Exposure to high doses of alkylating agents is associated with increased risk of impaired spermatogenesis among nonirradiated male survivors of childhood cancer, but there is substantial variation in this risk. Here we conducted a genetic study for impaired spermatogenesis utilizing whole-genome sequencing data from 167 nonirradiated male childhood cancer survivors of European ancestry from the St. Jude Lifetime Cohort treated with cyclophosphamide equivalent dose (CED) ≥4,000 mg/m2. Sperm concentration from semen analysis was assessed as the primary outcome. Common variants (MAF > 0.05) were adjusted for age at cancer diagnosis, CED, and top principal components. Rare/low-frequency variants (MAF ≤ 0.05) were evaluated jointly by various functional annotations and 4-kb sliding windows. A novel locus at 7q21.3 containing TAC1/ASNS was associated with decreased sperm concentration (rs7784118: P = 3.5 × 10−8). This association was replicated in two independent samples of SJLIFE survivors of European ancestry, including 34 nonirradiated male survivors treated with 0 < CED < 4,000 mg/m2 (P = 3.1 × 10−4) and 24 male survivors treated with CED ≥4,000 mg/m2 and radiotherapy <40 Gray (P = 0.012). No association was observed among survivors not exposed to alkylating agents included in the CED (P > 0.29). rs7784118 conferred 3.48- and 9.73-fold increases in risk for clinically defined oligospermia and azoospermia and improved prediction of normospermic, oligospermic, and azoospermic states by 13.7%, 5.3%, and 21.7%. rs7784118 was associated with decreased testosterone level, increased levels of follicle stimulating and luteinizing hormones, and 8.52-fold increased risk of Leydig cell failure. Additional research is warranted to determine how this SNP influences spermatogenesis and to assess its clinical utility in characterizing high-risk survivors and guiding intervention strategies.

Significance:

The identified genetic markers harbor potential clinical utility in characterizing high-risk survivors and guiding intervention strategies including pretreatment patient counseling and use of fertility preservation services.

Survival rates for childhood cancer have improved dramatically, with a current five-year survival probability of approximately 85% (1). However, many survivors are at risk of long-term therapy-related morbidity and mortality (2–7). In male survivors, radiotherapy and chemotherapy, in particular alkylating agents, used in cancer treatment may result in temporary, long-term, or permanent gonadal toxicity, which significantly impacts fertility (8). The alkylating agents produce adducts and cross-links in DNA and induce apoptosis through covalent binding of alkyl groups to cellular macromolecules (9). These result in death of stem spermatogonia and a block in their differentiation, which lead to prolonged reductions in sperm count and/or duration of azoospermia after cytotoxic therapy.

Using the Childhood Cancer Survivor Study (CCSS; ref. 8), we have previously reported that male survivors aged 15 to 44 years, who were not surgically sterile, were 79% less likely to father a child compared with their siblings, with the risk strongly associated with testicular radiation of >7.5 Gray (Gy) and high cumulative cyclophosphamide dose (≥9,360 mg/m2). In a report by the St. Jude Lifetime Cohort (SJLIFE) Study (4), approximately 66% of male survivors treated with alkylating agents and/or radiotherapy experienced gonadal toxicity. In another SJLIFE Study (10) including 214 nonirradiated male survivors exposed to alkylating agent chemotherapy, the prevalence of oligospermia or azoospermia was approximately 10% among those with a cyclophosphamide equivalent dose (CED) of <4,000 mg/m2, which increased to 62% among those with CED ≥4,000 mg/m2. However, there was substantial variation in sperm concentration among survivors who received higher alkylating agent treatment, suggesting additional risk factors including genetic predisposition for greater cytotoxic damage in the testis and consequently decreased sperm concentration. Nevertheless, very little is known about the genetic factors that may influence fertility preservation after chemotherapy treatment among childhood cancer survivors.

To address this gap, we performed a genome-wide association (GWA) study using deep-coverage whole-genome sequencing (WGS) data in the SJLIFE study to comprehensively investigate the potential role of germline genetic factors on sperm concentration in nonirradiated adult male survivors of childhood cancer who had received high cumulative doses of alkylating agents. Genetic associations that achieved genome-wide significance were evaluated in two independent sets of samples from SJLIFE participants.

Study population

The SJLIFE Study (11), initiated in 2007, is a retrospective cohort study with the prospective clinical follow-up and ongoing enrollment of survivors of childhood cancer treated at St. Jude Children's Research Hospital (SJCRH) since 1962. Semen analysis was offered to men who had received gonadotoxic treatments [exposure to an alkylating agent, testicular irradiation (any dose), or hypothalamic–pituitary irradiation (≥40 Gy)]. To be eligible for the current study, SJLIFE participants had to be ≥18 years of age, have semen analysis performed, and have WGS data.

Whole-genome sequencing and quality control

To be included in the WGS (12), enrolled survivors were required to have consented for a SJLIFE clinical assessment before April 2016 and provided written consent for banking of a blood sample for use in future research. Survivors who had undergone allogeneic stem cell transplantation were excluded from sequencing. DNA was extracted from blood samples banked in the SJCRH biorepository using either the QIAamp DNA Blood Mini Kit (Qiagen, catalog no. 51106) or DNeasy Blood & Tissue Kit (catalog no. 69506). Following extraction, DNA concentration was fluorometrically measured using the Quant-it dsDNA Assay Kit (Life Technologies, catalog no. Q33130) and DNA integrity verified visually by agarose gel electrophoresis (E-gel, Life Technologies, catalog no. G8008-01). Unique samples were barcoded and WGS was performed at the HudsonAlpha Institute for Biotechnology Genomic Services Laboratory using the Illumina HiSeq X10 sequencers to generate 360 million paired-ends reads, each 150 base pairs (bp) in length, for each sample. Whole-exome sequencing (WES) was performed internally at SJCRH by the genomics laboratory in the Department of Computational Biology using Illumina HiSeq 4000 sequencing platform and used for genomic concordance quality control (QC) checks for sample verification.

A total of 3,026 DNA samples were sent to HudsonAlpha for WGS, and sequencing raw reads were received for all samples except for one that failed library preparation. Samples that failed the coverage QC, that is, >70% of coding exons covered at >20×, underwent “top off” to ensure the minimum coverage criteria were met. WES data of 3,013 survivors were used to check for potential sample swaps and to confirm identification of coding variants from the WGS. A total of 19 samples failed additional QC filtering. The failures were attributed to contamination (n = 5), gender discordance (n = 7), and unexpected discordance, with previously genotyped Affymetrix 6.0 data for the same sample leading to sample identity concerns (n = 8). The mean coverage per sample was 36.8×.

Details of genotype data QC, mapping, variant identification, and annotation among 3,006 SJLIFE participants are described in the Supplementary Methods. Of the 2,986 SJLIFE survivors with WGS data passing QC, sperm concentration was available for 742 male survivors who had not undergone vasectomy. No survivor underwent bilateral orchiectomy. We excluded 69 survivors receiving androgen treatment, leaving 673 survivors in the study. Of these, 529 survivors were of European ancestry based on their genotype data (Supplementary Fig. S1; Supplementary Methods), including 201 who had received alkylating agents without radiation, 219 who had received a combination of both alkylating agents and radiotherapy, and the remaining 109 who had not received alkylating agents with or without radiotherapy. Exposure to alkylating agents was calculated as CED (Supplementary Methods; ref. 13). Among the 201 participants exposed to only alkylating agents, 167 received high CED exposure (≥4,000 mg/m2), forming the discovery sample for the current GWAS. A flowchart of patient selection is shown in Fig. 1. This study was approved by the institutional review board of the SJCRH and all participants provided a written informed consent.

