Stem Cell-Derived Heterocellular Atrial Engineered Cardiac Tissue With Comparisons to Native Human Atrial Myocardium

Abstract

There is a need for robust in vitro models of human atrial tissue to empower mechanistic disease research, drug discovery, toxicity screening, and precision medicine. In the present study, we employed atrial-like human-induced pluripotent stem cell-derived cardiomyocytes (hiPSC-aCM) and hiPSC-cardiac fibroblasts to produce and evaluate atrial-like engineered cardiac tissue (aECT) constructs compared with adult native human atrial myocardium. Utilizing various techniques to evaluate ultrastructure, molecular makeup, contractile function, and electrophysiology, we compare these aECT to ventricular-like engineered cardiac tissue (vECT) and native atrial myocardium. First, aECT demonstrated higher spontaneous beating rate, lowered IRX4 mRNA expression, and an atrial-like expression of contractile mRNA and protein with higher MYL7/MLC2A and lower MYL2/MLC2V, compared to vECT, following similar patterns exhibited by native myocardium. Secondly, aECT exhibited ultrastructural features like native atrial myocardium, including lower cardiomyocyte circularity, higher dimensional cardiomyocyte anisotropy (i.e., rod-shaped), higher caveolae abundance, and higher sarcomere alignment. Importantly, aECT showed contractile parameters similar to that previously observed in native atrial myocardium with minimal differences between the two in twitch force, contraction and relaxation times, and contraction kinetics. Electrophysiological data also showed that aECT exhibit atrial-like action potential morphology, with shorter action potential duration, lower APD20/80 ratio, and higher repolarization fraction. Electrophysiological data were accompanied by elevated potassium channel mRNA expressions, compared to vECT. Overall, we have generated aECT with atrial-like phenotypes, compared to vECT and native atrial myocardium that can be leveraged for drug testing and disease modeling of atrial electro-anatomical remodeling and contractile dysfunction that occurs during atrial pathology.

New & Noteworthy:

We demonstrate an atrial-like engineered cardiac tissue that recapitulates adult human atrial contraction, for both kinetics and force production. Furthermore, we include numerous previously unmeasured ultrastructure metrics such as sarcomere alignment, cardiomyocyte anisotropy, and caveolae abundance. Our constructs may provide a human cardiac tissue platform for drug testing to identify combinatorial therapies to address atrial contractile dysfunction and diseases linked to cardiomyocyte ultrastructural defects.

Introduction:

Engineered heart tissue generated from human induced pluripotent stem cell - cardiomyocytes (iPSC-CMs) has proven to be a powerful model for the human heart enabling studies ranging from basic research to disease modeling (1, 2). However, efforts to generate and evaluate atrial-specific engineered cardiac tissue (aECT) compared with adult human atrial myocardium have lagged, with the resulting constructs being less characterized, despite the large burden of atrial-specific pathology such as atrial fibrillation (3). hiPSC-aCM are commonly generated by including retinoic acid during differentiation, which promotes atrial-like phenotypes while limiting ventricular-like phenotypes both in vitro and in vivo (4-7). To generate ECT, not only are CMs required but a mesenchymal or fibroblast-like population is included to help remodel the extracellular matrix and the construct (8). However, some prior studies have generated aECT without adding a defined population of fibroblasts, rather relying on noncardiomyocytes present in the differentiate population of CMs (9-10). Furthermore, published work has extensively examined molecular composition and electrophysiology with respect to arrhythmogenesis (11, 12), but have been less focused on examining contractile function in detail, a significant pathogenic component of various atrial diseases (13-15).

In the present study, we generated a human cell-based model that expands on previous work (6, 14, 16-18) by utilizing a robust monolayer protocol to differentiate atrial-like iPSC-CMs and subsequently combine them with iPSC-derived second heart field cardiac fibroblasts to form highly reproducible aECTs and by uniquely examining direct comparisons between our aECT and both native human atrial myocardium and hiPSC-aCM in monolayer. We provide contractile data demonstrating atrial-like contraction which is known to play an important role in blood clot development and stroke that commonly occurs during AF (17, 18), where atrial contractile dysfunction is common (19). Importantly, in our ECT we included hiPSC-cardiac fibroblasts (CF) which are known to play a critical role in modulating atrial stress response and arrhythmogenesis in adult atria. We also provide novel ultrastructural comparisons between atrial-like hiPSC-aCM in monolayer (ML), hiPSC-aCM in ECT constructs, and native healthy human atrial tissue demonstrating improved structural maturity in ECT-incorporated hiPSC-aCM.

Overall, we demonstrate that hiPSC-aCM maintain atrial-like phenotypes when cocultured with second heart field-like hiPSC-CF in ECT and that atrial-like ECT possess an improved atrial contractile and ultrastructural phenotype compared to atrial-like ML, with some features comparable to native human atrial tissue. In the future, we believe these atrial-like ECT may be suitable for modeling atrial cardiomyopathy and atrial hypocontractility associated with atrial fibrillation where force production can be hindered (20–22), mechanistic research underlying atrial force production and ultrastructure, and atrial-related drug development and toxicity testing.

Methods:

hiPSC-CM Differentiation

HiPSCs derived from the DF19–9-11T.H line were differentiated into CMs using a small molecule-directed (GiWi) protocol as previously described (23). Briefly, hiPSCs maintained on the StemFlex/Matrigel system were dissociated into single cells and seeded onto Matrigel-coated six-well plates at 2.0 × 10⁶ cells/well in StemFlex medium. Cells were cultured for 5 days in StemFlex medium with daily media changes. For ventricular differentiation of hiPSC (hiPSC-vCM), on day 0, StemFlex medium was replaced with 2 mL/well RPMI supplemented with B27 without insulin (Gibco) supplemented with 10 μM CHIR99021 (GSK-3 inhibitor, Biogems). Precisely 24 h later (day 1), cell culture medium was changed to 3 mL/well RPMI + B27 without insulin and for 48 h (day 3). On day 3, the medium was changed to 3 mL/well RPMI + B27 without insulin supplemented with 5 μM IWP-4 (Biogems), with half the media consisting of D1–3 hiPSC-conditioned medium. Forty-eight hours later (day 5), the medium was replaced with 3 mL RPMI + B27 without insulin. The medium was changed to RPMI + B27 complete supplement (with insulin) (Gibco) on day 7. Differentiated cells were maintained in this medium until day 15 with medium changes every 24–48 h. On day 15, differentiated cells were dissociated with 10× TrypLE (Thermo Fisher Scientific) according to the manufacturer’s protocol. Following resuspension in StemFlex medium, cells were replated on SyntheMax (Corning) coated six-well plates at 2.0 × 10⁶ cells/well. Forty-eight hours after replating hiPSC-vCMs were purified using Lactate media, made with RPMI 1640 no glucose (Life Technologies), B27 supplement, and 0.02% (2.66 mM) lactate (Sigma-Aldrich) for 7 days with media changes every 24–48 h. After selection, CMs were maintained in RPMI with B27 supplement until day 30 at which point hiPSC-vCMs were dissociated for hiPSC-ECT generation. For hiPSC-aCM generation, all-trans retinoic acid was supplemented (0.75 μM) to differentiating hiPSC between days 4–6 in addition to other small molecules (16).

