Clade C MERS-CoV camel strains vary in protease utilization during viral entry

Significance

Clade A/B Middle East Respiratory Syndrome coronavirus (MERS-CoV) outbreaks have caused over 957 deaths since the first spillover in 2012; meanwhile, Clade C strains have been found in camels across Africa but have not yet been reported to cause outbreaks. Investigating why these viruses do not successfully transmit to humans will be key to understanding the pandemic potential of the African MERS-CoV camel reservoir. Our study indicates that clade C viruses exhibit less spike cleavage and that East African clade C isolates are less able to utilize the TMPRSS2 for viral entry of both human cell lines and primary nasal cells. Differences in viral entry pathways could alter cellular and organ tropism and contribute to differential pandemic potential.

Abstract

Middle East Respiratory Syndrome coronavirus (MERS-CoV) is a lethal pathogen with pandemic potential. Clade A and B MERS-CoV viruses have caused outbreaks in the Middle East since 2012 when they initially spilled over from camels to humans. Clade C viruses, however, are only found in camels across Africa and the spillover potential of these viruses seems to be lower than for clade A/B strains but remains to be fully understood. Here, we report that clade C spikes are less well-cleaved at the S1/S2 boundary than clade A or B viral spikes and that most clade C spikes induce reduced syncytium formation. Additionally, we demonstrate that several East African clade C strains are less able to utilize the TMPRSS2-mediated pathway for viral entry in both cell lines and primary nasal epithelial cultures. We map the molecular basis of this reduced TMPRSS2 usage to the N-terminal domain and subdomain 2 of East African clade C MERS-CoV. We suggest that reduced usage of the TMPRSS2-mediated entry pathway may underlie the reduced replication of East African clade C strains in humans, while the reduced replication of West African strains remains to be further investigated. Altered protease usage may contribute to differential tropism of East African clade C strains and indicate geographically distinct selection pressures on spike between MERS-CoV strains circulating in camels.

Middle East Respiratory Syndrome Coronavirus (MERS-CoV) is a betacoronavirus in the merbecovirus subgenus which has caused outbreaks in the Middle East since 2012. There have been over 2,627 cases as of April 2025 with a lethality rate of 35% (1). Like SARS-CoV and SARS-CoV-2, MERS-CoV ancestral strains originated in bats; however, most human cases are attributed to infections acquired from interactions with the amplifying reservoir host, camels (2). There are three clades of MERS-CoV which are geographically distinct: A, B, and C (3). Clade A includes the prototypic Erasmus Medical Center (EMC) strain which caused the 2012 outbreak; however, these viruses have since been outcompeted by clade B viruses (4). Clade B viruses circulate in camels across the Middle East and cause sporadic spillover events, with hundreds of cases reported in Saudi Arabia between 2013 and 2020 (5). Clade C viruses, however, are found only in camels across Africa and there is currently no evidence of human outbreaks attributed to these viruses. These viruses are phylogenetically distinct from clade A/B strains, with amino acid substitutions or deletions present in both the viral entry protein and key accessory genes previously shown to antagonize innate immunity (3).

MERS-CoV entry is mediated by the viral entry protein, spike, binding to the cellular receptor DPP4 (6). MERS-CoV spike is a trimer of heterodimers which are composed of two subunits, S1 and S2. S1 contains the N-terminal domain (NTD), which has been shown to bind sialic acids and facilitate attachment (7, 8), and the receptor-binding domain (RBD), which binds to DPP4. The S2 subunit contains the fusion peptide which promotes fusion of the cellular and viral membranes following receptor engagement. Like several other coronaviruses, MERS-CoV spike requires activation in order to bind DPP4 in the form of proteolytic cleavage. Spike contains two cleavage sites, one at the S1/S2 boundary, and one at the S2’ site in the S2 subunit. The first cleavage event, at the S1/S2 site results in conformational changes that allow the RBD to engage DPP4. The second cleavage event at the S2’ site is required for further conformational changes which ultimately release the fusion peptide and initiate fusion of the viral and host membranes (9).

During viral biogenesis, the S1/S2 site can be cleaved by furin as spike traffics through the Golgi. If not sufficiently cleaved, and if endosomal proteases are available, endosomal proteases during infection of the target cell can also cleave the S1/S2 site. It has been reported that both furin and cathepsin L can cleave both the MERS-CoV S1/S2 site and subsequently the S2’ site, allowing the virus to enter the cell via late endosomes. If sufficient S1/S2 cleavage has occurred during spike biogenesis, and transmembrane trypsin-like serine proteases (TTSPs), such as transmembrane protease, serine 2 (TMPRSS2), are available on the target cell surface, these proteases can mediate the S2’ cleavage at the cell surface and the virus can fuse at the cell membrane (9–11). While TMPRSS2 has been the most well characterized TTSP during the viral entry of SARS-CoV-1, SARS-CoV-2, and MERS-CoV, other TTSP’s including TMPRSS4, TMPRSS11D, and TMPRSS13D have been reported to be able to mediate this cleavage event for SARS-CoV-2 (12, 13). Presumably, these proteases could also cleave the S1/S2 of MERS-CoV spike. Thus, relative protease bioavailability across cell-types and prior S1/S2 cleavage can alter the cellular tropism of MERS-CoV (11). In addition to altering susceptibility to MERS-CoV infection, protease bioavailability has been reported to influence the efficiency of viral entry. For SARS-CoV-2, it has been reported that the “early” pathway of viral entry mediated by TMPRSS2 is faster (10 to 15 min) than the “late” pathway mediated by endosomal proteases (45 to 60 min) (14).

While the viral entry of clade A and clade B strains of MERS-CoV has been well characterized, to date the impact of amino acid substitutions in the spike gene of clade C strains on viral entry has not yet been fully examined. It has been confirmed that several clade C spike proteins do not differ from wild-type (WT) MERS in their ability to enter 293T cells overexpressing the receptor DPP4 as pseudotyped vectors (15). However, differences in protease usage, and entry into more physiologically relevant cell types, have not yet been investigated. While none of the amino acid (aa) substitutions in clade C spikes are within DPP4-contacting residues or the S1/S2 cleavage site, previous studies of other coronaviruses’ spike have indicated that sequence changes outside the RBD can impact spike cleavage, protease usage, and cell–cell fusion (9, 16–19).

