SARS-CoV-2 neuroinvasion selects for FCS mutations
To better understand the determinants driving viral evolution and compartmentalization, we performed viral whole genome sequencing of SARS-CoV-2 in the lungs and brains of mice which had received different vaccine formulas. We hypothesized that vaccine-induced immunity against S and/or nucleocapsid (N) would drive viral evolution towards variants that escape antibody-dependent neutralization in a compartment-specific manner. To address this hypothesis, K18-hACE2 mice were intramuscularly vaccinated with adenovirus type-5 (Ad5) vector vaccines that encoded either the SARS-CoV-2 S open reading frame (Ad5-S) or the N open reading frame (Ad5-N) from the SARS-CoV-2 isolate USA-WA1/2020, similar to our previous studies26–30. Mice were vaccinated with either 1x109 PFU of Ad5-S, 1x109 PFU of Ad5-N, 1x109 PFU of both Ad5-S and Ad5-N each, or with PBS as a control (n = 5 mice per condition). After three weeks, mice were challenged intranasally with 5x104 plaque forming units (PFU) of USA-WA1/2020 and ultimately euthanized at 5-days post infection (dpi) (Fig. 1a). Total RNA from lung or brain homogenate was analyzed for the presence of viral RNA by quantitative reverse transcription PCR (qRT-PCR) and used for viral whole genome sequencing using an amplicon based approach as previously 31–33.
Whole genome sequencing and phylogenetic analysis of the viral isolates from each condition revealed that, while the lung isolates were highly similar across all vaccination groups, the brain isolates were much more diverse (Fig. 1b). To better understand the distribution of these changes, Shannon entropy (a measure of viral diversity) was calculated at every position along the genome for the brain isolates, the lung isolates, and the USA-WA1/2020 input inoculate (Fig. 1c). While several positions varied within each compartment, there was a notable enrichment in diversity in S with the most diversity occurring in and around the FCS. Examining the consensus sequences of each isolate in this region, we found all 20 lung isolates had a majority consensus sequence that matched the reference with an intact FCS (Fig. 1d) regardless of vaccination. In stark contrast to this, 15 of the 20 brain isolates had a majority consensus sequence with a substitution mutation, frameshift, or deletion in or near the FCS.
Notably, the parental isolate USA-WA1/2020 used in the above experiment was found to be polymorphic at position 23606 in the FCS. Sequencing of the USA-WA1/2020 inoculate yielded 47.9% reference sequence at that position while 52.1% of reads contained a C23606T (R682W) mutation that renders the FCS nonfunctional. Furthermore, SARS-CoV-2 is known to accumulate to very high titers in the tissues of K18-hACE2 mice, which may impact the selective pressures underlying compartmentalization. To test our findings in an alternate model, we intranasally challenged immunocompetent BALB/c mice with a mouse adapted strain of SARS-CoV-2 (MA10)24 and euthanized at 5 dpi (n = 5 mice, Fig. 2a). Total RNA was extracted from lung or brain homogenate to determine viral load and perform viral whole genome sequencing. While virus was detected in the brains of the MA10 infected BALB/c mice (Fig. 2b), titers were lower than in the USA-WA1/2020 infected K18-hACE2 mice. Whole genome sequencing and phylogenetic analysis of the viral isolates from each compartment revealed that the lung isolates again more closely matched the initial inoculate compared to the brain isolates (Fig. 2c). Two out of the five brain isolates showed substantial divergence from the input inoculate with one isolate’s majority consensus sequence harboring an R682G mutation in the FCS (Fig. 2d). Furthermore, Shannon entropy per position over the genome was significantly increased in the brain compared to the lung isolates (linear mixed effect model p < 0.0005, Fig. 2e). This diversity was again concentrated in the S open reading frame, with the highest peaks in entropy at sites in and around the FCS (Fig. 2f). Taken together, these data - in two independent mouse models with varying levels of CNS pathology and viral replication - suggest that neuroinvasion elicits a selective pressure for deletion or mutation of the FCS independent of prior immunity.