Stem Cell Cardiac Fibroblast Culture

DF19–9-11T.H hiPSC-cardiac fibroblasts (CFs) were differentiated as previously described and cultured in FibroGRO-LS medium (SCMF001; Millipore Sigma) in uncoated six-well culture plates (Corning) with passaging every 4–5 days. The medium was replaced every 24–48 h. Low passage numbers (<12) were used for hiPSC-ECT generation (24).

Flow Cytometry

As previously described (23), cells were dissociated with 10 Tryple (Thermo Fisher Scientific) and neutralized with an equal volume of EB20 media. Cells were counted, and 500,000 cells were apportioned for immunolabeling. Cells were fixed in 1% paraformaldehyde, washed with FACS buffer (DPBS, 0.5% BSA, 0.1% NaN₃), centrifuged (200 g at 4°C for 5 minutes), and resuspended in roughly 50 μL of FACS buffer. Fixed cells were labeled with monoclonal anti-cTnT (IgG1, Thermo Fisher Scientific, A-21121, 1:200 dilution), according to the manufacturer’s instructions. Cells were then washed with FACS buffer plus 0.1% Triton X-100 and centrifuged (200 g at 4°C for 5 minutes), and all but 50 μL supernatant was aspirated. Secondary antibody (IgG, Thermo Fisher Scientific, MS-113-P1) and was diluted at 1:1,000 in FACS buffer plus 0.1% Triton X-100 and samples were incubated at room temperature for 30 minutes, washed in FACS buffer, and then resuspended in FACS buffer for analysis. Data were collected on an Attune Nxt (Thermo Fisher Scientific) flow cytometer and collected data were analyzed using Floreada.io and FlowJo.

ECT Generation

Day 30 DF19–9-11T.H hiPSC-CMs were dissociated with 10x TrypLE (Thermo Fisher Scientific) according to the manufacturer’s protocol and counted using a hemocytometer. hiPSC-CMs were subsequently resuspended in fibrin ECT media (60.3% high-glucose DMEM; 20% F12 nutrient supplement; 1 mg/mL gentamicin; 8.75% FBS; 6.25% horse serum; 1% HEPES; 1× nonessential amino acid cocktail; 3 mM sodium pyruvate; 0.004% (wt/vol) NaHCO3; 1 μg/mL insulin; 400 μM tranexamic acid; and 17.5 μg/mL aprotinin) incubated for at least 1 h on a rotating platform at 37°C to form small and uniform clusters of viable CMs. DF19–9-11T.H CFs were dissociated using 1× TrypLE (Thermo Fisher Scientific) according to the manufacturer’s protocol and counted using a hemocytometer. Following rotational culture, 2 × 10⁶ hiPSC-CMs were mixed with 2 × 10⁵ hiPSC-CFs in 200 μL fibrin ECT media per hiPSC-ECT, a ratio similar to that previously used in the generation of 3D cardiac constructs. To this cell mixture, 1.25 mg/mL fibrinogen and 0.5 unit of thrombin were added. This cell-matrix mixture was rapidly mixed and loaded onto a 20 × 3-mm cylindrical mold of FlexCell Tissue Train silicone membrane culture plate and incubated under pre-programmed vacuum condition for 60 min at 37°C supplied with 5% CO₂ to allow for attachment of the ECT constructs to the nylon tabs at each end of the Tissue Train well (Fig. 1A). Following polymerization of the fibrin matrix, ECTs were fed with ECT medium, carefully separated from the plate surface with a sterile pipette, and cultured for 30 days, with fibrin ECT medium changes every 2–3 days (8).

Figure 1.

aECT maintain cardiac phenotype. A: The GiWi protocol is used to differentiate hiPSC-CM, with all-trans retinoic acid being used to enrich atrial-like hiPSC-CM. All hiPSC-CM undergo lactate purification for 7 days, at D30 they are combined with hiPSC-CF (cardiac fibroblasts) to generate ECTs which are cultured and subsequently tested at D60. B-D: Flow cytometry demonstrates high hiPSC-CM purity at D30 prior to ECT generation for both ventricular-like (Ven) hiPSC-CM (C, n=3) and atrial-like hiPSC-CM (D, n=3). E: Representative image of whole ECT and at 10x magnification. F: Atrial-like ECT demonstrate higher automaticity in culture (E, n=16). G-J: RT-qPCR analysis of cardiac-, ventricular-, and atrial-specific mRNAs for human left ventricle (HsLV), human right atrial appendage (HsRAA) as well as ventricular-like (Ven) and atrial-like ECT. n=4–6 healthy human samples and n=15 ECTs per group. Significances determined by student’s t-test.

Calcium and Contraction Measurements

Contraction traces were measured in hiPSC-ECT using protocols similar to those previously describe (8). Briefly, each hiPSC-ECT construct was transferred from the culture dish to a model 801B small intact fiber test apparatus (Aurora Scientific) in Krebs–Henseleit buffer: 119 mmol/L NaCl, 12 mmol/L glucose, 4.6 mmol/L KCl, 25 mmol/L NaHCO₃, 1.2 mmol/L KH₂PO₄, 1.2 mmol/L MgCl₂, 1.8 mmol/L CaCl₂, gassed with 95% O₂-5% CO₂ (pH 7.4). hiPSC-ECT constructs were attached with sutures between a model 403A force transducer (Aurora Scientific) and a stationary arm and perfused with 37°C Krebs–Henseleit buffer at a rate of 1 mL/min, and field stimulation initiated at 1.5 Hz (2.5 ms, 12.5 V). The longitudinal length of each construct was increased stepwise until maximal twitch force was achieved in order to establish the Frank–Starling relationship. Constructs were allowed to equilibrate for 20 min with constant perfusion. After equilibration, twitch force production was measured with pacing at a frequency of 1.5 Hz both at baseline and following 5 min preincubation with 1 μM isoproterenol. Data from force measurements was analyzed using IonWizard 6.0 software (IonOptix). Under each condition, contractions transients of 40–60 successive contractions were collected and averaged. These data were exported to Microsoft Excel and the kinetics of force generation and relaxation were calculated.