Here, we utilize a panel of six clade C MERS-CoV strains to demonstrate that clade C spikes are less cleaved at the S1/S2 cleavage site. Additionally, we show that East African clade C strains HKU270, CAC9690, and CAC10200 are less able to utilize the TMPRSS2-mediated pathway for viral entry, which is mirrored by increased usage of cathepsins for endosomal entry by HKU270 and CAC10200. We identify that amino acid substitutions in the NTD, which are shared by all publicly accessible East African camel-derived MERS-CoV strains to date, confer reduced TMPRSS2 usage. Structural predictions of the NTD show that these amino acid substitutions abolish hydrogen bonds between NTD loops. Separately, the V677L substitution in the SD2 region of HKU270 spike also reduces TMPRSS2 usage. We suggest that these differences in viral entry pathway usage may contribute to differential tropism of these clade C viruses and indicate differential selection pressures on spike between MERS-CoV strains circulating in camels.

Results

Clade C Viruses Are Attenuated for Viral Replication in A549^DPP4 and Calu-3 Cells Relative to Clade A/B Viruses.

To begin to determine factors in the differential spillover potential of clade C isolates, we utilized six clade C isolates, three from West Africa (Nig1657, Mor213, and BF785) and three from East Africa (HKU270, CAC9690, CAC10200), along with WT EMC, a clade A strain, and AH13, a clade B strain. All the clade C viruses contain 6-8 amino acid substitutions in the spike gene (Fig. 1A). None of these substitutions are in direct DPP4-contacting residues or the S1/S2 cleavage site; however, it has previously been reported that amino acid substitutions in the NTD, SD2, and S2 regions of spike can alter spike conformation, cleavage, and fusion (9, 16, 17, 19). Mor213 does contain a mutation in the S2’ site, R884L. AH13, a clade B virus, also contains five amino acid substitutions relative to EMC; however, none of these are the same as those in the clade C spikes. We therefore hypothesized that clade C spike substitutions may alter clade C viral entry in the human airways. Viral stocks of these strains were generated and titered in Vero-CCL81 cells (Fig. 1B). Interestingly, most of the clade C viruses exhibited a smaller plaque morphology than the wild-type EMC strain or the clade B strain AH13 (Fig. 1B). This was particularly apparent for Nig1657, Mor213, and CAC9690. BF785 also exhibited a peculiar plaque phenotype, with large fuzzy plaques. Despite their different plaque morphology, most of the clade C viruses produced sufficient titer in Vero-CCL81 cells (Fig. 1C). Mor213 and CAC9690, however, were reduced by 2-logs at peak replication in these cells. We also measured the particle-to-plaque (P:PFU) ratio of this panel of viruses. We used genome copies in viral supernatant as a proxy for viral particles from 24 h of viral infection in Vero-CCL81 cells, and measured plaque-forming units from the same supernatant (Fig. 1D). We found that, under these conditions, the P:PFU ratio varied across these MERS strains but were not significantly different from each other. We did however observe a slightly lower P:PFU for the clade B strain AH13 than the WT EMC clade A strain, perhaps suggesting that clade B MERS strains, which have been reported to outcompete clade A strains in competition studies (4), have a slight increase in viral fitness.

Fig. 1.

Clade C viruses contain amino acid substitutions in the spike gene. (A) Amino acid changes found in clade B strain AH13 and six clade C viruses are shown. NTD = N-terminal domain, RBD = receptor binding domain, SD1 = subdomain 1, SD2 = subdomain 2, FP = fusion peptide, HR1 = heptad repeat 1, HR2 = heptad repeat 2. (B) Representative images of plaque morphology of clade A, B, and C strains in Vero-CCL81 cells. (C) Vero-CCL81 cells were infected with viruses for 24 h at MOI 0.1 and infectivity measured by plaque assay. (D) Vero-CCL81 cells were infected at MOI 0.1 and supernatant and cellular RNA harvested 48 h later. Plaque forming units (PFU)/µL was measured by plaque assay and genomes/µL measured by qPCR. All experiments are N = 3, *P < 0.05 tested by one-way ANOVA showing statistical significance against WT EMC.

Next, we examined the replication of this panel of viruses in respiratory cell lines. A549^DPP4 and Calu-3 cells were infected for 16, 24, or 48 h with EMC (clade A), AH13 (clade B) and the six clade C viruses shown in Fig. 1 (Fig. 2 A and B). EMC and AH13 replicated comparably, as expected, in both A549^DPP4 and Calu-3 cells. HKU270, an East African clade C strain isolated in Egypt, also replicated fairly comparably to EMC and AH13 in A549^DPP4 but was attenuated in Calu-3 cells. All other clade C viruses replicated lower than EMC/AH13 in both A549^DPP4 and Calu-3 cells at all time points, however, were only statistically significant at 48 hpi. Consistent with the literature (3, 20), this confirms these camel-derived isolates do not replicate as efficiently as wild-type EMC in human lung-derived cell lines. The reduced replication of Mor213 and CAC9690 in Vero-CCL81 cells also suggests these two clade C viruses also contain replication defects that are not necessarily specific to human airway cells.

Fig. 2.

Clade C viruses are attenuated for replication and viral entry in human cell lines. (A and B) A549^DPP4 (A) or Calu-3 (B) were infected for 16, 24, or 48 h (MOI 0.1) and infection quantified by plaque assay. (C and D) Pseudotyped lentiviral vectors (PLVs) bearing spikes as shown in Fig. 1A were used to transduce A549^DPP4 (C) or Calu-3 (D) and transduction measured by luciferase activity 48 h later. Transduction is normalized to that of WT EMC. (E and F) The relative inhibition, plotted as percentage of WT EMC for infection at 24 h or transduction at 48 h is plotted for A549^DPP4 (E) or Calu-3 (F). (G) A549^DPP4 were mock or TMPRSS2 transfected 18 h prior to transduction with PLVs bearing EMC or HKU270 spikes. (H) A549^DPP4 were transfected with TMPRSS2 and transduced with PLVs bearing different spikes 18 h later. Transduction was measured by luciferase activity 48 h later and normalized to that of WT EMC. All experiments are N = 3, *P < 0.05 tested by one-way ANOVA showing statistical significance against WT EMC.

