Selective protection of human cardiomyocytes from anthracycline cardiotoxicity by small molecule inhibitors of MAP4K4

Given the poor track record to date of animal models for creating cardioprotective drugs, human pluripotent stem cell-derived cardiomyocytes (hPSC-CMs) have been proposed as a therapeutically relevant human platform to guide target validation and cardiac drug development. Mitogen-Activated Protein Kinase Kinase Kinase Kinase-4 (MAP4K4) is an “upstream” member of the MAPK superfamily that is implicated in human cardiac muscle cell death from oxidative stress, based on gene silencing and pharmacological inhibition in hPSC-CMs. A further role for MAP4K4 was proposed in heart muscle cell death triggered by cardiotoxic anti-cancer drugs, given its reported activation in failing human hearts with doxorubicin (DOX) cardiomyopathy, and its activation acutely by DOX in cultured cardiomyocytes. Here, we report successful protection from DOX in two independent hPSC-CM lines, using two potent, highly selective MAP4K4 inhibitors. The MAP4K4 inhibitors enhanced viability and reduced apoptosis at otherwise lethal concentrations of DOX, and preserved cardiomyocyte function, as measured by spontaneous calcium transients, at sub-maximal ones. Notably, in contrast, no intereference was seen in tumor cell killing, caspase activation, or mitochondrial membrane dissipation by DOX, in human cancer cell lines. Thus, MAP4K4 is a plausible, tractable, selective therapeutic target in DOX-induced human heart muscle cell death.

H9c2 cell protection by each of these MAP4K4 inhibitors was associated both with a block to activation of executioner caspases (Fig. 1D-F) and with preservation of mitochondrial membrane potential (Fig. 1G-I). A small increase in TMRM fluorescence over time was observed in control cardiomyocytes, ascribed to equilibration of the dye. Cyclosporine A (CsA) was used for comparison, given its known inhibition of the mitochondrial permeability transition pore, by targeting cyclophilin D 37 . Thus, small molecule inhibitors of MAP4K4 Figure 1. Inhibitors of MAP4K4 confer protection from DOX in rat H9c2 cardiomyocytes. H9c2 cells were treated with the MAP4K4 inhibitors shown, or DMSO as the vehicle control, beginning 1 h prior to DOX. (A-C) Viability, measured by the CellTiter-Glo assay. (A,B) Dose-response for DOX cardiotoxicity at 24 h (A) and 48 h (B), in the absence or presence of 10 μM F1386-0,303, the second-generation MAP4K4 inhibitor 11 . Data are triplicates, plotted as the mean ± SEM, and are representative of 3 independent dose-response experiments. (C) Protection from 333 nM DOX at 48 h by F1386-0303 versus DMX-5804, a third-generation inhibition 11 . Data are from a single experiment, plotted as the mean ± SEM, and are representative of ≥ 20 independent dose-response experiments across multiple batches of each compound. (D-F) Activation of executioner caspases. (D,E) Dose-response for caspase-3/7 activity triggered by DOX, at 6 h (D) and 24 h (E), in the absence or presence of 10 μM F1386-0303. Data are triplicates, shown as the mean ± SEM. (F) Dose-response for protection from 1 μM DOX at 24  www.nature.com/scientificreports/ reduce DOX cardiotoxicity in this convenient non-human surrogate, acting at least in part through by preventing the dissipation of mitochondrial potential and cardiomyocyte apoptosis. These H9c2 data provided a direct justification for, and specifically guided, the experiments we next performed in hPSC-CMs.
