In vitro model of ischemic heart failure using human

In vitro model of ischemic heart failure using human inducedpluripotent stem cell-derived cardiomyocytes

Justin Davis hellip Katherine F Campbell Todd J Herron

JCI Insight 2021 httpsdoiorg101172jciinsight134368

In-Press Preview

Graphical abstract

Technical Advance Cardiology Cell biology

Find the latest version

httpsjcime134368pdf

Title In vitro model of ischemic heart failure using human induced pluripotent stem cell-derived 1

cardiomyocytes 2

Justin Davis1 Ahmad Chouman1 Jeffery Creech1 Andre Monteiro da Rocha1 2 4 Daniela Ponce 3

Balbuena2 Eric Jimenez Vazquez2 Ruthann Nichols3 Andrey Lozhkin4 Nageswara Madamanchi4 4

Katherine Campbell12 Todd J Herron1245 5

1University of Michigan Frankel Cardiovascular Regeneration Core Laboratory Internal Medicine-6

Cardiovascular Medicine Ann Arbor MI 48109 7

2University of Michigan Center for Arrhythmia Research Ann Arbor MI 48109 8

3University of Michigan Department of Biological Chemistry Medical Science Research Building III 9

1150 W Medical Center Drive Ann Arbor MI 48109 10

4University of Michigan Internal Medicine-Cardiology MSRBIII Ann Arbor MI 48109 11

5University of Michigan Department of Molecular and Integrative Physiology Ann Arbor MI 48109 12

Conflict of Interest Statement Jeffery Creech and Andre Monteiro da Rocha are consultants to 13

CARTOX Inc Todd J Herron is co-founder of and has financial interest in CARTOX Inc CARTOX 14

Inc is a Michigan company focused on the development of novel hiPSC-CM based assays for 15

cardiotoxicity screening TJH is also a consultant to StemBioSys Inc 16

Corresponding Author 17

Todd J Herron PhD 18 University of Michigan 19

Frankel Cardiovascular Regeneration Core Laboratory 20 2800 Plymouth Road 21 Building 26 223N 22 Ann Arbor MI 48109-2800 23 toddherrumichedu 24

Abstract 26

Human induced pluripotent stem cell derived cardiomyocytes (hiPSC-CMs) have been used 1

extensively to model inherited heart diseases but hiPSC-CM models of ischemic heart disease are 2

lacking Here our objective was to generate an hiPSC-CM model of ischemic heart disease To this 3

end hiPSCs were differentiated to functional hiPSC-CMs and then purified using either a simulated 4

ischemia media or by using magnetic antibody based purification targeting the non-myocyte 5

population for depletion from the cell population Flow cytometry analysis confirmed that each 6

purification approach generated hiPSC-CM cultures of gt94 cTnT+ cells Following purification 7

hiPSC-CMs were re-plated as confluent syncytial monolayers for electrophysiological phenotype 8

analysis and protein expression by Western blotting Metabolic selected hiPSC-CM monolayersrsquo 9

phenotype recapitulated many of the functional and structural hallmarks of ischemic cardiomyocytes 10

including elevated diastolic calcium diminished calcium transient amplitude prolonged action 11

potential duration depolarized resting membrane potential hypersensitivity to chemotherapy induced 12

cardiotoxicity depolarized mitochondrial membrane potential depressed SERCA2a expression 13

reduced maximal oxygen consumption rate and abnormal response to β1-adrenergic receptor 14

stimulation These findings indicate that metabolic selection of hiPSC-CMs generates cell populations 15

with phenotype similar to what is well known to occur in the setting of ischemic heart failure and thus 16

provides a novel opportunity for study of human ischemic heart disease 17

Introduction 24

The use of human induced pluripotent stem cell derived cardiomyocytes (hiPSC-CMs) makes it 1

possible now to generate a virtually unlimited supply of human cardiomyocytes that can be used for 2

research purposes(1 2) hiPSC-CMs have been utilized to model many monogenic cardiac diseases 3

such as long QT syndrome(3 4) hypertrophic cardiomyopathy(5) catecholaminergic polymorphic 4

ventricular tachycardia (CPVT)(6) and Timothy syndrome(7) On the other hand there are few reports 5

on the generation of hiPSC-CM models of acquired heart disease such as ischemic heart disease 6

and failure Ischemic heart failure is a significant health problem that affects over 5 million patients in 7

the US and currently the only cure is cardiac transplantation The development of new and effective 8

therapies to treat ischemic heart failure will be hastened by the availability of an in vitro hiPSC-CM 9

model of ischemic cardiomyopathy 10

Importantly the in vitro hiPSC-CM ischemic heart failure model should recapitulate the 11

molecular and functional hallmarks that are well known to occur in patients and research models of 12

heart failure For example contractile function of the individual cardiomyocytes is depressed in 13

human heart failure and this has been confirmed in numerous animal models of ischemic heart 14

failure(8) Molecular and genetic modification of key contractile and calcium handling proteins 15

underlie the poor contractile function of the failing heart and cardiomyocytes Specifically heart failure 16

is associated with reduced SERCA2a gene expression and reduced phospholamban (PLN) 17

phosphorylation(9 10) These proteins directly regulate the amount of calcium available to trigger 18

contraction and therefore represent molecular targets for heart failure therapy SERCA2a gene 19

overexpression in the heart rapidly hastens the removal of cytosolic calcium during diastole thus 20

alleviating diastolic dysfunction In fact SERCA2a gene therapy was shown in vitro to improve the 21

contractile function of human cardiomyocytes isolated from patients afflicted with heart failure and is 22

being tested as a new therapy for heart failure patients(11-13) Thus an in vitro model of human 23

ischemic heart failure should present with reduced contraction reduced calcium flux amplitudes 24

reduced SERCA2a protein expression prolonged calcium transient duration prolonged action 25

potential duration and reduced PLN phosphorylation Additionally the functional phenotype should be 1

rescued using established SERCA2a gene therapy approaches known to improve cardiomyocyte 2

contractile function Further the ischemic heart failure in vitro model should recapitulate elevated 3

sensitivity to cardiotoxic therapies such as doxorubicin chemotherapy Our goal here was to 4

demonstrate these features of human heart failure phenotypes using hiPSC-CMs in vitro 5

Currently the most widely utilized approach for purification of hiPSC-CMs relies on marked 6

metabolic differences in glucose and lactate metabolism between cardiomyocytes and non-7

cardiomyocytes(14) Tohyama et al(14) cultured pluripotent stem cell derivative cells with glucose-8

depleted culture medium supplemented with abundant lactate and found that only cardiomyocytes 9

survived This method of hiPSC-CM purification has been adopted by many laboratories and is a 10

highly efficient method for large scale purification(15 16) However this same media composition has 11

been utilized historically to simulate ischemia pre conditioning in cultured adult cardiomyocytes(17) 12

Also similar glucose free solution has been referred to as ldquoinjury solutionrdquo before to produce localized 13

ischemic injury to cardiac monolayers(18 19) Here we hypothesized that this simulated ischemia 14

media (0 glucose abundant lactate) generates purified hiPSC-CMs with an ischemic heart failure-like 15

phenotype To test this hypothesis we compared the phenotype of metabolic stress purified hiPSC-16

CMs with the phenotype of hiPSC-CMs purified using a commercially available magnetic activated 17

cell sorting approach (MACS Miltenyi Biotec) The MACS approach and subsequent enrichment of 18

the hiPSC-CMs was recently described and validated for use in cardiotoxicity proarrhythmia 19

assays(20 21) 20

Figure 1A outlines the general cardiac directed differentiation and purification approaches used 21

here All functional and structural phenotype analysis was performed on syncytial hiPSC-CM 22

monolayers The electrophysiological dysfunction (calcium transient alternans red traces of figure 23

1C) apparent during metabolic selection is depicted in figure 1C and provided us rationale for further 24

study 25

Results 2

Human induced pluripotent stem cell-derived cardiomyocytes (hiPSC-CMs) were enriched for 3

preparation of syncytial monolayers with a metabolic selection protocol or with magnetic activated cell 4

sorting (MACS) targeting non-cardiomyocytes for immunodepletion (Figure 1A) Two separate 5

established hiPSC lines were used here 1) 19-9-11 hiPSC line (WiCell iPS DF19-9-11TH)(22) and 6

2) BJ-hiPSC line (generated in house using mRNA reprogramming)(23 24) 7

Cardiomyocyte enrichment was successful with both methods across the cell lines used and the two 8

methods yielded similar percentages of cardiomyocytes (figure 2A metabolic selection=9703plusmn29 9

purity n=4 and MACS selection=952plusmn30 cTnT+ cells n=4)Cardiomyocytes selected with each of 10

the two different methods formed functional syncytial cardiac monolayers with cell-cell propagation of 11

electrical impulses as described before(21 23-26) nevertheless there were several phenotypic 12

differences between cardiomyocytes obtained by the two methods such as heterogeneous 13

expression and mislocalization of connexin 43 (Figure 2B) decrease in sarcomere length and 14

organization (Figure 2C) and the appearance of actin stress fibers in metabolic stress selected 15

hiPSC-CMs (Figure 2D) Additionally functional phenotyping of cardiomyocytes indicated changes in 16

important functional characteristics of cardiomyocytes like intracellular calcium handling and 17

electrophysiology 18

Metabolic selection alters intracellular calcium handling and blunts β-adrenergic response of 19

cardiomyocytes 20

Figure 1C shows that metabolic selection induces calcium transient alternans in hiPSC-CM 21

monolayers prior to re-plating whereas calcium transient alternans were not observed in non-22

selected hiPSC-CM monolayers Next hiPSC-CM monolayers selected with the two purification 23

methods were loaded with calcium sensitive dye (Fura2AM) for quantification of intracellular calcium 24

transients and displayed divergent functional phenotypes The frequency of spontaneous intracellular 1

calcium activations was higher in syncytial monolayers of cardiomyocytes prepared after metabolic 2

selection (052plusmn007Hz n=9 vs MACS=027plusmn005Hz n=7 significant difference-t-test Plt0001 3

figure 3A) in comparison to those prepared after MACS sorting Furthermore metabolic selected 4

hiPSC-CM monolayers had higher diastolic cytosolic calcium concentrations in comparison to MACS 5

enriched cardiomyocytes (004plusmn0008 fura ratio units n=15 vs MACS=001plusmn0004 n=13 t test 6

