Prevalence of Pathogenic and Potentially Pathogenic Inborn Error of Immunity Associated Variants in Children with Severe Sepsis

Children with severe sepsis between 44 weeks and 18 years old admitted to one of nine PICUs in the Eunice Kennedy Shriver National Institute of Child Health and Human Development Collaborative Pediatric Critical Care Research Network (NICHD-CPCCRN) between 2015 and 2017 without advance directives limiting care were eligible. The study was approved by the central and all nine local institutional review boards. Written informed consent was obtained from one or more parents/guardians, and child assent was garnered when able. At each center, all PICU admissions were screened twice weekly for the presence of sepsis, \(\ge\) 1 organ failure and a central venous or arterial catheter. Study enrollment occurred a median of 2 days after ICU admission (IQR: + 1 to + 4 days after ICU admission). Per-site enrollment was limited to 80 subjects. Sepsis was defined as suspected or documented infection and two systemic inflammatory response criteria: (1) tachycardia (heart rate > 90th percentile for age); (2) tachypnea (respiratory rate > 90th percentile for age); (3) temperature (< 36 °C or > 38.5 °C); and (4) abnormal WBC count (> 12,000/µL or < 4000/µL or > 10% immature neutrophils). Severe sepsis was defined by the previously mentioned criteria and at least one organ failure (cardiovascular: need for cardiovascular infusion support; pulmonary: need for mechanical ventilation support with PaO2/FIO2 ratio < 300 without support; hepatic: total bilirubin > 1.0 mg/dl and alanine aminotransferase > 100 IU/L; renal: serum creatinine > 1.0 mg/dl and oliguria (urine output < 0.5 mL/kg/h); hematologic: thrombocytopenia < 100,000/uL and international normalized ratio > 1.5 × normal; and CNS: Glasgow Coma Scale < 12 without sedatives) [15]. No geneticists or intensivists were involved in participant selection and suspicion for immunologic disorders was not considered at enrollment. Sequencing was performed between 2018 and 2020 and was not available to treating physicians.

A total of 401 pediatric patients were enrolled, for whom 381 parents (95%) provided WES consent. 2 mL of whole blood were collected for DNA extraction using standard methods. A total of 330 children completed WES (\(\mathrm{median}\;\mathrm{DNA}\;\mathrm{yield}\;39.24\mu\mathrm g,\;\mathrm{IQR}:\;20.19\mu\mathrm g-71.49\mu\mathrm g\)) and 51 had insufficient DNA (\(\mathrm{median}\;\mathrm{yield}\;1.120\mu\mathrm g,\;\mathrm{IQR}:\;0.205\mu\mathrm g-3.455\mu\mathrm g,\;p<0.0001\); Fig. 1). Those with insufficient DNA extraction had a lower median lymphocyte count on the day of sequencing blood draw (230 vs. 1200 cells/μl, p = 1.43 × 10⁻¹⁰), were older (median 8.5 y vs. 5.3 y, p = 0.0008), less likely to be previously healthy (13.7% vs. 47.9%, p = 2.8 × 10⁻⁶), more likely to have malignancy (47.1% vs. 7.0%, p = 9.21 × 10⁻¹²), and suffered higher mortality (23.5% vs. 8.5%, p = 0.0028).

The IUIS IEI report currently lists 430 genes as causes of 403 monogenic primary immunodeficiencies [9] categorized in nine groups: complement disorders, autoinflammatory disorders, combined immunodeficiencies with associated or syndromic defects, congenital defects of phagocyte number and function, disorders of immune dysregulation, defects in innate and intrinsic immunity, marrow failure, predominantly antibody deficiencies, and immunodeficiencies affecting cellular and humoral immunity (Table S1). Candidate variants were restricted to this list, with minor allele frequency (MAF) < 0.05 in ExAC, 1000 Genome, NHLBI-ESP 6500, and gnomAD databases. Variants were required to exhibit a disease-consistent inheritance pattern (one variant for autosomal dominant or X-linked in males, and two variants for autosomal recessive). We required only a single heterozygous variant for disorders with evidence supporting both recessive and dominant inheritance. Additionally, we limited missense variants to those classified as disease mutation (DM) or disease mutation? (DM?) in QIAGEN’s Professional Human Gene Mutation Database (HGMD) based on peer-reviewed literature of the variant in human disease. The DM? designation indicates a potential but not definitive association and is an uncertain link between variant and disease. In this manuscript, DM and DM? variants are defined as pathogenic and potentially pathogenic, respectively. As our dataset is without functional validation, we utilized HGMD for annotation where evidence for variant classification can be reviewed. We also aimed to increase power to detect genotype–phenotype associations in line with this resource’s aim to minimize false negatives [16]. While HGMD professional is commonly cited as evidence for variant pathogenicity, it is not equivalent to clinical sequencing which integrates additional annotation resources including ClinVar and others. Therefore, while the amino acid changes in this manuscript are reported in peer-reviewed literature, they are not equated to immunodeficiency diagnosis. Complete references for individual variants are listed in Table S2. Unique null (nonsense, frameshift, + / − 1 or 2 splice site and initiation codon) variants found in a disease-causing inheritance pattern were treated as pathogenic per the American College of Medical Genetics (ACMG) standards and guidelines for variant interpretation [17]. Highly recurrent null variants were filtered from the dataset, due to the high likelihood that they represented sequencing error or had no impact on function.

