Genetic Associations with Smoking Relapse and Proportion of Follow-up in Smoking Relapse throughout Adulthood in Pre- and Postmenopausal Women

Cancer is the second leading cause of death in U.S. adults and cigarette smoking causes 52.3% and 43.6% of all cancer deaths annually among U.S. men and women, respectively (1). Smoking not only increases cancer risk, it is associated with first-line cancer treatment failure, secondary metastasis, and decreased cancer survival rates (2). Risk of smoking-caused morbidity and mortality begin to decrease upon smoking cessation (3) with some rapid health benefits of quitting smoking, such as improved pulmonary function. Nonetheless, several years of smoking abstinence are required to experience many significant health benefits of smoking cessation. For example, cancer risk decreases by 30% to 50% after 10 years of abstinence (1, 2, 4). Despite the considerable benefits of long-term smoking cessation, sustained smoking abstinence is difficult to achieve, with less than 3% of quit attempts resulting in permanent cessation (5).

Smoking abstinence is typically unstable over multiple and episodic attempts to quit smoking, especially during early periods of a quit attempt. Relapse rates exceed 50% throughout the first year after cessation, dropping to 36% and 25%, respectively, after 2 and 5 fully abstinent years (6). Smokers regularly have several unsuccessful quit attempts before, if ever, permanently quitting smoking (7). Understanding the risk factors associated with relapse, including late relapse one or more years after quitting, is important for promoting sustained abstinence and reducing the smoking-caused burden on public health. However, these factors cannot be fully understood from studies with follow-up of 1 year or less. Despite this, prior studies have typically evaluated prolonged smoking cessation by measuring smoking abstinence 6 to 12 months after quit date. Thus, long-term studies are needed to identify which smokers have the highest risk for smoking relapse throughout their lifetime as they could inform the development of long-term interventions designed to prevent relapse, modeled after chronic disease management approaches.

Twin studies indicate that nicotine dependence, withdrawal symptoms, and smoking cessation are between 31% and 60% heritable (8–11), suggesting that research to identify genetic factors associated with smoking relapse holds promise for informing precision medicine approaches for smoking cessation. However, genetic research on this topic is still in its nascent stages and studies have typically been limited by follow-up less than 1 year. Despite these limitations, genes influencing nicotinic acetylcholine receptors (nAChR) and properties of the mesolimbic dopaminergic (DA) system are promising candidates for associations with smoking relapse due to their respective contributions to both nicotine response and cognitive-behavioral phenotypes such as associative learning, cue-primed behavior (12), and inhibitory control (13–16).

In addition, sex differences in smoking cessation and relapse have been identified (17, 18) with evidence of heterogeneity in mechanisms of nicotine addiction by biological sex. For example, women are more responsive to sensory stimuli of cigarette smoking and environmental cues (19). Furthermore, hormonal changes due to menstrual phase and menopause impact the level of craving, withdrawal, and negative affect experienced during a quit attempt (20–22).

Given these unique factors influencing smoking behavior among women, characterizing genetic susceptibility to smoking relapse specifically among women is essential. Yet, the few genetic studies focused on women have commonly been on smoking during pregnancy (23, 24). Thus, prior studies have had limited capacity to ascertain associations that may be unique to women. The aim of this study was to examine genetic associations with smoking relapse throughout adulthood within two all-female cohort studies with decades of follow-up. The study goal was to determine genetic associations with the amount of relapse among women who previously demonstrated the ability to successfully quit smoking to generate information unique to women that could be clinically useful for identifying smokers who may benefit from long-term relapse prevention following cessation. Furthermore, we assessed the strength of genetic associations by factors unique to women, such as menopausal status.

The Nurses’ Health Study (NHS; N = 121,700) and NHS-2 (N = 116,429) are prospective cohorts of female registered nurses with data collected every 2 years beginning in 1976 for NHS and 1989 for NHS-2 (25). Consented whole-genome genotype data were available for 10,049 and 2,804 former or current smokers (at baseline) in the NHS and NHS-2 cohorts, respectively. These data were previously collected and included genotype data that met quality control criteria during prior data review (26, 27). Of those baseline ever-smokers with genotype data, there were 32 and 16 women in the NHS and NHS-2 cohorts, respectively, that were of non-European ancestry; these women were excluded from the analysis to avoid confounding due to population stratification. We further excluded 505 and 171 participants from the NHS and NHS-2 cohorts, respectively, because they failed to quit smoking during follow-up and were not at risk for relapse. Finally, 63 women in NHS and 11 women in NHS-2 did not have follow-up observations following their initial report of smoking cessation; these participants were excluded from the analysis because it was not feasible to observe relapse. The final analysis population included 9,449 women from the NHS cohort and 2,611 women from the NHS-2 cohort for a total study population of 12,060.