Plt0001 figure 3A) Calcium transient amplitude was significantly greater in the MACS enriched 7

hiPSC-CM monolayers compared to metabolic selected monolayers (MACS=014plusmn005fura ratio 8

units n=13 vs metabolic selection=009plusmn003fura ratio units n=15 t test P=0014 figure 3A) 9

Next we sought to determine the molecular mechanism to account for the differences in 10

calcium flux observed between metabolic and MACS based purification approaches Western blotting 11

was employed to determine the relative expression levels of SERCA2a-the primary calcium pump of 12

cardiomyocytes responsible for cytosolic calcium sequestration into the sarcoplasmic reticulum 13

Impaired intracellular calcium handling in metabolic selected cardiomyocytes was mechanistically 14

associated with significantly lower expression of SERCA2a protein (094plusmn021au n=6 vs 15

231plusmn056au n=6 significant difference t-test P=0002 Figure 3B) in comparison to MACS selected 16

cardiomyocytes Furthermore in figure 3C imaging intracellular calcium flux (with fura-2) after 17

isoproterenol treatment demonstrated that MACS selected cardiomyocytes present expected 18

chronotropic and inotropic responses to β-adrenergic stimulation as both spontaneous frequency and 19

calcium transient amplitude significantly increased after isoproterenol treatment (from 027plusmn005Hz to 20

05plusmn01Hz n=3 and amplitude increased from 009plusmn006 Fura2 ratio units to 015plusmn010 fura ratio 21

units n=3) On the other hand in figure 3C β-adrenergic stimulation of cardiomyocytes purified with 22

metabolic selection did not induce an inotropic effect as spontaneous frequency significantly 23

increased (from 05plusmn008 up to 10plusmn04Hz n=3 red traces) without concomitant increase in 24

intracellular calcium amplitude (pre isoproterenol=008plusmn005 Fura2 ratio units and post isoproterenol 1

amplitude=006plusmn006 Fura2 ratio units n=3) 2

To further explore molecular mechanisms at play we examined the phosphorylation state of 3

two key cardiomyocyte proteins following isoproterenol (500nmolL-1) stimulation of β1 adrenergic 4

receptors namely cTnI and phospholamban Cardiomyocytes harvested from syncytial monolayers 5

prepared after metabolic selection had higher base-line cTnI phosphorylation in comparison to MACS 6

sorted cardiomyocytes but presented lower cTnI phosphorylation after isoproterenol treatment (figure 7

3D) Investigation of phospholamban phosphorylation levels by Western blotting provided 8

complementary mechanistic insight on metabolic selected cardiomyocytes intracellular calcium 9

dysregulation Untreated metabolic selected hiPSC-CM monolayers presented lower levels of 10

phosphorylated phospholamban in comparison to MACS selected cardiomyocytes (figure 3E) After 11

isoproterenol treatment MACS selected cardiomyocytes had a ~15-fold increase in phosphorylated 12

phospholamban whilst metabolic selected cardiomyocytes did not have any detectable increase in 13

phosphorylated phospholamban 14

Metabolic stress selection impacts hiPSC-CM electrophysiology 15

Following hiPSC-CM purification using MACS or metabolic stress selection cells were re-16

plated as electrically and mechanically confluent monolayers for electrophysiological analysis as 17

described before(20 21 27) First electrophysiology of hiPSC-CM monolayers was investigated using 18

sharp microelectrode recordings or optical mapping with a voltage sensitive dye Figure 4A shows 19

original action potential recordings and quantification of Maximal Diastolic Potential (MDP) and Action 20

Potential Amplitude (APA) MDP was less negative (depolarized) in metabolic selected monolayers 21

compared to MACS purified hiPSC-CM monolayers (MACS= -714plusmn484mV n=11 vs metabolic 22

stress hiPSC-CMs=-599plusmn759mV n=9 denotes significance unpaired t-test Plt0001) This 23

baseline membrane potential difference translated to greater frequency of spontaneous activations in 24

metabolic stress selected monolayers (MACS=041plusmn016Hz n=24 monolayers vs metabolic 25

selection=097plusmn029Hz n=24 significant difference unpaired t-test Plt0001) and lower action 1

potential amplitude (Figure 4B) To precisely investigate action potential duration hiPSC-CM syncytial 2

monolayers were paced at different frequencies during optical mapping with voltage sensitive dye 3

(Figure 4C) Action potential duration at 80 of repolarization (APD80) was significantly longer in 4

metabolic stress selected hiPSC-CM monolayers in comparison to MACS selected monolayers at 10 5

15 20 and 25Hz pacing frequencies (Figure 4C) Finally metabolic stress selected and MACS 6

selected hiPSC-CM syncytial monolayers were submitted to an arrhythmia test consisting of abrupt 7

decrease in pacing from 25Hz to 10Hz while recording optical action potentials All syncytia 8

prepared with MACS selected cardiomyocytes were able to follow the decrease in pacing frequency 9

with 11 capture however seventy-two percent of metabolic stress selected hiPSC-CM monolayers 10

displayed early afterdepolarizations (Figure 4D) 11

Effect of hiPSC-CM purification approach on mitochondrial membrane potential (Δψm) 12

Mitochondria have been implicated in several human diseases including cardiac disease and 13

are sensitive to energy substrates(28) Next we determined the impact of hiPSC-CM purification 14

approach on mitochondrial membrane potential (Δψm) using the carbocyanine compound JC-1 15

Figure 5A shows confocal images of hiPSC-CM monolayers made from metabolic stress selected or 16

MACS selected CMs loaded with JC-1 It is apparent in this image that there is a greater amount of 17

red staining in the MACS sorted cardiomyocytes relative to metabolic stress selected hiPSC-CMs 18

Indeed quantification of the ratio of redgreen fluorescence shows that metabolic selected hiPSC-19

CMs have a lower ratio and thus have depolarized Δψm (057plusmn006 au n=5 vs MACS=065plusmn004 20

n=4 significant difference-unpaired t-test Plt0001) Δψm depolarization is indicated by the reduction 21

of the red to green fluorescence intensity ratio and this difference persisted even following treatment 22

with the mitochondrial uncoupler FCCP (metabolic selected hiPSC-CM=021plusmn001 au n=5 vs 23

MACS=034plusmn006 n=4) Further complementary analysis of mitochondrial membrane potential was 24

done using MitoTracker Red CMXRos dye (Figure 5B) MitoTracker Red CMXRos dye accumulates 25

in the mitochondria and its accumulation is dependent on mitochondrial membrane potential(29) 1

Figure 5B shows greater MitoTracker Red dye accumulation in MACS purified hiPSC-CMs indicating 2

more hyperpolarized active mitochondria compared to CMs purified using metabolic stress selection 3

Metabolic selection increases susceptibility of cardiomyocytes to doxorubicin induced 4

cardiotoxicity (DOX-TOX) 5

Doxorubicin (Adriamycin) is a chemotherapy drug with well-documented cardiotoxicity referred 6

to as DOX-TOX DOX-TOX has been linked to direct effects on mitochondrial function and 7

intracellular calcium flux We hypothesized that metabolic selected hiPSC-CMs are more sensitive to 8

DOX-TOX owing to the baseline differences in mitochondrial membrane potential electrophysiology 9

and intracellular calcium flux identified in figures 3-5 To test this metabolic stress and MACS 10

selected cardiomyocyte monolayers were challenged with doxorubicin concentrations ranging from 11

100 to 1000nM First cell survival was assessed in cardiomyocyte monolayers using annexin V 12

staining (figure 6A) AnnexinV live cell staining demonstrated similar DOX-TOX of metabolic stress 13

selected and MACS selected cardiomyocytes to doses equal to or below 750nM of doxorubicin 14

However the number of cells undergoing apoptosis after exposure to 1000nM of doxorubicin was 15

~3X higher in metabolic stress selected cardiomyocyte syncytial monolayers compared to MACS 16

selected monolayers (1694plusmn751 AnnexinV+cellsmm2 n=4 vs 518plusmn51 AnnexinV+cellsmm2 n=4 17

meanplusmnsem t-test Plt0001 figure 6A) 18

Optical mapping with calcium sensitive dye (rhod2AM figure 6BampC) was performed after 19

chronic exposure (48h) to doxorubicin and percentage of monolayers with arrhythmias was calculated 20

for each group Arrhythmia phenotypes included early after depolarization extra calcium release and 21

tachycardia Doxorubicin concentration of 250nM or below did not induce arrhythmias in monolayers 22

from either group 500nM or greater concentrations of doxorubicin induced arrhythmias and 23

significantly higher arrhythmia burden in metabolic stress selected cardiomyocytes syncytia than in 24