Figure 1.

Flow chart of patient selection in the study.

Figure 1.

Flow chart of patient selection in the study.

Close modal

Semen analysis

Procedures for sperm collection and semen analysis have been described previously (10). Briefly, sperm samples were collected by masturbation after a planned minimum of two days and a maximum of seven days of sexual abstinence and were processed within 30 minutes of collection following the WHO guidelines. Samples were allowed to liquefy and time to liquefaction was recorded. The fresh sperm sample was microscopically assessed before any freezing. If no sperm were detected, the sample was centrifuged, concentrated, and reassessed again before being classified as azoospermic. Specimens that contained more than zero and less than 15 million sperms/mL were classified as oligospermic and those with ≥15 million sperms/mL were classified as normospermic. In semen specimens, that were not azoospermic, several additional characteristics including motility (%) (14), progressive motility (0–4; refs. 15, 16), and morphology (≥4% Kruger strict; ref. 14) were assessed.

Hormone measurements

Serum testosterone, follicle stimulating hormone (FSH), and luteinizing hormone (LH) were measured using electro-chemiluminescent immunometric assay (Roche Cobas 6000 Analyzer; Roche Diagnostics); blood samples were drawn prior to 10 AM. Serum testosterone ≥250 ng/dL, FSH ≤11.6 IU/L, and LH ≤9.85 IU/L were considered normal. These cut-off values were validated using testosterone, LH, and FSH levels measured in community controls who were not on sex hormone replacement (n = 160; ages 18.3–62.7 years; Supplementary Table S1). Specifically, the normal value for testosterone (≥250 ng/dL), also the suggested cutoff by the assay manufacturer, was between the 5th–10th percentiles in controls and was consistent with clinical guidelines (17). The FSH of 11.6 IU/L and LH of 9.85 IU/L were the 97th percentile in controls and were used as the cutoffs in the absence of consensus regarding normal values for these hormones. Leydig cell failure (LCF) was defined as morning serum levels of total testosterone <250 ng/dL (or 8.67 nmol/L) and LH >9.85 IU/L (18). These hormone levels were examined as a secondary analysis following findings from the primary GWAS analysis of semen concentration. Among the 167 participants in the discovery sample, 128 (77%) survivors had their hormonal lab tests performed at the same time or within a week before the semen analysis. For 38 survivors, lab tests were performed more than a week (range: 1.01–7.20 years) before their semen analysis, and only one survivor had hormonal lab tests conducted almost a year after semen analysis.

Measurement of physical activity and dietary intake

Physical activity during the past seven days was assessed by asking survivors to complete the National Health and Nutrition Examination Survey Activity Questionnaire. Survivors were classified as physically active if they participated in vigorous physical activities, such as running, aerobics, wheelchair basketball, heavy yard work, or anything else that causes large increases in breathing or heart rate, for at least 150 minutes per week. Dietary intake was collected using a self-administered Blood 2005 Food Frequency Questionnaire asking about the usual dietary intake of 110 food items during the past 12 months. Diet was estimated using the Healthy Eating Index (HEI) ‐2010 (19), which measures adherence to the 2010 Dietary Guidelines for Americans and has 12 components. For each component, intake of foods and nutrients is represented on a density basis (per 1,000 kcal). For adequacy components [foods recommended for increased consumption for healthy nutrition (total fruit, whole fruit, total vegetables, greens and beans, whole grains, dairy, total protein foods, seafood and plant proteins, and fatty acids)], a score of 0 is assigned for no intake, and the scores increase proportionately as intake increases to the recommended level. For moderation components [foods recommended for limited consumption for healthy nutrition (refined grains, sodium, and calories from added sugars and saturated fats)], intake not exceeding the recommended level is assigned the maximum score and the score decreases as intake increases. The total HEI‐2010 score ranges from 0 (nonadherence) to 100 (perfect adherence).

GWA analysis

Association analyses were performed as described previously (20). Common variants (∼6.6 million) were examined using linear regression, assuming an additive genetic model, adjusting for age at cancer diagnosis, CED and the top 20 principal components (PC). PCs are eigenvectors obtained from principal component analyses of genotype data to represent the genetic ancestry of individuals. We chose to adjust for age at childhood cancer diagnosis and CED following our previous study (10) of spermatogenesis, which identified their associations with impaired spermatogenesis (P < 0.10). We also adjusted for the top 20 PCs to account for finer population stratification among survivors of European ancestry (20, 21). Burden test and Sequence Kernel Association Test (SKAT) were used initially with the large-sample-approximate test whose top results followed by randomization test with 100 million permutations, to assess associations of rare/low-frequency variants (∼10.8 million), adjusting for the same covariates used in the analysis of common variants above. Rare/low-frequency variants were aggregated into four analytic sets with respect to the RefSeq gene model (protein truncating variants, nonsynonymous-strict variants, nonsynonymous-broad variants, and regulatory variants), and 4-kb sliding window across the genome (Supplementary Methods). Common variants showing associations with P < 5 × 10−8 were considered statistically significant at the genome-wide level. Rare/low-frequency variant associations were considered significant at P < 9.6 × 10−7, accounting for the 26,290 analytic variant sets created with respect to the RefSeq gene model and the two tests performed for each set. For the sliding window approach, we considered two statistical tests for 666,920 contiguous nonoverlapping windows, leading to a significance threshold at P < 3.7 × 10−8.

Replication samples, analyses, and meta-analysis

We utilized two replication datasets, both consisting of SJLIFE participants, but independent of the discovery sample. The primary replication sample (n = 34) fulfilled the same eligibility criteria as the participants in the discovery analysis, except that they were exposed to low doses of alkylating agents (CED <4,000 mg/m2). The secondary replication sample (n = 24) also fulfilled the same eligibility criteria as the survivors in the discovery sample, except they were also exposed to radiotherapy. To minimize the radiation exposure influences, we excluded: (i) survivors with radiation exposure to body regions previously known to have an impact on sperm concentration, namely, testes, brain/head, spine, pelvis, abdomen, or total body irradiation (8, 10, 22); and ii) survivors who had a maximum radiation dose of ≥40 Gy to any body region (Supplementary Methods).

Replication analyses employed the same statistical framework and adjustment covariates including age at diagnosis, CED, and top PCs except associations were further adjusted for the maximum radiation dose in secondary replication analyses. We combined the association results from discovery and replication analyses using the inverse variance-weighted fixed effect model in METAL (23). Heterogeneity of allelic associations was examined using Cochran's Q Test (24) and the I2 index (25). For variants showing evidence of heterogeneity (Phet < 0.05), meta-analysis was carried out using the Han-Eskin random-effects model (RE2) implemented in METASOFT (26).

Estimation of clinical utility of genetic findings

We used a polytomous logistic regression to evaluate associations for azoospermia and oligospermia relative to normospermia, adjusting for the same covariates used in the discovery analyses including age at diagnosis, CED, and top 20 PCs. To assess the potential clinical utility of the genetic findings, we obtained predicted probabilities of the three sperm-concentration states for each survivor based on the polytomous logistic regression model with and without the genetic variants. A sperm concentration state with a prediction probability >0.5 was considered the predicted state (note a survivor with a prediction probability ≤0.5 would not have a predicted state). We then performed McNemar test for assessing correct prediction of each sperm concentration state (normospermia, oligospermia, or azoospermia) with and without the genetic variants in the prediction model.