For calcium measurements, hiPSC-ECTs were subsequently incubated with Fura-2 loading solution for 30 min at 37°C consisting of Krebs–Henseleit buffer supplemented with 5 μM Fura2-AM (Invitrogen) and 1% (vol/vol) Chremophor EL (Sigma) with constant oxygenation (95% O₂, 5% CO₂). After Fura-2 loading, ECTs were rested for 40 min with perfusion with Krebs–Henseleit buffer at a rate of 1 mL/min, and field stimulation initiated at 1 Hz. Simultaneous twitch force and calcium transients (CaTs) were recorded at pacing frequencies ranging from 0.5 to 2.5 Hz. Fura-2 fluorescence was measured by alternately illuminating the preparation with 340- and 380-nm light (at a frequency of 250 Hz) while measuring the emission at 510 nm using IonOptix hardware and software (IonOptix Corporation, Milton, MA). Emitted fluorescence and force data were stored for analysis as the 340- and 380-nm counts and as the ratio R = F₃₄₀/F₃₈₀ with data from force and intracellular free CaT measurements analyzed with IonWizard 6.0 software (IonOptix). For each condition, force and/or intracellular CaTs of 40–60 successive contractions were collected and averaged. These data were then exported to Microsoft Excel and the magnitude of force generated and/or CaTs, as well as the kinetics of force generation and relaxation and/or calcium release and sequestration calculated.

Optical Mapping

As previously described (25), optical action potentials were captured using the FluoVolt membrane potential probe (F10488; Life Technologies). The staining solution contained 1 μM FluoVolt supplemented with a 1:1,000 dilution of 20% pluronic acid (P2443; Sigma-Aldrich). ECTs were incubated with the staining solution for 30 minutes in the 5% CO₂ incubator at 37°C and subsequently washed and incubated in Tyrode’s solution containing 154 mM NaCl, 5.4 mM KCl, 1 mM MgCl₂, 1.8 mM HEPES-NaOH, 1.8 CaCl₂, and 5.5 mM D-glucose (pH 7.4) at 36.5 ± 0.5°C. Excitation light was generated by LED (X-Cite XYLIS, Excelitas Technologies, CA, USA) with an excitation filter (475/35 nm; Thorlabs, NJ, USA). The emitted light was long-pass filtered (520/35 nm; Thorlabs, NJ, USA) for the action potential signal. Emission was captured by a MiCAM Ultima-L CMOS camera (SciMedia, CA, USA) offering high spatial (100 × 100 pixels, 60 ± 10 μm/pixel) and temporal (200 frames/second) resolution. The fluorescence signals were digitized, amplified, and visualized through proprietary software (SciMedia, CA, USA). ECTs were field stimulated with a rectangular pulse of 2 ms at 1 Hz rate, 50% above threshold.

Collection of Human Cardiac Tissues

Human heart collection protocols were approved by the University of Wisconsin Institutional Review Board. Non-failing human hearts that went unused for organ transplant were obtained from the University of Wisconsin Organ Procurement Organization, Madison, WI. At the time of harvest, hearts were aseptically excised, cardioplegically arrested (cardioplegic solution, in mmol/L: NaCl 110, CaCl₂ 1.2, KCl 16, MgCl₂ 16, NaHCO₃ 10; pH=7.65 ± 0.05; 4°C) in the operating room following crossclamping of the aorta and then transported to the research laboratory on ice as described previously in detail (26, 27). Cardioplegic perfusion removed all the blood and protected the hearts from ischemia during the subsequent period of tissue isolation. Human right atrial and left ventricular tissues of both sexes were isolated and either frozen in liquid nitrogen for Westen Blot and RT-qPCR or fixed for immunohistological staining.

Western Blot

As previously described (28), using bicinchoninic acid assays to determine protein concentration, 10 μg of protein was loaded to each gel (Bio-Rad, 4%–20% Tris-glycine) lane to not exceed the quantifiable range of glyceraldehyde 3-phosphate dehydrogenase (GAPDH). Gels were transferred to polyvinylidene fluoride (PVDF) via the Power Blotter system (Invitrogen) set for mixed-range molecular mass. Membranes were blocked for 1 h at room temperature with 3% non-fat milk and incubated overnight (∼16 h) with a primary antibody (1:500–1:1,000) at 4°C. Corresponding Alexa Fluor-488 fluorescent secondary antibodies were used to evaluate protein expression. Primary antibodies for GAPDH (mouse monoclonal, G8795, Millipore Sigma), MLC2A (mouse monoclonal, 311–011, Synaptic Systems), and MLC2V (rabbit polyclonal, 10906–1-AP, Proteintech) have been validated previously (23, 28).

Real time quantitative reverse transcription polymerase chain reaction (RT-qPCR) analysis

Whole hiPSC-ECTs were snap frozen in liquid nitrogen post physiological testing and stored at −80°C. RNA was isolated as previously described (29). hiPSC-ECTs were homogenized using TRIzol reagent (Ambion) and molecular-grade chloroform was added according to the manufacturer’s instructions. After mixing, incubation, and centrifugation, the aqueous phase containing RNA was collected. RNA was further purified using the Qiagen Miniprep kit according to the manufacturer’s instructions. RNA was quantified and quality was assessed using a NanoDrop spectrophotometer (Fisher Scientific). 50–100 ng total RNA was reverse transcribed into first-strand cDNA with the iScript Reverse Transcription Supermix for RT-qPCR (Bio-Rad) following manufacturer’s protocols. 1 ng of resulting cDNA was used for qPCR analysis. Taqman probes for assayed genes and appropriate controls were arrayed in MicroAmp Optical 96-well Reaction Plates (Applied Biosystems), and the PCR performed using the TaqMan Gene Expression Master Mix (Applied Biosystems). Real-time monitoring of TaqMan fluorescence was performed on the Bio-Rad RT-qPCR system CFX96. An initial activation step of 2 min at 50°C and 95°C for 10 min was followed by 40 cycles of 15 s of denaturation at 95°C and 60 s of annealing/extension at 60°C. Data were analyzed in Excel, using a ΔΔCT method as previously described (29), with GAPDH used as a housekeeping control gene. TaqMan assay IDs used were GAPDH (Hs02786624_g1), Cav3 (Hs00154292_m1), IRX4 (Hs00212560_m1), TNNT2 (Hs00943911_m1), NPPA (Hs00383230_g1), KCNJ3 (Hs04334861_s1), KCNA5 (Hs00969279_s1), NR2F2 (Hs00819630_m1), MYH6 (Hs01101425_m1), MYL7 (Hs01085598_g1), MYL2 (Hs00166405_m1), MYH7 (Hs01110632_m1), MYL4 (Hs00975401_m1), and MYBPHL (Hs02786624_g1). All TaqMan assays were obtained from Applied Biosystems unless indicated otherwise.

