The Autoimmune Risk R262W Variant of the Adaptor SH2B3 Improves Survival in Sepsis

Autoimmune risk alleles have been associated in genome-wide association studies for multiple autoimmune diseases, but the underlying molecular mechanisms remain elusive (1). The same loci may improve fitness against a known pathogens and thus be enriched in the population (2). The 12q24.12 region is an example of a large linkage disequilibrium region shown to have undergone a selective sweep in humans of European descent (3). The allele has progressed to become the predominant haplotype in several human populations despite associating with numerous autoimmune diseases, including type 1 diabetes (4), celiac disease, and rheumatoid arthritis (5), as well as inflammatory conditions, such as coronary artery disease (6) and hypertension (7).

The only credible coding variant in the region is the nonsynonymous single-nucleotide polymorphism (SNP) rs3184504*T encoding a hypomorphic variant in the adaptor protein SH2B3 (also termed LNK) (6). SH2B3 is a negative regulator of a range of receptor/non–receptor tyrosine kinases and cytokine receptors (including JAK2, c-fms, c-kit, mpl, and IL-7R) (8). Loss-of-function or deletion of human SH2B3 associates with acute lymphoblastic leukemia and several myeloid neoplasms (9). Missense somatic and germline pleckstrin homology (PH) domain mutations in SH2B3 have previously been identified in subjects with myeloproliferative disorders and in juvenile myelomonocytic leukemia (10, 11), but the 262W variant has not been implicated (12). The human rs3184504*T allele did associate with enhanced JAK2-dependent cytokine receptor signaling in vitro (6), suggesting that the degree of augmented signaling is not pathogenic. Furthermore, the rs3184504*T allele associates with multiple blood indices across numerous cell lineages, including progenitor cell populations (13–16). SH2B3 variants may alter the immune response by skewing hematopoiesis during infection. However, any fitness variant effect on immune function may be tempered by malignancy risk.

Sh2b3-deficient mice have been extensively studied and demonstrate baseline augmentation of several hematopoietic lineages, including hematopoietic stem cells, pre-B to mature B cells, dendritic cells, lymphocytes, monocytes, and megakaryocytes (17). Sh2b3^−/− animals additionally develop a myeloproliferative-like disease with a progressive phenotype in aged mice showing extramedullary hematopoiesis and splenomegaly (18–20). Mice on a high-fat inflammatory diet demonstrated augmented myelopoiesis, but direct testing of the role of SH2B3 in sepsis and during bacterial infections has not been performed (6). No hypomorphic SH2B3 animal studies have been reported.

In this article, we directly explore the role of the risk variant in bacterial sepsis. We tested the association of the human rs3184504*T allele with human sepsis outcomes and modeled this genetic variant in murine models for mechanistic insight. We used several cohorts of extensively genotyped patients, both healthy controls and patients with sepsis, to assess circulating human peripheral blood subsets and human survival in sepsis in relation to the genetic allele. Next, we challenged mice with global deletion of Sh2b3 or mice harboring a hypomorphic murine allele with cecal-ligation puncture (CLP), a model of polymicrobial sepsis. We tested the hypothesis that reduced function of SH2B3 promotes survival in sepsis.

Healthy control adult subjects provided informed consent, and studies were approved by the Institutional Review Board at Benaroya Research Institute (IRB07109-372). Subjects were genotyped and complete blood count (CBC) data analyzed with gPLINK-2.050. Deidentified sepsis genotypes were obtained from the identification of SNPs Altering ALI Risk (iSPAAR) consortium, a large National Heart, Lung, and Blood Institute–funded multicenter genomics study to study genetic risk factors for acute respiratory distress syndrome. Subjects were limited to whites with sepsis (n = 1542). Genomic and clinical variables for these subjects are publicly available (dbGaP study accession number phs000631.v1.p1). Subjects were originally genotyped using the Illumina 660Quad BeadChip (San Diego, CA). TagSNPs in the gene for SH2B3 were identified using the University of Washington genome variation server. Primary and secondary outcomes were inpatient mortality and peak 24-h WBC count with select plasma cytokines measured by immunoassay (Meso Scale Discovery), respectively, with a (n = 554) obtained within 48 h of enrollment. Continuous outcome measures were natural log-transformed prior to association testing in linear models. We performed multiple logistic regression adjusting for subject age and sex using an additive genetic model performed in Golden Helix (Bozeman, MT) or Stata 15 (StataCorp, LLC, College Station, TX).

