Correlation between frataxin expression and contractility revealed by in vitro Friedreich's ataxia cardiac tissue models engineered from human pluripotent stem cells

Abstract

Background: Friedreich's ataxia (FRDA) is an autosomal recessive disease caused by a non-coding mutation in the first intron of the frataxin (FXN) gene that suppresses its expression. Compensatory hypertrophic cardiomyopathy, dilated cardiomyopathy, and conduction system abnormalities in FRDA lead to cardiomyocyte (CM) death and fibrosis, consequently resulting in heart failure and arrhythmias. Murine models have been developed to study disease pathology in the past two decades; however, differences between human and mouse physiology and metabolism have limited the relevance of animal studies in cardiac disease conditions. To bridge this gap, we aimed to generate species-specific, functional in vitro experimental models of FRDA using 2-dimensional (2D) and 3-dimensional (3D) engineered cardiac tissues from FXN-deficient human pluripotent stem cell-derived ventricular cardiomyocytes (hPSC-hvCMs) and to compare their contractile and electrophysiological properties with healthy tissue constructs.

Methods: Healthy control and FRDA patient-specific hPSC-hvCMs were derived by directed differentiation using a small molecule-based protocol reported previously. We engineered the hvCMs into our established human ventricular cardiac tissue strip (hvCTS) and human ventricular cardiac anisotropic sheet (hvCAS) models, and functional assays were performed on days 7-17 post-tissue fabrication to assess the electrophysiology and contractility of FRDA patient-derived and FXN-knockdown engineered tissues, in comparison with healthy controls. To further validate the disease model, forced expression of FXN was induced in FXN-deficient tissues to test if disease phenotypes could be rescued.

Results: Here, we report for the first time the generation of human engineered tissue models of FRDA cardiomyopathy from hPSCs: FXN-deficient hvCTS displayed attenuated developed forces (by 70-80%) compared to healthy controls. High-resolution optical mapping of hvCAS with reduced FXN expression also revealed electrophysiological defects consistent with clinical observations, including action potential duration prolongation and maximum capture frequency reduction. Interestingly, a clear positive correlation between FXN expression and contractility was observed (ρ > 0.9), and restoration of FXN protein levels by lentiviral transduction rescued contractility defects in FXN-deficient hvCTS.

Conclusions: We conclude that human-based in vitro cardiac tissue models of FRDA provide a translational, disease-relevant biomimetic platform for the evaluation of novel therapeutics and to provide insight into FRDA disease progression.

Fig. 1

In vitro modeling of FRDA by engineered cardiac tissue constructs of FXN-deficient hPSCs. a Experimental timeline for generating cardiac tissue models, human ventricular cardiac anisotropic sheet (hvCAS), and human ventricular cardiac tissue strip (hvCTS), for electrophysiological and contractile assessment, respectively. b Representative images of (i) hvCAS and (ii) hvCTS. c FXN (i) transcript and (ii) protein expression and (iii) representative Western blot image of hESCs (n = 8–9), hiPSCs (n = 5–9), and FRDA-hiPSC lines 68 (n = 8–10) and 03665 (n = 3–4), normalized to GAPDH expression. Samples were independent replicates. Data are shown as mean ± SEM. Statistical significance indicated by *p < 0.05

Fig. 2

Isogenic FRDA cardiac model derived from hESCs. a FXN (i) transcript and (ii) protein expression (normalized to GAPDH expression) and (iii) representative Western blot image of hESC-hvCMs transduced with Lv-shFXN1 and Lv-shFXN2, relative to batch-matched control transduced with Lv-shNT (n = 3–6). Transcript and protein data were generated from 3 to 6 and 3 to 5 independent differentiation batches, respectively. Data are shown as mean ± SEM. b Representative contractile force traces on day 12 for hESC-hvCTS transduced with Lv-shFXN1 and Lv-shFXN2, with Lv-shNT as control. c Developed force generation on (i) day 7, (ii) day 9, and (iii) day 12 from 1 Hz-paced hESC-hvCTS transduced with Lv-shFXN1 (n = 12) and Lv-shFXN2 (n = 8), compared to control transduced with Lv-shNT (n = 17). hESC-hvCTS force generation data originated from 3 to 6 independent batches of differentiation. d Kinetics analysis of (i) contractile rate and (ii) relaxation rate of hESC-hvCTS force generation on day 12. All force generation data are shown as median with interquartile range. Statistical significance indicated by *p < 0.05 and **p < 0.01

