel et al., 2017; Eletto et al., 2016) . To assess the possible pathogenicity of the mutation and its impact on JAK1 function, we transduced WT JAK1, S703I JAK1, and empty vector lentiviruses into U4C cells, a fibrosarcoma cell line previously selected to lack endogenous JAK1 (Pellegrini et al., 1989) . Transduction with S703I JAK1, but not WT JAK1 or Luciferase, led to basal phosphorylation of STAT proteins and active target gene transcription in the absence of cytokine stimulation ( Figures  2B-2E ). S703I-transduced cells hyperresponded to IFN-a, in ll Article terms of both the proximal phosphorylation of STAT1 and STAT2 and the induction of IFN-stimulated genes (ISGs) (Figures 2B and 2C) . Similarly, these cells hyperphosphorylated downstream STATs in response to 2E, and S2A) .

For direct confirmation of the pathogenicity of the mutation in cells from the patient, we derived an EBV-immortalized B cell (B-EBV) line from the patient's PBMCs. Given the mosaicism for JAK1 in the patient's cells, individual lines were cloned from single cells to derive purely wild-type (WT) or S703I heterozygous mutant cells ( Figure S2B ). A comparison of STAT phosphorylation in patient WT and mutant B-EBV cells supported the GoF role of S703I JAK1, both at baseline and in response to cytokines (Figures 2F and 2G) . The isogenic control derived from the same patient pinpointed S703I JAK1 as the probable pathogenic mutation in the patient's genome. These results indicate that S703I is GoF for basal-and cytokine-induced STAT signaling.

To dissect the mechanisms underlying upregulated STAT signaling, we assessed the impact of S703I on JAK autophosphorylation. Consistent with the increase in STAT phosphorylation, S703I JAK1 was itself hyperphosphorylated (Figure 3A ). In addition, JAK2, TYK2, and JAK3 phosphorylation (H) Proportion of cells carrying the heterozygous JAK1 mutation, as estimated by digital droplet PCR with WT-and mutation-specific probes. DNA was obtained from bilateral cheek swabs, Ficoll-fractionated whole blood, and epithelial tissue isolated from a colonic biopsy (n = 1). (I) Model for the development of the de novo mutation and its distribution into all 3 germ layers. See also Figure S1 .

were upregulated (Figures 3B-3D and S2C), suggesting that the interacting JAK partners may also play a role in the GoF. We reasoned that partnering JAK activity could be overactivated by JAK1 S703I from two mechanisms: increased formation of the receptor complex or direct crosstalk between JAK proteins. We assessed whether JAK1 S703I allowed for more cytokine receptor at the surface, as the structural domains of JAK1 scaffold the receptor complex (Li et al., 2013) . However, surface staining for the type I IFN receptor subunit (IFNAR2) demonstrated equivalent receptor expression in WT JAK1 and S703I JAK1 cells, indicating that JAK1 scaffolding of the receptor complex was unaffected by the patient's mutation ( Figures  3E and S2D ). Next, we hypothesized that the mutant JAK1 pseudokinase domain transactivated the kinase activity of JAK2 and TYK2. To investigate this mechanism, we mutated the ATP-binding site (K908A) of JAK1 to render it catalytically inactive. This well-characterized mutation retains the signaling capability of the receptor complex, making it possible to study signaling by the partnering JAKs in isolation (Eletto et al., 2016; Li et al., 2013) . As expected, STAT phosphorylation was largely reduced in the absence of JAK1 activity ( Figures 3F and 3G ). However, following inactivation of the kinase domain of S703I JAK1 (S703I/K908A), an aberrant increase in STAT phosphory-lation relative to kinase-inactivated JAK1 without the S703I mutation was observed upon cytokine stimulation (Figures 3F and 3G) . This result indicates that S703I JAK1 transactivated JAK2 and TYK2, revealing that pseudokinase domains can regulate partnering JAKs in trans, in addition to traditionally understood cis-regulation (Babon et al., 2014) . This mechanism was conserved with the other reported JAK1 GoF mutation (A634D, Del Bel e al 2017) ( Figure 3H ), but not to the JAK2 GoF mutation (V617F) that is common in hematologic malignancy ( Figure S2E ).

Finally, to clinically substantiate the importance of partnering JAK activity in JAK1 S703I (and other JAK-mediated diseases), we compared the potency of selective and non-selective JAK inhibitors, filgotinib and tofacitinib, respectively (Changelian et al., 2003; Rompaey et al., 2013) . Unlike a pan-JAK inhibitor, a JAK1-specific inhibitor was unable to completely abrogate JAK1 S703I signaling owing to the continued signaling of partnering JAKs transactivated by the JAK1 pseudokinase ( Figures  3I and S2F , schematic). These findings underscore the biochemical and clinical importance of trans-regulation and indicates that JAK selectivity need not mean clinical efficacy. This notion not only challenges the current dogma in drug development but it also proves that precision medicine must be used on a mutation-specific basis. ll Article functional immaturity . This analysis revealed that JAK1 GoF NK cells aligned phenotypically with prototypical immature ''CD56 bright '' NK cells, unlike those from STAT1 GoF patients ( Figure S3I ). However, whether this phenotype results from the primary pathology or clinical intervention remains to be determined in this case. We then performed phospho-CyTOF to analyze the phosphorylation of all STATs downstream of JAK1 in all of the major immune cells of whole blood. Given the ubiquitous expression and diverse signaling capabilities of JAK1, as well as the basal activity of S703I observed in vitro, we hypothesized that all STATs within all immune cells would be hyperphosphorylated at baseline. Heightened phosphorylation of STAT1, STAT3, STAT4, STAT5, and STAT6 was observed, but not universally. Certain immune subsets, but not others, exhibited high basal phosphorylation of specific STATs ( Figure 4C ). For example, granulocytes from the patient displayed baseline STAT1 phosphorylation, but not STAT3 phosphorylation, whereas T cells displayed basal STAT3 phosphorylation but not STAT1 phosphorylation. STAT6 phosphorylation, however, was not upregulated in any immune subset, except B cells.

