Background

As major advances are being made to understand the biology of the disease and optimize therapeutic benefit of various treatment strategies, side effects associated with those treatments must also be understood in order to minimize the dose/treatment limiting effects of such complications. One of the most impactful therapies for breast cancer is the humanized monoclonal antibody, trastuzumab (Herceptin®), which specifically recognizes the HER2 protein encoded by the ERBB2 gene. HER2 is part of a family of transmembrane receptors, is overexpressed in about 15–20% of invasive breast cancers, and is associated with aggressive biology and a natural history of shortened survival.

The original pivotal trial of trastuzumab in patients with HER2-positive metastatic breast cancer created concern for the cardiac safety of patients receiving HER2 inhibitors after 27% of patients treated with concurrent anthracycline (when given at a cumulative dose of >300 mg/m2)/cyclophosphamide/trastuzumab and 13% of those treated with trastuzumab/paclitaxel experienced some degree of cardiotoxicity [1]. Data from multiple randomized adjuvant trials and observational studies suggest that the rate of discontinuation of trastuzumab treatment ranges between 6 and 31%, mainly due to cardiotoxicity [2–10]; and in patients aged ≥65 years old, this figure increases to a notable and clinically relevant rate of 41% [11].

Trastuzumab is a monoclonal antibody that targets the protein HER2, but the mechanisms that account for trastuzumab-mediated cardiotoxicity are largely unknown. The expert consensus statement from the American Society of Echocardiography and the European Association of Cardiovascular imaging, classifies Trastuzumab as a Type II Cancer therapeutics-related cardiac dysfunction (CTRCD) agent because it does not directly cause cell damage in a cumulative dose-dependent fashion (unlike Type 1 CTCRD/anthracycline cardiotoxicity), and observed cardiac dysfunction is often reversible when treatment is discontinued [12, 13]. However classification of cardiotoxicity as Type I or II is confounded by the fact that such treatments are often given sequentially or concurrently and that echocardiography may miss subtle cardiac defects which do not change left ventricular ejection fraction (LVEF) measurement.

To some extent, these factors have been addressed in animal models, which allow single treatment with trastuzumab and more invasive testing methods. ErbB2-deficient conditional mutant adult mice were viable and displayed no overt phenotype, but physiological analysis revealed a phenotype consistent with dilated cardiomyopathy [14, 15]. Isolated cardiomyocytes from conditional mutants were more susceptible to anthracycline toxicity, demonstrating that ErbB2 signaling in cardiomyocytes is requisite for the prevention of dilated cardiomyopathy (DCM) [14].

More recent pharmacological studies in both mouse and zebrafish suggest that the link between ErbB2 inhibition and DCM is not directly linked to cardiomyocyte survival; rather that pharmacological inhibition of ErbB2/erbb2 (without combination treatment with anthracycline), can result in myofibril remodeling. Treatment of wild type mice with trastuzumab, resulted in impaired ventricular function, ultrastructural damage of heart tissue (stretched appearance and reduced thickness of ventricular cardiac myofibers), and differential expression of 15 genes involved in adaptability to cardiac contractility, hemodynamic stress, DNA repair mechanisms, apoptosis, and mitochondrial function [16]. Taken together, mouse and zebrafish studies suggest that ErbB2/erbb2 inhibition plays a role in cardiomyocyte survival following anthracycline treatment, but also directly affects the structural organization of specific subpopulations of ventricular myofibrils, even in the absence of anthracycline.

The limitations of these studies is their reliance on genetic models of erbb2 inhibition, or in the case of pharmacological experimentation, trastuzumab is a humanized monoclonal antibody and may not have the same efficacy in the mouse. Trastuzumab is not an option in zebrafish models, and the pharmacological inhibition in the literature [17–19] used PD168393, a compound that inhibits both EGFR and ERBB2 but is not used in the clinic.

Human induced pluripotent stem cell (hiPSC) derived cardiomyocytes are now a convenient option for the assessment cardiotoxicity of clinically relevant drugs in humans. On the basis that trastuzumab has been shown to alter the expression of genes essential for cardiac contractility in a mouse model [16] and that different gene expression profiles in multiple models of heart failure are well known [20–25], we hypothesized that detailed examination of trastuzumab induced differential gene expression in human iPSC-derived cardiomyocytes could provide insight into the genes, pathways and potential biomarkers of trastuzumab induced cardiotoxicity.

