key: cord-0758015-if748w2n
authors: Bailey, Adam L.; Dmytrenko, Oleksandr; Greenberg, Lina; Bredemeyer, Andrea L.; Ma, Pan; Liu, Jing; Penna, Vinay; Lai, Lulu; Winkler, Emma S.; Sviben, Sanja; Brooks, Erin; Nair, Ajith P.; Heck, Kent A.; Rali, Aniket S.; Simpson, Leo; Saririan, Mehrdad; Hobohm, Dan; Stump, W. Tom; Fitzpatrick, James A.; Xie, Xuping; Shi, Pei-Yong; Hinson, J. Travis; Gi, Weng-Tein; Schmidt, Constanze; Leuschner, Florian; Lin, Chieh-Yu; Diamond, Michael S.; Greenberg, Michael J.; Lavine, Kory J.
title: SARS-CoV-2 Infects Human Engineered Heart Tissues and Models COVID-19 Myocarditis
date: 2020-11-05
journal: bioRxiv
DOI: 10.1101/2020.11.04.364315
sha: 59d5d4b6002f2f20b4c820238bb9e1a346483c87
doc_id: 758015
cord_uid: if748w2n

Epidemiological studies of the COVID-19 pandemic have revealed evidence of cardiac involvement and documented that myocardial injury and myocarditis are predictors of poor outcomes. Nonetheless, little is understood regarding SARS-CoV-2 tropism within the heart and whether cardiac complications result directly from myocardial infection. Here, we develop a human engineered heart tissue model and demonstrate that SARS-CoV-2 selectively infects cardiomyocytes. Viral infection is dependent on expression of angiotensin-I converting enzyme 2 (ACE2) and endosomal cysteine proteases, suggesting an endosomal mechanism of cell entry. After infection with SARS-CoV-2, engineered tissues display typical features of myocarditis, including cardiomyocyte cell death, impaired cardiac contractility, and innate immune cell activation. Consistent with these findings, autopsy tissue obtained from individuals with COVID-19 myocarditis demonstrated cardiomyocyte infection, cell death, and macrophage-predominate immune cell infiltrate. These findings establish human cardiomyocyte tropism for SARS-CoV-2 and provide an experimental platform for interrogating and mitigating cardiac complications of COVID-19.

To explore whether human cardiomyocytes might be susceptible to SARS-CoV-2 136 infection, we examined the expression of angiotensin converting enzyme 2 (ACE2) within the 137 human heart. Previous studies have established that ACE2 serves as a cell-surface receptor for 138 SARS-CoV-2 through interactions with the spike protein in numerous human cell types 15, 16 . 139

Immunostaining of human left ventricular myocardial tissue revealed evidence of ACE2 140 expression in cardiomyocytes (Fig. 1a) . We observed significant variation in ACE2 expression 141 between individual cardiomyocytes within the same myocardial specimen. ACE2 mRNA was 142 abundantly expressed in the healthy human heart and further increased in the context of chronic 143 heart failure (Fig. 1b) . RNA sequencing of human pediatric and adult heart failure specimens 144 revealed robust expression of ACE2 mRNA within the human heart across the spectrum of age 145 (Fig. 1c) . Consistent with our immunostaining findings, primary human cardiomyocytes obtained 146 from the left ventricle and atria expressed ACE2 mRNA (Fig. 1d) . These data are consistent with 147 prior single cell and bulk RNA sequencing analyses of human myocardium and suggest that 148 cardiomyocytes might be permissive to SARS-CoV-2 infection 17 . 149

To ascertain whether human pluripotent stem cell-derived cardiomyocytes (hPSC-derived 150 CMs) can serve as an appropriate model to study cardiac SARS-CoV-2 infection, we measured 151 ACE2 mRNA expression in hPSC-derived CMs. Quantitative RT-PCR revealed that hPSC-152 derived CMs abundantly expressed ACE2 mRNA. In contrast, minimal ACE2 mRNA was detected 153 in human dermal fibroblasts, hPSC-derived cardiac fibroblasts, or human fetal cord blood derived-154 macrophages (Fig. S1a-c) . Human engineered heart tissues (EHTs) self-assembled between two 155 deformable polydimethylsiloxane (PDMS) posts after mixing cells in an extracellular matrix 156 composed of collagen and matrigel (Fig. S1d) . EHTs composed of either hPSC-derived CMs and 157 fibroblasts or hPSC-derived CMs, fibroblasts, and macrophages also expressed ACE2 mRNA 158 (Fig. S1e) . Immunostaining of EHTs confirmed the presence of ACE2 protein specifically in 159 hPSC-derived CMs (Fig. S1f) . These data suggest that hPSC-derived CMs might be susceptible 160 to SARS-CoV-2 infection and serve as a suitable experimental model to study cardiac 161 manifestations of COVID- 19 . (Fig. S1g) . In contrast, two independent lines of hPSC-derived cardiomyocytes (hPSC-171 derived CMs) were permissive to SARS-CoV-2 infection (Fig. 1f) . Undifferentiated hPSC lines did 172 not demonstrate evidence of infection (Fig S3) . 173