Genetic associations with respect to endocrine parameters, age, physical activity, and nutrition

Serum hormone levels are indirect markers of gonadal function and hence we first examined genetic associations with levels of testosterone, FSH, LH, and LCF in the discovery sample, adjusting for the same set of covariates as the sperm concentration analysis including age at diagnosis, CED, and top 20 PCs. Second, the associations between the top SNP and sperm concentration were further adjusted, one at time, for testosterone, FSH, and LH values, to assess if the genetic associations with sperm concentration were confounded by serum hormone levels. Finally, stratified analyses were carried out with respect to normal cut-off values of hormonal levels to assess potential context-specific genetic effects. In addition, we performed subgroup analysis by the median age at semen analysis in the discovery sample to evaluate potential age-specificity of the top SNP's association with sperm concentration. We also assessed potential confounding on the genetic finding by physical activity and nutrition, by further adding them as covariates in the model.

Genetic effect among survivors of childhood cancer not exposed to alkylating agents

To assess the specificity of the genetic findings identified among survivors exposed to alkylating agents included in the CED, we examined the top SNP's association in an independent group of childhood cancer survivors of European ancestry who were not exposed to alkylating agents. There was a total of 109 survivors who also met the study eligibility criteria, including 21 nonirradiated survivors and 88 who were exposed to radiotherapy. Analyses were adjusted for age at diagnosis and top PCs among 21 nonirradiated survivors and with further adjustment for radiation dose in 88 survivors also treated with radiotherapy.

Data availability

Aligned binary map files for all the 334 survivors and the joint genotype calls are accessible through the St. Jude Cloud (https://stjude.cloud). All data generated or analyzed during the study are included in the published article (and its Supplementary Information files). Summary statistics from the discovery sample are available from the corresponding author upon request.

Demographic and treatment characteristics of all participants included in the discovery and replication cohorts are provided in Table 1, which also shows the characteristics of SJLIFE male survivors who met all the eligibility criteria of the discovery cohort but did not have both WGS and semen data available. The median time from cancer diagnosis to semen analysis was comparable across both discovery (21.2 years) and replication samples (primary replication sample: 20.4 years; and secondary replication sample: 19.8 years). Among the 201 nonirradiated participants exposed to alkylating agents, 167 survivors with a high CED dose (≥4,000 mg/m2) in the discovery sample had a significantly lower adjusted mean sperm concentration (−38.44 million sperm/mL; 95% CI = −59.18,−17.70; P = 3.7 × 10−4) than the 34 participants with a low CED dose (<4,000 mg/m2) in the primary replication sample (Supplementary Table S2). Among survivors in the discovery sample, those with normal sperm concentration were more likely to have normal motility, progressive motility, and morphology compared with those with oligospermia, although the difference was statistically significant only for progressive motility (P = 6.5 × 10−5; Supplementary Table S3).

Table 1.

Sociodemographic and treatment characteristics of SJLIFE survivors included in this study.