Methods

Compounds

Trastuzumab, was received from Genentech (South San Francisco, CA) as a gift. Lapatinib Ditosylate was purchased from LC Laboratories (Woburn, MA). We used 100 µg/mL trastuzumab to represent a clinically relevant dose. In studies of patients in which an initial dose of 4 mg/kg was followed by a weekly maintenance dose of 2 mg/kg, in combination with chemotherapy, the mean ± S.D. peak serum trastuzumab concentration at week 8 was 101.0 ± 30.6 µg/mL (n = 115). Among patients receiving trastuzumab alone, mean ± S.D. peak concentration at week 8 was 95.6 ± 35.9 µg/mL [26]. We used 2 µM lapatinib Ditosylate based on previous observations [27].

Breast cancer cell culture

SKBR3, BT474, HCC1954, MDA-MB-453, HCC1569, and HCC1419 breast cancer cell lines were purchased from ATCC (Manassas, VA). SKBR3 and BT474 cell lines were maintained at in IMEM supplemented with 10% fetal bovine serum (FBS) and cultured at 37 °C, 5% CO2. HCC1954, HCC1569, and HCC1419 cells were cultured in RPMI with 10% FBS at 37 °C, 5% CO2. MDA-MB-453 cells were cultured in L-15 with 10% FBS at 37 °C, no CO2. All cells passaged with 0.25% trypsin as needed.

Cardiomyocyte cell culture

iCell® cardiomyocytes were purchased from CELLular Dynamics International, CDI (Madison, WI) maintained according to the manufacturer’s recommended instructions. Cardiomyocytes were seeded with CDI’s plating media into fibronectin (5 µg/mL) coated wells at the appropriate density (2.4e5 cells per well of 12 well plate; 2e4 per well of 96 well plate. After 2 days, media was changed to CDI’s maintenance media and replaced every other day. Cells were maintained at 37 °C, 7% CO2 until 10 days after plating at which time cells were treated with the appropriate drug for 48 h and endpoint assays (RNA extraction, viability/apoptosis) performed.

Western blot analysis

15 µg total cell lysates were prepared by briefly sonicating cell pellets in 1× lysis buffer (Cell Signaling) supplemented with a protease inhibitor mixture (Roche) and phosphatase inhibitors (PhosSTOP, Roche). Protein samples were resolved on Bolt 8% Bis–Tris Plus gels (ThermoFisher) and were electrophoretically transferred to a polyvinylidene difluoride membrane and blocked for 1 h at room temperature with 5% dry milk in TBST (50 mM Tris, 150 mM NaCl, 0.1% Tween, pH 7.5). Blots were stained with antibodies against EGFR (D38B1), HER2/ERRB2 (D8F12), HER3/ERRB3(D22C5), and HER4/ERBB4 (111B2),and GAPDH (D16H11) purchased from Cell Signaling Technology and used a dilution of 1:1000 in 5% dry milk in TBST for 1 h at 22 °C. For detection of PHLDA1 expression, blots were stained with 1;1000 dilution of anti-PHLDA1 antibody (EPR6674, abcam®). For detection of PDK4 expression, blots were stained with 2 µg/mL dilution of anti-PDK4 (NBP1-07049, Novus Biologicals). All blots were washed with three changes of TBST for a total of 45 min. Blots were then incubated in TBST buffer with 5% dry milk containing GAR-HRP (1:10,000, Cell Signaling Technology) for 1 h at room temperature. After three washes with TBST, antigen–antibody complexes were detected with the ECL Plus chemiluminescent system (Amersham Biosciences) and visualized with film.

Transcriptome sequencing

RNA was extracted from breast cancer cell lines and iPSC-derived cardiomyocytes with the RNeasy minikit (Qiagen). RNASeq libraries were prepared with the TruSeq v2 (Illumina Inc, San Diego, Ca), and sequenced by 101 base paired-end sequencing on an Illumina HiSeq 2000 (Illumina Inc, San Diego, Ca), mRNA as per the manufacturer’s instructions. Quality control, sequence alignment and gene expression counts were obtained using MAP-R seq workflow [28]. Transcriptome data is deposited in GEO, accession number GSE91383.

