A Genome-Wide Association Study Identifies Five Novel Genetic Markers for Trastuzumab-Induced Cardiotoxicity in Japanese Population

Abstract

Trastuzumab has been administered to patients with human epidermal growth factor receptor 2 (HER2)-positive cancer, however, the cardiotoxicity is identified as one of the life-threatening toxicities. Clinically useful biomarker for trastuzumab-induced cardiotoxicity has been expected to be developed. To identify a novel genetic marker(s) determining the risk of trastuzumab-induced cardiotoxicity, we performed a first genome-wide association study (GWAS) in Japanese population. We enrolled 481 patients who had been treated with trastuzumab and carried out a GWAS using 11 cases (with cardiotoxicity) and 257 controls (without cardiotoxicity). Top 100 single nucleotide polymorphisms (SNPs) which revealed the smallest p values in GWAS (p = 7.60 × 10⁻⁷ − 2.01 × 10⁻⁴) were further examined using replication samples consisted of 14 cases and 199 controls. The combined analysis of the GWAS and replication study indicated possible association of five loci with trastuzumab-induced cardiotoxicity (rs9316695 on chromosome 13q14.3, rs28415722 on chromosome 15q26.3, rs7406710 on chromosome 17q25.3, rs11932853 on chromosome 4q25, and rs8032978 on chromosome 15q26.3, P_combined = 6.00 × 10⁻⁶, 8.88 × 10⁻⁵, 1.07 × 10⁻⁴, 1.42 × 10⁻⁴, 1.60 × 10⁻⁴, respectively). Furthermore, we developed a risk prediction model for trastuzumab-induced cardiotoxicity using the five marker SNPs. The incidence of trastuzumab-induced cardiotoxicity in patients with risk score ≥5 was significantly higher (42.5%) compared to that in patients with score ≤ 4 (1.8%) (p = 7.82 × 10⁻¹⁵, odds ratio = 40.0). These findings suggest the potential to improve the ability of physicians to avoid the trastuzumab-induced cardiotoxicity for patients with HER2-positive cancer.

INTRODUCTION

Human epidermal growth factor receptor (EGFR) type2 (HER2) is a member of the EGFR family, which is a transmembrane tyrosine kinase receptor involved in cell differentiation and proliferation through the activation of intracellular signaling pathways in response to homo- and hetero-dimerization of HER2 with EGFR family members. It is reported that HER2 is overexpressed or amplified in approximately 25–30% of breast cancer patients.^1–3) Furthermore, patients with HER2-positive cancer have a higher risk of recurrence and mortality than those with HER2-negative cancer.^1,4) Trastuzumab is a recombinant humanized monoclonal antibody and selectively targets the extracellular domain of the HER2 protein.^1,3–5) Trastuzumab induces antibody-dependent cellular cytotoxicity (ADCC) and prevents HER2 receptor from its dimerization, resulting in inhibiting downstream signaling pathways: the phosphoinositide 3-kinase (PI3K) pathway, AKT pathway and mitogen-activated protein kinase (MAPK) pathway, which are associated with cell survival and proliferation.^3,5,6) It is reported that patients treated with chemotherapy plus first-line trastuzumab showed a significantly better response than chemotherapy alone, and prolonged survival.^7–9) However, it is reported that one of the most serious side effects of trastuzumab is cardiotoxicity, which causes cardiac dysfunction, either asymptomatic or symptomatic.^6,10,11) Although trastuzumab is an effective drug for HER2-positive cancer, some patients cannot continue the trastuzumab therapy due to the cardiotoxicity. The mechanism of trastuzumab-induced cardiotoxicity is not yet fully understood, however, inhibition of HER2 signaling in cardiomyocytes has been suggested to be involved in this toxicity.^12,13) Anthracycline-induced cardiotoxicity is thought to be cumulative, dose-dependent, and irreversible, whereas trastuzumab-induced cardiotoxicity is mostly reversible and dose-independent, suggesting that genetic factors play important role in this toxicity.^6,14,15)

The association studies of candidate genes including Erb-B2 receptor tyrosine kinase 2 (ERBB2), which encodes HER2, have been reported, however, these results have not yet been sufficiently validated,^16–18) and clinically available predictive marker for trastuzumab-induced cardiotoxicity has not yet been developed. Hence, it has been suggested that the other unknown genetic factors which could contribute to the risk of trastuzumab-induced cardiotoxicity should exist. In this study, to identify common genetic variants determining the risk of trastuzumab-induced cardiotoxicity, we carried out a genome-wide association study (GWAS) and replication study of 481 patients treated with trastuzumab, and identified five loci that were likely to be associated with risk of trastuzumab-induced cardiotoxicity.

PATIENTS AND METHODS

Patients

To identify a genetic marker(s) for the risk of trastuzumab-induced cardiotoxicity, we selected 268 patients treated with trastuzumab from the samples registered in National Cancer Center (NCC) Biobank (http://www.ncc.go.jp/jp/biobank/) from February 2010 to December 2015 for GWAS stage. In the replication study, 213 patients (14 cases and 199 controls) treated with trastuzumab were collected from St. Marianna University School of Medicine Hospital and Nakagami Hospital from October 2017 to March 2018, and collected from the samples registered in NCC Biobank from January 2016 to October 2017. We defined trastuzumab-induced cardiotoxicity as left ventricular ejection fraction (LVEF) < 45% or LVEF <50% with an absolute decrease of 10% from baseline according to the criteria in HERceptin Adjuvant (HERA) trial.^10,12,15,19) Transthoracic echocardiography was performed for assessment of LVEF at baseline and during follow-up (every 1.0–37.7 months, median; 3.1 months). Echocardiography was performed according to standard procedures by one or two sonographers and analyzed by one cardiologist using the iE33 (Philips, Netherlands), SONOS5500 (Philips) or VividE95 (GE, U.S.A.) system. All participants provided written informed consent to this study. The research protocol of the study was reviewed and approved by the Research Ethics Committee of the NCC (Tokyo, Japan), St. Marianna University School of Medicine (Kawasaki, Japan), Nakagami Hospital (Okinawa, Japan) and Japanese Foundation for Cancer Research (Tokyo, Japan), and the members of the committee included experts in the genetic research and ethics.

