Increased susceptibility of human endothelial cells to infections by SARS-CoV-2 variants

Human umbilical vein endothelial cells (HUVEC; CC-2935), human coronary artery endothelial cells (HCAEC; CC-2585), human cardiac microvascular endothelial cells (HCMVEC; CC-7030), and human lung microvascular endothelial cells (HLMVEC; CC-2527), and human pulmonary arterial cells isolated from diabetics (D-HPAEC; CC-2924) were purchased from Lonza and cultured in endothelial basal medium (EBM; CC-3156, Lonza) supplemented with 10% fetal calf serum (FCS; 4133, Invitrogen), amphotericin-B, ascorbic acid, bovine brain extract, endothelial growth factor, gentamycin sulfate, and hydrocortisone (EGM-Bullet Kit; CC-3124, Lonza) or EBM-2 supplemented with 10% FCS, hydrocortisone, FGF, VEGF, R3-IGF, ascorbic acid, EGF and GA-1000 (EGM-2-Bullet kit; CC-3162, Lonza) at humidified atmosphere, at 37 °C/5% CO₂.

Primary human umbilical vein endothelial cells were isolated and purified using CD144 antibody-coated magnetic beads (Dynal Biotech, Hamburg, Germany) and cultured as reported previously [6]. The human umbilical cords were obtained from local hospitals in Frankfurt am Main and the use of human material in this study conforms to the principles outlined in the Declaration of Helsinki. The isolation of human cells was approved by the ethics committee at the Goethe University in Frankfurt am Main. HUVEC were cultured in endothelial basal medium (EBM; CC-3156, Lonza) as mentioned above for the commercial HUVEC.

Living human heart slices were generated and cultured as recently described [9]. Samples of left ventricular myocardium were obtained from failing hearts during transplantation in the Clinic of Thoracic and Cardiovascular Surgery, Heart and Diabetes Center, Bad Oeynhausen, Germany. The procedure has been approved by the institutional ethics board, and patients have provided informed consent to the scientific use of the explanted tissue. In brief, heart slices were generated from the explanted failing human myocardium by cutting 300-µm-thick vibratome sections. Slices were mounted and cultured in biomimetic culture chambers in a standard incubator (37 °C, 5% CO₂, 20% O_2, 80% humidity) [9]. Pacing was performed at 0.5 Hz with bipolar 50 mA pulses comprised of 1 ms charging and discharging pulses separated by a 1 ms interval. Slices were cultured in Medium 199 supplemented with penicillin/streptomycin, insulin/transferrin/selenite and 50 µM 2-Mercaptoethanol. Medium was exchanged in part (1.6 ml of 2.4 ml total volume in each biomimetic cultivation chamber) at 36–48 h intervals.

SARS-CoV-2 (strains: D614, G614, B.1.1.7, B.1.351, and P.2) were isolated and propagated in CaCo2 cells as previously described [4, 12]. For infection of endothelial cells, the viral stock was diluted to the desired MOI (multiplicity of infection) in the respective medium supplemented with 1% FCS and incubated for 2 h. Then, the medium was changed to the respective culture media (see above). Five days after infection, endothelial cells were fixed in 4% paraformaldehyde (PFA) for 10 min or lysed for RNA isolation. CaCo2 cells were fixed after 24 h. Living human heart slices were incubated with 200 µl of viral stock (1.10⁷ TCID 50/ml) for 3–5 days using the above-mentioned medium without 2-Mercaptoethanol.

For experiments that mimic pro-inflammatory conditions, endothelial cells were treated for 24 h with 30 ng/ml TNFα (210-TA-005, R&D Systems), 10 ng/ml IFNβ (8499-IF-010, R&D Systems) or 10 ng/ml IFNγ (285-IF-100, R&D Systems) prior to viral infection.

To block proteasomal degradation, cells were incubated with the virus together with 10 µM MG132 (M7449, Sigma-Aldrich; solved in DMSO) for 24 h.

