Adults with septic shock and extreme hyperferritinemia exhibit pathogenic immune variation

Abstract

Post-hoc subgroup analysis of the negative trial of interleukin-1β receptor antagonist (IL1RA) for septic shock suggested that patients with features of macrophage activation syndrome (MAS) experienced a 50% relative risk reduction for mortality with treatment. Here we seek a genetic basis for this differential response. From 1341 patients enrolled in the ProCESS trial of early goal directed therapy for septic shock, we selected 6 patients with MAS features and the highest ferritin, for whole exome sequencing (mean 24,030.7 ηg/ml, +/SEM 7,411.1). Eleven rare (minor allele frequency <5%) pathogenic or likely pathogenic variants causal for the monogenic disorders of Familial Hemophagocytic Lymphohistiocytosis, atypical Hemolytic Uremic Syndrome, Familial Mediterranean Fever, and Cryopyrin-associated Periodic Fever were identified. In these conditions, seven of the identified variants are currently targeted with IL1RA and four with anti-C5 antibody. Gene-targeted precision medicine may benefit this subgroup of patients with septic shock and pathogenic immune variation.

Introduction

A landmark study of Danish adoptees showed a near 6-fold increase in the risk of death from infection before age 50 for adoptees whose biological parents also died from infection under 50.¹ Multiple familial, case control and genome wide association studies have sought to identify genetic variation that contributes to sepsis outcome.^2–6 However, despite great variation in host response, attempts to identify genetic variants that contribute to sepsis outcomes has proven challenging. Typically, genomic studies in sepsis have treated all patients as a single group, assuming shared genetic risk factors. They have also focused on correlations between common polymorphisms and sepsis outcome with limited functional studies to support associations.^4,7,8

MAS is a fulminant form of multi-organ dysfunction presenting with fever, cytopenia, hepatosplenomegaly, hemophagocytosis, and extremely elevated serum ferritin, that can be triggered by sepsis.⁹ Although classically described in children, septic hyperferritinemic adults are also at risk of poor outcome.^9,10 Additionally, while overall, interleukin-1β receptor antagonist (IL1RA) therapy was ineffective for septic shock, post-hoc subgroup analysis showed that patients with MAS features experienced a 50% relative risk reduction for mortality when treated with IL1RA.^11,12 As the hyperferritinemic sepsis phenotype overlaps with other inherited immunologic disorders, including hemophagocytic lymphohistiocytosis (HLH), Cryopyrin Associated Periodic Syndromes (CAPS), Familial Mediterranean Fever, and atypical Hemolytic Uremic Syndrome (aHUS), we hypothesized that known pathogenic disease variants for these disorders would be identified in individuals with this sepsis phenotype. This work uses whole exome sequencing (WES) to identify causal mutations for these diseases.^13–17 Here we focus on a subset of adults with septic shock with MAS features to identify rare pathogenic variants known to cause monogenic immunologic disorders, potentially underlying a shared phenotype leading to multi-organ failure.

Results and Discussion

Because MAS is characterized by extremely high serum ferritin levels, we identified 6 patients (0.5%) with the highest concentrations from the ProCESS multicenter trial of protocolized early goal directed therapy for adult septic shock¹⁸ to undergo WES. We hypothesized that patients with hyperinflammation identified by the highest serum ferritin during macrophage activation syndrome would have pathogenic variants known to cause other single gene disorders of defective inflammation control, in a shared genotype-phenotype hypothesis. All variants with a minor allele frequency (MAF) >5% as reported by the 1000 Genomes, ExAC and NHLBI Esp6500 databases were treated as polymorphisms and removed from further analysis. Additionally, to bolster claims of disease relevance, previous literature reports of clinical phenotype related to pathogenic or likely pathogenic variant was required in the filtering scheme. Subsequently, only variants with MAF <5% that have been previously reported as pathogenic or likely pathogenic for heritable immunologic disorders are discussed here. Reference to the general population allows identification of all pathogenic and likely pathogenic variants, which are unlikely to be identified in a smaller sepsis controls sample. The mean ferritin among these 6 patients was 24,031 +/− 7,411 ηg/ml (+/− SEM). As hemophagocytic phenotypes are rare in adults, soluble IL-2 receptor, triglyceride and NK cell function studies are not available. The 6 patients had a mean APACHE II score of 27 +/− 4, corresponding to a mortality risk of 60.5% and multiple system organ failure including hepatic dysfunction and coagulopathy typical of MAS.¹⁹ Five of 6 patients died by 30 days, with subject 3 being the sole survivor. Halacli et al. recently reported a cohort of adults with severe sepsis where 0.7% had ferritin over 15,000 ηg/ml with an observed mortality was 100%.¹⁰ Clinical characteristics of these 6 patients are shown in Table 1.

Table 1.