Mice were bred and maintained at the Seattle Children’s Research Institute Animal Facility. All animal care and experimentation occurred with Institutional Animal Care and Use Committee approval. The Sh2b3^flox/flox and Sh2b3^tm3Draw mice were two separate lines created (Biocytogen) by inserting LoxP sites engineered into intron 3 and downstream of 3' untranslated region using C57BL6 mixed (75% B6/J and 25% B6/N) embryonic stem cells (Biocytogen). Sh2b3^tm3Draw mice additionally had a single nucleotide change (NM_008507.4:c701C > G, p.P234R) in the coding region. Mice crossed to B6.C-Tg(CMV-cre)1Cgn/JCMV-Cre mice (JAX #06054; N11) generated the SH2B3^KO (SH2B3^−/−) mouse line or the B6.Cg-Tg(ACTFLP3)9205Dym mice (JAX #005703) to remove the FRT-flanked neomycin-resistance cassette and maintained backcrossed to C57BL/6 mice (The Jackson Laboratory). Bone marrow (BM) chimeric mice were generated by two equal daily 500-rad irradiation doses followed by i.v. injection of 5 × 10⁶ purified BM cells from either Sh2b3^WT or Sh2b3^KO mice. Murine CBCs were performed at Phoenix Lab veterinary services facility using a Siemens Advia 2120 flow cytometry device with peroxidase methodology followed by confirmatory manual differentials. Baseline BM and spleen from age-matched (16–18 mo) littermate mice were preserved in 10% neutral buffered formalin with or without decalcification prior to paraffin-embedding and preparing of 4-µm sections and then stained with H&E. Blinded enumeration of megakaryocytes was performed in five high-power fields (×40) and summed per animal.

High-grade CLP modeling was performed as described previously (21) with survival rates in control mice expected to be ∼20% similar to published protocols (22, 23). All sepsis experiments had a blinded experimental design. Briefly, the distal two thirds of the cecum from a 12- to 18-wk-old male mouse was ligated and punctured with a 22-gauge needle prior to placing the cecum back into the abdomen. Sterile saline (1 ml) was administered into the peritoneal cavity, and the incision was closed using sutures and 9-mm steel wound clips. Antibiotics were not administered secondary to experimental design. Mouse rectal temperatures were measured at 24 h after CLP. Clinical monitoring of mice was performed at least four times daily for the first 3 d and then twice daily for 7 d. Animals considered to be moribund or below body temperature threshold were euthanized by CO₂ inhalation. Tissues from CLP animals were harvested at 24 h postprocedure. An all-inclusive histologic semiquantitative severity score was used that a blinded comparative pathologist used to score lesions in liver and lung based on a scale of 1+ (minimal) to 4+ (severe). Criteria included the distribution of lesions, severity of the inflammatory cell accumulation, and severity of hepatocellular degeneration (for liver sections) or perivascular edema (for lung sections). The overall severity score is a sum of all scores for each mouse.

Murine TNF and IL-6 levels were measured using ELISA MAX Kits (BioLegend) with detection limits of 4 and 2 pg/ml, respectively. Murine keratinocyte-derived chemokine (KC) amounts were measured with Mini ABST ELISA Kits (PeproTech) with detection limits of 16 pg/ml and 4 pg/ml, respectively. Luminex Cytokine & Chemokine Panel 1 (26-plex) (Thermo Fisher Scientific) analysis was performed. Human cytokines and chemokines were measured by immunoassay (Meso Scale Discovery). Alanine aminotransferase and blood urea nitrogen (BUN) levels were measured using a colorimetric method with detection limits of 0.5 IU/l and 100 mg/dl, respectively (Teco Diagnostics).