To evaluate the functional consequences of the observed basal phosphorylation, we assessed the expression of downstream genes from bulk PBMCs. We detected the elevated expression of downstream pSTAT target genes, including IFI27, IFIT1, MX1, and SIGLEC1 ( Figure S4A ). In addition, we tested non-hematopoietic tissues for baseline STAT phosphorylation by immunohistochemistry. In skin and gastrointestinal biopsies, but not renal biopsy, we detected highly phosphorylated STAT1 and STAT3 as compared to healthy samples ( Figure S4B ). This basal phosphorylation was observed in the apparent absence of any overt increase in circulating JAK-STAT cytokine concentrations ( Figure S4C ), suggesting that intrinsic S703I JAK1 activity drove this process, consistent with our in vitro results ( Figures 2B-2F ).

Whole blood was then stimulated ex vivo with a series of cytokines that engage JAK1 with various cytokine receptors, JAK partners, and downstream STAT targets. In response to IFN-a or IL-2, patient leukocytes hyperphosphorylated STAT1 and STAT3 or STAT5, respectively (Figures 4D and S4D and S4E, represented plots). By contrast, STAT6 phosphorylation in response to IL-4 was similar to that in healthy control cells, as seen in the baseline STAT6 data ( Figure 4D ). Likewise, IL-5 stimulation led to normal STAT5 phosphorylation ( Figure S5F ). These differential responses were also noted in patient B-EBV cells ( Figures S5G-S5I ). However, stimulation with either IL-2 or IL-4 induced the phosphorylation of STAT1 exclusively in patient cells, contrasting the canonical signaling cascade induced by these cytokines ( Figure 4D ). This non-canonical response suggests that S703I confers promiscuity onto JAK1, allowing it to transverse traditional signaling axes and establish non-canonical pathways. These results indicate that S703I JAK1 is a GoF mutation ex vivo, displaying both unexpected pathway promiscuity and specificity in the activation of STAT signaling.

To follow up on the differences across immune cell types, we performed single-cell RNA sequencing (scRNA-seq) of the patient's PBMCs. We aimed to specifically map and evaluate the impact of S703I JAK1 across immune cell subsets by relating single-cell resolution gene expression patterns to cell type and, given the mosaicism, JAK1 genotype. However, sequence data from droplet-based scRNA-seq platforms are typically restricted to 5 0 or 3 0 transcript regions and therefore do not include coverage of the S703I site (c.2108) in JAK1 mRNA (Figure S5A) . Therefore, using custom barcode microbeads and a modified library preparation procedure, we adapted the inDrops scRNA-seq methodology (Klein et al., 2015; Zilionis et al., 2017) to target exon 16 (containing the S703I site) of JAK1 in addition to the standard mRNA 3 0 regions ( Figure S5B ; Method Details). We used this custom inDrops scRNA-seq to analyze the patient's PBMCs, and concurrently processed the same sample on the 10X Genomics Chromium platform ( Figures S5C and S5D ). Clustering and manual annotation of cell types based on gene expression patterns distinguished the expected PBMC populations, with relative frequencies consistent with the corresponding CyTOF analyses (Figures 5A and S5D; Table S3 ). Again, we observed a relative increase in CD56 hi NK cell frequency. Using sequence data from JAK1-targeted libraries, we assigned putative per-cell JAK1 genotypes (based on transcript sequences) to a subset of PBMCs for which transcript and read depths were sufficient in the JAK1-targeted inDrops dataset (Figures S5E and S5F) . Within this subset, we found that the mutant allele was not evenly distributed across immune cell-type clusters ( Figures 5B and 5C ). For example, B cells and monocytes mostly carried WT transcrip, whereas 69% of CD56 hi NK cells contained JAK1 mRNA with the S703I mutation. This variable distribution may reflect differences in the intrinsic tolerance to the mutation in these cell types, which is consistent with the expansion of the CD56 hi NK cell subset and the higher expression of JAK1 in NK cells than in other cell types ( Figure S5G ). We next performed differential gene expression analysis, comparing patient cells expressing the WT and mutant alleles of JAK1. Although this analysis was constrained by a limited number of cells for which transcript-level genotype information was available, the intrasample comparison can be conducted on an inherently isogenic background with identical exposure history. We detected statistically significant differences in ISG expression for IFI44L and for an ISG set between WT and mutant monocytes ( Figures 5D and 5E ). These data suggest that tolerance to S703I may differ between cell types and confirm the cell-intrinsic nature of the GoF described above.

Given the mosaicism of a heterozygous mutation, we expected to observe some cells (i.e., homozygous WT JAK1) containing only WT JAK1 transcripts and others (i.e., heterozygous S703I JAK1) containing both WT and S703I JAK1 transcripts. However, we found that expression of the 2 alleles seemed almost mutually exclusive, as very few cells expressed both transcripts, as opposed to the~50% that was expected given our genomic estimates ( Figures 1D and 5F , left panel). By contrast, other genes ll Article from our dataset containing heterozygous variants exhibited the expected biallelic distribution ( Figure 5F , right panel). This result suggests that JAK1 may be subject to monoallelic bias, a pattern that has only recently been recognized in a fraction of the autosomal transcriptome (Borel et al., 2015; Deng et al., 2014; Gimelbrant et al., 2007; Jeffries et al., 2012) . We further tested this hypothesis of biased allele expression by sorting single cells from healthy donor PBMCs heterozygous for a synonymous SNP (rs2230587) of JAK1, adjacent to S703I. qPCR of isolated RNA with allele-specific probes revealed that relative expression of the two alleles was not normally distributed, but rather biased to one allele or the other ( Figure 5G ), unlike in a control gene (Figure 5H) . A query of publicly available murine expression data revealed a similar allele restriction for JAK1, which, at least in mice, remains fixed over time (Savova et al., 2016) . Whether the observed phenomenon in this patient represents transcriptional bursting or mitotically stable monoallelic expression and the potential impact on immune dysfunction remains to be fully determined. In either sense, these data perhaps indicate a departure from the classic genetic interpretation of heterozygosity and allow for a shift in understanding of the genetic penetrance of disease.