VariablesDiscovery sample (n = 167)Primary replication sample (n = 34)Secondary replication sample (n = 24)Nonparticipantsa (n = 109)
Age at diagnosis, years 
 Median (range) 5.4 (0.0–20.3) 8.8 (0.7–19.6) 12.6 (0.0–19.0) 12.2 (0.0–23.6) 
 ≤5 77 (46.1%) 6 (17.6%) 4 (16.7%) 32 (29.4%) 
 >5–10 35 (21.0%) 13 (38.2%) 5 (20.8%) 13 (11.9%) 
 >10–15 35 (21.0%) 8 (23.5%) 7 (29.2%) 32 (29.4%) 
 >15 20 (12.0%) 7 (20.6%) 8 (33.3%) 31 (28.4%) 
Age at assessment, years 
 Median (range) 28.5 (18.9–56.1) 29.5 (20.1–53.3) 32.0 (19.7–51.5) NA 
Elapsed time from diagnosis to assessment, years 
 Median (range) 21.2 (9.7–43.9) 20.4 (10.5–37.8) 19.8 (11.4–38.8) NA 
Cyclophosphamide equivalent dose (CED; mg/m2
 Median (range) 8,060 (4149–31894) 2,268 (938–3553) 12,834 (4288–30216) 7,996 (4161–30290) 
Maximum radiation exposure (Gray) 
 Median (range) NA NA 29.5 (2.0–38.0) NA 
Primary cancer diagnosis 
 Acute lymphoblastic leukemia 66 (39.5%) 9 (26.5%) 1 (4.2%) 38 (34.9%) 
 Non-Hodgkin lymphoma 30 (18.0%) 0 (0.0%) 0 (0.0%) 20 (18.3%) 
 Neuroblastoma 24 (14.4%) 20 (58.8%) 1 (4.2%) 8 (7.3%) 
 Osteosarcoma 19 (11.4%) 3 (8.8%) 0 (0.0%) 24 (22.0%) 
 Retinoblastoma 6 (3.6%) 0 (0.0%) 0 (0.0%) 2 (1.8%) 
 Acute myeloid leukemia 4 (2.4%) 0 (0.0%) 0 (0.0%) 3 (2.8%) 
 Ewing sarcoma family of tumors 4 (2.4%) 0 (0.0%) 12 (50%) 2 (1.8%) 
 Germ cell tumor 3 (1.8%) 0 (0.0%) 0 (0.0%) 0 (0.0%) 
 Melanoma 3 (1.8%) 0 (0.0%) 0 (0.0%) 1 (0.9%) 
 Soft tissue sarcoma 2 (1.2%) 0 (0.0%) 1 (4.2%) 0 (0.0%) 
 Central nervous system (CNS) tumors 1 (0.6%) 0 (0.0%) 0 (0.0%) 1 (0.9%) 
 Histiocytosis 1 (0.6%) 1 (2.9%) 0 (0.0%) 0 (0.0%) 
 Hodgkin lymphoma 1 (0.6%) 0 (0.0%) 9 (37.5%) 3 (2.8%) 
 Liver malignancies 1 (0.6%) 0 (0.0%) 0 (0.0%) 0 (0.0%) 
 Other leukemia 1 (0.6%) 0 (0.0%) 0 (0.0%) 1 (0.9%) 
 Rhabdomyosarcoma 1 (0.6%) 1 (2.9%) 0 (0.0%) 3 (2.8%) 
Physical activity 
 Active 121 (72.5%) 20 (58.8%) 12 (50.0%) 62 (56.9%) 
 Inactive 40 (24.0%) 13 (38.2%) 12 (50.0%) 41 (37.6%) 
 Missing 6 (3.6%) 1 (2.9%) 0 (0.0%) 6 (5.5%) 
Education 
 1–8 years (grade school) 3 (1.8%) 0 (0.0%) 0 (0.0%) 2 (1.8%) 
 9–12 years (high school), but did not graduate 11 (6.6%) 4 (11.8%) 2 (8.3%) 5 (4.6%) 
 Completed high school/GED 26 (15.6%) 4 (11.8%) 2 (8.3%) 32 (29.4%) 
 Training after high school, other than college 10 (6.0%) 1 (2.9%) 1 (4.2%) 7 (6.4%) 
 Some college 45 (26.9%) 8 (23.5%) 5 (20.8%) 27 (24.8%) 
 College graduate 35 (21.0%) 9 (26.5%) 9 (37.5%) 22 (20.2%) 
 Post-graduate level 18 (10.8%) 4 (11.8%) 3 (12.5%) 1 (0.9%) 
 Other 6 (3.6%) 1 (2.9%) 0 (0.0%) 3 (2.8%) 
 Missing 13 (7.8%) 3 (8.8%) 2 (8.3%) 10 (9.2%) 
Median household income 
 Less than $19,999 19 (11.4%) 4 (11.8%) 1 (4.2%) 15 (13.8%) 
 $20,000–$39,999 38 (22.8%) 4 (11.8%) 4 (16.7%) 28 (25.7%) 
 $40,000–$59,999 32 (19.2%) 7 (20.6%) 6 (25.0%) 16 (14.7%) 
 $60,000–$79,999 19 (11.4%) 7 (20.6%) 2 (8.3%) 9 (8.3%) 
 $80,000–$99,999 9 (5.4%) 1 (2.9%) 4 (16.7%) 7 (6.4%) 
 Over $100,000 19 (11.4%) 5 (14.7%) 6 (25.0%) 9 (8.3%) 
 Don't know 24 (14.4%) 3 (8.8%) 1 (4.2%) 16 (14.7%) 
 Missing 7 (4.2%) 3 (8.8%) 0 (0.0%) 9 (8.3%) 
Employment 
 Unable to work due to illness or disability 6 (3.6%) 1 (2.9%) 2 (8.3%) 6 (5.5%) 
 Never had a job, or not currently working or unemployed and looking for work 8 (4.8%) 1 (2.9%) 0 (0.0%) 1 (0.9%) 
 Caring for home or family (not seeking paid work), or student, or retired, or working part or full-time 129 (77.2%) 30 (88.2%) 18 (75%) 80 (73.4%) 
 Others 4 (2.4%) 1 (2.9%) 0 (0.0%) 1 (0.9%) 
 Missing 20 (12.0%) 1 (2.9%) 4 (16.7%) 21 (19.3%) 
Distance to St. Jude Children's Research Hospital (miles) 
 <100 31 (18.6%) 5 (14.7%) 7 (29.2%) 17 (15.6%) 
 100–299 49 (29.3%) 9 (26.5%) 7 (29.2%) 32 (29.4%) 
 300+ 82 (49.1%) 20 (58.8%) 9 (37.5%) 56 (51.4%) 
 Missing 5 (3.0%) 0 (0.0%) 1 (4.2%) 4 (3.7%) 
VariablesDiscovery sample (n = 167)Primary replication sample (n = 34)Secondary replication sample (n = 24)Nonparticipantsa (n = 109)
Age at diagnosis, years 
 Median (range) 5.4 (0.0–20.3) 8.8 (0.7–19.6) 12.6 (0.0–19.0) 12.2 (0.0–23.6) 
 ≤5 77 (46.1%) 6 (17.6%) 4 (16.7%) 32 (29.4%) 
 >5–10 35 (21.0%) 13 (38.2%) 5 (20.8%) 13 (11.9%) 
 >10–15 35 (21.0%) 8 (23.5%) 7 (29.2%) 32 (29.4%) 
 >15 20 (12.0%) 7 (20.6%) 8 (33.3%) 31 (28.4%) 
Age at assessment, years 
 Median (range) 28.5 (18.9–56.1) 29.5 (20.1–53.3) 32.0 (19.7–51.5) NA 
Elapsed time from diagnosis to assessment, years 
 Median (range) 21.2 (9.7–43.9) 20.4 (10.5–37.8) 19.8 (11.4–38.8) NA 
Cyclophosphamide equivalent dose (CED; mg/m2
 Median (range) 8,060 (4149–31894) 2,268 (938–3553) 12,834 (4288–30216) 7,996 (4161–30290) 
Maximum radiation exposure (Gray) 
 Median (range) NA NA 29.5 (2.0–38.0) NA 
Primary cancer diagnosis 
 Acute lymphoblastic leukemia 66 (39.5%) 9 (26.5%) 1 (4.2%) 38 (34.9%) 
 Non-Hodgkin lymphoma 30 (18.0%) 0 (0.0%) 0 (0.0%) 20 (18.3%) 
 Neuroblastoma 24 (14.4%) 20 (58.8%) 1 (4.2%) 8 (7.3%) 
 Osteosarcoma 19 (11.4%) 3 (8.8%) 0 (0.0%) 24 (22.0%) 
 Retinoblastoma 6 (3.6%) 0 (0.0%) 0 (0.0%) 2 (1.8%) 
 Acute myeloid leukemia 4 (2.4%) 0 (0.0%) 0 (0.0%) 3 (2.8%) 
 Ewing sarcoma family of tumors 4 (2.4%) 0 (0.0%) 12 (50%) 2 (1.8%) 
 Germ cell tumor 3 (1.8%) 0 (0.0%) 0 (0.0%) 0 (0.0%) 
 Melanoma 3 (1.8%) 0 (0.0%) 0 (0.0%) 1 (0.9%) 
 Soft tissue sarcoma 2 (1.2%) 0 (0.0%) 1 (4.2%) 0 (0.0%) 
 Central nervous system (CNS) tumors 1 (0.6%) 0 (0.0%) 0 (0.0%) 1 (0.9%) 
 Histiocytosis 1 (0.6%) 1 (2.9%) 0 (0.0%) 0 (0.0%) 
 Hodgkin lymphoma 1 (0.6%) 0 (0.0%) 9 (37.5%) 3 (2.8%) 
 Liver malignancies 1 (0.6%) 0 (0.0%) 0 (0.0%) 0 (0.0%) 
 Other leukemia 1 (0.6%) 0 (0.0%) 0 (0.0%) 1 (0.9%) 
 Rhabdomyosarcoma 1 (0.6%) 1 (2.9%) 0 (0.0%) 3 (2.8%) 
Physical activity 
 Active 121 (72.5%) 20 (58.8%) 12 (50.0%) 62 (56.9%) 
 Inactive 40 (24.0%) 13 (38.2%) 12 (50.0%) 41 (37.6%) 
 Missing 6 (3.6%) 1 (2.9%) 0 (0.0%) 6 (5.5%) 
Education 
 1–8 years (grade school) 3 (1.8%) 0 (0.0%) 0 (0.0%) 2 (1.8%) 
 9–12 years (high school), but did not graduate 11 (6.6%) 4 (11.8%) 2 (8.3%) 5 (4.6%) 
 Completed high school/GED 26 (15.6%) 4 (11.8%) 2 (8.3%) 32 (29.4%) 
 Training after high school, other than college 10 (6.0%) 1 (2.9%) 1 (4.2%) 7 (6.4%) 
 Some college 45 (26.9%) 8 (23.5%) 5 (20.8%) 27 (24.8%) 
 College graduate 35 (21.0%) 9 (26.5%) 9 (37.5%) 22 (20.2%) 
 Post-graduate level 18 (10.8%) 4 (11.8%) 3 (12.5%) 1 (0.9%) 
 Other 6 (3.6%) 1 (2.9%) 0 (0.0%) 3 (2.8%) 
 Missing 13 (7.8%) 3 (8.8%) 2 (8.3%) 10 (9.2%) 
Median household income 
 Less than $19,999 19 (11.4%) 4 (11.8%) 1 (4.2%) 15 (13.8%) 
 $20,000–$39,999 38 (22.8%) 4 (11.8%) 4 (16.7%) 28 (25.7%) 
 $40,000–$59,999 32 (19.2%) 7 (20.6%) 6 (25.0%) 16 (14.7%) 
 $60,000–$79,999 19 (11.4%) 7 (20.6%) 2 (8.3%) 9 (8.3%) 
 $80,000–$99,999 9 (5.4%) 1 (2.9%) 4 (16.7%) 7 (6.4%) 
 Over $100,000 19 (11.4%) 5 (14.7%) 6 (25.0%) 9 (8.3%) 
 Don't know 24 (14.4%) 3 (8.8%) 1 (4.2%) 16 (14.7%) 
 Missing 7 (4.2%) 3 (8.8%) 0 (0.0%) 9 (8.3%) 
Employment 
 Unable to work due to illness or disability 6 (3.6%) 1 (2.9%) 2 (8.3%) 6 (5.5%) 
 Never had a job, or not currently working or unemployed and looking for work 8 (4.8%) 1 (2.9%) 0 (0.0%) 1 (0.9%) 
 Caring for home or family (not seeking paid work), or student, or retired, or working part or full-time 129 (77.2%) 30 (88.2%) 18 (75%) 80 (73.4%) 
 Others 4 (2.4%) 1 (2.9%) 0 (0.0%) 1 (0.9%) 
 Missing 20 (12.0%) 1 (2.9%) 4 (16.7%) 21 (19.3%) 
Distance to St. Jude Children's Research Hospital (miles) 
 <100 31 (18.6%) 5 (14.7%) 7 (29.2%) 17 (15.6%) 
 100–299 49 (29.3%) 9 (26.5%) 7 (29.2%) 32 (29.4%) 
 300+ 82 (49.1%) 20 (58.8%) 9 (37.5%) 56 (51.4%) 
 Missing 5 (3.0%) 0 (0.0%) 1 (4.2%) 4 (3.7%) 