MACS selected syncytial monolayers (Figure 6BampC) Arrhythmia burden reached 100 in metabolic 25

stress selected hiPSC-CM syncytial monolayers at 750nM of doxorubicin treatment with absolute 1

predominance of EADs (Figure 6BampC) and severe tachyarrhythmia at 1000nM of doxorubicin 2

SERCA2a over expression corrected calcium transient duration prolongation associated with 3

heart failure 4

Finally we tested the effectiveness of SERCA2a gene therapy to improve the function of the 5

metabolic stress selected hiPSC-CMs Recombinant adenovirus (Vector Biolabs) designed to deliver 6

SERCA2a gene and mCherry promoter was applied to MACS sorted and metabolic stress treated 7

monolayers (100 moi multiplicity of infection) mCherry expression indicated robust expression 8

throughout the monolayers (figure 7A) and Western Blotting confirmed elevated SERCA2a 9

expression levels (figure 7A) Functional analysis measuring intracellular calcium flux (fluo-4) 10

indicated spontaneous arrhythmias in the metabolic stress media treated cells and prolonged calcium 11

transient duration 50 (CaTD50) The quantification in Figure 7C indicates correction of the CaTD50 12

with SERCA2a gene therapy to control levels 13

Metabolic stress media affects the electrophysiological phenotypes of MACS Purified hiPSC-14

CMs amp Antibiotic Resistance Purified hiPSC-CMs 15

Next we tested the effect of the CDML3 metabolic selection media on hiPSC-CMs that were 16

purified using the MACS approach This enabled testing the effect of the metabolic stress condition of 17

0 glucose and abundant lactate on hiPSC-CM function without interference of non-cardiomyocytes 18

death Here we used the same hiPSC line (19-9-11) differentiated to cardiomyocytes as outlined in 19

figure 1 and purified hiPSC-CMs using MACS selection hiPSC-CMs were re-plated into two separate 20

96 well plates 1 plate was cultured in maintenance media (RPMIB27+insulin) and the other plate 21

was cultured in CDML3 media each for 10 days prior to electrophysiological recordings Figure 8 22

outlines the effects of CDML3 media on hiPSC-CM electrophysiological function assessed using 23

voltage sensitive dye (fluovolt) Figure 8A shows representative action potential traces from hiPSC-24

CMs cultured in media containing glucose (RPMI+B27) Figure 8B and C show examples of 1

spontaneous arrhythmias recorded in hiPSC-CMs that were cultured in zero glucoseabundant lactate 2

media (CDML3) for 10 days following MACS sorting and re-plating Importantly all recordings were 3

made in the same physiological salt solution (HBSS) following the media treatments-indicating that 4

the effects of each media persist following media changes Quantification of action potential duration 5

and action potential triangulation are presented in figure 8D-F APD30 was significantly prolonged in 6

hiPSC-CM monolayers cultured in CDML3 (3097plusmn 290 ms n=95 vs 2847plusmn 278 n=94 unpaired t-7

test Plt00001) APD90 was also significantly prolonged in hiPSC-CM monolayers cultured in CDML3 8

media (12667plusmn4999 ms n=95 vs 8232plusmn 3661 ms n=94 unpaired t-test Plt00001) Action 9

potential triangulation is a unit less parameter calculated by the equation 119860119875 119879119903119894119886119899119892119906119897119886119905119894119900119899 =10

(11986011987511986390 minus 11986011987511986330)11986011987511986390 Increased AP Triangulation occurs when the APD is prolonged and is 11

used as an index to describe the potential to cause arrhythmias hiPSC-CM monolayers cultured for 12

10 days in CDML3 media had greater AP triangulation values (072plusmn012 au n=95 vs 060plusmn013 au 13

n=94 unpaired t-test Plt00001) Increased AP triangulation increases the odds for observing 14

spontaneous arrhythmias in these monolayers cultured in zero glucose and abundant lactate 15

(CDML3) Indeed figure 8G shows that the increased APD values and increased AP triangulation 16

manifested as greater number of spontaneous arrhythmias occurring in the CDML3 treated group 17

(2295 well with arrhythmias vs 0094 wells with arrhythmias) 18

To rigorously test the effect of CDML3 media on a separate hiPSC-CM source high purity 19

commercially available cardiomyocytes (iCell cardiomyocytes2) were cultured in CDML3 media for 10 20

days Action potential duration (APD50) was significantly increased by metabolic stress induced 21

with CDML3 (APD5049924plusmn 93ms n=24 p=001) in comparison to RPMI media 22

(APD504649plusmn991ms n=24 Figure 8H) Intracellular calcium transient duration (CaTD50) was 23

significantly increased in syncytia treated with CDML3 (59677plusmn948 ms n=24) in comparison to 24

syncytia cultured in control media (53682plusmn889 ms n=24 plt00001 Figure 8I) Further in two 25

separate 96 well plates where one full plate was maintained in CDML3 media and the other plate in 1

RPMI media we found reduced intracellular calcium amplitude induced by the CDML3 media 2

treatment (RPMI B27=00866plusmn00013 au n=96 and CDML3=00369plusmn00012 au n=96 plt00001 3

Figure 8J) These data collectively support the hypothesis that media containing zero glucose and 4

abundant lactate can be used to simulate cardiac injury and create an arrhythmogenic substrate in 5

syncytia of commercially available hiPSC-CMs 6

At last in another set of experiments MACS purified hiPSC-CMs were cultured in CDML3 or 7

RPMI media prior to bioenergetics assessment with Seahorse XFe96 analyzer (19-9-11 hiPSC-CMs 8

7 days of CDML3 exposure) Oxygen consumption rate was affected by 7 days of exposure of MACS 9

separated cardiomyocytes to CDML3 media Mitochondrial oxygen consumption rate (OCR) was 10

significantly lower in cardiomyocytes after 7 days of exposure to metabolic stress although there are 11

no differences in ATP-driven and non-mitochondrial OCR (Figure 8K-M) Bioenergetics analysis 12

showed significant reduction of maximal respiration [RPMI B27=3414plusmn358 (pmolesmin)mg of 13

protein and CDML3=2534plusmn19 (pmolesmin)mg of protein p=002 Figure 8L] and decrease in 14

respiratory reserve capacity (RPMI B27=2304plusmn325 (pmolesmin)mg of protein and 15

CDML3=1456plusmn105 (pmolesmin)mg of protein p=0006 Figure 8M) in cardiomyocytes cultured 16

under metabolic stress conditions Seahorse XFe96 analyzer data concatenate with mitochondrial 17

membrane potential data described in the previous section 18

Discussion 19

To date the most widely used method for hiPSC-CM purification relies on metabolism 20

differences between cardiomyocytes and non-cardiomyocytes The widely used hiPSC-CM 21

purification media contains 0 glucose and abundant lactate a media composition that has been used 22

historically to simulate ischemia in native adult cardiomyocytes in vitro(17-19) Here we tested the 23

hypothesis that this metabolic mediated hiPSC-CM selection subjects the CMs to a metabolic stress 24

that induces an ischemic-like phenotype To this end we completed a direct comparison of the 25

functional and structural phenotypes of hiPSC-CMs purified using either metabolic selection media (0 1

glucose abundant lactate) or magnetic activated cell sorting (MACS) using commercially available 2

reagents (figure 1) A robust feature of this approach is that we were able to differentiate a batch of 3

hiPSCs to hiPSC-CMs and then utilize either purification approach on the same batch of cells Thus 4

the comparisons are tightly time-matched 5

We report distinct structural and functional phenotypes induced by each hiPSC-CM purification 6

approach hiPSC-CM selected by means of alteration of energetic substrate presented a reduction in 7

membrane bound connexin43 in contrast to hiPSC-CMs that were not submitted to metabolic stress 8

(Figure 2B) Changes in connexin43 expression are not pathognomonic of cardiac ischemia 9

nevertheless it has been long known from studies using animal models of ischemiahypoxia and 10

post-mortem human specimens that ischemia induces alterations in connexin43 expression 11

localization and phosphorylation impacts on cardiac function and produce a substrate for reentrant 12

arrhythmias (30-35) 13

Cardiomyocytes exposed to 0 glucose and abundant lactate had sarcomere shortening in 14

comparison to hiPSC-CMs selected with MACS (Figure 2C) similarly to the decrease in sarcomere 15

length demonstrated a long ago in in vitro and animal models of myocardial infarction hypoxia and 16

hypoxia with glycolytic blockade(36-38) Similar sarcomere disorganization was apparent with cTnT 17

staining in a separate hiPSC cell line (supplemental figure 1) Furthermore staining with phalloidin 18

revealed the abundant presence of non-striated F-actin fibers amongst sarcomeres (also known as 19

stress fibers) in metabolic stress selected cells (Figure 2D) whilst MACs purified hiPSC-CMs 20

presented striated pattern of phalloidin staining characteristic of sarcomeres Stress fibers are present 21

in different cell types and have important roles in cell adhesion and migration In cardiomyocytes 22

stress fibers participate in early stages of sarcomere morphogenesis (39) and they can be observed 23

in dedifferentiated cardiomyocytes adjacent to immobilized areas of the heart (40) Whether the 24