A reported feature of human cardiomyocyte protection from acute oxidative stress by inhibitors of MAP4K4 was the preservation of spontaneous calcium oscillations 11 , a hallmark of cardiomyocyte function. Analogously, sub-maximal concentrations of DOX markedly impaired spontaneous calcium cycling in hPSC-CMs (beats min −1 , 13.1 ± 5.5 versus 44 ± 3.8; P ≤ 0.0001), as seen in prior studies 12 , and this was fully rescued by co-administration of either MAP4K4 inhibitor, F1386-0303 or DMX-5804 (P ≤ 0.0001 for each; Fig. 2J,K). By contrast to the loss of beat frequency and total peak area, only small changes occurred in median peak height and width (not shown). *P < 0.05; **P < 0.01; ****P ≤ 0.0001 versus DOX alone. non-selective in effect, might compromise the desired impact of DOX on killing cancer cells, a series of five human tumour lines was subjected to graded concentrations of DOX in the absence or presence of DMX-5804 ( Fig. 3A-D). The tested cell lines were HUT-78 (T cell non-Hodgkin's lymphoma), THP1 (acute monocytic leukemia), and U266, KMS-12-BM, and MM.1S (multiple myeloma). We chose to focus on these hemapoietic cancers in order to test multiple examples from a selected class, systematically, rather than canvas a wider array of cell types superficially. Second, we were guided by these specific lines' known susceptibility to DOX [46][47][48][49][50] . Third, we chose to compare the three myeloma lines to test for potential differences in protection by our compound dependent on the status of p53 (U266 and KMS-12-BM, mutated; MM1.S, wild-type) 51 . No interference was seen in the DOX response, in any of these five human cancer lines. Analysis of hypodiploid DNA and annexin staining by flow cytometry further substantiated the lack of protection of cancer cells by DMX-5804 from DOXinduced cell death (Fig. 3E,F). Additional studies were performed, to monitor upstream events in the THP1 line (Fig. 3G,H). Executioner caspase activity was induced tenfold by 1 μM DOX, and was unaffected by either F1386-0303 or DMX-5804 (Fig. 3G). Likewise, the loss of mitochondrial membrane potential provoked by 1 μM DOX was unimpeded (Fig. 3H); rather, additional deterioration was seen, at the highest concentrations of both compounds. Thus, the inhibitors of MAP4K4 promote cardiomyocyte resistance to DOX without confounding effects on tumour cell death.

Discussion
The unmet need for cardioprotection against the toxic effects of cancer chemotherapy poses a daunting challenge for the cardiologist and oncologist alike. As is true for cardiac muscle cell protection more generally, such as with acute ischemic injury, cardiac drug discovery has been slowed or stymied by the lack of human preclinical models for target validation and compound development 17,18 . The advent of hPSC-CMs provides an auspicious alternative to previous technologies, which has proven its predictive power at least in safety pharmacology 13,16 and has justifiably raised expectations about its ability to distinguish or prioritise among potential remedies, in the years to come. Here, based upon the activation of endogenous MAP4K4 in DOX-induced cardiomyopathy and in cardiomyocytes treated acutely with DOX, we prove that DOX toxicity in two independent lines of hPSC-CMs is suppressed by two small-molecule inhibitors of MAP4K4. On the basis of these highly encouraging findings, we postulate MAP4K4 to be a well-posed target toward suppressing human cardiac cell death and dysfunction in drug-induced cardiomyopathies due to DOX and, perhaps, other chemotherapeutic agents. No agent for cardioprotection in cancer chemotherapy has ever entered human trials based on human preclinical proof of effect. Of course, future whole-animal studies in small and large mammals will be needed, to complement human cell-based evidence. These findings are consistent with the proven benefits of blocking MAP4K4 to promote cell survival not only in hPSC-CMs subjected to alternative death signals (H 2 O 2 , menadione, hypoxia-reoxygenation) 11 , but also in adult mouse myocardium after experimental myocardial infarction 11 , human islet cells in palmitate-induced apoptosis 52 , and hPSC-derived motor neurons in amyotrophic lateral sclerosis 53 . Whereas this highly generalizable benefit speaks to the likely utility of MAP4K4 inhibitors beyond merely DOX-induced cardiotoxicity, such results also point to a reciprocal concern, that MAP4K4 inhibitors, if capable of acting as a broad survival signal, might interfere with tumour cell killing, notwithstanding current interest in the cancer field regarding the participation of MAP4K4 in tumor angiogenesis, tumor cell motility, and metastasis [54][55][56][57][58][59] .
A priori, the mechanistic basis for selective protection of cardiomyocytes but not tumor cells, strictly dichotomous across the seven human cell lines tested thus far, might relate to (i) DOX killing cardiomyocytes versus cancer cells in distinguishable ways, MAP4K4 inhibitors engaging preferentially the former, or (ii) DOX killing cardiomyocytes and cancer cells in identical ways, but with differing susceptibility to MAP4K4 inhibitors-in short, differences in the dying versus differences in the rescue, contingent on cell type. Provisionally, we favor the former of these overarching scenarios. At clinically relevant concentrations, the lethal effect of DOX in cancer cells is ascribed to interference with TOP2A, which is required during DNA replication, leading to DNA doublestrand breaks 60 . By contrast, cardiomyocytes lack TOP2A, TOP2B largely mediates cardiotoxicity 61 , and greater emphasis is given instead to DOX-induced mitochondrial dysfunction and reactive oxygen species 7 . But, some overlap exists in these reported effector pathways, and the separation based on cell type is at present incomplete. Independently of TOP2A, DNA intercalation by DOX, DNA adduct formation, and activation of DNA damage responses by this alternative route can drive tumor cell death 62 , and DOX activates the DNA damage pathway in cardiomyocytes as well 63 . What can be said, thus far, is that protection from DOX occurred in both quiescent and proliferative cardiomyocytes (hPSC-CMs, versus H9c2 myoblasts), and therefore is unlikely due merely to the absence or presence of cell cycling, and that susceptibility to DOX remained in cancer lines, both in the absence and presence of wild-type p53, a gene known to mediate the cardiac toxicity 64 .
