SARS-CoV-2 infection leads to cardiac pericyte loss, fibrosis, cardiomyocyte hypertrophy, and diastolic dysfunction

a Centre for Molecular and Vascular Biology, KU Leuven, Herestraat 49, bus 911, 3000 Leuven, Belgium b Laboratory of Virology and Chemotherapy, Department of Microbiology, Immunology and Transplantation, Rega Institute for Medical Research, KU Leuven Herestraat 49, bus 911, 3000 Leuven, Belgium c Department of Cardiology, Maastricht University, CARIM School for Cardiovascular Diseases, Universiteitssingel 50, 6229 ER Maastricht, Netherlands d Berlin Institute of Health Center for Regenerative Therapies (BCRT) Charité University Medicine Berlin Campus Virchow Clinic Berlin Germany. e German Center for Cardiovascular Research (DZHK), partner site Berlin, Germany f ICIN-Netherlands Heart Institute, Holland Heart House, Moreelsepark 1, 3511 EP Utrecht, Netherlands g Department of Cardiology and Pneumology, Charité, University Medicine Berlin, Campus Virchow Klinikum, Berlin, Germany


MAIN
With the number of COVID-19 cases increasing every day, the long-term consequences of SARS-CoV-2 infection, both after clinically mild or more severe manifestations, are not yet understood. SARS-CoV-2 infection has several implications beyond the traditional respiratory syndrome COVID19, including the risk of developing severe cardiac injury, which in turn may exacerbate its severity [1][2][3] . Over 60% of patients hospitalized for COVID19 present with clear signs of myocardial injury 4 , with between 20 and 63% of patients showing increased biomarker levels, depending on the study 2,4,5 . Hospitalized patients present with a range of cardiac dysfunctions, including left ventricular wall motion abnormalities (23.7%), ventricular dysfunction (RV 26.3%, LV18.4%), diastolic dysfunction (13.2%), and pericardial effusions (7.2%) 4 . Additionally, long term consequences on cardiac function are suggested, as evidenced by the presence of cardiac fibrosis 6,7 .
The underlying mechanism by which SARS-CoV-2 infection causes cardiac injury is poorly understood. Cardiac damage could occur indirectly, due to thrombotic events or hyperinflammation targeted against the heart, or directly through viral infection of the heart 8 . In autopsy specimens from patients who died from COVID19, significant amounts of viral RNA were found in the heart 9 . Similarly, in a SARS-CoV-2-infected hamster model, the heart is among the tissues with the highest viral loads, after the respiratory tract 10 .
Internalization of the virus depends on its binding to the angiotensin converting enzyme 2 (ACE2) receptor. In the human heart, ACE2 is mostly expressed by pericytes, and to a lesser extent by cardiomyocytes 11,12 . As pericytes are crucial for the microvascular function and integrity, SARS-CoV-2 infection may induce capillary dysfunction indirectly by attacking pericytes. The combination of microvascular dysfunction, inflammation, and the presence of post-infective cardiac fibrosis highlights the potential of developing diastolic dysfunction after recovering from COVID19. In fact, most patients with COVID19 that suffer cardiac injury present with preserved ejection fraction, but show either right ventricular dilatation (29%) or left ventricular diastolic dysfunction (16%) 13 . The former is likely indirectly caused by effects on the lung, whereas the ventricular dysfunction could be a direct effect of the cardiac SARS-CoV-2 infection itself. As such, COVID19 may be an independent risk factor for the development of heart failure with preserved ejection fraction (HFpEF) 14 .
Here, we use a SARS-CoV-2-infected hamster model to investigate the cardiac effects of COVID19. A reduced cardiac pericyte coverage is present very early after infection.
Additionally, we show that in vitro ACE2 internalization renders pericytes more prone to vasodilatation. Finally, we demonstrate that in the presence of metabolic comorbidities (obesity, hyperlipidemia, and hypertension), SARS-CoV-2-infected hamsters fail to recover their reduced cardiac pericyte coverage and present with fibrosis and cardiomyocyte hypertrophy. Using echocardiography, we demonstrate that in the presence metabolic comorbidities, SARS-CoV-2 infection accelerate the development of diastolic dysfunction. To confirm the relevance for COVID19 patients, we investigated microvascular structure in biopsies from recovered COVID19 patients who had developed cardiac complications. All patients showed reduced pericyte coverage of capillaries, as indicated by αSMA low expression.