Quantitative PCR

Two-step quantitative reverse transcriptase-mediated real-time PCR (qPCR) was used to measure abundance of individual mRNAs. Equal aliquots of total RNA from samples were converted to cDNA with the High-Capacity cDNA Archive kit (Applied Biosystems), and qPCR reactions were performed in triplicate with 10 ng of cDNA and the TaqMan Universal PCR master mix (Applied Biosystems). Primer/probe sets were purchased from Applied Biosystems. Amplification data were collected with an Applied Biosystems ViiA7 detector and analyzed with ViiA7 v 1.2.4 software (Life Technologies). Data were normalized to the endogenous control POLR2A [29] and mRNA abundance was calculated using the ΔΔCT method [30].

Metabolic assays

Glucose was measured in cell culture supernatant by Amplex® Red Glucose/Glucose Oxidase Assay (ThermoFisher). Fluorescence was measured at 530/590 nm, and glucose concentration calculated via standard curve. Lactate was measured in cell culture supernatant using the Lactate Colorimetric/Fluorometric Kit (Biovision). Fluorescence was measured at 535/587 nm.

Apoptosis assay

Apoptosis was measured by caspase-3/7 activation with the ApoTox-Glo assay (Promega) as per the manufacturer’s instructions.

Statistical analyses

RNASeq statistical analyses were conducted with R version 3.1.1. Transcriptome-wide differential expression analyses were undertaken via the edgeR package, version 3.8.6. Cell lines and cardiomyocytes treated with trastuzumab or lapatinib were compared to untreated samples (four replicates each group). Genes with less than two samples expressed at two counts-per-million or greater were considered beneath the threshold of detection and filtered from all analyses. To correct for multiple testing, Benjamini and Hochberg’s FDR method was applied and a threshold of q < 0.05 was considered significant.

Gene enrichment analyses were performed with the Gene Ontology (GO) Consortium database [31, 32] in each cell type following ERBB2 inhibition. For each transcriptome, we took those genes differentially expressed, p < 0.05 and searched for enrichment of gene ontology categories by biological process.

Significance of qPCR fold change, glucose uptake, lactate production and apoptosis following treatment with trastuzumab or lapatinib was analyzed by T test in Graphpad Prism.

Results

ERBB expression in iPSC-derived cardiomyocytes and cancer cell lines overexpressing ERBB2

To put into context, ERBB2 expression levels in cancer cell lines overexpressing ERBB2 and iPSC-derived cardiomyocytes, we examined RNA and protein expression levels of the ERBB family. Gene expression of all members of the ERBB family of tyrosine kinase receptors was detected in cardiomyocytes and cancer cell lines overexpressing ERBB2 (Fig. 1). As expected, ERBB2 expression was significantly higher in the ERBB2-overexpressing cancer cell lines (mean 12.9 log2 counts per million), but ERBB2 was also expressed in iPSC-derived cardiomyocytes, 8.96 log2 counts per million. ERBB2 gene expression is shown separately for each of the six cancer cell lines in Fig. 1.

Fig. 1
figure 1

RNA expression of ERBB family in iPSC-derived cardiomyocytes and cancer cell lines overexpressing ERBB2. At the level of RNA, ERBB2-overexpressing cancer cell lines express higher levels of ERBB2 by ~1–5 orders of magnitude relative to iPSC-derived cardiomyocytes. iPSC-derived cardiomyocytes expressed moderate levels of ERBB4, and low levels of EGFR and ERBB3

Full size image

Protein levels of each ERBB family member in iPSC-derived cardiomyocytes are shown in Fig. 2. Using 15 µg of cell lysate, only ERBB2 and ERBB4 were detected at the protein level in iPSC-derived cardiomyocytes. The lack of EGFR expression in the iPSC-derived cardiomyocytes suggests that the action of lapatinib which targets EGFR and ERBB2 is not through inhibition of EGFR, but either ERBB2 or ERBB2 in combination with any off-target effects. ERBB2 protein was down-regulated by trastuzumab in iPSC-derived cardiomyocytes, but not by lapatinib. Trastuzumab did not down-regulate ERBB2 protein expression in ERBB2 overexpressing cancer cell lines (Fig. 3).