Genotyping and Quality Control

Genomic DNA was extracted from peripheral blood or formalin-fixed paraffin-embedded (FFPE) lymph nodes which did not contain cancer cells. For the GWAS analysis, 268 patients were genotyped using the Infinium OmniExpressExome-8 v1.4 DNA Analysis Kit (Illumina, San Diego, CA, U.S.A.). We applied SNP quality control (call rate of ≥99%) in both cases and controls and a Hardy–Weinberg equilibrium (p > 1.0 × 10⁻⁶) in controls and SNPs with minor allele frequency of ≥0.01. A total of 543807 SNPs in autosomal chromosomes passed the quality control filters. We utilized the identity-by-state method to evaluate cryptic relatedness for the samples included in this study. The population stratification was examined by principal component analysis (PCA) using the EIGENSTRAT software 6.0.1 (https://www.hsph.harvard.edu/alkes-price/software/). The PCA was carried out by comparing the distribution of the sample populations with three reference populations from the 1000 Genomes Project Phase3 database that included Europeans (represented by Caucasian from UTAH, CEU), Africans (represented by Yoruba from Ibadan, YRI) and East Asians (represented by Japanese from Tokyo, JPT, Han Chinese from Beijing, CHB, Southern Han Chinese, CHS, Chinese Dai in Xishuangbanna, CDX, and Kinh in Ho Chi Minh City, Vietnam, KHV). The PCA was performed on the basis of the genotype information from the samples included in this study. The quantile–quantile (Q–Q) plot was generated between observed p values of Fisher’s exact test against expected p values and showed that genomic inflation factor was 1.172, representing no significant population stratification (Supplementary Fig. 1).

Genotype Imputation

Genome-wide imputation was performed for the 268 patients used in GWAS. Because 43 SNPs located in the 3 genomic regions (chromosome 13q14.3 and 2 independent loci on 15q26.3) showed greater significance than marker SNPs (tag-SNPs) at each locus in the genome-wide imputation analysis, these genomic regions (chromosome 13q14.3, 54593774–54618139, 15q26.3, 98578726–98646496, and 15q26.3, 101796748–101800094, respectively) were also imputed for 213 patients used in replication study. For imputation, we used the reference panel which was based on the 1000 Genomes Project Phase 3 integrated release version 5 for individuals of East Asian descent comprising Japanese from Tokyo, Chinese from Beijing and Chinese from southern China. Missing genotypes were imputed with Minimac3 software. We extracted variants with imputation quality threshold of RSQ >0.3. For imputation analysis, we performed SNP quality control by excluding SNPs that had low genotyping rate <99%, and showed deviations from Hardy–Weinberg equilibrium (p ≤ 1.0 × 10⁻⁶) in controls and SNPs with minor allele frequency of <0.01.

Statistical Analysis

In the GWAS and the replication study, a case-control study by Fisher’s exact test was applied to three genetic models: an allelic frequency model, a dominant-inheritance model, and a recessive-inheritance model. Odds ratios (ORs) and confidence intervals (CIs) were calculated for the genetic model with smallest p value in those three genetic models, using a non-risk allele or a non-risk genotype as a reference. Genome-wide significance levels after adjustment of Bonferroni correction for multiple testing of three genetic models were p = 3.06 × 10⁻⁸ [0.05/(543807 × 3)] in the GWAS and p = 1.67 × 10⁻⁴ [0.05/(100 × 3)] in the replication study. For combination analysis, the genotype count of the replication study was added to that of the GWAS. The differences in the distribution of age, primary cancer site, pretreatment with anthracycline, HER2 status and hormone receptor status were evaluated by logistic regression analysis (Supplementary Table 1). For the predictive scoring system of the risk of trastuzumab-induced cardiotoxicity, we assigned a score of 2 to individuals homozygous for risk allele, 1 to individuals heterozygous for risk allele, and 0 to individuals homozygous for non-risk allele at rs9316695, rs11932853 and rs8032978, respectively. We further assigned a score of 2 to individuals homozygous for the risk allele and 0 to individuals with the other genotypes at rs28415722 and rs7406710 and summed up the scores for each SNP to obtain individuals’ scores. Based on this system, each patient was classified into any of the nine prediction score groups (score 0, 1, 2, 3, 4, 5, 6, 7, or 8). All the statistical analyses were carried out using R statistical environment version 3.3.1 (http://www.r-project.org/), PLINK version 1.07,²⁰⁾ or the BellCurve for Excel (Social Survey Research Information Co., Ltd., Tokyo, Japan). Regional association plots were generated using Locus Zoom (http://locuszoom.sph.umich.edu/).

RESULTS

Patient Characteristics

To identify a genetic marker(s) determining the risk of trastuzumab-induced cardiotoxicity, we recruited 481 patients treated with trastuzumab, including 25 cases (with trastuzumab-induced cardiotoxicity) and 456 controls (without trastuzumab-induced cardiotoxicity). Table 1 showed the characteristics of these 481 patients receiving trastuzumab treatment. Their median age at the beginning of the trastuzumab treatment was 56 years old in both case and control groups. The distributions of gender (percentage of female patients) were 100 and 76.7% in the case and control groups, respectively, in the GWAS, and 100 and 91.5%, respectively, in the replication samples (Supplementary Table 2). The incidence of cardiotoxicity in gastric and breast cancers were 1.1% (1/90) and 6.2% (24/387), respectively (Table 1). Among the characteristics listed in Table 1, none of them showed significant association with trastuzumab-induced cardiotoxicity in logistic regression analysis (Supplementary Table 1).