Having identified JAK1 hyperactivity as the putative driver of clinical disease in the patient, we then considered the use of JAK inhibitors for the clinical treatment of this patient. We compared the ability of the two US Food and Drug Administration (FDA)approved pan-JAK inhibitors available at the time to reduce basal STAT phosphorylation in S703I-transduced U4C cells. Despite its lower relative potency against JAK1, tofacitinib inhibited STAT phosphorylation in a comparable dose response to ruxolitinib ( Figure 6A ), again reflecting the independence of JAK1 catalysis and STAT hyperphosphorylation. Similar results were obtained with patient-derived B-EBV cells, which, unlike the transduced fibrosarcoma U4C cells, are hematopoietic in origin and therefore express JAK3 ( Figure 6B ). Next, we treated patient blood ex vivo with the two compounds at equimolar doses that mimic physiological dosing (Chen et al., 2014; Krishnaswami et al., 2014; Lamba et al., 2016) , and we further assessed the inhibition of IFN-a stimulation. Analysis of phospho-STAT inhibition across whole-blood immune subsets by phospho-CyTOF revealed that tofacitinib attenuated the response more potently than ruxolitinib in nearly all cell types ( Figure 6C ). Following these extensive functional studies, we treated the patient with low-dose tofacitinib (5 mg daily). Within 8 weeks, circulating inflammatory markers (erythrocyte sedimentation rate [ESR] and C-reactive protein) normalized ( Figures 6D and  6E ). Near-complete improvement in dermatitis followed, both grossly and histologically ( Figure 6F ). By 6 months, the patient reported complete resolution of gastrointestinal symptoms (decrease in modified Pediatric Ulcerative Colitis Activity Index [PUCAI] from 35-50 initially to 0). Biopsy of colonic tissue revealed the restoration of crypt architecture and the complete resolution of eosinophilic infiltrates ( Figure 6G ). The patient remained stable for 2 years after initiation of therapy until, unfortunately, the patient succumbed to acute respiratory failure due to coronavirus disease 2019 (COVID-19) .

Finally, we confirmed the pharmacological rescue of JAK hyperactivity in the patient's cells after tofacitinib treatment. RNA isolated from PBMCs revealed that the expression of ISGs, which was elevated before treatment, progressively declined to normal amounts ( Figure 6H ). CyTOF analysis was then performed to confirm the decrease in basal STAT phosphorylation ( Figure S6 ). Reductions were observed across cell types in STAT3, STAT4, STAT5, and STAT6 phosphorylation, whereas STAT1 phosphorylation was reduced in some, but not all cell types ( Figure 6I ). Overall, these results validate S703I JAK1 as the etiology in vivo of the widespread immune dysregulation (summarized in Table S4 ), and they illustrate the power of precision medicine both as a treatment approach for patients with rare diseases and as a means of discovering fundamental physiological mechanisms.

Undiagnosed disease programs have shown promise for the detection of potential causative variants of disease (Lee et al., 2014; Splinter et al., 2018; Yang et al., 2014) . This report demonstrates the value of the in-depth study of select patients identified in these programs. Most immediately, the findings described herein directed the successful treatment of a complex immunodysregulatory disease in a personalized molecular fashion. More broadly, this case implicates JAK1 dysfunction in common forms of multifactorial diseases, including dermatitis, enteritis, colitis, and eosinophilic disorders. These features align with the other reported JAK1 GoF mutation recently published (Del Bel et al., 2017) , and together provide strong justification for the expanding use of JAK inhibitors in these disorders (O'Shea and Gadina, 2019; O'Shea et al., 2015) . However, to date, MN has not been recognized to involve JAK-STAT dysregulation. Identifying more patients with JAK1 GoF mutations will be critical to determine whether the link between JAK1 GoF and MN is causal. If true, then MN, the most common cause of nephrotic syndrome, may be amenable to early treatment with JAK inhibitors. In fact, baricitinib, a JAK1 and JAK2 inhibitor, demonstrated recent success in a clinical trial for diabetic nephropathy, a distinct but related nephrotic syndrome (Tuttle et al., 2018) .

The absence of MN in the other reported JAK1 GoF mutation (A634D) may represent important distinctions in the behavior of different mutated forms of JAK1, rather than variable penetrance of the same genetic etiology. Disruptions of the pseudokinase domain may have vastly divergent functional consequences, which can already be gleaned by comparing the JAK1 mutations identified to date: P733L and P832S (LoF) versus A634D and S703I (GoF). Moreover, our findings indicate that GoF may not necessarily lead to the universal activation of downstream pathways, as S703I caused the hyperactivation of some pathways, but not others. Furthermore, disruption of the pseudokinase domain by S703I enabled cells to respond promiscuously via non-canonical signaling pathways. These findings suggest that the pseudokinase domain is not a simple ''on/off'' switch. This regulatory complexity may stem from the ability of the JAK1 pseudokinase to modulate the activity of JAK2 and TYK2, as demonstrated here. Consequently, careful study of each mutation is warranted, each yielding valuable information on fundamental JAK1 function. The complex regulation becomes ll Article Immunity 53, 1-13, September 15, 2020 9 especially important, given the high incidence of oncogenic JAK mutations, as well as the expanding therapeutic use of JAK inhibitors. Regarding the latter, the evidence presented here and elsewhere (Eletto et al., 2016; Haan et al., 2011; Li et al., 2013) of the highly cooperative action of JAKs challenges the strategic wisdom of increasing the selectivity of JAK inhibitors.