SARS-CoV-2 infects cardiac pericytes and causes a reduction in cardiac pericyte coverage in hamsters
To study the effect of SARS-CoV-2 infection on cardiac function, we used a previously described COVID19 hamster model 15,16 . In this model, hamsters are inoculated intranasally with SARS-CoV-2 virus (2x10 6 TCID50), which leads to a peak in lung viral load at 4 days post infection (dpi). To determine whether SARS-CoV-2 infects the heart tissue, we quantified the viral load in the heart, relative to lung and muscle tissue, by using qPCR at 4 dpi. In the heart, the viral load reached a median of ~4 log10 RNA copies per mg tissue, around 3 log10 less compared to the lung, while in the muscle only minimal amounts of viral RNA was detected (Fig. 1a). Titrations of homogenized cardiac tissue confirmed the presence of infectious particles in the heart of around 2 log10 TCID50/mg tissue (Supplemental Fig. 1).
We next investigated the expression of viral RNA in the heart using in situ hybridization at 4 dpi. Our in situ hybridization demonstrates that the SARS-CoV-2 virus was present in cells adjacent to capillaries. This indicates that SARS-CoV-2 replication occurs primarily in the pericytes surrounding the capillaries (Fig. 1b).
Given the presence of virus in the pericytes, we next investigated cardiac microvascular changes in the infected hamster hearts at 4 dpi. No difference in cardiac arteriolar density was observed, as defined by high αSMA expression (Supplemental Fig. 2). In order to identify all cardiac pericyte subsets, we used both PDGFR-β and NG2 as markers. Infected hamsters showed a significantly reduction in the percentage of capillaries with NG2 + pericytes ( Fig. 1c-d), while there was no difference observed in PDGFR-β + pericytes ( Fig. 1cd). Double-positive NG2/PDGFR-β vessels showed a trend (p = 0.0542) towards a reduced pericyte density in the heart (Fig. 1c-d). These changes in cardiac pericyte coverage were not associated with any changes in vascular density at 4 dpi ( Fig. 1c-d).
Pericytes in vitro are more susceptible to pro-relaxation cues after ACE2 internalization As SARS-CoV-2 infection is associated with cardiac pericyte loss in hamsters, we questioned whether it would affect pericyte functioning as well. As such, we simulated SARS-CoV-2 infection by exposing human immortalized pericytes in vitro to recombinant SARS-CoV-2 spike protein (Fig. 2). The SARS-CoV-2 spike protein causes an internalization of ACE2, resulting in a reduced ACE2 surface expression 17 . Exposure of human immortalized pericytes with SARS-CoV-2 spike protein did not induce apoptosis (Fig. 2a, b), but significantly affected pericyte function. We analyzed the expression of genes involved in pericyte contraction and relaxation ( Fig. 2c-f) and their interaction with endothelial cells (Supplemental Fig. 3). We found a two-fold and three-fold increase in gene expression of adenosine A1 receptor and adenosine A2A receptor, respectively (encoded by ADORA1 and ADORA2, Fig. 2c-d). These genes are involved in adenosine-induced pericyte relaxation 18 .
Similarly, transcript levels of prostaglandin E2 receptor 4 (PTGER4), a receptor involved in prostaglandin-induced pericyte relaxation, were upregulated upon SARS-CoV-2 spike protein exposure (Fig. 2e) 19 . We observed no change in the Adrenoceptor Alpha 2A receptor, encoded by ADRA2A, which induces noradrenaline-induced pericyte contraction ( Fig. 2f) 19,20 . In contrast to the strong effects on genes involved in pericyte relaxation, most of the growth factors produced by pericytes sustaining endothelial cell function were only slightly affected by exposure to the SARS-CoV-2 spike protein, with the exception of Heparin-binding EGF-like growth factor (HBEGF), which showed an almost 2-fold increase (Supplemental Fig. 3). These results show that ACE2 internalization as a consequence of SARS-CoV-2 spike protein exposure promotes growth factor production by pericytes slightly, but makes them more sensitive to relaxation cues, possibly causing a dysregulation in microvascular flow patterns.