To confirm cardiomyocyte tropism, we inoculated various combinations of hPSC-derived 174 CMs, fibroblasts, and macrophages grown in monolayer culture with wild-type SARS-CoV-2 175 (USA_WA1/2019). We analyzed tissue culture supernatants for production of infectious virus 176 using a Vero cell infection-based focus forming assay, and we measured intracellular viral RNA 177 transcript levels using RT-qPCR at 3 days post-inoculation. These assays revealed the production 178 of infectious virus (Fig. 2a) and viral RNA (Fig. 2b) selectively in cultures that contained hPSC-179 derived CMs. Cultures lacking hPSC-derived CMs contained viral loads that were equivalent to 180 media-only controls. A time course of hPSC-derived CM infection showed that cardiomyocytes 181 rapidly produced infectious virus with peak titers observed at day 3 post-inoculation. These 182 kinetics were closely mirrored by SARS-CoV-2-mNeonGreen (Fig. 2c) . 183

Using SARS-CoV-2-mNeonGreen, we examined the relationship between viral replication 184 and cell death using flow cytometry. Although the percentage of hPSC-derived CMs that were 185 mNeonGreen-positive peaked at day 3 post-inoculation, significant levels of hPSC-derived CM 186 cell death were not observed until 4-5 days post-inoculation (Fig. 2d) indicating that viral infection 187 precedes hPSC-derived CM cell death. SARS-CoV-2-infected cardiomyocytes also displayed 188 characteristics of cytopathic effects. Cellular rounding, clumping, and syncytium formation first 189 were observed on day 3 post-inoculation. Distortion of cellular morphology was evident by day 4 190 post-inoculation and cultures contained largely dead cells and debris by days 5-6 post-inoculation 191 (Fig. 2e) . 192

To verify that cardiomyocytes are the primary target of SARS-CoV-2 in a simulated cardiac 193 environment, we infected two-dimensional tissues assembled with hPSC-derived CMs (80%), 194 fibroblasts (10%), and macrophages (10%) with SARS-CoV-2-mNeonGreen. Flow cytometry 195 performed 3 days following infection revealed mNeonGreen expression only in CD90 -CD14 -196 TNNT2 + cardiomyocytes. mNeonGreen was not detected in CD90 + fibroblasts or CD14 + 197 macrophages within infected two-dimensional tissues ( Fig. 2f-g, Fig. S4 ). These data To examine viral transcription and the host immune response to SARS-CoV-2 infection, 210

we performed RNA sequencing. Cultures containing either hPSC-derived CMs, fibroblasts, or 211 macrophages were either mock-infected or inoculated with SARS-CoV-2. We also examined two-212 dimensional co-culture tissues assembled with 80% cardiomyocytes, 10% fibroblasts, and 10% 213 macrophages. Cells and tissues were harvested on day 3 post-inoculation. Principal component 214 analysis revealed separation between experimental groups consistent with their distinct cellular 215 composition (Fig. 3a) . Classification of transcript types demonstrated that infected hPSC-derived 216

CMs and two-dimensional tissues comprised of hPSC-derived CMs and fibroblasts or hPSC-217 derived CMs, fibroblasts, and macrophages contained abundant viral transcripts (Fig. S5a) . We 218 then assessed the expression of specific viral transcripts by aligning the RNA sequencing data to 219 the SARS-CoV-2 genome and transcriptome. Subgenomic RNAs were identified based on the 220 presence of 5' leader sequences 21 . We observed robust expression of most SARS-CoV-2 221 genomic and subgenomic RNAs in infected hPSC-derived CMs and two-dimensional tissues with 222 the exception of ORF7b (Fig. 3b, Fig. S5b) . 223

To facilitate differential expression analysis of host genes, we censored viral RNAs from 224 the RNA sequencing computational model. This was necessary given the asymmetric prevalence 225 of viral transcripts across samples. We identified numerous host genes that were differentially 226 regulated upon SARS-CoV-2 infection in each of the examined cell types and two-dimensional 227 tissues (Fig. 3c) . Conditions that supported viral replication (hPSC-derived CMs and two-228 dimensional tissues) displayed the greatest overlap in differentially expressed genes. Cell types 229 that did not support viral replication (fibroblasts and macrophages) also demonstrated numerous 230 differentially expressed host genes, indicating that SARS-CoV-2 might elicit changes in host gene 231 expression in the absence of direct viral infection. Notably, host genes differentially expressed in 232 fibroblasts and macrophages exposed to SARS-CoV-2 were largely distinct (Fig. 3d) . These 233 findings suggest that elements within or on the surface of SARS-CoV-2 virions may serve as 234 pathogen-associated molecular patterns (PAMPs) and stimulate distinct gene expression 235 programs in differing cell types. 236 GO pathway analysis revealed that infected hPSC-derived CMs and two-dimensional co-237 culture tissues showed upregulation of genes associated with immune cell activation, stress-238 induced transcription, and responses to pathogens including viruses. Genes associated with 239 metabolism, oxidative phosphorylation, and mitochondrial function were downregulated by 240 infection. Two-dimensional tissues displayed alterations in other pathways including upregulation 241 of cellular responses to cytokines and downregulation of genes involved in muscle contraction 242 ( Fig. 3e-f) . Host genes differentially expressed in macrophages and fibroblasts were associated 243 with pathways involved in innate immune cell activation, migration, and cytokine responses (Fig.  244 3g-h). 245