Fig. 2
figure 2

Protein expression of ERBB family in iPSC-derived cardiomyocytes. Western blot of 15 µg of total cell lysate of iPSC-derived cardiomyocytes stained for members of the ERBB family. GAPDH was used as loading control. ERBB2 and ERBB4 are expressed in iPSC-derived cardiomyocytes at the level of protein. We did not observe protein expression of EGFR and ERBB4

Full size image
Fig. 3
figure 3

Trastuzumab down-regulates ERBB2 in iPSC-derived cardiomyocytes, but not in ERBB2 overexpressing cancer cell lines. a Western blot of 15 µg of total cell lysate stained for ERBB2 in MDA-MB-231 (non-ERBB2 overexpressing breast cancer cell line negative control), and iPSC-cardiomyocytes treated with trastuzumab and lapatinib. ERBB2 expression was down-regulated by trastuzumab but not by lapatinib. b ERBB2 expression was not altered by trastuzumab in the ERBB2 overexpressing cancer cell lines, BT474 and SKBR3

Full size image

Trastuzumab induced significant differential gene expression in iPSC-derived cardiomyocytes

517 genes were differentially expressed in iPSC-derived cardiomyocytes treated with trastuzumab at p < 0.05, of which 137 showed fold change >1.5 (49 down-regulated, 88 up-regulated). Only one gene, IL31RA remained significant following correction for multiple testing, fold change 3.85, p = 1.56 × 10−6, q = 0.025. Full differential expression results are shown in Additional file 1: Table S1.

Overlap of trastuzumab induced differentially expressed genes is relatively small between ERBB2-overexpressing cancer cell lines and iPSC-derived cardiomyocytes

We observed a high degree of overlap in the genes expressed in iPSC-derived cardiomyocytes and ERBB2-overexpressing cancer cell lines. Of the 14,946 genes expressed above background in the cancer cell lines, 11,466 (77%) were also expressed above background in the iPSC cardiomyocytes (Additional file 2: Figure S1A). As further validation of changes in gene expression mediated by ERBB2 inhibition, we next assessed the overlap of trastuzumab induced differential gene expression between the ERBB2-overexpressing cancer cell lines and iPSC-derived cardiomyocytes. 76/11,466 (0.66%) genes expressed in both cell types were differentially expressed, p < 0.05, (Additional file 2: Figure S1B). 76 genes were differentially expressed, p < 0.05 in both cancer cell lines and iPSC-derived cardiomyocytes following treatment with trastuzumab (Table 1). 63/76 genes showed fold changes in the same direction in both cancer cell lines and iPSC-derived cardiomyocytes.

Table 1 76 genes were differentially expressed in both ERBB2-overexpressing cancer cell lines and iPSC-derived cardiomyocytes, following treatment with trastuzumab, p < 0.05
Full size table

Validation of trastuzumab induced differentially expressed genes in iPSC-derived cardiomyocytes by additional ERBB2 targeted compounds

We also performed transcriptome sequencing on iPSC-derived cardiomyocytes following 48 h treatment with the small molecule tyrosine kinase inhibitor, lapatinib, and examined differential gene expression of treated versus untreated cells. We observed a total of 1358 genes that were differentially expressed following treatment with lapatinib at p < 0.05, of which 533 showed fold change >1.5 (271 up-regulated and 262 down-regulated). 75 genes remained significant q < 0.05 after correction for multiple testing, the most significant of which was PHLDA1, p = 8.89 × 10−9, q = 0.0001, fold change 0.29. Differential gene expression data for all genes following treatment with lapatinib is shown in Additional file 3: Table S2.

38/137 (28%) of genes that were differentially expressed in iPSC cardiomyocytes following treatment with trastuzumab, were also differentially expressed following treatment with lapatinib, p < 0.05 and fold change >1.5, (Table 2), including the most significantly differentially expressed genes from both drugs, IL31RA from trastuzumab and PHLDA1 from lapatinib. 93% of genes that were differentially expressed following treatment with lapatinib were not significantly changed by treatment with trastuzumab, demonstrating the wider range of effect of the tyrosine kinase inhibitor relative to monoclonal antibody.

Table 2 Differentially expressed genes in cardiomyocytes following ERBB2 inhibition
Full size table

To validate the differentially expressed genes identified by transcriptome sequencing, we selected genes with fold changes >1.5, p < 0.05, following treatment with both trastuzumab and lapatinib: EGR1, PHLDA1, RGS4, SLC6A6 for qPCR in RNA extracted from two separate batches of iPSC-derived cardiomyocytes (Fig. 4). To further evaluate ERBB2 inhibition of these genes, we also treated cells from a new batch of iPSC-derived cardiomyocytes with additional ERBB2 inhibitors, neratinib, afatinib and poziotinib. PHLDA1 and SLC6A6 were significantly down-regulated by all inhibitors (Additional file 4: Figure S2). We also confirmed down-regulation of PHLDA1 at the level of protein expression by Western blot, (Fig. 5).

