Genetic variants predictive of chemotherapy-induced peripheral neuropathy symptoms in gynecologic cancer survivors

Abstract

Objective:

To identify genetic variants associated with chemotherapy-induced peripheral neuropathy (CIPN) symptoms among gynecologic cancer survivors and determine the variants’ predictive power in addition to age and clinical factors at time of diagnosis.

Methods:

Participants of a prospective cohort study on gynecologic cancers provided a DNA saliva sample and reported CIPN symptoms (FACT/GOG-Ntx). Genotyping of 23 single nucleotide polymorphisms (SNPs) previously identified as related to platinum- or taxane-induced neuropathy was performed using iPLEX Gold method. Risk allele carrier frequencies of 19 SNPs that passed quality checks were compared between those with/without high CIPN symptoms using logistic regression, adjusting for age. Receiver operating characteristic (ROC) curves using clinical risk factors (age, diabetes, BMI, Charlson Comorbidity Index, previous cancer diagnosis) with and without the identified SNPs were compared.

Results:

107 individuals received platinum or taxane-based chemotherapy and provided sufficient DNA for analysis. Median age was 65.1 years; 39.6% had obesity and 8.4% diabetes; most had ovarian (58.9%) or uterine cancer (29.0%). Two SNPs were significantly associated with high CIPN symptomatology: rs3753753 in GPX7, OR=2.55 (1.13, 5.72) and rs139887 in SOX10, 2.66 (1.18, 6.00). Including these two SNPs in a model with clinical characteristics led to an improved AUC for CIPN symptomatology (0.65 vs. 0.74, p=0.04).

Conclusions:

Genetic and clinical characteristics were predictive of higher CIPN symptomatology in gynecologic cancer survivors, and combining these factors resulted in superior predictive power compared with a model with clinical factors only. Prospective validation and assessment of clinical utility are warranted.

Introduction

One of the most common and debilitating side effects of treatment for gynecologic cancers is chemotherapy-induced peripheral neuropathy (CIPN) (1). Platinum and taxane-based regimens, both frequently implicated in CIPN, are commonly used for treatment of gynecologic malignancies (2). Conservative estimates suggest up to 36% of older patients with ovarian cancer and 20% of patients under the age of 70 experience grade 2 or higher peripheral neuropathy during and following treatment (3). Prevalence is reported to be even higher in some studies; a recent international, internet-based survey of patients with ovarian cancer who had received chemotherapy found 78% of patients report symptoms of peripheral neuropathy (4). Further, almost half of female cancer survivors report persistent or worsening neuropathy more than 5 years after treatment, leading to worsening physical function, increased disability, increased risk of falling, and lower quality of life (5,6). There are currently no established effective preventative methods and limited treatments available for CIPN (7); but with a growing population of over 1.3 million survivors of gynecologic cancer in the United States, addressing treatment-related CIPN preventatively is imperative (8–10).

Polymorphisms in genes relevant to drug metabolism, transport and excretion can influence a patient’s sensitivity to chemotherapeutic agents (11). This has led to a growing body of literature evaluating single nucleotide polymorphisms (SNPs) and their association with susceptibility to CIPN (12). To date, most of these studies have evaluated patients diagnosed with lung, breast or colorectal cancers (13). Our objective was to determine if genetic variants previously implicated in platinum- or taxane-induced neuropathy could help predict the development of CIPN in a cohort of gynecologic cancer survivors.

Methods

Study Design and Population

We conducted a secondary analysis of a prospective cohort study. Adults aged 18 years and older with a history of a gynecologic cancer (ovarian, uterine, cervical, vaginal or vulvar cancer) who could read/write in English were recruited at time of diagnosis or during surveillance from the University of Minnesota into the Gynecologic Oncology: Life after Diagnosis (GOLD) study. Detailed methods of this prospective cohort study are described elsewhere (14). All participants who were alive and remaining in the GOLD study as of July 2019 (N=338) were mailed information about participating in an optional saliva DNA sample collection for future studies. Eligible participants were mailed introductory letters and Oragene 600 collection kits. All participants provided written informed consent at the time of enrollment in both the cohort and again at the time of their saliva sample collection. The study received approval from the University of Minnesota Institutional Review Board (#1612S01581).