Examination of specific genes differentially regulated in infected hPSC-derived CMs and 246 two-dimensional tissues (Fig. 3i ) revealed marked reduction in components of the electron 247 transport chain (ATP synthase, mitochondrial cytochrome C oxidase, NADPH dehydrogenase) 248

and key upstream metabolic regulators (glycerol-3-phosphate dehydrogenase, pyruvate 249 dehydrogenase, succinate dehydrogenase complex). PDK4, an inhibitor of pyruvate 250 dehydrogenase was upregulated in infected hPSC-derived CMs and two-dimensional tissues. We 251 also observed marked downregulation of numerous components and regulators of the contractile 252 apparatus including cardiac actin, troponin subunits, myosin light and heavy chains, desmin, 253 phospholamban, and calsequestrin in infected two-dimensional tissues. Infected hPSC-derived 254

CMs displayed similar changes, albeit to a lesser extent. ACE2 expression was diminished in 255 infected cardiomyocytes and two-dimensional tissues. 256

Infected hPSC-derived CMs and two-dimensional tissues also displayed upregulation of 257 key regulators of innate immunity (Fig. 3i) . Type I interferon (IFN) activation was apparent by the 258 increased expression of IFNB1 and numerous IFN stimulated genes including IFIT1, IFIT2, IFIT3, 259 ISG15, MX1, and OAS1. Stress response programs (FOS) and cytokine expression (TNF) were 260 similarly upregulated in these cell types. Consistent with a greater innate immune response in 261 two-dimensional tissues, we found that several chemokines (CCL3, CCL4, CCL7, CCL8, and 262 CXCL8) and cytokines (IL1B, IL6, and CSF3) were selectively upregulated in infected two-263 dimensional tissues. Macrophages and fibroblasts contributed to enhanced chemokine and 264 cytokine responses in two-dimensional tissues. CCL3, CCL4, and CCL8 were selectively 265 expressed in infected macrophages and CSF3, CXCL8, IL1B, and IL6 were induced in infected 266 fibroblasts (Fig. S5c) . inhibitor of the SARS-CoV-2 RNA-dependent RNA polymerase 22-24 ( Fig. 4a-b) . 277

After binding to ACE2, the SARS-CoV (and SARS-CoV-2) spike protein must undergo 278 proteolytic activation to initiate membrane fusion 25 . Host proteases located at the plasma 279 membrane (TMPRSS2) or within endosomes (cathepsins) most commonly perform this function. 280

The relative contributions of each of these protease families to SARS-CoV-2 infection varies by 281 cell-type 15,25 . RNA sequencing data revealed that hPSC-derived CMs express robust levels of 282 ACE2 and multiple endosomal proteases including cathepsins and calpains (Fig. 4c) . ACE2 283 mRNA was not abundantly expressed in either macrophages or fibroblasts. While TMPRSS2 284 expression was present at the lower limit of detection for RNAseq, we detected low, but 285 measurable levels of TMPRSS2 by RT-qPCR in hPSC-derived CMs, but not in fibroblasts or 286 macrophages ( Fig. 4c-d) . 287

To determine whether SARS-CoV-2 enters cardiomyocytes through an endosomal or 288 plasma membrane route, we inoculated hPSC-derived CMs with SARS-CoV-2-mNeonGreen and 289 administered either the endosomal cysteine protease inhibitor E-64, which blocks cathepsins, or 290 the serine protease inhibitor camostat mesylate, which blocks TMPRSS2 (and possibly 291 TMPRSS4) 25 . Notably, E-64 abolished SARS-CoV-2 infection of hPSC-derived CMs as 292 demonstrated by reduced mNeonGreen expression and viral RNA within the supernatant (Fig.  293 4e-f). In contrast, camostat had no effect on cardiomyocyte infection over a range of doses ( Fig.  294 4g-h). Thus, SARS-CoV-2 enters cardiomyocytes through an endosomal pathway that requires 295 cathepsin but not TMPRSS2-mediated cleavage. 296 297

Myocarditis is characterized by direct viral infection of cardiomyocytes and accumulation 299 of immune cells at sites of active infection or tissue injury 26,27 . To examine whether SARS-CoV-2 300 infection of cardiomyocytes in a three-dimensional environment mimics aspects of viral 301 myocarditis, we generated EHTs containing either hPSC-derived CMs and fibroblasts or hPSC-302 derived CMs, fibroblasts, and macrophages. EHTs were seeded in a collagen-Matrigel matrix 303 between two PDMS posts, infected with SARS-CoV-2, and harvested 5 days after inoculation. 304