Clinical phenotypes of subjects enrolled in the study. Lab values represent baseline values at time of enrollment. BSI: blood stream infection, PNA: pneumonia, UTI: urinary tract infection. Culture negative indicates that no positive culture from any site was obtained.

Subject	Age	Sex	SBP (mmHg)	Lactate (mmol/L)	WBC (×109 /L)	Hgb (g/dL)	Plt (x109/L)	INR	PTT (s)	Tbili (mg/dL)	Cr (g/dL)	Ferritin (ηg/ml)	Infection	APACHE II	Dead at 30d
1	32	M	80	3.9	2.9	8.4	44	1.5		2.5	3.1	14,949	Culture negative	24	Yes
2	73	M	83	16	10.5	17.4	57	1.2	26.0	1.5	2.7	36,240	UTI/BSI	42	Yes
3	64	F	91	7.4	2.9	14.8	33			1.7	3.3	7,259	BSI	18	No
4	44	F	140	9.5	6.4	9.1	25	1.8		6.2	0.8	8,329	PNA/BSI	20	Yes
5	51	M	70	6.3	4.5	13.9	50		47.1	1.8	3.5	55,314	PNA/BSI	37	Yes
6	70	F	102	3.9	8.4	5.1	88	3.2	48.0	6.4	5.1	11,850	Culture negative	22	Yes
MEAN	56		94	7.8	5.9	11.5	50	1.9	40.4	3.4	3.1	24,031		27
+/− SEM	6.6		10	1.9	1.3	1.9	9	0.4	0.9	0.9	0.6	7,411		4

As shown in Table 2, 100% of subjects inherited at least one pathogenic or likely pathogenic variant previously reported in the literature as causal for heritable immunologic diseases. Despite the similarity in phenotype, the variants found on WES represent diverse genetic disorders. Three of the six patients had UNC13D variants, where mutations are known to cause abnormal NK cell degranulation and cytolytic killing²⁰ and when inherited as autosomal recessive AR (Bi-allelic mutations) cause Familial HLH Type 3. Accumulating evidence suggests that adults with sporadic HLH¹⁶ or MAS in the context of rheumatic disease,^17,21 or infections^22–24 often carry heterozygous mutations in NK cell degranulation pathways including UNC13D. The heterozygous UNC13D (c.1579C>T; p.Arg527Trp) mutation in subject 1 has been reported in an 18-year-old female with familial HLH with normal PRF1 and STXBP2 sequencing²⁵ as well as a 12 year HLH patient who also carried a PRF1 c.50delT variant²⁶. The UNC13D (c.2782C>T; p.Arg928Cys) variant in subject 3 has been described by Aricò et al in 3 individuals with ALPS²⁷ a predisposing condition for FHLH, and by Kaufman in a heterozygous individual with MAS related to systemic Juvenile Idiopathic Arthritis²⁸. The variants described in subject 5, UNC13D (c.2983G>C; p. Ala995Pro) and (c.2542A>C; p.Ile848Pro) have also been described in an ALPS patient without identified FAS, FASL, or CASP10 mutation with functional studies of transfected HMC-1 cells showing decrease in secretory granule fusion with the plasma membrane as measured by a CD63 expression assay following stimulation with fMLP, a macrophage activating chemokine²⁷.

Table 2.

Table of previously reported pathogenic variants identified during whole exome sequencing of sepsis patients with extreme hyperferritinemia. Numbers 1–6 indicate study subjects where rows represent individual variants. Columns indicate genes with mutations identified and the specific variant. Reference numbers in column 3 refer to original reports of human pathogenicity, and column 4 the genetic disorder associated with it. In order to support claims of disease relevance, previous literature reports of clinical phenotype was required as part of the variant filtering scheme. Subsequently, all eleven variants have been previously reported in individuals with either aHUS, HLH, or the periodic fever syndromes of CAPS or FMF as indicated. MAF per the ExAC database is provided. PhyloP computes conservation or evolutionary acceleration. Values are between −11.764 and +6.424. Positive scores indicate conservation; negative scores fast-evolution⁶¹. SIFT scores are between 0–1 with scores less than 0.05 predicted to be deleterious, those greater than or equal to 0.05 are predicted to be tolerated⁶². Putative targeted therapies have been suggested based on the identification of these variants in the context of aHUS, HLH, MEFV and CAPS, with citations provided. aHUS: atypical hemolytic uremic syndrome; CAPS: cryopyrin associated periodic fever syndrome, FMF: familial Mediterranean fever, HLH: hemophagocytic lymphohistiocytosis; MAF: Minor allele frequency.