Following transactivation of partnering JAKs, S703I JAK1 constitutively upregulated STAT1, STAT2, STAT3, STAT4, and STAT5 phosphorylation and their downstream target genes, all in the apparent absence of circulating cytokine. These STATs mediate signaling for >25 cytokine pathways, and thus the clinical disease that results is likely driven by a complex mixture of ll Article these pathways. Most predominantly, the pathophysiology in this patient appeared to be a combined (1) autoinflammatory and (2) atopic disease process. The presence of severe, earlyonset unspecified inflammation throughout the gastrointestinal tract and skin with drastically high acute phase reactants is evocative of primary autoinflammatory disorders. In this regard, the JAK1 GoF here represents a partial phenocopy of other systemic autoinflammatory diseases that overactivate the IFN and IL-6 axes. These include STAT3 GoF mutations (Fabre et al., 2019) and the type I interferonopathies, disorders of overactive STAT1 and STAT2 activity (Rodero and Crow, 2016 )-pathways that were highly upregulated by JAK1 S703I. Other disease features suggested an atopic syndrome. Eosinophils were elevated in peripheral circulation and infiltrating into the gastrointestinal lamina propria and crypt epithelium. Asthma, food and environmental allergies, and a skin disease similar to atopic dermatitis were severe and early in onset. Lastly, her renal disease (membranous nephropathy) could only be ameliorated by an elemental diet of strictly liquid nutrients, suggesting a dietary hypersensitivity trigger. Considering the high basal phosphorylation of STAT5 in this patient, these features seem to phenocopy GoF STAT5b (Ma et al., 2017) mutations, which result in earlyonset eosinophilia, urticaria, dermatitis, and diarrhea. While it is possible the pathological inflammatory stage of COVID-19 pathogenesis was exaggerated in this patient, other risk factors must be considered, including the comorbidities (e.g., asthma, renal failure) and the combined immunosuppression (i.e., tofacitinib, tacrolimus, and oral glucocorticoids) in this patient. In particular, there is an urgent need to understand how JAK inhibition modulates both the protective and pathological inflammatory immune responses to SARS-CoV-2, as JAK inhibitors are in clinical trials for COVID-19. Finally, this study advances our understanding of mosaicism-at both a genomic and transcriptomic level-in personalized medicine. Genomically, this JAK1 mutation was uncovered, mapped across tissues to trace its embryological origin, and directly linked to phenotypic consequences by deriving JAK1-WT and JAK1-S703I cell lines from the patient. These findings underscore how asymmetric clinical manifestations like those observed here (leg-length discrepancy and dermatitis along lines of embryological migration) should prompt suspicions of mosaicism and guide genetic analysis with carefully chosen techniques. In particular, the analyses described here have only recently become possible with technological advances in single-cell assays. Here, by adapting the inDrops scRNA-seq platform, we implement an approach for the detection and analysis of a specific mutation within a transcript region not readily accessible by standard methods. The resulting single-cell resolution data allowed us to determine mutant allele frequency in different cell populations, identify mutation-associated gene expression patterns, and, unexpectedly, to observe allelic bias in the transcription of JAK1. This bias is consistent with recent evidence of widespread transcriptional bursting and monoallelic expression of autosomal genes (Borel et al., 2015; Gimelbrant et al., 2007; Jeffries et al., 2012; Reinius and Sandberg, 2015) . This report demonstrates monoallelic expression of a mutated gene. Biased allelic expression, in conjunction with mosaicism, may prove an important point of focus for future genetic studies of variable penetrance, affected carriers, and undiagnosed disease.

In conclusion, intense basic mechanistic investigations of a single mutation identified pan-JAK inhibition, as opposed to highly selective JAK1 inhibition, as the optimal therapy for personalized medicine, leading to biological and clinical rescue. This approach constitutes a workflow for alike monogenic syndromes.

Detailed methods are provided in the online version of this paper and include the following: 

The authors declare no competing interests.

Kofoed, E.M., Hwa, V., Little, B., Woods, K.A., Buckway, C.K., Tsubaki, J., Pratt, K.L., Bezrodnik, L., Jasper, H., Tepper, A., et al. (2003) . 

Further information and requests for reagents may be directed to the Lead Contact, Dusan Bogunovic (dusan.bogunovic@ mssm.edu).

All unique/stable reagents generated in this study are available from the Lead Contact with a completed Materials Transfer Agreement.

The raw data for experiments performed, including the PCR, multiplex ELISA, mass cytometry, flow cytometry, scRNaseq data, are available upon request from the lead contact. For privacy concerns of the study participants, raw scRNA-Seq data are not available, and the complete data files from whole exome sequencing will be restricted to the variants in Table S1 . scRNA-Seq gene x cell matrices, python scripts and associated Seurat code (R-based) used for the tailored analysis of JAK1-specific scRNA-Seq are available at https://github.com/jorgcalis/JAK1-alleles-pipeline.

This study reports an 18-year-old female who presented during early childhood with persistent, recurrent cutaneous and gastrointestinal inflammatory disease with eosinophilic infiltration and peripheral eosinophilia. Notable physical features include a leg length discrepancy, short stature and low body weight. In parallel, she developed refractory membranous nephropathy leading to end stage renal disease. A kidney transplant was complicated by disease recurrence in the graft, progressing over several years, as well as acute rejection ultimately rendering the patient dialysis-dependent. Specifically, her past medical history by organ involvement is further detailed, as follows: Renal: At 3 years of age, she developed rapid weight gain, edema and proteinuria. Renal biopsy at age 7 demonstrated membranous nephropathy, which was refractory to treatment with corticosteroids, and later cyclosporine and tacrolimus. By history the nephrotic syndrome was ameliorated by use of an elemental diet, but this was not able to be consistently maintained. Serological testing for anti-PLA-2R receptor, anti-thrombospondin and anti-bovine serum albumin was all negative. A gradual decline in renal function was observed and at age 11 a living-donor kidney transplantation was performed. One year later, nephrotic range proteinuria recurred, and biopsy confirmed relapsing membranous nephropathy. The graft function further declined, and an episode of acute antibody-mediated rejection resulted in transplant failure at age 16. She has required long-term hemodialysis since that time, delivered via an AV fistula, and is currently being evaluated for a second transplant.