Transmission Electron Microscopy (TEM)

ECTs were fixed and prepared for transmission electron microscopy similarly to previous work (8, 29-30). Briefly, samples were immersion fixed in 2.5% glutaraldehyde, 2.0% paraformaldehyde buffered in 0.1 M sodium phosphate buffer for 2 h at room temperature. After osmium post-fixation, dehydration, and embedding, each sample was sectioned on a Reichert-Jung Ultracut E ultramicrotome at 80 nm. The sections were collected on formvar coated 2×1 mm slot Cu grids (EMS Hatfield, PA), and post-stained with uranyl acetate and lead citrate. The sectioned samples were viewed at 80 kV on a Philips CM120 transmission electron microscope, equipped with AMT BioSprint12 digital camera (AMT Imaging Systems, Woburn, MA). Images for caveolae and convolution index were obtained at 40,000x magnification. Caveolae were considered 50–100 nm flask-shaped membrane invaginations located within 50 nm from the cell surface. Caveolae were counted in each of the 40,000x magnification images and identified as sub-sarcolemmal (no visible connection to the sarcolemma in the image plane) or visibly integrated into the sarcolemma (flask-like connection to the sarcolemma observed). Total density of caveolae and their localization were quantified, with respect to the membrane length. Caveolae density was calculated from each image by counting the number of caveolae and normalizing to the membrane length.

ECT Immunostaining

As similarly performed (8), ECTs were first embedded in OCT and frozen via dry ice on a flat surface and stored at −80^oC. ECT were then cryosectioned at 6 μm onto charged glass slides and dried for 30 min in a laminar flow hood. Once dry, ECTs were fixed in 4% PFA at room temperature for 15 min and then were dried similarly as before. ECTs were then washed with running tap water for 10 min to remove residual OCT and rehydrated with 3 × 5 min TBS washes. Permeabilization was performed with 0.15% Triton-X-100 in TBS for 10 min, subsequently washed 2 × 5 min with TBST, and then blocked with TBST with 0.15% Triton-X-100, 5% normal goat serum and 2 mg/ml bovine serum albumin for 2 h at room temperature. ECTs were stained with primary antibodies for 1 h (1:200) in a humidity chamber and then washed 3 × 5 min in PBST. Secondary antibodies were incubated for 1 h in blocking buffer (1:200) and then washed 3 × 5 min in TBST. ECTs were then cover slipped with Prolong Gold Antifade Reagent with DAPI which was sealed with clear nail polish.

Monolayer Immunostaining

As previously performed (31), prior to fixation, hiPSC-aCM in monolayer were replated on coverslips at 100% confluency. 48 h after plating, monolayers were fixed in 4% PFA for 15 min, washed 2x with sterile PBS, and then stored at 4^oC in sterile PBS. Monolayers were permeabilized with 0.2% Triton X-100 for 1 h at room temperature followed by 2 washes of PBS. Primary antibodies (1:200) were added in 0.1% Triton X-100, 1% BSA in PBS solution and incubated overnight at 4°C. Following primary incubation, samples were washed with 0.2% Tween 20 in PBS twice and 1X PBS twice. Appropriate secondary antibodies specific to the primary IgG isotype were diluted (1:1,000) in the same solution as the primary antibodies and incubated at room temperature for 1.5 h in dark on a rotator. Samples were washed with 0.2% Tween 20 in PBS twice and 1X PBS twice. Coverslips were sealed with Gold Anti-fade Reagent with DAPI (Invitrogen) and clear nail polish on glass slides.

Human tissue immunostaining

At the time of harvest, hearts were aseptically excised, perfused and stored in cold cardioplegia solution (in mmol/L: NaCl 110, CaCl₂ 1.2, KCl 16, MgCl₂ 16, NaHCO₃ 10; pH=7.65±0.05; 4°C) while being transported to the experimental laboratory on ice as described previously in details (32). After isolation, the preparations were fixed in 10% phosphate buffered formalin for 48 h, then moved to 70% ethanol for at least 24 h at 4°C. Tissues were dehydrated by applying a series of organic solutions, paraffin embedded, and sectioned. Slides were stained as previously described (30). Briefly, paraffin sections were dewaxed and heated in a citrate-based buffer for antigen retrieval before the straining protocol. After blocking the slides with bovine serum agglutinin and goat serum (5%) for 2 h, solution was changed to primary antibodies for 24 h. Ryanodine receptors (1:500, Sigma, anti-rabbit) antibodies were used as primary staining. On the next day, slides were washed with phosphate buffer saline (PBS) and loaded with both WGA-Alexa 488 (20 μg/ml) and Alexa-fluor 568 (1 μM, raised in rabbit) for 2 h. Human heart tissue membrane and proteins were visualized by confocal imaging using Leica SP5 confocal microscopy system following previously published methods (33)

Immunostaining Imaging and Data analysis

Images for both monolayers and ECT were collected using a Leica confocal laser microscope with base laser power set at 25%. Laser powers for DAPI and 488 nm signals were set to 50% and 30%, respectively, for all samples. Human tissues were imaged using 2 lasers 488 for WGA and 594 for RyR channel with a similar power as for monolayers. Z- stacks of images were taken at 40x oil- objective with a 250 μM step between frames (34). Using ImageJ, cardiomyocytes were first identified by the presence of sarcomeres and cell edges were measured for anisotropy and roundness. Cardiomyocyte edges in donor hearts and ECT were then identified longitudinally by the termination of sarcomeric staining and then transversally by a ~90° angle in respect to the longitudinal edges. Cardiomyocyte edges in monolayers were only evaluated on the edge of the monolayer where isolated cells were present (35).

Statistical Analysis

Statistical analyses were performed in Origin or GraphPad Prism software and P-values less than 0.05 were considered statistically significant. All values presented in graphs and in-text are Average±S.D. For all figures, *, **, and *** denote p-values <0.05, <0.01, and <0.001, respectively. If no units are described, the quantification uses absolute units.

