Shifts in receptors during submergence of an encephalitic arbovirus

,


Article
re-emerged in Argentina and Uruguay in November 2023, with more than 2,400 equid cases (clinically diagnosed and laboratory confirmed), 103 human cases and 10 human fatalities as of 31 March 2024 12,[15][16][17] .The molecular determinants of WEEV infectivity in mammalian cells remain unknown, as do factors that drive the marked phenotypic change in WEEV strains isolated over the past century.A better understanding of determinants of WEEV infection of mammalian cells is urgently needed for risk assessment of re-emerging strains and to facilitate outbreak preparedness.

PCDH10 is a WEEV cellular receptor
The alphavirus genome encodes four nonstructural proteins (nsP1-nsP4) and six structural proteins (capsid, E3, E2, 6K, TF and E1).WEEV strains isolated in North America are divided into two lineages: group A and group B 1 (Extended Data Fig. 1 and Supplementary Table 1).Group B is further subdivided into B1, B2 and B3 1 .Group A contains ancestral WEEV strains, several of which have been demonstrated to be highly pathogenic in mammals 2,9,18 .The B1, B2 and B3 sublineages contain moderately pathogenic strains, and B3 contains at least one strain that is nonpathogenic in animal models [1][2][3]18 . An volutionary analysis suggests rapid displacement of the earlier groups by later ones, with B3 being the only group that is known to circulate in North America today 1,3 .
To identify WEEV cellular receptors on human cells, we performed a pooled CRISPR-Cas9 knockout screen with guide RNAs (10 per gene) targeting membrane-associated proteins in HEK 293T (human kidney epithelial) cells stably expressing Cas9 (Extended Data Fig. 2a).We used a single-cycle reporter virus particle (RVP) system 4 in which a genomic RNA that encodes the Old World alphavirus Ross River virus nsP1-nsP4, capsid and a reporter were packaged into virions bearing surface glycoproteins of heterologous alphaviruses.For the screen, we chose group B2 WEEV strain 71V1658 (71V), which was originally isolated from an infected horse in Oregon, USA in 1971 19 .This strain does not recognize VLDLR, ApoER2 or low-density lipoprotein receptor (LDLR) class A domain-containing 3 (LDLRAD3), which serve as receptors for other encephalitic alphaviruses 4 .The CRISPR knockout screen identified PCDH10 as the top candidate receptor according to robust rank aggregation 20 (Fig. 1a and Supplementary Table 2).PCDH10 is a δ2-protocadherin that is expressed in several peripheral tissues, but is notably enriched in the brain, where it participates in synapse development [21][22][23][24] .Mutations in PCDH10 have been linked to autism-spectrum disorders 25 .Protocadherins have no structural similarity to any previously known alphavirus receptors 4,26,27 .
Clonal PCDH10-knockout HEK 293T cells became resistant to infection by WEEV 71V RVPs encoding GFP, and infection could be rescued by PCDH10 overexpression (Fig. 1b and Extended Data Fig. 3a,b).Infection by WEEV 71V RVPs of HEK 293T cells, SK-N-SH (a human neuroblast cell line) and SVG-A (a human astrocyte cell line) could also be blocked by polyclonal antibodies against PCDH10, but not by an isotype control antibody (Fig. 1c,d and Extended Data Fig. 3c).K562 cells, a human lymphoblast cell line 28 that is refractory to entry of all tested alphaviruses 4 , do not express PCDH10 on their surface (Extended Data Fig. 3d).PCDH10 overexpression resulted in robust WEEV 71V RVP infection of these cells (Fig. 1e).Overexpression of VLDLR, a receptor for EEEV, Semliki Forest virus (SFV) and Sindbis virus 4 (SINV) or of LDLRAD3, a VEEV receptor 26 , did not affect WEEV 71V RVP entry (Fig. 1e,f and Extended Data Fig. 3e).
Owing to the expected rigidity of the PCDH10 ectodomain, we hypothesized that the most membrane-distal repeats (EC1 or EC2) would be the most likely sites of WEEV attachment.A PCDH10 construct in which EC1 is deleted could not support WEEV 71V RVP infection when ectopically expressed on K562 cells, suggesting that WEEV binds PCDH10 EC1 (Fig. 2a,d and Extended Data Fig. 4b).We generated single extracellular cadherin constructs in which the ectodomain was replaced by either EC1 or EC2 (Fig. 2a).Overexpression of the single EC1, but not the single EC2 construct or a stalk-only control construct, rendered K562 cells susceptible to infection by WEEV 71V RVPs (Fig. 2e and Extended Data Fig. 4c).
Using biolayer interferometry, we tested whether PCDH10 EC1 -Fc could directly bind purified WEEV virus-like particles (VLPs).The source sequence strain for the WEEV VLPs (CBA87 29 ) was isolated in 1958 from an infected horse in Córdoba, Argentina 11 .This strain has been used in investigational VLP-based vaccines against encephalitic alphaviruses modified for high-yield expression 29 .Sensor tips coated with PCD-H10 EC1 -Fc, but not VLDLR LBD -Fc or a control human IgG, bound WEEV CBA87 VLPs (Fig. 2g).Thus, PCDH10 EC1 is the site of WEEV attachment.

