Potential Gene Association Studies of Chemotherapy-Induced Cardiotoxicity: A Systematic Review and Meta-Analysis

Chemotherapy is widely used in the treatment of cancer patients, but the cardiotoxicity induced by chemotherapy is still a major concern to most clinicians. Currently, genetic methods have been used to detect patients with high risk of chemotherapy-induced cardiotoxicity (CIC), and our study evaluated the correlation between genomic variants and CIC. The systematic literature search was performed in the PubMed, Cochrane Central Register of Controlled Trials (CENTRAL), China Biology Medicine disc (CBMdisc), the Embase database, China National Knowledge Internet (CNKI) and Wanfang database from inception until June 2020. Forty-one studies were identified that examined the relationship between genetic variations and CIC. And these studies examined 88 different genes and 154 single nucleotide polymorphisms (SNPs). Our study indicated 6 variants obviously associated with the increased risk for CIC, including CYBA rs4673 (pooled odds ratio, 1.93; 95% CI, 1.13–3.30), RAC2 rs13058338 (2.05; 1.11–3.78), CYP3A5 rs776746 (2.15; 1.00–4.62) ABCC1 rs45511401 (1.46; 1.05–2.01), ABCC2 rs8187710 (2.19; 1.38–3.48), and HER2-Ile655Val rs1136201 (2.48; 1.53–4.02). Although further studies are required to validate the diagnostic and prognostic roles of these 6 variants in predicting CIC, our study emphasizes the promising benefits of pharmacogenomic screening before chemotherapy to minimize the CIC.


INTRODUCTION
The burgeoning field of cardio-oncology is continuing to grow in step with major scientific developments in oncology that have improved cancer prognosis and survivorship (1). Cardiotoxicity has long been considered one of the main side effects of chemotherapy in cancer patients (2)(3)(4)(5)(6). Now, more effective therapies and some forms of radiotherapy may also have multiple cardiovascular (CV) secondary effects, in particularly left ventricular dysfunction (LVD) and heart failure (HF) (7)(8)(9)(10)(11)(12). Some of the data from studies on the genetic defects and pharmacological interventions suggested that many molecules, primarily those regulating oxidative stress (OS), autophagy, apoptosis, and metabolism, contributing to the pathogenesis of cardiotoxicity induced by cancer treatment. Treatment with anthracyclines has been reported to increase the risk of cardiotoxicity and death by more than five times (13)(14)(15). Early diagnosis and treatment of cardiotoxicity can increase the chances of recovery (16,17), which highlights the urgent need to develop new technologies and programmes for the early management of cardiotoxicity caused by chemotherapy and for a multidisciplinary clinical approach throughout the chemotherapy process (18)(19)(20).
Early identification of the CIC is essential to minimize the harmful side effects of cancer treatment, and could provide oncologists and cardiologists with an ideal choice to allow personalized antitumor treatment strategies or interventions (21). Although some factors may make some patients more susceptible to the severity of toxicity, individual differences in toxicity manifestations still remain large, which will significantly exacerbate these toxic. Genetics, therefore, could provide insights into the development of toxicity induced by some cancer treatments. Identification of the genetic biomarkers that are able to predict whether the patient is at risk of developing cancer therapies-induced cardiac dysfunction will allow for the minimization of cardiotoxicities during cancer treatment by careful monitoring, applying cardioprotective drugs or using optimized cancer therapies. Several recent studies have showed the role of genetic variation as a biomarker for the early detection of CIC (22)(23)(24)(25). The aim of this study was to provide an overview of studies focusing on the relationship between polymorphic gene variants and CIC.

Search Strategy
Our study searched the PubMed, CENTRAL, CBMdisc, the Embase database, Wanfang database and CNKI from the beginning to June 2020. The search terms include chemotherapy, anthracyclines, doxorubicin, daunorubicin, epirubicin, idarubicin, trastuzumab, cyclophosphamide, 5-fluorouracil, methotrexate, adriamycin, cisplatin, cytoxan, cardiotoxicity, HF, cardiomyopathy, arrhythmia, genetic, pharmacogenomics, variant and polymorphism. The search is limited to clinical trials involving human participants. Then, literature titles, abstracts and subject words are carefully analyzed to further identify keywords for document retrieval. If the abstract is relevant to our research, we will read the full text. References in the research were also analyzed to find out some studies that might have been missed in the original search.

