DNA methylation of a novel PAK4 locus influences ototoxicity susceptibility following cisplatin and radiation therapy for pediatric embryonal tumors

Abstract

Background

Ototoxicity is a common adverse side effect of platinum chemotherapy and cranial radiation therapy; however, individual susceptibility is highly variable. Therefore, our objective was to conduct an epigenome-wide association study to identify differentially methylated cytosine-phosphate-guanine (CpG) sites associated with ototoxicity susceptibility among cisplatin-treated pediatric patients with embryonal tumors.

Methods

Samples were collected for a discovery cohort (n = 62) and a replication cohort (n = 18) of medulloblastoma and primitive neuroectodermal tumor patients. Posttreatment audiograms were evaluated using the International Society of Paediatric Oncology (SIOP) Boston Ototoxicity Scale. Genome-wide associations between CpG methylation and ototoxicity were examined using multiple linear regression, controlling for demographic and treatment factors.

Results

The mean cumulative dose of cisplatin was 330 mg/m² and the mean time from end of therapy to the last available audiogram was 6.9 years. In the discovery analysis of 435233 CpG sites, 6 sites were associated with ototoxicity grade (P < 5 × 10⁻⁵) after adjusting for confounders. Differential methylation at the top CpG site identified in the discovery cohort (cg14010619, PAK4 gene) was replicated (P = 0.029) and reached genome-wide significance (P = 2.73 × 10⁻⁸) in a combined analysis. These findings were robust to a sensitivity analysis evaluating other potential confounders.

Conclusions

We identified and replicated a novel CpG methylation loci (cg14010619) associated with ototoxicity severity. Methylation at cg14010619 may modify PAK4 activity, which has been implicated in cisplatin resistance in malignant cell lines.

Importance of the study

Pediatric patients treated for embryonal tumors experience high rates of ototoxicity, which leads to diminished academic performance and quality of life. While the major drivers of ototoxicity are cisplatin chemotherapy and cranial radiation therapy, there remains significant variability in hearing loss among these patients. As other established modifiers of ototoxicity only explain a small fraction of the risk, we sought to evaluate the role of DNA methylation, an important epigenetic regulator of gene expression, on susceptibility to ototoxicity. Specifically, we conducted the first epigenome-wide association study of ototoxicity susceptibility among cisplatin-treated pediatric patients with embryonal tumors. We identified and replicated differential DNA methylation in cg14010619, which appears to regulate expression of the p21 activating kinase 4 (PAK4) gene. Notably, PAK4 expression is associated with chemotherapy resistance and has been identified as a potential therapeutic target for other malignancies. This study provides the first evidence that epigenetic modification of PAK4 may affect susceptibility to treatment-related ototoxicity.

Cisplatin is a key chemotherapeutic agent used in the treatment of various solid tumors; however, exposure to cisplatin is associated with substantial toxicity, including ototoxicity. Cisplatin ototoxicity typically manifests as bilateral sensorineural hearing impairment that is permanent and progressive in nature. Children and adolescents are particularly vulnerable to the adverse effects of treatment-related hearing impairment, with greater than half of cisplatin-treated children developing moderate to severe ototoxicity.¹ Hearing impairment in children is associated with poor academic performance, social-emotional dysfunction, and decreased quality of life.^2,3 Despite the serious long-term complications linked to cisplatin-induced hearing impairment and the prevalence of these adverse outcomes, few modifying factors have been identified to inform prevention strategies and precision medicine efforts in this population.

New approaches are needed to better manage cisplatin ototoxicity. Currently, standard treatment protocols include monitoring hearing thresholds and adjusting cisplatin therapy if excessive deterioration of hearing is detected. When cisplatin ototoxicity is detected, dose reductions and treatment delays may reduce treatment efficacy. Even if detected early, hearing impairment is typically permanent and treatment modification will not restore normal hearing. Several potential otoprotective agents have been evaluated in randomized controlled trials, including amifostine, sodium thiosulfate, dexamethasone, and N-acetylcysteine. While each agent initially demonstrated some protective potential,^4–6 other studies have raised concerns over the limited protection afforded by these agents and their possible interaction with anticancer therapies.^7–9 As a result, there are currently no FDA-approved treatments to reduce the risk of ototoxicity in patients treated with cisplatin. The lack of effective otoprotective treatment options highlights the need for new insight into biomarkers of cisplatin-induced ototoxicity.