FCS regulates the host cell entry pathway.
A previous study reported that deletion of the FCS in SARS-CoV-2 attenuates infection and reduces replication in the respiratory tract34, but it is not clear if this occurs in other organs of the body. The CNS has lower expression levels of ACE2 and TMPRSS2 compared to lung epithelial cells 35–37. The lack of TMPRSS2 in the CNS suggests that the virus must use an alternative entry pathway to target cells in the brain. The brain has high levels of cathepsin B and L expression and ACE2 can be found on epithelium, neurons, and vascular cells38. We thus hypothesized that the deletion of the FCS in the CNS is due to tissue specific selective pressure and utilization of the endosomal entry pathway.
In order to evaluate the viral entry pathway of a virus lacking the FCS, we generated luciferase expressing pseudoviruses with an HIV backbone and SARS-CoV-2 S glycoprotein on the surface. The pseudovirus was generated with the S amino acid sequence from the parental WA-1 strain of SARS-CoV-2 or S with the amino acids 681-PRRA-684 deleted (ΔFCS). Loss of the FCS prevents furin-mediated processing of the S protein, resulting in higher levels of full-length S protein compared to the cleaved S1 and S2 domains as demonstrated by Western blot on the purified WA-1 and ΔFCS pseudoviruses (Fig. 3a).
To interrogate whether the ΔFCS pseudovirus exploits an alternative entry pathway, we transduced cells in the presence of TMPRSS2 and/or cathepsin B/L protease inhibitors. Expression of luciferase was used to determine transduction efficiency in VeroE6 cells overexpressing hACE2 and TMPRSS2. Cells were treated with 10µM of aloxostatin (E64d), which inhibit cathepsins, and/or 10µM of camostat mesylate, which inhibits TMPRSS2 activity. Luciferase expression remained unchanged with aloxostatin, indicating that both WA-1 and ΔFCS can enter through the canonical TMPRSS2 mediated entry (Fig. 3b). However camostat mesylate treatment caused a 26-fold decrease in WT pseudovirus entry while the ΔFCS mutant entry was unaffected. Combinatorial treatment with both aloxostatin and camostat mesylate lowered transduction of the ΔFCS pseudovirus. This data indicates that the ΔFCS mutant can efficiently utilize either the endosomal entry or TMRPSS2 mediated entry.
ΔFCS virus is attenuated in the lung but not in the brain.
Due to the increased prevalence of the ΔFCS mutant in the CNS, we hypothesized that the ΔFCS has a selective advantage within the CNS compared to the WA-1 virus. To address this hypothesis, we first obtained an infectious clone of the USA-WA1/2020 virus along with an infectious clone of USA-WA1/2020 with the ΔFCS mutation18. We confirmed that the S protein of the WT infectious clone and the ΔFCS infectious clone had differential cleavage of the S protein by western blot (Fig. 3c). After amplifying both viruses in VeroE6 cells expressing hACE2 and TMPRSS22, the WT viral stock contained 91.5% of reads with the WT FCS sequence and 8.5% of reads with mutations in the FCS region. The ΔFCS stock had 93.6% of reads with the predicted ΔPRRA deletion and 6.4% of reads with wild-type FCS. With these stocks we inoculated 5-week-old K18-hACE2 mice intranasally with 6x103 PFU of either virus (Fig. 4a). As seen previously, the ΔFCS mutant caused significantly less weight loss in intranasally inoculated mice. At 4dpi, WT infected mice lost 17% body weight while ΔFCS infected mice gained ~ 5% body weight (Fig. 4b). Intranasally inoculated mice were euthanized at 2- and 5-dpi to provide a time course of viral dissemination during early infection. The ΔFCS mutant had an impaired growth rate compared to the WT virus with a 25-fold reduction in genomes/lung at 2-dpi (p-value = 0.0001) (Fig. 4c). This impaired growth phenotype was maintained through 5-dpi with 6-fold reduction in viral RNA titers in the lung of ΔFCS inoculated mice (p-value = 0.0002). We then asked if the deletion of the FCS provided improved entry into the CNS where it could then replicate. However, we found that at 2-dpi the ΔFCS had 58-fold lower genome/organ in the CNS compared to the WT virus (p-value = 0.0015). By 5 dpi the ΔFCS virus had slightly higher viral titers, although these differences were not significant (Fig. 4d). The delayed morbidity of the ΔFCS infected mice is likely due to the delayed viral kinetics in the respiratory system.