Survey Measures

Participants in the GOLD study complete baseline and semiannual surveys evaluating numerous measures of quality of life, psychosocial outcomes, and physical symptoms following a cancer diagnosis. Self-reported neuropathy symptoms were collected at baseline, and at 6, 18 and 30 months using the validated FACT/GOG-Ntx subscale (15,16). Scores were calculated using standard methods; however, the scores were not reverse coded; higher scores indicate greater CIPN symptomology. The maximum neuropathy score from among all follow-up surveys was used to determine neuropathy symptomatology. There is no established cutoff for this measure for significant or clinically relevant neuropathy symptoms on the FACT/GOG-Ntx subscale. We classified individuals as having high neuropathy symptomatology based on the median score in our sample (<10 versus 10+).

Age and clinical data from the time of diagnosis, including disease site, International Federation of Gynecology and Obstetrics (FIGO) stage, and treatments received (chemotherapy, surgery and/or radiation) were abstracted from patients’ electronic health records. Number of cycles of chemotherapy and regimens received were collected up to the date of report of maximum neuropathy symptoms. We also collected diagnosis of diabetes prior to cancer diagnosis, previous cancer experience and other comorbidities to calculate the Charlson Comorbidity Index score (17,18).

SNP Selection, Genotyping, and Quality Control

Between September 2019 and January 2020, a literature review was performed in PubMed database using search terms: “SNP chemotherapy neuropathy” and “single nucleotide polymorphism chemotherapy neuropathy”. Search results in PubMed database were reviewed for publications specifically evaluating chemotherapy-induced neuropathy in patients who had received platinum or taxane chemotherapy regimens. Publications in which SNPs were found to have a statistically significant association with chemotherapy-induced neuropathy were extracted. We identified 23 SNPs in 15 genes to pursue for analysis. Genomic coordinates were aligned to GRCh38 assembly hgx38.

Returned saliva samples were processed by the University of Minnesota Genomics Center. DNA isolation from the saliva was performed per the manufacturer’s protocol (DNA Genotek, USA) and quantified by both Nanodrop UV/VIS spectrophotometry (ThermoFisher, USA) and fluorometry using PicoGreen staining (BioTek, USA). Genotyping was performed using the iPLEX Gold method. iPLEX reagents and protocols for multiplex PCR, single base primer extension (SBE) and generation of mass spectra were used, as per the manufacturer’s instructions. Multiplexed PCR for the 23 SNPs was performed in 5-μl reactions on 384-well plates containing 10 ng of genomic DNA. Reactions contained 0.5 U HotStar Taq polymerase (QIAGEN), 100 nM primers, 1.25X HotStar Taq buffer, 1.625 mM MgCl2, and 500 nM dNTPs. Following enzyme activation at 95 °C for 5 min, DNA was amplified with 45 cycles of 95 °C × 20 sec, 56 °C × 30 sec, 72 *C × 1 min, followed by a 3-min extension at 72 °C. Unincorporated dNTPs were removed using shrimp alkaline phosphatase (0.3 U, Agena). Single-base extension was carried out by addition of SBE primers at concentrations from 0.625 μM (low MW primers) to 1.25 μM (high MW primers) using iPLEX enzyme and buffers (Agena, San Diego) in 9-μl reactions. Reactions were desalted and SBE products measured using the MassARRAY system, and mass spectra analyzed using TYPER software (Agena, San Diego) in order to generate genotype calls.

Four SNPs (rs7637888, rs1058930, rs35599367, and rs6442150) were dropped from our analysis because less than 80% of samples produced genotyping calls. The average genotyping call rate was 98.5% across the remaining 19 SNPs. Samples were considered to have passed genotyping analysis by producing genotype calls in greater than 70% of the SNPs; 23 samples were dropped for falling below this metric.

Statistical Analysis

Since the majority (>97%) of the study participants were non-Hispanic white, analyses were limited to those of European-descent to avoid confounding by ancestry (19). Participant demographic and clinical characteristics were summarized using descriptive statistics. Carrier frequencies between individuals with and without high neuropathy symptoms (FACT/GOG-Ntx score <10, ≥10) were compared using logistic regression models adjusting for age. Odds ratios (OR) and 95% confidence intervals (CIs) adjusting for age were calculated using logistic regression models. In order to explore how identified statistically significant SNPs improved predictive value over known demographic and clinical factors alone, we conducted a receiver operating characteristic (ROC) analysis. Two ROC curves were generated: the first for a model including demographic and clinical risk factors for CIPN known prior to the start of chemotherapy [age, BMI (kg/m²), Charlson Comorbidity Index, diagnosis of diabetes mellitus (yes/no), history of prior cancer diagnosis (yes/no) (20,21)], and the second for a model including these factors along with SNPs statistically significantly associated with CIPN. The area under the curve (AUC) and associated 95% confidence intervals were used as a measure of the diagnostic performance and compared between the two models accounting for the correlation between curves(22). Due to sample size limitations, we did not adjust for multiple comparisons; p-values <0.05 were considered statistically significant. Analyses were conducted using SAS 9.4 (Cary, NC).