Cumulative cisplatin dose, exposure to additional ototoxic therapy (ie, aminoglycosides, loop diuretics, carboplatin, and cranial radiation therapy), and younger age at treatment are well-established risk factors for cisplatin-induced ototoxicity.¹⁰ However, considerable variability exists in ototoxicity susceptibility even among similarly treated patients. Heritable factors explain nearly half of the variability in cisplatin sensitivity in lymphoblastoid cell lines.¹¹ In fact, nearly 45% of the variation in cisplatin cytotoxicity is attributed to the influence of common genetic variants on gene transcription in experimental cell lines.¹² Numerous studies have identified candidate single nucleotide polymorphisms associated with cisplatin-induced hearing impairment in patient populations using candidate gene, pathway, or genome-wide approaches.^13–15 The observed association between these genetic variants and ototoxicity risk has been somewhat inconsistent across studies and the actual impact of many of these genes remains unclear.¹⁶ A genome-wide association study recently identified and replicated a single locus in the ACYP2 gene associated with ototoxicity susceptibility¹⁷; however, the risk allele was present in fewer than 15% of patients who developed ototoxicity, and nearly 60% of patients without the risk allele still experienced hearing impairment. More work is needed to understand the role of genetic variation in cisplatin-induced ototoxicity susceptibility, while alternative biologic mechanisms must also be evaluated to better predict these adverse treatment-related outcomes.

Epigenetic mechanisms are important regulators of gene activity. DNA methylation, the most widely studied epigenetic mechanism, involves the methylation or demethylation of cytosine bases at cytosine-guanine dinucleotides (CpG). Because DNA methylation plays a central role in gene expression, variation in DNA methylation may explain variation in the response to environmental stimuli, including treatment exposures. Therefore, the objective of this study is to identify epigenetic profiles associated with ototoxicity susceptibility among cisplatin-treated pediatric embryonal brain tumor patients using an epigenome-wide approach.

Materials and Methods

Study Population

Study participants were treated at Texas Children’s Cancer Center, Houston, between 2005 and 2012. The Epidemiology Program at Texas Children’s Cancer Center enrolls patients in research studies, administers questionnaires, and obtains biologic samples from willing participants. Peripheral blood samples were initially collected on most patients (2005–2009), though less invasive saliva samples were frequently collected on more recently enrolled patients (2010–2012). In this study, eligible patients were those with biologic samples available, who were less than 18 years of age at diagnosis, treated for medulloblastoma or primitive neuroectodermal tumor with cisplatin-containing protocols, presented with normal hearing at diagnosis, and received at least one posttreatment (≥6 months following treatment cessation) auditory assessment. To avoid potential confounding that may arise when evaluating DNA methylation from a mixed sample of older peripheral blood samples and more recently collected saliva samples, the study population was divided into a discovery cohort (n = 62) of eligible patients with DNA derived from peripheral blood samples and an independent replication cohort (n = 18) with DNA derived from saliva samples. Written informed consent was obtained from participants or legal guardians. The study protocol was reviewed and approved by the institutional review board at Baylor College of Medicine.