Table 1. Patient Demographics and Clinical Characteristics

Characteristic	GWAS (N = 268)		Replication (N = 213)		Total (N = 481)
	No. of patients (%)		No. of patients (%)		No. of patients (%)
	Case (N = 11)	Control (N = 257)	Case (N = 14)	Control (N = 199)	Case (N = 25)	Control (N = 456)
Age (years)
Median	59.0	57.0	55.5	55.0	56.0	56.0
Range	46–76	29–86	36–81	27–83	36–81	27–86
Primary cancer site
Breast	10 (90.9)	185 (72.0)	14 (100.0)	178 (89.4)	24 (96.0)	363 (79.6)
Gastric	1 (9.1)	68 (26.5)	0 (0.0)	21 (10.6)	1 (4.0)	89 (19.5)
Unknown	0 (0.0)	4 (1.6)	0 (0.0)	0 (0.0)	0 (0.0)	4 (0.9)
Pretreatment with anthracycline
Yes	9 (81.8)	148 (57.6)	10 (71.4)	128 (64.3)	19 (76.0)	276 (60.5)
No	2 (18.2)	109 (42.4)	4 (28.6)	71 (35.7)	6 (24.0)	180 (39.5)
HER-2
Positive	10 (90.9)	249 (96.9)	14 (100.0)	199 (100.0)	24 (96.0)	448 (98.2)
Negative	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)
Unknown	1 (9.1)	8 (3.1)	0 (0.0)	0 (0.0)	1 (4.0)	8 (1.8)
ER status (Breast cancer)
Positive	6 (60.0)	119 (64.3)	11 (78.6)	117 (65.7)	17 (70.8)	236 (65.0)
Negative	3 (30.0)	64 (34.6)	3 (21.4)	61 (34.3)	6 (25.0)	125 (34.4)
Unknown	1 (10.0)	2 (1.1)	0 (0.0)	0 (0.0)	1 (4.2)	2 (0.6)
PR status (Breast cancer)
Positive	6 (60.0)	104 (56.2)	8 (57.1)	92 (51.7)	14 (58.3)	196 (54.0)
Negative	4 (40.0)	80 (43.2)	6 (42.9)	86 (48.3)	10 (41.7)	166 (45.7)
Unknown	0 (0.0)	1 (0.5)	0 (0.0)	0 (0.0)	0 (0.0)	1 (0.3)

ER, estrogen receptor; PR, progesterone receptor.

GWAS and Replication Analysis

In this study, trastuzumab-induced cardiotoxicity was defined as LVEF <45% or LVEF <50% with an absolute decrease of 10% from baseline according to the criteria in HERA trial.^10,12,15,19) We carried out a GWAS of 268 patients who received trastuzumab therapy using the Infinium OmniExpressExome-8 v1.4 DNA Analysis Kit (Illumina). After the standard quality control, an association analysis was carried out for 543807 SNPs by Fisher’s exact test on the basis of three genetic models: allelic, dominant, and recessive. We generated a Q–Q plot and observed the genomic inflation factor of 1.172, indicating no population stratification (Supplementary Fig. 1). However, we could not observe the SNP which reached the genome-wide significance level (Supplementary Fig. 2). The top 100 SNPs which revealed the smallest p value in GWAS showed possible associations (p = 7.60 × 10⁻⁷ − 2.01 × 10⁻⁴, Supplementary Table 3). To further validate the results of the GWAS analysis, we carried out a replication study of the 100 SNPs using 213 independent patients, including 14 cases and 199 controls. Although no SNP reached the significance level for three genetic models after the Bonferroni correction (p = 1.67 × 10⁻⁴ [0.05/(100 × 3)]) in the replication study, we identified 16 SNPs with suggestive associations (p = 3.57 × 10⁻² − 9.37 × 10⁻², Supplementary Table 4). A combined analysis of the GWAS and the replication study for the 16 SNPs indicated that nine SNPs in five loci showed stronger association (smaller p value) than those in GWAS results (rs9316695, rs9527156, rs12583122 and rs11617903 on chromosome 13q14.3, P_combined = 6.00 × 10⁻⁶, 8.64 × 10⁻⁶, 1.92 × 10⁻⁵ and 2.22 × 10⁻⁵, respectively, rs1383149 and rs12372962 on chromosome 15q26.3, P_combined = 1.01 × 10⁻⁴ and 1.15 × 10⁻⁴, respectively, rs7406710 on chromosome 17q25.3, P_combined = 1.07 × 10⁻⁴, rs11932853 on chromosome 4q25, P_combined = 1.42 × 10⁻⁴, and rs8032978 on chromosome 15q26.3, P_combined = 1.60 × 10⁻⁴, Supplementary Table 5, Fig. 1), however, none of them reached the genome-wide significance level in this study (p = 3.06 × 10⁻⁸).

Fig. 1. Regional Association Plots of Five Loci Associated with Trastuzumab-Induced Cardiotoxicity

Upper panel; p values of genotyped SNPs (circle) and imputed SNPs (square) are plotted (as −log₁₀ p value) against their physical location on chromosome 13q14.3(A), 15q26.3(B), 17q25.3(C), 4q25(D) and 15q26.3(E). The genetic recombination rates estimated from 1000 Genomes samples (JPT + CHB) are shown with a blue line. SNP’s color indicates LD with rs9316695 (A), rs28415722 (B), rs7406710 (C), rs11932853 (D) and rs8032978 (E) according to a scale from r² = 0 to 1 based on pair-wise r² values from ASN data in 1000 Genomes Project. Lower panel; gene annotations from the University of California Santa Cruz genome browser.

Furthermore, we carried out subgroup analysis according to cancer type and hormone receptor status (Supplementary Table 6), and found that rs8032978 on chromosome 15q26.3 showed stronger association in patients with hormone receptor positive breast cancer compared to that in original combined analysis (p = 6.55 × 10⁻⁵, OR = 8.53, 95% CI = 2.91–23.39, Supplementary Table 6). This finding suggests that rs8032978 might also be a promising marker of trastuzumab-induced cardiotoxicity in patients with hormone receptor positive breast cancer. Association analysis was also carried out by logistic regression analysis with adjusting medication for cardiovascular disease and gender which could be associated with the risk of cardiotoxicity (Supplementary Table 2), however, none of the above SNPs did not reach genome-wide significance level (Supplementary Table 7).