Results

Saliva samples from 213 participants (63.0% of those approached) were returned prior to the time of this analysis; 190 (89.2%) passed quality checks. Among those, 170 participants had associated neuropathy data available, and of those, 110 had received platinum or taxane chemotherapy regimens prior to the survey data; 107 were non-Hispanic white and therefore included in this analysis.

The median age of the cohort was 65.1 years. Most patients had a diagnosis of ovarian (58.9%) or uterine cancer (29.0%) and had advanced stage disease (59.6%). Almost 40% were overweight or had obesity, and 8.4% had diabetes. There were no statistically significant differences in age and clinical characteristics known prior to the start of chemotherapy between participants who had low versus high neuropathy scores (Table 1). Almost all participants (99.1%) had received platinum-based chemotherapy, 96 (89.7%) had received taxane-based chemotherapy, and 95 (88.8%) had received both; the median number of cycles was 6 (range 3–15 cycles) prior to the time of survey when greatest neuropathy symptoms were reported. Additionally, most (93.5%) underwent surgery as part of their cancer treatment and 30.8% received radiation.

Table 1.

Demographic and clinical characteristics by neuropathy symptoms (<10 vs. 10+).

	Low neuropathy symptom score (<10)		High neuropathy symptom score (10+)
Variable	N	Median (Min, Max)	N	Median (Min, Max)	p-value
Age at time of survey, years	52	65.1 (32.7, 79.4)	55	64.4 (36.1, 89.0)	0.97
BMI at diagnosis, kg/m2	52	27.0 (20.0, 43.7)	54	28.5 (20.4, 59.0)	0.12
Charlson Comorbidity Index	52	2.0 (0–9)	55	2.0 (0–10)	0.22
	N	%	N	%	p-value
History of Diabetes					0.16
No	50	96.2	48	87.3
Yes	2	3.9	7	12.7
History of Liver disease					1.00
No	50	96.2	53	96.4
Yes	2	3.9	2	3.6
History of other cancer					0.09
No	50	96.2	47	85.5
Yes	2	3.9	8	14.6
Diagnosis					0.98
Ovarian	30	57.7	33	60.0
Uterine	15	28.9	16	29.1
Cervical	5	9.6	4	7.3
Vaginal/vulvar	2	3.9	2	3.6
Cancer Stage					0.20
Early stage (I/II)	17	34.0	25	46.3
Advanced stage (III/IV)	33	66.0	29	53.7

Two SNPs [rs139887 in SOX10, OR: 2.66 (95% CI: 1.18, 6.00), p=0.02; and rs3753753 in GPX7, OR: 2.55 (95% CI: 1.13, 5.72), p=0.02] were identified as being statistically significantly associated with high neuropathy scores in our cohort (Table 2). Adding these two SNPs to the clinical risk factor model significantly improved the AUC compared to the model with clinical risk factors alone [0.74 (95% CI: 0.65, 0.84) versus 0.65 (95% CI: 0.54, 0.76); p=0.04; Figure 1).

Discussion

We identified two genetic variants, rs139887 in SOX10 and rs3753753 in GPX7, that were associated with neuropathy symptoms in a cohort of non-Hispanic White gynecologic cancer survivors treated with chemotherapy; and adding these variants to a model with clinical risk factors resulted in superior predictive power for neuropathy symptoms. Identification of patients at high risk for CIPN using clinical risk factors alone has been previously noted to be unreliable (2). Our findings suggest that the genetic risk factors reported here could help identify patients with gynecologic cancer who are at highest risk of CIPN, which could guide clinical decision making, justify closer monitoring during treatment, and inform preventative measures.