SARS-CoV-2 infected hamsters do not recover their cardiac pericyte coverage when exposed to metabolic comorbidities
Though our results indicate that cardiac pericytes were lost, and possibly dysfunctional, after SARS-CoV-2 infection, it was unclear whether pericytes could recover post-infection.
Pericytes are also lost in other pathologies, notably diabetes 21 . Here, insulin treatment can partly rescue the reduced pericyte density observed 22 . Thus, pericytes have at least some regenerative capacities. As such, we investigated whether the heart could also recover after SARS-CoV-2 infection (Fig. 3). Since the presence of metabolic comorbidities is associated with both HFpEF and a more severe outcome after COVID19 23,24 , we investigated whether hyperlipidemia and hypertension affects the cardiac phenotype after COVID19 recovery. To this end, infected hamsters and uninfected controls received either high fat diet (HFD) combined with hypertension-inducing Nγ-Nitro-L-arginine methyl ester (L-NAME) or a regular chow diet 25 . Hamsters were put on HFD/L-NAME one week before infection ( Fig.   3a).
At 14 dpi, viral RNA was still detected in the lungs, but had decreased substantially compared to levels present at 4 dpi (Fig. 3b). In the heart, no viral RNA was detected at 14 dpi. Uninfected hamsters on HFD/L-NAME gained weight after being three weeks on diet, while the weight of SARS-CoV-2 infected animals remained stable ( Fig. 3c-d).
In uninfected chow-fed hamsters, approximately 80% of cardiac capillaries are covered by pericytes expressing either PDGFR-β and/or NG2 at 14 dpi ( Fig. 3e-i). Both HFD/L-NAMEtreated uninfected hamsters and SARS-CoV-2-infected hamsters on a chow diet showed a slight but unsignificant reduction in pericyte coverage, compared to uninfected chow-fed animals (Fig 3e-f). Infected animals on HFD/L-NAME, however, presented with a reduced NG2 + pericyte density compared to chow-fed hamsters ( Fig. 3e-i). Approximately 60% of vessels were covered with pericytes, which is similar to the observed level of pericyte coverage of infected hamsters at 4 dpi ( Fig. 1c,d,f). Overall, our results indicate that cardiac pericyte coverage does recover, but is prevented by metabolic comorbidities. When we investigated αSMA coverage and microvascular density, no differences were observed between the groups (Supplemental Fig. 4).
Cardiac damage remaining post-infection could be indicative of future cardiac dysfunction.
We therefore investigated cardiomyocyte hypertrophy and cardiac fibrosis in hamster hearts at 14 dpi (Fig. 4). Cardiomyocytes from SARS-Cov-2 infected hearts were 20% larger compared to cardiomyocytes from uninfected hamster hearts ( Fig. 4a-b). This is the a slightly smaller range of hypertrophy reported for animal HFpEF models, where increases between 35% and 45% have been reported 25,26 . Other viral myocarditis models, such as CVB3 infection in mice, show approximately 15% increases in cardiomyocyte size after five weeks 27 , further highlighting the unique nature of this cardiac hypertrophy. We found a significant increase in fibrosis, both due to HFD/L-NAME, as well as infection with SARS-CoV-2 (Figure 4c-d). These effects were not cumulative, as infected hamsters on treatment showed the same level of fibrosis as hamsters on treatment or infected alone.
Lastly, we investigated cardiac function at 14 dpi by echocardiography. At this stage, few defects in cardiac function were present (Fig. 5). No differences in heart rate (HR) were observed between the groups (Fig. 5a). Systolic function was assessed by measuring ejection fraction (EF), fractional shortening (FS), and isovolumic contraction time (IVCT) (Fig.   5b, Supplemental Table S1). No difference in FS was observed (Fig. 5b). IVCT was increased in chow diet-fed infected hamsters compared to non-infected hamsters (Supplemental Table S1). Cardiac hypertrophy was also suggested by the echocardiography, with an increased LV mass present in both infected hamster groups, although statistically insignificant (Fig. 5c). Some evidence of the onset of diastolic dysfunction could be observed in the infected HFD/L-NAME hamsters. Although the E-wave increases drastically in severe, late-stage diastolic dysfunction, it decreases early in diastolic dysfunction. Here, a small but statistically significant decrease in the E-wave was present in HFD/L-NAME-treated infected hamsters compared to control uninfected hamsters (Fig. 5d). Similarly, the IVRT was slightly increased (Fig. 5e). However, the most common indicators of diastolic dysfunction, E/E', E/A, and mitral valve deceleration time (MV DT), were unaffected in all groups ( Fig. 5f- Table S1). Complete echocardiographic analysis is available in Supplemental Table S1. Together with the histological data, our echocardiography data suggest that the infected hamster exposed to comorbidities are showing early signs of diastolic dysfunction.