Results:

aECT exhibit an atrial-like molecular phenotype

Differentiation of ventricular-like (Ven) and atrial-like hiPSC-CMs (Fig. 1A) yielded similarly pure cardiomyocyte populations (Fig. 1B-D). After ECT generation (Fig. 1E), aECT exhibited a ~3-fold higher level of spontaneous automaticity compared to vECT (2.16 ± 0.52 Hz vs. 0.75 ± 0.27 Hz, p<0.001, Fig. 1F). Post-ECT generation, both vECTs and aECTs exhibited similar expression of muscle-specific mRNAs, including caveolin-3 (Cav3, Fig. 1G) and troponin T (TnnT2, Fig. 1H). Importantly, aECTs demonstrated significantly limited expression of ventricular development mRNA IRX4 (p<0.001; Fig. 1I), but similar expression of NPPA, an atrial-like marker (Fig. 1J). In adult human cardiac tissue, IRX4 mRNA was also expressed higher in human left ventricle (LV) compared to human right atrial appendage (RAA) (Fig. 1I). However, unlike in ECTs, NPPA was found to be more highly expressed (p<0.05) in RAA and TnnT2 exhibited higher expression (p<0.05) in LV (Fig. 1J).

RT-qPCR analysis also revealed an atrial-like expression of myosin mRNA and protein, with aECT exhibiting ~90% lower MYL2 (0.35 ± 0.16 A.U. in aECT vs. 5.14 ± 1.63 A.U. in vECT, p<0.001; Fig. 2A), ~70% higher MYL7 (6.61 ± 2.48 A.U. in aECT vs. 3.54 ± 0.78 A.U. in vECT, p<0.001; Fig. 2B), ~3-fold higher MYH6 (2.17 ± 0.68 A.U. in aECT vs. 0.80 ± 0.67 A.U. in vECT, p<0.001; Fig. 2C), ~60% lower MYH7 (2.70 ± 1.65 A.U. in aECT vs. 4.81 ± 1.90 A.U. in vECT, p<0.01; Fig. 2D) and ~9-fold higher MYBPLC (p<0.001, Fig. 2E) mRNA expression, with no change in MYL4 (p>0.05, Fig. 2F). Importantly, quantification of various myosin mRNAs and proteins revealed striking similarities between native adult human myocardium and corresponding ECT types (Fig. 2). In both RAA and aECT, we observed a significantly higher expression for atrial-enriched myosin mRNAs MYL7 (p<0.01; Fig. 2B) and MYH6 (p<0.001; Fig. 2D). Conversely, significantly higher expressions of ventricular-enriched myosin mRNAs MYL2 (p<0.001; Fig. 2A) and MYH7 (p<0.05; Fig. 2C) were found in both LV and vECT groups. Ratiometric analysis of myosin mRNAs revealed that vECTs and aECTs have significantly different MYH6/MYH7 (α/β heavy chain) ratios of 0.28 ± 0.41 and 1.63 ± 1.16 (p<0.001), respectively, and MYL7/MYL2 ratios of 1.33 ± 0.09 and 22.14 ± 6.03 (p<0.001), respectively. These metrics are similar, but more extreme in adult myocardium, with human LV and RAA also having significantly different MYH6/MYH7 (α/β heavy chain) ratios of 0.06 ± 0.07 and 4.67 ± 1.29 (p<0.001), respectively, and MYL7/MYL2 ratios of 0.10 ± 0.08 and 606.29 ± 655.25 (p<0.001), respectively. To evaluate MLC protein expression, Western blotting for MLC2A revealed expression only in RAA and aECT, with no bands detected for LV or vECT (Fig. 2G). Following subsequent staining with anti-MLC2V antibody, we observed bands only for LV and vECT, with no additional bands revealed for RAA or aECT (Fig. 2G). Quantitative analysis demonstrated higher expression of myosin light chain proteins in adult myocardium, compared to their ECT counterparts relative to GAPDH (Fig. 2G, p<0.05).

Figure 2.

Myosin mRNAs and proteins in atrial-like (a) and ventricular-like (v) ECTs express similar differences compared to their adult human counterparts. A-D: Quantitative analysis relative to GAPDH mRNA of myosin mRNAs MYL2 (A), MYL7 (B), MYH6 (C), MYH7 (D), MYL4 (E), and MYBPHL (F). n=4–6 human (Hs) myocardium samples, n=15 ECTs/group for A-D and 8–11 ECTs/group for E-F. G: Representative Western Blot (left) and quantitative analysis (right) for myosin proteins MLC2V and MLC2A, normalized to GAPDH. n=3 per group. Significances determined by student’s t-test.

hiPSC-aCM in aECT demonstrate improved structural maturity

To characterize structural maturity of aECTs in comparison with atrial monolayers and adult atrial myocardium, we performed immunostaining of sarcomeric proteins as well as transmission electron microscopy (TEM) to evaluate CM ultrastructure and morphology (Figs. 3 and 4). Prior to ECT generation, we determined that atrial-like D30 hiPSC-CM exhibit a ~4-fold higher number of COUPTFII-positive cells (88.81 ± 1.33 vs. 24.47 ± 8.31, p<0.001) suggesting higher atrial purity (Fig. 3A-B). We found that unlike predominantly round-shaped hiPSC-aCM in monolayers (Fig. 3C), in ECT hiPSC-aCM developed into rod-shaped cells (Fig. 3D), a characteristic of adult CM (Fig. 3E). Quantitatively, we calculated ~2.4-fold higher relative length to width ratio (i.e., dimensional CM anisotropy: 6.23 ± 2.19 vs. 2.65 ± 1.31, p<0.001; Fig. 3F) and ~3-fold lower CM roundness (0.23 ± 0.07 vs. 0.72 ± 0.07, p<0.01; Fig. 3G) in hiPSC-aCM in ECT, compared to monolayer. These differences were less stark between aECT and RAA with ~1.3-fold higher CM anisotropy in RAA (7.93 ± 2.21 vs. 6.23 ± 2.19, p<0.05; Fig. 3F) and ~1.5-fold lower CM roundness in RAA (0.35 ± 0.08 vs. 0.23 ± 0.07, p<0.01; Fig. 3G).

Figure 3.