Hematoxylin and eosin (H&E) staining revealed evidence of tissue injury and increased interstitial 305 cell abundance within the periphery of SARS-CoV-2-infected EHTs (Fig. 5a) . Immunostaining for 306 the viral nucleocapsid protein demonstrated evidence of prominent infection at the periphery of 307 the EHT. Nucleocapsid staining was localized within hPSC-derived CMs. Staining for CD68 308 demonstrated macrophage accumulation corresponding to sites of interstitial cell accumulation 309 and viral infection (Fig. 5b, Fig. S6 ). Enrichment of nucleocapsid staining at the periphery of the 310 tissue suggests that viral diffusion might be limited by the three-dimensional EHT environment. 311

Consistent with our immunostaining results, infected EHTs (with and without macrophages) 312 accumulated high levels of viral RNA, as detected by quantitative RT-PCR (Fig. 5c) . In situ 313 hybridization for viral spike sense and antisense RNA was also indicative of viral replication within 314 EHTs (Fig. 5d, Fig. S7) . EHTs consisting of hPSC-derived CMs and fibroblasts were assembled and allowed to 325 mature for 7 days prior to infection. EHTs were inoculated with SARS-CoV-2, and contractile 326 function was analyzed daily. From days 0 to 3 post infection, the average maximal displacement 327 generated during beating did not differ between the mock and SARS-CoV-2-infected tissues. 328

However, on days 4 to 5 post infection, the SARS-CoV-2 inoculated tissues showed reduced 329 contraction relative to the mock-infected tissues ( Fig. 5e-f ). On day 5 after inoculation, the 330 maximal displacement produced during contraction by the SARS-CoV-2 inoculated tissues was 331 markedly lower than mock infected-tissues. Moreover, the tissues show reduced speed of 332 contraction and relaxation, consistent with systolic dysfunction (Fig. 5g) . 333 334

To examine whether cardiomyocyte cell death might serve as a mechanism explaining 336 reduced EHT contractility on days 4 to 5 post inoculation, we performed TUNEL staining. 337

Consistent with the temporal course of SARS-CoV-2 cardiomyocyte infection and cell death in 338 our two-dimensional hPSC-CM cultures (Fig. 2d) , we observed increased numbers of TUNEL 339 positive cardiomyocytes in SARS-CoV-2 infected EHTs on day 5 post infection ( Fig. 6a-b) . Our 340 RNA sequencing data suggest that other mechanisms also may contribute to reduced EHT 341 contractility, including decreased expression of genes important for sarcomere function and 342 metabolism as well as activation of host immune responses (Fig. 3i) . Consistent with the 343 possibility that disrupted sarcomere gene expression might contribute to reduced EHT 344 contractility, immunostaining of hPSC-derived CMs infected with SARS-CoV-2 revealed evidence 345 of sarcomere loss 3 days following infection (Fig. 6c) , a time point that preceded cell death. 346

Furthermore, immunostaining of EHTs demonstrated loss of Troponin T expression in infected 347 cardiomyocytes. (Fig. 6d-e) . Thus, the reduction in contractile function may be multifactorial with 348 contributions from virus-induced cardiomyocyte cell death and loss of sarcomere elements. 349

We then examined the mechanistic relationship between cardiomyocyte infection, 350 inflammatory signaling, sarcomere breakdown, and cell death. Inhibition of viral entry (ACE2 351 neutralizing antibody) or viral replication (remdesivir) was sufficient to prevent type I IFN and TNF 352 expression following SARS-CoV-2 infection ( Fig. 6f-g) . Remdesivir similarly reduced 353 inflammatory gene expression in 3D EHTs ( Fig. S8a-b) . These data establish that viral infection 354 represents the upstream driver of inflammation in our model system. 355

To examine the impact of cardiomyocyte inflammatory signaling on cardiomyocyte cell 356 death, sarcomere gene expression, and sarcomere structure, we focused on inhibiting viral 357 nucleic acid sensing. TBK1 (TANK-binding kinase 1) is an essential mediator of numerous nucleic 358 acid sensing pathways including RIG-I, MAVS, STING, and TLRs 29,30 . Inhibition of TBK1 activity 359 was sufficient to reduce type I IFN activity (primary inflammatory signature identified in infected 360 cardiomyocytes, Fig 3i) without impacting viral load or cardiomyocyte infectivity ( Fig. 6f-h) . 361

Inhibition of TBK1 activity during SARS-CoV-2 cardiomyocyte infection had no impact on 362 cardiomyocyte cell death (Fig. 6i) . While TBK1 inhibition prevented reductions in TNNT2 and 363 MYH7 mRNA expression following cardiomyocyte SARS-CoV-2 infection, sarcomere breakdown 364 remained prevalent in infected cardiomyocytes treated with the TBK1 inhibitor. In contrast, 365 remdesivir prevented both reductions in TNNT2 and MYH7 mRNA expression and sarcomere 366 loss following SARS-CoV-2 infection ( Fig. 6j-k, Fig. S8c-d) . These data indicate that SARS-CoV-367 2 elicits an inflammatory response in cardiomyocytes that is at least partially dependent on viral 368 nucleic acid sensing and TBK1 signaling. However, TBK1-dependent cardiomyocyte 369 inflammation does not appear responsible for sarcomeric disassembly or cardiomyocyte cell 370 death. These findings do not rule out the possibility that other inflammatory pathways or cross-371 talk between infected cardiomyocytes and immune cells contributes to reduced EHT contractility. 372 373