presence of stress fibers in cardiomyocytes submitted to metabolic stress is indicative of 25

dedifferentiation or immature phenotype is out of the scope of this manuscript nevertheless 1

cardiomyocytes purified with glucose deprivation and high availability of lactate present nuclear 2

expression of Ki67 and retain proliferation potential (supplemental figure 2)(14) and suggest the 3

hypothesis that metabolic selection delays hiPSC-CM maturation Although we have not further 4

explored the expression of genes associated to maturation of cardiomyocytes Ordono et al (41) has 5

shown that exposure of hiPSC-CMs to 6mM of lactate for 3 days is able to induce upregulation of 6

genes associated to dedifferentiation and proliferation 7

Functionally metabolic stress selected hiPSC-CM monolayers showed electrophysiological 8

and intracellular calcium handling properties reminiscent of cardiomyocytes isolated from ischemic 9

failing hearts Figure 3 summarizes the differences in hiPSC-CM calcium handling between the two 10

purification approaches Metabolic selected syncytial monolayers had higher diastolic calcium 11

concentrations reduced calcium transient amplitudes and reduced SERCA2a expression levels at 12

baseline relative to MACS sorted hiPSC-CMs (figure 3) These are all features reminiscent of the 13

phenotypes well documented for heart failure cardiomyocytes 14

The expression and phosphorylation status of calcium handling proteins regulate cardiac 15

function and are implicated as causal for poor function of the failing heart SERCA2a expression for 16

example is significantly reduced in animal models of heart failure and this molecular defect 17

contributes to diastolic dysfunction and poor contraction of the failing heart(10) In fact increasing 18

SERCA2a expression levels has shown to be effective to restore cardiac function in animal models of 19

heart failure and clinical trials have been conducted to deliver the SERCA2a gene to patients with 20

heart failure(42) Here we have discovered that metabolic stress selection of hiPSC-CMs produces 21

cardiomyocytes with reduced SERCA2a expression compared to hiPSC-CMs purified using MACS 22

approach (figure 3B) Furthermore we used SERCA2a gene therapy approach in vitro to restore the 23

function and arrhythmia phenotypes of metabolic stress selected hiPSC-CM monolayers (figure 7) 24

Another calcium handling protein consistently implicated in contributing to heart failure 1

phenotypes is phospholamban(43) It is well established that phospholamban is phosphorylated upon 2

myocardial β-adrenergic stimulation and this contributes to positive inotropic and lusitropic effects in 3

healthy hearts In cases of heart failure however phosphorylation of phospholamban is reduced and 4

this is attributed to dysfunctional adrenergic signaling(9 43 44) Here we have found that metabolic 5

selection generated hiPSC-CMs in which phospholamban phosphorylation did not occur upon 6

isoproterenol treatment (figure 3E) MACS selected hiPSC-CM monolayers however showed 7

significant phosphorylation of phospholamban upon isoproterenol stimulation (figure 3E) This 8

suggests a dysfunctional or underdeveloped β-adrenergic signaling pathway in metabolic stress 9

selected hiPSC-CMs These molecular differences most likely underlie the disparate functional 10

responses to isoproterenol presented in figure 3C MACS selected hiPSC-CM monolayers responded 11

to isoproterenol with positive chronotropic and inotropic responses On the other hand metabolic 12

selected hiPSC-CM monolayers only showed a positive chronotropic effect of isoproterenol treatment 13

with no significant positive inotropic effect cTnI phosphorylation however was induced by 14

isoproterenol treatment in both groups These molecular and cellular differences of intracellular 15

calcium flux in hiPSC-CMs purified by distinct methods resemble the extensive comparisons that 16

have been made historically between native cardiomyocytes isolated from healthy and failing hearts 17

Key electrophysiological parameters including MDP APD and rate adaptation were also 18

impacted by purification approach Figure 4 outlines the key electrophysiological differences that we 19

observed between metabolic selected hiPSC-CMs and MACS selected hiPSC-CMs Microelectrode 20

recordings in figure 4A show differences in maximal diastolic potential (MDP) and action potential 21

amplitude (APA) between hiPSC-CMs purified by each method MDP was significantly depolarized in 22

metabolic selected hiPSC-CMs syncytial monolayers compared to MACS selected hiPSC-CMs 23

Depolarization of MDP and smaller APA may be attributed to elevated diastolic calcium 24

concentrations observed in metabolic selected hiPSC-CMs relative to MACS selected hiPSC-CM 25

monolayers Depolarization of MDP also contributes to the differences of baseline spontaneous 1

contraction rate between the groups (figure 4B)-depolarized MDP sets the membrane potential closer 2

to threshold for firing an action potential Action potential duration (APD) was significantly prolonged 3

in metabolic selected hiPSC-CM monolayers even over a range of electrical pacing frequencies 4

(figure 4C) APD prolongation is a feature well known to occur in heart failure contributes to QT 5

prolongation of the EKG and increases heart failure patientsrsquo risk for suffering a fatal arrhythmia(45) 6

Arrhythmia testing by rapid reduction of pacing frequency from 25 to 05Hz revealed significantly 7

greater incidence of EADs and extra contractions in metabolic selected hiPSC-CM monolayers 8

relative to MACS selected monolayers (figure 4D) The electrophysiological differences between 9

hiPSC-CM monolayers purified by each method outlined in figure 1 resemble the well-established 10

differences between healthy and heart failure cardiomyocytes(45) 11

Furthermore we have observed changes in mitochondrial function that recapitulate 12

bioenergetic features of cardiomyocytes submitted to ischemia (Figure 5 and Figure 8K-M) namely 13

an overall reduction of mitochondrial membrane potential and function This was measured using 14

fluorescent mitochondrial membrane potential sensitive dyes and with a respirometry assay 15

performed with a Seahorse XFe96 analyzer Figure 8K-M show the bioenergetic differences between 16

hiPSC-CM monolayers that were maintained in either CDML3 (metabolic selection media) or glucose 17

containing RPMI media for seven days Metabolic selection media produced hiPSC-CM monolayers 18

with reduced maximal oxygen consumption rate and respiratory reserve capacity The reduction of 19

maximal oxygen consumption rate is similar to clinical data showing that human cardiomyocytes 20

isolated from the left atrial appendage from patients suffering of chronic myocardial ischemia have 21

loss of maximal oxygen consumption rate(46) Inherent mitochondrial functional differences also 22

underlie the enhanced sensitivity to DOX-TOX observed in metabolic selected hiPSC-CMs (figure 6) 23

In combination with elevated calcium concentrations in metabolic stress selected hiPSC-CMs the 24

mitochondrial dysfunction combines to increase sensitivity to DOX-TOX Metabolic selected hiPSC-25

CMs showed increased apoptosis after doxorubicin treatment compared to MACS sorted hiPSC-CMs 1

(figure 6B) Also DOX-TOX indexed by arrhythmia occurrence was more apparent in metabolic 2

stress selected hiPSC-CMs relative to MACS sorted hiPSC-CM monolayers (figure 6CampD) This 3

indicates that the purification approach should be carefully considered when using hiPSC-CMs to 4

predict patient specific sensitivity to chemotherapy induced cardiotoxicity 5

The approach of metabolically starving non-myocytes in order to enrich cardiomyocyte 6

populations derived from stem cells has been employed for over seven years(14) and has proven to 7

be an effective approach The metabolic starvation technique is attractive owing to its ease of use 8

applicability to bulk processing and extremely low cost Here we have used the same metabolic 9

starvation approach with subtle modification for comparison to a novel extracellular receptor-magnetic 10

bead based approach that is commercially available (Miltenyi Biotec) Compared to the MACS based 11

purification metabolic selection generates cardiomyocytes with a heart failure like molecular and 12

functional phenotype This provides a novel acquired heart failure model system for in vitro testing 13

and research There are some slight differences in protocols that require discussion The first reports 14

of using metabolic starvation used 8 days of treatment for purification of cells and later publications 15

using this method applied longer durations up to 10 days (16) Figure 1 shows that we used 10 days 16

of metabolic media selection to obtain hiPSC-CMs for our comparisons except for oxygen 17

consumption rate experiments that were performed after 7 days of continuous exposure to CDM3L 18

Although we demonstrated that 7 days of treatment with CDM3L affects energy metabolism of 19

cardiomyocytes we cannot state that shorter exposures to CDM3L such as used by Sharma et al 20

(47) induce an ischemic phenotype and this should be subjected to further investigation During the 21

differentiation protocol previous reports began the metabolic media purification treatment at day 10 22

however we have initiated the metabolic selection at day 14 The timing of introducing metabolic 23

selection and the duration of exposure may impact on cellular phenotypes Despite the subtle 24

protocol differences the rigorous experimental design using time matched controls in each 25

differentiation experiment for each purification approach here enabled the comparisons between 1

hiPSC-CM purification approaches 2

Conclusion 3

Heart failure affects approximately 65 million Americans(48) the majority of these heart failure 4

patients suffer ischemic heart failure which generally occurs secondary to coronary heart disease 5

Current heart failure treatments are palliative in nature and cardiac transplantation is possible but 6

inadequate to address a significant portion of the population A major obstacle to development of new 7

and effective heart failure treatments has been the lack of viable human cardiac tissue and myocytes 8

available for research Our results indicate that human heart failure molecular and functional 9

phenotypes can be modeled in vitro using metabolic selection of hiPSC-CMs The use of acquired 10

heart failure in vitro models using human cardiac syncytial monolayers offers advantages over the 11

use of animal models and provides a testing platform for development of new heart failure therapies 12