Imputation Analysis

To identify additional candidate loci associated with trastuzumab-induced cardiotoxicity, we examined the associations by using genome-wide imputed genotypes of GWAS samples, however, we could not observe novel candidate locus associated with the trastuzumab-induced cardiotoxicity. Of the five candidate loci identified in GWAS (chromosome 4q25, 13q14.3, 2 independent loci on 15q26.3, and 17q25.3), imputation analysis identified 43 novel SNPs in three loci (chromosome 13q14.3 and 2 independent loci on 15q26.3) of which significance were equal to or greater than the marker SNPs (tag-SNPs) identified in GWAS (Figs. 1A, B, E, Supplementary Table 8). We further performed a replication study using 213 independent patients for the 43 SNPs and found that no SNP showed stronger association than each marker SNP in the above three loci (Supplementary Table 8). However, a combined result of the GWAS and replication stage for the 43 SNPs revealed that the association of an imputed SNP, rs28415722 on chromosome 15q26.3, was the strongest in the region (P_combined = 8.88 × 10⁻⁵, Supplementary Tables 5, 8). Hence, we regarded rs28415722 as a marker SNP in the region (Table 2, Fig. 1B)

Table 2. Summary of Association Results of GWAS and Replication Study for Five Loci

Chromosomal region	Positiona)	SNP	Gene	Alleleb) [1/2]	Risk allele	Stage	Genotyped/ Imputed	Case				Control				p Value			Odds ratio (95% CI)c)
								Genotype			RAF	Genotype			RAF	Allelic	Dominant	Recessive
								11	12	22		11	12	22
13q14.3	54615773	rs9316695	No gene	C/A	A	GWAS	Genotyped	3	5	3	0.50	199	55	3	0.12	2.59 × 10−5	8.22 × 10−4	9.71 × 10−4	7.38 (2.77–19.66)
						replication		7	6	1	0.29	153	43	3	0.12	3.79 × 10−2	4.80 × 10−2	2.40 × 10−1	2.84 (1.02–7.19)
						combined		10	11	4	0.38	352	98	6	0.12	6.00 × 10−6	1.22 × 10−4	9.73 × 10−4	4.46 (2.30–8.47)
15q26.3	98609746	rs28415722d)	No gene	G/A	A	GWAS	Imputed	1	2	8	0.82	95	119	43	0.40	1.97 × 10−4	1.04 × 10−1	1.10 × 10−4	13.08 (2.99–79.58)
						replication		6	3	5	0.46	82	85	32	0.37	4.21 × 10−1	1.00	7.31 × 10−2	2.88 (0.71–10.34)
						combined		7	5	13	0.62	177	204	75	0.39	1.62 × 10−3	3.98 × 10−1	8.88 × 10−5	5.48 (2.21–13.69)
17q25.3	79505624	rs7406710	No gene	C/T	C	GWAS	Genotyped	11	0	0	1.00	111	119	27	0.66	2.38 × 10−4	6.10 × 10−1	1.35 × 10−4	NA
						replication		10	3	1	0.82	90	93	16	0.69	2.01 × 10−1	1.00	9.37 × 10−2	3.01 (0.83–13.61)
						combined		21	3	1	0.90	201	212	43	0.67	4.55 × 10−4	7.17 × 10−1	1.07 × 10−4	6.64 (2.19–27.01)
4q25	112024388	rs11932853	No gene	T/C	T	GWAS	Genotyped	7	3	1	0.77	31	129	97	0.37	2.26 × 10−4	6.04 × 10−2	1.47 × 10−4	12.54 (2.99–61.88)
						replication		4	8	2	0.57	27	99	73	0.38	7.05 × 10−2	1.45 × 10−1	1.28 × 10−1	2.13 (0.92–5.08)
						combined		11	11	3	0.66	58	228	170	0.38	1.42 × 10−4	9.56 × 10−3	2.10 × 10−4	3.20 (1.70–6.23)
15q26.3	101799790	rs8032978	No gene	A/G	G	GWAS	Genotyped	6	4	1	0.27	239	18	0	0.04	1.95 × 10−4	9.84 × 10−4	4.10 × 10−2	10.22 (2.93–31.90)
						replication		11	3	0	0.11	185	13	1	0.04	1.07 × 10−1	8.87 × 10−2	1.00	3.57 (0.57–15.86)
						combined		17	7	1	0.18	424	31	1	0.04	1.60 × 10−4	4.29 × 10−4	1.01 × 10−1	5.83 (2.30–13.51)

a) Based on GRCh37 genome assembly. b) Reference allele (GRCh37) was defined as allele 1. c) Odds ratios were shown for the model with minimum p values. d) Genotype of rs28415722 was validated using TaqMan SNP genotyping assay. GWAS, genome-wide association study; SNP, single nucleotide polymorphism; RAF, risk allele frequency; CI, confidence interval; NA, not available.