To validate the myocarditis phenotype generated by SARS-CoV-2 infection of the EHT 375 model, we obtained autopsy and endomyocardial biopsy specimens from four subjects with 376 confirmed SARS-CoV-2 infection and clinical diagnoses of myocarditis. Evidence of myocardial 377 injury (elevated troponin) and left ventricular systolic dysfunction were present in each case 378 (Table 1) were noted accompanied by a mixed mononuclear cell infiltrate (Fig. 7a) . These changes are 384 distinct from postmortem autolytic changes. Examination of the coronary arteries from the COVID-385 19 myocarditis autopsy cases demonstrated non-obstructive mild atherosclerotic changes, 386 consistent with the angiogram findings. There was no evidence of microvascular injury or 387 thromboembolic events. Two autopsy heart samples from subjects with metastatic carcinoma and 388 an inherited neurodegenerative disease with similar tissue procurement times were included as 389 negative controls. 390 RNA in situ hybridization for SARS-CoV-2 spike and nucleocapsid genes revealed 391 evidence of viral RNA within the myocardium of each COVID-19 myocarditis subject. Viral 392 transcripts were located in cytoplasmic and perinuclear locations within cells that were 393 morphologically consistent with cardiomyocytes (Fig. 7b, Fig. S9a-b) . Viral transcripts also were 394 identified in airway epithelial cells within the lung of this subject and other myocardial cell types 395 including perivascular adipocytes and pericytes (Fig. S9c) . Immunostaining for the nucleocapsid 396 protein further demonstrated presence of viral protein in cardiomyocytes Fig. 7c) . The COVID-19 397 myocarditis immune cell infiltrate was characterized by accumulation of a mixed population of 398 CCR2and CCR2 + macrophages within injured areas of the myocardium (Fig. 7d) . Minimal 399 evidence of T-cell infiltration was noted (Fig. 7e) . Macrophage abundance was highest in areas 400 that demonstrated evidence of cardiomyocyte injury as depicted by complement deposition (C4d 401 staining), a pathological marker of cardiomyocyte cell death 31-33 (Fig. S9d) . Together, these 402 observations provide initial pathological evidence that SARS-CoV-2 infects the human heart and 403 may contribute to cardiomyocyte cell death and myocardial inflammation that is distinct from 404 lymphocytic myocarditis. (4 mg/mL), 10% FBS, 1% non-essential amino acids, 1% GlutaMAX Supplement, and 1% Pen-682

Strep. All drug compounds were purchased from Selleckchem (ruxolitinib, catalog number S8932; 683 MRT67307, catalog number S7948; E64, catalog number S7379; camostat, catalog number 684 S2874) and resuspended to a stock concentration of 10μM in PBS or DMSO (depending on the 685 solubility profile), then diluted to working concentration in culture media (described above) and 686 sterile-filtered. 687 688

Immunostaining was performed as previously described with a few modifications 56 . Briefly, 690 cardiomyocytes were fixed for 20 minutes in 4% formaldehyde in phosphate buffered saline 691 (PBS). Cells were then permeabilized with 0.4% Triton X-100 for 20 minutes at room temperature. 692

The cells were blocked for 1 hour using a blocking solution containing 3% bovine serum albumin, 693 5% donkey serum, 0.1% Triton X-100, and 0.02% sodium azide in PBS. Primary antibodies (rabbit 694 anti Troponin T, 1:400, Abcam, ab45932) were added for 1-2 hours at room temperature or 695 overnight at 4 °C. Cells were then washed with PBS before incubating for 1 hour in secondary 696 antibody (Cy3 donkey anti-rabbit, Jackson Immunoresearch, 711165152). 4′,6-diamidino-2-697 phenylindole (DAPI) was used at a 1:50000 dilution to stain for nuclei. Cells were visualized using 698 a Nikon A1Rsi confocal microscope (Washington University Center for Cellular Imaging). Z-699 stacks of cells with 40x magnification were recorded in sequential scanning mode. Images were 700 processed in ImageJ and Z-stacks were converted to standard deviation projections 61 . 701 to test for changes in expression of the reported log 2 fold-changes reported by Limma in each 806 term versus the background log 2 fold-changes of all genes found outside the respective term. 807

The R/Bioconductor package heatmap3 71 was used to display heatmaps across groups of 808 samples for each GO or MSigDb term with a Benjamini-Hochberg false-discovery rate adjusted 809 p-value less than or equal to 0.05. Perturbed KEGG pathways where the observed log 2 fold-810 changes of genes within the term were significantly perturbed in a single-direction versus 811 background or in any direction compared to other genes within a given term with p-values less 812 than or equal to 0.05 were rendered as nnotated KEGG graphs with the R/Bioconductor package 813