The use of syncytial monolayers of human cardiomyocytes can only be accomplished using 13

pluripotent stem cell derived cardiac myocytes (PSC-CMs) and here we have validated that these 14

monolayers are sensitive to metabolic stress and can recapitulate many molecular and functional 15

features well known to occur in cardiomyocytes obtained from failing human hearts 16

Methods 19

hiPSC maintenance Vector free hiPSCs were maintained on matrigel coated 6 well plates using 20

Xeno-free media (iPS Brew Miltenyi Biotec) essentially as described before(20 49) Two vector free 21

control cell lines were utilized to determine efficiency of cardiomyocyte purification approaches (19-9-22

11 and BJ iPSCs) The 19-9-11 hiPSC line was obtained from WiCell and the BJ iPSC line was 23

generated from commercially available human dermal fibroblasts in the laboratory using mRNA 24

reprogramming hiPSCs were maintained as colonies and passaged once every week Pluripotent 1

stem cell use was approved by the University of Michigan HPSCRO Committee 2

Cardiac Directed Differentiation and hiPSC-CM purification Figure 1A outlines the cardiac 3

directed differentiation and purification protocols We utilized the highly efficient small molecule 4

approach relying on temporal modulation of the Wnt signaling pathway(50) On day 14 cardiac 5

differentiation cultures were switched to metabolic stress selection media (CDM3L 0 glucose 5mM 6

sodium DL-lactate) or maintained in cardiomyocyte maintenance media (RPMIB27+insulin) Phase 7

contrast images of the differentiated cardiomyocytes in CDM3L or RPMI maintenance media are 8

shown in figure 1B and indicate significant non-myocyte cell death in CDM3L media-as expected 9

based on previous reports(16) Figure 1C data shows time matched spontaneous calcium transient 10

recordings of hiPSC-CM monolayers either with no selection or on day 10 of metabolic stress 11

selection process prior to replating The non-myocyte cell death creates holes in the monolayer that 12

in addition to the high lactate concentrations contribute to the observed dysfunction For metabolic 13

selection of hiPSC-CMs cell cultures were kept in CDM3L for 10 days and on day 24 purified hiPSC-14

CMs were trypsinized and re-plated as confluent monolayers for phenotype analysis This is a slight 15

deviation from the original reports of Tohyama et al which used 6 to 8 days of metabolic stress 16

selection with CDM3L to purify stem cell derived cardiomyocytes to 93 and 98 respectively(14 17

51 52) There is however a precedence for our use of 10 days of CDM3L metabolic stress selection 18

media in the work of Burridge et al(16) In 2014 Burridge et al(16) reported that 6-10 days of glucose 19

deprivation using CDM3L was optimal to purify cardiomyocytes from 85 to gt95 purity In fact 10 20

days of selection was found to purify hiPSC-CMs to a greater extent than shorter time periods (6 21

days) For MACS hiPSC-CM purification differentiation plates were kept in normal maintenance 22

media and processed for cardiomyocyte purification on day 24 of the protocol (Miltenyi Biotec PSC-23

derived cardiomyocyte isolation kit human) The MACS purification approach uses antibody to target 24

a non-myocyte extracellular epitope to magnetize the non-myocytes which are trapped in a magnetic 25

field while non-labelled cardiomyocytes flow through the magnetic field and are collected similarly to 1

the alternative approach to the use of SIRPA antibodies for positive selection of cardiomyocytes 2

suggested by Dubois et al (53) In each group the media was changed every other day Flow 3

cytometry was performed to assess the relative purity of cardiomyocyte cultures purified with the two 4

methods by probing cTnT expression (Anti-cardiac Troponin T-APC Miltenyi Biotec 130-120-403) 5

On the day of MACS purification hiPSC-CMs were re-plated in parallel to the metabolic selected 6

hiPSC-CMs for phenotype comparison Plating procedures have been described recently in 7

detail(20 26 49) In each condition phenotype analysis to study the structure and function of the 8

hiPSC-CM monolayers was done 7 days after re plating This approach enabled phenotype analysis 9

of the same batch of hiPSC-CMs that were time-matched but purified by distinct methods 10

Experiments were done to determine the effectiveness of an established gene therapy reported to 11

reverse heart failure phenotypes This data is in figure 8 Ad-mCherry-hATP2A2 was obtained from 12

Vector Biolabs (Malvern PA) stored at -80C and applied to hiPSC-CM monolayers plated in 96 well 13

plates using MOI=100 (MOI=Multiplicity of Infection 50000 hiPSC-CMs per well) mCherry 14

expression was detected using live cell time lapse imaging (Incucyte Zoom Essen Bioscience Ann 15

Arbor MI) and expression was robust by 48h following gene transfer Following optical mapping of 16

intracellular calcium flux using fluo4AM (10μM) monolayers were collectedsolubilized in SDS sample 17

buffer for Western blot detection of SERCA2a protein 18

Cardiac monolayer optical mapping and electrophysiology hiPSC-CM monolayer 19

electrophysiology was measured using optical mapping or micro electrode recording as previously 20

described(20 26 49) Action potential properties were measured using the voltage sensitive dye 21

fluovolt (ThermoFisher) or by microelectrode recording Action potential duration (APD80) restitution 22

was quantified using field stimulation of monolayers (20V 5ms duration various Hz) Arrhythmia 23

screening was done using a dynamic stimulation frequency assay as previously described(54) 24

Briefly monolayers were paced at 25Hz and action potential responses to immediate slowing of 1

pacing frequency (10Hz) were quantified 2

Calcium transients were recorded using either rhod2AM fluo-4AM or the more quantitative 3

calcium dye-Fura 2AM For fura2AM measurements monolayers were re-plated on to glass bottom 4

dishes for microscopic imaging using an IonOptix system with alternating excitation between 340nm 5

and 380nm ultraviolet light with emission recordings in the green spectrum (515nm) Fura2 6

measurements are independent of cellular motion and amount of dye loaded and thus reflect 7

intracellular calcium concentration changes Fura2 measurements are expressed as ratio units In 8

each purification approach intracellular calcium response to β-adrenergic receptor stimulation 9

(500nM isoproterenol) was quantified using Fura2 loaded monolayers Rhod2 loaded monolayers 10

(10microM) were used to determine sensitivity to doxorubicin induced arrhythmias in large monolayers 11

Control Experiments using Commercially Available Purified hiPSC-CMs In control experiements 13

we tested the hypothesis that CDML3 metabolic stress media also induces electrophysiological 14

abnormalities in commercially available human iPSC-CMs To test the effect of CDML3 metabolic 15

stress media on hiPSC-CMs from another source we applied this media to commercially available 16

human cardiomyocytes (iCell2Cellular Dynamics International Madison WI) that are purified by 17

mechanisms other than those used in this study These cells were handled plated and used for 18

optical mapping as outlined before(20 21) Results of these experiments are presented in 19

supplemental figure 5 First we generated a 96 well plate of these cells using 50000 cells per well to 20

form monolayers One half of the plate was cultured in RPMI (glucose containing) media and the 21

other half of the plate was cultured in CDML3 (zero glucose high lactate) media After 10 days in 22

culture the plate was used for functional analysis with half of the plate being used for action potential 23

measurements (Fluovolt) and the other half being used for calcium transient analysis (fluo-4AM) 24

Next to determine the effect of CDML3 media on calcium transient amplitudes two additional 96 well 25

plates of iCell2 cardiomyocytes were generated In one plate all 96 wells were cultured in RPMI 1

media for 10 days while the other plate was maintained in CDML3 media for the same 10 days On 2

day 10 cells were loaded with rhod2AM (10microM for 15min) and spontaneous calcium flux was 3

recorded using the LED based optical mapping system 4

Mitochondrial membrane potential measurements Mitochondrial membrane potential (Δψm) was 5

determined using the carbocyanine compound JC-1 the optimal dye to use for quantification of 6

changes of ψ in cardiomyocytes(55 56) JC-1 dye exhibits potential dependent accumulation in 7

mitochondria indicated by a shift in fluorescence emission from green to red As a result 8

mitochondrial depolarization is indicated by a decrease of the redgreen fluorescence intensity ratio 9

JC-1 dye was loaded into cardiomyocytes (2microM 30 min) in the incubator at 37˚C Red and green 10

emission was recorded in live cells using a Nikon A1R confocal microscope with the appropriate laser 11

and detection wavelengths FCCP (50 microM carbonilcyanide p-trifluoromethoxyphenylhydrazone) a 12

potent uncoupler of mitochondrial oxidative phosphorylation was applied to validate the JC-1 13

measurements 14

Western blotting analysis Protein expression was analyzed by Western blotting essentially as 15

previously described(20 49) For each purification approach SERCA2a protein expression was 16

determined using a monoclonal antibody (ThermoFisher MA3-919 1300) Phospho-specific 17

antibodies were used to determine phosphorylation levels of cardiac troponin I (cTnI) (Cell Signaling 18

Technology 4004S 1100) and phospholamban (PLB) (Millipore 07-052 1150) with or without 19

isoproterenol treatment (500nM) 20

Immunofluorescence Monolayer structure was observed by immunofluorescence as previously 21

described 14 For each purification approach sarcomeric α-actinin (Sigma A7811 1300) and 22

connexin 43 (Sigma C6219 1100) protein expression were visualized using monoclonal antibodies 23