Recovered COVID19 patients present with reduced number of αSMA low pericytes.
Lastly, we investigated whether similar effects were observed in patients that recovered from COVID19 (Fig. 6). Biopsies were obtained from patients that developed cardiac complications, warranting catheterization. We limited our analysis to patients who presented with cardiac injury but had a preserved ejection fraction. As control tissue, we used biopsies from patients who had been suspected of developing viral myocarditis but tested negative, as well as patients diagnosed with HFpEF. We found no differences in cardiac microvascular density between control patients, HFpEF patients, and recovered COVID19 patients, based on CD34 endothelial staining ( Fig. 6a-b). Human cardiac pericytes express several markers, including platelet-derived growth factor receptor beta (PDGFR-β), neuron-glial antigen 2 (NG2), as well as low alpha smooth muscle actin (αSMA) levels 28 .
However, not all pericytes express all these markers, suggesting that each marker may stain a different subset of pericytes. None of the tested PDGRF-β antibodies worked on our samples, possibly due to the fixation protocol necessary for working with infected cardiac biopsies. We found no statistical difference in the percentage of NG2 + capillaries for any of the patient groups, even the HFpEF patients. The data from recovered COVID19 patients was extremely variable (Supplemental Fig. 5). However, we found a significant reduction in the percentage of αSMA low capillaries in both HFpEF and COVID19 recovered patients ( Fig.   6a-c). Together, these data demonstrate that both recovered COVID19 patients and HFpEF patients lose the same amount of αSMA low pericytes, independently of SARS-CoV-2 infection, suggesting that pericytes are indeed targeted by SARS-CoV-2 infection.
Though cardiomyocyte expression of ACE2 is low, several reports show that the expression increases in patients with certain comorbidities 29 . We therefore investigated whether COVID19 affects cardiomyocytes by measuring cardiomyocyte hypertrophy, but could not detect any difference in cardiomyocyte size between either patient group ( Supplemental   Fig. 6). We cannot, however, say with certainty that our control patients lacked hypertrophy since they presented with viral myocarditis-like symptoms.