Note: All survivors in the discovery and replication samples and the nonparticipants are of European ancestry based on their genotype data.

Abbreviation: NA, not applicable.

aNonparticipants fulfilled the same criteria as survivors in the discovery samples, except they did not have both WGS data and semen analysis.

Genetic variants associated with sperm concentration

Common variant analyses identified three SNPs showing genome-wide significance. The quantile-quantile plot and genomic inflation factor of 1.01 (Supplementary Fig. S2) indicate no or little influence of population stratification on the association results. These three SNPs represent two different loci on chromosome 7, including two SNPs (rs10225965 and rs2237572) on 7q21.2 and another SNP (rs7784118) on 7q21.3; Table 2; Fig. 2A). The minor allele (T) of the top SNP on 7q21.2 (rs10225965) was associated with increased sperm concentration (39.14 million sperms/mL per minor allele; 95% CI = 26.25, 52.03; P = 1.9 × 10−8). Another SNP rs2237572 on 7q21.2 in very high linkage disequilibrium (LD) with rs10225965 (r2 = 0.98) also showed a similar association with sperm concentration. SNP rs7784118 on 7q21.3 was associated with decreased sperm concentration (−37.30 million sperms/mL per minor allele; 95% CI = -49.83, −24.77; P = 3.5 × 10−8; Fig. 2B). Of the 167 survivors, over 50% carried the minor allele of rs7784118, including 78 and 7 survivors who were heterozygotes and homozygous for the minor allele (Supplementary Fig. S3). Analyses of 10.8 million rare/low-frequency variants identified CDC42BPG and 13 sliding windows as significantly associated with sperm concentration (Supplementary Table S4). However, none of these remained significant based on the 100 million permutations.

Table 2.

Genome-wide significant loci on chromosome 7q21.3 for sperm concentration and their results in independent replication samples.

Discovery sample (n = 167)Primary replication sample (n = 34)Secondary replication sample (n = 24)
SNPBPMA/OAMAFCrude mean sperm concentration (OO/OM/MM)Change in sperm concentration, million/mL (95% CI)PMAFChange in sperm concentration, million/mL (95% CI)PMAFChange in sperm concentration, million/mL (95% CI)P
rs2237572 92630946 C/T 0.21 21.93/51.28/86.71 38.45 (25.44 to 51.46) 4.2 × 10−8 0.16 −73.51 (−183.60 to 36.58) 0.22 0.26 −0.22 (−13.31 to 12.86) 0.974 
rs10225965 92644264 T/C 0.20 21.52/53.22/86.71 39.14 (26.25 to 52.03) 1.9 × 10−8 0.16 −73.51 (−183.60 to 36.58) 0.22 0.24 0.80 (−13.11 to 14.70) 0.912 
rs7784118 97627780 A/G 0.28 52.26/17.99/6.00 −37.30 (−49.83 to −24.77) 3.5 × 10−8 0.34 −73.27 (−99.97 to −46.58) 3.1 × 10−4 0.36 −14.88 (−25.11 to −4.64) 0.012 
Discovery sample (n = 167)Primary replication sample (n = 34)Secondary replication sample (n = 24)
SNPBPMA/OAMAFCrude mean sperm concentration (OO/OM/MM)Change in sperm concentration, million/mL (95% CI)PMAFChange in sperm concentration, million/mL (95% CI)PMAFChange in sperm concentration, million/mL (95% CI)P
rs2237572 92630946 C/T 0.21 21.93/51.28/86.71 38.45 (25.44 to 51.46) 4.2 × 10−8 0.16 −73.51 (−183.60 to 36.58) 0.22 0.26 −0.22 (−13.31 to 12.86) 0.974 
rs10225965 92644264 T/C 0.20 21.52/53.22/86.71 39.14 (26.25 to 52.03) 1.9 × 10−8 0.16 −73.51 (−183.60 to 36.58) 0.22 0.24 0.80 (−13.11 to 14.70) 0.912 
rs7784118 97627780 A/G 0.28 52.26/17.99/6.00 −37.30 (−49.83 to −24.77) 3.5 × 10−8 0.34 −73.27 (−99.97 to −46.58) 3.1 × 10−4 0.36 −14.88 (−25.11 to −4.64) 0.012 

Abbreviations: CI, confidence interval; Chr, chromosome [genomic positions are shown relative to GRCh38 (hg38)]; MA, minor allele; MAF, minor allele frequency; MM, survivors homozygous for the minor allele; OA, other (major) allele; OM, survivors with heterozygote genotypes; OO, survivors homozygous for other allele.

Figure 2.

A, Results from the GWA analysis showing Manhattan plot for genome-wide associations with sperm concentration in the discovery sample. The horizontal axis shows the chromosomal position, and the vertical axis shows the statistical significance of the tested autosomal common variants, using multivariable linear regression. Red horizontal line, genome-wide significant threshold (P = 5 × 10−8). B, LocusZoom plot of the novel risk locus for sperm concentration at the 7q21.3 locus showing association with the sperm concentration expressed as −log10(P) obtained from analysis of common variants in the discovery sample. Results for the SNPs within ± 500 kb of the top SNP rs7784118 are shown. SNPs are shown as circles, with the top SNP represented by a purple diamond. All other SNPs are color coded according to the strength of LD with the top SNP (as measured by r2 in the European 1000 Genomes project data).

Figure 2.

A, Results from the GWA analysis showing Manhattan plot for genome-wide associations with sperm concentration in the discovery sample. The horizontal axis shows the chromosomal position, and the vertical axis shows the statistical significance of the tested autosomal common variants, using multivariable linear regression. Red horizontal line, genome-wide significant threshold (P = 5 × 10−8). B, LocusZoom plot of the novel risk locus for sperm concentration at the 7q21.3 locus showing association with the sperm concentration expressed as −log10(P) obtained from analysis of common variants in the discovery sample. Results for the SNPs within ± 500 kb of the top SNP rs7784118 are shown. SNPs are shown as circles, with the top SNP represented by a purple diamond. All other SNPs are color coded according to the strength of LD with the top SNP (as measured by r2 in the European 1000 Genomes project data).