To find the most critical genes, the raw counts were variance stabilized with the R/Bioconductor 815 package DESeq2 54 and then analyzed via weighted gene correlation network analysis with the 816 R/Bioconductor package WGCNA 73 . Briefly, all genes were correlated across each other by 817 Pearson correlations and clustered by expression similarity into unsigned modules using a power 818 threshold empirically determined from the data. An eigengene was created for each de novo 819 cluster and its expression profile was then correlated across all coefficients of the model matrix. 820

Because these clusters of genes were created by expression profile rather than known functional 821 similarity, the clustered modules were given the names of random colors where grey is the only 822 module that has any pre-existing definition of containing genes that do not cluster well with others. 823

These de novo clustered genes were then tested for functional enrichment of known GO terms 824 with hypergeometric tests available in the R/Bioconductor package clusterProfiler 74 . Significant 825 terms with Benjamini-Hochberg adjusted p-values less than 0.05 were then collapsed by similarity 826 into clusterProfiler category network plots to display the most significant terms for each module 827 of hub genes in order to interpolate the function of each significant module. The information for 828 all clustered genes for each module were combined with their respective statistical significance 829 results from Limma to identify differentially expressed genes. Figure 1 : ACE2 is expressed in the human heart and in stem cell derived cardiomyocytes.

Clinical course and risk factors for mortality of adult inpatients with COVID-19

China: a retrospective cohort study

Association of Cardiac Injury With Mortality in Hospitalized Patients With 880

COVID-19 and Cardiac Arrhythmias

Cardiac Involvement in Patients Recovered From

Using Magnetic Resonance Imaging

Outcomes of Cardiovascular Magnetic Resonance Imaging in 887

COVID-19)

Cardiovascular Magnetic Resonance Findings in Competitive Athletes 890

Recovering From COVID-19 Infection

Elevated Troponin in Patients With Coronavirus Disease

Animal models of mechanisms of SARS-CoV-2 infection and COVID-19 895 pathology

Functional assessment of cell entry and receptor usage 897 for SARS-CoV-2 and other lineage B betacoronaviruses

Coronavirus from Wuhan: an Analysis Based on Decade-Long Structural Studies of SARS 900

Engineering Approaches to Modeling the Mechanics of Human Heart Failure for Drug 903

Engineering Cardiac Muscle Tissue: A 905

Maturating Field of Research

Cardiomyocyte maturation: advances in knowledge and implications for 907 regenerative medicine

SARS-CoV-2 Cell Entry Depends on ACE2 and TMPRSS2 and Is 909

The pathogenicity of SARS-CoV-2 in hACE2 transgenic mice

The ACE2 expression in human heart 913 indicates new potential mechanism of heart injury among patients infected with SARS-CoV-914 2

An Infectious cDNA Clone of SARS-CoV-2

Ultrastructural characterization of SARS coronavirus

Electron 920 microscopy of SARS-CoV-2: a challenging task

The Architecture of SARS-CoV-2 Transcriptome

SARS-CoV-2 infects and induces cytotoxic effects in human 950 cardiomyocytes

Myocardial localization of coronavirus in COVID-19 cardiogenic shock

Multiorgan and Renal Tropism of SARS-CoV-2

Association of Cardiac Infection With SARS-CoV-2 in Confirmed

Autopsy Cases

SARS-CoV-2 productively infects human gut enterocytes

Inhibition of SARS-CoV-2 Infections in Engineered Human Tissues Using 960

A Human Pluripotent Stem Cell-based Platform to Study SARS-CoV-2

Tropism and Model Virus Infection in Human Cells and Organoids

Marked Up-Regulation of ACE2 in Hearts of Patients With Obstructive 965

Hypertrophic Cardiomyopathy: Implications for SARS-CoV-2-Mediated COVID-19

Angiotensin-converting enzyme 2 is an essential regulator of heart 968 function

HEART DISEASE. Titin mutations in iPS cells define sarcomere 970 insufficiency as a cause of dilated cardiomyopathy

Comparison of the effects of a truncating and a missense MYBPC3 972 mutation on contractile parameters of engineered heart tissue

Generation of Quiescent Cardiac Fibroblasts From Human Induced 975

Pluripotent Stem Cells for In Vitro Modeling of Cardiac Fibrosis

Efficient differentiation of human pluripotent stem cells to endothelial 978 progenitors via small-molecule activation of WNT signaling

Phosphomimetic cardiac myosin-binding protein C partially rescues a 981 cardiomyopathy phenotype in murine engineered heart tissue

Increased Afterload Augments Sunitinib-Induced Cardiotoxicity in an 983

Engineered Cardiac Microtissue Model

TMPRSS2 and TMPRSS4 promote SARS-CoV-2 infection of human small 985 intestinal enterocytes

Infection of bat and human intestinal organoids by SARS-CoV-2

COVID-19 ARTIC v3 Illumina library construction and sequencing 989 protocol v4 (protocols.io.bgxjjxkn)

Human monoclonal antibody combination against SARS coronavirus: 991 synergy and coverage of escape mutants

Growth, detection, 993 quantification, and inactivation of SARS-CoV-2

Moderated estimation of fold change and dispersion for 995 RNA-seq data with DESeq2