PCDH10 mediates attachment and uptake
We tested whether PCDH10 could support cell surface attachment and internalization of WEEV VLPs (strain CBA87) using live-cell confocal microscopy.As a control, we included VLPs for chikungunya virus (CHIKV), an Old World alphavirus that recognizes human MXRA8 as a receptor 27 .We incubated fluorescently labelled VLPs with transduced K562 cells and found that expression of human PCDH10, but not human MXRA8, allowed WEEV VLP binding to cell surface membranes (Fig. 2h,i and Extended Data Figs. 5 and 6).An increase in particles in the cytoplasm of cells was detected at 37 °C versus 4 °C, suggesting internalization.Conversely, ectopic expression of human MXRA8, but not human PCDH10, increased the number of CHIKV VLPs in the cytoplasm of cells at 37 °C versus 4 °C.Thus, PCDH10 can bind WEEV E2-E1 to mediate virus cell surface attachment and internalization.

VLDLR and ApoER2 as alternative receptors
Our phylogenetic analysis using the structural polyprotein coding sequences of 44 WEEV strains collected over 90 years matched a previous study by Bergren et al. 1 that used the full-length genomic sequences of 33 strains (Extended Data Fig. 1 and Supplementary Table 1).We included the group A ancestral strain California, a WEEV strain retrieved from a horse (San Joaquin Valley, 1930), and the Fleming and McMillan strains, isolated from human cases (1938 and 1941, respectively).We also included two strains that were not examined by Bergren et al.: Y62-33 and CU71-CPA, which are closely related to group A strains and are thought to originate outside of North America, bringing their origins into question.The South American isolates of WEEV that we examined are phylogenetically distinct from North American lineages (Extended Data Fig. 1).
The E2 subunit of the E2-E1 spike protein typically contains binding sites for cellular receptors [30][31][32][33] .Among viral proteins, WEEV E2 sequences vary most in strains that differ in virulence in mammals 1 , and E2 substitutions have been implicated in causing pathogenicity differences between McMillan, a virulent group A strain, and Imperial 181, a group B3 strain that appears to have lost pathogenicity in mammals 18 .To test whether WEEV recognition of PCDH10 may have drifted during viral evolution, we tested whether RVPs bearing E2-E1  spike proteins of select WEEV strains representing groups A, B1, B2 and B3 (14 in total) differ in receptor dependencies using K562 cells expressing human PCDH10.We also included K562 cells that express human VLDLR and ApoER2 because they serve as cellular receptors for EEEV 4 , another encephalitic alphavirus, and used the shorter isoform of ApoER2, which does not contain LDLR class A repeats 4-7 and is a presumed dominant form 34,35 (Extended Data Fig. 3f).
Remarkably, group A strains California (1930), Fleming (1938) and McMillan (1941) could engage not only PCDH10, but also VLDLR and ApoER2 to infect K562 cells (Fig. 3a,b).We observed the same receptor dependencies for the group A strains Y62-33 and CU71-CPA (Extended Data Fig. 7a).All RVPs tested for group B1 and B2 strains and South American strain CBA87 recognized PCDH10, but not VLDLR or ApoER2 (Fig. 3a,b and Extended Data Fig. 7a).Group B3 strains isolated in 1985 and 2001 recognized PCDH10, but two group B3 strains from 2005 did not: Imperial 181, isolated from mosquitoes in California, and R02PV003422B, isolated from mosquitoes in Texas (Fig. 3a,b and Supplementary Table 1).To confirm that RVPs for these two strains had been generated successfully, we used them to infect Vero E6 cells.They yielded similar levels of infection as McMillan strain (group A) RVPs (Extended Data Fig. 7b).
To confirm that PCDH10, VLDLR and ApoER2 can support infection by group A WEEV strains, we rescued replication-competent WEEV Fleming and McMillan from molecular clones, and found they replicated faster and to higher levels in K562 cells expressing PCDH10, VLDLR or ApoER2 than in control cells (Fig. 3c).Alphaviruses usually interact with LDLR-related proteins through the receptors' LBD 4,[31][32][33] .To test whether group A strain WEEV spike proteins interact with the LBD of VLDLR, we purified McMillan VLPs.In biolayer interferometry binding experiments, unlike results with WEEV CBA87 VLPs, sensor tips coated with VLDLR LBD -Fc and PCDH10 EC1 -Fc bound WEEV McMillan VLPs (Extended Data Fig. 8a).
In entry-blocking assays with soluble receptor-Fc fusion proteins using K562 cells ectopically expressing PCDH10, VLDLR LBD -Fc blocked infection by all five group A strains (Extended Data Fig. 8b), suggesting that these strains bind VLDLR in a manner that competes with PCDH10 binding.In the same assays, infection by group B2 strain 71V, which only uses PCDH10 as a receptor, was unaffected, and an Fc fusion protein comprising the ectodomain of human MXRA8 (MXRA8 ect -Fc), used as a negative control, had no effect on viral entry (Extended Data Figs.4f  and 8b).
Because WEEV strains may differ in the efficiency with which they use each alternative cellular receptor, we tested the effect of genetic disruption of PCDH10 or VLDLR on RVP entry for group A strains on HEK 293T cells.HEK 293T cells express PCDH10 and VLDLR (Extended Data Figs.3b and 8c,d) but do not express ApoER2 4 .RVPs for California, McMillan and CU71-CPA were impacted in clonal PCDH10-knockout HEK 293T cells but not in clonal VLDLR-knockout HEK 293T cells, suggesting that they depend on PCDH10 to infect HEK 293T cells (Extended Data Fig. 8e).Fleming was only partially affected by PCDH10 knockout or VLDLR knockout, whereas Y62-33 was minimally affected by knockout of either receptor (Extended Data Fig. 8e).Thus, Fleming and Y62-33 are likely to use PCDH10 and VLDLR with similar efficiency on this cell type.Group B strains BFS932 and 71V, which can bind PCDH10 but not VLDLR, were affected by PCDH10 knockout but not by VLDLR knockout.

