Myelodysplastic syndromes (MDS) occurring in Agent Orange exposed individuals carry a mutational spectrum similar to that of de novo MDS

Myelodysplastic syndromes (MDS) include a group of clonal hematopoietic disorders characterized by blood cytopenias, dysplastic hematopoietic differentiation, and a risk of progression to acute myeloid leukemia (AML) [1]. While most cases of MDS arise without a known antecedent exposure (“de novo MDS”), a variety of causative agents have been linked to the development of MDS (“therapy-related MDS”), including exposure to organic solvents such as benzene, ionizing radiation and cytotoxic chemotherapy [2].

Agent Orange (AO) was a phenoxy herbicide mixture used extensively by the United States military as a strategic defoliant, particularly during the Vietnam War, when it was commonly diluted in the field by military staff with kerosene, gasoline, or JP-4 jet fuel to facilitate aerosolization and spraying from aircraft [3]. During the manufacturing process, AO was contaminated with dioxins, highly toxic organic pollutants for which exposure to even low levels has been associated with increased risk for the development of a variety of cancers including prostate cancer, soft tissue sarcomas, multiple myeloma and non-Hodgkin lymphoma [4–6]. While no clear link between the development of MDS and exposure to AO has yet been documented [3], the long latency for development and relative rarity of MDS has made this difficult to assess in epidemiological studies, and the potential hematopoietic toxicity of AO has raised the possibility of an association [7]. Of note, a long latency between disease diagnosis and genotoxic exposure is still consistent with AO being a risk factor for MDS; the incidence of MDS continued to be increased in exposed Japanese populations compared to unexposed populations more than 60years following the 1945 Hiroshima and Nagasaki atomic events [8].

In order to assess whether MDS that developed in AO-exposed individuals has a common etiologic link we asked whether there might be a distinctive molecular signature in these patients. We performed next generation exome sequencing to assess acquired somatic mutations in a panel of 29 Vietnam-era military veterans diagnosed with MDS and prior reported exposure to AO that met the US Veterans Administration definition of presumptive AO exposure [9] (Table 1). After obtaining IRB-approval for the study we established a recruitment website in order to identify military veterans who carried a diagnosis of MDS and a self-reported history of exposure to AO meeting US Department of Veterans Affairs location-specific criteria (https://www.publichealth.va.gov/exposures/agentorange/locations/index.asp). Eligible veterans were asked to complete an online form and to provide a peripheral blood sample. The median age of the cohort at diagnosis was 65 years (range 59–77); all were white males, and most had served in the US Army. As an unexposed comparator group, we examined the exome sequences of nine patients with MDS treated at our institution without known military service or suspected AO exposure (Table 1) and also compared to a dataset of more than 3000 patients with MDS assessed with a 95-gene molecular profiling tool [10,11].

Table 1.

Patient characteristics for Agent Orange exposed and control subjects.

Sample #	Gender	Race	Branch of military	Age at diagnosis	MDS subtype	Prior chemo/RT
AO01	Male	White	Navy	61	MDS	No
AO02	Male	White	Army Reserve	59	MDS-U	No
AO03	Male	White	Army	67	RAEB-2	Yes
AO04	Male	White	Air Force	64	MDS-MLD	Yes
AO05	Male	White	Army	65	RAEB-1	Yes
AO06	Male	White	Air Force	77	MDS-U	No
AO07	Male	White	Air Force	61	MDS-MLD	No
AO08	Male	White	Army	61	MDS-U	No
AO09	Male	White	Army	62	MDS-U	Unknown
AO10	Male	White	Army	66	MDS	No
AO11	Male	White	Army	65	RARS	No
AO12	Male	White	Army	73	MDS-U	No
AO13	Male	White	Army	67	RAEB-2	Yes
AO14	Male	White	Army	76	RAEB-2	No
AO15	Male	White	Army	63	MDS-MLD	No
AO16	Male	White	Army	59	MDS-U	No
AO17	Male	White	Navy	64	MDS-U	Unknown
AO18	Male	White	Army	67	RAEB-1	Yes
AO19	Male	White	Army	64	MDS-U	No
AO20	Male	White	Army	68	MDS-U	No
AO21	Male	White	Army	64	MDS-U	No
AO22	Male	White	Army	73	MDS	No
AO23	Male	White	Air Force	68	MDS-MLD	No
AO24	Male	White	Air Force	65	MDS-U	No
AO25	Male	White	Navy	61	RARS	No
AO26	Male	White	Army	61	RARS	No
AO27	Male	White	Army	71	RARS	No
AO28	Male	White	Army	71	MDS-U	Yes
AO29	Male	White	Navy	65	MDS-U	No
AO30 (control)	Female	White	N/A	92	RAEB-2	No
AO31 (control)	Female	Hispanic	N/A	53	RARS	No
AO32 (control)	Female	White	N/A	48	RAEB-2	No
AO33 (control)	Female	White	N/A	66	RCMD	Yes
AO34 (control)	Male	White	N/A	39	MDS-U	No
AO35 (control)	Male	White	N/A	73	RAEB-1	No
AO36 (control)	Male	White	N/A	52	RAEB-1	No
AO37 (control)	Male	White	N/A	57	RAEB-1	No
AO38 (control)	Male	White	N/A	66	RAEB-2	No

RT: radiation therapy.