Epigenome-Wide Association Study Identifies Cardiac Gene Patterning and 997

a Novel Class of Biomarkers for Heart Failure

Disrupted mechanobiology links the molecular and cellular 999 phenotypes in familial dilated cardiomyopathy

Robust cardiomyocyte differentiation from human pluripotent stem cells via 1002 temporal modulation of canonical Wnt signaling

Directed cardiomyocyte differentiation from human pluripotent stem cells by 1005

modulating Wnt/β-catenin signaling under fully defined conditions

Derivation of highly purified cardiomyocytes from human induced 1008 pluripotent stem cells using small molecule-modulated differentiation and subsequent 1009 glucose starvation

Directed Differentiation of Primitive and Definitive 1011

Hematopoietic Progenitors from Human Pluripotent Stem Cells

Fiji: an open-source platform for biological-image analysis

Differentiation of cardiomyocytes and generation of human engineered 1016 heart tissue

STAR: ultrafast universal RNA-seq aligner

featureCounts: an efficient general purpose program for 1020 assigning sequence reads to genomic features

Salmon provides fast and 1022 bias-aware quantification of transcript expression

RSeQC: quality control of RNA-seq experiments

edgeR: a Bioconductor package for 1026 differential expression analysis of digital gene expression data

limma powers differential expression analyses for RNA-sequencing and 1029 microarray studies

Why weight? Modelling sample and observational level variability improves 1031 power in RNA-seq analyses

GAGE: generally 1033 applicable gene set enrichment for pathway analysis

Advanced heat map and clustering analysis using 1035 heatmap3

Pathview: an R/Bioconductor package for pathway-based data 1037 integration and visualization

WGCNA: an R package for weighted correlation network 1039 analysis

clusterProfiler: an R package for comparing 1041 biological themes among gene clusters

RNAscope 2.5 HD Detection Reagent -RED

The authors would like to acknowledge funding support from the National Institutes of Health 841

Anti-human CD31 BV421 Biolegend RRID: AB_2563810 a, Immunohistochemistry of human heart tissue showing ACE2 (red) expression in cardiomyocytes (green, sarcomeric actin). Representative images from 5 analyzed specimens. b, RNA sequencing demonstrating ACE2 mRNA expression in myocardial biopsies obtained from adult controls and heart failure patients. Data is displayed as counts per million (CPM). n=18 pediatric, n=33 adult. Each data point indicates an individual sample. n=60 controls, n=35 heart failure. c, RNA sequencing demonstrating ACE2 mRNA expression in adult and pediatric heart tissue. Data is displayed as counts per million (CPM 