Study Selection
Studies that met the criteria were as follows: (i) most of the SNPs were considered as dominant inheritance models unless specifically notified; (ii) original studies that determine the relationship between the genetic polymorphism (including different SNPs in each gene deletions, duplication, and copynumber variants) and cardiotoxicity; (iii) chemotherapy was used regardless of the cycle regimen type, timing, and duration of administration; (iv) in human studies and (vi) in English language. Exclusion criteria: (i) laboratory studies, case series and reports, interim studies; (ii) republication literature; (iii) data with obvious error. The relevance of the article titles and abstract was filtered by two independent reviewers, and the full text was retrieved based on inclusion criteria. Any disagreement will be settled by the third author by decision.

Data Extraction
The authors (GPL and MKG) extracted the data and checked the qualifications and the methodological quality of each included study. Any disagreements will be discussed and if the discussion is not finalized, the disagreements would be resolved by the third author (XYY). This information extracted from each document included the name of the first author, the year of publication, the sample size of the trial, the type of participant, the age of the participant, the type of cancer, the genotyping technique and the definition of cardiotoxicity. At last, we assessed the relationship between different genomic polymorphisms and cardiotoxicity.

Quality Assessment
The STREGA reporting guide was used to assess the quality of each study report (26,27). STREGA includes five main categories of information: reporting possible genotyping methods and errors, addressing population stratification methods, methods used to inference haplotypes or genotypes, whether the Hardy-Weinberg equilibrium was considered and whether this study is the first to report genetic associations, replication work or both. The quality of the report is assessed independently by two investigators, and differences are resolved through the discussion or through the third author if no consensus is reached between the two investigators.

Statistical Analysis
The meta-analysis was analyzed using the Review Manager 5.3 packages (http://comunity.cochrane.org/tools/reviewproduction-tools/revman-5) (28) and STATA version 13.0. In studies evaluating the same genotype polymorphism, we performed the meta-analysis using the fixed effects (FE) model and the random effects model (29), and this study heterogeneity was evaluated using the I 2 statistics. Overall heterogeneity was quantified with the I 2 , with p < 0.01 used to indicate significance (30). Compared with the sampling error in the study, the true variance ratio of the estimated effects between the included studies was calculated by using the I 2 statistic and moderate heterogeneity is considered when I 2 is between 50 and 75% and high heterogeneity is considered when I 2 >75% (30). The sensitivity analysis was conducted to evaluate the stability of this study, namely meta-analysis is performed again after the exclusion of abnormal results, and the results of meta-analysis were compared with those of studies that did not exclude abnormal results.

The Quality of the Reporting in the Studies
Among the studies reviewed ( Table 2), there was only one study met the five main standard for reporting data from the genetic association studies in the STREGA guidelines. Most researches (N = 32) did not report the error rates or call rates related to the genotyping methods. Twenty-six researches did not indicate whether the genotyping was done in the batches or simultaneously. Twenty-nine studies did not provide any information on whether the population stratification was evaluated in our analysis.

Chemotherapy-Induced Cardiotoxicity (CIC) and Susceptibility Genes
Our study reported a total of 154 SNPs involving eighty-eight genes (Supplementary Table 1). Most of the research has focused on genetic variations linked to chemical synthesis or heart function. Accessable data was found in 17 studies out of the included 41 studies and 14 SNPs were consistently detected. Then these 14 SNPs were subjected to our quantitative analysis (Figures 2-4). And we searched for these six variants in the protein-coding region of the gene in the Pubmed ( Table 3). The genes included in the meta-analysis are discussed below.