Seven SNPs within nAChR genes and three SNPs within dopaminergic genes were selected on the basis of existing evidence of potential associations with either short-term smoking abstinence (8 to 24 weeks after cessation) or cross-sectional smoking status (28). Six of the nAChR SNPs were located within the CHRNA5-A3-B4 gene complex, encoding for the α-5, α-3, and β-4 nAChR subunits, and the seventh nAChR SNP selected was within CHRNB2, encoding for the β-2 nAChR subunit. The specific nAChR SNPs identified were CHRNA5 SNPs rs16969968 G>A (28, 29), rs680244 T>C (29), rs588765 T>C (30, 31), CHRNA3 SNPs rs1051730 G>A (23, 31) and rs578776 G>A (30), CHRNB4 SNP rs12914008 G>A (32), and CHRNB2 SNP rs2072661 G>A (33). The three dopaminergic variants identified were SNPs rs1800497 G>A (34) and rs6277 G>A (35) within DRD2, encoding for D2 dopamine receptors, and SNP rs4680 G>A (36, 37) within COMT, encoding for the catechol-o-methyltransferase enzyme that metabolizes dopamine. Two pairs of SNPs were in linkage disequilibrium (e.g., highly correlated in inheritance; D’ = 1.00, R² = 1.00): CHRNA5 rs16969968 with CHRNA3 rs1051730 and CHRNA5 rs588765 with CHRNA5 rs680244. All four were considered; however, SNPs in linkage disequilibrium are expected to have similar results.

The independent variables were the selected SNPs. Genetic associations were tested using the dominant, recessive, and additive genetic models for all SNPs except CHRNB4 rs12914008, which has a minor allele frequency (MAF) less than 0.05 making it feasible to evaluate only the dominant model.

The dependent variable was the proportion of timepoints returned to smoking (relapsed) throughout adulthood to quantify the amount of follow-up relapsed after an initial quit. For simplicity, we term this “proportion of follow-up in relapse.” We calculated proportion of follow-up in relapse as the number of observed timepoints in which the participant reported current smoking status divided by the total number of observed timepoints after the first-reported former smoker status. For current smokers at baseline, observations were considered after the first transition to former smoker during follow-up. For example, the three NHS participants represented in Fig. 1 were current smokers at baseline, first reported former smoker status in 1980, and spent 0%, 100%, and 24% of the follow-up in relapse, respectively. For former smokers at baseline, time since quit smoking for the current period of abstinence was accounted for in the calculation under the assumption that the current abstinence period was also the first quit period for these participants because the analyses could only account for time under observation. There were a total of 162,182 and 33,402 follow-up observations within the NHS and NHS-2, respectively. Of these, 11,267 and 356 observations were missing smoking status; however, after comparing with death records, all but 102 of those missing smoking status within NHS and 2 of those missing smoking status within NHS-2 were attributable to the participant's loss to follow-up attributable to death. We considered the remaining 104 observations missing smoking status as missing data and these observations were not utilized in the dependent variable calculations.

Additional variables considered included smoking intensity, age first reported quit smoking, number of follow-up observations, cohort (NHS or NHS-2), menopausal status, and body mass index (kg/m²); the latter two were measured at the time of the first report of smoking cessation. Participants were classified as either lifetime light smokers [<5 cigarettes per day (CPD)] or moderate to heavy smokers (≥5 CPD). Age quit smoking did not satisfy linearity in the logit, thus age quit was classified in categories of ≤20, 21 to 30, 31 to 40, 41 to 50, 51 to 60, 61 to 70, and ≥71 years of age. The prevalence of missing data at the timepoint first indicating smoking cessation was 0.7% for BMI and 1.3% for menopausal status. These values were imputed using the data values the participant reported for BMI and menopausal status at a data collection timepoint proximal to the timepoint with missing data.