Dermatologic: At birth, a pustular rash involving face and extremities was noted. The rash persisted and worsened after discharge. Skin biopsy at 3 months suggested Inflammatory Linear Verrucous Nevus. Skin involvement spread and worsened in intensity with age, manifesting as a diffuse, erythematous rash involving the face, trunk and extremities, with prominence on the left side. Biopsies later demonstrated a subacute or chronic spongiotic dermatitis. The epidermis was acanthotic and showed varying degrees of intercellular edema (spongiosis) with widening of the intercellular spaces; the stratum corneum was thickened and focally compact; the dermis contained a perivascular lymphohistiocytic infiltrate which extended around the superficial and deep vascular plexus. Of note, the biopsies did not show the changes commonly associated with epidermal nevi: alternating ortho and parakeratosis, epidermolytic hyperkeratosis, or acantholytic dyskeratosis. Rather, these clinical and histologic changes in the skin likely represent a form of blaschkitis, an inflammatory skin condition, presenting as papules or vesicles, occurring along the lines of Blaschko (which represent somatically distinct bands of ectodermal migration).

Gastrointestinal: In infancy, she experienced recurrent emesis and diarrhea unresponsive to formula changes. Bloody stools were noted at 10 months of age and watery diarrhea and abdominal pain became persistent. Repeat endoscopic biopsies demonstrated chronic, unspecified inflammation at various sites (most frequently colonic, but also gastric, duodenal, ileal and esophageal regions). Eosinophilic infiltration of the colon was consistently noted. Symptoms were only marginally responsive to treatment with corticosteroids, chronic antibiotics and a severely restricted diet.

Growth disturbances: Growth impairment was reflected by short stature (Z score < À3) and low body weight (Z score À2 to À8), currently 138 cm and 31 kg. Nutritional etiologies were addressed by placement of a G-tube at age 10 with some improvement in growth. Growth hormone was administered for 5 years with moderate benefit. Leg length discrepancy was identified at birth, with left extremity smaller than right in girth and length.

Immunologic: Allergic reactions were observed to enalapril (anaphylaxis), milk (rash), soy (rash) and wheat (rash). She also experienced occasional episodes of dyspnea along with lip and leg angioedema, without an identifiable inciting allergen. Asthma was diagnosed and managed with bronchodilators. Acute phase reactants were noted to be consistently elevated, including ESR and C-reactive protein. Complement (C3 and C4) levels were within reference range. Likewise, quantitative immunoglobulin testing for IgG, IgA, IgE and IgM was within normal limits. Seroconversion after immunization was observed for all vaccinations except varicella virus, hepatitis A virus and hepatitis B virus.

Family history was largely unremarkable, with no family history of consanguinity or gastrointestinal, renal, immunologic or dermatologic disease. Mother, father and older brother are alive and well.

Response to tofacitinib: After 8 weeks of tofacitinib there was complete normalization of acute phase reactants, ESR and C-reactive protein upon laboratory assessment. The patient was previously unable to tolerate dairy, soy and gluten due to severe abdominal pain and diarrhea. After initiation of tofacitinib, she liberalized her diet without restriction and remained asymptomatic, with formed stools and without abdominal pain.

At baseline the endoscopic findings included altered vascularity and friable mucosa from rectum to descending colon with microscopic patchy, active colitis with eosinophilic infiltration noted in the ascending colon. After 6 months of treatment on tofacitinib 5mg daily, the colon was grossly normal, with microscopic active colitis and complete resolution of eosinophilia. The dose was further increased to 7.5 mg daily thereafter. The dermatitis significantly improved, as seen in the images.

Tofacitinib is 30% renally-excreted and the dose administered was limited due to chronic kidney disease. Presumably, after retransplantation the dose may be escalated with a potentially greater effect. To date, the drug has been very well tolerated with no evidence of adverse effects with close monitoring.

Coronavirus Disease 2019 (COVID-19): Two years after initiation of tofacitinib treatment, the patient developed an acute respiratory infection that was later diagnosed as COVID-19. She was admitted to the hospital where she was treated with azithromycin and hydroxychloroquine and subsequently intubated. Her immunosuppressive treatment regiment, including tofacitinib, was held constant over this time. Unfortunately, despite intensive supportive care, the patient expired 7 days after admission.

Cell lines U4C cells (JAK1 À/À ) and g2A cells (JAK2 À/À ) were obtained from S. Pellegrini and cultured in DMEM supplemented with 10% fetal bovine serum (FBS) (Invitrogen), GlutaMAX (350 ng/ml; GIBCO), and penicillin/streptomycin (GIBCO). All cell lines expressing ectopic JAK1 and JAK2 variants were generated by lentiviral transduction and were subsequently selected with puromycin and FACS for matched RFP expression. EBV-transformed lymphoblastoid cell lines (EBV-B cells) were generated by infecting PBMCs from healthy controls or the patient with EBV supernatants. EBV-B cells were cultured in RPMI with 10% FBS, 1% glutamine and 1% penicillin/ streptomycin. Single-cell clones were isolated by limiting dilution analysis on OP9 feeder cells and expanded in conditioned media. After genotyping by Sanger sequencing, WT/WT and WT/S703I clones were selected.