To test whether sequence differences in spike contributed to these attenuated phenotypes, we generated pseudotyped lentiviral vectors (PLVs) bearing the different clade C spike proteins (Fig. 1A). Transduction of A549^DPP4 cells with these PLVs demonstrated that, while there were minor differences in transduction efficiency, all clade C spikes were capable of transducing A549^DPP4 cells to a similar extent (Fig. 2C, raw data shown in SI Appendix, Fig. S1A). This is consistent with previous reports which found spikes from Morocco, Nigeria, and Burkina Faso transduced 293T-DPP4 cells comparably to, if not better than, wild-type EMC (15). We next transduced Calu-3 cells and found that Nig1657, Mor213, and BF785 showed a twofold decrease in viral entry in these cell lines (Fig. 2D, raw data shown in SI Appendix, Fig. S1B). Strikingly, however, HKU270 and CAC10200 showed a fivefold decrease in viral entry relative to EMC. To compare the magnitude of replication versus transduction defect, % inhibition relative to EMC was plotted for each virus or PLV for both A549^DPP4 and Calu-3 (Fig. 2 E and F). This demonstrates that, while all clade C spikes could transduce A549^DPP4 cells, replication is severely attenuated in this cell line, with the exception of HKU270. In addition to mutations in the spike gene, Nig1657, Mor213, and BF785 also contain deletions of varying extents in ORF4b. Additionally, Nig1657, CAC9690, and CAC10200 all contain a deletion in ORF3. Given the importance of accessory genes in antagonizing innate sensing, it is highly likely that all the clade C viruses in this panel with the exception of HKU270, are attenuated in A549-DPP4 cells due to postentry effects on viral replication.

However, in Calu-3 cells, the % inhibition of both replication and transduction for HKU270 and CAC10200 was comparable. This suggests that, for HKU270 and CAC10200, the reduced viral replication in Calu-3 cells can be attributed to defects in viral entry. Given that a major difference in A549 and Calu-3 cells is the presence of TMPRSS2, we hypothesized that the presence of TMPRSS2 in Calu-3 cells and the preferred “early” plasma membrane entry route may explain the difference in viral entry for HKU270 and CAC10200. To test this, we transiently expressed TMPRSS2 in A549^DPP4 prior to transduction with PLVs bearing the spike of EMC and one of the clade C panel, HKU270 (Fig. 2G). The presence of TMPRSS2 conferred a 35-fold increase in transduction efficiency for EMC, consistent with previous literature demonstrating EMC spike can utilize the TMPRSS2-mediated pathway for entry (11, 21). HKU270 PLVs, however, only conferred a nonsignificant sevenfold increase in transduction, consistent with the reduced transduction of this PLV in Calu-3 cells which endogenously express TMPRSS2. Next, we examined the transduction efficiency of the entire panel of clade C spike bearing PLVs (Fig. 2H). As expected, the relative transduction efficiency relative to EMC PLVs phenocopied that of Calu-3 cells when TMPRSS2 is present in A549^DPP4 cells. Altogether, these data suggest that, while there is a postentry restriction of clade C viruses in human lung cell lines, there may be a difference in viral replication due to differences in the ability to utilize TMPRSS2 for East African clade C viruses.

Clade C Spike Proteins Are Less Cleaved at S1/S2 Relative to Clade A/B Spikes.

It has previously been reported that cleavage at the S1/S2 of MERS-CoV spike alters entry pathways which may lead to differential viral tropism, with precleavage at the S1/S2 boundary allowing the S2’ site to be cleaved by TMPRSS proteases at/near the plasma membrane, and that spikes lacking the furin cleavage site are instead reliant on the endosomal entry pathway for both S1/S2 and S2’ cleavage events (11, 22, 23). S1/S2 cleavage therefore permits infection of TMPRSS2-expressing cells (11, 24). As overexpression of TMPRSS2 resulted in reduced transduction by East African clade C PLVs, we hypothesized that the spike aa substitutions present in clade C spikes may alter spike cleavage and therefore TMPRSS2 usage. Thus, we next sought to determine the relative spike cleavage of our viral stocks by western blotting both the cellular lysate and purified viral supernatant after 48 hpi in Vero CCL81 (Fig. 3 A–D). Strikingly, all the clade C spikes appeared to be less well cleaved at the S1/S2 boundary than EMC. Surprisingly, AH13 also appeared to be less cleaved than EMC in this cell-type. We next analyzed the spike cleavage in the PLV producer 293T cellular lysates and purified PLVs (Fig. 3 E–H). In both the cellular lysates and purified PLVs, all spikes were cleaved at the S1/S2 site. While there appeared to be a small difference in S1/S2 cleavage when band intensity was quantified, all spikes were similarly cleaved in the PLV system. Differences in cleavage of coronavirus spike proteins in PLV systems relative to authentic virus, or virus-like particles where E, M, and N are also expressed, has previously been reported by other groups (25, 26). Additionally, SARS-CoV-2 virus-like particles (VLPs) produced in 293T cells compared to Vero cells have been demonstrated to have more variable morphology and differences in spike trimer length (27). Taken together, we suggest that while useful to interrogate the viral entry step alone, PLV systems may not fully recapitulate authentic differences in spike cleavage in coronaviruses.

Fig. 3.