Peritoneal fluids or blood were diluted on Luria-Bertani agar (for peritoneal fluids) or tryptic soy agar with 5% sheep blood plates (for blood). Colonies were visually scored after overnight incubation at 37°C. The oxidative stress indicator CM-H2DCFDA (Thermo Fisher Scientific) and FITC-labeled Escherichia coli (K-12 strain) BioParticles (Thermo Fisher Scientific) were used for reactive oxygen species (ROS) and phagocytosis assays, respectively. Cells were stained with markers for neutrophils (CD11b⁺, Ly6G⁺, Ly6C⁻), monocytes (CD11b⁺, Ly6G⁻, Ly6C⁺), or macrophages (CD11b⁺, F4/80⁺) and evaluated via flow cytometry. As a control for ROS reactivity of each sample, the cells were treated with or without 100 ng/ml PMA for 7 min at 37°C.

5-Ethynyl-2’-deoxyuridine (EdU; Thermo Fisher Scientific) was dissolved in sterile 1× PBS (10% DMSO) and 1 mg in a volume of 100 μl was injected i.p. 1 h prior to sacrifice. EdU detection was performed with the Click-iT Plus EdU Pacific Blue Flow Cytometry Assay Kit (Thermo Fisher Scientific). Cells from total BM were stained for surface markers or for myeloid progenitors in the BM; total BM was subject to lineage depletion prior to surface staining followed by fixation and incubation with the Click-iT Plus reaction mixture according to the manufacturer’s instructions.

Whole-cell lysates were analyzed by Western blot using primary Abs: mouse anti-Lnk (Santa Cruz Biotechnology) and rabbit anti–β-actin (Sigma-Aldrich). Anti-mouse and rabbit secondaries (LI-COR Biosciences) were used and imaged on an Odyssey Infrared Imager (LI-COR Biosciences) and quantified by ImageJ software.

Murine cells were stained using PE-Ly-6G (1A8); eFluor 450-CD11b (M1/70); allophycocyanin-F4/80 (BM8); PE-Cy7-Ly6C (HK1.4); Pacific Blue-CD117 (c-kit) (2B8); PE-CD115 (AFS98); Ly6C PE-Cy7 (HK1.4); Ly6G PE (1A8); and CD16/32 PerCP-Cy5.5 (93) (BioLegend). Live/Dead Fixable (near-IR or AF350; Invitrogen) or DAPI was used to detect dead cells. EdU detection was performed with the Click-iT Plus EdU Pacific Blue Flow Cytometry Assay Kit (Thermo Fisher Scientific). Cells were acquired on a BD LSR II with FACSDiva software and analyzed with FlowJo software (version 10.2; Tree Star).

All statistical analysis used GraphPad Prism version 9.0 except where noted above. All specific statistical tests and p values are indicated in the relevant figures. Data were tested for normality using Anderson-Darling, D’Agostino-Pearson, Shapiro-Wilk, and Kolmogorov-Smirnov test depending on sample size. To assess statistical significance between two groups with normal distribution data sets, the Student t test was used. Nonnormal data were analyzed by Mann–Whitney U test. When three groups were analyzed, we used one-way ANOVA with Tukey multiple-comparisons post hoc analysis or Kruskal-Wallis test for nonparametric data with post hoc Dunn multiple-comparisons analysis.

The data that support the findings of this study are available from the corresponding author upon request.