Receptors determine WEEV neurotropism
PCDH10, VLDLR and ApoER2 are expressed on cells in the central nervous system 21,22,36 .We tested whether PCDH10 and receptors in the LDLR family could redundantly direct neurotropism of group A WEEV strains and whether PCDH10 is the determinant of neurotropism for group B strains that recognize only this receptor.We infected cortical neurons isolated from postnatal day 1 or day 2 wild-type and Pcdh10 −/− mice 21 (Extended Data Fig. 9a) with WEEV RVPs for the group A strain McMillan and the group B strain 71V.We included RVPs for SFV, which depend only on LDLR-family receptors to infect mouse cortical neurons 4 , as a control virus.Infection was performed in the absence or presence of receptor-associated protein (RAP) (Extended Data Fig. 4g), a near-universal ligand antagonist for LDLR family members that can block alphavirus E2-E1 spike protein binding to VLDLR or ApoER2 4,37,38 .
WEEV 71V RVPs robustly infected wild-type neurons, but infection was almost completely abolished in Pcdh10 −/− neurons (Fig. 4a,b and Extended Data Fig. 9b).RAP addition did not affect WEEV 71V RVP infection of wild-type neurons.Thus, for a WEEV strain that only recognizes PCDH10 and not LDLR-family receptors, PCDH10 is the sole determinant of neuronal infection.McMillan RVPs robustly infected wild-type neurons; however, infection was reduced but not abolished in Pcdh10 −/− neurons (Fig. 4a,b and Extended Data Fig. 9b), suggesting that LDLR-related proteins could compensate for the absence of PCDH10 on   9b), indicating that access to PCDH10 as a receptor was sufficient for infection.As expected, RAP blocked SFV infection of both genotypes of primary cortical neurons (Fig. 4a,b and Extended Data Fig. 9b).
Collectively, these results show that PCDH10 and LDLR-family receptors redundantly support neurotropism of the McMillan strain and that 71V, a group B2 strain that does not bind LDLR-family members, depends on PCDH10 as a primary receptor on mouse cortical neurons.

Decoy protects against lethal challenge
Our finding that VLDLR LBD -Fc could block WEEV McMillan RVP infection of K562 cells stably expressing PCDH10 suggests that group A strains recognize PCDH10 and VLDLR through an overlapping surface on their E2-E1 spike proteins (Extended Data Fig. 8b).Additionally, we found that PCDH10 EC1 -Fc could block McMillan RVP infection of K562 cells expressing VLDLR or ApoER2 (Fig. 4c) and infection of Vero E6 cells by replication-competent WEEV McMillan (Fig. 4d), confirming that PCDH10 EC1 -Fc can block access to multiple receptors that redundantly support WEEV infection.
We chose the WEEV McMillan strain for in vivo experiments because it is the most well-studied group A strain in vivo 2,11,18 .Subcutaneous challenge with McMillan strain is lethal in CD1 mice, with most if not all animals succumbing within five days 2 .When five-to six-week-old CD1 mice were subcutaneously inoculated in the footpad with 1,000 plaque-forming units (PFU) 6 h after treatment with phosphate-buffered saline or an isotype control IgG, most mice succumbed (became moribund, meeting euthanasia criteria) within 6 days (Fig. 4e).However, all mice treated with PCDH10 EC1 -Fc survived infection.Therefore, PCDH10 EC1 -Fc can protect mice against a highly virulent WEEV strain that can engage multiple receptors on brain cells.
Orthologues of VLDLR and ApoER2 from various species can support infection by EEEV, SFV and SINV to varying degrees, potentially explaining the wide host range of these alphaviruses 4 .We tested whether the group A WEEV strains could have engaged VLDLR and ApoER2 during infection of human, avian, equine and mosquito hosts.We found that RVPs of two group A WEEV strains, McMillan and Fleming, could infect K562 cells expressing the VLDLR orthologues of horses, starlings (Sturnus vulgaris) and mosquitoes (Aedes aegypti), as well as ApoER2 orthologues of mice, horses and starlings (Extended Data Fig. 10d-f).Group B2 WEEV strain 71V did not recognize any tested VLDLR or ApoER2 orthologues (Extended Data Fig. 10g,h).

Imperial 181 strain binds avian PCDH10
Group B3 WEEV strain Imperial 181, isolated in 2005 from Culex tarsalis mosquitoes in Imperial County, California, causes no mortality in inoculated mice or Syrian hamsters 2 .This strain does not recognize human PCDH10, VLDLR, or ApoER2 (Fig. 3a,b).Because the Imperial 181 strain has similar fitness in house sparrows as the group B1 WEEV strain BFS932 3 , which binds human PCDH10 (Fig. 3a,b), we suspected that avian hosts express cellular receptors that Imperial 181 recognizes.Indeed, K562 cells transduced to express sparrow PCDH10 were rendered permissive to infection with Imperial 181 RVP strain (Fig. 5c).K562 cells expressing common garter snake PCDH10 were similarly rendered permissive to WEEV Imperial 181 RVPs (Fig. 5c), suggesting that recognition of reptilian PCDH10 also has been preserved in this strain.Whereas mammalian orthologues of MXRA8 are receptors for CHIKV and some other alphaviruses 27 , a recent study found that avian orthologues of MXRA8 can serve as receptors for WEEV, SINV and other WEE complex alphaviruses with avian reservoirs 43 .That study examined only WEEV E2-E1 spike protein sequences for CBA87 (South America) and McMillan (group A) strains.We found that overexpression of sparrow MXRA8, but not human MXRA8, could mediate entry of WEEV Imperial 181 RVPs and 71V (group B2) strain RVPs (Fig. 5d and Extended Data Fig. 10c), suggesting that recognition of avian MXRA8 has similarly been preserved.

