LAUSR

Dear Editor, Human pluripotent stem cell (hPSC)-derived cardiomyocytes (CMs) are of significant translational value to in vitro studies of human cardiac development, drug and cardiotoxicity testing and cardiac disease modelling. Differentiation of hPSCs to CMs, however, yields mixed cultures of atrial-, ventricular-, and pacemaker-like cells as well as non-CMs in variable proportions. Strategies to enrich CMs from non-CMs and to generate ventricular versus atrial cells have been successful; however, enriched hPSC-CMs are developmentally immature and fail to recapitulate key functional traits that are fundamental to the (patho)physiology of adult CMs. Moreover, these strategies do not adequately address experimental variabilities caused by differences in the genomes and differentiation capabilities of diverse hPSC lines. One validated approach that overcomes issues of cell heterogeneity and experimental variability is immunophenotyping; however, accessible markers suitable for defining mature, live CMs are lacking. Although cell surface markers such as SIRPA(CD172a) and VCAM1(CD106) have been used to sort for hPSC-CMs, they do not distinguish between maturation states. Here, we report the identification of CD36 as a cell surface marker of maturation, which can be used to reduce experimental variability and improve drug screens. To identify cell surface proteins informative of more mature CMs, we qualitatively profiled the surfaceome of cardiac Troponin T (TNNT2)-positive (>95%) human embryonic stem cell (hESC) (H7)-derived CMs (Supplementary information, Fig. S1) using mass spectrometry-based cell surface capture (CSC) technology. We identified 525 N-glycoproteins composed of membrane (85.5%), glycosylphosphatidylinositol-anchored (5.9%), and extracellular matrix (ECM) (5.0%) proteins. Cluster of differentiation (CD) molecules comprised 14% of the total glycoproteins identified (Supplementary information, Table S1). Comparisons with the Cell Surface Protein Atlas (CSPA) revealed 12 CD molecules with cardiac-prevalent expression (Supplementary information, Fig. S2). Among these, only CD36, a fatty acid translocase implicated in the uptake of long-chain fatty acids (FAs) in adult heart, was upregulated in CMs as a function of differentiation time and differentially present on the surface of sub-populations of CMs (Fig. 1a, b, Supplementary information, Fig. S3). CD36 is present on the surface of multiple cell types including CMs, hematopoietic cells and adipocytes (Supplementary information, Fig. S2), but it is absent from undifferentiated hPSCs (Fig. 1b) and fibroblasts (not shown). Comparisons with public datasets show that CD36 transcripts increase in mouse and human hearts during in vivo development and are more abundant in ventricles than in atria from late embryonic mouse hearts (Supplementary information, Fig. S4). Human PSC-CMs could be sorted into CD36 or CD36 cell populations with unique traits. Differentiation day (D) 45 CD36 cells were enriched for α-actinin CMs, whereas CD36 cells contained both α-actinin CMs and α-actinin non-CMs (Supplementary information, Fig. S5). Sorting, after co-staining of CD172a (Supplementary information, Fig. S6) or pre-treatment with lactate, could eliminate most non-CMs. The percentages of CMs positive for CD36 at D45 ± 5 ranged from 34–78, 18–72, 30–81%, to 42–68% in differentiated cultures of H7 and H9 ESC lines and MD1and JHU001-induced pluripotent stem cell (iPSC) lines, respectively (Supplementary information, Fig. S7). CMs derived from hPSC lines thus display a large degree of intraand inter-line variability with respect to CD36. In atrial-like CMs induced with retinoic acid (RA), the fluorescent signal intensities for CD36 CMs were lower than those observed from ventricular CMs derived from H7 or JHU001 lines (Supplementary information, Fig. S8). Sorted CD36 CMs (Supplementary information, Fig. S6 for gating) have traits characteristic of a more mature state. CD36 CMs have similar morphology but on average a 10% greater mean surface area, compared to CD36 CMs (Supplementary information, Fig. S9a, b). At D45, little to no cell proliferation could be demonstrated, and higher percentages of binucleate cells negative for Ki67 were observed in the CD36 (9.9 ± 0.8) versus CD36 (4.5 ± 0.8) CMs (Supplementary information, Fig. S9c–e). This latter trait correlated with elevated centromeric/kinetochore RNAs for CENPH and CENPM in CD36 CMs (Supplementary information, Tables S2 and S3). Structurally, CD36 CMs have more organized sarcomeres with elevated sarcomeric scores (Fig. 1c) relative to CD36 CMs, but myofibril widths did not significantly differ (CD36: 1.17 ± 0.23, n= 14; CD36: 1.28 ± 0.21, n= 16). Organized sarcomeres with H zones, A bands, and Z-disks could be clearly discerned in CD36 CMs (Supplementary information, Fig. S9f). Early ventricular hPSC-CMs express both the ventricular and atrial myosin regulatory light chain isoforms (MLC2V and MLC2A), but in more mature ventricular CMs, the cells express predominantly MLC2V. Consistently, the MLC2V to MLC2A ratio was higher in CD36 than in CD36 CMs (Supplementary information, Fig. S9g). Transcripts encoded from developmentally regulated genes such as cardiac troponin I and MLC2V were also increased by 1.40 ± 0.14 and 6.80 ± 3.56 fold, respectively in CD36 versus CD36 CMs (Supplementary information, Fig. S9h). T-tubules were not observed in any cell subpopulation. Functionally, CD36 CMs (±RA) from all hPSC lines tested had significantly lower spontaneous beating frequencies (Supplementary information, Fig. S10a) and a decreased incidence of beating rate variability (i.e., arrhythmias) relative to CD36 CMs. Electrophysiologically, all cells differentiated without RA, displayed ventricular-like action potentials (AP). CD36 CMs had a prolonged plateau phase, a decreased diastolic depolarization (DD) slope and similar maximum diastolic potential (MDP) compared with CD36 CMs (Fig. 1d, Supplementary information, S10b). Relative to CD36 CMs, CD36 CMs had significantly increased AP amplitudes, reduced spontaneous firing frequencies, longer AP durations and a trend towards higher maximal upstroke velocities (Fig. 1e and Supplementary information, Fig. S10b). Monolayer