Predictive Scoring System for Trastuzumab-Induced Cardiotoxicity Using the Five Marker SNPs

Because the five SNPs, which showed smallest p value in the combined study at each candidate locus (rs9316695, rs28415722, rs7406710, rs11932853 and rs8032978, Table 2), were independent predictors of trastuzumab-induced cardiotoxicity by logistic regression analysis (p = 2.82 × 10⁻⁴ − 4.15 × 10⁻³, Supplementary Table 1), we investigated combined effects of the five loci on the risk of cardiotoxicity in patients treated with trastuzumab by using a scoring system. For the predictive scoring system, each patient was scored considering the total number of risk alleles and genotypes. We gave a score of 2 to individuals homozygous for risk allele and 0 to individuals with the other genotypes at rs28415722 and rs7406710 because these SNPs showed smallest p value in recessive-inheritance model. We further gave a score of 2 to individuals homozygous for risk allele, 1 to individuals heterozygous for risk allele and 0 to individuals homozygous for non-risk allele at rs9316695, rs11932853 and rs8032978. The patients’ scores were obtained by summing up the scores of the five SNPs. The patients used in this study were classified into nine prediction score groups (score 0–8, Table 3). The proportion of patients with trastuzumab-induced cardiotoxicity was likely to be increased in groups with higher prediction scores; the incidences of the cardiotoxicity were 1.8% (8/441) in the score 0–4 group, 36.4% (8/22) in the score 5 group, 22.2% (2/9) in the score 6 group, 75.0% (3/4) in the score 7 group and 80.0% (4/5) in the score 8 group (Table 3). We further classify the patients into two groups; score 0–4 group and 5–8 group, considering the best threshold for the diagnostic test (sensitivity, 68.0% and specificity, 95.0%). The incidences of trastuzumab-induced cardiotoxicity in score 5–8 group were 45.8 (11/24), 37.5 (6/16) and 42.5% (17/40), whereas those in score 0–4 group were 0 (0/244), 4.1 (8/197) and 1.8% (8/441) in the GWAS, replication and combined stage, respectively, suggesting that the predictive scoring system using the above five SNPs could predict the risk of cardiotoxicity prior to initiation of trastuzumab therapy (P_combined = 7.82 × 10⁻¹⁵, OR = 40.0, 95% CI = 15.6–102.3, Table 3, Fig. 2). Although we investigated the relationship between the prediction score and cardiotoxicity grade (LVEF) or time-to-event (cardiotoxicity), we could not observe significant relationship between them (Supplementary Fig. 3). Moreover, we performed risk stratification analysis based on medication for cardiovascular disease because it showed significant association with cardiotoxicity (p = 1.80 × 10⁻⁴, Supplementary Table 2). However, this predictive scoring system appeared to be medication-nonspecific (independent of medication) because the associations of cardiotoxicity with prediction scores were comparable between medicated and unmedicated patients (medicated; p = 6.82 × 10⁻⁶, unmedicated; p = 1.01 × 10⁻⁶, Supplementary Table 9).

Table 3. Prediction Scores of Trastuzumab-Induced Cardiotoxicity Using rs9316695, rs28415722, rs7406710, rs11932853 and rs8032978

CI, confidence interval.

Fig. 2. Predictive Risk Score and Incidence of Trastuzumab-Induced Cardiotoxicity

The incidence of trastuzumab-induced cardiotoxicity in patients with score ≥5 was significantly higher compared to those in patients with score ≤4 (Fisher’s exact test; p = 7.82 × 10⁻¹⁵). Stippled bar: GWAS stage, striped bar: replication stage, and black bar: combined stage.

DISCUSSION

The safe and effective treatment with trastuzumab based on individual germline and/or somatic genetic information is important to provide personalized medicine for patients with HER2-positive cancer. Although a GWAS of cardiotoxicity in Adjuvant Trastuzumab Trial (NCCT G N9831) in Caucasian population has already been reported,²¹⁾ this study is the first GWAS which attempts to identify common genetic variants associated with trastuzumab-induced cardiotoxicity in Japanese population. The genome-wide significant SNP was not identified in this study, however, we identified five independent loci showing possible association with trastuzumab-induced cardiotoxicity (Table 2, Fig. 1). Interestingly, of the above loci, proprotein convertase subtilisin/kexin type 6 (PCSK6) and small nuclear ribonucleoprotein polypeptide A (SNRPA1), which have been suggested to promote the proliferation of breast cancer and colon cancer, respectively, were included in the genomic region of chromosome 15q26.3^22,23) (Fig. 1E). We observed significantly higher incidence of trastuzumab-induced cardiotoxicity in patients with higher score (≥ 5) compared to those with lower score (≤ 4) in predictive scoring system using the five marker SNPs (Fig. 2, Tables 2, 3). This result suggests the potential to improve the ability of physicians to avoid the risk of trastuzumab-induced cardiotoxicity based on the result of this predictive scoring system for the treatment of patients with HER2-positive cancer.

rs9316695, the marker SNP on chromosome 13q14.3 (P_combined = 6.00 × 10⁻⁶, OR = 4.46, Table 2), which showed the strongest association in a combined analysis of GWAS and replication study, was located in intergenic region, and long intergenic non-protein coding RNA 558 (LINC00558) was included in this region (Fig. 1A). NOTCH3, which encodes the notch receptor 3, has been reported to be one of the target genes of LINC00558,²⁴⁾ and NOTCH3 signaling pathways are significantly activated in ErbB2-negative cells.^25,26) The Notch signaling system is present in the cardiomyocytes and suggested to play a role in myocardial remodeling in heart failure.^27,28) Hence, genetic variation in LINC00558 might contribute to interindividual variability in trastuzumab-induced cardiotoxicity through effects on Notch signaling in the stressed (inactivation of HER2 signaling) cardiomyocyte, although further validation studies will be needed to prove the true association.

rs8032978 on chromosome 15q26.3, which is one of the five marker SNPs in this study (P_combined = 1.60 × 10⁻⁴, OR = 5.83, Table 2), was located 44 kb downstream from the PCSK6 gene (Fig. 1E). PCSK6 activates corin which is a type-II transmembrane serine protease found mainly in the heart and converts it to a two-chain active enzyme through cleavage at Arg-801.^29–32) Corin is known to play an essential role in the regulation of water and salt balance by converting natriuretic peptides to their active form,^30,33) which promotes vasodilation, diuresis and natriuresis, and prevent cardiac remodeling, in patients with heart failure.^29,33) These lines of evidence suggest that impaired activity and/or gene expression of PCSK6 might contribute to the inability to recover the cardiac function in damaged (inactivation of HER2 signaling) heart due to insufficient activation of natriuretic peptides.^29–33)