Immunostaining reveals that hiPSC-aCM within aECT are more similar to native human myocardium than 2D monolayer. A: Representative immunostaining images of D30 ventricular-like (Ven) and atrial-like hiPSC-CM in monolayer (COUPTFII – Pink, DAPI - blue, a-actinin – green). B: Quantification of COUPTFII positive hiPSC-CM between ventricular-like and atrial-like D30 hiPSC-CM. C-E: Representative images from immunohistochemistry (DAPI - blue, a-actinin - green) from D60 hiPSC-aCM in monolayer (C), D60 ECT (D), and adult atria (E) with binarized color scheme below. Zoomed in binarized images demonstrating sarcomere alignment between monolayer and ECT atrial hiPSC-CM, as well as adult atria. Analysis of anisotropy (relative length to width ratio, F) and roundness (4π*Area/Perimeter²) (G) in aCM between monolayer, aECT, and adult atria. N=3 per group, 3–6 cell measurements per biological replicate.

Figure 4.

Electron microscopy demonstrates improved maturity of hiPSC-aCM within ECT. A-B: Representative TEM images of hiPSC-aCM in monolayer (A) and aECT (B) at 11500x (top) and 19,500x magnification (bottom). Red asterisks indicate sarcomere z-disks and red arrows indicate caveolae. C-E: Quantitative analysis of sarcomere alignment (C), sarcomere length (D), and caveolae abundance (E). n=3 unique monolayer and ECT per group. Significances determined by student’s t-test.

As determined by TEM analysis (Fig. 4A), structural maturity of hiPSC-aCM in ECT was further complemented by higher sarcomere alignment (as estimated by hiPSC-aCM sarcomere angle: 3.52±2.85º in aECTs vs. 10.86±8.97º in atrial monolayers, p<0.05; Fig. 4A, C). We observed no significant differences in sarcomere length between hiPSC-aCM in ECT vs. monolayer (Fig. 4D); however, these sarcomere lengths fall within the physiological range of adult native CM (36). We calculated 2-fold higher caveolae abundance (1.03 ± 0.57 caveolae/mm in aECT vs. 0.41 ± 0.44 caveolae/mm in atrial monolayers, p<0.01; Fig. 4E) which is suggested to be elevated in more mature iPSC-CMs (37) and adult animal cardiomyocytes (30, 33). Overall, our ultrastructural findings suggest a more developed structural maturity of atrial-like hiPSC-CM in ECTs than in monolayers.

aECT exhibit atrial-like contraction

Analysis of baseline contraction traces paced at 1.5 Hz revealed significant differences between atrial-like and ventricular-like ECTs (Fig. 5), with aECT having a ~80% lower twitch force (0.59 ± 0.04 mN in aECT vs. 2.37 ± 0.86 mN in vECT, p<0.01; Fig. 5B), ~50% lower time to 100% maximum contraction (CT 100: 0.13 ± 0.004 s in aECT vs. 0.22 ± 0.02 s in vECT, p<0.001; Fig. 5C), and ~20% lower time to 50% maximum relaxation (RT 50: 0.10 ± 0.006 s in aECT vs. 0.13 ± 0.007 s in vECT, p<0.001; Fig. 5D). During contraction testing in Krebs-Henseleit buffer, automaticity slowed for both aECTs (1.08 ± 0.65 Hz) and vECTs (0.47 ± 0.36 Hz) (p<0.05, Fig. 5E), allowing for contraction testing at 1.5 Hz.

Figure 5.

aECT demonstrate atrial-like contraction dynamics. A: Representative contraction traces from atrial-like (a) and ventricular-like (v) ECT collected at 1.5 Hz. B-E: Analysis of ECT twitch force (B), time to 100% contracted (CT 100, C), time to 50% relaxed (RT 50, D), and automaticity (E) measured at 1.5 Hz pacing. n=4–8 ECT/group.

Functionally, while paced at 1.5 Hz aECT maintained a similar response to b-adrenergic stimulation (100 mM isoproterenol), remaining non-significantly changed compared to vECT for twitch force and CT100 (p>0.05; Figs. 6A-C), with only time to 90% relaxed (RT 90) demonstrating ~2-fold higher isoproterenol response (8.17 ± 6.42% in vECT vs. 18.38 ± 2.82% in aECT, p<0.05; Fig. 6D). Under b-adrenergic stimulation, aECT exhibited a ~3-fold higher spontaneous beating rate (1.79 ± 0.21 Hz vs. 0.64 ± 0.47 Hz in vECT, p<0.001; Fig. 6E).

Figure 6.

ECT response to b-adrenergic stimulation. A: Representative atrial-like (a) and ventricular-like (v) ECT contraction traces with and without 100 mM isoproterenol stimulation at 1.5 Hz. B-E: Analysis of ECT response to isoproterenol for twitch force (TF, B), time to 100% contracted (CT 100, C), time to 90% relaxed (RT 90, D) at 1.5 Hz pacing, and automaticity (E). n=4 ECT/group. Significances determined by student’s t-test.

For calcium handling (Fig. 7A), while paced at 2 Hz no significant differences were observed between aECT and vECT for maximum CaT and time to 100% maximum calcium level (CaT100) (Fig. 7B and 7C, respectively, p>0.05). Significant differences between aECT and vECT were observed for CaT decay time at 50% (CaTD50: 0.13 ± 0.016 ms in aECT vs. 0.20 ± 0.015 ms in vECT, p<0.001; Fig. 7D).

Figure 7.

Calcium handling in atrial-like (a) and ventricular-like (v) ECTs. A: Averaged ECT CaT traces at 2.0 Hz. B-D: Analysis of CaT for maximum CaT (B), CaT 100 (C), and CaT decay time at 50% (CaDT50, D) at 1.5 Hz. n=4 ECT/group. Significances determined by student’s t-test.