We tested for association between any IEI variant or IUIS group and infection site and inflammatory markers. All microbiologic/virologic data were limited to 48 h before or after enrollment. All results were reviewed, and likely contaminants were omitted by site including coagulase-negative staphylococci and viridans group streptococci blood cultures, Candida albicans and viridans group streptococci respiratory cultures, and mixed flora urine cultures. Culture-negative sepsis was defined as the absence of positive viral or bacterial testing from any site during the 48 h before or after enrollment after contaminant exclusion. For markers of inflammation, the most abnormal value per subject was used to define lymphopenia (minimum lymphocyte count < 1000/µL), hyperferritinemia (ferritin > 500 ng/mL), both markers of inflammation and risk for sepsis mortality [18, 19], thrombocytopenia [20] (platelet count < 150,000/µL) and extreme elevations in C-reactive protein (CRP > 10 mg/dL, normal 0.04–0.79 mg/dL) [21].

Comparisons between baseline characteristics and outcome were made between children with and without IEI using \(\chi^2\) testing for categorical variables and Wilcoxon rank sum for continuous variables. For IUIS group comparisons to children without identified variants, Fisher exact testing was performed. For IUIS group comparisons, both unadjusted and adjusted p-values following multiple testing correction for the number of groups are reported (Benjamini–Hochberg method).

For individual identified variants, a cohort minor allele frequency (MAF) was computed and compared using \(\chi^2\) testing to gnomAD (https://​gnomad.​broadinstitute.​org/​v2.​1.​1), a sequencing database of 141,456 individuals that excludes those with severe pediatric diseases [22]. This approach has been used to identify overrepresented variants in rare disease cohorts [23]. Additional MAF comparisons were made based on the self-reported race of any ethnicity between (1) African American and non-African American children in the sepsis cohort and (2) African American children in the sepsis cohort and African participants in gnomAD. p-values were multiple-test corrected using the Benjamini–Hochberg method for the number of identified variants.

Among the 330 sequenced, variants occurred in 200 children (61%) including 89 previously healthy children (Table 1). More than 3/4 of African American children with severe sepsis were found to have a genetic variant previously associated with IEI (p = 0.004). While most individuals harbored single IEI (N = 104, 52% of positive WES), 57 children (29%) had two variants, 23 had three (12%), and 16 had five or more (8%, Fig. 2, Table S3). Additionally, children with IEI had increased odds of requiring extracorporeal membrane oxygenation (ECMO, OR 4.19, 95% CI: 1.21–14.50, p = 0.019). Other interventions including mechanical ventilation, continuous renal replacement therapy, and plasma exchange did not differ between groups.

The variants exhibited both locus and allele diversity including 333 total variants at 131 loci (Fig. 3.). Among these loci, 69 were associated with autosomal dominant disorders (N = 144 individuals), 25 with autosomal recessive conditions (homozygous or compound heterozygous in N = 21 individuals), and two with X-linked recessive disorders (N = 2 individuals). A total of 35 loci were associated with disorders that can be inherited as either dominant or recessive conditions (N = 88 individuals).

In this cohort, we identified diverse IEI variants across all nine IUIS categories. Complement variants implicated in atypical hemolytic uremic syndrome (aHUS) were the most common, identified in 92 individuals (Fig. 3). The most common variants included CFH (c.3148A > T; p.Asn1050Tyr), C3 (c.1407G > C; p.Glu469Asp), and CFHR3 (c.424 C > T; p.Arg142Cys), all previously described in aHUS [24, 25, 26, 27, 28]. We also identified other alternative complement variants reported in aHUS in CFHR5, CFI, CD46, CFB, CFHR3, THBD, and CFHR4.

The next most frequent variant group was related to autoinflammatory conditions (Fig. 3). They occurred in 77 individuals and included variants reported in familial Mediterranean fever (MEFV), familial cold inflammatory syndrome type 2 (NLRP12), TNF receptor-associated periodic syndrome (TNFRSF1A), Blau syndrome (NOD2), pyogenic sterile arthritis, pyoderma gangrenosum acne syndrome (PSTPIP1), and CARD14-mediated pustular psoriasis (CARD14). The most common variant in this group, NLRP3 c.2113C > A; p.Gln705Lys was carried by 21 individuals and homozygous in one.