All samples were collected with informed consent in accordance with IRB-approved protocols (Study ID# IF2349568). For wholeexome sequencing, DNA was isolated (QIAGEN Cat No 69504) from Ficoll-isolated granulocytes from whole blood of the proband and her healthy parents. Library preparation, sequencing (150 bp paired-end reads) and alignment for whole-exome sequencing were performed with the Genewiz exome-sequencing package. Potential disease-causing variants were investigated by Ingenuity Variant Analysis (QIAGEN). Hight-quality variants were identified by filtering as follows: exclusion of common variants (> 0.1% allele frequencies in public databases); retention of variants in coding regions resulting in substitutions, premature stops, frameshifts or altered splicing; exclusion of variants with CADD scores below MSC thresholds. For recessive models of inheritance, only homozygous or compound heterozygous mutations were assessed. For de novo inheritance, all variants in the parents were excluded, and alleles with known haploinsufficiency, hemizygous, or dominant-negative effects were included. For the validation of WES results, JAK1 was then amplified from PBMC DNA by PCR and Sanger sequenced.

For the determination of mosaic allele fractions, DNA was isolated from bilateral buccal swabs, fractionated blood and a gastrointestinal endoscopic biopsy specimen in which the epithelial layer was isolated by chemical dissociation. Digital droplet PCR was performed with JAK1 amplification primers, mutation-specific probes (IDT), and with ddPCR Supermix (Biorad 1863026). Amplification and quantification were performed on a QX100 Droplet Digital PCR system (BioRad). Cellular genotypes were estimated with QX100 software (BioRad), assuming heterozygosity.

We analyzed the presence of circulating autoantibodies using IgG autoantibody arrays (RayBiotech PAH-AIDG-G1-16). In addition to samples from the patient in this study, plasma from five healthy donors, two patients with systemic lupus erythematosus and Immune dysregulation, polyendocrinopathy, enteropathy, X-linked (IPEX) were included. All samples were run in duplicate. Before processing, samples were was first clarified and diluted at a ratio of 1:200. Genepix Pro 7.0 software was used to analyze the images.

The JAK1 plasmid was obtained from Addgene (Plasmid #23932) in a Gateway-compatible backbone, pDONR 223. Site-directed mutagenesis with specific primers (Quikchange II, Agilent 200521) was performed to obtain the WT coding sequence in the same plasmid, and plasmids encoding the S703I, K908A or S703I/K908A forms were then generated. A control vector containing the luciferase gene was also cloned. Plasmids were then subcloned into a lentivirus-compatible pTRIP-X-IRES-RFP backbone with ll Article puromycin resistance. Pseudotyped lentiviral particles were produced by the transfection of HEK293T cells with pCAGGS-VSV-G, pCMV-Gag/Pol and genes of interest.

For the analysis of STAT phosphorylation, cells were stimulated with the indicated doses of recombinant IFNa-2b (Intron-A, Merck), IFNg, IL-2, IL-4, and IL-6, IL-13 (Biolegend) for 15 minutes and then lysed for western blotting. For the analysis of gene induction, cells were stimulated for 8 hours, after which cells were lysed for RNA isolation. For JAK inhibitor treatment, cells were incubated with ruxolitinib (Selleckchem S1378), filgotinib (SelleckchemS7605) or tofacitinib (Selleckchem S5001) at the indicated doses for 4 hours.

Cells were lysed in RIPA buffer (Thermo Fisher 89900) supplemented with protease/phosphatase inhibitor cocktail (Cell Signaling #5872). Lysates were sonicated, centrifuged to remove insoluble complexes, then boiled with NuPage sample buffer (Thermo Fisher NP0007) containing 20 mM DTT. The samples were subjected to gel electrophoresis and semi-dry transfer, and the resulting immunoblots were blocked in 5% BSA, then incubated overnight with primary antibody followed by HRP-conjugated secondary antibodies. Primary antibodies against the following targets were used: GAPDH (Cell Signaling D16H11), STAT1 (Santa Cruz C-111), phospho-STAT1 (Cell Signaling 58D6), phospho-STAT2 (Cell Signaling D3P2P), phospho-STAT3 (Cell Signaling D3A7), phospho-STAT5 (Cell Signaling C11C5), phospho-STAT6 (Cell Signaling 9361), JAK1 (Santa Cruz B3), JAK2 (Cell Signaling D2E12), TYK2 (Cell Signaling D4I5T), phospho-JAK1 (Cell Signaling 3331), phospho-JAK2 (Cell Signaling 3771), and phospho-TYK2 (Cell Signaling 9321). For the analysis of JAK phosphorylation, lysates were first incubated overnight with antibodies against total JAK protein conjugated to Protein G Dynabeads (Thermo Fisher 10007D). Immunoblotting was then performed as above.

Cell lines or isolated PBMCs were lysed and RNA was isolated with RNeasy spin columns (QIAGEN 74104). Reverse transcription was performed with the High-Capacity RT Kit (Applied Biosystems 4368814). The resulting cDNA was then subjected to qPCR with the TaqMan Master Mix II with UNG (Thermo Fisher 4440038), on a Roche LightCycler 480, with the following primers/probes: 18S (4318839), MX1 (hs00895608), RSAD2 (hs00369813), SIGLEC1 (hs00988063) IFIT1 (hs01911452). The relative expression of each transcript was normalized relative to 18S by the DDCt method.

Whole blood was collected, by venipuncture, into sodium heparin vacutainer tubes. For immunophenotyping, blood was immediately stained and processed for mass cytometry. For intracellular staining, whole blood was stimulated by incubation for 15 minutes with the cytokines indicated and was then immediately stabilized with Proteomic Stabilizer PROT1 (SmartTube) and frozen at À80 C. Similarly, for inhibition with ruxolitinib and tofacitinib, blood was treated for 4 hours with 500 nM inhibitor and was then stimulated by incubation for 15 minutes with 1000 IU/mL IFNa, before stabilization and freezing.