CIPN is a common reason for dose reduction or early discontinuation of chemotherapy, which has prompted pharmacogenetic studies to explore genetic variants and their association with CIPN in various types of cancer. Identified genetic variants to date have been related to drug disposition, metabolism and action, and are thought to affect a patient’s exposure and sensitivity to drugs. Cliff et al. performed a systematic review and meta-analysis of 93 studies exploring genetic variants and evidence for association with CIPN (13). Among the studies evaluated, 85 included platinum or taxane-associated CIPN; genes identified in these studies regulate the cell cycle, cell signaling, apoptosis, glutathione metabolism, and DNA repair pathways. Unfortunately, many pharmacogenetic studies are underpowered, which makes it difficult to interpret results and apply them to clinical practice, highlighting the need for larger and more broadly inclusive analyses (34).

The SRY-Box Transcription Factor 10 SOX10 gene encodes a nucleocytoplasmic shuttle protein that is important for neural crest and peripheral nerve development (35). McWhinney-Glass et al. found that rs139887 in SOX10, as well as variants in BCL2, OPRM1 and TRPV1, were significantly associated with CIPN in an analysis of 1261 SNPs in 404 patients with ovarian cancer treated with platinum-taxane chemotherapy from the Scottish Randomized Controlled Trial in Ovarian Cancer (SCOTROC1) (28). They observed that patients carrying three risk alleles among these genes were 4.5 times more likely to have neuropathy compared to patients with no risk alleles. ECOG performance status and treatment arm (Carboplatin/Paclitaxel vs. Carboplatin/Doxectaxel) were the clinical covariates included in their model, and as in the present study, the authors found that including risk variants in the neuropathy model improved its predictive ability. Uniquely, we included patient age and clinical risk factors available at diagnosis in our model.

The other SNP in our study found to be significantly associated with higher neuropathy symptoms was rs3753753 in Glutathione peroxidase 7 GPX7. GPX7 is a member of the GPX gene family, which is involved in reactive oxygen species production and scavenging, as well as mitochondrial function. Johnson et al. studied platinum and platinum-taxane associated CIPN in 950 patients with lung cancer (33), and similar to our study, compared epidemiological and genetic risk factors in patients with and without CIPN. Among the 174 SNPs analyzed, SNPs in GPX7 and in ATP-binding cassette sub-family C member 4 (ABCC4) gene were found to be significantly associated with CIPN (33). In addition to genetic risk factors, Johnson et al. reported that combination platinum-taxane chemotherapy, greater number of chemotherapy cycles, and diagnosis of diabetes mellitus conferred a higher risk of CIPN. Patients with lung cancer in the Johnson et al. study treated with a platinum-taxane regimen were more than twice as likely to develop CIPN compared to patients treated with only a platinum agent. These findings are particularly important for our patients as most patients with advanced stage gynecologic cancers will be treated with a combination platinum-taxane regimen.

Strengths of our study include the evaluation of CIPN specifically in gynecologic cancer survivors, including various gynecologic cancer diagnoses, and of genetic risk factors [previously validated in other studies; Table 2] in combination with known clinical factors at time of diagnosis. Symptomatology was collected prospectively without patients’ knowledge of how it would be analyzed limiting response bias. However, our study also has limitations, including being a single institution study with a small sample size, particularly for a genetics study. Our analysis was limited to non-Hispanic White patients, and results may differ in non-European ancestry patient populations. We did not have data regarding neuropathy symptoms prior to cancer diagnosis and treatment, and therefore, we could not ascertain whether participants’ neuropathy symptoms developed as a result of their cancer treatment. Recruitment of patients was at diagnosis or at any point in their cancer treatment or surveillance period; therefore, time from exposure to assessment differed between participants. Of note, there is no objective measure for CIPN. The FACT-GOG/Ntx subscale was used in this study to evaluate neuropathy symptoms, and though the instrument is both simple to use and has been validated, there is no standardized, “severe” or “clinically meaningful” cut-off score that correlates with commonly used clinical grading instruments. It is also important to note that there is a lack of consensus regarding neuropathy outcome measures, with significant discrepancies between clinician and patient-reported neuropathy symptom severity (13,36).

Table 2.