Cytochrome b-245, Alpha Polypeptide (CYBA) Gene
The CYBA gene encodes a major component of the phagocytes microbicidal oxidase system. Meanwhile, our study confirmed six studies that assessed the correlation between CYBA rs4673 and cardiotoxicity, and four studies (22,24,31,47) were included in quantitative analysis. In our study, this SNP was found to increase the risk of developing cardiotoxicity (OR: 1.93; 95% CI: 1.13-3.30; p = 0.04) with high heterogeneity (I 2 = 65%) (Figure 2.1). We used the random effects model and excluded individual studies to conduct sensitivity analysis to account for

Genotyping methods and errors
Describe the laboratory methods: state the source and storage of DNA, the genotyping methods and the platforms Describe the laboratory methods: state the error rates and call rates State the laboratory/center where the genotyping was done Specify whether genotypes were assigned using all of the data from the study simultaneously or in smaller batches Report the numbers of individuals for whom genotyping was attempted and the numbers of individuals for whom genotyping was successful

Modeling population stratification
Describe any methods used to assess or address population stratification

Modeling haplotype variation
Describe any methods used for inferring genotypes or haplotypes

Hardy-Weinberg equilibrium
State whether the Hardy-Weinberg equilibrium was considered

Replication
State if the study is the first report of a genetic association, a replication effort or both Studies a  15  16  17  18  19  20  21  22  23  24  25  26  27 28

Genotyping methods and errors
Describe the laboratory methods: state the source and storage of DNA, the genotyping methods and the platforms Describe the laboratory methods: state the error rates and call rates State the laboratory/center where the genotyping was done Specify whether genotypes were assigned using all of the data from the study simultaneously or in smaller batches Report the numbers of individuals for whom genotyping was attempted and the numbers of individuals for whom genotyping was successful

Modeling population stratification
Describe any methods used to assess or address population stratification

Modeling haplotype variation
Describe any methods used for inferring genotypes or haplotypes

Hardy-Weinberg equilibrium
State whether the Hardy-Weinberg equilibrium was considered

Replication
State if the study is the first report of a genetic association, a replication effort or both  Report the numbers of individuals for whom genotyping was attempted and the numbers of individuals for whom genotyping was successful

Modeling population stratification
Describe any methods used to assess or address population stratification

Modeling haplotype variation
Describe any methods used for inferring genotypes or haplotypes

Hardy-Weinberg equilibrium
State whether the Hardy-Weinberg equilibrium was considered

Replication
State if the study is the first report of a genetic association, a replication effort or both the heterogeneity, and these results show that excluding the study of Armenian et al. can alter heterogeneity (I 2 = 0%).

Ras-Related C3 Botulinum Toxin Substrate 2 (RAC2) Gene
The RAC2 gene encodes proteins that regulate a variety of processes including secretion, phagocytosis, cell polarization, and ROS production. Four of the six studies indicated that the SNP rs13058338 on the RAC2 obviously increased the risk of cardiotoxicity (22,24,31,47). We used the random effects model and the analysis of this variant in four studies suggested that the RAC mutation significantly increased the risk of cardiotoxicity by nearly twice (OR: 2.05; 95% CI: 1.11-3.78; p = 0.02), but with moderate heterogeneity (I 2 = 68%) (Figure 2.2). Sensitivity analyses were conducted to explore the potential sources of heterogeneity, but these results did not change substantially [I 2 = 74% (47), 74% (22), 60% (24), 66% (31)].

Cytochrome P450 Family 3 Subfamily A Member 5 (CYP3A5) Gene
The CYP3A5 gene is involved in the metabolism and clearance of daunorubicin (DNR), but the activity of CYP3A5 is impacted on the gene polymorphism, which has individual differences (68). Meta-analysis of two studies (22,64) reported that the rs776746 on CYP3A5 increased cardiotoxicity risk, including a total of 232 patients (54 in the case group and 178 in the control group). Using the FE model, our study suggested that the missense mutation was related to an obviously increase in the risk of cardiotoxicity [OR = 2.15 [95% CI (1.00-4.62)], P = 0.05] (Figure 2.3).

ATP Binding Cassette (ABC) Genes
The ABC transporter gene encodes a superfamily of transmembrane proteins that can use adenosine triphosphate to actively transport substrates including doxorubicin through the membrane (69). Meta-analysis of three studies (24,31,47) reported that the ABCC1 rs45511401 increased cardiotoxicity risk, including a total of 1,023 patients (363 in the case group and 660 in the control group). Using the random effects model, we found that missense mutations were related to a significant increase in risk [OR = 1.46 [95% CI (1.05-2.01)], P = 0.02] with moderate heterogeneity (I 2 = 70%) (Figure 2.4). The study conducted by Semsei et al. (38) were excluded, changing the results and addressing heterogeneity (I 2 = 0%). And three studies (24,31,38) indicated that the ABCC2 rs8187710 significantly augmented the risk of cardiotoxicity (OR: 2.19; 95% CI: 1.38-3.48; p = 0.0009) (Figure 2.5).