To effectively clarify the mechanism of trastuzumab-induced cardiotoxicity, we performed expression quantitative trait locus (cis-eQTL) analysis for the five candidate loci. We observed cis-regulatory effect of rs7406710, which is located 1.5kb downstream from fascin actin-bundling protein 2, retinal (FSCN2) on chromosome 17q25.3 (Fig. 1C), on gene expression level of FSCN2 (p = 3.20 × 10⁻¹⁰, Supplementary Fig. 4). FSCN2 is actin-binding proteins that cross-link filamentous actin into tightly packed parallel bundles.^34,35) FSCN2 can bind with multiple actions,^34,36) and has been reported to bind with beta-, gamma-, and alpha-actin which is highly expressed in cardiomyocyte.^34,37,38) Although the relationship between HER2 signaling and actin crosslinking protein, FSCN2, in myocardium is not yet clarified, lower expression of FSCN2 in cardiomyocyte might impair cardiac contraction under the condition of HER2 inhibition (Supplementary Fig. 4). Further studies will be required to clarify the functional associations of FSCN2 with cardiotoxicity in patients treated with trastuzumab.

The candidate gene association studies previously reported that rs1058808 (P1170A) and rs1136201 (I655V) in ERBB2 (HER2) could be candidates that confer susceptibility to trastuzumab-induced cardiotoxicity.^16–18) In our GWAS, the two SNPs showed no association with trastuzumab-induced cardiotoxicity (rs1058808; p = 1.21 × 10⁻¹, rs1136201; p = 6.86 × 10⁻¹). Moreover, a first GWAS of trastuzumab-induced cardiotoxicity in Caucasian population suggested that SNPs at six loci, LIM domain binding 2 (LDB2), BMP/retinoic acid inducible neural specific 1 (BRINP1), chr6 intergenic, RAB22A, member RAS oncogene family (RAB22A), transient receptor potential cation channel subfamily C member 6 (TRPC6) and long intergenic non-protein coding RNA 1060 (LINC01060), were associated with a decline in LVEF (p = 8.93 × 10⁻⁸ − 7.73 × 10⁻⁶) in patients who received chemotherapy plus trastuzumab.²¹⁾ In our GWAS, the SNPs in these candidate genes showed no or weak association with trastuzumab-induced cardiotoxicity (p = 3.33 × 10⁻³ − 9.99 × 10⁻¹). Although the small sample size in our study might be one of the causes of the discrepant results, interethnic difference in allele frequency of the above SNPs could also affect the discrepancy. Moreover, mechanism of action involved in the trastuzumab-induced cardiotoxicity might differ between ethnic groups. The p values of the SNPs in ATP binding cassette subfamily B member 1 (ABCB1), solute carrier family 28 member 3 (SLC28A3), carbonyl reductase 3 (CBR3), neutrophil cytosolic factor (NCF4), Rac family small GTPase 2 (RAC2), retinoic acid receptor gamma (RARG) and uridine 5′-diphosphate (UDP) glucuronosyltransferase family 1 member A6 (UGT1A6), which were reported to be associated with anthracycline-induced cardiotoxicity,^39–42) were from 1.68 × 10⁻² to 9.99 × 10⁻¹, indicating no or weak association in our GWAS. The sample size used in our study might not be enough to detect true associations of the above SNPs because the effect sizes of these candidate genes were not large.^{16,17,21,39–42)}

In conclusion, our GWAS using 481 patients treated with trastuzumab identified five novel candidate loci, chromosome 4q25, 13q14.3, 17q25.3, and 2 independent loci on 15q26.3, which were associated with trastuzumab-induced cardiotoxicity. Furthermore, the combined analysis of the five SNP loci revealed that the predictive score based on risk genotypes was significantly associated with trastuzumab-induced cardiotoxicity. These findings provide new insights into personalized trastuzumab therapy for patients with HER2 positive cancer. However, larger-scale replication study and further functional analysis are required to verify our results and to clarify their biological mechanisms which have effects on trastuzumab-induced cardiotoxicity.

Acknowledgments

We express our heartfelt gratitude to all the study participants. We thank Mr. Satoshi Kasahara, Ms. Aya Imai and Ms. Emi Shigino for their contribution to the collection of clinical information, and all staffs in Nakagami Hospital, St. Marianna University School of Medicine, and National Cancer Center Biobank for their contribution to the sample collection.

Conflict of Interest

The authors declare no conflict of interest.

Supplementary Materials

The online version of this article contains supplementary materials.