Outcome Ascertainment

Since 2004,¹⁸ the Children’s Oncology Group has published guidelines recommending routine auditory assessments for all survivors of pediatric malignancies exposed to cisplatin chemotherapy.¹⁰ Patients included in this study had hearing thresholds evaluated by trained pediatric audiologists. Most patients received a pure-tone air conduction hearing test. Bone conduction was assessed in patients with abnormal tympanograms or to confirm hearing impairment detected with pure-tone air conduction testing. Audiograms were abstracted from electronic medical records. Approximately 10% of the abstracted audiograms were evaluated by a second blinded investigator and no discrepancies were detected between audiograms abstracted by the 2 independent investigators. Audiograms were assigned a grade according to the International Society of Paediatric Oncology’s (SIOP) Boston Ototoxicity Scale.¹⁹ The left and right ears were scored separately. Ears with a hearing threshold <20 dB across all frequencies received a grade of 0; ears with a hearing threshold >20 dB at ≥6 kHz received a grade of 1; ears with a hearing threshold of >20 dB at ≥4 kHz received a grade of 2; ears with a hearing threshold of >20 dB at ≥2 kHz received a grade of 3; and ears with a hearing threshold of >40 dB at ≥2 kHz received a grade of 4. Each patient was assigned a single grade equal to the highest grade from either the left or right ear. The last available audiogram for each patient was treated as the primary endpoint for the current study.

Covariate Data Collection

Trained research coordinators administered questionnaires to collect guardian- or patient-reported demographic (ie, race, ethnicity), medical history, and family information (ie, residential history, parental occupation). Research coordinators also abstracted clinical patient characteristics from electronic medical records, including age, gender, cumulative cisplatin chemotherapy received, and exposure to amifostine therapy. Radiation dose to the craniospinal axis was used to approximate cochlear exposure to radiation, and individuals were assigned to high (≥34 Gy) and low (<26 Gy) dose categories.

Sample Collection and Processing

Peripheral blood samples were collected by clinical phlebotomists following treatment cessation for individuals in the discovery cohort. DNA was extracted from peripheral blood samples using the Gentra Puregene Blood Kit (Qiagen). Trained research coordinators collected posttreatment saliva samples from participants in the replication cohort using the Oragene saliva collection kit (DNA Genotek). DNA from blood and saliva was bisulfite converted using the EZ DNA Methylation Kit (Zymo Research) following manufacturer’s instructions.

DNA Methylation Data Processing and Quality Control

Genome-wide DNA methylation at CpG sites was measured using the Illumina Infinium HumanMethylation450 BeadChip Kit. Quality control and preprocessing of the raw intensity data were performed separately for the discovery and replication sets as well as the combined cohort using Methylumi and ChAMP Bioconductor packages.^20,21 Prior to analysis, we excluded CpG sites with a detection P-value >0.01 in greater than 1% of the discovery cohort samples (n = 1731), a bead count <3 in greater than 5% of the discovery cohort samples (n = 87), CpG sites associated with previously identified single nucleotide polymorphisms (n = 28762), cross-reactive CpG probes (n = 8506), and non-autosomal CpG sites (n = 11193).²² Among the remaining 435233 CpG sites which passed the initial quality filtering, beta-mixture quantile normalization was used to correct for possible bias resulting from the design of type I and type II probes.²³ To reduce the potential influence of technical artifacts, the ComBat method was used to correct for batch effects.²⁰ Finally, to explore the possible impact of interindividual differences in the cellular composition of blood samples, validated algorithms were used to estimate the heterogeneity of each sample (eg, proportion of lymphocytes, monocytes).²⁴ The same criteria used in the discovery cohort for CpG filtering, processing, and quality control were applied to the replication set. The batch-corrected, normalized β-values were retained for statistical analyses.

Data Analysis

Descriptive statistics, including medians and interquartile ranges for continuous variables and counts and proportion of the total for categorical variables, were calculated for the discovery and replication cohorts. Methylation β-values were transformed to improve normality (M-values) prior to statistical modeling. The associations between M-values and ototoxicity grade were evaluated using multiple linear regression, considering M-values as the dependent variable. A test for linear trend was conducted between each CpG site and ototoxicity by treating ototoxicity grade as a continuous variable (range: 0–4) in regression models. Clinical and demographic factors which were associated with ototoxicity grade (P < 0.20) in univariate models were included as covariates in multiple linear regression models to control for confounding. Any CpG site that was associated with ototoxicity below the P = 5.0 × 10⁻⁵ significance threshold in adjusted models was considered a candidate to be evaluated in the replication cohort. CpG sites which were nominally associated with ototoxicity (P < 0.05) in the replication cohort and retained the same direction of association as the discovery cohort were considered replicated. The discovery and replication cohorts were then combined to evaluate the overall contribution of methylation at replicated CpG sites in the entire study population. A principal components analysis of the CpG sites which passed quality control and filtering was conducted in the combined cohort to identify major sources of variation in the methylation data (Supplementary Figures S1–S2). The top two principal components were retained to adjust for these sources of variation in regression models. A Bonferroni-corrected P-value was calculated to account for multiple comparisons, with P < 1.15 × 10⁻⁷ indicating genome-wide statistical significance. Regression diagnostics (eg, linearity, residual normality, influential observations) were performed for the top CpG sites. No violations of the regression assumptions were identified.