The rs3184504*C allele encodes for SH2B3 Arg262 (CGG, Arg, R) and represents the ancestral allele or “nonrisk” for autoimmune disease. The autoimmune-associated “risk allele” (TGG, Trp, W) is a derived allele that is rare in East Asian and African populations (gnomAD maf T = 0.001 and T = 0.073, respectively), while approaching the major allele in cohorts of non-Finnish European subgroups (T = 0.487 in gnomAD). The positive selection of the haplotype has been previously noted (3), but further testing of the host response to infection is lacking. In this study, we analyzed the rs3184504 allele frequency in a known sepsis cohort of genotyped patients. We used the National Heart, Lung, and Blood Institute–funded iSPAAR consortium of genotyped patients with sepsis defined using the 2001 consensus definition used at the time of enrollment (24). Reanalysis showed these subjects (1527 of 1542 or 99%) also meet the Sepsis-3 guideline definition for patients in the intensive care unit with an admission of sequential organ failure assessment score ≥2 along with a suspected infection (Supplemental Table I) (25). Analysis was limited to white subjects (n = 1050) given the allele frequency bias. Initial analysis revealed subjects with the rs3184505*T/T genotype had 10.42% mortality (39 of 374) compared with 17.30% (135 of 780) and 17.26% (67 of 388) mortality with rs3184504*T/C or C/C genotypes, respectively.

Next, we performed a multivariate logistic regression model to consider variants within 2 kb upstream or downstream of the SH2B3 gene, with a minor allele frequency >0.05 in the HapMap CEU population and chose an R² threshold of 0.5. This identified four SNPs previously genotyped in this data set (rs2239196, rs2239194, rs3184504, and rs739496), all with Hardy-Weinberg equilibrium p > 0.001 and call rate >0.99 (Supplemental Fig. 1). The primary outcome measure for our analysis was inpatient mortality, whereas secondary outcomes included WBC count and plasma cytokine levels. In a multiple logistic regression model adjusted for age and sex, the rs3184504*T allele uniquely associated with improved survival (odds ratio [OR] 0.76 [95% confidence interval (CI) 0.62–0.92]; adjusted p = 0.0054] (Table I). Given that genome-wide data were available for the subjects in iSPAAR, next, we calculated principal components using the genomic data adjusting for the first three principal components, in addition to age and sex, to account for unmeasured confounding and further adjustment for APACHE III score, a measure for severity of illness in patients in the intensive care unit. Similar estimates and degree of significance were found (OR 0.80 [95% CI 0.64–0.99]; adjusted p = 0.04). The rs3184504*T allele also associated with an increased peak absolute WBCs in the first 24 h of sepsis (β= 0.05; adjusted p = 0.032) (Supplemental Fig. 1). Conversely, IL-8 levels, a known mediator associated with poor outcomes (26), were reduced with the rs3184504*T allele (β= −0.19; p = 0.025). IL-6, G-CSF, and sTNFR1 were not different among rs3184504 genotype groups. Thus, the rs3184504*T/T genotype associated with reduced sepsis mortality, elevated peak WBCs, and lower circulating IL-8 levels in this cohort. Collectively, homozygosity of the rs3184504*T risk allele was associated with reduced sepsis mortality in this cohort.

To better understand the improved sepsis response in humans, we used a murine model to directly test the role of SH2B3 during a polymicrobial sepsis response. The rs3184504*T allele has been shown to be hypomorphic as a negative regulator (6); thus, we first performed studies in SH2B3-deficient mice (Fig. 1A). Using a CLP model, we found Sh2b3^KO mice exhibited decreased mortality over 7 d (Fig. 1B) and demonstrated reduced morbidity measured by drop in body temperature at 24 h, an accurate surrogate for mortality (Fig. 1C) (27). Endothelial cell expression of SH2B3 has been reported (28); thus, we tested whether SH2B3 deficiency in hematopoietic cells was sufficient for the protective phenotype. We generated BM chimeras using Sh2b3^WT or Sh2b3^KO BM transferred into irradiated wild-type hosts. Chimeras were rested 12 wk to allow reconstitution, and then CLP surgeries were performed. We found SH2B3 deficiency in hematopoietic cells was sufficient for improved morbidity (Fig. 1D). For the remainder of our studies, we thus used nonirradiated mice. We next analyzed mice for early signs of tissue damage postsepsis. At 24 h following CLP, lung and liver tissue showed early signs of tissue damage uniquely in the Sh2b3^WT mice, including neutrophil infiltration with edema and hepatocellular degeneration and perivascular edema (Fig. 1E, 1F). Average semiquantitative severity across mice was scored. Additionally, BUN was elevated in the Sh2b3^WT mice compared with Sh2b3^KO mice (Fig. 1G). Overall, Sh2b3^KO mice lacked significant organ injury at 24 h.