DISCUSSION
Our results demonstrate that SARS-CoV-2 infection profoundly affects the cardiac microvascular structures by reducing pericyte coverage. This can predispose SARS-CoV-2 infected individuals to developing diastolic heart failure, especially in the presence of metabolic comorbidities. Previous reports stated that ACE2 was largely expressed by cardiac pericytes and to a lesser extent in cardiomyocytes 11,12 . Our in situ hybridization shows that pericytes have the highest levels of viral replication in the heart. This does not mean that the virus is not present in cardiomyocytes, but only that the virus replicates at a much higher level in pericytes.
COVID19 can in some cases lead to viral myocarditis-like syndromes 30  could not yet be detected, as assessed by changes in E/E' and E/A. However, some measures of diastolic dysfunction, specifically IVRT and the E wave velocity, began to indicate dysfunction. Therefore, our results demonstrate that COVID19 recovered patients with metabolic comorbidities should be followed closely to investigate diastolic dysfunction.
The level of pericyte loss observed in our hamster model was much milder compared to the recovered COVID19 patients (around 30% reduction when using αSMA low as a marker) and our animal HFpEF models (41% pericyte loss after 14 weeks in ZSF1 rat model, Cuijpers, et al., in preparation). Pericyte loss in patients was also marker specific, possibly indicated that a specific subset of cardiac pericytes is targeted. Only hamsters with pre-existing metabolic comorbidities exhibited a difference in all pathophysiological markers (hypertrophy, fibrosis, and pericyte loss) and showed some level of diastolic dysfunction by echocardiography. Therefore, we suspect that a model with increased cardiac pericyte loss would have an even more pronounced phenotype. Given that patients experience different viral loads, it would be important to investigate different viral loads and the effect on the development of HFpEF after recovery.
Our results highlight that the cardiac pericyte population can recover in the absence of metabolic comorbidities. Pericytes can migrate and make extensions to replace lost pericytes or recover by proliferation 28,32 . Which of these mechanisms is occurring after SARS-CoV-2 infection is not clear. We can, however, show that pericyte recovery is inhibited by metabolic comorbidities. As such, our results indicate that controlling co-morbidities in COVID19 patients is of primary importance.
Cardiac fibrosis was present in our HFpEF hamster model, as well as in the infected hamsters, regardless of co-treatment with HFD/L-NAME. Cardiac fibrosis has been demonstrated by cardiovascular magnetic resonance techniques after SARS-CoV-2 infection, even months later in young patients with mild or no symptoms during infection 33 . Similarly, a third of COVID19 patients show increased late gadolinium enhancement (LGE) 6,7 . LGE increase can indicate myocyte necrosis, myocardial oedema, myocardial scar tissue, and focal areas of fibrosis 34 . At the same time, increased LGE predicts mortality in viral myocarditis 34 . Our results suggest that the increased LGE observed after SARS-CoV-2 infection is likely due to increased cardiac fibrosis.
Cardiac pericytes were lost, but they may also be hypersensitive to pro-relaxation signals.
Pericytes provide a baseline tone to capillaries and loss of pericytes induces capillary dilatation 32 . Localized loss of pericyte tone increases blood flow to a region 35 , however, massive loss of microvascular tone induces reductions in blood flow 36 . Interestingly, loss of vessel perfusion has been noted upon SARS-CoV-2 infection 37 . Circulating levels of the pericyte vasodilator adenosine increase rapidly during any viral infection 38,39 . Thus, the combination of reduced pericyte numbers and a hypersensitivity to pro-relaxation stimuli may not be beneficial in this case. The receptors for adenosine are also expressed on immune cells, where their activation reduces the intensity of the immune response 40 . If immune cells also upregulated ADORA1/2 in response to SARS-CoV-2 spike protein exposure, then adenosine would lead to a more moderate immune response, something of benefit in the response to SARS-CoV-2 infection. Overall, it is unclear whether ADORA1/2 expression changes represent an overall benefit or detriment. A common variant in adenosine monophosphate deaminase (rs17602729, AMPD1) leads to increased circulating adenosine levels 41 and is present in 20% of the population 42 . The possible significance of these in vitro results could be confirmed by studying this population and whether they are protected or more sensitive to SARS-CoV-2 infection.
Overall, our results show that SARS-CoV-2 infection primarily targets pericytes in the heart. Microvascular dysfunction and fibrosis are hallmarks of HFpEF. We therefore emphasize the need to manage metabolic comorbidities in COVID19 patients. with L-NAME (0.5 g/L; N5751, Sigma Aldrich) dissolved in the drinking water, as indicated.

Organ collection
Hamsters were sacrificed at 4 dpi, or after echocardiography at 14 dpi. Hamsters were euthanized by excision of the heart. Approximately 5-30 mg of the apex of the heart was placed in RTK buffer for RNA extraction, the remaining was fixed overnight in 4% paraformaldehyde (PFA).
Viral titers were calculated with the Reed and Muench method and expressed as 50% tissue culture infectious dose (TCID50) per mg tissue.