Frozen samples were thawed according to the manufacturer's recommended protocol. The thawed samples were washed with barcode permeabilization buffer (Fluidigm) and barcoded with Fluidigm's Cell-ID 20-Plex Pd Barcoding Kit. Samples were then washed and pooled into a single tube, Fc-blocked (FcX, Biolegend) and heparin-blocked to prevent non-specific binding. Cells were then stained with a cocktail of markers to identify major immune populations. All antibodies in the panel were either conjugated in-house with X8 MaxPar conjugation kits (Fluidigm) or purchased from Fluidigm. The antibody cocktail was filtered through an Amicon filter with 0.1 mm pores before staining.

After surface staining, the samples were permeabilized with methanol and stored for at least 12 hours in methanol at À80 C. Samples were then washed, heparin-blocked and stained with a cocktail of phosphorylation and signaling antibodies. The stained samples were washed and incubated in freshly diluted 2.4% formaldehyde containing 125nM Ir Intercalator (Fluidigm), 0.02% saponin and 30 nM OsO4 (ACROS Organics) for 30 minutes at room temperature. Samples were then washed and acquired immediately after staining.

Samples were washed once with PBS+0.2% BSA, once in PBS, and once in CAS buffer (Fluidigm). They were and resuspended at a concentration of 1 million cells per mL in CAS buffer containing a 1/20 dilution of EQ beads (Fluidigm). Following routine instrument tuning and optimization, samples were run at an acquisition rate of < 300 events per second on a Helios mass cytometer (Fluidigm) with a modified wide-bore injector (Fluidigm). FCS files were then normalized and concatenated with Fluidigm acquisition software, and the barcoded samples were deconvoluted with a MATLAB-based debarcoding application (''Palladium-based mass tag cell barcoding with a doublet-filtering scheme and single-cell deconvolution algorithm''). The FCS files were then uploaded to Cytobank for analysis. Cell events were identified as Ir191/193-positive and Ce140-negative events. Doublets were excluded on the basis of Mahalanobis distance and barcode separation and with the Gaussian parameters acquired with Helios CyTOF software. Downstream data analysis was performed on Cytobank, by both tSNE analysis and biaxial gating of immune populations, as shown in Figure S3 . Mean signal intensities were calculated, and relative induction was determined by normalization relative to the mean for healthy control samples. ll Article Immunity 53, 1-13.e1-e11, September 15, 2020 e8

The molecular regulation of Janus kinase (JAK) activation

JAK1 gain-of-function causes an autosomal dominant immune dysregulatory and hypereosinophilic syndrome

Biased allelic expression in human primary fibroblast single cells

Integrating single-cell transcriptomic data across different conditions, technologies, and species

Pharmacokinetics and pharmacodynamics of orally administered ruxolitinib (INCB018424 phosphate) in renal and hepatic impairment patients

Single-Cell RNA-Seq Reveals Dynamic, Random Monoallelic Gene Expression in Mammalian Cells

Impairment of Mycobacterial But Not Viral Immunity by a Germline Human STAT1 Mutation

Biallelic JAK1 mutations in immunodeficient patient with mycobacterial infection

A novel activating, germline JAK2 mutation, JAK2R564Q, causes familial essential thrombocytosis

Clinical Aspects of STAT3 Gain-of-Function Germline Mutations: A Systematic Review

Activating germline mutations in STAT3 cause early-onset multi-organ autoimmune disease

Widespread monoallelic expression on human autosomes

Jak1 has a dominant role over Jak3 in signal transduction through gc-containing cytokine receptors

STAT2 deficiency and susceptibility to viral illness in humans

STAT3 mutations in the hyper-IgE syndrome

Stochastic choice of allelic expression in human neural stem cells

Droplet barcoding for single-cell transcriptomics applied to embryonic stem cells

Dominant-negative mutations in the DNA-binding domain of STAT3 cause hyper-IgE syndrome

Before We Are Born: Essentials of Embryology and Birth Defects

Selective Janus kinase inhibitors come of age

The JAK-STAT pathway: impact on human disease and therapeutic intervention

CD56bright natural killer (NK) cells: an important NK cell subset

Random monoallelic expression of autosomal genes: stochastic transcription and allele-level regulation

edgeR : a Bioconductor package for differential expression analysis of digital gene expression data

Type I interferon-mediated monogenic autoinflammation: the type I interferonopathies

Preclinical Characterization of GLPG0634, a Selective Inhibitor of JAK1, for the Treatment of Inflammatory Diseases

Genetic variation at IFNL4 influences extrahepatic interferon-stimulated gene expression in chronic HCV patients

Mutation of Jak3 in a Patient with SCID : Essential Role of Jak3 in Lymphoid Development

dbMAE: the database of autosomal monoallelic expression

Bias, robustness and scalability in single-cell differential expression analysis

Effect of Genetic Diagnosis on Patients with Previously Undiagnosed Disease

JAK1/JAK2 inhibition by baricitinib in diabetic kidney disease: results from a Phase 2 randomized controlled clinical trial

Ruxolitinib partially reverses functional NK cell deficiency in patients with STAT1 gain-of-function mutations

STAT1 mutations in autosomal dominant chronic mucocutaneous candidiasis

Camera: a competitive gene set test accounting for inter-gene correlation

Molecular findings among patients referred for clinical whole-exome sequencing

Flow cytometry

and three unrelated healthy donors were thawed and allowed to rest briefly in complete RPMI medium supplemented with 10% FCS. Cells were immunostained with antibodies in 2% FBS in PBS for 45 minutes. For the panel assessing effector function and activation, cells were stimulated with phorbol 12-myristate 13-acetate (10 ng/ml, Sigma-Aldrich) and ionomycin (1 mg/ml, Sigma-Aldrich) for 4 hours at 37 C in the presence of brefeldin A (10 mg/ml, Sigma-Aldrich) and anti-CD107a antibody. For the panels evaluating effector function and transcription factors, cells were then permeabilized with BD Cytofix/Cytoperm (BD Biosciences) or FoxP3 buffer (Tonbo), and were stained by incubation with antibody for 45-60 minutes. Activated cells were stained for surface markers for 20-25 minutes after the four hours of incubation. Data were acquired on a FACSAria machine (BD Biosciences) with the capacity to detect 18 fluorescent parameters and exported to FlowJo 10.5.3 (BD Biosciences) for analysis. The frequency of cells positive for each parameter was compared with the mean and standard deviation for three healthy donors analyzed in parallel with the samples from the patient. IFNAR2 staining was performed on transduced U4C cells surface stained with anti-IFNAR2 antibody (PBL 21385-1) at a 1:100 dilution for 1 hour on ice. Secondary antibody staining was performed with an anti-murine IgG antibody with an Alexa Fluor 647 tag (Thermo Fischer A-21235)

