Cohort Characteristics:
All protocols were conducted in compliance with the relevant guidelines and regulations, and approved by institutional review boards (IRB) at each institution. Informed consent was obtained from all participants. These secondary data analyses were largely facilitated through data sharing by the Boston Biorepository and Integrative Network (BBRAIN) for the GWI repository. All biological and clinical data utilized within this study were from the following case-control veteran cohorts: (1) the Roskamp Neurology Clinic (RNC) cohort (n = 69), (2) the Boston Gulf War Illness Consortium (GWIC) (n = 38), (3) the Fort Devens cohort (n = 57)[22], (4) the San Francisco Veterans Affairs Health Care System (SFVAHCS) cohort (n = 142) that consisted of consecutive Gulf War veteran recruitment from 2014 to 2018[25, 26], and (5) the Georgetown University cohort (n = 45)[27]. Among all cohorts, either the Centers for Disease Control (CDC) chronic multi-symptom illness definition or the Kansas GWI criteria determined GWI status[28, 29]. More specifically, the Kansas GWI criteria requires that GW veterans must show symptoms in at least three of six symptom domains (fatigue/sleep problems, somatic pain, neurological/cognitive/mood symptoms, gastrointestinal symptoms, respiratory symptoms, and skin abnormalities) to ascertain a GWI diagnosis, whereas the CDC case criteria requires that a veteran must exhibit at least two chronic symptoms that present longer than six months from the following categories: pain, fatigue, and mood/cognition. Controls were either veterans deployed to the 1990-91 GW or healthy sedentary civilian controls that did not meet the Kansas GWI criteria or any of the exclusionary criteria. Primarily, participants were excluded if they reported a confounding diagnosis in their medical history [30]. Self-report of pesticide exposure was based on field use of pesticides, including sprays, fogs, pest-strips, fly baits, and the personal use of flea collars and pesticide-sprayed uniforms. These variables were combined into a binary pesticide exposure questionnaire where reporting yes for any of the pesticide categories established pesticide exposure. Furthermore, self-reported exposure to chemical alarms, oil well fires, SCUD missiles, CARC paint, and the consumption of PB pills were also each coded as separate binary variables. Data on the sex of the participants were gathered via self-report as well, with the options being male and female. As per the National Institute of Health, sex refers to the biological differences between males and females, as previously described [30]. Similarly, self-reported information of symptoms pertaining to fatigue, cognitive problems, depression/depressed mood, and sleep disturbances was also collected and analyzed as a dichotomous binary variable (yes/no). Participant recruitment, blood collection, and blood storage procedures have been described previously [30]. All samples were collected using similarly written standard operating procedures for performing phlebotomy for blood collection.
Apoe Genotyping:
The Ft. Devens, GWIC, and RINC samples: Within the Ft. Devens and RINC locations, blood samples were drawn into EDTA tubes to prepare plasma and preserve the blood cells for DNA extraction. With the GWIC and Georgetown University samples, DNA from white blood cells was extracted using the Gentra Puregene Blood Kit (Qiagen) as per the manufacturer’s instructions (Qiagen). Genotyping of APOE using extracted DNA was performed using the EzWay direct ApoE genotyping kit (Cosmo Bio) as described by the providers. Amplified DNA fragments corresponding to different genotypes were separated by electrophoresis in Ethidium Bromide (EtBr) stained 2% agarose and 1% MetaPhor agarose gels. The presence of the ε4 allele identified with the EzWay direct ApoE genotyping kit was confirmed using polymerase chain reaction-restriction fragment-length polymorphism as follows: a section of the ApoE gene (218bp) was amplified by PCR using the following primers: 5′-TCCAAGGAG-GTGCAGGCGGCGCA-3′ (upstream) and 5′-GCCCCGGCCTGGTACACTGCCA − 3′ (downstream), ~ 250 ng of DNA, 5µL of Expand High Fidelity Buffer, with 15 mM MgCl2 10x concentrated (Sigma), 2.5 µL of each primer (10 µM), 1 µL of dNTP (10 mM), 5% dimethyl sulfoxide, 4 µL of MgCl2 (25 mM) and 0.75 µL of Expand High Fidelity Enzyme mix (Sigma) in a final volume of 50 µL. The DNA was denatured at 95°C for five minutes, followed by 35 cycles of denaturation (95°C, 1 minute), annealing (65°C, 1 minute), and extension (72°C, 1 minute), with a final extension at 72°C for four minutes. 15 µL of PCR product was used to confirm the presence of the expected band (218bp) by electrophoresis in EtBr-stained 2% agarose gels, while the remaining 35 µL of PCR product was precipitated by adding 3.5 µL of Sodium Acetate (3 M, pH 5.2) and 100 µL of 100% ethanol to 35 µL of amplified DNA. After incubating for 30 minutes at -80°C, samples were centrifuged at 14,000 rpm at 4°C for 10 minutes, and DNA-containing pellets were resuspended in nuclease-free water. Enzymatic digestion using 0.5 µL of Cfo-I restriction enzyme (10 µg/µL), 1.5 µL of restriction enzyme 10X buffer, 0.2 µL of acetylated BSA (10 µg/µL), and 13 µL of precipitated DNA was performed for 1.5 hours at 37°C to identify digestion profiles reflecting the presence of the ε4 allele. The restriction fragments corresponding to one or two copies of the ε4 allele were then separated by electrophoresis in 2% EtBr-stained and 1% MetaPhor agarose gels. For the SFVAHCS, DNA was isolated from saliva samples using Oragene kits (DNA Genotek, Ottawa, ON, Canada) in the laboratory of Dr. Joachim Hallmayer at Stanford University, as previously described [31]. For the Georgetown University cohort, APOE genotyping was performed using the TaqMan assay as previously described [32].
Statistical analyses
The Pearson χ2 statistics were used to compare demographic differences in race/ethnicity, exposure status, and sex, and determine significant differences in the Hardy-Weinberg equilibrium (HWE) for APOE genotypes. The age of the controls and veterans with GWI was examined using a t-test for independent samples. Binary logistic regression modeling was used to calculate the odds ratios (OR) and associated 95% confidence intervals (CI). To evaluate possible interactions between GW exposures and the ε4 allele status, stepwise forward likelihood ratio selection in logistic regression was used where the final step was retained in the model. Terms in this model included independent GW exposures and their interactions with PB and the presence of the ε4 allele, while still adjusting for age, sex, and race/ethnicity. A post-hoc power analysis of the data, conducted using the G-Power software, revealed a power of > 95% based on a sample of 351 individuals for the observed OR regarding the association between APOE and GWI diagnosis. Alpha levels were set at 0.05, and SPSS version 26 was used to analyze these data.