Results

Hearing thresholds were evaluated with pure-tone air conduction alone in 76.3% of patients (n = 61), bone conduction alone in 1.2% (n = 1), or a combination in 22.5% (n = 18). The demographic and treatment characteristics of individuals included in the discovery and replication cohorts are presented in Table 1. In the discovery cohort (n = 62), 72.6% of participants were male, 48.4% non-Hispanic white, 30.6% Hispanic, 91.1% with diagnosis of medulloblastoma (91.9%), and 67.7% treated between 3 and 12 years of age. Patients included in this study were treated on or according to SJMB96, SJMB03, CCG-9961, or similar regimens consisting of cisplatin, vincristine, and cyclophosphamide with or without etoposide. All patients received craniospinal radiation and cisplatin chemotherapy (median cumulative dose = 300 mg/m², range: 100–720 mg/m²). Moderate to severe ototoxicity was common among participants in the discovery cohort, with nearly half (46.8%) developing grade 3–4 ototoxicity. The only significant differences (P < 0.05) between the discovery cohort and the replication cohort were on mean age at sample collection and mean age at last audiogram. Specifically, as participants included in the discovery sample were treated and recruited onto the study prior to the replication set, the median time between diagnosis and last available audiogram is significantly less (P < 0.01) for the replication cohort (median = 4.4 y) than the discovery cohort (median = 6.6 y). There were also slight differences in treatment exposures between the cohorts, such as an increase in amifostine use (discovery = 40.3%, replication = 66.7%, P = 0.05) and a decrease in the prevalence of high-dose craniospinal radiation (discovery = 33.9%, replication = 17.7%, P = 0.20), which likely reflect temporal changes in treatment regimens. Similarly, the lower prevalence of grade 3–4 ototoxicity detected in the replication cohort (26.8%) may be an artifact of the shorter time elapsed between treatment and auditory evaluation, increased exposure to potential otoprotective therapy (ie, amifostine), reduced exposure to ototoxic therapy (ie, high-dose radiation), or a combination of these factors. Differences in demographic and treatment characteristics by SIOP Boston Ototoxicity grade are presented in Supplementary Table S1.

Table 1.

Demographic and clinical characteristic of discovery (n = 62) and replication (n = 18) cohorts of cisplatin-treated pediatric brain tumor patients

	Discovery Set (n = 62)	Replication Set (n = 18)
	Median (IQR)	Median (IQR)	P
Age at diagnosis, y	6.8 (4.6–11.1)	6.8 (3.7–10.1)	0.80
Age at sample, y	12.2 (7.9–14.9)	9.4 (5.5–12.0)	0.03
Age at last audiogram, y	15.0 (12.2–18.2)	11.8 (9.1–13.5)	<0.01
Time to last audiogram from diagnosis, y	6.6 (4.0–10.9)	4.4 (2.0–6.0)	<0.01
Cumulative cisplatin, mg/m2	300 (300–375)	300 (300–332)	0.51
	N (%)	N (%)
Gender			0.63
Female	17 (27.4)	6 (33.3)
Male	45 (72.6)	12 (66.7)
Race/ethnicity			0.63
White	40 (48.4)	11 (61.1)
Hispanic	19 (30.6)	4 (22.2)
Other	13 (21.0)	3 (16.7)
Craniospinal radiation			0.20
<26 Gy	41 (66.1)	14 (82.4)
≥34 Gy	21 (33.9)	3 (17.7)
Amifostine			0.05
None	37 (59.7)	6 (33.3)
Yes	25 (40.3)	12 (66.7)
Disease			0.28
PNET	5 (8.1)	3 (16.7)
Medulloblastoma	57 (91.9)	15 (83.3)
SIOP ototoxicity grade			0.22
0	8 (12.9)	4 (22.2)
1	19 (30.6)	5 (27.8)
2	6 (9.7)	4 (22.2)
3	13 (21.0)	4 (22.2)
4	16 (25.8)	1 (4.6)