Cell nuclei were also labeled using DAPI (11000) α-actinin spacing was measured using 24

fluorescence intensity-space plots as shown in figure 2C Confocal images were collected using a 25

Nikon A1R confocal microscope (60X objective) from five separate monolayers generated using each 1

purification method Average sarcomere spacing was quantified from at least seven intensity-space 2

plots drawn perpendicular to separate sarcomere patterns for each image 3

Apoptosis Assay The extent of hiPSC-CM apoptosis immediately following each purification 4

approach was quantified using Annexin V Green reagent (Essen Bioscience Cat No 4642) 5

Monolayers resulting from each purification method were plated on PDMS and incubated with 6

annexin V (ThermoFisher A13201) and doses of doxorubicin ranging from 0nM to 1000nM for 48 7

hours Cells were then imaged using the Incucyte Zoom time lapse microscope (Essen Biosciences 8

Ann Arbor MI) and annexin V fluorescence analyzed with its software 9

Author Contributions 10

Experimental design was by all authors Cell culture experiments including stem cell 11

maintenance cardiac differentiation purification and replating were done by JD AC JC AMR and 12

KC Microelectrode recordings of action potentials were done by DPB and EJV Optical mapping of 13

action potentials and calcium transients were done by JD TJH JC and AMR Mitochondrial 14

membrane potentials were measured by JD TJH and AMR AL AMR and NM executed bioenergetic 15

analysis using the SeaHorse All authors contributed to writing and editing the manuscript 16

Acknowledgments 18

The authors wish to acknowledge Haydon Herron Fiona Patrick Carter Herron and James 19

Patrick for assistance with electrophysiology optical mapping data analysis This study was supported 20

by funds from the Lefkofsky Family Foundation (TJH) from the University of Michigan Frankel 21

Cardiovascular Center and NIH (R44ES027703-03 from NIEHS TJH) 22

References 2

1 Kamp TJ and Lyons GE On the road to iPS cell cardiovascular applications Circ Res 3 2009105(7)617-9 4

2 Zhang J Wilson GF Soerens AG Koonce CH Yu J Palecek SP et al Functional cardiomyocytes 5 derived from human induced pluripotent stem cells Circ Res 2009104(4)e30-41 6

3 Kamp TJ An electrifying iPSC disease model long QT syndrome type 2 and heart cells in a dish Cell 7 Stem Cell 20118(2)130-1 8

4 Itzhaki I Maizels L Huber I Zwi-Dantsis L Caspi O Winterstern A et al Modelling the long QT 9 syndrome with induced pluripotent stem cells Nature 2011471(7337)225-9 10

5 Lan F Lee Andrew S Liang P Sanchez-Freire V Nguyen Patricia K Wang L et al Abnormal Calcium 11 Handling Properties Underlie Familial Hypertrophic Cardiomyopathy Pathology in Patient-Specific 12 Induced Pluripotent Stem Cells Cell Stem Cell 201312(1)101-13 13

6 Novak A Barad L Lorber A Gherghiceanu M Reiter I Eisen B et al Functional abnormalities in iPSC-14 derived cardiomyocytes generated from CPVT1 and CPVT2 patients carrying ryanodine or 15 calsequestrin mutations J Cell Mol Med 201519(8)2006-18 16

7 Yazawa M Hsueh B Jia X Pasca AM Bernstein JA Hallmayer J et al Using induced pluripotent stem 17 cells to investigate cardiac phenotypes in Timothy syndrome Nature 2011471(7337)230-4 18

8 Herron TJ Devaney E Mundada L Arden E Day S Guerrero-Serna G et al Ca2+-independent 19 positive molecular inotropy for failing rabbit and human cardiac muscle by α-myosin motor gene 20 transfer The FASEB Journal 201024(2)415-24 21

9 Chu G and Kranias EG Phospholamban as a therapeutic modality in heart failure Novartis Found 22 Symp 2006274156-71 discussion 72-5 272-6 23

10 Gianni D Chan J Gwathmey JK del Monte F and Hajjar RJ SERCA2a in heart failure role and 24 therapeutic prospects J Bioenerg Biomembr 200537(6)375-80 25

11 del Monte F Harding SE Schmidt U Matsui T Kang ZB Dec GW et al Restoration of contractile 26 function in isolated cardiomyocytes from failing human hearts by gene transfer of SERCA2a 27 Circulation 1999100(23)2308-11 28

12 Park WJ and Oh JG SERCA2a a prime target for modulation of cardiac contractility during heart 29 failure BMB Rep 201346(5)237-43 30

13 Zhai Y Luo Y Wu P and Li D New insights into SERCA2a gene therapy in heart failure pay attention 31 to the negative effects of B-type natriuretic peptides J Med Genet 201855(5)287 32

14 Tohyama S Hattori F Sano M Hishiki T Nagahata Y Matsuura T et al Distinct metabolic flow 33 enables large-scale purification of mouse and human pluripotent stem cell-derived cardiomyocytes Cell 34 Stem Cell 201312(1)127-37 35

15 Burridge PW Li YF Matsa E Wu H Ong SG Sharma A et al Human Induced Pluripotent Stem Cellndash36 Derived Cardiomyocytes Recapitulate the Predilection of Breast Cancer Patients to Doxorubicinndash37 Induced Cardiotoxicity Nat Med 201622(5)547-56 38

16 Burridge PW Matsa E Shukla P Lin ZC Churko JM Ebert AD et al Chemically Defined and Small 39 Molecule-Based Generation of Human Cardiomyocytes Nature methods 201411(8)855-60 40

17 Diaz RJ and Wilson GJ Studying ischemic preconditioning in isolated cardiomyocyte models 41 Cardiovasc Res 200670(2)286-96 42

18 Vanden Hoek TL Shao Z Li C Zak R Schumacker PT and Becker LB Reperfusion injury on cardiac 43 myocytes after simulated ischemia American Journal of Physiology-Heart and Circulatory Physiology 44 1996270(4)H1334-H41 45

19 Arutunyan A Webster DR Swift LM and Sarvazyan N Localized injury in cardiomyocyte network a 46 new experimental model of ischemia-reperfusion arrhythmias American journal of physiology Heart 47 and circulatory physiology 2001280(4)H1905-H15 48

20 da Rocha AM Campbell K Mironov S Jiang J Mundada L Guerrero-Serna G et al hiPSC-CM 49 Monolayer Maturation State Determines Drug Responsiveness in High Throughput Pro-Arrhythmia 50 Screen Sci Rep 20177(1)13834 51

21 da Rocha AM Creech J Thonn E Mironov S and Herron TJ Detection of Drug-Induced Torsades de 1 Pointes Arrhythmia Mechanisms Using hiPSC-CM Syncytial Monolayers in a High-Throughput 2 Screening Voltage Sensitive Dye Assay Toxicol Sci 2020173(2)402-15 3

22 Yu J Hu K Smuga-Otto K Tian S Stewart R Slukvin II et al Human induced pluripotent stem cells 4 free of vector and transgene sequences Science (New York NY) 2009324(5928)797-801 5

23 Herron TJ Rocha AM Campbell KF Ponce-Balbuena D Willis BC Guerrero-Serna G et al 6 Extracellular Matrix-Mediated Maturation of Human Pluripotent Stem Cell-Derived Cardiac Monolayer 7 Structure and Electrophysiological Function Circulation Arrhythmia and electrophysiology 8 20169(4)e003638 9

24 Bizy A Guerrero-Serna G Hu B Ponce-Balbuena D Willis BC Zarzoso M et al Myosin light chain 2-10 based selection of human iPSC-derived early ventricular cardiac myocytes Stem Cell Research 11 201311(3)1335-47 12

25 Lee P Klos M Bollensdorff C Hou L Ewart P Kamp TJ et al Simultaneous voltage and calcium 13 mapping of genetically purified human induced pluripotent stem cell-derived cardiac myocyte 14 monolayers Circ Res 2012110(12)1556-63 15

26 Monteiro da Rocha A Guerrero-Serna G Helms A Luzod C Mironov S Russell M et al Deficient 16 cMyBP-C protein expression during cardiomyocyte differentiation underlies human hypertrophic 17 cardiomyopathy cellular phenotypes in disease specific human ES cell derived cardiomyocytes J Mol 18 Cell Cardiol 201699197-206 19

27 Zhang J Tao R Campbell KF Carvalho JL Ruiz EC Kim GC et al Functional cardiac fibroblasts 20 derived from human pluripotent stem cells via second heart field progenitors Nat Commun 21 201910(1)2238 22

28 Rosca MG and Hoppel CL Mitochondrial dysfunction in heart failure Heart Fail Rev 201318(5)607-23 22 24

29 Zheng X Halle S Yu K Mishra P Scherr M Pietzsch S et al Cardiomyocytes display low 25 mitochondrial priming and are highly resistant toward cytotoxic T-cell killing Eur J Immunol 26 201646(6)1415-26 27

30 Peters NS Coromilas J Severs NJ and Wit AL Disturbed connexin43 gap junction distribution 28 correlates with the location of reentrant circuits in the epicardial border zone of healing canine infarcts 29 that cause ventricular tachycardia Circulation 199795(4)988-96 30