Resolving the molecular mechanism for cardioprotection by inhibitors of MAP4K4 will require identification of the responsible downstream MAP3Ks, terminal MAPKs, alternative downstream kinases, and non-canonical substrates, among other complementary strategies 53,55,56,[65][66][67][68][69] . It is unknown whether MAP4K4 operates nonredundantly in driving human cardiac cell death from lethal chemotherapeutic drugs (our default hypothesis), equivalent to the unique role demonstrated by gene silencing in some other contexts: cardiomyocytes subjected to H 2 O 2 as a model oxidative stress 11 and motor neuron degeneration in amyotrophic lateral sclerosis 53 . The counter-hypothesis, redundancy with its closest relatives, MINK1 (MAP4K6) and TNIK (MAP4K7), which are identical in the ATP-binding pocket, is asserted for neuronal cell death after NGF withdrawal, based on combinatorial loss-of-function mutations 68 . A genetic dissection of DOX toxicity in hPSC-CMs could be instructive in distinguishing between these.  www.nature.com/scientificreports/ MAP4K4 has been implicated as a strong biomarker of cancer severity or prognosis, including evidence from histochemical levels 68,70 , gene expression 57,71 , microRNA networks 72 , and next-generation proteomics 73 . While the results shown here bear solely on cell survival, the demonstrated lack of a confounding effect on tumour cell killing by DOX gives obvious credence to further exploration of MAP4K4 inhibitors in these further aspects of cancer therapeutics, beyond just cardioprotection.

Methods cell lines.
Two complimentary hPSC-CM lines were used, both with a proven role for MAP4K4 in cardiomyocyte death in other disease models 11 . Human vCor.4U PSC-derived ventricular myocytes were obtained from Ncardia. IMR90-4 hPSC-CMs were produced in-house using chemically defined medium, the Wnt inhibitors CHIR99021 and Wnt-C59, and metabolic selection in glucose-free medium 74  compounds. The MAP4K4 inhibitors F1386-0303 and DMX-5804 were synthesized as previously reported 11 . cell culture. Human vCor.4U cardiomyocytes were cultured on white (viability, CellTox Green) or black (FLIPR) clear-bottom 384-well plates (781098/781091, Greiner Bio-One), treated for 1 h with 50 μM fibronectin (Sigma-Aldrich), as described 11 . Thawed cells first were transferred from cryovials at 10,000 well −1 into 384-well plates containing pre-warmed Cor.4U maintenance medium for 4 days, with medium changes every 48 h. For viability and FLIPR experiments, cells were subjected to DOX (or comparator stress signals) ± test compounds in 40 µl well −1 of maintenance medium on day 4, and were assayed on day 5 or at other time-points as noted.
Human IMR90-derived cardiomyocytes were differentiated from the iPS(IMR90)-4 cell line (WiCell) 42 by a modification of published methods 74,75 , as follows. Undifferentiated hiPSCs at 65-85% confluency were passaged using 0.5 mM EDTA and replated in E8 medium (Gibco) with 10 µM ROCK inhibitor (Y-27632; Selleckchem). At day 0 of differentiation, the medium was changed to RPMI supplemented with B27 without insulin (Thermo Fisher Scientific) and 6 µM CHIR99021 (LC Laboratories). On day 2, medium was changed to RPMI supplemented with B27 without insulin and the next day supplemented with 2.5 µM Wnt-C59 (Selleck Chemicals). On days 5, 7 and 9, the medium was replaced with fresh RPMI supplemented with B27 without insulin. For metabolic selection, the medium was replaced on days 11 and 13 with RPMI without D-glucose supplemented with B27, and then replaced on day 15 with RPMI supplemented with B27. The resulting cardiomyocytes were maintained for three weeks in serum-free RPMI-1640 (R8758, Sigma-Aldrich), supplemented with B27 and Antibiotic-Antimycotic (Thermo Fisher). The final concentration of amphotericin B was 270 nM, well below the threshold concentrations reported for toxicity of this compound 76,77 . The IMR90 cardiomyocytes were reseeded at a density of 12,500 cells well −1 onto half-area 96-well plates (675096, Grenier Bio-One), coated as above, for 1 week prior to treatment. The maintenance medium was replaced every 2 days, and treatments were performed in the maintenance medium.