Clade C viral spikes are less cleaved at S1/S2 than WT EMC. (A–D) Cellular lysates (A and B) or viral supernatant purified through 20% sucrose (C and D) from Vero-CCL81 and immunoblotted for spike and nucleocapsid. Relative S2 to S0 was quantified in ImageJ. (E and F) 293T cells transfected to produce PLVs were immunoblotted for spike, p24, and HSP90. Relative S2 to S0 was quantified in ImageJ. (G and H) PLVs were purified through 20% sucrose and immunoblotted for spike and p24. Relative S2 to S0 was quantified in ImageJ. (I) EMC and HKU270 were used to infect A549^DPP4 or Calu-3 cells at MOI 0.1 for 24 h and infection quantified by plaque assay. (J) Cellular lysates of A549^DPP4 48 hpi were western blotted for spike, Nucleocapsid, and HSP90. (K) Cellular lysates of A549^DPP4 48 hpi were western blotted for spike and nucleocapsid with 3× as much HKU270 sample loaded onto the gel. (L) Cellular lysates from infected Calu-3 48 hpi were immunoblotted for spike, nucleocapsid, and HSP90. All experiments are N = 3, *P < 0.05 tested by one-way ANOVA indicating statistical significance against WT EMC.

We noted that spike incorporation of the clade C spikes appeared to be reduced relative to clades A and B spikes (Fig. 3G). As it has been previously reported that increased viral entry protein incorporation can increase fusion at the cell membrane for enveloped viruses (19, 28), we normalized total (S0+S2) spike incorporation for both viral and PLV spike (SI Appendix, Fig. S2A). This showed that while there was a mostly nonsignificant trend of reduced spike incorporation in authentic viral particles, there was a significant decrease in spike incorporation for Mor213, HKU270, CAC9690, and CAC10200 PLVs compared to EMC (SI Appendix, Fig. S2B). To investigate if the differences in viral entry we previously observed with PLVs were still present when differences in spike incorporation were taken into account, we normalized the transduction of our PLVs to their relative spike incorporation (SI Appendix, Fig. S2 C and D). This demonstrated that there were no significant differences between PLV transduction in A549^DPP4 cells, consistent with the data in Fig. 2. In Calu-3 cells, we observed that HKU270 and CAC10200 entry was still reduced relative to EMC even when spike incorporation was normalized, however the magnitude of this phenotype was significantly decreased. These data suggest that some of the reduced viral entry of clade C spikes in Calu-3 cells is attributable to reduced spike incorporation; however, for HKU270 and CAC10200, there is still a defect in viral entry when spike incorporation is accounted for. This suggests there are multiple mechanisms by which clade C spikes exhibit reduced viral entry efficiency. Given that we also observed a trend toward reduced spike incorporation with our viral particles, it could be that clade C strains indeed incorporate less spike compared to clade A/B spikes, which remains to be further investigated.

Due to these differences between spike cleavage during authentic virus infection and PLV generation, we next wanted to analyze spike processing of clade C viruses in A549^DPP4 and Calu-3 cells. We chose to compare the cleavage of EMC (clade A), AH13 (clade B), and HKU270 (clade C), as HKU270 replicated comparably to EMC and AH13 in A549^DPP4 but displayed attenuation in Calu-3 at 16 and 24 hpi (Fig. 3I). In A549^DPP4, HKU270 appeared less well cleaved (Fig. 3J). Interestingly, despite the comparable viral titers in this cell line, we observed less nucleocapsid and spike of HKU270 in the cellular lysates. To confirm that the reduced spike cleavage of HKU270 was not merely an artifact of less detectable spike, more protein of HKU270 was loaded and the spike cleavage again analyzed by western blot (Fig. 3K). This demonstrated that even with comparable S0 input, HKU270 spike is less cleaved at the S1/S2 boundary in A549^DPP4 cells. Next, we immunoblotted cellular lysates of Calu-3 cells for spike and again observed reduced spike expression and cleavage of HKU270 (Fig. 3L). This suggests that the S1/S2 site is less accessible by furin-like proteases for HKU270 spike in both A549^DPP4 and Calu-3 cells. Altogether, these data show clade C spikes are less well cleaved at the S1/S2 junction than clades A and B spikes during native viral infection, despite no sequence differences directly within the S1/S2 cleavage site.

Clade C Viruses Are Generally Reduced for Syncytium Formation Relative to Clades A and B Viruses.

S1/S2 cleavage has previously been shown to be required for efficient cell–cell fusion (29). Additionally, mutations across both the NTD and S2 of spike have previously been reported to alter syncytia formation of both SARS-CoV-2 variants of concern (VOCs) and MERS-CoV strains (19, 30). We hypothesized that syncytium formation during infection with these viruses may be altered due to the reduced S1/S2 cleavage of these spike proteins. To test this, we infected A549^DPP4 cells at MOI 0.1 and stained for nucleocapsid at 16 hpi and quantified number of nuclei per syncytium (Fig. 4A, quantified in Fig. 4C). Besides BF785, which formed syncytia comparably to EMC and AH13, all other clade C viruses were impaired for syncytium formation. To confirm if the reduced syncytial formation was simply an artifact of differences in replication at 16 hpi, we transfected spike alone into A549^DPP4 cells and again quantified syncytium formation 16 h later (Fig. 4B, quantified in Fig. 4D). While the magnitude of difference in syncytium formation was smaller than observed with live viral infection, clade C spikes, with the exception of BF785, still formed syncytia to a reduced extent compared to EMC and AH13. In particular, the HKU270 spike was severely attenuated for syncytium formation. Interestingly, HKU270 replicates comparably to EMC at 16 hpi in A549^DPP4 (Fig. 2A); however, syncytium formation is markedly decreased (Fig. 4). This suggests the relationship between viral replication and cell to cell spread is not a direct correlation as may be expected.

Fig. 4.

Clade C viruses are varyingly attenuated for cell–cell fusion. (A) A549^DPP4 cells were infected at MOI 0.1 for 16 h and stained for Nucleocapsid and DAPI. Images were acquired on a Nikon Eclipse Ti 2 and number of nuclei per syncytium counted. (B) A549^DPP4 cells were transfected with 500 ng of spike plasmid of interest and stained for spike and DAPI 16 h later. Images were acquired on a Nikon Eclipse Ti 2 and number of nuclei per syncytium counted. All experiments are N = 3 with six images taken per condition per experiment. (C) Quantification of A. (D) Quantification of B. *P < 0.05 tested by one-way ANOVA indicating statistical significance against WT EMC.