Discussion
WEEV submergence has been marked by a decline in mammalian virulence 2,3 (Extended Data Fig. 11e), but the precise molecular drivers remain unknown.Here we show that PCDH10 is a general receptor for WEEV, but a WEEV strain that was isolated recently in North America has lost the ability to recognize human PCDH10 but retains the ability to bind avian and reptilian PCDH10.Although PCDH10 is not structurally related to the LDLR-related receptors that are implicated in the entry of multiple alphaviruses 4,26 , PCDH1 (a δ1-protocadherin expressed in the lung) is a receptor for certain hantaviruses that cause pulmonary syndromes 44 .Thus, tissue distribution of protocadherins may direct tissue-specific tropism and pathogenesis of viruses that recognize this family of proteins as receptors.We tested individual strains of EEEV, VEEV, SFV, SINV and CHIKV, but cannot exclude that other strains of these alphaviruses may bind PCDH10.
On the basis of available sequences, we found pathogenic ancestral WEEV strains that caused large-scale outbreaks in the 1930s and 1940s could recognize PCDH10, VLDLR and ApoER2.Of note, EEEV, an alphavirus that continues to cause outbreaks, also binds VLDLR and ApoER2 4 .Source group A WEEV sequences that we used to generate RVPs or rescue molecular clones were derived from isolates that had been serially passaged in suckling mouse brains 1,45 , a common practice in the early twentieth century.It is possible that some group A strains acquired the ability to bind LDLR-related proteins during this passaging.However, additional strains we used that only bind human PCDH10 had been similarly passaged, albeit less extensively (Supplementary Table 1).Nonetheless, our experiments suggest that the ability to acquire the use of VLDLR and ApoER2 as cellular receptors is likely to be an intrinsic, ancestral property of the WEEV spike protein that should be considered during risk assessment of emerging strains and in the development of countermeasures.
The inability to bind mammalian PCDH10, VLDLR or ApoER2 may be associated with decreased virulence in mouse models of recent WEEV strains compared with ancestral strains 2,3,18 .A study showed that three strains that we found to recognize PCDH10 only (BFS2005, 71V and 85-452NM) do not have significantly different virulence, despite being isolated over a period of three decades 2 (Extended Data Fig. 11e).Strains of distinct lineages (B1, B2, B3 and South American) that recognize PCDH10 but not LDLR-related receptors have been found to be similarly virulent in mice 2,11 (Extended Data Fig. 11e).These studies and our findings suggest that mammalian virulence of WEEV is associated with receptor usage patterns and not necessarily time of isolation or phylogenetic relationship.
A recent study using SFV suggests that binding to VLDLR and ApoER2 is important for neuroinvasion in mice 46 .Determining how the ability to bind mammalian PCDH10 versus LDLR-family members influences WEEV strain virulence would require additional experiments, including in vivo studies with strains that contain E2 and E1 spike protein mutations that selectively ablate the ability to bind PCDH10 or LDLR-related proteins.
Notably, the group B3 Imperial 181 strain (2005), which does not bind mammalian PCDH10, and group B1 BFS932 (1946)  were found to be equally fit in competition assays performed in house sparrows 3 .The ability to recognize avian orthologues of PCDH10 and MXRA8 potentially provides a mechanism for WEEV to maintain fitness in its avian reservoir upon losing the ability to bind mammalian receptors.Additional studies will be required to clarify the relative roles of WEEV binding to PCDH10 or MXRA8 in infection of avian hosts.
Neither Imperial 181 (California) nor R02PV003422B, another group B3 strain (Texas), binds human PCDH10, suggesting that this phenotypic variant has been geographically dispersed.Whether strains that cannot bind human PCDH10 have become dominant in North America awaits further environmental sampling and surveillance efforts.Of note, because fitness of WEEV for transmission by mosquito vectors among avian reservoir hosts has not significantly decreased 3 , strains that can recognize human PCDH10 as a receptor may be poised for re-emergence.
We also found that the Imperial 181 and R02PV003422B strains could infect Vero E6 cells (Extended Data Fig. 7b).The EC1 domains of African green monkey (Chlorocebus sabaeus) and human PCDH10 are identical (Extended Data Fig. 10a), suggesting that Imperial 181 and R02PV003422B strains are also unlikely to bind C. sabaeus PCDH10.PCDH10 polyclonal antibodies, which bind Vero E6 cells, and the LDLR-related receptor antagonist RAP, had little effect on infection of this cell type by McMillan, 71V or Imperial 181 RVPs (Extended Data Fig. 7c,d).Therefore, WEEV is likely to be able to enter Vero cells using a pathway that is independent from VLDLR, ApoER2 or PCDH10.
It is unclear whether the lack of binding of contemporary WEEV strains to mammalian orthologues of PCDH10 has occurred through genetic drift or as a result of evolutionary pressure.Over the course of its evolution, WEEV may have had decreasing ecological opportunities for epizootic circulation because of the industrialization of agriculture, fewer horses on farmlands, and vaccination in horses 3 .Because WEEV could be maintained in enzootic circulation independent of epizootic hosts, recognition of receptor orthologues in epizootic hosts no longer conferred an evolutionary advantage, if there was any (for example, through a potential role of equids as amplification hosts) 3,47,48 , and became susceptible to elimination owing to drift.
LDLR was recently implicated as a receptor for multiple alphaviruses, and as a low affinity receptor for WEEV (strain CBA87) and EEEV 49,50 .We tested whether WEEV McMillan, a group A strain that binds VLDLR and ApoER2, could also bind LDLR.We found that LDLR overexpression could enhance entry of WEEV McMillan RVPs, albeit to limited extents (Extended Data Fig. 11a,b), suggesting that McMillan, and possibly other group A WEEV strains, may engage LDLR with low affinity to promote viral entry into neuronal cell types that express this receptor.This observation is consistent with our observation that McMillan RVPs exhibited some residual entry in Pcdh10 −/− mouse neurons treated with RAP (Fig. 4a,b), a protein that antagonizes VLDLR and ApoER2 but not LDLR itself when added exogenously [51][52][53][54] .However, we observed little to no effects of ectopic LDLR expression on WEEV CBA87 (South American strain), 71V (group B2) or Imperial 181 (group B3) RVPs, or RVPs for two EEEV strains (Extended Data Fig. 11b-d).Differences in multiplicity of infection or systems used to study viral entry (for example, single-cycle RVPs instead of replication-competent chimeric SINV) may account for these differences.
Considering the recent outbreak in South America, our results may provide guidance for estimating the threat of re-emerging strains.We demonstrated that South American strain WEEV CBA87 (Argentina, 1958) binds human PCDH10, but not VLDLR or ApoER2, suggesting that human PCDH10 is a receptor for WEEV strains that have circulated in South America.
Several contemporary WEEV strains have lost the ability to bind multiple mammalian receptors over the course of WEEV evolution, providing a potential molecular basis for decreased mammalian virulence accompanying the decrease in epizootic activity of a major encephalitic arbovirus in North America.We propose that the inability of contemporary North American strains to recognize mammalian orthologues of PCDH10 is a molecular hallmark of the epizootic decline of WEEV.Outbreak preparedness could be bolstered by future studies to identify the key sequence polymorphisms in the WEEV spike E2 and E1 proteins that explain shifted receptor compatibilities.