Abbreviations: IQR, interquartile range; PNET, primitive neuroectodermal tumor.

P-value calculated from Wilcoxon–Mann–Whitney test or Fisher’s exact test.

Of the 435233 CpG sites which passed filtering and quality control in the discovery cohort (n = 62), 6 sites were associated with ototoxicity grade at P < 5 × 10⁻⁵ (Fig. 1) after adjusting for gender, amifostine therapy, craniospinal radiation dose, age at diagnosis, and age at follow-up. We attempted to confirm the observed associations in the replication cohort (n = 18). Of the 6 candidate methylation sites, one differentially methylated position was nominally associated with ototoxicity grade (P < 0.05) and retained the same direction of association in the replication cohort (Table 2). Specifically, methylation at cg14010619 was significantly and inversely associated (Fig. 2) with ototoxicity grade at the genome-wide level (P = 2.74 × 10⁻⁸) in an adjusted analysis of the combined cohorts (n = 80). The observed inverse association between ototoxicity and methylation at cg14010619 was robust to a sensitivity analysis adjusting for other potential confounders, including cumulative cisplatin dose, race/ethnicity, cellular composition of blood samples, and the first 2 principal components from a principal components analysis to account for variation in methylation across arrays (Supplementary Table S2). The cg14010619 CpG site is located on chromosome 19q and is positioned within a CpG island shelf associated with the p21-activating kinase 4 gene (PAK4). Because mRNA expression data were not generated for the current study, we evaluated the association between methylation at cg14010619 and PAK4 expression using data from the Gene Expression Omnibus (see Supplementary material). In these data, increasing cg14010619 methylation levels significantly (P = 0.002) correlated with higher levels of PAK4 transcription (Supplementary Figure S3).

Fig. 1.

Manhattan plot of association¹ between CpG methylation and ototoxicity grade in discovery cohort (n = 62) of cisplatin-treated survivors of pediatric brain tumors.

¹Reference line at threshold considered statistically significant for discovery cohort (P < 5 × 10⁻⁵).

Table 2.

Associations between methylation at top candidate CpG (P < 5 × 10⁻⁵ in discovery cohort) and SIOP ototoxicity grade in discovery, replication, and combined cohorts

			Discovery Cohort (n = 62)1			Replication Cohort (n = 18)			Combined Cohort (n = 80)2
CpG Site	Chr	Gene	β	t-Stat	P	β	t-Stat	P	β	t-Stat	P
cg14010619	19q	PAK4	−1.06	−5.63	6.38 × 10−7	−0.69	−2.38	0.029	−1.01	−6.25	2.74 × 10−8
cg05053403	12q	FBRSL1	−2.97	−5.39	1.52 × 10−6	1.45	2.08	0.053	−2.08	−3.36	1.27 × 10−3
cg04135540	4p	SHISA3	1.46	5.38	1.56 × 10−6	−1.55	−1.15	0.268	0.81	3.19	2.12 × 10−3
cg26771097	16p	MSLN	0.98	5.27	2.34 × 10−6	−2.06	−2.42	0.027	0.38	2.27	2.62 × 10−2
cg06566615	17q	TANC2	0.63	5.13	3.87 × 10−6	−0.57	−0.74	0.470	0.39	3.51	7.79 × 10−4
cg01299082	4q	HSP90AA4P	1.15	5.07	4.75 × 10−6	−0.69	−0.46	0.653	0.73	3.62	5.53 × 10−4

Abbreviations: β, regression coefficient.