In situ hybridization
In situ hybridization was performed as described previously 44

Patient characterization
After routine non-invasive diagnostic work-up and angiography had failed to elucidate any specific cause of heart failure like-symptoms, all patients underwent endomyocardial biopsy (EMB) and left heart catheterization in a standardized manner, as previously described 45  Three different groups of patients were investigated and compared: 1. Patients served as controls where a cardiac cause for their chest discomfort were excluded (n=3; age (57±14); male/female (1/2); normal EF (55±2.9%) without signs of LV diastolic dysfunction and regular NT-pro BNP levels.
2. Symptomatic patients with classical signs of HFpEF, LV diastolic dysfunction, and abnormal NT-proBNP levels, served as a group where a viral presence and / or a significant cardiac inflammation were excluded (n=3; age (54±7.0); male/female (2/1); preserved EF (55±3.2%)). Presence of other viruses was excluded, whereas subacute lymphocytic myocarditis was diagnosed in 1 out of the 3 patients.

Histological analysis
All antibodies and concentrations are listed in Supplemental Table S2. EMB were fixed in 10% neutral buffered formalin overnight at 4°C (classical protocol) or at room temperature to avoid potential SARS-CoV2 infection. Four µm sections were stained with anti-CD34 antibodies, αSMA-Cy3 antibody, and NG2 antibody. Epitope retrieval was performed for CD34 and NG2 stainings by using citrate buffer (0.1M Target Retrieval Solution; DAKO S1699, Agilent). Amplification of CD34 was performed using a Tyramide Superboost kit (Alexa Fluor 488 Tyramide SuperBoostTM Kit, B40922, Invitrogen).

Statistical analysis
Results are presented as mean ± SEM. Statistical analysis was performed using GraphPad 8 Prism. Variance was tested with F-test (p < 0.05) and normality was tested with Shapiro-Wilk test (p < 0.05). Biological outliers were picked up based on Grubb's test (p < 0.05). In vivo data were analyzed using a one-way ANOVA with Dunnett's post hoc test (p < 0.05 *, 0.01 **, 0.001 ***). EAVJ designed experiments, analysed data and wrote the manuscript.

DATA AVAILABILITY STATEMENT
All data are available upon request.   presented as mean ± SEM with n = 6. Statistical analysis was analyzed using a one-way ANOVA followed by a Dunnett's multiple comparisons test, with * p < 0.05, comparing HFD + L-NAME hamsters or infected hamsters with or without HFD + L-NAME diets to control hamsters. Scalebar represents 20 µm (a), 50 µm (c).

Figure 5. Echocardiography demonstrates an extremely mild diastolic dysfunction in SARS-
CoV-2 infected hamsters on HFD + L-NAME diet. Cardiac function of hamsters was investigated using echocardiography two weeks after SARS-CoV-2 infection. HR (a), FS (b), E wave (c), IVRT (d), E/E' (f), and E/A (g) were analyzed. A, late mitral inflow peak velocity; E, early mitral inflow peak velocity; E', early diastolic mitral annulus peak velocity; FS, fractional shortening; HFD, high fat diet; HR, heart rate; IVRT, isovolumic relaxation time; L-NAME, N γ Nitro L arginine methyl ester. Values are presented as mean ± SEM with n = 6. All data were analyzed using a one-way ANOVA followed by a Dunnett's multiple comparison test with * p < 0.05, ** p < 0.01, comparing HFD + L-NAME hamsters or infected hamsters with or without HFD + L-NAME diets to control hamsters. (green), and αSMA (red). Microvascular density (b) and the amount of αSMA low vessels (c) were analyzed. H&E, hematoxylin, and eosin; αSMA, alpha smooth muscle actin. Values are presented as mean ± SEM with n = 3 per group. Statistical analysis was performed using one-way ANOVA followed by a Dunnett's multiple comparisons test, with * p < 0.05, comparing HFpEF patients or recovered COVID19 patients to controls. Scalebar represents 20 µm.