Clonal Hematopoiesis in Gulf War Veterans

Abstract

Veterans of the 1990–1991 Gulf War (GW) endured exposures from oil well fire smoke, pesticides, burn pits, depleted uranium, jet fuel, chemical warfare, and nerve agents [1]. The National Academies of Science, Engineering, and Medicine have reported sufficient data to associate GW deployment with chronic multisystem illness and post-traumatic stress disorder, but owing to latency there are insufficient data to associate GW deployment with hematologic malignancies [1]. Given the difficulty attributing adverse health outcomes to wartime exposures, the Veterans Health Administration (VHA) expanded healthcare to veterans with presumptive conditions following burn pit exposure, and myeloid neoplasms (MN) were recently added as a presumptive condition. Clonal hematopoiesis (CH) is a precursor to MN and may be an indicator of clinically significant toxin exposure. Exposures to chemotherapy, cigarette smoke, and ionizing radiotherapy are known to heighten the risk of CH of indeterminate potential (CHIP) [2-5], and data suggest that toxic exposures to World Trade Center particulate matter after 9/11 can also increase CHIP [6]. Therefore, there is rationale that GW exposures could also increase the risk of CH. Unlike Agent Orange-exposed Vietnam veterans and atomic veterans who have already declared their long-term comorbidities [7], GW veterans are at an age that CH could be detected but clinical comorbidities including MN may not have begun to manifest.

A pilot study was performed to describe and characterize CH in a group of exposed and unexposed GW veterans enrolled in the VHA Gulf War Era Cohort and Biorepository (GWECB) [8]. At about 25 years post exposure between 2014 and 2016, GWECB collected a blood sample and administered a survey that included demographics, military service history, exposure to pesticides, oil well fire smoke, and subsequent deployments to Operation Enduring Freedom (OEF) and Operation Iraqi Freedom (OIF). Exposure was reported in durations of days (1–6, 7–30, or ≥ 31 days), and was characterized in three levels: level 1 (≥ 31 days exposure to oil well fire smoke, < 31 days exposure to pesticides, and subsequent OEF/OIF deployment), level 2 (< 31 days exposure to oil well fire smoke, ≥ 31 days exposure to pesticides and subsequent OEF/OIF deployment), and level 3 (≥ 31 days exposure to oil well fire smoke and pesticides during the GW but no OEF/OIF deployment). The exposed cohort included 30 male veterans who deployed to the GW and reported levels 1, 2, or 3 exposures. The unexposed cohort included 30 male veterans who served during the GW era but did not deploy. All veterans were ages 55–65 at the time of blood collection, and veterans in each cohort were matched by year of birth and smoking history.

Genomic DNA was used as input at 50 ng into the Archer VariantPlex Myeloid library preparation workflow following the manufacturer's protocol with minor modifications for semi-automated liquid handling. Amplification for PCR1 and PCR2 steps were 16 and 20 cycles, respectively. Libraries were assessed for absence of amplification-based artifacts by size distribution using the Agilent Fragment Analyzer and quantitated by qPCR using the Roche Light Cycler II 480. Libraries were normalized into 62-plex pools and sequenced on an Illumina NovaSeq 6000 using a S4 300 cycle kit, PhiX spike-in and the XP workflow. Run parameters were 151 + 8 + 8 + 151 targeting approximately 50 million reads per library. Sequencing data were aligned and variants called and annotated using Archer Analysis 6.0 software. Resulting variants were reduced to protein coding variants and were filtered to remove low- quality and/or germline variants. Variants were restricted to those with a population allele frequency less than 1%, depth greater than 250, no sequence direction bias, no sequence strand bias, no annotation of known germline or likely non-somatic, and no occurring in a sequenced control sample. CH was defined as the presence of a somatic mutation in a candidate gene with a VAF ≥ 0.5%. All statistics are descriptive. Thirty samples per cohort was selected extrapolating from The Nurse's Health Study, which demonstrated measurable CH in a sample size of 20 participants [9]. This study was exempt from full IRB review.