Fig. 4
figure 4

qPCR validation of differential expression of RGS4, EGR1, SLC6A6 and PHLDA1 by trastuzumab and lapatinib in iPSC-derived cardiomyocytes. Standard deviation (SD) calculated from three replicates. Down-regulation of: a RGS4, b SLC6A6, c EGR1 and d PHLDA1 by trastuzumab and lapatinib was confirmed by qPCR. Student’s T-test, unequal values, p < 0.05*; p < 0.01**

Full size image
Fig. 5
figure 5

Trastuzumab and lapatinib induced down-regulation of PHLDA1 protein expression and up-regulation of PDK4 protein expression. Western blot of 15 μg of total cell lysate of iPSC-derived cardiomyocytes stained for PHLDA1 and PDK4

Full size image

ERBB2 inhibition down-regulates genes involved in metabolism in iPSC-derived cardiomyocytes and ERBB2-overexpressing cancer cell lines

We used gene ontology (GO) [31, 32] to perform gene enrichment analyses in each cell type following ERBB2 inhibition. For each transcriptome, we took those genes differentially expressed at p < 0.05 and searched for enrichment of gene ontology categories by biological process. Following Bonferroni correction of each GO term, a total of 272 GO terms were associated at p < 0.05 with differentially expressed genes in ERBB2-overexpressing cancer cell lines treated with trastuzumab. Not surprisingly, the majority of these terms related to cell cycle processes, with mitotic cell cycle, GO:0000278 being the most significant, p = 2.62 × 10−42 (Additional file 5: Table S3). We observed a similar trend in cardiomyocytes treated with lapatinib. 88 GO terms were associated at p < 0.05 with differentially expressed genes, the most significant terms being, mitotic nuclear division, GO:0007067, p = 1.09 × 10−15 and mitotic cell cycle process, GO:1903047, p = 1.68 × 10−14 (Additional file 6: Table S4).

However, in cardiomyocytes treated with trastuzumab, following Bonferroni correction, only 20 GO terms were associated at p < 0.05 with the differentially expressed genes, all of which related to metabolism. The most significant GO term was small molecule metabolic process, p = 3.22 × 10−9 (which included the taurine transporter, SLC6A6, already shown to be significantly down-regulated by multiple ERBB2 inhibitors in qPCR validation) and multiple terms relating to cholesterol processing were associated [cholesterol biosynthetic process (GO:0006695), cholesterol metabolic process (GO:0008203) and sterol biosynthetic process (GO:0016126) (Additional file 7: Table S5)]. Post-hoc examination of the differentially expressed genes within these categories identified the same set of recurring genes (LDLR, SREBF1, ABCA1, CLN6, MSMO1, INSIG1, NSDHL, DHCR24, DHCR7, IDI2, FDPS, IDI1) detailed in Additional file 8: Table S6, 11/12 of which were down-regulated by trastuzumab.

The same GO terms relating to small molecule metabolic process and cholesterol metabolic process in cardiomyocytes treated with trastuzumab, were also significant in cardiomyocytes treated with lapatinib (GO:0044281, p = 4.94 × 10−4; GO:0008203, p = 3.90 × 10−3 respectively) and ERBB2-overexpressing cells treated with trastuzumab (GO:0044281, p = 1.23 × 10−10; GO:0008203, p = 1.35 × 10−3) (Table 3).

Table 3 Gene ontology enrichment analysis
Full size table

Trastuzumab induced ERBB2 inhibition down-regulates glucose uptake in iPSC-derived cardiomyocytes

Following the link to trastuzumab induced differential expression of genes involved in metabolism, which included up-regulation (confirmed at the level of protein, Fig. 5), of PDK4, an inhibitor of the pyruvate dehydrogenase complex and hence glycolysis, and down-regulation of multiple genes located in the mitochondria, perhaps reflecting a reduction in oxidative phosphorylation, (ALDH1L2, SLC27A3, SLC25A25, SLC25A5, SLC25A6 and UCP2, fold change and p values in Additional file 1: Table S1), we next examined the effect of trastuzumab on glycolysis. Following treatment with trastuzumab, we measured glucose uptake (based on measuring the amount of glucose in the media) up to 96 h. Cardiomyocytes treated with trastuzumab showed significantly higher amounts of glucose in the media (hence lower glucose uptake) as early as 24 h post-treatment (p = 0.001), and this difference remained significant up to 96 h post-treatment (p = 0.03), (Fig. 6).