Sh2b3^KO mice also demonstrated improved bacterial clearance and less organ damage following CLP. At 24 h following surgery, Sh2b3^KO mice exhibited improved systemic and local bacterial clearance measured by CFUs from blood and peritoneal cavity, respectively (Fig. 1H). Peritoneal granulocytes from Sh2b3^KO mice (gated separately on neutrophils [CD11b⁺Ly6C⁻Ly6G⁺] and “classical” inflammatory monocytes [CD11b⁺Ly6C^hi Ly6G⁻], respectively) demonstrated enhanced ex vivo phagocytosis compared with controls (Fig. 1I). Peritoneal cells were also tested for ROS following PMA stimulation, and SH2B3-deficient monocytes produced reduced ROS from each animal compared with control mice, whereas neutrophils demonstrated no difference (Fig. 1J). Thus, in the setting of polymicrobial CLP, deficiency of SH2B3 results in a more rapid bacterial clearance, enhanced phagocytosis, and less organ damage.

For most pathogens, monocyte recruitment to the site of inflammation enhances resistance to that infection (29). We next characterized and quantified peritoneal granulocytes ex vivo following the CLP (Fig. 2A). Total cellular influx and absolute neutrophils were not significantly different in the peritoneal cavity between genotypes, but Sh2b3^KO mice displayed augmented absolute monocyte counts at both 12 and 24 h compared with Sh2b3^WT mice (Fig. 2B). There was no significant difference in circulating blood monocytes in Sh2b3^KO mice following CLP, whereas the percentage of neutrophils was reduced (Fig. 2C). Thus, in the absence of SH2B3, inflammatory monocytes were increased at sites of peritoneal inflammation.

We hypothesized that augmented monocyte-related chemokine expression would be found in the peritoneal fluid of Sh2b3^KO mice during these models. We performed a broad cytokine and chemokine multiplex array and targeted ELISA testing on peritoneal fluid 24 h post-CLP from Sh2b3^WT (n = 6) and Sh2b3^KO mice (n = 6) (Fig. 2D, 2E). Overall, compared with controls, Sh2b3^KO mice had a broad reduction in the proinflammatory cytokines associated with tissue damage, including IL-6, TNF-α, TSLP, IL-1β, and IL-17A, findings that likely reflected earlier bacterial clearance in Sh2b3^KO mice. Consistent with reduced levels of IL-8 in septic human subjects with the rs3184504*T autoimmune risk allele, the murine surrogates for IL-8 (CXCL1 [KC] and CXCL2 [MIP-2]) were also decreased in the peritoneum of Sh2b3^KO mice in the setting of CLP. Surprisingly, monocyte chemotactic factors measured, including CCL2 (MCP-1) and CCL7 (MCP-3), were also reduced in Sh2b3^KO mice. Thus, although the reduction in proinflammatory cytokines, including IL-8 surrogates, correlated with the reduced morbidity post-CLP, altered chemokine levels could not explain the augmented monocytes in SH2B3-deficient mice.