To inhibit viral infection, HCAECs pre-incubated with 10 µM chloroquine (PHR1258, Sigma-Aldrich), 1 µM (in DMSO) cathepsin inhibitor N-Acetyl-l-leucyl-l-leucyl-l-methional (ALLM; 0384, Tocris) or 20 µM Furin Inhibitor I (344390-AMG, Merck). 5 µg/ml human recombinant ACE2 (933-ZN, R&D Systems) was mixed with the virus and incubated for 30 min, prior to infection. Cells were fixed 3 days after infection.

Cells were fixed with 4% PFA and were permeabilized with 0.1% Triton X-100/PBS for 10 min. Cells were blocked with blocking solution (5% donkey serum in PBS) for 1 h at RT. Primary antibodies were incubated in blocking solution overnight at 4 °C. After four 5-min washes with PBS, secondary antibodies and DAPI were incubated for 1 h in blocking solution. Finally, two 5-min washes of PBS were performed before cell observation.

Heart slices were fixed in 4% HistoFix (P087.4, Carl Roth GmbH) for 24 h at 4 °C. Slices were transferred to a series of ascending sucrose concentration: 4% sucrose in PBS (1 h, 4 °C), 15% sucrose in PBS (4 h, 4 °C) and finally 30% sucrose in PBS (overnight, 4 °C). The day after, slices were washed twice for 30 min with 100 mM glycine at 4 °C and once with PBS for 30 min at 4 °C. Cardiac slices were permeabilized with 1% Triton X-100 in PBS overnight at 4 °C. Slices were then washed three times with PBS for 30 min at 4 °C and blocked with blocking solution (3% BSA in 0.3% Triton X-100) overnight at 4 °C. Slices were washed three times in 0.3% Triton X-100 and incubated with primary antibodies that were diluted in blocking solution (overnight, 4 °C). Slices were again washed trice with 0.3% Triton X-100/PBS (30 min each) and incubated with the secondary antibodies that were diluted in PBS (overnight, 4 °C). Finally, slices were again washed trice with 0.3% Triton X-100/PBS before mounting (see Suppl. Tab. 1 for detailed antibody information).

Total RNA was isolated using RNeasy Mini Kit (217004, Qiagen) according to the manufacturer’s instructions including an on-column DNase I digestion step (79254, Qiagen). Reverse transcription was performed using 500 ng RNA, random hexamers and MuLV reverse transcriptase (N8080018, Thermo Fisher). Fast SYBR Green qPCR were carried out by StepOnePlus real-time PCR systems (4385617, Thermo Fisher). RPLP0 amplification was used for data normalization. Relative expression levels were calculated by 2^‒ΔCt. Primer sequences are provided in Suppl. Table 2. PCR products were visualized on a 1.5% agarose gel in 1xTAE buffer and visualized with Midori Green Advance (617004, Biozym Scientific GmbH).

Supernatants from infected endothelial cells were collected 5 days post-infection. Confluent layers of CaCo₂ cells in 96-well plates were infected with serially diluted supernatants. Cytopathogenic effect (CPE) was assessed visually 48 h after infection. The infectious titer was determined as TCID50/ml.

Single-nuclei RNA-sequencing data from human cardiac samples were derived from a previously published data set [21]. Data were analyzed with the Seurat (v3) package.