Median age at the time of blood collection was 59 years (range 55–65). Twenty-one of the 60 veterans (35%) were black (10 exposed and 11 unexposed), and the remainder were white. Current or prior smokers comprised about half of each cohort. Of the exposed cohort, 11, 10, and 9 veterans had level 1, 2, and 3 exposure, respectively. Later deployment to OEF/OIF occurred in 21/30 exposed and 5/30 unexposed veterans. Active-duty service was reported in 16/30 exposed and 25/30 unexposed veterans, reserve duty was reported in 5/30 exposed and 1/30 unexposed veterans, and both active-duty and reserve duty were reported in 9 exposed and 4 unexposed veterans, respectively.

The landscape of mutated genes, type of mutations, VAF of each mutation, and number of times a gene was mutated in the two cohorts is shown in the Figure 1, and a case listing of veterans with mutations is shown in the Table 1. There were 12 mutations in 10 exposed veterans and 28 mutations in 16 unexposed veterans and for an overall prevalence of CH of 43% (26/60). The median VAF of the group was low at 0.795% (0.52–10.73), of the exposed cohort 0.775% (0.53–10.73), and of the unexposed cohort 0.825% (0.52–5.54). Consistent with previously published data, DNMT3A and TET2 were the most frequently mutated genes, and missense mutations were the most common mutation type [3, 4]. Most base pair substitutions were A > G (7/31) and 4/31 were C > T transitions associated with age-related CH, possibly a reflection of the young age of the cohort. Mutations in genes high-risk for MN were present in three veterans, one exposed and two unexposed. These mutations included a missense mutation in TP53, VAF 0.53%, in the exposed group and an in-frame deletion in JAK2, VAF 1.92%, and frameshift in PPM1D, VAF 0.53%, in the unexposed group.

FIGURE 1

Open in figure viewerPowerPoint

Oncoplot of driver mutations identified in GW veterans by exposure status.

TABLE 1. Case listing of exposed and unexposed veterans with CH mutations.

Exposed cohort
Veteran ID	Age	Race	Smoking status	Active-duty vs. Reserves	Exposure level	OEF/OIF	Gene	Variant	Mutation type	Base substitution	DP	RO	AO	VAF %
E1	62	White	Never	AD and reserves	1	Yes	DNMT3A	NM_022552.4:c.2645G > A (p.Arg882His)	Missense	G > A	7272	6492	780	10.73
E2	62	White	Prior	AD and reserves	1	Yes	DNMT3A	NM_022552.4:c.2195 T > G (Phe732Cys)	Missense	T > G	6517	6255	254	3.90
							DNMT3A	NM_022552.4:c.1430-2A > G	Splice acceptor	A > G	5306	5260	45	0.85
E3	61	White	Prior	AD	1	Yes	DNMT3A	NM_022552.4:c.2204A > G (p.Tyr735Cys)	Missense	A > G	9235	9091	141	1.53
E4	61	Black	Current	AD and reserves	1	Yes	DNMT3A	NM_022552.4:c.1671C > A (p.Cys557Ter)	Stop gained	C > A	5630	5547	82	1.46
							RAD21	NM_006265.2:c.1580del (p.Kts527ArgfsTer85)	Frameshiift	n/a	4119	4083	23	0.56
E5	56	White	Current	AD and reserves	2	Yes	DNMT3A	NM_022552.4:c.1900A > T (p.Ile634Phe)	Missense	A > T	10 211	10 132	76	0.74
E6	57	Black	Prior	AD	2	Yes	NF1	NM_000267.3:c.460 T > C (p.PHe154Leu)	Missense	T > C	1031	1024	7	0.68
E7	61	White	Never	AD	3	No	TP53	NM_000546.5:c.578A > G (p.His193Arg)	Missense	A > G	5299	5270	28	0.53
E8	58	White	Never	AD	1	Yes	DNMT3A	NM_022552.4:c.2120G > A (p.Gly707Asp)	Missense	G > A	8914	8853	60	0.67
E9	61	Black	Prior	Reserves	2	Yes	DNMT3A	NM_022552.4:c.2320G > A (p.Glu774Lys)	Missense	G > A	7295	7236	59	0.81
E10	59	White	Never	AD	1	Yes	DNMT3A	NM_022552.4:c.889 T > G (p.Trp297Gly)	Missense	T > G	7680	7618	46	0.60