aECT demonstrate atrial-like electrophysiology

To assess electrophysiological characteristics of ventricular-like and atrial-like ECTs, we evaluated action potential (AP) morphology using fluorescent optical mapping (Fig. 8A-F). Under pacing conditions (1 Hz) optical action potentials (OAPs) were recorded from both vECT and aECT, and these results were compared with previously published OAPs recorded from coronary-perfused ventricular/atrial non-failing human heart preparations (n=4) (38). Both native RA myocardium and aECT exhibited a significantly lower APD20/80 ratio (Fig. 8B, p<0.05), as compared to their respective counterparts. While student’s t-test calculated a significantly higher repolarization fraction in RA compared to LV (p<0.05), two-way ANOVA analysis revealed that only aECT had a significantly higher repolarization fraction (calculated as APD50/(APD90-APD50), Fig. 8C, p<0.01), as compared to vECT. Both aECT and RA myocardium demonstrated significantly shorter APDs measured at 90% (APD90, p<0.01, Fig. 8D), 50% (APD50, p<0.01, Fig. 8E), and 20% of repolarization (APD20, p<0.001, Fig. 8F), as compared to their respective counterparts. Except for APD20, all APD quantifications demonstrated that there were no significant differences between ECTs and their adult human counterparts (Figs. 8B-F, p>0.05). These findings were complimented by RT-qPCR data demonstrating higher mRNA expression levels for predominantly atrial-specific ion channel proteins including KCNJ3 (responsible for G protein-activated inward rectifier potassium channel 1, or GIRK-1: 0.023 ± 0.011 A.U. in aECT vs. 0.014 ± 0.007 A.U. in vECT, p<0.05) and KCNA5 (responsible for voltage-gated potassium channel K_v1.5 that underlies the ultrarapid delayed rectifier I_Kur current: 0.023 ± 0.018 A.U. in aECT vs. 0.004 ± 0.001 A.U. in vECT, p<0.01) mRNA expressions (Fig. 8G). The elevated expression of atrial-specific ion channel mRNAs, among others previously described, may be partially due to elevated NR2F2 (0.048 ± 0.020 A.U. in aECT vs. 0.025 ± 0.026 A.U. in vECT, p<0.01, Fig. 8G), which is known to have a major effect on atrial gene expression (39). Finally, to assess absolute values of maximum diastolic potential (MDP) and AP amplitude (APA), we employed sharp microelectrode recordings (Fig. 8H). Analysis of spontaneous microelectrode recordings revealed that aECT show a trend for a more negative MDP (−71.64 ± 9.31 mV in aECT vs. −61.11 ± 5.17 mV in vECT, p = 0.095; Fig. 8I) and have a significantly higher APA compared to vECT (95.4 ± 4.7 mV vs. 86.1 ± 2.3 mV, p<0.05, Fig. 8J), which further support RT-qPCR data demonstrating higher mRNA expression levels for potassium ion channels in aECTs.

Figure 8.

aECT exhibits atrial-like electrophysiology. A through F: Representative optical action potentials (OAP) from atrial-like and ventricular-like ECTs as well as from coronary-perfused right/left ventricle and right/left atrium isolated preparations from non-failing explanted human hearts at 1 Hz pacing (A), accompanied by quantitative analyses (B through F). Human optical mapping OAPs were produced by re-analyzing data collected from Fedorov et al., 2011 (38). n=3–7 ECT per group and n=4 human ventricular/atrial preparations. Significances determined by two-way ANOVA. G: RT-qPCR of KCNJ3, KCNA5, and NR2F2 from atrial-like and ventricular-like ECT relative to GAPDH. n=4–5 human samples and n=12–14 ECTs per group. Significances determined by student’s t-test. H through J: Representative ECT spontaneous action potentials (AP) collected from sharp microelectrode recordings (H) with corresponding quantifications of maximum diastolic potential (MDP, I) and AP amplitude (APA, J). n=4 ECT per group. Significances determined by student’s t-test.

Discussion:

The main goal of this study was to compare heterocellular atrial-like engineered cardiac tissue (aECT) constructs consisting of both hiPSC-CM and hiPSC-CF, to adult human atrial myocardium. Overall, we successfully differentiated atrial-like hiPSC-CMs and demonstrated that their phenotypes are maintained in hiPSC-aCM when combined with hiPSC-CF to generate engineered cardiac tissues. This is supported by molecular data showing higher chamber-specific mRNA and protein expression levels (Figs. 1, 2, and 8) as well as atrial-like electrophysiology (Fig. 8) and contraction dynamics (Figs. 5–7). Furthermore, hiPSC-aCM when incorporated into aECT exhibit improved structural maturity compared to age-matched monolayers and are reminiscent of native healthy adult human atrial myocardium, with a higher tendency to form rod-shaped cardiomyocytes with improved sarcomere alignment (Figs. 3 and 4, and Table 1).

Table 1:

Summary table comparing functional measurements from Adult Human Atria and Atrial ECT throughout literature.

Measurement	Native Atria*	Atrial ECT**
Sarcomere Length (μm)	1.61 ± 0.12 (present study)	1.92 ± 0.11 (present study)
Twitch Force (mN/mm2)	7.07 ± 1.36 (66) and 3 (0.5 Hz) – 7 (2 Hz) (67)	4.23 ± 0.28 (1.5 Hz, present study), 0.2–0.5 (2 Hz) (10), and 0.5–1 (2 Hz) (4)
CT100 (ms)	140.9 ± 6.7 (66) and 91 ± 11 (67)	133.6 ± 4 (present study)
RT50 (ms)	69.4 ± 4.0 (66) and 85 ± 7 (67)	102.6 ± 5.9 (present study), 50–75 (spontaneous) (6)
Max dF/dt (mN/mm2/s)	102.9 ± 19.3 (66)	105.95 ± 1.71 (present study)

^*All contraction measurements from native atria were extracted from adult atrial trabecula paced at 1 Hz, unless otherwise indicated.

^**Contraction measurements from present study were acquired at 1.5 Hz pacing.

aECT were structurally similar to the previously generated ventricular-like ECT constructs by de Lange et al. (8), while also maintaining similar response to β-adrenergic stimulation (Fig. 6). Similar to what has been shown in previously reported atrial-like constructs (6, 16), our aECTs exhibit significantly weaker and faster contractions as well as an atrial-like electrophysiology compared to ventricular-like ECTs (Fig. 5). These functional differences are likely due to differences in mRNA and protein expression between aECT and vECT, similarly reported in previously reported atrial-like hiPSC-CM in monolayer (40) as well as in constructs (7). Indeed, our aECT exhibited higher atrial-like mRNAs and protein such as MYL7, KCNA5, KCNJ3, MYH6, and MYBPHL and low ventricular mRNA and protein such as MYL2, MYH7, and IRX4 while maintaining expression of cardiac mRNAs like Cav3 and TnnT2 (Figs. 2 and 8). We also found that while native adult human cardiac mRNAs are expressed at higher levels than ECT mRNA in general, we found that ratiometrically the differences between human right atrium and human left ventricle follow a similar pattern for aECT versus vECT (Figs. 1, 2, and 8). Contrary to what is observed in adult atrial and ventricular myocardium (41) and some other chamber-specific engineered cardiac tissues (7), we observed no difference in MYL4 or NPPA mRNA between our aECT and vECT. These discrepancies may exist due to differences in differentiation protocol where previous studies utilized an embryonic body protocol (GiWi was used in the present study) and the use of all-trans retinoic acid at Days 3–5 instead of Days 4–6.