Human Epidermal Growth Factor Receptor 2 (HER2) Gene
The HER2 is a proto-oncogene that encodes transmembrane proteins that have tyrosine kinase activity, but have not been identified as the physiological ligand. Four included studies (33,45,55,58) revealed Ile655Val rs1136201 on HER2 significantly increase risk for cardiotoxicity, including a total of 480 patients (118 in the case group and 362 in the control group). Using the FE model, the result indicated that the missense mutation was related to a significant increase in the risk of cardiotoxicity [OR = 2.48 [95% CI (1.53-4.02)], P = 0.0002] with low heterogeneity (I 2 = 12%) (Figure 2.6).

Solute Carrier Family 28 Member 3 (SLC28A3) Gene
It has been previously reported that SLC28A3 (rs7853758) has cardioprotective effects in the multiple patient cohorts with an odds ratio of 0.35-0.42 (39,51). We identified five studies that assessed the association between rs7853758 missense SNP on SLC28A3 and cardiotoxicity, and three studies (22,24,65) were included in the quantitative analysis. We used the random effects model. But the SNP rs7853758 was not statistically significant difference (OR: 1.77; 95% CI: 0.44-7.13; p = 0.42) (Figure 3.2) with high heterogeneity (I 2 = 70%). Therefore, we excluded individual studies to conduct sensitivity analyses to account for the heterogeneity, and the result suggested no significant differences between the selected studies.

Superoxide Dismutase II (SOD2) Gene
SOD2 exists in the mitochondria and metabolizes superoxide radicals formed when anthracycline compounds are oxidized to hydrogen peroxide. Meta-analysis of two included studies (36,47) revealed rs4880 on SOD2 significantly increase risk

NAD(P)H Quinone Dehydrogenase1 (NQO1) Gene
The NQO1 gene is involved in the protection of intracellular OS, and many pro-oxidant drugs induced basic NQO1 activity (34). Two included studies (34,47) revealed SNP rs1800566 on NQO1 significantly increase risk for cardiotoxicity, including a  (Figure 3.4).

Angiotensinogen (AGT) Gene
Association between AGT P.m.ET235THr gene polymorphism and cardiovascular disease concluded that there was a positive correlation between essential hypertension (72,73) and myocardial infarction (74). Two included studies (44,47) revealed SNP rs699 AGT significantly increase risk for cardiotoxicity, including a total of 303 participants (90 in the case group and 213 in the control group). The FE model indicated that the missense mutation was not statistically significant [OR = 1.93 [95% CI (0.60-1.79)], P = 0.91] (Figure 4.1).

Angiotensin II Type-1 Receptor (AGTR1) Gene
Angiotensin II is the main ligand of AGTR1A, which adjusts the intravascular volume and the blood pressure (75). And functional polymorphisms in the ACE, AGT, and AGTR1 can impact the expression or the function of encoded proteins, and are related to stroke, coronary heart disease, vascular dysfunction and diabetes (76). Meta-analysis of two included studies (47,50) revealed SNP rs5186 significantly increase risk for cardiotoxicity, including a total of 303 patients (90 in the case group and 213 in the control group). But the FE model indicated that the missense mutation was not statistically significant difference [OR = 0.66 [95% CI (0.40-1.10)], P = 0.11] (Figure 4.2).

Carbonyl Reductases (CBR) Genes
The enzyme encoded by the CBR genes catalyzes the reduction of the endogenous aliphatic aldehydes and ketones as well as various xenobiotic, and thus has a cardioprotective effect against cardiotoxicity. Four SNPs on CBRs were studied, one on carbonyl reductase 1 (CBR1) gene and three on carbonyl reductase 3 (CBR3) gene. However, two SNPs, rs9024 of CBR1 (22,37,47,59) (Figures 4.3,4.4).

Sensitivity Analyses
We excluded each study in order, and the rest reported inconsistent results between ABCC1 and CYBA. In this metaanalysis, the studies that included Semsei et al. (38) and Armenia et al. (47) obviously distorted these results, suggesting that the two reports may be statistically unstable (Figures 5A,B). Other results showed consistent results (Figures 5C,D).