¹Model adjusted for age at diagnosis, age at last follow-up, amifostine therapy, craniospinal radiation, and gender.

²Model adjusted for age at diagnosis, age at last follow-up, amifostine therapy, craniospinal radiation, gender, and cohort.

Fig. 2.

Predicted mean cg14010619 methylation level (95% CI) by ototoxicity grade.

Mean and 95% CIs for predicted methylation at cg14010619 by categories of ototoxicity grade in overall cohort (n = 80), adjusted for age at diagnosis, age at last follow-up, amifostine therapy, craniospinal radiation, gender, and cohort.

Discussion

We identified and replicated a novel DNA methylation locus associated with ototoxicity susceptibility among pediatric embryonal brain tumor patients treated with cisplatin-containing chemotherapy regimens. DNA methylation is an important epigenetic regulator of gene transcription. In this study, methylation levels at cg14010619 were inversely associated with ototoxicity severity, suggesting that PAK4 gene expression may impact treatment-related ototoxicity susceptibility. Notably, there were no adjacent loci associated with ototoxicity. However, the correlation between cg14010619 methylation and methylation at other CpG sites in the region was very low (median = 0.068, range: 0.001–0.283). While the current study utilized an epigenome-wide approach, it is possible that the array used in this analysis did not provide adequate coverage for the identification of a differentially methylated region that included PAK4.

Paired gene expression data were not available for the participants included in this study; however, analysis of publicly available data supports a link between cg14010619 methylation and PAK4 mRNA expression in peripheral blood samples (Supplementary material; Supplementary Figure S1). While these results provide some evidence of a possible relationship between epigenetic mechanisms, downstream PAK4 gene expression, and subsequent toxicity, additional research is needed to clarify potentially important tissue-specific difference in DNA methylation and PAK4 activity.

The PAK4 gene encodes a serine/threonine kinase, which is involved in a variety of biological pathways, including actin cytoskeleton organization and Ras signaling. PAK4 is expressed by the inner and outer cochlear hair cells.²⁵ These cells are characterized by the formation of stereociliary bundles, which function as the mechanotransduction system involved in sound detection. Notably, PAKs are important regulators of stereociliary bundle migration, orientation, and positioning during organ of Corti development.^26,27 Chemical inhibition of PAK impedes stereociliary bundle morphogenesis, resulting in abnormal fragmented bundles.²⁸ These studies highlight the critical role PAK plays in the development and function of sensory hair cells, the primary suspected target of cisplatin-induced ototoxicity.

The expression of PAK4 has also been implicated in numerous human malignancies and may contribute to tumorigenesis, increased tumor invasion, and reduced chemosensitivity. For example, PAK4 overexpression has been associated with poor outcomes in ovarian cancer,²⁹ breast cancer,³⁰ non-small-cell lung cancer,³¹ and clear cell renal cell carcinoma.³²PAK4 expression also appears to bolster chemotherapy resistance^29,33 through the upregulation of the nuclear factor-kappaB pro-survival pathway via activation of Akt and extracellular signal-regulated kinase.^34,35PAK4 expression was recently linked to cisplatin resistance in cervical and gastric cancer cell lines, while PAK4 inhibition restored cisplatin cytotoxicity in these cells.^36–38 Because of its apparent role in cancer progression and treatment resistance, PAK4 has been identified as an attractive therapeutic target. Our findings also suggest that PAK4 may affect individual susceptibility to treatment-related toxicity. Specifically, we identified a differentially methylated CpG site associated with the PAK4 gene that influences ototoxicity severity among cisplatin-treated patients. Methylation levels at cg14010619 were inversely associated with ototoxicity grade in our population of pediatric brain tumor patients. These results are consistent with the hypothesis that PAK4 expression confers protection against therapy-directed damage to both tumor and normal host tissue.