Similar to previous work focused on arrhythmogenicity of atrial constructs (11, 12), our aECT demonstrate an atrial-like electrophysiology highlighted by a lower APD20/APD80 ratio, higher repolarization fraction, and shorter various APDs, which is supported by higher potassium channel mRNA levels (Fig. 8). Supporting the electrophysiological relevance of engineered cardiac/heart tissues, we found that there were no significant differences between aECT and RA, as well as vECT and LV, for APD20/APD80 ratio, repolarization fraction, APD90, and APD50 (Figs. 8B-F and Table 2, p>0.05). However, these model systems do not recapitulate human electrophysiology entirely, as indicated by the spontaneous activity of our constructs and the apparent presence of I_f current in our vECTs (Fig. 8H).

Table 2:

Summary table comparing electrophysiological measurements from Adult Human Atria and Atrial ECT throughout literature.

Measurement	Native Atria*	Atrial ECT
Automaticity (Hz)	0.67–1 in AV Node (68)	2.16±0.52 (present study), 2.5–3.5 (6), 0.5–2.5 (4), and 0.5–1 (71)
APD90 (ms)	307.75±21.47 (38) (1 Hz), 234.2±32.3 (69) (2 Hz), 318±42 (70) (1 Hz), 317.36±3.82 (71) (1 Hz)	324.89±12.30 (present study), 221±2.4 (6) (2 Hz), 177±8 (12) (3 Hz)
APD80 (ms)	277.67 ± 26.52 (38) (1 Hz)	277.22±29.76 (present study)
APD50 (ms)	221.00 ± 13.69 (38) (1 Hz), 139±44 (70)(1 Hz), 141.93±3.12 (71) (1 Hz)	200.22±39.19 (present study), 117 ± 4 (12) (3 Hz)
APD20 (ms)	129.00 ± 6.05 (38) (1 Hz), 7±8 (70) (1 Hz), 8.48±1.16 (71) (1 Hz)	125.56±16.07 (present study), 80.4±6.1 (12) (3 Hz)
APD20/APD80	0.47 ± 0.04 (38) (1 Hz)	0.46±0.03 (present study)
Repolarization Fraction	0.28 ± 0.02 (38) (1 Hz)	0.39±0.10 (present study), 0.41±0.005 (6) (2 Hz), 0.39±0.02 (12) (3 Hz)

^*No significant differences were detected between AP measurements from native atria (38) and aECT, as measured by Student’s t-test.

Importantly, in our present study, we uniquely highlight improved structural maturity of aECTs regarding sarcomere alignment (Fig. 4C), caveolae enrichment (Fig. 4E), and cell length to width ratio (cellular anisotropy, Fig. 3D) that were similar to adult human atrial myocardium unlike those parameters measured in ML. These critical ultrastructural features have been found to be relevant to proper cardiac function, with sarcomere alignment demonstrated to be altered by mutations in sarcomeric proteins (42, 43), consistent with familial hypertrophic cardiomyopathy phenotypes in patients (44, 45), and being crucial to mitochondrial function (46). Cardiomyocyte anisotropy has been found to be essential to contraction force generation (47, 48), and affected by chronic cardiac overload (49) and heart failure (50). Our aECT also demonstrated elevated caveolae abundance (Fig. 4E), which are known to have a major role in cardiac physiology (51), being implicated in cytoprotection (28, 52), mechanosensitivity (29, 33) and β₂-adrenergic signaling (53-54). Caveolar scaffolding protein caveolin-3 regulates various ion channels located in caveolae, and mutations in caveolin-3 have been linked to long-QT syndrome (30, 55, 56). Higher caveolae density has been associated with hiPSC-CM maturation (37), and downregulation of caveolae has been found during chronic cardiac overload (58) and linked to cardiomyocyte fragility, cell death and elevated fibrosis (28, 59) as well as increased arrhythmogenesis (30, 56). Overall, these studies highlight the significance of ultrastructural myocyte features in cardiac physiology as well as pathological remodeling and arrhythmogenesis supporting their importance to cardiac modeling in vitro.

Various groups have also examined the effect of fibroblasts on aECT function. Brown et al. found that inclusion of adult human atrial fibroblasts improved various maturity metrics of hiPSC-CM such as sarcomere length, calcium handling, and metabolism (60). Our study saw no difference in sarcomere length between aCM in monolayer vs. in ECT but did reveal a more organized sarcomere organization in aECT compared to monolayers (Fig. 4). It should be noted that our hiPSC-aCM already exhibited sarcomere lengths comparable to native human myocardium which range from 1.6 – 2.2 μM (61). Our previous ECT work found that the inclusion of 10% hiPSC-CFs into these ECTs generates a more consistent 3D scaffold (8) and that these fibroblasts persist until functional tests, as evidenced by approximately 8% of fibrotic tissue as per Mason’s Trichrome staining and the presence of collagen deposition and CF-related mRNAs such as Col3a1, Col1a2, and TGFβ (29). For future work, the inclusion of fibroblasts should be carefully titrated, for example, Nakanishi et al. found that ECT formulated with a starting amount of 30% fibroblasts can induce conduction abnormalities (14).

Based on our data, we propose that our aECT could be utilized to model atrial remodeling via testing various disease-promoting components, such as excessive sympathetic and/or angiotensin II stimulation, mechanical stretch, individual inflammatory factors, among many others. This platform could be further employed for drug testing of both structural and functional remodeling and arrhythmogenesis. In addition, unlike other atrial-like ECT/EHT models, our evaluation of contraction lays the groundwork to explore atrial hypo-contractility that occurs during atrial myopathy and AT/AF (62). With this in mind, we anticipate that by evaluating changes in atrial contraction and fibrosis, our aECT model could be utilized to identify early mediators of atrial disease to develop early-intervention individual or combinatorial drug therapies.

Limitations

We acknowledge various limitations of this work. First, it is unknown which fibroblast population is optimal in aECT generation to promote manifestation of atrial-specific pathophysiology. Second, we acknowledge that our model lacks immune and smooth muscle cells which are known to play a crucial role in mediating pathological development in all chambers of the heart (76, 77) and that our model is also limited by the use of a single male-derived hiPSC line. Furthermore, while higher automaticity is expected in hiPSC-aCM, it is likely that the differentiation process produces nodal-like hiPSC-CM as many protocols to generate nodal-like hiPSC include all-trans retinoic acid during the differentiation process (78). For future atrial disease modeling, it may be useful to identify and utilize an inhibitor of both the nodal and ventricular differentiation processes to produce hiPSC-CM better enriched for atrial-like CM.