Generation and characterization of atrial-like vs. ventricular-like EHTs (RA-EHTs vs

88

89

enrichment of MLC2A⁺ CMs over MLC2V⁺ CMs upon RA-treatment was higher in EHT than in ML (Figure 27 and Figure 28). MLC2A and MLC2V distribution in RA- and Ctrl-EHTs reflects the difference observed in native human atrial and ventricular tissues (Figure 28). These data suggest that the combination of RA-treatment and 3D culture promotes atrial specification of hiPSC-CMs (Lemme et al. 2018).

4.1.2 Contractility of RA-EHTs resembles contractions of human atrial tissue

Previous studies have shown that contractile force per cross-sectional area in human EHTs ranges from 0.1 to 20 mN/mm² and absolute force ranges from 0.08 to 1.5 mN (Tulloch et al.

2011; Jackman et al. 2016) indicating the importance of 3D construct diameter for force development (Weinberger et al. 2017). In thin muscle strips of human heart, force values range from 40 to 80 mN/mm² (Van Der Velden et al. 1998). This discrepancy can be explained by the fact that the cross sectional area of most EHTs is not fully comprised of CMs, but it is also occupied by ECM and non-myocytes. Apart from the lower density of CMs, the lower force developed in EHTs may be also related to the lower sarcomere volume fraction and the general lower level of hiPSC-CM maturation (Zimmermann et al. 2006; Hirt et al. 2014a; Weinberger et al. 2017).

Atrial specific myosin is characterized by higher cross-bridge cycling rate than ventricular myosin. Developed force is inversely dependent on cross-bridge kinetics (Morano et al. 1991).

Therefore, atrial myocardium shows lower Ca²⁺ sensitivity and lower tension generation compared to ventricular myocardium (Morano et al. 1991). In fact, human atrial skinned fibers exhibited two times lower Ca²⁺ sensitivity and force per cross-sectional area than ventricular skinned fibers (Ruf et al. 1998; Piroddi et al. 2007; Ng et al. 2010). In accordance with these data, RA-EHTs developed two times lower force than Ctrl-EHTs (Figure 30B). Differences in Ca²⁺ sensitivity due to different myosin expression in atrial and ventricular myocardium might influence the FFR. Both atrial and ventricular myocardium show a positive FFR, an increase in frequency translates into an increase in force of contraction (Schwinger et al. 1993). However, the degree of the increase in force upon increase in frequency was significantly lower in atrial than in ventricular myocardium (Schwinger et al. 1993). In contrast, Ctrl- and RA-EHTs both showed a similarly positive FFR until 1.8 Hz stimulation and a negative FFR at higher frequencies (Figure 29).

Contraction and relaxation kinetics are faster in human atrial myofibrils compared to ventricular myofibrils, probably due to the faster cross-bridge cycling rate (Piroddi et al. 2007; Ng et al.

90

2010). This difference in contraction kinetics could be related to the fact that atrial light chain 1 has faster cross-bridge kinetics than ventricular light chain 1 (Lowey et al. 1993; Morano et al. 1996; Ng et al. 2010). In fact, isometric contractions of human atrial myocardium showed two time faster shortening velocity compared to ventricular myocardium (Ng et al. 2010;

Molenaar et al. 2013; Berk et al. 2016). Therefore, shorter TTP and RT in human atrial myocardium compared to ventricular myocardium was in line with 47% shorter TTP-50% and 35% shorter RT50% in RA-EHTs compared to Ctrl-EHTs (Figure 30). Ctrl- and RA-MLs showed qualitatively similar differences in contraction kinetics, but these differences were smaller in ML than in EHTs (Figure 30). Faster contraction kinetics of RA-EHTs compared to Ctrl-EHTs might be due to different expression of myosin light chain isoforms (Figure 28) or different levels of sarcomere maturation (Figure 26).