Fig. 3

FRDA cardiac model derived from FRDA-hiPSCs. a FXN (i) transcript and (ii) protein expression (normalized to GAPDH expression) and (iii) representative Western blot image of FRDA(68)-hiPSC-hvCMs and FRDA(03665)-hiPSC-hvCMs relative to healthy control hiPSC-hvCMs (n = 3–5). Transcript and protein data were generated by 3–4 and 3–5 independent differentiations, respectively. b Correlation of developed force to FXN transcript expression in hESC- and hiPSC-hvCTS transduced with Lv-shNT, Lv-shFXN1, Lv-shFXN2, and FRDA-hiPSC-hvCTS relative to the hESC-shNT group (on day 12, at 1 Hz pacing). Force generation and FXN transcript expression data were generated from 3 to 8 and 3 to 6 independent cell differentiation batches, respectively. Data are shown as mean ± SEM. Statistical significance indicated by *p < 0.05 and **p < 0.01

Fig. 4

Electrophysiological measurements from FXN-deficient hvCAS derived from hESCs and hiPSCs relative to their respective healthy controls. a Representative action potentials and isochronal maps from control hESC-hvCAS transduced with non-targeting lentivirus (Lv-shNT), hESC-hvCAS transduced with Lv-shFXN, control hiPSC-hvCAS, and FRDA-hiPSC-hvCAS. b Maximum capture frequency (MCF) for (i) Lv-shNT (n = 32) vs. Lv-shFXN (n = 33) hESC-hvCAS and (ii) healthy control hiPSC-hvCAS (n = 20) vs. FRDA-hiPSC-hvCAS (n = 30). c Action potential duration at 50% repolarization (APD50) and 90% repolarization (APD90) derived from optical mapping of (i and iii) Lv-shNT vs. Lv-shFXN-transduced hESC-hvCAS and (ii and iv) healthy control vs. FRDA-hiPSC-hvCAS. Data were generated from 3 to 9 independent batches of differentiation. All electrophysiology data are shown as median with interquartile range. Statistical significance indicated by *p < 0.05, **p < 0.01, and ***p < 0.001

Fig. 5

FXN expression rescue of FRDA hiPSC-hvCTS model. a FXN (i) transcript and (ii) protein expression (normalized to GAPDH expression) and (iii) representative Western blot image of FRDA-hiPSC-hvCTS after Lv-FXN transduction relative to Lv-RFP control (n = 3–4). Transcript and protein data were generated from 3 to 4 independent differentiation batches, respectively. b Representative contractile force traces on days 17–18 for FRDA-hiPSC-hvCTS transduced with Lv-FXN with Lv-RFP as control. c (i) Developed force, (ii) contractile rate, and (iii) relaxation rate generated at 1 Hz pacing in FRDA-hiPSC-hvCTS transduced with Lv-FXN (n = 10) relative to Lv-RFP control (n = 8). Data were generated by 4 independent batches of cell differentiation. FXN expression data are shown as mean ± SEM, and force generation data are shown as median with interquartile range. Statistical significance indicated by *p < 0.05 and **p < 0.01

Fig. 6

FXN expression rescue of FRDA hESC-hvCTS model. a FXN (i) transcript and (ii) protein expression (normalized to GAPDH expression) and (iii) representative Western blot image of hESC-hvCTS double-transduced with Lv-shFXN and Lv-FXN relative to Lv-shFXN and Lv-RFP control (n = 5–6). Transcript and protein data were generated from 5 to 6 independent batches of cell differentiation, respectively. b Representative contractile force traces from isometric force measurement relative to % L_max for hESC-hvCTS double-transduced with Lv-shFXN and Lv-FXN, relative to Lv-shFXN and Lv-RFP control. c Developed force, contractile rate, and relaxation rate for % L_max measured at 1 Hz stimulation using isometric force measurement for hESC-hvCTS double-transduced with Lv-shFXN and Lv-FXN, relative to Lv-shFXN and Lv-RFP control (Lv-shFXN+Lv-FXN: n = 16; Lv-shFXN+Lv-RFP: n = 11). Data were generated by 4 independent differentiation batches. Data are shown as mean ± SEM. Statistical significance indicated by *p < 0.05 and **p < 0.01