Circulating cytokine levels were determined in magnetic Luminex assays with the Bio-Plex Pro Human Inflammation (BioRad 171al001m) and custom Human Cytokine Panel (R&D LXSAHM), according to the manufacturer's protocol. Samples were quantified on a MAGPIX xMAP Instrument (Luminex)

Slides were incubated with primary antibodies for 60 minutes at 37 C. The following primary antibodies were used: anti-phospho-STAT1 (Tyr701) 58D6 (Cell Signaling) and anti-phospho-STAT3 (Tyr 705) D3A7 XP (Cell Signaling) at 1:100 dilution in blocking agent. The slides were then incubated with Omni Map anti-rabbit HRP-conjugated secondary antibody (Multimer HRP) (Cat # 760-4311) (Roche) for 32 minutes

After downstream high throughput sequencing, reads can be assigned to individual cells by barcode sequences, thereby enabling expression quantification at single cell resolution. As the cellular barcode is introduced downstream of transcript polyA tails, sequence data is typically restricted to transcript regions immediately proximal to 3 0 termini. Because the S703I site is located at position 2402 from 5 0 transcript start and 2690 from the 3 0 transcript polyA tail, it is not accessible by standard droplet microfluidics single cell RNA-Seq platforms. Therefore, we adapted the inDrops method by generating custom barcode microbeads containing JAK1-specific primers (flanking the S703I site) in addition to polyT primer sequences, enabling more efficient JAK1 target capture and access to the S703I site. Following within-droplet reverse transcription, second strand synthesis and in vitro transcription amplification, samples are split into two parallel library preparations: one for standard polyT-primed libraries, and one for JAK1-targeted libraries. During data processing of resulting high throughput sequencing data

2019) with the following modifications. For the second round of split-and-pool primer extension for barcode synthesis, hydrogel beads were added to microplate wells (n = 384) containing oligonucleotide templates for both standard polyT primers and for an additional primer complementary to an S703I-adjacent region of the JAK1 transcript (11.53uM polyT oligonucleotide template, 2.3uM JAK1 oligonucleotide template) Within a given well, both polyT and JAK1 oligonucleotide templates carried identical ''barcode 2'' sequences, ensuring that extended primers on hydrogels contained matching barcodes

20 ul) were then split to two parallel library preparations: 10 ul of IVT product was prepped for polyT-primed gene expression libraries according to the standard inDrops protocol, and 10 ul of IVT product (typically reserved as a ''backup'' aliquot) was prepped according to a JAK1-targeted protocol as follows. IVT products were reverse transcribed by SuperScript III with a JAK1-specific primer (with 5 0 extension containing Illumina adaptor sequence) flanking the S703I site (55C for 1 hr, 70C for 15 min). RT reactions were treated with RNase H (37C for 30 min, 65C for 20 min) to remove RNA template from cDNA heteroduplexes

GCTCGGAGATGTGTA TAAGAGACAG[bc2,8nt]NNNNNNTTTTTTTTTTTTTTTTTTTV-3 0 Final ''on bead

We next used per cell JAK1 WT and S703I transcript counts to assign putative genotypes to individual cells. As cell free RNA in suspension can co-encapsulate with cells thereby generating unwanted background signal in droplet microfluidics scRNA-Seq methods, we aimed to apply a stringent, data driven threshold for genotyping assignment. To evaluate the potential influence of cell free RNA, we quantified the frequency of JAK1 transcripts detected in empty (i.e., cell free) droplets (defined as barcodes with 1000 -2000 reads in polyT-primed libraries) using the same barcode/UMI strategy described above. We found that more than 99% of empty droplets had less than 3 JAK1 transcripts. Guided by these data, cells were only considered for JAK1 genotype assignment if at least 5 JAK1 transcripts were detected. For JAK1 genotype assignment, individual cells were classified in one of the following four categories: cells that carried the S703I allele were classified as ''MUT+'' regardless of carrying the wild-type allele; the remainder (S703I allele negative) cells that carried the wild-type allele were classified as ''WT+MUTneg'' if no S703I transcripts were detected, and as ''WT+MUTnonZero'' if 1-to-4 S703I transcripts were detected; cells that carried neither the wild-type nor the S703I allele were classified as ''WTnegMUTneg

2018) was performed with CAM-ERA (Wu and Smyth, 2012) and the linear model described above. heterozygous SNP in JAK1 (rs2230587) were isolated by Ficoll gradient. Cells were stained with antibodies against CD3, CD19, CD14 and CD56 (Biolegend), and 100 single cells were FACS-sorted into single cell lysis buffer (Ambion 4458235). Following DNase treatment, cDNA was generated using SuperScript VILO RT kit (ThermoFisher 11754050). A linear preamplification was then performed using JAK1 primers (gtcctctggatctcttcatgca, gctgtttggcaactttgaatttcc) and primers to a negative control gene NACA (cccaggcaaccacacaac, ccgactctgttttgctttactgact)

as indicated in the legend for each figure. Statistical parameters including the methods implemented, corrections for multiple comparisons, exact values of n, identity of replicates, definitions of center and dispersion and statistical significance are reported in the Figure Legends when necessary. Statistical testing for the analysis of single cell RNA sequencing is discussed in the corresponding section of the Method Details