Online content
Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41586-024-07740-2.
Full-length infectious clones of WEEV Fleming and WEEV McMillan have been previously described 55 and were provided by W. Klimstra.Plasmids were transformed into TOP10 Escherichia coli (Invitrogen) and prepared using the Plasmid Plus Midi or Maxi kits according to the manufacturer's protocol (Qiagen).Linearization of 10 μg of plasmid was achieved with the NotI-HF restriction enzyme (NEB), followed by phenol-chloroform extraction.WEEV RNA was in vitro transcribed using the mMESSAGE mMACHINE T7 kit (Invitrogen) with 1 μg linearized plasmid.Following RNA transcription, two confluent T-150 or T-175 flasks of Vero 81 (for Fleming) or Vero E6 (for McMillan) cells were detached with 0.25% trypsin-EDTA (Gibco) and washed three times with Dulbecco's phosphate-buffered saline (DPBS).Following the final wash, cells were resuspended in DPBS and combined with the entire volume of transcribed RNA in a 4 mm gap cuvette.The cells and RNA were subjected to three 250 V, 10 ms pulses at 1 s intervals in an ECM 830 square wave electroporation system (BTX).Cells were allowed to rest for approximately 10 min at room temperature before being transferred to a T-75 flask in the presence of medium with reduced FBS and maintained at 37 °C with 5% CO 2 .Upon onset of cytopathic effect two days post-electroporation, cellular debris was pelleted by centrifugation and viral stocks were collected and stored at −80 °C.