REFERENCES

• 1) Bonifazi M, Franchi M, Rossi M, Moja L, Zambelli A, Zambon AT, Corrao G, La Vecchia C, Zocchetti C, Negri E. Trastuzumab-related cardiotoxicity in early breast cancer: a cohort study. Oncologist, 18, 795–801 (2013).
• 2) Hsieh AC, Moasser MM. Targeting HER proteins in cancer therapy and the role of the non-target HER3. Br. J. Cancer, 97, 453–457 (2007).
• 3) Valabrega G, Montemurro F, Aglietta M. Trastuzumab: mechanism of action, resistance and future perspectives in HER2-overexpressing breast cancer. Ann. Oncol., 18, 977–984 (2007).
• 4) Pondé NF, Lambertini M, de Azambuja E. Twenty years of anti-HER2 therapy-associated cardiotoxicity. ESMO Open, 1, e000073 (2016).
• 5) Jones KL, Buzdar AU. Evolving novel anti-HER2 strategies. Lancet Oncol., 10, 1179–1187 (2009).
• 6) Chen ZI, Ai DI. Cardiotoxicity associated with targeted cancer therapies. Mol. Clin. Oncol., 4, 675–681 (2016).
• 7) Jeyakumar A, Younis T. Trastuzumab for HER2-positive metastatic breast cancer: clinical and economic considerations. Clin. Med. Insights Oncol., 6, 179–187 (2012).
• 8) Slamon DJ, Leyland-Jones B, Shak S, Fuchs H, Paton V, Bajamonde A, Fleming T, Eiermann W, Wolter J, Pegram M, Baselga J, Norton L. Use of chemotherapy plus a monoclonal antibody against HER2 for metastatic breast cancer that overexpresses HER2. N. Engl. J. Med., 344, 783–792 (2001).
• 9) Montemurro F, Donadio M, Clavarezza M, Redana S, Jacomuzzi ME, Valabrega G, Danese S, Vietti-Ramus G, Durando A, Venturini M, Aglietta M. Outcome of patients with HER2-positive advanced breast cancer progressing during trastuzumab-based therapy. Oncologist, 11, 318–324 (2006).
• 10) Sengupta PP, Northfelt DW, Gentile F, Zamorano JL, Khandheria BK. Trastuzumab-induced cardiotoxicity: heart failure at the crossroads. Mayo Clin. Proc., 83, 197–203 (2008).
• 11) Advani PP, Ballman KV, Dockter TJ, Colon-Otero G, Perez EA. Long-term cardiac safety analysis of NCCTG N9831 (alliance) adjuvant trastuzumab trial. J. Clin. Oncol., 34, 581–587 (2016).
• 12) Dias A, Claudino W, Sinha R, Perez CA, Jain D. Human epidermal growth factor antagonists and cardiotoxicity—A short review of the problem and preventative measures. Crit. Rev. Oncol. Hematol., 104, 42–51 (2016).
• 13) Florido R, Smith KL, Cuomo KK, Russell SD. Cardiotoxicity from human epidermal growth factor receptor-2 (HER2) targeted therapies. J. Am. Heart Assoc., 6, e006915 (2017).
• 14) Wilkinson EL, Sidaway JE, Cross MJ. Cardiotoxic drugs Herceptin and doxorubicin inhibit cardiac microvascular endothelial cell barrier formation resulting in increased drug permeability. Biol. Open, 5, 1362–1370 (2016).
• 15) Bloom MW, Hamo CE, Cardinale D, Ky B, Nohria A, Baer L, Skopicki H, Lenihan DJ, Gheorghiade M, Lyon AR, Butler J. Cancer therapy-related cardiac dysfunction and heart failure: Part 1: definitions, pathophysiology, risk factors, and imaging. Circ Heart Fail, 9, e002661 (2016).
• 16) Stanton SE, Ward MM, Christos P, Sanford R, Lam C, Cobham MV, Donovan D, Scheff RJ, Cigler T, Moore A, Vahdat LT, Lane ME, Chuang E. Pro1170 Ala polymorphism in HER2-neu is associated with risk of trastuzumab cardiotoxicity. BMC Cancer, 15, 267 (2015).
• 17) Beauclair S, Formento P, Fischel JL, Lescaut W, Largillier R, Chamorey E, Hofman P, Ferrero JM, Pagès G, Milano G. Role of the HER2 [Ile655Val] genetic polymorphism in tumorogenesis and in the risk of trastuzumab-related cardiotoxicity. Ann. Oncol., 18, 1335–1341 (2007).
• 18) Lemieux J, Diorio C, Cote MA, Provencher L, Barabe F, Jacob S, St-Pierre C, Demers E, Tremblay-Lemay R, Nadeau-Larochelle C, Michaud A, Laflamme C. Alcohol and HER2 polymorphisms as risk factor for cardiotoxicity in breast cancer treated with trastuzumab. Anticancer Res., 33, 2569–2576 (2013).
• 19) Cameron D, Piccart-Gebhart MJ, Gelber RD, Procter M, Goldhirsch A, de Azambuja E, Castro G Jr, Untch M, Smith I, Gianni L, Baselga J, Al-Sakaff N, Lauer S, McFadden E, Leyland-Jones B, Bell R, Dowsett M, Jackisch C. 11 years’ follow-up of trastuzumab after adjuvant chemotherapy in HER2-positive early breast cancer: final analysis of the HERceptin Adjuvant (HERA) trial. Lancet, 389, 1195–1205 (2017).
• 20) Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MA, Bender D, Maller J, Sklar P, de Bakker PI, Daly MJ, Sham PC. PLINK: a tool set for whole-genome association and population-based linkage analyses. Am. J. Hum. Genet., 81, 559–575 (2007).
• 21) Serie DJ, Crook JE, Necela BM, Dockter TJ, Wang X, Asmann YW, Fairweather D, Bruno KA, Colon-Otero G, Perez EA, Thompson EA, Norton N. Genome-wide association study of cardiotoxicity in the NCCTG N9831 (Alliance) adjuvant trastuzumab trial. Pharmacogenet. Genomics, 27, 378–385 (2017).
• 22) Wang P, Wang F, Wang L, Pan J. Proprotein convertase subtilisin/kexin type 6 activates the extracellular signal-regulated kinase 1/2 and Wnt family member 3A pathways and promotes in vitro proliferation, migration and invasion of breast cancer MDA-MB-231 cells. Oncol. Lett., 16, 145–150 (2018).