Clade C Viruses Vary in Their Ability to Utilize TMPRSS2 for Viral Entry.

Having established that several clade C viruses appear to be attenuated in a TMPRSS2-dependent manner, we next wanted to investigate if there were differences in the sensitivity of our clade C panel to protease inhibitors of both the endosomal and TMPRSS2-mediated pathway to further characterize if there are differences in protease usage. Calu-3 cells were pretreated with increasing doses of Concanamycin A, an inhibitor of the vacuolar ATP-ase which acidifies endosomes, for 1 h prior to infection with EMC, AH13, and clade C viruses (Fig. 5A). All viruses were inhibited by Concanamycin A, however HKU270 and CAC10200 appeared to be the most sensitive at 1 and 10 nM, albeit nonsignificantly. Next, we tested the sensitivity of these viruses to E64d, which inhibits cathepsins B and L which can cleave spike during entry via the endosomal pathway (Fig. 5B). Again, HKU270 and CAC10200 appeared the most sensitive to this inhibitor, in particular at 10 μM, however differences between the viruses overall were minimal. We next measured the sensitivity of these viruses to Nafamostat, an inhibitor of TMPRSS2 (Fig. 5C). EMC and AH13 were very sensitive to Nafamostat pretreatment, consistent with previous literature on the utilization of TMPRSS2 by the EMC strain (11, 21). West African clade C isolates Nig1657 and Mor213 were also sensitive to Nafamostat pretreatment. As these viruses also exhibited only a twofold reduction in viral entry as PLVs in both Calu-3 and A549^DPP4-TMPRSS2 cells (Fig. 2D), this suggests that these viruses utilize TMPRSS2 comparably to EMC/AH13 for entry, despite apparently poor S1/S2 cleavage. This could suggest that perhaps minimal S1/S2 cleavage is sufficient for TMPRSS2 usage for these West African viral isolates. BF785 appeared slightly less sensitive than the other West African isolates in Calu-3 cells; however, was still inhibited twofold by the lowest dose of Nafamostat. We overexpressed TMPRSS2 in A549^DPP4 cells as before and pretreated these with Nafamostat, to test if this could recapitulate the differences in Nafamostat sensitivity observed in Calu-3 cells (Fig. 5D). In A549^DPP4-TMPRSS2 cells, BF785 appeared resistant to the lowest dose of Nafamostat, but gained sensitivity comparable to the other West African strains at the highest dose. This could suggest that while BF785 does not utilize TMPRSS2 as well as wild-type MERS-CoV, BF785 can utilize other TTSPs available in Calu-3 cells that are also inhibited by Nafamostat. The East African clade C isolates, however, which all exhibited the lowest viral entry measured by PLV transduction in Calu-3 and A549^DPP4-TMPRSS2 cells, were all relatively resistant to Nafamostat until the highest dose in both Calu-3 and A549^DPP4-TMPRSS2.

Fig. 5.

East African clade C viruses are less able to utilize TMPRSS2 for viral entry. (A–C) Calu-3 cells were pretreated with Concanamycin A (A), E64d (B), or Nafamostat (C) for 1 h prior to infection at MOI 0.1 with viruses indicated. Infection was quantified by plaque assay and infectivity normalized to mock treatment per virus. (D) A549^DPP4 pretransfected with TMPRSS2 were pretreated with Nafamostat and infected as in A–C. (E) Primary nasal cells derived from three pooled donors (pool 1 = donors 1627, 1628, 1615, pool 2 = donors 1485, 1578, 1592) were cultured at the air–liquid interface and apically treated with 1 μM of Nafamostat for 1 h prior to infection with EMC or HKU270 at MOI 1. Two biologically independent infections from two independent pools of three different donors are shown. (F) EMC and HKU270 serum-free stocks were pretreated with trypsin prior to infection of Calu-3 cells (MOI 0.1) and infectivity quantified by plaque assay 24 hpi. (G) EMC and HKU270 were pretreated with trypsin as in A and used to infect Calu-3 cells (MOI 0.1) which had been pretreated with Nafamostat for 1 h. Infection was measured by plaque assay 24 h later. Experiments are N = 3, *P < 0.05 tested by one-way ANOVA indicating statistical significance against WT EMC (A–D and G) or mock treatment (E).

While A549 and Calu-3 cells are lung-derived cell lines, we wanted to confirm if differences in TMPRSS2 usage were detectable in primary human nasal cells. Two pooled cultures of primary human nasal cells, each derived from three different donors, were cultured at the air–liquid interface and pretreated with Nafamostat prior to infection with EMC and HKU270 and infectivity measured by plaque assay 48 hpi (Fig. 5E). Interestingly, we observed differences in the titers of EMC in the mock-treated condition between the two different pools of donors, with pool 1 producing higher titer EMC than pool 2. This may be due to donor-dependent differences in receptor or protease expression, as we observed previously (31). Despite the differences in titer of the mock-treated condition, EMC was inhibited threefold by Nafamostat. HKU270 was, however, entirely unaffected by the addition of Nafamostat, confirming that HKU270 is less able to utilize TTSPs during infection of primary human airway cultures.

Next, we wanted to test if East African strains can be forced to utilize the TMPRSS2-mediated pathway by altering spike cleavage. As shown in Fig. 3, all clade C strains appeared less well cleaved at the S1/S2 site than EMC. To test if an East African clade C spike could be made to utilize the TMPRSS2-mediated pathway by precleavage at the S1/S2 boundary, we pretreated EMC and HKU270 with trypsin prior to infection of Calu-3 cells (Fig. 5F). Trypsin pretreatment resulted in a modest boost of twofold to EMC infectivity, and a fivefold increase in infectivity of HKU270. Importantly, pretreatment of both viruses with trypsin prior to infection of Calu-3 cells pretreated with Nafamostat rendered the previously resistant HKU270 sensitive to Nafamostat (Fig. 5G). This confirms that HKU270 can utilize the TMPRSS2-mediated pathway if sufficient S1/S2 cleavage occurs. While some of the West African clade C strains were also poorly cleaved at the S1/S2 boundary and yet still capable of utilizing TMPRSS2, this suggests that for HKU270 cleavage at the S1/S2 site is a prerequisite to TMPRSS2-mediated entry.