SNP associations by neuropathy symptoms (<10 vs. 10+)

Gene	Reference	rsid	Chromosome, position†	Non-Risk Allele	Risk Allele*	Low neuropathy symptom score (<10) Carrier frequency (risk allele)	High neuropathy symptom score (10+) Carrier frequency (risk allele)	Odds Ratio (95% CI)**	p-value
CYP2C8	Leskela (23) Hertz (24,25)	rs11572080	10:95067273	C	T	0.19	0.16	0.82 (0.30, 2.21)	0.69
CYP3A5	Leskela (23)	rs776746	7:99672916	C	T	0.10	0.21	2.44 (0.77, 7.70)	0.13
XKR4	Baldwin (26) Leandro-Garcia (27)	rs4737264	8:55198762	A	C	0.33	0.45	1.75 (0.79, 3.86)	0.17
SOX10	McWhinney-Glass (28)	rs139887	22:37975389	C	G	0.50	0.72	2.66 (1.18, 6.00)	0.02
BCL2	McWhinney-Glass (28)	rs2849380	18:63312127	T	C	0.90	0.98	5.60 (0.63, 49.72)	0.12
OPRM1	McWhinney-Glass (28)	rs544093	6:154136358	T	G	0.14	0.16	1.20 (0.41, 3.54)	0.74
TRPV1	McWhinney-Glass (28)	rs879207	17:3563302	A	G	0.55	0.51	0.83 (0.38, 1.80)	0.64
FANCD2	Sucheston (29)	rs7648104	3:10031627	C	A	0.23	0.31	1.52 (0.64, 3.61)	0.35
		rs6786638	3:10076391	C	G	0.98	1.00	NE
ABCB1	Abraham (30) NCCTG N08C1 (31)	rs2032582	7:87531302	C/T	A	0.67	0.74	1.39 (0.60, 3.23)	0.44
		rs1045642	7:87509329	A	G	0.75	0.63	0.56 (0.24, 1.30)	0.18
		rs3213619	7:87600877	G	A	1.00	1.00	NE
ABCG2	Lamba (32)	rs13120400	4:88112375	C	T	0.87	0.93	2.01 (0.55, 7.35)	0.29
ABCG1	Hertz (24,25)	rs492338	21:42281867	G	A	0.71	0.71	0.99 (0.43, 2.27)	0.97
ABCC4	Johnson (33)	rs1729786	13:95170985	T	C	0.94	0.89	0.50 (0.12, 2.10)	0.34
GPX7	Johnson (33)	rs3753753	1:52603842	G	C	0.50	0.72	2.55 (1.13, 5.72)	0.02
FGD4	Baldwin (26)	rs10771973	12:32640040	G	A	0.52	0.49	0.89 (0.41, 1.90)	0.76
FZD3	Baldwin (26)	rs7001034	8:28505861	A	G	0.92	0.89	0.66 (0.17, 2.49)	0.54
		rs7833751	8:28505275	T	G	0.92	0.85	0.49 (0.14, 1.74)	0.27

^*based on previous literature review

^**adjusted for age (years) at survey

^†Position is based on GRCh38 assembly hg38

In summary, we found two SNPs - rs139887 in SOX10 and rs3753753 in GPX7 - that were significantly associated with neuropathy symptomatology in a cohort of patients with gynecologic cancers. Furthermore, combining these SNPs with known clinical risk factors for CIPN significantly improved predictive power in a model designed to predict higher neuropathy symptoms. The impact of CIPN is underestimated in terms of prevalence and impact on quality of life and function in patients with gynecologic cancers. Identifying patients most at risk for CIPN could help tailor treatment decision making, inform recruitment of patients for clinical trial interventions specifically evaluating prevention or treatment of CIPN, and inform preventative care measures. While identifying patients at higher risk of neuropathy using SNPs along with clinical factors would be a relatively inexpensive and easy strategy to support treatment decisions (substitution of chemotherapy regimen based on previous GOG clinical trials; dose reduction; and/or use of prophylactic medications), it is important to note that our results are hypothesis-generating and require validation for clinical utility. Future prospective trials are needed to validate that these SNPs predict CIPN in individuals with gynecologic cancers, which could be evaluated in larger prospective trials to increase statistical power and inclusivity of individuals from diverse populations.

Figure 1.

Comparison of receiver operating characteristic (ROC) curves.

Highlights.

• Chemotherapy induced peripheral neuropathy is a common treatment-related side effect and associated with genetic variants

• 2 SNPs were associated with neuropathy symptoms in non-Hispanic white gynecologic cancer patients treated with chemotherapy

• Combining significant SNPs with clinical risk factors resulted in a superior predictive model of neuropathy symptoms