Descriptive statistics were calculated for the NHS and NHS-2 cohort studies and the combined cohort study population. We modeled the proportion of follow-up in relapse, a continuous proportion from 0 to 1, with a β regression model using a logit transformation of the proportion of follow-up in relapse. Because the proportion of follow-up in relapse was zero-inflated, we implemented a zero-inflated β regression model with the SAS Beta Regression macro (38). This joint model evaluated (i) SNP genotype associations with the log odds of having zero observations relapsed during follow-up (zero proportion of follow-up in relapse), and (ii) conditional on relapse, SNP genotype associations with the proportion of follow-up in relapse. We also stratified the analysis by menopausal status at cessation to determine differences in genetic associations with relapse in women prior to and postmenopause. Associations are reported based on a P-value threshold of 0.05. However, we also controlled for multiple testing of 10 SNPs across several genetic models (28 tests) using a FDR of 0.05 and note those associations that are statistically significant after adjusting for multiple testing. Model assumptions and fit were checked graphically.

Dissemination of data used for this manuscript will be shared per the NHS/NHS-2 Data Availability Policy. Access to the NHS data are available by request for access using the procedures described at https://www.nurseshealthstudy.org/researchers.

Descriptive statistics for the NHS and NHS-2 cohorts and the combined cohorts are shown in Table 1. A higher proportion of women in NHS-2 (24.9%) relapsed to smoking during follow-up compared with women in NHS (20.0%; P-value < 0.001). Women who relapsed on average spent 33.7% of the follow-up returned to smoking; the proportion of follow-up returned to smoking on average was greater among women in NHS-2 (39.8%) than among women in NHS (31.7%; P-value < 0.001). Consistent with the increasing prevalence of light smokers among U.S. adult smokers in general (39), the proportion of light smokers was higher in the more recent NHS-2 cohort (18.0%) compared with the NHS (6.1%; P-value < 0.001). On average, women in NHS-2 were 11 years younger than women in NHS and were less likely to be postmenopause at the time of smoking cessation (P-value for both < 0.001). The mean BMI at smoking cessation was slightly higher in NHS-2 than NHS (24.0 kg/m² vs. 23.7 kg/m², P-value = 0.005).

Prior to considering genetic associations, the associations of smoking intensity and menopausal status at cessation with the likelihood of relapse were estimated and, among women who relapsed, the proportion of follow-up in relapse. Compared with light smokers, moderate to heavy smokers had 1.45 (95% CI, 1.20–1.75) times the odds of at least one relapse episode during follow-up. Conditional on relapse, moderate to heavy smokers also spent a significantly higher proportion of follow-up time returned to smoking than light smokers, specifically 11.6% (95% CI, 7.5%–15.7%) and 11.0% (95% CI, 7.2%–14.9%) more time returned to smoking among those who quit smoking pre- and postmenopause, respectively.

The minor alleles of 2 of the 10 SNPs evaluated, CHRNA5 SNP rs16969968 G>A and CHRNA3 SNP rs1051730 G>A, were associated with lower likelihood of relapse throughout the follow-up. Compared with women with heterozygous or homozygous common allele genotypes, sustained abstinence (zero relapse) throughout adulthood was more likely among women with two copies of the minor allele of rs16969968 (OR, 1.15; 95% CI, 1.00–1.33) and rs1051730 (OR, 1.16; 95% CI, 1.01–1.34). The associations between rs16969968 and rs1051730 and the likelihood of relapse were not statistically significant after adjusting for multiple testing (FDR > 0.05). None of the other eight SNPs were associated with likelihood of relapse during follow-up.

Conditional on relapse to smoking, four SNPs were significantly associated with the proportion of follow-up in relapse. Women with two copies of the minor allele of CHRNA5 SNP rs588765 G>A had 2.6% (95% CI, 0.1%–5.0%) more follow-up and those with two copies of the minor allele of CHRNA5 SNP rs680244 G>A had 2.4% (95% CI, 0.2%–4.9%) more follow-up in relapse than women with heterozygous or homozygous common allele genotypes. Similarly, women with one or two copies of the minor allele of DRD2 SNP rs6277 G>A had 3.3% (95% CI, 0.8%–5.7%) more follow-up in relapse than women without the minor allele. In contrast, those with two copies of the minor allele of COMT SNP rs4680 G>A had 2.7% (95% CI, 0.2%–5.2%) lower proportion of follow-up in relapse than women with heterozygous or homozygous common allele genotypes. None of these associations remained statistically significant after adjusting for multiple testing using FDR <0.05.