Both the NTD and SD2 Region of HKU270 Confer Its Reduced TMPRSS2 Usage.

We next sought to determine the amino acid substitutions which determine the reduced TMPRSS2 usage of HKU270. First, we verified that differences in Nafamostat sensitivity could be recapitulated utilizing pseudotyped lentiviral vectors (Fig. 6 A and B). Similarly to measuring viral infection after 24 h, East African clade C isolates were more resistant to the effects of TMPRSS2-inhibition than EMC, AH13, or West African spikes. Rather than the intermediate phenotype of BF785 found in Calu-3 with virus, PLVs of BF785 appeared as Nafamostat-resistant as the East African spikes, further suggesting that certain biological processes are difficult to fully recapitulate with a PLV system. However, HKU270, CAC9690, and CAC10200 all remained relatively Nafamostat-resistant as PLVs compared to EMC, AH13, Nig1657, and Mor213, confirming these East African spikes are less able to utilize the TMPRSS2-mediated pathway for viral entry.

Fig. 6.

Amino acid substitutions in the NTD and SD2 region of spike govern the reduced TMPRSS2 usage of HKU270. (A and B) Calu-3 (A) or A549^DPP4 pretransfected with TMPRSS2 (B) were pretreated with Nafamostat for 1 h prior to transduction with PLVs bearing different spikes. Transduction was quantified by luciferase activity 48 h later and transduction normalized to mock treatment per spike. (C) Schematic of spike mutants generated into EMC background. (D and E) PLVs were purified through 20% sucrose and immunoblotted for spike and p24. Relative S2 to S0 was quantified in ImageJ. (F) Calu-3 cells were transduced with PLVs bearing different spikes and transduction measured by luciferase activity 48 h later. Transduction is normalized to that of WT EMC. (G) Calu-3 cells were pretreated with Nafamostat for 1 h prior to transduction with PLVs. Transduction is normalized to mock treatment per PLV. (H and I) PLVs were purified through 20% sucrose and immunoblotted for spike and p24. Relative S2 to S0 was quantified in ImageJ. (J) Calu-3 cells were transduced with PLVs bearing different spikes and transduction measured by luciferase activity 48 h later. Transduction is normalized to that of WT EMC. (K) Calu-3 cells were pretreated with Nafamostat for 1 h prior to transduction with PLVs. Transduction is normalized to mock treatment per PLV. All experiments are N = 3, *P < 0.05 tested by one-way ANOVA indicating statistical significance against WT EMC.

As HKU270 bore the most prominent TMPRSS2-inhibitor phenotype, we utilized this spike to generate a panel of spike mutants bearing different HKU270 aa substitutions in the background of EMC. HKU270 contains seven amino acid substitutions in its spike protein, three in the NTD, one in the RBD, one in the SD2, and two in the S2 domain (Fig. 6C). Interestingly, the NTD mutations V26A, D158Y, and H194Y are identical to that of the two other East African isolates in this study (Fig. 1A). Due to these NTD substitutions occurring in multiple spikes together, we swapped the NTD entirely of HKU270 into EMC: V26A/D158Y/H194Y (NTD), and the following aa substitutions individually: A431T (RBD), V677L (SD2), S856Y (S2), or A1158S (S2) Fig. 6C). We measured the relative spike cleavage of these mutant spikes, and while none of the differences were statistically significant, we observed the NTD and S856Y swaps trending toward less S1/S2 cleavage (Fig. 6 D and E). Next, we transduced Calu-3 cells with this panel of PLVs and found that the NTD and V677L substitutions exhibited a fivefold reduction in transduction efficiency, comparable to that of HKU270 (Fig. 6F). Meanwhile, the A431T substitution did not reduce viral entry at all and transduced comparably to EMC. The A1158S mutation, which is found in all other clade C spikes, reduced viral entry mildly, which may contribute to the mild reduction of viral entry of the West African clade C spikes. Next, we used these PLVs to transduce Calu-3 cells pretreated with Nafamostat (Fig. 6G). Again, the NTD and V677L substitutions appeared to confer HKU270-like properties to the EMC spike, with both substitutions separately resulting in reduced Nafamostat sensitivity. This suggests that differing regions of spike can independently contribute to a reduced ability to utilize the TMPRSS2-mediated pathway during entry.

Having established that the NTD and V677L of HKU270 independently confer reduced usage of TMPRSS2 proteases, we wanted to test if substituting the NTD of EMC and reversion of the V677L substitution could rescue the TMPRSS2 usage of the HKU270 spike. We generated an HKU270 spike bearing the NTD of EMC, and an HKU270 spike bearing the L677V substitution and generated PLVs of these mutant spikes (Fig. 6H). We examined the S1/S2 cleavage of these PLVs, and again while differences were not statistically significant, we observed a trend toward more cleavage for the L677V mutant (Fig. 6 H and I). Transduction of Calu-3 cells with these PLVs demonstrated a significant rescue of both of the HKU270 mutant spikes, comparable to that of wild-type EMC transduction (Fig. 6J). Additionally, the HKU270 NTD and L677V revertant mutants both exhibited an increased sensitivity to Nafamostat pretreatment in Calu-3 cells (Fig. 6K). While HKU270 PLVs were statistically significantly different to EMC at 1 and 10 nM of Nafamostat, both the HKU270(NTD EMC) and HKU270(L677V) PLVs were no longer significantly different from the EMC PLVs. Furthermore, both HKU270(EMC NTD) and HKU270(L677V) were significantly different from HKU270 at 10 nM (SI Appendix, Fig. S3). Altogether, these data demonstrate that the NTD of the East African clade C spikes tested here alter the protease usage of MERS spike such that it is less able to utilize TTSPs for viral entry. Additionally, these data show that the V677L substitution in the HKU270 spike also renders spike less able to utilize the TMPRSS2-mediated viral entry.

The aa Substitutions in the NTD of HKU270, CAC9690, and CAC10200 Alter Interloop Interactions.