Unexposed cohort
Veteran ID	Age	Race	Smoking status	Active-duty vs. reserves	Exposure level	OEF/OIF	Gene	Variant	Mutation type	Base substitution	DP	RO	AO	VAF %
U1	61	Black	Never	AD	N/A	No	DNMT3A	NM_022552.4:c.1430-2A > G	Splice acceptor	A > G	7171	6772	397	5.54
							DNMT3A	NM_022552.4:c.1628G > A (p.Gly543Asp)	Missense	G > A	9894	9820	74	0.75
							DNMT3A	NM_022552.4:c.2389A > C (p.Asn797His)	Missense	A > C	12 489	11 831	652	5.22
U2	63	White	Current	AD	N/A	No	TET2	NM_001127208.2:c.2462del (Gln821ArgfsTer3)	Frameshiift	n/a	14 101	13 391	686	4.86
							DNMT3A	NM_022552.4:c.1596del (Tyr533ThrfsTer118)	Frameshiift	n/a	9248	9155	81	0.88
							JAK2	NM_004972.3:c.1624_1629del (p.Asn542_Glu543del)	Inframe deletion	n/a	5482	5347	105	1.92
							TET2	NM_001127208.2:c.3803 + 5G > A	Splice region	G > A	6555	6438	116	1.77
U3	63	White	Never	AD and Reserves	N/A	No	LUC7L2	NM_016019.4:c.550C > T (p.Gln184Ter)	Stop gained	C > T	4621	4594	27	0.58
							DNMT3A	NM_022552.4:c.1913C > T (Ser638Phe)	Missense	C > T	12 000	11 773	227	1.89
U4	62	Black	Prior	AD	N/A	No	TET2	NM_001127208.2:c.4045-2A > G	Splice acceptor	A > G	3937	3910	21	0.53
							DNMT3A	NM_022552.4:c.2026C > T (p.Arg676Trp)	Missense	C > T	9899	9758	140	1.41
U5	63	White	Never	AD	N/A	Yes	DNMT3A	NM_022552.4:c.2544del (Pro849LeufsTer4)	Frameshiift	n/a	13 648	13 463	157	1.15
							GNAS	NM_000516.4:c.601C > A (Arg201Ser)	Missense	C > A	4604	4567	32	0.70
U6	60	White	Prior	AD	N/A	No	DNMT3A	NM_022552.4:c.1296C > G (Tyr432Ter)	Stop gained	C > G	8513	8465	46	0.54
							DNMT3A	NM_022552.4:c.2204A > G (p.Tyr735Cys)	Missense	A > G	10 217	10 107	110	1.08
U7	59	White	Current	AD	N/A	No	SMC3	NM_005445.3:c.2005 T > G (p.Tyr669Asp)	Missense	T > G	10 195	10 083	99	1.00
							DNMT3A	NM_022552.4:c.1015-3C > G	Splice region	C > G	6628	6551	75	1.13
U8	55	White	Current	AD	N/A	No	PPM1D	NM_003620.3:c.1643del (p.Lys548ArgfsTer8)	Frameshiift	n/a	8256	8196	44	0.53
							DNMT3A	NM_022552.4:c.2146G > T (p.Val716Phe)	Missense	G > T	8993	8941	48	0.53
U9	55	Black	Never	AD and reserves	N/A	Yes	DNMT3A	NM_022552.4:c.1723G > C (p.Ala575Pro)	Missense	G > C	8877	8824	52	0.59
U10	61	White	Never	AD	N/A	No	TET2	NM_001127208.2:c.4241A > G (p.Gln1414Arg)	Missense	A > G	4663	4627	35	0.75
U11	62	White	Never	AD	N/A	No	GNAS	NM_000516.5:c.602G > A (pArg201His)	Missense	G > A	5676	5632	44	0.78
							NF1	NM_000267.3:c.2133_2141del (p.His712_Cys714del)	Inframe deletion	n/a	5355	5275	41	0.77
U12	57	White	Never	AD	N/A	No	BCORL1	NM_001379451.1:c.3551A > C (p.Lys1184Thr)	Missense	A > C	2863	2845	15	0.52
U13	60	Black	Current	Reserves	N/A	No	DNMT3A	NM_022552.4:c.2711C > T (p.Pro904Leu)	Missense	C > T	5318	5285	30	0.56
U14	62	Black	Prior	AD	N/A	No	SH2B3	NM_005475.2:c.926 + 1G > T	Splice donor	G > T	4187	4144	40	0.96
U15	57	White	Prior	AD	N/A	No	DNMT3A	NM_022552.4:c.2645G > A (p.Arg882His)	Missense	G > A	8452	8400	51	0.60
U16	56	Black	Never	AD	N/A	No	DNMT3A	NM_022552.4:c.1578C > G (p.Tyr526Ter)	Stop gained	C > G	10 403	10 309	91	0.87