This study has several notable strengths, including the recruitment of a well-characterized cohort of pediatric patients with similar medical histories, treated under contemporary protocols, at high risk of ototoxicity, with extensive auditory follow-up and monitoring. Still, the results of this study should be considered in light of several limitations. First, this study evaluated DNA methylation profiles in blood and saliva samples, which may not reflect the epigenetic profiles of the tissue directly implicated in ototoxicity. Previous research, however, has demonstrated that DNA methylation profiles are highly conserved across multiple tissues.^39–41 Still, additional work is needed to characterize the role of epigenetic networks and PAK4 expression in cochlear hair cells and other systems directly involved in treatment-related hearing impairment. Second, the observed association between DNA methylation and ototoxicity susceptibility may be attributed to the actions of cisplatin, radiation, or a combination of the two. Most cases identified in this study showed evidence of sensorineural hearing loss; however, hearing impairment following ototoxic therapy, particularly radiation therapy, is often complex and may manifest as sensorineural, conductive, or mixed hearing loss.⁴² As a result, these findings may not be generalizable to populations receiving only cisplatin chemotherapy or cranial radiation therapy. Furthermore, while this is a relatively large cohort of this rare condition (ie, pediatric brain tumors), this study was restricted to a population of 80 pediatric patients (discovery cohort n = 62, replication cohort n = 18). As a result, this analysis was likely underpowered to detect modest differences in DNA methylation across several potentially important CpG sites. Expanding the inclusion criteria to capture additional cancer diagnoses or older individuals would increase power at the expense of possibly introducing confounding due to heterogeneous treatment exposures (eg, different cisplatin and radiation dose or treatment schedules, exposure to loop diuretics or aminoglycosides, exposure to compounds which modify ototoxicity susceptibility) and other sources of hearing damage (eg, noise exposure, age-related hearing loss, other ototoxic medications). Despite the limited sample size, we identified a DNA methylation locus significantly associated with ototoxicity grade even after accounting for multiple comparisons.

Conclusions

Ototoxicity is a common and debilitating consequence of cisplatin chemotherapy and craniospinal radiation therapy, resulting in long-term auditory complications and reduced quality of life. The absence of established otoprotective therapy and inability of ototoxicity monitoring programs to prevent hearing impairment underscore the need for new approaches to improve our understanding of the pathobiology of treatment-related toxicity and identify high-risk patients who may benefit from targeted interventions. This epigenome-wide association study identified and replicated a novel association between DNA methylation in the PAK4 gene and ototoxicity grade among cisplatin-treated survivors of pediatric embryonal brain tumors. PAK4 has recently been identified as a potential therapeutic target due to its association with tumorigenesis and treatment resistance in various malignancies.^36–38 The results of this study suggest that PAK4 expression may also confer some protection against treatment-related ototoxicity. Targeting PAK4, therefore, may have broader implications for both cancer and normal tissue treatment response. With the ultimate goal of achieving an acceptable balance between treatment efficacy and toxicity, this work highlights the need to conduct additional research and functional work to establish the biological significance of PAK4 DNA methylation and the role of PAK4 expression in evaluating therapeutic responses. This direction of research may eventually uncover powerful biomarkers of treatment-associated ototoxicity. Recognizing the need to reduce treatment-associated toxicity, clinical trials are currently investigating the impact of less intensive cranial radiation therapy and chemotherapy in medulloblastoma patients defined as low-risk based on clinical risk and tumor molecular subgroups. While less intensive therapy may reduce neurocognitive sequelae, it may compromise treatment efficacy in patients who could tolerate more intensive therapy. Ultimately, approaches to distinguish patients who can tolerate intensive treatment from patients who are susceptible to toxicity may result in clinical improvements by providing an opportunity to intervene or tailor treatment without sacrificing efficacy.

Supplementary Material

Supplementary material is available at Neuro-Oncology online.

Funding

This study was supported by funding from the National Cancer Institute (R25CA160078, PI: Scheurer) and Wipe Out Kids’ Cancer (PI: Murray).

Conflict of interest statement. All authors declare no conflict of interest.