To investigate the structural consequences of the East African clade C aa substitutions in the NTD, we modeled the WT EMC and HKU270/CAC9690/CAC10200 NTD in AlphaFold 3 (32). The NTD of MERS spike has previously been characterized to contain two loops which comprise the binding site of several NTD-neutralizing antibodies (33). We modeled the EMC and HKU270 NTD regions in AlphaFold and examined the presence of hydrogen bonds between the NTD loops in ChimeraX (Fig. 7 A and B). Interestingly, the presence of D158Y abolishes a hydrogen bond between this previously undescribed loop and the previously described loop 2 (K160-Y201) (Fig. 7 A and B). Additionally, the D158Y substitution results in the loss of a hydrogen bond within this loop (D158-K160) (Fig. 7C). We also measured the hydrophobicity of the surface of the NTD using ChimeraX and found that the D158Y substitution also resulted in the loss of a hydrophilic pocket (hydrophobic residues shown in yellow and hydrophilic residues are shown in cyan) (Fig. 7D). Both the loss of hydrogen bonds between NTD loops and a hydrophilic pocket on the surface of the NTD could alter intra-spike interactions and interactions with attachment factors during viral entry.

Fig. 7.

The D158Y substitution alters the predicted structure of the NTD. (A and B) EMC (A, red) and HKU270 (B, blue) spike sequences were modeled in AlphaFold 3 and hydrogen bonds displayed in ChimeraX. (C) EMC (red) and HKU270 (blue) NTD loops (aa 147-167) are displayed in ChimeraX with the 158-160 hydrogen bond annotated. (D) Hydrophobicity of the EMC (Left) and HKU270 (Right) NTDs are shown with the 158 position highlighted. (E and F) Hydrophobicity of the SD2 region of EMC (E) and HKU270 (F) are shown with key residues highlighted. Hydrophobic residues are shown in yellow, hydrophilic in cyan, and neutral in white.

We next explored the effects of the V677L substitution on the SD2 region of spike (Fig. 7 E and F). The V677L substitution was not predicted to alter any hydrogen bonds in the region. However, when we measured the hydrophobicity of the SD2 region it appears that in the presence of V677 a neutral pocket at T685 exists which is absent in the presence of L677 (Fig. 7F). Altogether, these structural models suggest that these NTD and SD2 substitutions may alter intra-spike interactions and interactions of the NTD with host attachment factors, which may contribute to the altered protease usage of East African clade C spikes.

Discussion

Here we show that East African clade C isolates are less able to utilize TMPRSS2 for viral entry into cell lines than clade A or B strains, or indeed West African clade C strains. Furthermore, we show that the East African strain HKU270 is less able to utilize TMPRSS2 in primary nasal cultures compared to WT EMC. We also identify that independent aa substitutions in the NTD and SD2 of these East African strain spikes confer reduced TMPRSS2 utilization.

The reduced TMPRSS2 usage of East African clade C strains is accompanied by an increase in sensitivity to cathepsin inhibitors and thus an increased reliance on the endosomal entry pathway. A preference for the endosomal route of entry could render East African MERS-CoV strains more sensitive to endosomally located antiviral proteins such as interferon-induced transmembrane (IFITM) proteins or nuclear receptor coactivator 7 (NCOA7), which have been shown to inhibit SARS-CoV-2 in an entry route-dependent manner (22, 34). Additionally, an inability to efficiently infect TMPRSS2-expressing cells in the airways could limit the cellular tropism of East African clade C viruses and thereby contribute to reduced replication and viral spread.

While the West African clade C isolates studied here appear to utilize TMPRSS2 similarly to wild-type EMC and AH13 strains, their spike proteins appear to be poorly cleaved at the S1/S2 boundary. Additionally, all the spike proteins appear more cleaved in 293T cells when generating PLVs than the live virus in Vero-CCL81, A549^DPP4, or Calu-3, despite differences in protease usage broadly being phenocopied between authentic infection and PLV transductions. Increased spike cleavage in 293Ts is consistent with previous reports on SARS-CoV-2 PLVs versus authentic infection/VLPs (25–27), and is likely explained by differences in protease expression, site of spike incorporation, and interactions with other viral structural proteins during viral assembly. Despite this more complex relationship between spike cleavage and TMPRSS2 usage, we demonstrate here that the East African clade C virus HKU270 can be sensitized to Nafamostat with pretreatment with trypsin. Additionally, albeit a small trend, we observe an increase in S1/S2 cleavage of the HKU270 spike when the V677L substitution is reverted. Altogether, these data suggest that the relationship between S1/S2 cleavage and TMPRSS2 usage can be complex. Perhaps subtle differences in S1/S2 cleavage are sufficient to alter spike conformation and therefore the accessibility of the S2’ cleavage site, which requires further examination and more quantitative proteomics techniques to measure differences in S1/S2 cleavage and spike conformation.

These data also suggest that reduced usage of TTSPs does not perfectly correlate with increased usage of endosomal proteases. We observed that while there were large differences in TMPRSS2 usage across this panel, only small differences in cathepsin L usage were measurable. It is highly possible there are other proteases contributing to MERS-CoV entry, such as matrix metalloproteases or trypsin-like proteases which have been shown to cleave other coronavirus spike proteins, which we have not accounted for and which requires further investigation (26, 35). Differences in S1/S2 cleavage observed between clade A/B and clade C spikes could also suggest differences in protease availability in the camelid airway relative to humans. Camelid TMPRSS2 has been reported to share 88% homology with human TMPRSS2; however, the prevalence of TMPRSS2, or indeed other proteases which could cleave MERS-CoV spike, in the camel airways has not yet been investigated (36).