To assess CNS specific replication dynamics, we inoculated K18-hACE2 mice directly into the brain with 1x102 PFU of WT or ΔFCS stocks (Fig. 5a). Following inoculation with either strain, mice reached humane endpoint criteria by 3 dpi; however, there was no clear difference in clinical symptoms or weight loss between the groups (Fig. 5b). We quantified viral titers in the brain and lung at 1 and 3 dpi. At 1 dpi, the ΔFCS mutant virus had increased viral loads in brain compared to the WT (4E6 vs. 9E5, p-value = 0.03), indicating that ΔFCS virus has an early growth advantage in the CNS (Fig. 5c). By 3 dpi, there was no difference in the viral loads in ΔFCS and WT infected mice. We also quantified viral genomes in the lungs of mice infected via the intracranial route to evaluate the possibility of leakage from the CNS into the respiratory tract. Low levels of viral genome copies in the lung were detected and these values increased over 1000-fold from day 1 to day 3, suggesting that viral dissemination can occur not only from the respiratory tract to the brain, but also from the brain to the lung (Fig. 5d). We could not detect infectious virus in the lung via plaque assays likely due to the limit of detection of this assay. However, in vitro infection of VeroE6 cells expressing ACE2 and TMPRSS2 with lung homogenate from intracranially infected mice resulted in cytopathic effect (Fig. 5e), indicating the presence of infectious virus. Collectively, these data demonstrate that, in contrast to the findings in the lung, the ΔFCS virus is not attenuated in the brain, and that virus originating from the brain can traffic back out to the respiratory tract.
To better understand the compartmental dynamics following intracranial or intranasal challenge, each isolate was again subject to viral whole genome sequencing and phylogenetic analysis. Mice infected intranasally with WT virus (left panels) displayed the previously observed pattern with the lung isolates (n = 5, green squares) grouping closely with the input inoculate (n = 1, blue circle) and with the brain isolates showing more divergence (n = 5, green triangles) (Fig. 6a). Following intranasal inoculation, the consensus sequence of the FCS in each lung isolate matched the reference whereas 3 of the 5 brain isolates had substitution mutations or deletions near or in the FCS (Fig. 6b). On the other hand, mice infected intracranially with WT virus had some lung isolates (n = 5, red squares) cluster more closely with the brain isolates (n = 5, red triangles) with more divergence from the input inoculate (Fig. 6a). Looking at the consensus sequences, 4 of the 5 brain and lung isolates maintained wild-type sequence across the FCS after intracranial inoculation, with one isolate in each compartment gaining a substitution or deletion mutation in that region (Fig. 6b). Quasispecies analysis confirmed that divergence of viral subpopulation after intracranial inoculation was maintained in the lung after trafficking (Fig. 6c).
Mice infected with the ΔFCS virus (right panels) showed a distinct pattern of viral evolution. Upon intranasal inoculation, some divergence is observed relative to the input inoculate (Fig. 6a), but all brain and lung isolates maintain the FCS deletion (n = 10, Fig. 6b). Upon intracranial inoculation, the FCS deletion is similarly conserved in all brain isolates (n = 5). However, the lung isolates of intracranially ΔFCS inoculated mice were highly divergent from the inoculating virus (red squares, Fig. 6a). Furthermore, each lung isolate was found to have partially (n = 2) or completely (n = 3) re-acquired the FCS sequence (Fig. 6b), which is similarly reflected in the quasispecies analysis (Fig. 6c). Taken together, these data suggest that trafficking of SARS-CoV-2 between the lung and CNS elicits a differential selective pressure for or against an intact FCS in Spike, respectively.