It was reported that not only contraction and relaxation processes are faster in atrial than ventricular tissue, but also calcium transients are shorter in human atrial compared to ventricular trabeculae (Maier et al. 2000). Faster calcium transients in the atrial tissue might be associated with higher calcium uptake from the SR and faster calcium removal from the cytosol. These observations are consistent with faster calcium transients in RA-EHTs compared to Ctrl-EHTs (Figure 31). Moreover, a recent study demonstrated that RA-treatment of hiPSC-CMs increases the rate of Ca²⁺ uptake and release, thus generating atrial-like myocytes characterized by Ca²⁺

handling comparable to adult atrial myocardium (Argenziano et al. 2018). This finding is particularly relevant to model AF linked to calcium channel mutations.

4.1.3 RA-EHTs display atrial-like AP

RA-EHTs showed a faster spontaneous beating rate than Ctrl-EHTs (Figure 30). It is difficult to compare this finding with human adult cardiac muscle, because, in contrast to hiPSC-CMs, ventricular and atrial tissues do not show automaticity. The difference in basal rate might be explained with differential ion channel expression between Ctrl- and RA-EHTs. RA-treatment affected EHT electrophysiology. In fact, RA-EHTs showed a less negative TOP than Ctrl-EHTs (-70±1.1 mV vs. -76±1.5 mV, Figure S6), reflecting a similar difference in RMP between RAA and LV (-74.0±0.5 mV vs. -78.5±1.0 mV; Burashnikov et al. 2008). The less negative TOP in RA-EHTs translates into a lower upstroke velocity compared to Ctrl-EHTs (97.6±2.4 V/s vs.

207.6±10.6 V/s, Figure S6). TOP is in the steep phase of the steady-state inactivation curve of sodium channels, where small changes in TOP determine large difference of sodium channel availability, which in turn affects upstroke velocity (Skibsbye et al. 2016; Lemoine et al. 2017).

91

As previously shown, EHTs casted from hiPSC-CMs showed shorter APD90 than human heart tissue (Horváth et al. 2018). RA-treatment induced an additional shortening of APD90 (Figure 32). The difference in APD was larger between RA- and Ctrl-EHTs than between RAA and LV. A shorter APD90 determines a shorter ERP, which in turn could facilitate induction of tachyarrhythmias. This result may represent an advantage to test new antiarrhythmic drugs.

As previously demonstrated (Du et al. 2015; Horváth et al. 2018), repolarization fraction could discriminate between atrial and ventricular APs. As observed in RAA and LV (Horváth et al.

2018), the repolarization fraction of Ctrl- and RA-EHTs did not overlap (Figure 32 and Figure S4), suggesting a marked effect of RA-treatment on the repolarization phase of APs. However, the AP shape of RA-EHTs did not fully resemble AP from atrial tissue, due to the lack of the steep initial repolarization phase resulting in the spike and dome phenomenon typical for human RAA. In fact, RA-EHTs displayed a triangular AP similar to patients in persistent AF. The different repolarization phase could underlie a low contribution of Ito and IKur in RA-EHTs.

4.1.4 RA-EHTs show responsiveness to atrial-selective drugs

KCNA5 and KCNJ3 encode for the ion channels Kv1.5 and Kir3.1, respectively. These ion channels conduct the potassium currents IKur and IK,ACh, which represent the major electrophysiological discriminators between atrial and ventricular CMs (Ravens et al. 2013).

CCh was used to identify IK,ACh in EHTs. Activation of muscarinic receptors by CCh decreased spontaneous beating rate of both RA- and Ctrl-EHTs (Figure 35). This finding cannot be taken as proof of IK,ACh activation, because activation of muscarinic receptors also decreases If (Dario DiFrancesco, Pierre Ducouret 1989) and If plays an important role in spontaneous pacemaking in EHTs (Mannhardt et al. 2016; Lemoine et al. 2018). On the other hand, APD90 shortening upon CCh exposure in RA-EHT can be considered a proof of IK,Ach activation. CCh did not affect APD90 in Ctrl-EHTs, in accordance with the previously reported absence of IK,ACh in cells isolated from Ctrl-EHTs (Horváth et al. 2018). The CCh effect on APD90 is known to show a reverse rate-dependency, meaning that APD90 shortening decreases at higher frequency (Figure 33). According with this observation, APD90 shortening induced by CCh was smaller in C25 which beat at 3 Hz (7%) than in ERC18 which beat at 2 Hz (20%, Figure S5). Nevertheless, CCh induced APD90 shortening was smaller in RA-EHTs than in RAA where the effect reached 50% (Figure 34), probably due to the lower amplitude of IK,Ach. CCh has a transient negative inotropic effect on human atrial trabeculae (Figure 36), but it does not affect contraction force