Subject	Gene	Variant	Amino Acid Change	Disease	MAF	PhyloP Score	SIFT Score	Putative Therapy
1	C3	c.1407G>C 52 NM_000064.2	p.Glu469Asp	aHUS	0.00394	−0.9	1	Anti-C5 ab 53–55
	UNC13D	c.1579C>T 25,26 NM_199242.2	p.Arg527Trp	HLH	0.00523	0.45	0.02	IL1-RA 56,17
2	CD46	c.1058C>T 57 NM_172359.2	p.Ala353Val	aHUS	0.01532	−3.19	0.47	Anti-C5 ab 53–55
	CFHR5	c.832G>A 58 NM_030787	p.Gly278Ser		0.00729	1.39	0.03
3	UNC13D	c.2782C>T 27,28 NM199242.2	p.Arg928Cys	HLH	0.02986	0.65	0.13	IL1RA 56,17
4	NLRP3	c.2113C>A 35 NM_004895.4	p.Gln705Lys	CAPS	0.0495	−0.17	0.22	IL1RA 15
	MEFV	c.250G>A 37 NM_000243.2	p.Glu84Lys	FMF	0.00012	1.48	0	IL1RA 59
5	UNC13D	c.2983G>C 27 NM_199242.2	p.Ala995Pro	HLH	0.00096	1.52	0.22	IL1RA 56,17
		c.2542A>C 27 NM_199242.2	p.Ile848Leu		0.00090	−0.72	0.10
6	CD46	c.1058C>T 57 NM_172359.2	p.Ala353Val	aHUS	0.01532	−3.19	0.47	Anti-C5 ab 53–55
	MEFV	c.2084A>G 60 NM_000243.2	p.Lys695Arg	FMF	0.00550	−0.05	0.13	IL1RA 59

Among those with pathogenic mutations for recurrent fever syndromes, causal mutations for CAPS and FMF were identified in MEFV and NLRP3.^29–33 Interestingly, subject 4 carried both NLRP3 (c.2113 C>A; p.Gln705Lys) and MEFV (c.250 G>A; p.Glu84Lys) variants. Screening of individuals with clinical suspicion for CAPS demonstrated that the NLRP3 (p.Gln705Lys) variant is associated with a mild autoinflammatory phenotype characterized by skin lesions, arthralgia and myalgia.^34,35 This variant increases IL-1β and IL-18 release by cultured monocytes in an IL-1 receptor-dependent manner.³⁶ Similarly, the MEFV (p.Glu84Lys) variant has been reported in both heterzogygous and compound heterozygous states in FMF.^37–39 Other studies have investigated potential interaction of NLRP3 and MEFV variants in digenetic inheritance of abnormal Il-1 driven FMF.^33,40 While there are multiple treatment strategies for CAPS/FMF¹⁵ and MAS,^16,17 both can be targeted with IL1RA.

Complement pathway mutations causal for aHUS were identified in three subjects: two in CD46, both with (c.1058C>T; p.Ala353Val) and one in C3 (c.1407G>C; p.Glu469Asp) and CFHR5 (c.832G>A; p.Gly278Ser). aHUS is a thrombotic microangiopathy resulting from uncontrolled complement activation causing anemia, thrombocytopenia and kidney failure.¹³ This is markedly similar to Thrombocytopenia Associated Multiple Organ Failure, a sepsis phenotype of new onset thrombocytopenia and acute kidney injury where we have recently reported increased incidence of hyperferritinemia.^41,42 When found in patients with aHUS, these variants are FDA approved indications for the C5 inhibitor eculizumab.^13,14

Our study also identified individuals with multiple pathogenic variants, affecting different aspects of the inflammatory response. This heterogeneity may explain the observed absence of effect of immunomodulatory agents in sepsis, where clinical trials have applied agents without consideration of potential variation in underlying disease process, leading to overall negative results.⁴³

While all the variants reported here have been classified as pathogenic or likely pathogenic in CAPS, aHUS, and HLH/MAS, their identification in hyperferritinemic sepsis is of interest, but cannot be claimed as causal. Additionally, WES sequencing will miss pathogenic variants in regulatory and intronic gene regions that could be relevant to clinical phenotype. This has been shown in some individuals with HLH.⁴⁴ Further, the application of immunomodulatory therapies to septic individuals with these variants is of unclear benefit or harm. However, these findings provide evidence that screening select sepsis patients can identify unappreciated heritable disease, and could facilitate a genome-driven precision medicine.