We next hypothesized that peripheral monocytes and/or their myeloid progenitors may be proliferating to account for the enhanced monocyte population at the site of infection. In the CLP model, immature myeloid cells and hematopoietic stem and progenitor cells are mobilized from the BM niches and can be measured in the spleen (30–32). Thus, we performed an i.p. injection of EdU at 24 h post-CLP and 1 h prior to sacrifice (Fig. 3A). EdU-positive cycling cells were identified in the BM and spleen. The EdU⁺B220⁻ fraction was similar between mice, indicating equal labeling of the BM (Fig. 3B). We focused on the splenic non-B, non-T cell fraction that contains myeloid cells and hematopoietic stem and progenitor cells. We analyzed the neutrophils, monocytes, and c-kit–positive cells and found a significant increase in CD11b⁻c-kit⁺ cells in the spleen from Sh2b3^KO mice, as well as Ly6C^hi monocytes (CD11b⁺Ly6G⁻Ly6C^hi) (Fig. 3C, 3D). Thus, reduced SH2B3 function correlates with increased peripheral myeloid progenitors and monocytes following CLP in mice. Overall, absence of the SH2B3 adaptor improved the immediate sepsis response in mice.

Human SH2B3 deficiency associates with malignancy and myelodysplasia. In contrast, the autoimmune risk allele involving SH2B3 is hypomorphic. Thus, we reasoned that a hypomorphic murine SH2B3 variant similar to the human autoimmune risk allele may be sufficient to improve the response to the CLP. The human ancestral arginine at position 262 is highly conserved in mammals outside of rodents with proline at the homologous position 234. Based upon in silico modeling from the murine SH2B2 (APS) PH domain (Protein Data Bank: 1V5M), we predicted an arginine substitution (P234R) in the murine SH2B3 would partially disrupt the PH domain. We created a novel knockin mouse using homologous recombination and introduced the 234R into the genomic murine locus (Fig. 4A). We included flanking loxP sites allowing cre-dependent excision of exons 3–8 as well. Founder mice were established that were healthy and fertile, and heterozygous mice were intercrossed to deliver littermate controls (Sh2b3^WT) and homozygous knockin (Sh2b3^KI) mice for all remaining experiments. We compared protein expression and found the 234R coding change did not impact protein expression (Fig. 4B).

We next compared littermates for in vivo pathology previously observed in SH2B3 knockout mice (18). Deficiency in SH2B3 augments numerous cytokine receptor signaling pathways (17). Sh2b3^−/− mice demonstrate augmented peripheral WBC, lymphocyte, and monocyte counts, and aged mice demonstrate myeloproliferative phenotypes with augmented megakaryocytes and extramedullary hematopoiesis (18). We hypothesized that the Sh2b3^KI mice would demonstrate an intermediate phenotype between Sh2b3^WT and Sh2b3^KO mice, including splenomegaly, leukocytosis, extramedullary hematopoiesis, and increased megakaryocytes (18). Thus, we analyzed the blood, BM, and spleen of aged mice (Sh2b3^WT, Sh2b3^KI, and Sh2b3^KO aged 15–18 mo) to compare the phenotype to the published phenotype from SH2B3-deficient mice ((Fig. 4C–E). Histologic analysis of BM revealed a higher megakaryocyte cell density in both Sh2b3^KI and Sh2b3^KO mice per high-power field (Fig. 4C), and low power showed increasing hypercellularity in Sh2b3^KI and Sh2b3^KO mice, respectively (Fig. 4D). Increases in other myeloid cells were also observed within the red pulp of the spleen. Murine peripheral blood predominantly contains lymphocytes (33), which were also observed in our studies. Peripheral blood analysis from SH2B3-deficient mice displayed leukocytosis with elevations in lymphocytes, monocytes, and neutrophils (Fig. 4E), similar to cell counts reported in the literature (18). The Sh2b3^KI mice also showed leukocytosis when compared with Sh2b3^WT mice, specifically with elevated circulating absolute lymphocyte counts intermediate between Sh2b3^WT and Sh2b3^KO mice (Fig. 4E). Thus, introduction of the hypomorphic variant into the murine PH domain mimicked Sh2b3^KO mice, but resulted in a less severe, intermediate phenotype.