In order to identify somatic variants, genomic DNA was isolated from whole blood and library preparation was performed using hybrid capture (Agilent, version 5). Sequencing was performed on the Illumina platform at a mean target coverage of 90X and genomic alignment and variant calling were performed as previously described [12]. This led to the identification of 55 driver mutations in the 29 patients (range 0–5) (Figure 1(a)). The majority of these mutations were in genes involved in RNA splicing (SF3B1, U2AF1 and SRSF2) and epigenetic regulation (TET2, DNMT3A and ASXL1). Three of the four patients with a self-reported diagnosis of MDS with ring sideroblasts did carry a mutation in SF3B1. The driver mutations identified in the AO cohort were similar to that seen in our control cohort and to that seen in previously published sequencing studies of MDS (Figure 1(a)) as well as our control cohort, including restricting the control cohort to males only and those with a similar age spectrum to the AO exposed population [1,13,14]. Thus, the spectrum of driver mutations seen in MDS exposed to AO is similar to that seen in de novo MDS. We did not identify recurrent frameshift and nonsense mutations in genes not previously reported to drive MDS.

Figure 1.

The mutational spectrum is similar between Agent Orange exposed and control MDS patients. (a) Co-mutation plot for driver mutations identified in Agent Orange exposed and control subjects. Syn: synonymous. (b) Mutational signature analysis for subjects exposed to Agent Orange and controls. Colored chart signifies the specific single nucleotide variant and 4 × 4 grid denotes the sequence context. (c) Mutational burden per Mbp for Agent Orange exposed veterans compared to controls.

Dioxins have been proposed to promote cancer formation through stimulation of a variety of transcriptional pathways, as opposed to through direct DNA damage [15]. Thus, we hypothesized that AO-associated MDS would not be enriched for mutations within the DNA damage response pathway. We identified only two mutations in TP53 and no mutations in PPM1D, mutations in which can be found in up to one-third of patients with therapy-related myeloid neoplasms [13].

MDS is primarily a disease of the elderly (median age at diagnosis 71–76years) with the major source of driver mutations being single nucleotide variants believed to arise from spontaneous cytidine deamination leading to C to T transition mutations [2,16]. In order to determine whether AO exposure might cause a unique form of DNA damage or promote MDS through increased replication or other cellular stress we examined the mutational signatures within our AO cohort as compared to our control cohort (Figure 1(b)). We identified all somatic single nucleotide variants with a variant allele fraction between 2% and 35%, and categorized them by the type of nucleotide change and sequence context. The vast majority of mutations occur within CpG dinucleotides leading to C to T transitions consistent with the aging related mutational signature that has been previously reported in MDS and acute myeloid leukemia [16]. We saw no significant difference between the mutational signature in AO exposed individuals and our control cohort (cosine similarity score of 0.95), nor in the number of mutations per Mb of DNA (Figure 1(c)) (p = 0.14, Wilcoxonn rank sum test). This pattern of C to T transitions is consistent with the aging-associated mutational signature seen in other MDS sequencing reports [14].

In summary, here we report the first detailed genetic characterization of MDS in patients exposed to AO during the Vietnam War era. The landscape of driver mutations and mutational events detected in this population is similar to that seen in de novo MDS and suggests that if AO promotes the development of MDS, it does so through a mechanism that is not associated with increased DNA damage. While we did not identify an association between any recurrently mutated genes or a clear mutational signature and AO exposure, the limited size of this study does not exclude the possibility of an AO-induced carcinogenic process that promotes the development of MDS. It is also possible that dioxins, such as AO, help select for mutant hematopoietic stem cell clones and thus promote disease formation without any direct genotoxic effects; analysis of exposed Veterans in the 1960s and 1970s shortly after AO exposure would likely have been necessary to find evidence for such changes in clonal architecture. This study was also not sufficiently powered to examine a direct association between AO exposure and the subsequent development of MDS. Further epidemiological and molecular studies will be needed to determine whether MDS is among the multitude of diseases that afflict military veterans exposed to AO more commonly than the unexposed population.