Fig. 6
figure 6

Down-regulation of glucose levels in the media of iPSC-derived cardiomyocytes following treatment with trastuzumab. Standard deviation (SD) calculated from three replicates. Time course of glucose concentration in the media of iPSC-derived cardiomyocytes treated with trastuzumab versus untreated (control). *p = 0.04 (72 h), p = 0.026 (96 h); **p = 0.001 (24 h), ***p < 0.0002 (48 h)

Full size image

Trastuzumab induced ERBB2 inhibition did not alter lactate production in iPSC-derived cardiomyocytes

Pyruvate can be converted to either lactate in the glycolytic pathway, or to acetyl co-enzyme A via the pyruvate dehydrogenase complex, linking glycolysis to the tricarboxylic (TCA) cycle and fatty acid synthesis. Our differential expression data showed up-regulation of PDK4, an inhibitor of the pyruvate dehydrogenase complex, suggesting that trastuzumab could increase lactate production. We measured lactate production in cardiomyocyte media following 48 h treatment with trastuzumab. However, we did not observe any change in lactate production, despite significantly lower glucose uptake (Fig. 7).

Fig. 7
figure 7

Glucose and lactate levels in the media of iPSC-derived cardiomyocytes following 48 h treatment with trastuzumab. Standard deviation (SD) calculated from three replicates. Glucose levels were significantly higher in the media of iPSC-derived cardiomyocytes treated with trastuzumab relative to untreated cells, *p = 0.004. Levels of lactate were not significantly different between iPSC-derived cardiomyocytes treated with trastuzumab and untreated cell, p = 0.329

Full size image

Discussion

Current standard of care requires patients to receive trastuzumab for 12 months during which time they are monitored by echocardiography. Those patients experiencing significant declines in LVEF are treated with standard heart failure medications such as beta blockers or ACE inhibitors, and if LVEF does not recover, trastuzumab may be withheld. We hypothesized that identifying early changes in gene expression mediated by trastuzumab could potentially identify novel therapeutic targets for prevention of cardiac side-effects or potential biomarkers of drug-induced cardiotoxicity. We used commercially available, human iPSC-derived cardiomyocytes from a “healthy” female individual, treated with clinically relevant ERBB2 targeted therapies, trastuzumab and lapatinib to identify early therapy induced changes. We identified two groups of genes that may be relevant in the management of cardiotoxicity.

Firstly, the small group of significantly differentially expressed genes in iPSC-derived cardiomyocytes treated with trastuzumab and lapatinib, related to cardiac dysfunction and ischemic injury. EGR1 and PHLDA1, both down-regulated in our experiments, are significantly upregulated in ischemic preconditioning [17], suggesting perhaps that prevention of this change in gene expression could be protective against ERBB2 therapy related cardiotoxicity. Absence of regulator of G-protein signaling 4 (RGS4) predisposes to atrial fibrillation and is associated with abnormal calcium handling [33] and activation of RGS4 mediates the cardioprotective effects of natriuretic peptides in the heart [34], of which natriuretic peptide A, NPPA, was also significantly down-regulated by trastuzumab and lapatinib. Solute carrier family 6, member 6 (SLC6A6), is a transporter of taurine, an endogenous sulfur-containing beta-amino acid, associated with calcium handling, protection against ischemia–reperfusion injury, heart failure ischemic heart disease, and diabetic cardiomyopathy [35]. Slc6a6 knock out mouse models demonstrate a phenotype of dilated cardiomyopathy with cardiac atrophy [36], lower fasting blood glucose, acceleration of glycolysis in skeletal muscle and susceptibility to development of obesity when placed on a high fat diet [37]. Taken together, our findings of trastuzumab induced down-regulation of SLC6A6 in human iPSC-derived cardiomyocytes, and the reported phenotypes of dilated cardiomyopathy and altered metabolism in Slc6a6 knock-out mice provide a potential mechanism towards trastuzumab related cardiotoxicity.