While the clade C MERS-CoV spikes do not contain any aa substitutions in the S1/S2 site, aa substitutions outside of this region could impact the accessibility of the cleavage site and spike dynamics. Mutations in the NTD of early omicron strains of SARS-CoV-2 were reported to result in a more “closed” spike that did not adopt the “RBD up” conformation as easily, despite efficient S1/S2 cleavage (17, 19). Furthermore, these early omicron strains exhibited reduced TMPRSS2 usage which was partially determined by the NTD. It could be that the sequence differences in the NTD between the West and East African spikes studied here confer different levels of “closed” spike. Of the three substitutions found in the NTD of the HKU270, CAC9690, and CAC10200 spikes, two substitutions (V26A and H194Y) are found in either all or most of the West African isolates. The D158Y substitution, however, is only found in the East African spikes in this panel. Interestingly, the D158Y substitution is found in all 34 publicly accessible East African isolates to date but in none of the 12 West African isolates. However, due to the limited number of West African sequences available, further clade C sampling is required before conclusions on geographical differences in spike mutations and protease usage can be drawn.

Our structural predictions of EMC and HKU270 spikes generated in AlphaFold suggest that the D158Y substitution abolishes several hydrogen bonds between two loops of the NTD, which could affect both interspike interactions and binding to host attachment factors. The NTD has been postulated to be important for initial attachment during MERS-CoV entry, in particular in binding to α2,3-linked sialic acids, which have been reported in the nasal epithelium of dromedary camels but predominantly in the lower airways of humans (7, 8). Whether clade C-specific NTD substitutions alter sialic acid binding remains to be examined.

Interestingly, the V677L substitution in the SD2 region of spike also independently confers reduced TMPRSS2 usage for HKU270, and reversion of this substitution in the HKU270 rendered HKU270 as sensitive to Nafamostat as WT EMC. While the CAC9690 and CAC10200 genomes do not encode this substitution, they encode R626P, which is located just upstream of where V677 is located in the SD2 region where V677 is located. The SD2 region of other coronavirus spikes has also previously been implicated in altering protease cleavage at the S1/S2 junction. The D614G aa substitution, which arose early in the SARS-CoV-2 pandemic, was reported to alter furin cleavage efficiency and the up/down ratio of the RBD (37). The H655Y substitution, found in the omicron SARS-CoV-2 lineage, is also in the SD2 region and has been reported to increase the preference of omicron on cathepsin-mediated viral entry (16, 17, 19). Last, introduction of a Y623H substitution found into SARS-like coronavirus RsSHC014 spike protein increases its entry into human cells by promoting an open spike conformation (38). While requiring further examination, it is possible that the V677L mutation operates in a similar manner and contributes to reduced TMPRSS2 usage by altering spike conformation. Intriguingly, it has also been reported that the SD2 region responds to the motion of the NTD and can adopt a different conformation depending on NTD movement, and that small changes in SD2 conformation could lead to large changes in NTD and RBD movement (37). It may be that the sequence changes in the NTD and SD2 regions of clade C spikes function cooperatively to reduce S1/S2 cleavage and TTSP usage. This may partially explain how both mutation of the NTD and SD2 independently both alter TTSP usage in our PLV experiments.

Somewhat surprisingly, we also find that BF785 is less able to utilize TMPRSS2 than the other West African strains or the WT EMC or AH13 despite lacking the D158Y substitution or substitutions in the SD2 region. This suggests there is an alternative mechanism underlying the differential protease usage of BF785. Furthermore, while BF785 appeared moderately sensitive to Nafamostat in Calu-3 cells, BF785 was resistant to Nafamostat in A549^DPP4-TMPRSS2 cells which could suggest that BF785 utilizes an alternative serine protease present in Calu-3 cells that is inhibited by Nafamostat (39). BF785 is also the only clade C virus in this panel to mediate cell–cell fusion comparably to WT, perhaps indicating BF785 has a more cell-associated phenotype than WT EMC. This has previously been observed with the mouse hepatitis virus (MHV) strain JHM, where increased receptor-independent cell–cell spread correlates with the increased neurovirulence of this MHV strain (40).

The attenuation of replication of West African clade C isolates appears to be independent of viral entry. Nig1657, Mor213, and BF785 are all attenuated in both A549^DPP4 and Calu-3 cells, despite robust entry into A549^DPP4 and only a twofold decrease in entry into Calu-3 cells. We suggest that defects in other viral genes, namely NS3 and NS4b, underlie the attenuation of Nig1657, BF785, and Mor213. NS4b has been well characterized as an antagonist of activation of the antiviral protein RNase L, which degrades host and viral RNA (41–43). An inability to disrupt RNase L activation may explain the reduced replication of West African isolates in A549 and Calu-3 cells observed here. The nuclear localization signal on NS4b has been shown to compete with and prevent nuclear localization of NFkB with potential effects on inflammatory responses (44). Additionally, Nig1657, CAC9690, and CAC10200 all contain a deletion in NS3, the role of which remains to be examined. Further studies are required to elucidate the entry-independent attenuation of replication of these viruses in human cells.

Altogether, we show here that camel-derived East African isolates of MERS-CoV are less able to utilize the TMPRSS2 protease for viral entry and map the molecular determinants of this differential protease usage. We suggest that limited cellular tropism of East African clade C viruses could contribute to their attenuation in the human airways.

Materials and Methods

Patient derived nasal epithelial cells, A549^DPP4 and Calu-3 cell lines were infected with MERS-CoV (EMC/2012, AH13, Nig1657, Mor213, BF785, HKU270, CAC9690, and CAC10200). Pseudotyped lentiviral vectors bearing spike proteins of these viruses were used to transduce A549^DPP4 and Calu-3 cell lines. Infected cells, pseudotype producer cells, and pseudotypes, were analyzed for infectious virus production/lentiviral transduction, spike/nucleocapsid expression by immunofluorescence staining and western immunoblotting, and for sensitivity to protease inhibitors. Nasal cells were obtained via cytologic brushing of patients in the Department of Otorhinolaryngology-Head and Neck Surgery, Division of Rhinology at the University of Pennsylvania and the Philadelphia Veteran Affairs Medical Center after obtaining informed consent. The full study protocol, including acquisition and use of nasal cells, was approved by the University of Pennsylvania Institutional Review Board (protocol #800614) and the Philadelphia VA Institutional Review Board (protocol #00781) with off-site waiver approval for BSL-3 work. All these techniques are described in SI Appendix, Materials and Methods. All study data are included in the article and/or SI Appendix.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Study data are included in the article and/or SI Appendix.