31 Peters NS Myocardial gap junction organization in ischemia and infarction Microsc Res Tech 31 199531(5)375-86 32

32 Lampe PD Cooper CD King TJ and Burt JM Analysis of Connexin43 phosphorylated at S325 S328 33 and S330 in normoxic and ischemic heart J Cell Sci 2006119(Pt 16)3435-42 34

33 Kawamoto O Michiue T Ishikawa T and Maeda H Immunohistochemistry of connexin43 and zonula 35 occludens-1 in the myocardium as markers of early ischemia in autopsy material Histol Histopathol 36 201429(6)767-75 37

34 Martins-Marques T Anjo SI Pereira P Manadas B and Girao H Interacting Network of the Gap 38 Junction (GJ) Protein Connexin43 (Cx43) is Modulated by Ischemia and Reperfusion in the Heart Mol 39 Cell Proteomics 201514(11)3040-55 40

35 Zeevi-Levin N Barac YD Reisner Y Reiter I Yaniv G Meiry G et al Gap junctional remodeling by 41 hypoxia in cultured neonatal rat ventricular myocytes Cardiovasc Res 200566(1)64-73 42

36 Bing OH and Fishbein MC Mechanical and structural correlates of contracture induced by metabolic 43 blockade in cardiac muscle from the rat Circ Res 197945(2)298-308 44

37 Schwartz P Piper HM Spahr R and Spieckermann PG Ultrastructure of cultured adult myocardial 45 cells during anoxia and reoxygenation Am J Pathol 1984115(3)349-61 46

38 Ait Mou Y Reboul C Andre L Lacampagne A and Cazorla O Late exercise training improves non-47 uniformity of transmural myocardial function in rats with ischaemic heart failure Cardiovasc Res 48 200981(3)555-64 49

39 Fenix AM Neininger AC Taneja N Hyde K Visetsouk MR Garde RJ et al Muscle-specific stress 50 fibers give rise to sarcomeres in cardiomyocytes Elife 20187 51

40 Dispersyn GD Mesotten L Meuris B Maes A Mortelmans L Flameng W et al Dissociation of 52 cardiomyocyte apoptosis and dedifferentiation in infarct border zones Eur Heart J 200223(11)849-57 53

41 Ordontildeo J Peacuterez-Amodio S Ball K Aguirre A and Engel E Lactate promotes cardiomyocyte 1 dedifferentiation through metabolic reprogramming bioRxiv 2020 2 httpswwwbiorxivorgcontent10110120200721213736v1 3

42 Greenberg B Butler J Felker GM Ponikowski P Voors AA Desai AS et al Calcium upregulation by 4 percutaneous administration of gene therapy in patients with cardiac disease (CUPID 2) a randomised 5 multinational double-blind placebo-controlled phase 2b trial The Lancet 2016387(10024)1178-86 6

43 Waggoner JR and Kranias EG Role of Phospholamban in the Pathogenesis of Heart Failure Heart 7 Fail Clin 20051(2)207-18 8

44 Frank K and Kranias EG Phospholamban and cardiac contractility Ann Med 200032(8)572-8 9 45 Wickenden AD Kaprielian R Kassiri Z Tsoporis JN Tsushima R Fishman GI et al The role of action 10

potential prolongation and altered intracellular calcium handling in the pathogenesis of heart failure 11 Cardiovasc Res 199837(2)312-23 12

46 Schipper DA Palsma R Marsh KM OHare C Dicken DS Lick S et al Chronic Myocardial Ischemia 13 Leads to Loss of Maximal Oxygen Consumption and Complex I Dysfunction Ann Thorac Surg 14 2017104(4)1298-304 15

47 Sharma A Marceau C Hamaguchi R Burridge PW Rajarajan K Churko JM et al Human induced 16 pluripotent stem cell-derived cardiomyocytes as an in vitro model for coxsackievirus B3-induced 17 myocarditis and antiviral drug screening platform Circ Res 2014115(6)556-66 18

48 Benjamin EJ Blaha MJ Chiuve SE Cushman M Das SR Deo R et al Heart Disease and Stroke 19 Statistics-2017 Update A Report From the American Heart Association Circulation 20 2017135(10)e146-e603 21

49 Herron TJ Rocha AMD Campbell KF Ponce-Balbuena D Willis BC Guerrero-Serna G et al 22 Extracellular MatrixndashMediated Maturation of Human Pluripotent Stem CellndashDerived Cardiac Monolayer 23 Structure and Electrophysiological Function Circulation Arrhythmia and Electrophysiology 20169(4) 24

50 Lian X Zhang J Azarin SM Zhu K Hazeltine LB Bao X et al Directed cardiomyocyte differentiation 25 from human pluripotent stem cells by modulating Wntβ-catenin signaling under fully defined conditions 26 Nat Protocols 20138(1)162-75 27

51 Tohyama S Fujita J Fujita C Yamaguchi M Kanaami S Ohno R et al Efficient Large-Scale 2D 28 Culture System for Human Induced Pluripotent Stem Cells and Differentiated Cardiomyocytes Stem 29 Cell Reports 20179(5)1406-14 30

52 Tohyama S Fujita J Hishiki T Matsuura T Hattori F Ohno R et al Glutamine Oxidation Is 31 Indispensable for Survival of Human Pluripotent Stem Cells Cell Metab 201623(4)663-74 32

53 Dubois NC Craft AM Sharma P Elliott DA Stanley EG Elefanty AG et al SIRPA is a specific cell-33 surface marker for isolating cardiomyocytes derived from human pluripotent stem cells Nat Biotechnol 34 201129(11)1011-8 35

54 Herron TJ Milstein ML Anumonwo J Priori SG and Jalife J Purkinje cell calcium dysregulation is the 36 cellular mechanism that underlies catecholaminergic polymorphic ventricular tachycardia Heart 37 Rhythm 20107(8)1122-8 38

55 Dedkova EN and Blatter LA Measuring mitochondrial function in intact cardiac myocytes J Mol Cell 39 Cardiol 201252(1)48-61 40

56 Mathur A Hong Y Kemp BK Barrientos AA and Erusalimsky JD Evaluation of fluorescent dyes for 41 the detection of mitochondrial membrane potential changes in cultured cardiomyocytes Cardiovasc 42 Res 200046(1)126-38 43

Figure 1 hiPSC-CM differentiation protocol and purification approaches A Timeline of small molecule based

cardiac directed differentiation and purification approaches B Phase contrast images showing the impact of

metabolic selection on monolayer confluence (19-1-11 hiPSC line day 24) C hiPSC-CM spontaneous calcium

transient recordings in 19-9-11 hiPSC-CMs without any selection (black n=6 monolayers) and with metabolic

selection (red n=5 monolayers) Spontaneous calcium transient alternans were induced by metabolic selection media

(red traces and bars) denotes significant difference of amplitude between even and odd numbered beats

Quantification in panel C shows the relative amplitude of each beat following beat 1-normalized to beat 1

amplitude Without selection the amplitude did not vary beat to beat (beat 2=097plusmn002 beat 3=098plusmn001 n=6)

However in lactate selection media (CDML3) the beat amplitude varied on a beat to beat basis (beat2=084plusmn011

beat 3=099plusmn001 n=6 paired t-test Plt005) indicating calcium transient alternans

Figure 2 hiPSC-CMs generated using two distinct purification approaches A Each purification approach is

equally effective to enrich hiPSC-CM population analyzed by flow cytometry with cTnT labelling B Connexin43

(Cx43) mislocalization was apparent in metabolic stress selected monolayers C Metabolic stress selected hiPSC-

CMs had shorter sarcomere length (MACS=185plusmn017microm n=35 vs metabolic stress selection=177plusmn013microm n=40

denotes difference unpaired t-test P=0009 Sarcomere length was quantified using the repeating fluorescence

pattern of the α-actinin staining D Phalloidin staining was used to examine the actin cytoskeleton Stress fibers

rather than sarcomeres were identified in the metabolic stress selected hiPSC-CMs

Figure 3 hiPSC-CM purification approach impacts on cardiomyocyte intracellular calcium flux A Metabolic stress selection significantly alters baseline calcium flux of hiPSC-CMs B SERCA2a expression levels were reduced in metabolic stress selected hiPSC-CMs C MACS purified cardiomyocytes (black traces symbols) responded as expected to isoproterenol metabolic stress selected hiPSC-CMs however only responded with positive chronotropy D cTnI phosphorylation state was elevated at baseline in metabolic stress selected hiPSC-CMs (019plusmn005 n=3 vs 003plusmn002au n=3) cTnI phosphorylation increased following isoproterenol stimulation (metabolic stress=075plusmn029 n=3 and MACS=099plusmn030au n=3) E PLN phosphorylation levels (PLN-P) were similarly low at baseline in each group and only in MACS sorted hiPSC-CMs was PLN-P detected at significant levels following isoproterenol treatment (metabolic stress=005plusmn006au n=3 vs MACS=077plusmn035au n=3 P=002)

Figure 4 hiPSC-CM purification approach impacts electrophysiology phenotypes A Microelectrode recordings

show Maximal Diastolic Potential (MDP) was depolarized in metabolic stress selected hiPSC-CMs Action potential

amplitude (APA) was greater in MACS selected hiPSC-CM monolayers (MACS=1035plusmn125mV n=11 vs metabolic

stress selection=815plusmn179mV n=9denotes significant difference unpaired t-test Plt0001) B Metabolic selected

monolayers had significantly greater spontaneous beating frequency C Metabolic stress selection monolayers had

significantly longer APD80 than MACS selected monolayers over a range of pacing frequencies (1Hz

metabolic=35712plusmn685ms 15Hz metabolic=31224plusmn703ms 20Hzmetabolic=27938plusmn621ms 25Hz

metabolic=25029plusmn420ms n=9-12 monolayers) vs (1Hz MACS= 30956plusmn685ms15Hz MACS= 27916plusmn383ms