Close modal

Replication results

Association of rs7784118 at 7q21.3 locus was replicated in both primary (−73.27 million sperm/mL per minor allele; 95% CI = −99.97, −46.58; P = 3.1 × 10−4) and secondary (−14.88 million sperm/mL per minor allele; 95% CI = −25.11, −4.64, P = 0.012) replication samples (Table 2); results were also consistent in the combined analyses (Supplementary Table S5). In contrast, the 7q21.2 locus did not show associations in both primary and secondary replication samples.

Effect of endocrine parameters, age, physical activity, and nutrition on the 7q21.3-sperm concentration

We observed associations between the top SNP rs7784118 on the 7q21.3 locus with the levels of various hormones among the 167 survivors in the discovery sample. After adjusting for age at diagnosis, CED, and the top 20 PCs, the minor allele showed associations with decreased testosterone level (−70.38 ng/dL; P = 1.8 × 10−3) and increased levels of both FSH (4.60 IU/L; P = 7.5 × 10−4) and LH (1.55 IU/L; 95% CI = 1.08 – 2.02; P = 1.2 × 10−3). Six out of the 167 survivors had LCF and the SNP rs7784118 showed a significant association with an increased risk of LCF (OR = 8.52; P = 0.0082).

The association between rs7784118 and sperm concentration remained consistent even after adjusting for serum levels of testosterone (−34.72 million sperms/mL per minor allele; P = 4.9 × 10−7), FSH (−28.52 million sperms/mL per minor allele; P = 1.2 × 10−6), and LH (−33.08 million sperms/mL per minor allele; P = 7.3 × 10−6). These observations were further supported by the results from stratified analyses, which showed the rs7784118 was predominantly associated with sperm concentration among survivors with normal hormone levels of testosterone (−44.34 million sperms/mL per minor allele; P = 7.6 × 10−8), FSH (−45.48 million sperms/mL per minor allele; P = 4.7 × 10−8), and LH (−57.67 million sperms/mL per minor allele; P = 8.7 × 10−9; Supplementary Table S6). Except nominal statistical significance among survivors with abnormal FSH levels (7.41 million sperms/mL per minor allele; P = 0.017), rs7784118 showed no statistically significant association among those with abnormal hormone levels.

Adjusting for age at diagnosis, CED, and the top 20 PCs, we found similar associations between the rs7784118 and sperm concentration in survivors above or below the median age of 28.5 years at semen analysis, −39.88 million sperms/mL per minor allele (P = 9.6 × 10−4) and −33.05 million sperms/mL per minor allele (P = 1.8 × 10−4), respectively. Further adjusting for physical activity and nutrition did not materially change the association between the rs7784118 and sperm concentration (−39.13 million sperm/mL per minor allele; P = 6.0 × 10−8).

Potential clinical utility of the 7q21.3 locus

In the discovery sample, 73 were normospermic, 48 were oligospermic, and 46 were azoospermic. Mean sperm concentrations were 3.43 million sperms/mL (range = 0.001 – 14.5) for survivors with oligospermia and 76.24 million sperms/mL (range = 17.0 – 245.0) for those with normospermia. SNP rs7784118 conferred increased risk of oligospermia (OR = 3.48; 95% CI = 1.46 – 8.33; P = 4.9 × 10−3) and azoospermia (OR = 9.73; 95% CI = 3.64 – 26.02; P = 5.8 × 10−6), respectively. Notably, the model with SNP rs7784118 successfully predicted 22 (47.8%) of the 46 azoospermic survivors, compared with only 12 individuals (26.1%) by the model without the SNP (Fig. 3), and this improvement was statistically significant (P = 0.016). The proportion of predicted survivors with oligospermia was also higher, albeit not statistically significant (P = 0.37), for the model with the SNP (20.8%), relative to that of the model without the SNP (14.6%). Of the 73 male survivors with normospermia, 55 (75.3%) were correctly predicted by the model with the SNP, whereas only 45 (61.6%) were predicted to be normospermic by the model without the SNP. The improvement in prediction of normospermia was also statistically significant (P = 0.044). Results in survivors not exposed to CED showed that rs7784118 was not associated with sperm concentration in 21 nonirradiated survivors not exposed to CED (25.56 million sperms/mL per minor allele; P = 0.29) and in 88 survivors exposed to radiotherapy but not exposed to CED (9.98 million sperms/mL per minor allele; P = 0.52).

Figure 3.

Proportion of male survivors in the discovery sample predicted (with >50% probabilities) to be normospermic, oligospermic, or azoospermic by the clinical model including only nongenetic factors (age at diagnosis, CED, and top principal components) and by the clinical model plus the top SNP rs7784118 at the 7q21.3 locus.

Figure 3.

Proportion of male survivors in the discovery sample predicted (with >50% probabilities) to be normospermic, oligospermic, or azoospermic by the clinical model including only nongenetic factors (age at diagnosis, CED, and top principal components) and by the clinical model plus the top SNP rs7784118 at the 7q21.3 locus.

Close modal

Expanding upon our previous findings (10), we performed the first comprehensive genetic investigation for impaired spermatogenesis in the high-risk clinical population of survivors of pediatric cancer, and identified a novel locus on chromosome 7q21.3 showing genome-wide significance with sperm concentration and replicated this finding in two independent samples. Our results showed appreciable improvement in the prediction of both impaired and normal spermatogenesis using the 7q21.3 locus, thereby supporting potential clinical utility in identifying patients at high-risk for impaired spermatogenesis. We also found a statistically significant (P = 5.5 × 10−4) interaction between rs7784118 at the 7q21.3 locus and rs10225965 at the 7q21.2 locus in the discovery sample, but this finding was not replicated in either of the replication samples (both P > 0.40). Further investigation in a larger sample size is required to accurately assess these nonadditive genetic effects contributing to impaired spermatogenesis in survivors.

Magnitudes of the rs7784118 associations in the primary and secondary replication samples were approximately 2-fold and half-fold, respectively, of that observed in the discovery sample. Furthermore, rs7784118 was not associated with sperm concentration in survivors of childhood cancer not exposed to CED, suggesting its specificity with decreased sperm concentration and impaired spermatogenesis in survivors exposed to CED. To assess this specificity more directly, we conducted an interaction test assessing a modification of the rs7784118 association with sperm concentration by alkylating agent exposure status. Specifically, we tested the SNP × CED-exposure interaction in a combined analysis of survivors exposed to CED (both discovery sample and primary replication sample) and those not exposed to CED: rs7784118 showed a significant interaction with CED exposure (beta = −36.06; P = 4.6 × 10−3). These data suggest rs7784118 may be associated with sperm concentration only in survivors with exposure to alkylating agents, although additional investigation in a larger sample is needed to confirm this observation and to further delineate the contribution of genetic risk factors and alkylating agents in impaired spermatogenesis.