Insets are high magnification images of the boxed areas. Representative images from 4 independent samples. b, Immunostaining of mock or SARS-CoV-2 infected three-dimensional EHTs for sarcomeric actin (cardiomyocytes, red), CD68 (macrophages, green), and nucleocapsid protein (white). EHTs were harvested 5 days after inoculation. Blue: DAPI. Images are representative of 4 independent experiments. Representative images from 4 independent samples. c, Quantitative RT-PCR of SARS-CoV-2 N gene expression in EHTs consisting of hPSC-derived cardiomyocytes (CM) and fibroblasts (Fb) or hPSC-derived cardiomyocytes, fibroblasts, and macrophages. EHTs were either mock infected or inoculated with SARS-CoV-2 (MOI 0.1) and harvested 5 days after inoculation. Each data point represents individual samples/experiments. Error bars denote standard error of the mean. Bar height represents sample mean. Dotted line: limit of detection. *p<0.05 compared to uninfected control (mock, Mann-Whitney test). d, In situ hybridization for SARS-CoV-2 ORF1ab RNA sense and anti-sense strands (red) in EHTs 5 days after mock or SARS-CoV-2 infection (MOI 0.1). Hematoxylin: blue. Representative images from 4 independent specimens. Insets are high magnification images of the boxed areas. e, Representative spontaneous beating displacement traces for an infected and an uninfected EHT on day 5 post-infection. Videos used to generate these traces can be found in Supplemental Videos 1 and 2. f, Displacement (relative to uninfected mock condition) generated by spontaneous beating of EHTs as a function of time following inoculation with SARS-CoV-2 (MOI 0.1). Each data point represents a mean value from 4-7 independent samples (2 independent experiments), error bars denote standard error of the mean. g, Quantification of absolute displacement (left) and contraction speed (right) generated by spontaneous beating of EHTs 5 days following mock or SARS-CoV-2 infection (MOI 0.1). Each data point denotes an individual EHT, bar height corresponds to mean displacement, error bars represent standard error of the mean, *p<0.05 compared to mock (Mann-Whitney test). Figure 6 . Mechanisms of reduced EHT contractility. a, Combined immunostaining for cardiomyocytes (cardiac actin, green) and TUNEL staining (red) of EHTs (CM+Fb+Mac) 5 days after mock or SARS-CoV-2 infection (MOI 0.1). DAPI: blue. Representative images from 4 independent experiments. b, Quantification of cell death (percent of TUNEL-positive cells) in areas of viral infection. Each data point denotes an individual EHT, bar height corresponds to the mean, error bars represent standard error of the mean, *p<0.05 compared to mock (Mann-Whitney test). c, Immunostaining of hPSC-derived cardiomyocytes for Troponin T (red) 3 days after inoculation with mock control or SARS-CoV-2-NeonGreen (MOI 0.1). Blue: DAPI. Arrows denote areas of sarcomere disassembly. d, Immunostaining of EHTs for Troponin T (red) and SARS-CoV-2 nucleocapsid (green) 5 days after inoculation with mock control or SARS-CoV-2-NeonGreen (MOI 0.1). Blue: DAPI. Arrows denote SARS-CoV-2 nucleocapsid positive cells with reduced Troponin T staining. e, Quantification of Troponin T staining in mock (white) and SARS-CoV-2 (red) infected EHTs. NP: nucleocapsid. Data is presented as mean florescence intensity (MFI). MFI was measured in infected (NP+) cardiomyocytes and uninfected (NP-) cardiomyocytes located proximal or remote to areas of infection. Each data point denotes an individual EHT, bar height corresponds to the mean, error bars represent standard error of the mean, *p<0.05 compared to mock (Mann-Whitney test). f, Quantitative RT-PCR measuring OAS1, MX1, and TNF mRNA expression in hPSC-derived cardiomyocytes 3 days after inoculation with mock control (white) or SARS-CoV-2 (green, MOI 0.1). Cells were treated with vehicle, ACE2 antibody (ACE2 Ab) (20µg/ml), remdesivir (10µM), or TBK inhibitor (MRT67307, 10µM). Each data point denotes a biologically unique sample, bar height corresponds to the mean, and error bars indicate standard error of the mean. * p<0.05 compared to mock control. g, Quantitative RT-PCR of SARS-CoV-2 N gene expression in hPSC-derived cardiomyocytes that were either mock infected (white) or inoculated with SARS-CoV-2 (green, MOI 0.1) and harvested 3 days after inoculation. Cells were treated with vehicle, ACE2 Ab (20µg/ml), remdesivir (10µM), or TBK1 inhibitor (MRT67307, 10µM). Each data point represents individual samples. Error bars denote standard error of the mean. Bar height represents sample mean. Dotted line: limit of detection. *p<0.05 compared to uninfected control. ***p<0.05 compared to uninfected control and vehicle infected (mock, Mann-Whitney test). h-i, Flow cytometry measuring the percent of infected (h) and viable (i) hPSCderived cardiomyocytes following either mock infection (white) or inoculation with SARS-CoV-2 (green, MOI 0.1). Cells were harvested and analyzed 3 days after inoculation. Cells were treated with vehicle, remdesivir (10µM), or TBK1 inhibitor (MRT67307, 10µM). Each data point represents individual samples. Error bars denote standard error of the mean. Bar height represents sample mean. *p<0.05 compared to uninfected control. **p<0.05 compared to vehicle infected (mock, Mann-Whitney test). j, Quantitative RT-PCR measuring TNNT2 mRNA expression in hPSC-derived cardiomyocytes 3 days after inoculation with mock control (white) or SARS-CoV-2 (green, MOI 0.1). Cells were treated with vehicle, ACE2 Ab (20µg/ml), remdesivir (10µM), or TBK inhibitor (MRT67307, 10µM). Each data point denotes a biologically unique sample, bar height corresponds to the mean, and error bars indicate standard error of the mean. * p<0.05 compared to mock control. k, Immunostaining of hPSC-derived cardiomyocytes for Troponin T (red) 3 days after inoculation with mock control or SARS-CoV-2-NeonGreen (MOI 0.1). hPSC-derived cardiomycoytes were treated with vehicle, remdesivir (10µM) or TBK inhibitor (MRT67307, 10µM). Blue: DAPI. Arrows denote areas of sarcomere disassembly. Merged images can be found in Fig. S8.   Figure 7 . Human autopsy and endomyocardial tissue from patients with suspected COVID-19 myocarditis show evidence of SARS-CoV-2 cardiomyocyte infection. a, Hematoxylin and eosin staining of cardiac autopsy (anterior left ventricular wall) and biopsy samples (right ventricular septum) from subjects without COVID-19 (control case) and patients with a clinical diagnosis of COVID-19 myocarditis (case 1-4) . b, In situ hybridization of cardiac autopsy and biopsy tissue for SARS-CoV-2 spike and nucleocapsid RNA (red) showing evidence of viral infection. Hematoxylin: blue. Arrows denotes viral RNA staining in cells with cardiomyocyte morphology. c, Immunostaining of control and COVID-19 myocarditis cardiac autopsy tissue for SARS-CoV-2 nucleocapsid (white) and cardiac actin (red). DAPI: blue. Arrows denotes nucleocapsid staining in cardiomyocytes .d, Immunostaining of control and COVID-19 myocarditis cardiac autopsy and biopsy tissue for CD68 (green) and CCR2 (red). DAPI: blue. e, Immunostaining of control and COVID-19 myocarditis cardiac autopsy and biopsy tissue for CD3 (brown). Hematoxylin: blue.