• 23) Zeng Q, Lei F, Chang Y, Gao Z, Wang Y, Gao Q, Niu P, Li Q. An oncogenic gene, SNRPA1, regulates PIK3R1, VEGFC, MKI67, CDK1 and other genes in colorectal cancer. Biomed. Pharmacother., 117, 109076 (2019).
• 24) Chen X, Jung JG, Shajahan-Haq AN, Clarke R, Shih I-M, Wang Y, Magnani L, Wang TL, Xuan J. ChIP-BIT: bayesian inference of target genes using a novel joint probabilistic model of ChIP-seq profiles. Nucleic Acids Res., 44, e65 (2016).
• 25) Hirose H, Ishii H, Mimori K, Ohta D, Ohkuma M, Tsujii H, Saito T, Sekimoto M, Doki Y, Mori M. Notch pathway as candidate therapeutic target in Her2/Neu/ErbB2 receptor-negative breast tumors. Oncol. Rep., 23, 35–43 (2010).
• 26) Yamaguchi N, Oyama T, Ito E, Satoh H, Azuma S, Hayashi M, Shimizu K, Honma R, Yanagisawa Y, Nishikawa A, Kawamura M, Imai J, Ohwada S, Tatsuta K, Inoue J, Semba K, Watanabe S. NOTCH3 signaling pathway plays crucial roles in the proliferation of ErbB2-negative human breast cancer cells. Cancer Res., 68, 1881–1888 (2008).
• 27) Øie E, Sandberg WJ, Ahmed MS, Yndestad A, Laerum OD, Attramadal H, Aukrust P, Eiken HG. Activation of Notch signaling in cardiomyocytes during post-infarction remodeling. Scand. Cardiovasc. J., 44, 359–366 (2010).
• 28) Rizzo P, Mele D, Caliceti C, Pannella M, Fortini C, Clementz AG, Morelli MB, Aquila G, Ameri P, Ferrari R. The role of notch in the cardiovascular system: potential adverse effects of investigational notch inhibitors. Front. Oncol., 4, 384 (2015).
• 29) Khoury EE, Kinaneh S, Aronson D, Amir O, Ghanim D, Volinsky N, Azzam Z, Abassi Z. Natriuretic peptides system in the pulmonary tissue of rats with heart failure: potential involvement in lung edema and inflammation. Oncotarget, 9, 21715–21730 (2018).
• 30) Yan W, Wu F, Morser J, Wu Q. Corin, a transmembrane cardiac serine protease, acts as a pro-atrial natriuretic peptide-converting enzyme. Proc. Natl. Acad. Sci. U.S.A., 97, 8525–8529 (2000).
• 31) Chen S, Cao P, Dong N, Peng J, Zhang C, Wang H, Zhou T, Yang J, Zhang Y, Martelli EE, Naga Prasad SV, Miller RE, Malfait AM, Zhou Y, Wu Q. PCSK6-mediated corin activation is essential for normal blood pressure. Nat. Med., 21, 1048–1053 (2015).
• 32) Yan W, Sheng N, Seto M, Morser J, Wu Q. Corin, a mosaic transmembrane serine protease encoded by a novel cDNA from human heart. J. Biol. Chem., 274, 14926–14935 (1999).
• 33) Wu F, Yan W, Pan J, Morser J, Wu Q. Processing of pro-atrial natriuretic peptide by corin in cardiac myocytes. J. Biol. Chem., 277, 16900–16905 (2002).
• 34) Liu X, Zhao M, Xie Y, Li P, Wang O, Zhou B, Yang L, Nie Y, Cheng L, Song X, Jin C, Han F. Null mutation of the fascin2 gene by TALEN leading to progressive hearing loss and retinal degeneration in C57BL/6J mice. G3 Bethesda, 8, 3221–3230 (2018).
• 35) Hashimoto Y, Kim DJ, Adams JC. The roles of fascins in health and disease. J. Pathol., 224, 289–300 (2011).
• 36) Avenarius MR, Saylor KW, Lundeberg MR, Wilmarth PA, Shin JB, Spinelli KJ, Pagana JM, Andrade L, Kachar B, Choi D, David LL, Barr-Gillespie PG. Correlation of actin crosslinker and capper expression levels with stereocilia growth phases. Mol. Cell. Proteomics, 13, 606–620 (2014).
• 37) Perrin BJ, Strandjord DM, Narayanan P, Henderson DM, Johnson KR, Ervasti JM. β-Actin and fascin-2 cooperate to maintain stereocilia length. J. Neurosci., 33, 8114–8121 (2013).
• 38) Shin JB, Longo-Guess CM, Gagnon LH, Saylor KW, Dumont RA, Spinelli KJ, Pagana JM, Wilmarth PA, David LL, Gillespie PG, Johnson KR. The R109H variant of fascin-2, a developmentally regulated actin crosslinker in hair-cell stereocilia, underlies early-onset hearing loss of DBA/2J mice. J. Neurosci., 30, 9683–9694 (2010).
• 39) Hertz DL, Caram MV, Kidwell KM, Thibert JN, Gersch C, Seewald NJ, Smerage J, Rubenfire M, Henry NL, Cooney KA, Leja M, Griggs JJ, Rae JM. Evidence for association of SNPs in ABCB1 and CBR3, but not RAC2, NCF4, SLC28A3 or TOP2B, with chronic cardiotoxicity in a cohort of breast cancer patients treated with anthracyclines. Pharmacogenomics, 17, 231–240 (2016).
• 40) Aminkeng F, Bhavsar AP, Visscher H, Rassekh SR, Li Y, Lee JW, Brunham LR, Caron HN, van Dalen EC, Kremer LC, van der Pal HJ, Amstutz U, Rieder MJ, Bernstein D, Carleton BC, Hayden MR, Ross CJ. A coding variant in RARG confers susceptibility to anthracycline-induced cardiotoxicity in childhood cancer. Nat. Genet., 47, 1079–1084 (2015).
• 41) Aminkeng F, Ross CJ, Rassekh SR, Hwang S, Rieder MJ, Bhavsar AP, Smith A, Sanatani S, Gelmon KA, Bernstein D, Hayden MR, Amstutz U, Carleton BC. Recommendations for genetic testing to reduce the incidence of anthracycline-induced cardiotoxicity. Br. J. Clin. Pharmacol., 82, 683–695 (2016).
• 42) Visscher H, Ross CJ, Rassekh SR, Barhdadi A, Dube MP, Al-Saloos H, Sandor GS, Caron HN, van Dalen EC, Kremer LC, van der Pal HJ, Brown AM, Rogers PC, Phillips MS, Rieder MJ, Carleton BC, Hayden MR. Pharmacogenomic prediction of anthracycline-induced cardiotoxicity in children. J. Clin. Oncol., 30, 1422–1428 (2012).