DISCUSSION
Our study analyzed the association between genetic polymorphism and CIC. We included a total of 41 studies that screened 88 different genes. Gene variation may promote OS, metabolic disorders, mitochondrial dysfunction, calcium overload, myocardial fibrosis, sarcoplasmic reticulum structure and function destruction, cardiomyocyte autophagy and apoptosis in CIC (77,78). The results of our meta-analysis showed that the polymorphisms in six (6.8%) of eighty-eight genes were obviously associated with the risk of cardiotoxicity in patients receiving chemotherapy. For genes with positive associations, mechanism studies have indicated that these alleles can alter the encoded proteins expression or activity, leading to cardiotoxicity. The ABCC1 transporter plays an important function in the OS, and is involved in maintaining the adequate levels of glutathione, which is essential for ROS defense. In addition, ABCC1 also requires glutathione to transport anthracycline antibiotics (79)(80)(81), which may also affect the OS response induced by anthracycline antibiotics (82). The ABCC2 gene encoded proteins involved in the efflux of substances from cells, and ABCC2 mutation obviously decrease the ATPase-activity, leading to reduce in the efflux activity resulting in the accumulation of anthracyclines in cells (83). CYBA encodes p22phox, which is one of the two subunits of the NOX located in the cell membrane. And the reduced activity of inherited NOX may lead to impaired ROS defense capacity, thus increasing ROS levels under chemotherapy exposure. Similarly, RAC2 encoded by the RAC2 gene is a mitochondrial protein required for the electron transfer reaction of NOX (84) during OS formation (85). Genetic alteration leads to mitochondrial dysfunction, which leads to an increase in ROS production and ultimately cardiomyocyte damage. There is also evidence that the CYP3A5-mediated oxidative metabolism of anthracyclines may induce drug-induced cardiotoxicity by generating OS (86). Taken together, these gene mutations are believed to cause cardiotoxicity on account of the accumulation of chemotherapy and excessive ROS in cardiomyocytes.
Some other genes were also observed to be associated with CIC, such as HER2-Ile655Val rs1136201. The most studied germline polymorphism in the clinical level is associated with the transmembrane domain of HER2 protein 655 A>G Ile/Val (87,88), which may be related to a high risk of BC (89). The Val allele presence may make cardiomyocytes especially dependent on HER2-Ile655Val rs1136201 signaling and the highly sensitive to trastuzumab (33). This mechanism of cardiotoxicity induced by trastuzumab is unknown, however, HER2/neu has been shown to be critical for cardiomyocytes in animal model. Our meta-analysis also confirmed the role of HER2-Ile655Val polymorphism as a genetic predictor of cardiac toxicity induced by trastuzumab in the BC patients.
There are some limitations to this study that deserve discussion. Firstly, the sample size of the study was small. This is particularly important for genome-wide association study (GWAS), with some literature advocating 10,000 cases to gain sufficient statistical power to detect causality through meta-analysis and data aggregation. Sixty one percentage of the included studies had fewer than 200 people. Secondly, there is selection bias in this study. Most literature used a retrospective method, usually by convenience sampling to recruit patients who were still in the hospital system, thereby further limiting the choice to participants who are still alive. In addition, the analysis was based on previous reports, which may not be complete or accurate. Finally, due to the lack of raw data, studies on drug dosage protocols and study periods varied widely. In addition, patient cohorts were often heterogeneous in terms of disease, drug dose, drug route, and administration, all of which may confuse toxicity associated with the target drug.

CONCLUSIONS
This study suggests that the polymorphisms in multiple pharmacogenetic in the biochemical and cardiotoxicity pathways may be predictors of CIC. However, for limited quantitative analysis, the evidences are limited and too diverse. Further researches are needed to produce reliable genetic predictors of CIC in order to achieve the goal of individualized chemotherapy.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author/s.

AUTHOR CONTRIBUTIONS
YX, HS, and SD designed the idea for drafting this review. XY, GL, MG, and AB collected the documents and wrote the paper. TY, QD, CZ, WL, and NA contributed to discussion. FY, HP, PW, CL, YGo, and YGa revised and edited the review. All authors commented on the manuscript.