H9c2 cells were provided at passage 2 from ATCC and were used only up through passage 16 or one month of culture. Cells were cultured to a maximum of 75% confluency before passaging in DMEM and 10% FBS. Cells were removed from the culture flask (typically, 75 cm 2 ) using 0.25% (w/v) Trypsin-0.53 mM EDTA and passaged 1:2 every 3 days. Cells (1,000 well −1 ) were plated in white (viability, CellTox Green, caspase) or black (TMRM) clear-bottom 384-well plates (781098/781091, Greiner Bio-One) in DMEM and 10% FBS and were allowed to attach for 6-24 h before treatment.
THP-1 and HUT-78 cells were cultured with addition of fresh medium every 2-3 days. Cells were subcultured at a concentration of 1 × 10 6 ml −1 with addition of media to a concentration of 3 × 10 5 cells ml −1 . Cells were used only up through passage 15 or one month of culture. The THP-1 and HUT-78 media comprised RPMI-1640 (30-2001, Sigma) with 10% or 20% FBS, respectively. Cells were seeded at 3 × 10 5 ml −1 in 384-well clear-bottom plates as above, using RPMI-1640 and 10% FBS.
MM.1S, U266 and KMS-12-BM cells were cultured in RPMI-1640 containing 10% FBS and Antibiotic-Antimycotic. Cells were seeded at a density of 25,000 cells well −1 in half-area 96-well plates (675096, Greiner Bio-One) and were serum-starved for 2 h before treatment.
Except where otherwise indicated, cells were pre-treated for 1 h with DMX-5804, then DOX (Sigma-Aldrich) was added for 24 h at the indicated concentrations. All cell lines were maintained in 5% CO 2 and at 37 °C.
For in-cell dose-response experiments, MAP4K4 inhibitors were prepared in 96-well polypropylene plates as serial dilutions in DMSO of 10 mM stock solutions 11 . To a sterile 96-well intermediate plate, 1-3 µl of each well was added to 99 µl of medium, the samples were mixed, and 5 µl well −1 from the intermediate plate were transferred to the cells (final concentration, 0.1-0.3% DMSO; the final top concentration of inhibitor was, typically, 10-30 μM). Subsequently, DMX-5804 was used at 10 μM, added as 5 µl well −1 of a 100 μM stock solution. The death triggers including DOX were prepared fresh on the day of treatment at 10 × the final concentrations, and 5 µl well −1 were added as appropriate. Assay plates were incubated at 37 °C for the duration of treatment. cell viability assays. ATP generation. Cell ATP levels were measured as described 11 . Assay plates were removed from the incubator and allowed to reach room temperature. CellTiter-Glo (CTG) reagent (Promega) was added (20 µl well −1 ), with gentle agitation for 30 min. Luminescence as a measurement of cellular ATP levels was read on a PHERAstar Plus microplate reader (BMG Labtech). Results were normalised to those of untreated www.nature.com/scientificreports/ control cells (no death signal, no MAP4K4 inhibitor) and to 100% cell death (addition of 0.1% Triton X-100, 2 h before CTG). Normalised values were plotted against the log concentration of the death inducer or inhibitor.
Caspase activity. The Caspase-Glo 3/7 Assay (Promega) was used to measure executioner caspase activity. Luminescence was measured 6-24 h after treatment (PHERAstar Plus) and is expressed as absolute luminscence or the ratio relative to vehicle-treated control cells.