While semen analysis remains the gold standard of fertility assessment, serum hormone levels provide an indirect assessment (27), although serum markers do not consistently correlate with sperm concentration (28). We demonstrate significant associations of rs7784118 with decreased levels of testosterone and increased levels of FSH and LH, the levels suggestive of LCF and consistent with gonadal injury (18, 29). Our results show that the association between rs7784118 and sperm concentration was not confounded by serum levels of testosterone, FSH, and LH, and the genetic association was observed chiefly among the survivors with normal hormone levels. The rs7784118′s association with sperm concentration was not modified by age at semen analysis, nor was it confounded by physical activity or nutrition. While residual confounding by factors, we did not measure or adjust for, including unmeasured elements of physical activities and nutrition, these data suggest that the underlying mechanisms by which the 7q21.3 locus increases risk of impaired spermatogenesis in survivors are likely independent of the hormonal influence, physical activity, and nutrition. Furthermore, rs7784118 showed 3.48-fold and 9.73-fold increased risk of oligospermia and azoospermia, respectively. The clinical prediction model with the rs7784118 genotype was able to correctly predict an additional 13.7%, 5.3%, and 21.7% male survivors with normospermia, oligospermia, and azoospermia, respectively, relative to the clinical model that did not include the SNP. However, our results of the clinical prediction model with the SNP should be interpreted with caution because performance of this model could not be evaluated in an independent cohort of survivors due to the unavailability of a comparable dataset.

SNP rs7784118 is located near TAC1 (tachykinin precursor 1) and ASNS (asparagine synthetase) and overlaps with the FOXC1-binding site. FOXC1 is a member of the forkhead family of transcription factors that is involved in a broad range of cellular and development of processes and certainly could influence spermatogenesis. ASNS converts aspartate and glutamine to asparagine and glutamate (30), influencing the glutamine-glutamate ratio and exchange between cells (referred to as the glutamine glutamate cycle). We speculate that ASNS could affect the germ cell supply of glutamine and influence spermatogenesis. Research is ongoing to elucidate the complete molecular mechanisms underlying the 7q21.3 association.

Limitations should be considered when interpreting the results of this study. GWASs have been very successful in identifying genotype-phenotype associations for thousands of complex traits and diseases over the past decade (31). These associations have provided insights into the genetic architecture of disease susceptibility, identification of new therapeutics, and risk prediction and optimization of therapies based on genotype. However, GWASs have limitations, such as their inability to identify all genetic determinants of complex traits, especially the rare variants, and the difficulty in defining the true causal associations. Sample size of this GWAS was not sufficiently large to identify genetic associations with modest effect size or with lower allele frequencies, and hence future GWASs are needed to discover rarer variants contributing to impaired spermatogenesis in male survivors of childhood cancer. Our study relied upon a single semen sample because of the logistics of the SJLIFE study, which requires many participants to travel long distances to the St. Jude Children's Research Hospital for multiple days (average 3 days) of evaluation. However, a handful of survivors in the discovery sample had subsequent semen analysis and the results were generally consistent with those from the previous semen analysis used in this study (Supplementary Table S7). Our study population represents a highly selected group of long-term male survivors diagnosed and treated at a single institution, whose study eligibility was based on their previous cancer treatment (i.e., absence of radiotherapy and exposure to CED ≥4,000 mg/m2). Participation in the sperm analysis might have been affected by other factors such as history of established fertility, known infertility, inability, or unwillingness to provide a sample, and being a SJLIFE participant, which may be affected by residing closer to St. Jude.

In summary, we identified a novel risk locus at 7q21.3 associated with sperm concentration in survivors of childhood cancer treated with alkylating agents. The top SNP rs7784118 also showed appreciable improvements on identification of male survivors with impaired spermatogenesis, and, if confirmed in independent datasets, the 7q21.3 locus may have potential utility in identifying male childhood cancer patients who are at increased risk of impaired spermatogenesis and guiding intervention strategies including pretreatment patient counselling and use of fertility preservation services. For genotype-informed high-risk male patients with cancer treated with alkylating agents, options for preservation of fertility, including choice of less gonadotoxic chemotherapy regimens (as feasible), semen banking in postpubertal patients, and cryopreservation of stem spermatogonia and/or testicular tissue in prepubertal patients for later use, may be discussed.

M.M. Hudson reports grants from NCI U01CA195547 (grant to institution for research) during the conduct of the study. L.L. Robison reports grants from National Cancer Institute during the conduct of the study. D.M. Green reports grants from NIH during the conduct of the study. No potential conflicts of interest were disclosed by the other authors.

Y. Sapkota: Conceptualization, data curation, software, formal analysis, supervision, validation, investigation, methodology, writing-original draft, writing-review and editing. C.L. Wilson: Writing-review and editing. A.K. Zaidi: Validation, writing-review and editing. W. Moon: Software, writing-review and editing. K. Fon Tacer: Investigation, writing-review and editing. L. Lu: Data curation, writing-review and editing. Q. Liu: Formal analysis, writing-review and editing. J. Baedke: Writing-review and editing. R. Dhaduk: Data curation, writing-review and editing. Z. Wang: Data curation, writing-review and editing. W. Chemaitilly: Data curation, writing-review and editing. M.J. Krasin: Data curation, writing-review and editing. F.B. Berry: Validation, writing-review and editing. J. Zhang: Data curation, funding acquisition, writing-review and editing. M.M. Hudson: Conceptualization, resources, supervision, funding acquisition, writing-review and editing. L.L. Robison: Conceptualization, resources, supervision, funding acquisition, project administration, writing-review and editing. D.M. Green: Data curation, writing-review and editing. Y. Yasui: Conceptualization, supervision, funding acquisition, investigation, methodology, writing-review and editing.