The associations of SNPs rs16969968 and rs1051730 with the likelihood of sustained abstinence (zero relapse) were concentrated among women who quit smoking premenopause [rs16969968 OR, 1.17 (95% CI, 1.00–1.37), rs1051730 OR, 1.19 (95% CI, 1.02–1.38)], and were absent among women who quit postmenopause [rs16969968 OR, 1.01 (95% CI, 0.71–1.43); rs1051730 OR, 1.00 (95% CI, 0.70–1.41); genotype by menopause interaction P-value = 0.65 for both SNPs]. Associations were not statistically significant after correcting for multiple testing.

Conditional on relapse, the association of SNPs rs588765 and rs680244 with a higher proportion of follow-up in relapse was significantly stronger among women who quit smoking postmenopause [rs588765 difference in proportion of follow-up in relapse (risk difference) = 9.5% (95% CI, 3.8%–15.2%), rs680244 risk difference = 10.1% (95% CI, 4.4%–15.9%)]. These associations remained statistically significant after correction for multiple testing with FDR corrected P-values of 0.014. No difference in proportion of follow-up in relapse by genotype was observed among women who had quit smoking prior to menopause [rs588765 risk difference = 0.8% (95% CI, −1.6%–3.3%), P-value of genotype by menopause interaction = 0.003; rs680244 risk difference = 0.6% (95% CI, −1.9%–3.0%), P-value of genotype by menopause interaction = 0.001]. Similarly, the genetic association of SNP rs6277 with more follow-up in relapse was greater among women who reported quitting smoking postmenopause [risk difference = 7.6% (95% CI, 1.5%–13.6%)] and was not statistically significant among women quitting pre-menopause (risk difference = 2.1%; 95% CI, −0.3%–4.5%), P-value of genotype by menopause interaction = 0.04]. The association of rs6277 with proportion of follow-up in relapse among post-menopausal women was not statistically significant after adjustment for multiple testing.

This study is the first to examine genetic associations in women with the likelihood of smoking relapse and the proportion of follow-up in relapse in a long-term prospective cohort. To our knowledge, this is also the first study to investigate differences in genetic associations by menopausal status by stratifying analysis by menopausal status at the first report of smoking cessation. None of the SNPs were associated with both the likelihood of relapse and, conditional on relapse, the proportion of follow-up in relapse. However, associations were observed with either likelihood of relapse or proportion of follow-up in relapse for four SNPs within nAChR genes and two within genes influencing dopaminergic function. Associations for several of these SNPs were significantly stronger among women who first reported quitting smoking postmenopause.

The four SNPs within nAChR genes comprised two pairs of SNPs in high linkage disequilibrium (correlation in inheritance): SNP rs16969968 with rs1051730, and SNP rs588765 with rs680244. The first pair, rs16969968 and rs1051730, were associated with sustained abstinence (zero relapse) during follow-up but were not associated with the proportion follow-up in relapse. This result is surprising given consistent prior evidence that these SNPs are associated with increased nicotine dependence, smoking more CPD, and lower likelihood of smoking cessation (40). Smokers with these alleles may be more nicotine-dependent increasing cessation difficulty. However, people with these alleles may also be more responsive to varenicline pharmacotherapy interventions (41). Thus, gene-by-pharmacotherapy interactions could have plausibly confounded these associations in this study. Furthermore, our findings should be viewed cautiously as the associations were weak and no longer statistically significant after correcting for multiple testing. Nonetheless, investigating genetic associations with various smoking outcomes (e.g., nicotine dependence, cessation, and relapse), especially in studies with long-term follow-up, has merit as the same allele may distinctly impact different smoking outcomes.

In contrast to SNPs rs16969968 and rs1051730, CHRNA5 SNPs rs588765 and rs680244 were not associated with sustained abstinence throughout follow-up but, conditional on relapse, were associated with the proportion of follow-up in relapse. Animal models suggest the α-5 nAChR subunit, encoded by CHRNA5, may partially mediate the effect of nicotine on mood and anxiety (42). Variation in rs588765 and rs680244 could possibly contribute to the mediating effects on mood leading to increased response among women with these minor alleles. Women are more likely to report smoking to improve negative mood (19) thus, women with these alleles may be susceptible to prolonged duration of smoking after relapse due to mediation of negative mood.