Five mutations had a VAF ≥ 2% and were classified as CHIP. Two veterans in each cohort had CHIP for an overall prevalence of 6.7% (4/60), which is similar to the 5.6% rate demonstrated by Jaiswal et al. [3] using whole exome sequencing and is comparable to the 7% rate demonstrated by Vlasschaert et al. [10] in the All of Us whole genome data set for ages 60–69. The two CHIP mutations in the exposed cohort occurred in two veterans, both with level 1 exposure, both white, one a never smoker and one a prior smoker, both active-duty and reserves, and both of whom deployed to OEF/OIF. One had a missense mutation in DNMT3A, p.Arg882His, VAF 10.73%, and the other had a missense mutation in DNMT3A, VAF 3.9% along with an additional second splice acceptor mutation in DNMT3A, VAF 0.85%. The three CHIP mutations in the unexposed cohort occurred in two veterans, one black, one white, one a never smoker and one a current smoker, both active-duty and neither deployed to OEF/OIF. One had a DNMT3A splice acceptor mutation, VAF of 5.54%, along with another missense mutation in DNMT3A, VAF 0.75%. The other had two CHIP mutations, one a missense DNMT3A, VAF 5.22%, and a frameshift in TET2, VAF 4.86%, along with two additional mutations, one in-frame deletion in JAK2, VAF 1.92%, and the second a frameshift in DNMT3A, VAF 0.88%.

The median age of exposed veterans with and without mutations was 61 and 59 years, respectively. Among the unexposed veterans, the median age with and without mutations was 60.5 and 58.5 years, respectively. Among exposed GW veterans, CH was prevalent among veterans with subsequent OEF/OIF deployment, particularly among exposed veterans who reported high levels of oil well fire smoke exposure. CH was present in 6/11 veterans with level 1 exposure, compared to 3/10 with level 2 exposure and 1/9 with level 3 exposure. Subsequent OEF/OIF deployment was reported in 9/10 exposed veterans with mutations, in comparison to 12/20 exposed veterans without mutations, 2/16 unexposed veterans with mutations, and 3/14 unexposed veterans without mutations. Both exposed veterans with CHIP reported level 1 exposure, which is ≥ 31 days of exposure to oil well fire smoke plus subsequent OEF/OIF deployment. It is possible that cumulative exposure to inhaled carcinogens in theater, such as from oil well fire smoke and burn pit exposure, contributed to these observations. Because this study was not powered to make comparisons between cohorts or by OEF/OIF deployment, these observations warrant further investigation.

There are limitations to this study. We had a single point in time with incomplete medical records preventing calculation of CHIP risk scores [11] or characterization of the dynamics of low-VAF CH needed to put our findings into clinical context [2]. In veterans who have endured toxic exposures, it would be important to understand the clinical consequences and clonal dynamics of low-VAF mutations in high-risk genes such as TP53, PPM1D, and JAK2, especially since low-VAF TP53 mutations predating other toxic exposures (chemotherapy, radiotherapy) can expand under selective pressure and lead to therapy-related MN [2, 12-14]. Additionally, in our study, more exposed veterans were reservists and unexposed active-duty. Active-duty are full-time the military, whereas reservists are largely civilians who are activated for deployments and otherwise work 2 weeks per year in the military. We suspect some active-duty veterans were activated reservists. Therefore, future studies should delineate duty status a priori.

The VHA is the largest integrated healthcare system in the U.S. [15] and could support large-scale studies needed to further investigate the impact of deployment and toxic exposures during war on clinically meaningful endpoints such as high-risk CHIP, as well as the clonal dynamics and clinical significance of low-VAF CH in veterans.