Patients and Methods

Subjects were taken from the ProCESS multicenter trial cohort of protocolized resuscitation strategies in the emergency department as DNA was collected from the enrollees.¹⁸ Immunomodulatory therapy was not systematically studied in this trial. While the Shakoory study of interleukin-1 receptor blockade in sepsis patients with MAS features laid the groundwork for this study, no DNA samples were available from this initial study. Subsequently, DNA from individuals from the ProCESS trial who met the Shakoory et al. MAS definition underwent WES in the present study.¹¹ The trial was approved by Institutional Review Board at each enrolling institution. All patients or their legally authorized representatives provided written informed consent. Briefly, individuals were 18 years of age or older and enrolled based on 1.) clinical suspicion for septic shock and 2.) either refractory hypotension (SBP less than 90mmHg or vasopressor requirement following fluid challenge) or evidence of poor perfusion (lactate level > 4mmol per liter).¹⁸ We defined features of MAS as the combination of coagulation dysfunction (platelet count < 100K or INR > 1.5) plus hepatobiliary dysfunction (total bilirubin level > 1.2 mg/dL).¹³ Eighty two of the 1341 ProCESS patients met these criteria. While serum ferritin levels were not available a priori, the levels were measured retrospectively from banked serum with the highest level recorded. Among those meeting MAS criteria, the median serum ferritin was 601.9ng/ml (IQR 268.33–2013.45). From these we selected the six patients with the highest ferritin levels (range 7,259–55,314ng/mL) for whole exome sequencing. A recent study from the Hellenic Sepsis Study Group identified serum ferritin greater than 4420 ng/ml as a marker for mortality and increased inflammation as measured by levels of IL-18, INF-y and sCD163.⁴⁵ The individuals with the 6 highest ferritin level samples were selected as a proof of concept that pathogenic variants for immunologic disorders with overlapping phenotype could be identified among enriched populations.

DNA was extracted from whole blood samples using standard methods and WES performed at the University of Pittsburgh Genomic Research Core with the Ion Torrent Platform. Libraries were constructed using Ion Ampliseq Exome RDY (ThermoFisher) with target of 100X coverage per sample. FASTQ files were aligned to homo sapiens reference sequence hg19 to create VCF files. VCF files were analyzed in the Fabric Genomics Opal 5.2.2 software platform (Fabric Genomics Inc, CA, https://www.fabricgenomics.com) to identify missense, nonsense or frameshift mutations. Variants were filtered for minor allele frequency less than 5% in the ExAC ⁴⁶, 1000 Genome ⁴⁷ and NHLBI-ESP 6500 databases ⁴⁸. ExAC database frequencies are reported in Table 2. While no control group was sequenced, all identified variants are previously reported as pathogenic or likely pathogenic, and are rare in the general population.^49,50 Identified variants were restricted to candidates in an immune disorder panel to enhance relevance (Table 3). Each identified variant was evaluated in the HGMD professional database ⁵¹ with manual literature review. All identified variants were confirmed via Sanger sequencing. Specific primer pairs can be found in the supplementary table S1.

Table 3.

This table shows the gene panel examined in our study. All 6 subjects underwent whole exome sequencing. Identified variants were filtered for minor allele frequency less than 5% based on the 1000 Genomes, ExAC and NHLBI-ESP 6500 databases. Variants were then filtered for those genes in the panel of interest, that were previously reported as pathogenic or likely pathogenic in the corresponding immunologic disorder as reported in the HGMD professional database.

Disease Class	Disease	Genes
Primary Immunodeficiencies	Chronic Granulomatous Disease	CYBA, CYBB, NCF1, NCF2, NCF4
	WHIM Syndrome	CXCR4
	Bruton's Agammaglobulinemia	BTK
	Activated PI3K-Delta Syndrome	PIK3CD
	Common Variable Immunodeficiency	BLK, CD19, CD81, CR2, CTLA4, ICOS, IKZF1, IL21, IL21R, LRBA, IRF2BP2, MS4A1, NFKB1, NFKB2K, PIK3R1, PLCG2, PRKCD, RAC2, TNFRSF13B, TNFRSF13C, TNFSF12, VAV1
	SCID	ADA, AK2, CD3D, CD3E, DCLRE1C, FOXN1, IL2RG, IL7R, JAK3, NHEJ1, ORAI1, PNP, PTPRC, RAG1, RAG2, RMRP, STAT5B, STIM1, TBX1, ZAP70
	HLH	PRF1, UNC13D, AP3B1, BLOC1S6, CD27, ITK, LYST, RAB27A, SLC7A7, STX11, STXBP2, SH2D1A, XIAP
Lymphoproliferative Syndromes	ALPS	CASP10, CASP8, FADD, FAS, FASLG, KRAS, MAGT1, NRAS
Recurrent Fever Syndromes	Crypopyrin-Associated Periodic Syndrome	NLRP3
	Familial Mediterranean Fever	MEFV
Complement Coagulation Disorders	aHUS	PLG, C3, CD46, CBF, CFH, CFHR5, CFI, DGKE, THBD
	TTP	ADAMTS13
Disorders of Iron Handling	Hemochromatosis	FTH1, HFE, SLC40A1, TFR2
	Juvenile hemochromatosis	HAMP, HFE2