Mitochondrial membrane potential (∆ψ m ). Tetramethylrhodamine, methyl ester (TMRM; Sigma-Aldrich) was added to cells at a final concentration of 30 nM, for 30 min in the incubator prior to treatment. TMRM fluorescence was measured using an IncuCyte S3 system (EssenBio; excitation 585 nm, emission 635 nm, 20 × objective) or CLARIOstar (BMG Labtech; excitation 544 nm, emission 645 nm, direct optic bottom reading). Using the IncuCyte, each plate was examined serially, every hour for 20 h. Data were determined as the integrated fluorescence intensity for 2 sites per well, using IncuCyte S3 software. The CLARIOstar readout was total fluorescence across the well, measured by spiral averaging. Data are shown as the integrated fluorescence (IncuCyte), total fluorescence (CLARIOstar), or ratio relative to vehicle-treated control cells. Alternatively, ∆ψ m was assessed in IMR-90 cardiomyocytes using the JC-10 Mitochondrial Membrane Potential Assay Kit-Microplate (Abcam). Cardiomyocytes were cultured on half-area, 96-well, black-walled, clearbottom, fibronectin-coated plates (675096, Greiner Bio-One). JC-10 dye-loading solution was added at 25 µl well −1 and incubated in a CO 2 incubator at 37 °C for 30 min followed by addition of 25 µl well −1 of assay buffer B. The fluorescence intensities of J-aggregates and monomeric forms were determined in a microplate reader using the excitation/emission wavelengths 490/525 nm and 540/590 nm, respectively. Data are expressed as the ratio of aggregate/monomeric forms, relative to untreated control cells.
TUNEL staining. vCor.4U cells were seeded at a density of 25,000 well -1 in full-area 96-well plates (Greiner Bio-One) and treated as detailed for the ATP viability assay. The cultures were then washed in cold PBS, fixed using 4% PFA for 15 min, and stained using the Click-iT Plus TUNEL Assay (Invitrogen). Fluorescence was scored using a Cellomics ArrayScan VTI High Content Screening platform and analysed using HCS Studio Software.
Flow cytometry. U266 and MM1.S cells were treated as above and fixed with 4% PFA for 15 min at 4 °C. Staining was performed using SYTOX Blue (Thermo Fisher) for hypodiploid DNA or the Pacific Blue Annexin V Apoptosis Detection Kit with PI (Biolegend), according to the respective manufacturer's instructions. Flow cytometry was performed using an LSRII flow cytometer (Becton Dickinson) equipped with 355 nm ultraviolet, 405 nm violet, 488 nm blue, 561 nm yellow-green and 638 nm red lasers. Data were analysed using FlowJo v10, using the Cell Cycle plug-in.
QRT-PCR. Gene expression was quantified as described 11 . RNA extraction was performed using PureLink RNA Mini Kits (Life Technologies). RNA quality and quantity were assessed using a NanoDrop 1000 spectrometer (Thermo Fisher Scientific). RNA was converted to cDNA using High-Capacity cDNA Reverse Transcription Kits (Applied Biosystems). QRT-PCR was performed using TaqMan Gene Expression Assays, MicroAmp Optical 384-well plates, 2X TaqMan Gene Expression Master Mix, and a QuantStudio G Flex Real-Time PCR System (Thermo Fisher Scientific). Results were normalized to UBC as the loading control. calcium transients. Human cardiac calcium oscillation was assessed using FLIPR Tetra instrumentation and FLIPR calcium assay kits (Molecular Devices), as described 11 . vCor.4U cells were plated for 4 days, then were subjected to graded concentrations of DOX ± MAP4K4 inhibitors, as described for the viability assays. Cells were incubated for 24 h at 37 °C in 5% CO 2, using 25 μl well −1 of loading dye concentrate as the fluorescent calcium indicator (EarlyTox Cardiotoxicity Kit, R8210; Molecular Devices). Plates were then removed from the incubator and allowed to reach room temperature. Intracellular calcium oscillations were monitored using the excitation LED bank 470-495 nm and emission filter set 515-575 nm. Fluorescence intensity signals were acquired for 150 s at 2 ms intervals. Beat frequency, total peak area, median peak height, and median peak width were calculated and compared with the baseline control and with DOX alone. Peak parameters were evaluated using GraphPad Prism 6. Baseline fluorescence was removed for each well, and data were analysed as the area under the curve. Peaks were defined as ≥ 20% of the difference from minimum to maximum intensity.

Quantitation and statistical analysis
Data are reported as the mean ± standard error, using a significance level of P < 0.05. Data were analyzed by oneor two-way ANOVA, using the Sidak or Bonferroni test for multiple comparisons and Welch's t-test for pairwise comparisons (GraphPad Prism 7-8) 11 . Where multiple experiments are pooled, the technical replicates (separate cultures in each study) were averaged and treated as a single data point; where representative data are shown,