Human umbilical vein endothelial cells (HUVEC), human coronary artery endothelial cells (HCAEC), human cardiac microvascular endothelial cells (HCMVEC) and human lung microvascular endothelial cells (HLMVEC) were incubated with isolated SARS-CoV-2 [4] for 2 h and viral infection was measured 5 days post-infection by detection of intracellular viral spike protein (Fig. 1a). Spike protein was only weakly detected in HCAEC and not in any of the other endothelial cells (Fig. 1b–d), even when extending the duration of infection to 24 h (data not shown). The activity of the viral strain used [5] was evidenced by its ability to infect CaCo2 cells (Fig. 1b, d). Since we used expanded endothelial cell cultures that might change their gene expression during passaging, the ability of the virus to infect freshly isolated HUVEC was also tested. In line with the results obtained with commercially available HUVEC, no viral spike protein could be detected following incubation with SARS-CoV-2 (Suppl. Fig. 1a–c). Of note, endogenous degradation pathways might also prevent significant virus accumulation in HCAEC. To test this assumption, we treated HCAEC with the proteasomal inhibitor MG132. MG132 significantly increased spike protein accumulation suggesting that indeed endogenous degradation pathways limit viral protein accumulation (Suppl. Fig. 2).

Since diabetes mellitus is a risk factor for complicated courses of COVID-19, we additionally infected pulmonary artery endothelial cells from subjects with diabetes (D-HPAEC). However, no spike protein was detected (Fig. 1d, e). In addition, we investigated whether or not a pro-inflammatory environment could sensitize endothelial cells to SARS-CoV-2 infection. To mimic a pro-inflammatory environment as it would occur in COVID-19 patients, we pre-incubated endothelial cells with TNF-α or interferons for 24 h prior to infection. However, the pre-exposure of endothelial cells to TNF-α, IFNβ or IFNγ did not change the levels of spike protein in any of the endothelial cells studied (Fig. 1f–h and Suppl. Fig. 2). Rather, the inflammatory cytokine TNF-α reduced spike protein levels in SARS-CoV-2 infected HCAECs (Fig. 1f, g, Suppl. Fig. 3). Given the discussion, that hydrocortisone may be antiviral [17] and given the fact that the used media contain hydrocortisone, we repeated the SARS-CoV-2 infection experiments in hydrocortisone-depleted media. However, no differences in spike detection were found in the presence and absence of hydrocortisone (data not shown).

We additionally explored whether the cardiac microvascular endothelium can be infected in human heart slides ex vivo. However, only cardiomyocytes and other interstitial cells were positive for spike protein (Fig. 2, white arrows). There was no endothelial cell found to be positive for spike protein. Together these data suggest that only coronary artery endothelial cells but, not cardiac microvascular or pulmonary endothelial cells are susceptible to SARS-CoV-2 infection in vitro.

Next, we determined whether the different susceptibility of endothelial cell types to SARS-CoV-2 infection might be explained by expression of putative SARS-CoV-2 receptors. The well-established co-receptor ACE2 was expressed by HCAEC but not by any of the other endothelial cells (Fig. 3a, Suppl. Fig. 1a). However, unlike CaCo2 cells, where ACE2 proteins was concentrated at the cell membrane, in HCAECs ACE2 was largely present in the perinuclear area with little or no protein detectable at the plasma membrane (Fig. 3a, b). Expression of the mRNA encoding the protease TMPRSS2, which was shown to act as an activator of SARS-CoV-2 entry, was below detection level in all of the endothelial cells studied (CT-values of water control: 36, TMPRSS2: 37.8 ± 0.7 in endothelial cells compared to 22 ± 0.2 in CaCo2 cells). Other proteases known to be involved in SARS-CoV-2 uptake, such as cathepsin B and cathepsin L, were expressed in the endothelial cells tested (Fig. 3c). Neuropilin-1 (NRP1) was recently reported to bind to Furin-cleaved substrates and to potentiate SARS-CoV-2 infectivity [7]. NRP1 and FURIN were clearly expressed in heart tissue as determined by single nucleus RNA sequencing (Suppl. Fig. 4) and in all of the cultured endothelial cells (Fig. 3c). CD209L was reported to be a further receptor for SARS-CoV-2 in epithelial and endothelial cells [1], and surprisingly, was weakly expressed by HLMVEC and D-HPAEC only (Fig. 3c). Interestingly, bulk RNA sequencing analysis of human aortic endothelial cells, HCMVEC and HUVEC showed elevated levels of ACE2, CTSL and NRP in aortic endothelial cells, which may explain why arterial or coronary endothelial cells might be more prone to SARS-CoV-2 infection (Suppl. Fig. 5). To assess the involvement of these proteins in SARS-CoV-2 entry, SARS-CoV-2 infected HCAECs were co-incubated with human recombinant ACE2, FURIN-inhibitor I, chloroquine and the non-selective cathepsin inhibitor ALLM. Interestingly, 3 days post-infection, we found a significant reduction in spike protein staining under all of these conditions (Fig. 3d).