92

of ventricular heart muscle (Jakob et al. 1989). On the other hand, CCh (10 µmol/L) slightly decreased contraction force in both RA- and Ctrl-EHTs (Figure 35).

Ikur is another atrial-selective current involved in the repolarization phase of AP, which can be useful to discriminate between atrial and ventricular CMs. It was previously demonstrated that the Ikur blocker Xention D-0101 was effective in RA-treated, but not in control hESC-CMs (Ford et al. 2013; Devalla et al. 2015). Due to the unavailability of this compound, a low concentration of 4-AP (50 µmol/L) was tested to block Ikur in Ctrl-and RA-EHTs. A high concentration of 4-AP (1 mmol/L) did not have any effect on LV (Figure 38). This finding was unexpected because it is known that high concentrations of 4-AP block Ito. The absence of Ito

blockade could be explained by the fact that these experiments were performed with subendocardial preparations from patients suffering from HF characterized by low Ito

amplitudes (Wettwer et al. 1994). In contrast, a low concentration of 4-AP (50 µmol/L) had the same effect as D-0101 in RAA (Wettwer et al. 2004; Ford et al. 2013). Specifically, 4-AP (50 µmol/L) prolonged APD20 and shortened APD90 in RAA (Figure 38 and Figure 37). This apparently contradictory finding can be explained by the fact that APD20 prolongation leads to a more positive plateau voltage, which induces a stronger activation of IKr and thereby APD90

shortening (Wettwer et al. 2004). In RA-EHTs, 4-AP (50 µmol/L) prolonged APD20, but it did not decrease APD90 (Figure 38 and Figure 37), probably due to the lack of a clear plateau phase (Lemme et al. 2018).

At the Institute of Experimental Pharmacology and Toxicology, UKE, Hamburg (UKE, Hamburg) RA-EHTs were established in parallel to rat atrial EHTs (Krause et al. 2018). As RA-EHTs, rat atrial EHTs exhibited faster spontaneous beating, smaller contraction force, faster contraction kinetics and shorter APD90 compared to rat ventricular EHTs (Krause et al.

2018). The decrease in force observed in the FFR from rat atrial EHTs started at higher frequency than in the FFR from rat ventricular EHTs, whereas FFR did not differ between Ctrl- and RA-EHTs (Figure 29). CCh showed a negative inotropic effect only in rat atrial EHTs, while CCh induced a force reduction, even though not significant, in both RA- and Ctrl-EHTs (Figure 35). Atrial myocyte purity is higher in rat atrial EHTs obtained from cells directly isolated from the rat atrium than in RA-EHTs generated from hiPSC differentiated into atrial-like myocytes without any purification step. On the other hand, the clinical relevance of RA-EHTs is greater than rat atrial RA-EHTs, since rat physiology is different from human.

In conclusion, the results of the first part of the project overcome limitations related to cardiac chamber specification. In fact, RA-treatment could successfully direct cardiac differentiation

93

of hiPSC towards an atrial phenotype. 3D culture further contributed to atrial specification of hiPSC-CMs. RA-EHTs exhibited distinctive features of human atrial tissue in terms of atrial-specific gene expression, protein levels, contraction force and kinetics, AP parameters and response to atrial-selective drugs. Although quantitative differences compared to RAA persist, RA-EHTs can be considered an experimental model of human atrium to be used in disease modeling and preclinical drug development (Lemme et al. 2018). Moreover, highly enriched populations of atrial or ventricular CMs are useful for cardiac repair, because they could reduce the risk of arrhythmias upon transplantation of mixed population of hPSC-CMs (Shiba et al.

2016).