The remaining other pathogenic or potentially pathogenic variants according to IUIS Classification of Primary Immunodeficiencies include combined immunodeficiencies with associated or syndromic features (KMT2D, CHD7, ERBB2IP, TTC37, SPINK5, TBX1, KMT2A); congenital defects of phagocyte number and function (ELANE, CFTR, ITGB2, GATA2); disorders of immune dysregulation (UNC13D, AIRE, PRF1, FAS, CASP10, XIAP, CTLA4); defects of innate and intrinsic immunity (IRF3, STAT1, IL17RA, POLR3C, IRF4, TICAM1); bone marrow failure (TERT, RTEL1, SRP72, SKIV2L, FANCG, FANCD, TINF2, ACD); predominantly antibody deficiencies (TNFRSF13B, TCF3, SLC39A); and mutations affecting cellular and humoral immunity (TAP1, REL, POLD1, IL2RG) as shown in Fig. 3. Variant data including zygosity, predicted in silico impact, and MAF can be found in Supplementary Table S3. Phenotypic details for novel variants are found in Table S4.

IEI-variants associated with infection site (Fig. 4a). Children with variants had increased odds of positive blood or urinary culture (blood: OR 2.82, 95% CI 1.12–7.10, p = 0.022; urine: OR: 8.23, 95% CI 1.06–64.11, p = 0.016) and were less likely to have culture-negative sepsis (OR: 0.59, 95% CI: 0.38–0.92, p = 0.020). There was no significant difference in odds of positive respiratory culture or respiratory viral testing. As seen in Fig. 4b, children with IEI-associated variants were more likely to be lymphopenic (OR: 1.67, 95% CI: 1.05–2.60, p = 0.027), hyperferritinemic (OR: 2.16, CI: 1.28–3.67, p = 0.004), thrombocytopenic (OR: 1.76, 95% CI: 1.12–2.76, p = 0.013), and have a CRP > 10 mg/dl (OR: 1.71, 95% CI: 1.09–2.68, p = 0.017).

In comparison to children without identified variants, infection type and inflammatory markers also varied by the IUIS group (Fig. 4c). Individuals with variants from multiple IUIS classes had increased odds of positive blood, respiratory, or urine culture and increased odds of hyperferritinemia, lymphopenia, thrombocytopenia, and CRP > 10 mg/dl. Children with complement variants were less likely to have culture-negative sepsis and had increased odds of hyperferritinemia, thrombocytopenia, and CRP > 10 mg/dl. Individuals with autoinflammatory variants had increased odds of positive urine culture. Individuals with combined immunodeficiencies with syndromic features had increased odds of positive blood or respiratory cultures, hyperferritinemia, thrombocytopenia, and CRP > 10 mg/dl. Those innate immune disorder variants had increased odds of positive blood culture and hyperferritinemia, while those with phagocytic disorder variants had increased odds of positive urine culture. Individuals with immune dysregulation variants had the highest odds of hyperferritinemia seen in any group (OR 7.00: 95% CI: 2.3–21.1, p = 0.0005). Those with antibody deficiencies had increased odds of positive urine culture, and those with variants affecting cellular and humoral immunity had decreased odds of culture-negative sepsis. Following adjustments for multiple testing, the following comparisons remained significantly different from children with no variants: (1) any variant: hyperferritinemia; (2) multiple variants: culture negative, blood culture positive, and hyperferritinemia; (3) complement: hyperferritinemia; (4) combined syndromic: hyperferritinemia; (5) innate: hyperferritinemia; (6) phagoctye: positive urine culture; and (7) dysregulation: hyperferritinemia (Tables S5 and S6).

After candidate variant identification, we sought statistical evidence of variant overrepresentation in children with severe sepsis compared to gnomAD [22] (Table 2, Table S7). Three complement variants were significantly overrepresented: C3 c.1407G > C; p.Glu469Asp (adjusted p = 1.2 × 10⁻⁷), C3 c.443G > A; p.Arg148Gln (adjusted p = 4.5 × 10⁻⁷) both activating variants of the main complement convertase [27] and CFHR3 c.424C > T; p.Arg142Cys (adjusted p = 2.6 × 10⁻³), a negative regulator of complement [29]. In addition, IRF3 (c.829G > A; p.Ala277Thr), a transcriptional regulator of type I interferon (IFN)-dependent immune responses [30], was overrepresented in the sepsis cohort (adjusted p = 0.012).

After demonstrating that African American children had increased odds of variant identification, we sought to identify specific contributing variants. In African American children with severe sepsis, complement variants represented a higher proportion of variants than in the complete cohort (Fig. 5, N = 38/70 vs. Figure 3, N = 92/330, p < 0.0002). When comparing African American to non-African American children in our cohort, the activating convertase variant C3 (c.1407G > C; p.Glu469Asp), complement regulatory variants CFHR3 (c.424C > T; p.Arg142Cys) and CFHR5 (c.434G > A; p.Gly145Glu), and TNFRSF1A (c.224C > T; p.Pro75Leu) an autoinflammation variant associated with increased NF-kB p65 activity and IL-8 secretion [31] were overrepresented (Table 2). Next, we compared MAF in African American children with sepsis to individuals of African descent in gnomAD, highlighting potential predisposing factors for severe sepsis among groups of similar ancestral backgrounds. While limited by small numbers, we saw that C3 (c. 443G > A; p.Arg148Gln) and CFH (c.2850G > T; p.Gln950His) were both more common than expected in African American children with sepsis (Table 2).