We next tested the hypothesis that homozygous Sh2b3^KI mice would exhibit reduced mortality following induction of CLP. We compared mortality over 7 d post-CLP induction. Homozygous Sh2b3^KI mice showed less mortality compared with heterozygous knockin mice (Sh2b3^234P/R) and Sh2b3^WT littermate control mice (Fig. 5A). Morbidity post-CLP, as measured by drop in body temperature in the first 24 h, revealed an allele effect with more homozygous animals maintaining body temperature compared with the Sh2b3^234P/R mice (Fig. 5B). Bacterial clearance in the homozygous Sh2b3^KI and Sh2b3^234P/R mice was not significantly different from the Sh2b3^WT littermate control mice, although a trend for lower bacterial burden in the peritoneal cavity and systemically was observed stepwise with allele burden (Fig. 5C). Lastly, comparing the local cellular infiltrate in the peritoneal cavity, inflammatory monocytes (CD11b⁺Ly6C⁺Ly6G⁻) were significantly increased in Sh2b3^KI mice compared with Sh2b3^WT control mice (Fig. 5D). In contrast, the peritoneal cells in Sh2b3^234P/R mice were not discernibly different from wild-type mice, suggesting again that homozygosity was necessary to improve the sepsis response in the CLP animal model. Overall, our data suggest that reduced SH2B3 function increased inflammatory monocyte recruitment, resulting in improved bacterial clearance and reduced mortality in hypomorphic Sh2b3^KI mice.

In this study, we show that reduced function of the SH2B3 adaptor protein results in an improved host response to bacterial sepsis in both humans and mice. First, we found in human subjects with sepsis, homozygosity for the rs3184504*T autoimmune risk allele significantly associated with reduced mortality and higher peak leukocyte counts. Previous studies using human cord blood cells suggested that expression of the rs3184504*T allele correlated with increased thrombopoietin signaling, consistent with a hypomorphic or reduced function of this inhibitory adaptor protein (6). Given the linkage disequilibrium at this locus, we transitioned to murine modeling to examine the role of the SH2B3 adaptor protein in sepsis. We found that SH2B3 deficiency decreased mortality and morbidity and demonstrated less organ damage in a CLP model. The SH2B3-deficient mice achieved pathogen control with enhanced phagocytosis and increased monocytes at the inflammatory site. Finally, we created a knockin mouse model with a variant in the PH domain, demonstrating similar features as the Sh2b3^−/− mice, including increased megakaryocytes, augmented baseline leukocytes, and tissue pathology findings, all suggestive of a hypomorphic allele. To model the human variant, we challenged the Sh2b3 knockin mice with the CLP model showing decreased mortality, increased bacterial clearance, and increased monocytes at the site of inflammation. Taken together, these findings suggest that the SH2B3^262W autoimmune risk variant provides protection during acute bacterial challenge in both mice and humans.

Negative regulators of cytokine signaling have emerged as targets of positive natural selection, often with improved fitness against infections (1). Simultaneously, these alleles are also important potential autoimmune risk variants (34). Our data suggest that SH2B3 fits within this growing paradigm. Previous work and ours have consistently found rs3184504*T allele associates with many increased hematopoietic traits, including elevated monocytes (Supplemental Fig. 1) (13, 14, 35). Consistent with these findings, human monocyte progenitors and terminal monocytes highly express SH2B3 transcripts (36, 37). Given the plasticity of the myeloid lineages and short neutrophil life span, genetic risk factors that regulate aspects of myelopoiesis are likely impactful and only starting to be explored.