Two additional SNPs, both influencing dopaminergic function, were not associated with sustained abstinence throughout adulthood but, if relapse occurred, were associated with the proportion of follow-up in relapse. The first is SNP rs4680 within COMT, the gene that encodes for catechol-o-methyltransferase (COMT), an enzyme that metabolizes dopamine. Each copy of the minor allele of rs4680 functionally decreases COMT enzyme activity resulting in higher dopamine levels in the prefrontal cortex (43). Prefrontal cortex dopamine is important for cognitive function, and people with the minor allele of rs4680 have been found to have improved goal-directed behavioral control (44). This could explain why women with the homozygous minor allele genotype of SNP rs4680 had a significantly lower proportion of follow-up in relapse. Higher dopamine levels, presumably from lower COMT activity in women with this genotype, could also minimize the nicotine reward experienced through nicotine-stimulated dopamine release during relapse episodes, reducing the incentive to continue smoking.

Several prior studies found no association between COMT rs4680 genotype and smoking cessation, but most of these studies did not consider men and women separately thus raising confounding by sex as a limitation (36, 45). Sex differences in COMT enzyme activity, possibly due to estrogen downregulating activity, have been identified, as have differences in COMT associations with psychiatric disorders (43). A few prior studies have considered women specifically. Among these, Colilla and colleagues found increased smoking abstinence among women with the minor allele of SNP rs4680, with the strongest association observed among those with two copies of the allele (46). A study of women during pregnancy found no association between the minor allele and quitting smoking during pregnancy (24). In contrast, a study of older smokers (at least 55 years of age at recruitment) found that both men and women with the minor allele were less likely to quit smoking (47). These heterogeneous results could be due to the differences in circulating estrogen concentrations according to stage of life, as higher estrogen concentrations downregulate COMT activity and estrogen concentrations increase during pregnancy and decrease after menopause. These circumstances highlight the complexity of understanding smoking behavior among women and the need to consider factors, such as circulating hormone concentrations, that might interact with genetics to influence smoking-related outcomes.

The second dopaminergic SNP is rs6277 within DRD2, which encodes for the D2 dopamine receptor; the minor allele of rs6277 is associated with increased striatal D2 receptor density (48). Typically, decreased D2 receptor availability has been associated with risk for addiction. Thus, our observation that women with the minor allele of rs6277 had a higher proportion follow-up in relapse than women without the minor allele was surprising. However, we did not observe an association between SNP rs6277 and the likelihood of relapse throughout adulthood and prior studies found no association with smoking cessation (35). People with the minor allele have been shown to have a bias toward negative learning; that is, a bias toward avoiding behaviors with high probability of negative feedback rather than learning to choose behavior with probabilistic positive feedback (49). Bias toward negatively learning is one plausible explanation for the association with continued smoking after relapse to avoid the negative effects of quitting and withdrawal. More research is needed to understand the degree to which these mechanisms influence smoking behavior.

Significant differences in the associations of SNPs rs588765, rs680244, and rs6277 with the proportion follow-up in relapse by menopausal status at cessation were observed. The differences in proportion follow-up in relapse between genotypes of these SNPs were considerably larger among women who first reported quitting smoking postmenopause. In contrast, among women who quit premenopause, there was no difference in proportion follow-up in relapse by genotype of SNPs rs588765 and rs680244 and the difference by genotype of rs6277 was significantly smaller in premenopausal compared with postmenopausal women. The credibility of these observations is buttressed by the strength and biological plausibility of the associations among postmenopausal women, but replication of these findings is needed.

This phenomenon of greater genetic effect postmenopause could be partially due to decreased estrogen levels postmenopause, which can lead to decreased dopaminergic functioning, a possible explanation for higher risk of negative affect (50). The α-5 nAChR subunit may mediate nicotine's effect on mood and anxiety (42), thus some CHRNA5 variants, including rs588765 and rs680244, may play a stronger role in smoking relapse among postmenopausal women by influencing the effect of nicotine intake on mood. For some postmenopausal women, regulation of mood could be a more important motivator for smoking. In addition, the bias toward avoiding negative feedback that may be associated with rs6277 could more strongly influence older women who quit smoking as some negative consequences of cessation, such as negative affect and weight management concerns, may have a stronger influence postmenopause (50). Furthermore, chronic nicotine exposure leads to neuroadaptations in nicotinic and dopaminergic receptors (51). The interplay of neuroadaptation and genetics could contribute to greater nicotine dependence and behavioral addiction and play an important role in older smokers with greater long-term, chronic nicotine exposure.