Although we detected spike protein in HCAECs after SARS-CoV-2 infection, this does not necessarily document permissive infection of the endothelial cells. Therefore, we further assessed the levels of double-stranded RNA, as a sign of viral RNA synthesis [25, 29], and the presence of infectious virus in the cell supernatant, which would be indicative of viral replication. However, neither double-strand RNA (Fig. 4a) nor infectious virus in HCAEC supernatant could be detected 5 days after SARS-CoV-2 infection (Fig. 4b, Suppl. Table 3). Intracellular RNA copies were only transiently increased during the initial 3 days (Fig. 4c), suggesting that no new virus is generated by the tested endothelial cells and that the spike protein may originated from the virus that was originally taken up.

Indeed, SARS-CoV-2 has been reported to enter cells via receptor-mediated endocytosis [29], and the spike protein detected in HCAECs was localized to endosomal calnexin-positive areas (Fig. 4d). To determine if endosomal localized virus may elicit an ER stress response, we measured various genes known to be induced by ER stress. Interestingly, only EDEM1, which is involved in the clearance of misfolded proteins [18], was transiently up-regulated at day 1 and day 2 post-infection (Fig. 4e), whereas BiP, DDIT3 and ATF4 mRNA expression were not affected (Fig. 4f–h).

Finally, we assessed putative cytotoxic effects in the SARS-CoV-2-infected endothelial cells. However, we did not find evidence for cytopathic effects in the endothelial cells studied (Fig. 4i, Suppl. Figs. 1d, 3c). Moreover, expression of the inflammatory cytokine IL-6, the adhesion molecule ICAM1 and the pro-angiogenic and permeability inducing VEGF were not changed by SARS-CoV-2 infection. VCAM1 was down-regulated by active and UV-irradiated SARS-CoV-2 virus (Fig. 4j).

Recently, we have isolated and characterized several novel SARS-CoV-2 variants including B.1.1.7 (mutations include N501Y and del69/70), B.1.351 (mutations include E484K and N501Y) and P.2 (mutations include E484K in the absence of a N501Y mutation) [31]. Emergence of these variants has raised concerns regarding their host infection capability. We have performed comparative experiments to test whether the novel variants may infect and replicate in HCAECs. Cells were infected with different SARS-CoV-2 variants (D614, G614, B.1.1.7, B.1.351, P.2) at MOI 1. Viral replication was assessed by measurement of viral RNA at different time points post-infection. The production of infectious viral particles was determined by titration of infected cell supernatants on confluent CaCo2 cell layers. Although intracellular viral RNA was detected after infection of HCAECs with the variants, particularly the variant B1.1.7, no increase was observed over time (Fig. 5a), indicating that the virus enters the cell, but does not replicate in the endothelial cells. Consistently, none of the variants produced infectious viral particles (Fig. 5b). These data demonstrate that SARS-CoV-2 variants are not able to replicate in HCAEC. However, we observed enhanced virus uptake in endothelial cells by novel SARS-CoV-2 variants (B.1.1.7, B.1.351, and P.2) in association with increased spike staining compared to early variants (D614 and G614) (Fig. 5c). Interestingly, cell toxicity was not observed, despite B.1.1.7, which significantly reduced HCAEC counts (Fig. 5c).