Hematopoietic progenitors are rapidly mobilized following CLP. During the early phases post-CLP in murine sepsis, distinct neutrophil progenitors are expanded in the BM at the expense of monocytes (38). Our data showed increased cell cycling of c-kit⁺ as well as the CD11b⁺Ly6C⁺ fraction in the spleen uniquely in the SH2B3-deficient animals as evidence for altered myelopoiesis. Emergency or demand-adapted myelopoiesis has been a well-described observation of shift in BM hematopoiesis favoring myelopoiesis at the expense of lymphopoiesis. However, growing evidence suggests that the neutrophil versus monocyte-dendritic cells lineage decision is equally critical in the modulation of the immune response (39). In human studies, pooled genomic signatures of sepsis mortality revealed whole-blood gene expression of several neutrophil-specific genes (DEFA4, MPO, CTSG, and BPI) increased with high mortality and several monocyte-specific genes (CCR2, EIF5A, CX3CR1, and EMR3) decreased with high mortality (40). Although it is likely that the ratio of myeloid subpopulations plays an important role in the sepsis response, further studies are needed to assess potential differences in these progenitors in the absence of SH2B3 and whether expanded progenitor populations are sufficient for an improved response.

Pathogen-imposed selection pressures have also been linked to altered genetic alleles involved in innate sensing pathways (41, 42). Previous work using PBMCs from human subjects stimulated with muramyl dipeptide, a known NOD2 ligand, demonstrated increased IL-8 secretion in association with the rs3184504*T/T autoimmune risk genotype (3). In contrast, we found, similar to studies in macrophages from SH2B3-deficient mice (43), that reduced human SH2B3 function based on the rs3184504-T allele did not significantly alter TLR signaling on a per-cell basis (Supplemental Figs. 2 and 3). Augmented monocyte absolute numbers may explain the previously reported findings, as we found consistently higher absolute monocyte numbers associating with the rs3184504 allele in a healthy control cohort (Supplemental Fig. 1). Unfortunately, our sepsis cohort only collected peak total white blood counts and not further differential data to determine if the monocytosis was augmented during human sepsis.

In this study, we also demonstrated that reduced function of SH2B3 associated with rapid bacterial clearance and enhanced phagocytic abilities. Previous work with BM-derived macrophages from SH2B3-deficient mice showed no significant difference in zymosan-labeled particle phagocytosis (43), but minimal follow-up has been performed with primary cells from Sh2b3^−/− mice. In vivo studies of mice with augmented JAK2 signaling (JAK2^V617F mice) showed increased erythrophagocytosis in the macrophages at baseline, suggesting that a JAK2-dependent cytokine receptor may augment this function (44). Hematopoietic myeloid progenitor cells and the developmental cues have been found to produce functionally distinct monocytes based upon gene signature, but direct assessment of phagocytic function was not performed (40). In the early phase following CLP, we found the microbicidal function of granulocytes augmented with enhanced phagocytosis, but not altered ROS. The rapid bacterial clearance was in the absence of antibiotic prophylaxis in our animal model and thus represents a possible gain-of-function effect that may explain the selective sweep that occurred with the rs3184504*T genetic allele (3). Further studies will be needed to understand the detailed mechanism of the enhanced phagocytic function following a bacterial challenge.

In summary, our data show that reduced function of the SH2B3 adaptor improves the early sepsis response and suggest an important role for the SH2B3^262W coding variant in immune regulation. Our findings support the hypothesis that the SH2B3^262W risk protein is a causative variant that underwent positive selection to promote improved monocyte function and accumulation in the absence of severe tissue damage. Future work using the models described in this study will help to comprehensively assess the impact of the SH2B3^262W risk in autoimmune pathogenesis, including whether regulating life span or function of innate myeloid responses may contribute to the initiation and/or acceleration of human autoimmunity.

We thank S. Khim for animal husbandry, the University of Washington Histology and Imaging Core, and Benaroya Research Institute Translational Core Laboratory.

The authors have no financial conflicts of interest.