Despite the strength of the long-term longitudinal follow-up in this study, the study had limitations. The existing genotype data was primarily for European ancestry women thus, it was not feasible to determine SNP associations among women of non-European ancestry. Genetic associations may differ by ancestry thus, examination of associations across people of non-European ancestry is necessary to fully understand genetic susceptibility for long-term smoking behavior. In addition, though the all-female study populations were important for evaluating genetic factors among women, it is necessary to investigate these SNP associations among men to understand if the associations differ by biological sex.

Furthermore, though well-suited for determining genetic associations, these data did not include important nongenetic factors contributing to smoking such as motivation to quit, use of cessation aids, detailed smoking behavior during pregnancy, and comprehensive repeated measures of exposure to other smokers at home, work, and within social networks (6). Likewise, these data were inclusive of only registered nurses and thus were homogeneous in socioeconomic status and education level, both factors associated with smoking prevalence and cessation (6). This homogeneity avoids confounding due to socioeconomic factors, but also precludes examining SNP interactions with socioeconomic factors.

This study is the first to quantify genetic associations with relapse and proportion of time in relapse among female smokers throughout adulthood. Regardless of the number and distribution of relapse events throughout adulthood, the proportion of follow-up in relapse is a measure of smoking dose, analogous to pack-years of smoking. The adverse health effects of smoking are dose-dependent, making proportion of follow-up in relapse clinically meaningful for identifying smokers who are prone to greater duration of lifetime smoking and thus greater risk of smoking-caused disease (52). The findings showed that some SNPs not associated with odds of relapse are associated with the amount of follow-up in relapse suggesting that genetic risk markers differ by stage of smoking cessation. The increased proportion of follow-up in relapse for individual variants may be small for some SNPs (e.g., 2.7% for rs4680), but the combined risk of multiple variants could have a larger effect on risk for long-term smoking.

Further, we found that some SNP associations with a higher proportion follow-up in relapse were considerably stronger in women who first quit smoking postmenopause. Postmenopausal women can experience increased cancer risk due to factors such as age and weight gain (53), and, postmenopausal women who also have a higher likelihood of continued smoking may be at even higher risk for morbidity and mortality. Thus, our findings emphasize the importance of studying women and accounting for changes pre- and postmenopause. Research on genetic risk of smoking in women by menopausal phase is warranted. This line of inquiry is important because long-term smoking cessation could potentially prevent up to 63,400 cancer deaths annually among women (1).

The identification of smokers at increased risk for relapse and long-term persistent smoking could inform precision medicine approaches. Long-term cessation management and relapse prevention could reduce smoking rates and prevent smoking attributable morbidity and mortality. The large number of people who smoke cigarettes and the limited resources available to implement long-term cessation and relapse prevention strategies reinforces the importance of research to identify the characteristics of cigarette smokers who are most likely to benefit from long-term interventions.

S.K. Jones reports other support from NIH TL1 Fellowship, MUSC Hollings Cancer Center, and NIH NIAMS during the conduct of the study. B. Wolf reports grants from NIH during the conduct of the study; grants from NIH outside the submitted work. No disclosures were reported by the other authors.

S.K. Jones: Conceptualization, data curation, formal analysis, funding acquisition, visualization, methodology, writing–original draft. A.J. Alberg: Conceptualization, supervision, visualization, writing–review and editing. K. Wallace: Conceptualization, supervision, writing–review and editing. B. Froeliger: Conceptualization, visualization, writing–review and editing. M.J. Carpenter: Conceptualization, supervision, writing–review and editing. B. Wolf: Conceptualization, formal analysis, supervision, validation, writing–review and editing.

B. Wolf and S.K. Jones received support from NIH/NIAMS P30AR072582 (CCCR Improving Minority Health in Rheumatic Diseases), S.K. Jones received funding support from NIH TL1 grant (TL1 TR001451) and the Medical University of South Carolina Hollings Cancer Center Abney Graduate Fellowship. The NHS received support from UM1 CA186107 (NHS infrastructure grant) and U01 CA176726 (NHS II Infrastructure grant).

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.