Primary mouse cortical neuron culture and infection
Mouse experiments were approved under the Harvard Medical School Institutional Animal Care and Use Committee (protocol number IS00002530-3), and under the Boston Children's Hospital Institutional Animal Care and Use Committee (protocol number 00001725).The Pcdh10-knockout mouse line was maintained on a C57BL/6J background 21 .Postnatal day 0 or day 1 pups were genotyped by genomic PCR, in which a fragment of the wild-type allele was amplified by primer P1 (5′-GCTCGCGTTTGCCAGCCGTTGATATC-3′) and primer P2 (5′-AGAGCGTCTCCAAATCGAGCCTCATT-3′), and a fragment of the mutant (null) allele was amplified by primer P1 and primer P3 (5′-ACTGGTACACGCGACTGAAAACAGTG-3′).Cortical neurons were dissected and dissociated from postnatal day 1 or 2 neonates using methods adapted from ref. 56.In brief, pups were anaesthetized on ice and euthanized by decapitation.The cortices were then isolated in cold HBSS and dissociated in HBSS supplemented with 20 units ml −1 of papain (Worthington Biochemicals) and 2000 units ml −1 of DNase I (Roche).During dissociation, the cortices were first incubated at 37 °C for 5 min following trituration.Following dissociation, the papain was neutralized with 10 mg ml −1 ovomucoid inhibitor (Worthington Biochemicals) in HBSS.Cells were then washed once with neurobasal medium by centrifugation at 600g for 3 min and plated at a density of 100,000 cells per well in 96-well plates (Cellvis) coated with 20 μg ml −1 poly-l-lysine (Sigma) and 4 μg ml −1 laminin (Thermo Fisher).The neurons were maintained in neurobasal medium supplemented with B27 (Thermo Fisher), l-glutamine and penicillin-streptomycin, unless specified otherwise.The plated neurons were treated with 3 μM cytosine arabinoside (AraC) from day 1 post-plating (day in vitro 1 (DIV 1)) to DIV 3 to reduce non-neuronal cell outgrowth.On DIV 4 we pre-incubated WEEV or SFV RVPs with 100 μg ml −1 transferrin or RAP in culture medium containing 5 μg ml −1 polybrene for 30 min at 37 °C.We then added the mixtures to cells.Cells were imaged every 4 h for 24 h using the Incucyte S3 Live Cell Imaging system (Sartorius) with Incucyte S3 Software version 2022B Rev2 (Sartorius) using a 20× objective.GFP-positive neurons were scored as cells with a threshold signal greater than 5 green calibrated units (GCU) above background, using a Top-hat background subtraction method.The neuronal cell body area in each image was obtained by analysing phase-contrast images using the Incucyte S3 Software.To calculate the percentage of positive cells, at the time point of 24 h post-infection, the area of GFP signal above background was divided by the total area covered by neuronal cell bodies and was multiplied by 100.We calculated relative infection as follows: Relative infection (%) for wild-type neurons = (percentage of GFP-positive wild-type cells in the presence of transferrin or RAP)/(percentage of GFP-positive wild-type cells in the absence of transferrin or RAP) × 100; relative infection (%) for Pcdh10 −/− neurons = (percentage GFP-positive Pcdh10 −/− cells in the presence or absence of transferrin or RAP)/(percentage GFP-positive wild-type cells in the absence of transferrin or RAP) × 100.

Reporter virus particle titration
Titration of GFP-expressing RVPs was performed on Vero E6 cells seeded in 96-well plates using a serial twofold or tenfold dilution of the RVP stocks.At 24 h post-infection, numbers of GFP-positive cells were counted using fluorescence microscopy and used to calculate RVP titre as infectious unit per millilitre (IU ml −1 ), assuming that at high are found in the image metadata and image legend where applicable).Fluorescence from each fluorophore was acquired sequentially at each z-step of the confocal to improve the axial precision of the measurements.Nikon NIS-Elements Advanced Research (AR) 5.02 acquisition software was used to acquire the data, and the files were exported in ND2 file format.Figures were generated using Fiji 66 .A Gaussian filter of σ = 1 was applied to the image to smooth single pixel noise before adjustment of brightness and contrast.Max intensity projection (MPI) renderings were created by using the 3D projection function (Stacks>3D Project) with 10° increments and interpolation selected to smooth the 3D rendering.
3D image analysis was performed using custom pipelines built in Arivis 4DFusion 4.0 analysis software.We detected VLPs through a particle enhancement denoising filter of diameter 0.6 μm followed by a dilation morphology filter of diameter 0.13 μm (sphere shaped).We then applied a Blob Finder segmentation filter of diameter 0.52 μm, a probability threshold of 21.24%, and a split sensitivity of 50%.We segmented cellular compartments by first applying an enhance edges filter within the membrane detection operation, with a membrane width of 0.9 μm and a gap size of 0.6 μm, to enhance the AF488 signal.A discrete Gaussian denoising filter of diameter 0.2 μm was then applied.The membrane-based segmentation operation with a split sensitivity of 30% and a maximum diameter of 50 μm was executed to segment the processed image, and the whole cell masks were obtained by additional feature filters of sphericity >0.58 and volume >20 μm 3 .The cytoplasm mask was created by eroding the cell mask by two pixels.The membrane + cytoplasm mask was created by dilating the cell mask by three pixels, and the membrane mask was obtained by subtracting the membrane + cytoplasm mask with the cytoplasm mask.Finally, to remove segments created based on cells cut off at the edges of the imaged volume as well as cellular blebs that were segmented as independent cells, we applied a volume filter to exclude cytoplasm segments with a volume of <500 μm 3 as well as their corresponding membrane segments.The number of VLPs in each compartment was then calculated by combining all masks.

Phylogenetic analysis
Sequences encoding the structural polyprotein (C-E3-E2-(6 K/TF)-E1) of 44 WEEV strains with full genome sequences available (Supplementary Table 1) were aligned in MEGA11 using the built-in MUSCLE algorithm 67 .A maximum-likelihood phylogenetic tree was constructed using the aligned sequences with the Tamura 3-parameter nucleotide substitution model.The bootstrap method was used to test phylogeny with 1000 bootstrap replications.
Replication kinetics assay with authentic WEEV K562 cells (2.5 × 10 6 ) transduced with an empty lentiGuide-Puro vector or transduced to overexpress MXRA8, PCDH10, VLDLR, or ApoER2 were pelleted by centrifugation at 450g for 2 min.The medium was discarded and the cell pellets were gently resuspended in 1 ml of solution containing WEEV (strain McMillan or Fleming) diluted to 2.5 × 10 4 PFU/ml in maintenance medium (RPMI1640 supplemented with 2% (v/v) FBS, 25 mM HEPES, 1% (v/v) penicillin-streptomycin) for a MOI of 0.01.The infection was allowed to proceed for 1 h at 37 °C with 5% CO 2 .Following the infection, the cells were washed three times with 5 ml DPBS (Sigma) by centrifugation as above.Finally, the cells were resuspended in 5 ml maintenance medium.Immediately following this final resuspension, and again at 6, 12, 24 and 48 h post-infection, 500 μl supernatant was collected from each sample and stored at −80 °C.The removed volume was replaced with 500 μl fresh maintenance medium, and the samples were returned to the incubator.Sample titres were determined by plaque assay on Vero E6 cells.