The second group of trastuzumab regulated genes in our iPSC-derived cardiomyocyte model involved metabolism. Enrichment analyses identified gene ontology categories: small molecule metabolism, cholesterol and sterol metabolism and cholesterol and sterol biosynthesis, and expression of the majority of genes in these categories was down-regulated across iPSC-derived cardiomyocytes treated with trastuzumab and lapatinib and also in cancer cell lines overexpressing ERBB2. In the case of iPSC-derived cardiomyocytes treated with trastuzumab, specifically, only metabolic gene ontology categories were significantly enriched. This is in contrast to ERBB2-overexpressing cancer cell lines treated with trastuzumab, in which the most significantly enriched categories related to cell cycle. We note at this point, a similar pattern in trastuzumab-induced down-regulation of ERBB2 protein levels between cancer cells and cardiomyocytes. In our data, and that of others, ERBB2 was down-regulated by trastuzumab in cardiomyocytes [38] but not in ERBB2 overexpressing cancer cell lines [39]. This is not surprising given the different mechanisms of inhibition of ERBB2 signaling and trafficking between cancer cells and cardiomyocytes [40, 41], in which treatment results in reduced proliferation and induction of apoptosis in cancer cell lines [42, 43] and reduced contractility without apoptosis in cardiomyocytes, (Additional file 9: Figure S3) [12, 44].

Regards enrichment of differential expression of metabolism related genes, the largest metabolic category, was small molecule metabolism (N = 98 genes), which included the taurine transporter, SLC6A6, key enzymes in glucose metabolism (PDK4 and GLA) and mitochondrial transporters (SLC27A3, SLC25A25, SLC25A5, SLC25A6). Due to the high energy demand of the heart, cardiac metabolism is highly flexible. In the ‘normal/healthy’ heart, mitochondria are primarily fueled by the metabolites of fatty acids and carbohydrates, but under chronic pathological conditions, the heart has the capacity to remodel metabolic pathways, and subsequently, myocardial energetics and contractile function [45], although these data relate mostly to reprogramming in response to pathological hypertrophy, characterized by increased reliance on glucose metabolism and decreased fatty acid oxidation. The effect of trastuzumab, on glucose uptake in iPSC-derived cardiomyocytes suggests a different metabolic change to that observed in cardiac hypertrophy. We observed significantly more glucose in the media of iPSC-derived cardiomyocytes treated with trastuzumab within 24 h and constantly up to 96 h, indicative of a lower uptake of glucose as a result of trastuzumab, with negligible change in lactate. These observations are in line with studies of “lipid status” responsiveness, whereby up-regulation of PDK4, facilitates fatty acid oxidation by “sparing” pyruvate for oxaloacetate formation [46, 47]. In heart and skeletal muscle, increased anaplerotic entry of pyruvate into the TCA cycle as oxaloacetate facilitates entry of acetyl co-A derived from fatty acid β-oxidation into the TCA cycle through increased citrate formation, which in turn, acts as a sensor of fatty acid abundance and suppresses glucose uptake [46].

Observations of trastuzumab induced changes in metabolism are clinically relevant in light of potential metabolic therapies for heart failure. For example, dichloracetate (DCA), a drug shown to improve cardiac function in right ventricular hypertrophy and failure [48, 49], increases pyruvate dehydrogenase activity by inhibiting PDK, and perhaps could potentially reverse the effects of trastuzumab. The efficacy of DCA treatment has also been demonstrated in functional recovery during reperfusion in multiple animal models [50–53]. Lewandowski and White [52] demonstrated that counteracting depressed pyruvate oxidation enhanced contractile recovery in post-ischemic rabbit hearts. Of note, two of the most consistently down-regulated genes following ERBB2 inhibition, were EGR1 and PHLDA1, which have also been shown to be up-regulated in models of ischemic preconditioning [17].

In summary our transcriptome based approach of trastuzumab induced changes in iPSC-derived cardiomyocytes identified significant down-regulation of expression of genes associated with ischemic preconditioning and cardiac dysfunction, and more significantly with genes involved in metabolism. The main caveats of the study are the reliance on an in vitro model of iPSC-derived cardiomyocytes which may not reflect fully mature differentiated human cardiomyocytes and does not address potential immune response, and lack of functional data. However, our reported changes of trastuzumab mediated differential gene expression in cardiomyocytes were validated with additional ERBB2 targeted compounds and we also observed considerable overlap in ERBB2-overexpressing cancer cell lines which we interpret as strong evidence that ERBB2 inhibition is responsible for our reported changes in gene expression. We acknowledge, it is not proven that these specific genes functionally contribute to dysregulation of glycolysis in iPSC-derived cardiomyocytes, but we note that knock out of slc6a6, one of the most consistently, differentially expressed genes in our human study, resulted in a phenotype of accelerated glycolysis in skeletal muscle in a mouse model [36].