The St. Jude Lifetime Cohort (SJLIFE) study is supported by the United States National Cancer Institute (NCI) of the NIH (U01 CA195547 to M.M. Hudson and L.L. Robinson, principal investigators; Cancer Center Support CORE grant CA21765 to C. Roberts, principal investigator). This study is also supported by NIH/NCI R01 CA216354 (to J. Zhang and Y. Yasui, principal investigators) from the National Cancer Institute and the American Lebanese Syrian Associated Charities, Memphis, Tennessee. We thank William H Kutteh, M.D., Ph.D. and Raymond W Ke, M.D. for providing the semen analysis data.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Howlader
N
,
Noone
AM
,
Krapcho
M
,
Miller
D
,
Bishop
K
,
Kosary
CL
, et al
(
eds
).
SEER cancer statistics review, 1975–2014
,
Bethesda, MD
:
National Cancer Institute
,April 
2017
.
Available at
: https://seer.cancer.gov/csr/1975_2014/.
2.
Armstrong
GT
,
Chen
Y
,
Yasui
Y
,
Leisenring
W
,
Gibson
TM
,
Mertens
AC
, et al
Reduction in late mortality among 5-year survivors of childhood cancer
.
N Engl J Med
2016
;
374
:
833
42
.
3.
Bhatia
S
,
Sklar
C
. 
Second cancers in survivors of childhood cancer
.
Nat Rev Cancer
2002
;
2
:
124
32
.
4.
Hudson
MM
,
Ness
KK
,
Gurney
JG
,
Mulrooney
DA
,
Chemaitilly
W
,
Krull
KR
, et al
Clinical ascertainment of health outcomes among adults treated for childhood cancer
.
JAMA
2013
;
309
:
2371
81
.
5.
Olsen
JH
,
Moller
T
,
Anderson
H
,
Langmark
F
,
Sankila
R
,
Tryggvadottir
L
, et al
Lifelong cancer incidence in 47,697 patients treated for childhood cancer in the Nordic countries
.
J Natl Cancer Inst
2009
;
101
:
806
13
.
6.
Reulen
RC
,
Frobisher
C
,
Winter
DL
,
Kelly
J
,
Lancashire
ER
,
Stiller
CA
, et al
Long-term risks of subsequent primary neoplasms among survivors of childhood cancer
.
JAMA
2011
;
305
:
2311
9
.
7.
Robison
LL
,
Hudson
MM
. 
Survivors of childhood and adolescent cancer: life-long risks and responsibilities
.
Nat Rev Cancer
2014
;
14
:
61
70
.
8.
Green
DM
,
Kawashima
T
,
Stovall
M
,
Leisenring
W
,
Sklar
CA
,
Mertens
AC
, et al
Fertility of male survivors of childhood cancer: a report from the Childhood Cancer Survivor Study
.
J Clin Oncol
2010
;
28
:
332
9
.
9.
Meistrich
ML
. 
Male gonadal toxicity
.
Pediatr Blood Cancer
2009
;
53
:
261
6
.
10.
Green
DM
,
Liu
W
,
Kutteh
WH
,
Ke
RW
,
Shelton
KC
,
Sklar
CA
, et al
Cumulative alkylating agent exposure and semen parameters in adult survivors of childhood cancer: a report from the St Jude Lifetime Cohort Study
.
Lancet Oncol
2014
;
15
:
1215
23
.
11.
Hudson
MM
,
Ehrhardt
MJ
,
Bhakta
N
,
Baassiri
M
,
Eissa
H
,
Chemaitilly
W
, et al
Approach for classification and severity grading of long-term and late-onset health events among childhood cancer survivors in the St. Jude Lifetime Cohort
. Cancer Epidem Biomar
2017
;
26
:
666
74
.
12.
Wang
ZM
,
Wilson
CL
,
Easton
J
,
Thrasher
A
,
Mulder
H
,
Liu
Q
, et al
Genetic risk for subsequent neoplasms among long-term survivors of childhood cancer
.
J Clin Oncol
2018
;
36
:
2078
87
.
13.
Green
DM
,
Nolan
VG
,
Goodman
PJ
,
Whitton
JA
,
Srivastava
D
,
Leisenring
WM
, et al
The cyclophosphamide equivalent dose as an approach for quantifying alkylating agent exposure: a report from the Childhood Cancer Survivor Study
.
Pediatr Blood Cancer
2014
;
61
:
53
67
.
14.
Department of Reproductive Health and Research W
.
WHO laboratory manual for the examination and processing of human semen, Fifth edition
.
Geneva, Switzerland
:
World Health Organization
; 
2010
.
15.
Bollendorf
A
,
Check
JH
,
Lurie
D
. 
Evaluation of the effect of the absence of sperm with rapid and linear progressive motility on subsequent pregnancy rates following intrauterine insemination or in vitro fertilization
.
J Androl
1996
;
17
:
550
7
.
16.
Comhaire
FH
,
Vermeulen
L
,
Hinting
A
,
Schoonjans
F
. 
Accuracy of sperm characteristics in predicting the invitro fertilizing-capacity of semen
.
J In Vitro Fertil Em
1988
;
5
:
326
31
.
17.
Bhasin
S
,
Brito
JP
,
Cunningham
GR
,
Hayes
FJ
,
Hodis
HN
,
Matsumoto
AM
, et al
Testosterone therapy in men with hypogonadism: an Endocrine Society* Clinical Practice Guideline
.
J Clin Endocr Metab
2018
;
103
:
1715
44
.
18.
Chemaitilly
W
,
Liu
Q
,
van Iersel
L
,
Ness
KK
,
Li
Z
,
Wilson
CL
, et al
Leydig cell function in male survivors of childhood cancer: a report from the St Jude Lifetime Cohort Study
.
J Clin Oncol
2019
;
37
:
3018
31
.
19.
Guenther
PM
,
Kirkpatrick
SI
,
Reedy
J
,
Krebs-Smith
SM
,
Buckman
DW
,
Dodd
KW
, et al
The Healthy Eating Index-2010 is a valid and reliable measure of diet quality according to the 2010 Dietary Guidelines for Americans
.
J Nutr
2014
;
144
:
399
407
.
20.
Sapkota
Y
,
Cheung
YT
,
Moon
W
,
Shelton
K
,
Wilson
CL
,
Wang
Z
, et al
Whole-genome sequencing of childhood cancer survivors treated with cranial radiation therapy identifies 5p15.33 locus for stroke: a report from the St. Jude Lifetime Cohort Study
.
Clin Cancer Res
2019
;
25
:
6700
8
.
21.
Sapkota
Y
,
Turcotte
LM
,
Ehrhardt
MJ
,
Howell
RM
,
Arnold
MA
,
Wilson
CL
, et al
Genome-wide association study in irradiated childhood cancer survivors identifies HTR2A for subsequent basal cell carcinoma
.
J Invest Dermatol
2019
;
139
:
2042
5
.
22.
Green
DM
,
Zhu
L
,
Wang
M
,
Chemaitilly
W
,
Srivastava
D
,
Kutteh
WH
, et al
Effect of cranial irradiation on sperm concentration of adult survivors of childhood acute lymphoblastic leukemia: a report from the St. Jude Lifetime Cohort Study
.
Hum Reprod
2017
;
32
:
1192
201
.
23.
Willer
CJ
,
Li
Y
,
Abecasis
GR
. 
METAL: fast and efficient meta-analysis of genomewide association scans
.
Bioinformatics
2010
;
26
:
2190
1
.
24.
Cochran
WG
. 
The combination of estimates from different experiments
.
Biometrics
1954
;
10
:
101
29
.
25.
Ioannidis
JPA
,
Patsopoulos
NA
,
Evangelou
E
. 
Heterogeneity in meta-analyses of genome-wide association investigations
.
PLoS One
2007
;
2
:
e841
.
26.
Han
B
,
Eskin
E
. 
Random-effects model aimed at discovering associations in meta-analysis of genome-wide association studies
.
Am J Hum Genet
2011
;
88
:
586
98
.
27.
Kelsey
TW
,
McConville
L
,
Edgar
AB
,
Ungurianu
AI
,
Mitchell
RT
,
Anderson
RA
, et al
Follicle Stimulating Hormone is an accurate predictor of azoospermia in childhood cancer survivors
.
PLoS One
2017
;
12
:
e0181377
.
28.
Green
DM
,
Zhu
L
,
Zhang
N
,
Sklar
CA
,
Ke
RW
,
Kutteh
WH
, et al
Lack of specificity of plasma concentrations of inhibin B and follicle-stimulating hormone for identification of azoospermic survivors of childhood cancer: a report from the St Jude Lifetime Cohort Study
.
J Clin Oncol
2013
;
31
:
1324
8
.
29.
Anderson
RA
,
Wallace
EM
,
Groome
NP
,
Bellis
AJ
,
Wu
FC
. 
Physiological relationships between inhibin B, follicle stimulating hormone secretion and spermatogenesis in normal men and response to gonadotrophin suppression by exogenous testosterone
.
Hum Reprod
1997
;
12
:
746
51
.
30.
Lomelino
CL
,
Andring
JT
,
McKenna
R
,
Kilberg
MS
. 
Asparagine synthetase: function, structure, and role in disease
.
J Biol Chem
2017
;
292
:
19952
8
.
31.
Buniello
A
,
MacArthur
JAL
,
Cerezo
M
,
Harris
LW
,
Hayhurst
J
,
Malangone
C
, et al
The NHGRI-EBI GWAS Catalog of published genome-wide association studies, targeted arrays and summary statistics 2019
.
Nucleic Acids Res
2019
;
47
:
D1005
12
.