Plaque reduction neutralization assay
For plaque neutralization assays with infectious WEEV (strain McMillan), PCDH10 EC1 -Fc or MXRA8 ect -Fc fusion proteins were serially diluted in PBS supplemented with 2% (v/v) FBS to concentrations of 150-0.073μg ml −1 .Diluted proteins (or diluent alone as a control) were combined with an equal volume of virus containing approximately 30 PFU of WEEV McMillan.The mixture of protein and virus was incubated at 37 °C with 5% CO 2 for one hour.Following this incubation, samples were processed as described for the plaque assay.Per cent neutralization was calculated as follows: per cent neutralization = (1 - (number of plaques in an experimental well/average number of plaques in diluent-only control wells)) × 100%.

In vivo protection study
Mouse studies were performed as approved by the University of Texas Medical Branch Institutional Animal Care and Use Committee (protocol number 1708051) in accordance with the NIH Guidance for the Care and Use of Laboratory Animals.Mice were fed a 19% protein diet (Teklad, 2919, Irradiated), had 12 h light:dark cycle (06:00-18:00), and were housed in a facility maintained at a temperature range of 20 to 26 °C with a humidity range of 30 to 70%.Food and water were provided ad libitum.Sample sizes for mouse studies were determined based on previously published results for similar in vivo experiments 4 .Randomly assigned mixed-sex cohorts (n = 5 female and n = 5 male) of 6-week-old CD1 IGS mice (Charles River) received a 50 mg kg −1 dose of either PCDH10 EC1 -Fc, control IgG (C1A-H12 anti-SARS-CoV-2 spike antibody) 65 , or PBS via the intraperitoneal route.Six hours later, all mice were infected with 1,000 PFU WEEV McMillan via the subcutaneous route in the left rear footpad.Weights were recorded and health checks were performed daily up to 14 days post-infection.We were not blinded to the treatment or infection status of the mice, also for safety reasons, since WEEV can cause severe disease in humans.

Statistical analysis
Data were deemed statistically significant when P values were <0.05 using version 10 of GraphPad Prism.Experiments were analysed by one-way or two-way ANOVA with multiple comparison correction, or by log-rank (Mantel-Cox) test in GraphPad Prism.P values are indicated in each of the figure legends.Statistical methods were not used to predetermine sample sizes.

Fig. 4 |
Fig. 4 | PCDH10 is a determinant of WEEV neurotropism and pathogenicity.a, Cortical neurons isolated from wild-type (Pcdh10 +/+ ) or PCDH10-knockout (Pcdh10 −/− ) C57BL/6J mice on postnatal day 1 or 2 were infected with GFPexpressing RVPs bearing E2-E1 spike proteins of WEEV 71V, WEEV McMillan or SFV in the presence of 100 μg ml −1 RAP or transferrin (Tf; control).Representative images taken at 24 h post-infection.Scale bars, 100 μm.b, Quantification of infection of wild-type or Pcdh10 −/− mouse cortical neurons in a using a live-cell imaging system.Relative infection is normalized to infection levels in wild-type neurons without transferrin or RAP.See Methods for additional details.c, K562 cells expressing VLDLR or ApoER2 were infected with GFP-expressing WEEV McMillan RVPs in the presence of 316 μg ml −1 MXRA8 ect -Fc or PCDH10 EC1 -Fc.Infection was quantified by flow cytometry.d, WEEV McMillan plaque reduction neutralization assay with the indicated proteins performed on Vero E6 cells.e, Six-week-old CD1 mice were administered PCDH10 EC1 -Fc fusion protein, an isotype control antibody, or phosphate-buffered saline (PBS) intraperitoneally 6 h before subcutaneous inoculation with 1,000 PFU of WEEV McMillan rescued from a molecular clone.Survival of the mice was monitored daily.Infection of mouse cortical neurons was performed in two independent experiments, each consisting of neurons from two wild-type mice and one Pcdh10 −/− mouse.Data are mean ± s.d.b, Pcdh10 +/+ , n = 8; Pcdh10 −/− n = 6; Twoway ANOVA with Tukey's multiple comparisons test.c,d, Data are mean ± s.d.from 2 experiments performed in triplicates (n = 6).Two-way ANOVA with Dunnett's multiple comparisons test (c) or Šídák's multiple comparisons test (d).e, For PCDH10 EC1 -Fc protection experiment: PBS, n = 10; PCDH10 EC1 -Fc, n = 10; isotype control, n = 10 mice.Log-rank (Mantel-Cox) test comparing PCDH10 EC1 -Fc or isotype control to PBS.NS, not significant.