20Hz MACS= 24860plusmn631ms 25Hz MACS=21693plusmn420ms n=12 monolayers meanplusmnsem) D Arrhythmia testing

was done by abrupt slowing of pacing from 25 to 10Hz with continuous recording MACS purified monolayers could

follow the pacing change with 11 capture however 722 of metabolic selected monolayers displayed arrhythmias

Figure 5 Metabolic selection induces mitochondrial dysfunction in hiPSC-CM monolayers A JC-1

staining indicates more polarized mitochondria in MACS purified monolayers Numeric values are in the

main text B Mitotracker Red CMX Ros staining supports the JC-1 resultsMitotracker signal

MACS=136911plusmn 35566 au vs metabolic stress selection= 63545plusmn 30212 t test Plt0001

Figure 6 Metabolic stress selected hiPSC-CM monolayers have increased sensitivity to

chemotherapy induced cardiotoxicity (DOX-TOX) A Annexin V staining indicates greater extent of

apoptosis in metabolic purified hiPSC-CMs treated with 1000nM Doxorubicin (DOX) B Calcium

transient optical mapping using rhod2 example traces show greater sensitivity of electrophysiological

function of metabolic stress selection hiPSC-CMs (red traces) C Metabolic stress selected hiPSC-

CM monolayers show arrhythmias in response to DOX-TOX with greater sensitivity than MACS

purified hiPSC-CM monolayers

Figure 7 SERCA2a calcium pump gene therapy reverses hiPSC-CM heart failure phenotype A

Viral transduction was verified by mCherry live cell expression in hiPSC-CM monolayers Further

SERCA2a protein expression was determined by Western Blotting Quantification indicates that SERCA2a

protein expression was reduced in metabolic stress treated hiPSC-CMs at baseline(MACS=211plusmn022 n=3

vs Metbolic stress=127plusmn006 n=3) AdSERCA2a treatment restonred SERCA2a levels to control levels

(Metabolic stress + Ad SERCA2a=215plusmn017 n=3 ANOVA Plt005) B Calcium transient measurements

indicate that metabolic stress selection media prolongs the cardiac action potenctial and spontaneous

arrhythmias were observed in the metabolic stess selection purification approach hiPSC-CM monolayers

treated with AdSERCA2a did not develop arrhythmias C Calcium transient duration 50 (CaTD50) was

prolonged in metabolic stress media treated cells at baseline and was corrected to control values with

AdSERCA2a gene therapy (CaTD50s MACS baseline=4755 plusmn181 ms metbolic stress baseline=57737 plusmn

177 ms MACS + AdSERCA2a=3966 plusmn112 ms Metabolic stress + Ad SERCA2a=49367 plusmn199 ms

ANOVA P=001 P=0002 ns=not significant)

Figure 8 CDML3 metabolic selection media induces arrhythmia phenotypes in MACS purified 19-9-11

hiPSC-CM monolayers and antibiotic resistance purified hiPSC-CMs A Action potential traces (fluovolt) of

hiPSC-CMs maintained in RPMI+B27 maintenance media BampC CDML3 induced arrhythmia phenotypes observed

using voltage sensitive dye D APD30 E APD50 and F AP triangulation are greater in CDML3 treated hiPSC-CMs

RPMI+B27 APD30=2847plusmn278 ms CDML3 APD30=3097plusmn290 ms unpaired t-tests Plt0001 n=94 and n=95

RPMI APD90=8232plusmn3661 ms CDML3 APD90=12667 plusmn 4999 ms unpaired t-test Plt0001 RPMI

Triangulation=060plusmn013 CDML3 Triangulation=072plusmn012 unpaired t-test Plt0001 G zero of 94 wells showed

spontaneous arrhythmias in RPMI+B27 monolayers while 2495 monolayers presented with spontaneous

arrhythmias in the CDML3 treated group H-J electrophysiology and calcium flux of iCell2 Cardiomyocytes is affected

by maintenance in CDML3 media APD50 and CaTD50 are prolonged in the CDML3 media treated monolayers

Calcium transient amplitude of CDML3 media treated monolayers plated in 96 well plates is reduced compared to

RPMI-B27 maintenance media K Seahorse measurement of oxygen consumption rate shows reduced maximal

OCR (L) and reduced reserve respiratory capacity in CDML3 treated hiPSC-CM monolayers

Graphical abstract

Title In vitro model of ischemic heart failure using human induced pluripotent stem cell-derived 1

cardiomyocytes 2

Justin Davis1 Ahmad Chouman1 Jeffery Creech1 Andre Monteiro da Rocha1 2 4 Daniela Ponce 3

Balbuena2 Eric Jimenez Vazquez2 Ruthann Nichols3 Andrey Lozhkin4 Nageswara Madamanchi4 4

Katherine Campbell12 Todd J Herron1245 5

1University of Michigan Frankel Cardiovascular Regeneration Core Laboratory Internal Medicine-6

Cardiovascular Medicine Ann Arbor MI 48109 7

2University of Michigan Center for Arrhythmia Research Ann Arbor MI 48109 8

3University of Michigan Department of Biological Chemistry Medical Science Research Building III 9

1150 W Medical Center Drive Ann Arbor MI 48109 10

4University of Michigan Internal Medicine-Cardiology MSRBIII Ann Arbor MI 48109 11

5University of Michigan Department of Molecular and Integrative Physiology Ann Arbor MI 48109 12

Conflict of Interest Statement Jeffery Creech and Andre Monteiro da Rocha are consultants to 13

CARTOX Inc Todd J Herron is co-founder of and has financial interest in CARTOX Inc CARTOX 14

Inc is a Michigan company focused on the development of novel hiPSC-CM based assays for 15

cardiotoxicity screening TJH is also a consultant to StemBioSys Inc 16

Corresponding Author 17

Todd J Herron PhD 18 University of Michigan 19

Frankel Cardiovascular Regeneration Core Laboratory 20 2800 Plymouth Road 21 Building 26 223N 22 Ann Arbor MI 48109-2800 23 toddherrumichedu 24

Abstract 26

Introduction 24

assays(20 21) 20

study 25

Results 2

cardiomyocytes 20

heart failure 4

Discussion 19

Conclusion 3

Methods 19

measurements 14

Acknowledgments 18

References 2

Graphical abstract

Introduction 24

assays(20 21) 20

study 25

Results 2

cardiomyocytes 20

heart failure 4

Discussion 19

Conclusion 3

Methods 19

measurements 14

Acknowledgments 18

References 2

Graphical abstract

assays(20 21) 20

study 25

Results 2

cardiomyocytes 20

heart failure 4

Discussion 19

Conclusion 3

Methods 19

measurements 14

Acknowledgments 18

References 2

Graphical abstract

assays(20 21) 20

study 25

Results 2

cardiomyocytes 20

heart failure 4

Discussion 19

Conclusion 3

Methods 19

measurements 14

Acknowledgments 18

References 2

Graphical abstract

Results 2

cardiomyocytes 20

heart failure 4

Discussion 19

Conclusion 3

Methods 19

measurements 14

Acknowledgments 18

References 2

Graphical abstract

heart failure 4

Discussion 19

Conclusion 3

Methods 19

measurements 14

Acknowledgments 18

References 2

Graphical abstract

heart failure 4

Discussion 19

Conclusion 3

Methods 19

measurements 14

Acknowledgments 18

References 2

Graphical abstract

heart failure 4

Discussion 19

Conclusion 3

Methods 19

measurements 14

Acknowledgments 18

References 2

Graphical abstract

heart failure 4

Discussion 19

Conclusion 3

Methods 19

measurements 14

Acknowledgments 18

References 2

Graphical abstract

heart failure 4

Discussion 19

Conclusion 3

Methods 19

measurements 14

Acknowledgments 18

References 2

Graphical abstract

Discussion 19

Conclusion 3

Methods 19

measurements 14

Acknowledgments 18

References 2

Graphical abstract

Discussion 19

Conclusion 3

Methods 19

measurements 14

Acknowledgments 18

References 2

Graphical abstract

Conclusion 3

Methods 19

measurements 14

Acknowledgments 18

References 2

Graphical abstract

Conclusion 3

Methods 19

measurements 14

Acknowledgments 18

References 2

Graphical abstract

Conclusion 3

Methods 19

measurements 14

Acknowledgments 18

References 2

Graphical abstract

Conclusion 3

Methods 19

measurements 14

Acknowledgments 18

References 2

Graphical abstract

Conclusion 3

Methods 19

measurements 14

Acknowledgments 18

References 2

Graphical abstract

Conclusion 3

Methods 19

measurements 14

Acknowledgments 18

References 2

Graphical abstract

measurements 14

Acknowledgments 18

References 2

Graphical abstract

measurements 14

Acknowledgments 18

References 2

Graphical abstract

measurements 14

Acknowledgments 18

References 2

Graphical abstract

measurements 14

Acknowledgments 18

References 2

Graphical abstract

Acknowledgments 18

References 2

Graphical abstract

References 2

Graphical abstract

Documents

In vitro model of ischemic heart failure using human