1 .
1) as an outgroup (not shown).Numbers at nodes indicate bootstrap values.In cases in which the branches are too small bootstrap values may not be shown.Scale bar represents 0.01 nucleotide substitutions per site.Groups A, B1, B2, and B3, are indicated.Taxon labels include strain name and year of isolation.Strains examined in this study are bolded and indicated by a triangle.Asterisks indicate strains reportedly isolated outside of North America.Extended Data Fig. 2 | Schematic diagram for CRISPR-Cas9 screen and gating strategy for flow cytometry.a, Screening strategy.HEK 293 T cells were transduced with vesicular stomatitis virus glycoprotein (G) pseudotyped lentivirus encoding the sgRNA library.Cells were infected with CD20-expressing WEEV RVPs (strain 71 V), and infected cells were depleted with magnetic beads coated with an anti-CD20 antibody.Uninfected cells were expanded and re-infected iteratively to increase signal-to-noise ratio.Enriched sgRNAs in CD20-negative cells were identified by next-generation sequencing.b, Flow cytometry gating strategy for quantifying GFP-expressing cells infected with WEEV RVPs (strain 71 V).Wild-type (WT) K562 cells (upper panels) or K562 cells overexpressing PCDH10 (lower panels) were infected with WEEV RVPs and subjected to flow cytometry analysis.c, Flow cytometry gating strategy for monitoring cell surface expression of receptors.Wild-type K562 cells (upper panels) and K562 cells overexpressing human PCDH10 (lower panels) were stained with an anti-PCDH10 antibody, and a PE-conjugated secondary antibody was used for detection.Staining is overlapped in the rightmost panel for comparison.Extended Data Fig. 5 | Gels of purified VLPs and representative confocal images for VLP cell attachment and internalization.a, SDS-PAGE gels of unlabeled or AF647-labeled CHIKV and WEEV VLPs, visualized using a stain-free system (left) or lasers to excite the fluorophore (right).As capsid proteins are protected within the viral membrane, lack of AF647 conjugation to the capsid proteins indicates that VLP membranes are intact.The experiment was performed three times and representative gels are shown.Positions of the E2-E1 glycoproteins and capsid (C) are indicated.NR, nonreducing; R, reducing.b, K562 cells stably expressing human PCDH10 or human MXRA8 were incubated with fluorescently labeled VLPs at 4 °C or 37 °C and then imaged by live cell confocal microscopy.WGA: wheat germ agglutinin.Scale bar is 10 μm.The experiment was performed twice, and representative images are shown.Gel source data for panels a and b are provided in Supplementary Fig. Extended Data Fig. 6 | Workflow diagram of 3D quantification of VLP cell attachment and internalization.a, 3D analysis of multi-color stacks (magenta, VLPs; green, cell membranes) using Arivis 4DFusion.Two custommade pipelines were used to detect VLPs and cellular compartments.b, 3D rendering of VLP stacks (left) and 3D rendering of detected VLPs (right).c, 3D rendering of cellular membranes stacks (left), 3D rendering of the detected cytoplasms (red) overlayed with an enhanced-membrane filter (white) (right, top), and 3D rendering of the detected membranes (yellow) (right, bottom).Objects obtained in each pipeline were combined to quantify the number of VLPs in each cellular compartment.d, 3D rendering of detected objects from the two pipelines combined, used to quantify the number of VLPs in each cellular compartment.Related to Fig. 2i.An example of small objects that could result from non-intact cells at the edge of the imaged volume, or cellular blebs segmented as complete cells, is indicated in the green box.These objects were removed by a volume filter and the VLPs they contain were not counted.See Methods for additional details.Extended Data Fig. 8 | Functional assessment of Group A WEEV strain interactions with PCDH10 and VLDLR.a, Biolayer interferometry sensorgram of WEEV McMillan VLP binding to sensor tips coated with human VLDLR LBD -Fc, PCDH10 EC1 -Fc, or MXRA8 ect -Fc.The experiment was performed twice, and a representative sensorgram is shown.b, K562 cells stably expressing human PCDH10 were infected with GFP-expressing RVPs in the presence of 31.6 μg ml −1 of VLDLR LBD -Fc or MXRA8 ect -Fc.Infection was measured by flow cytometry.c, Cell surface staining of VLDLR and PCDH10 for clonal VLDLR KO HEK 293T cells, as monitored by flow cytometry.d, Cell surface staining of VLDLR for wild-type or clonal PCDH10 KO HEK 293T cells, as monitored by flow cytometry.e, Wild-type, VLDLR KO, or PCDH10 KO HEK 293T cells were infected with GFP-expressing RVPs of the indicated WEEV strains.Infection was measured by flow cytometry.Data are mean ± s.d.from two experiments performed in triplicates (n = 6) (b), or three experiments performed in triplicates (n = 9) (e).Two-way ANOVA with Šídák's multiple comparisons test, ****P < 0.0001 (b).Two-way ANOVA with Dunnett's multiple comparisons test, ***P = 0.0005, ****P < 0.0001 (e).Extended Data Fig. 9 | Mouse genotyping and images for infectivity studies in primary murine cortical neurons.a, Genotyping PCR for two PCDH10 KO (Pcdh10 −/− ) mice used in Fig. 4a,b.Examples of genotyping PCR for wild-type (Pcdh10 +/+ ) and heterozygote (Pcdh10 +/− ) littermates are shown for reference.P1, P2, and P3 are PCR primers used for the reaction (see Methods for additional information).Band patterns for genotyping PCR of Pcdh10 −/− mice are interpreted as previously established 21 .Genotyping was performed once for each pup.b, Merged phase contrast and fluorescent channels for images shown in Fig. 4a, as well as representative images for the mock treatment groups.Images are representative of two independent experiments.Scale bars are 100 μm.Gel source data for panel a are provided in Supplemental Fig. 1.