Plxdc family members are novel receptors for the rhesus monkey rhadinovirus (RRV)

The rhesus monkey rhadinovirus (RRV), a γ2-herpesvirus of rhesus macaques, shares many biological features with the human pathogenic Kaposi’s sarcoma-associated herpesvirus (KSHV). Both viruses, as well as the more distantly related Epstein-Barr virus, engage cellular receptors from the Eph family of receptor tyrosine kinases (Ephs). However, the importance of the Eph interaction for RRV entry varies between cell types suggesting the existence of Eph-independent entry pathways. We therefore aimed to identify additional cellular receptors for RRV by affinity enrichment and mass spectrometry. We identified an additional receptor family, the Plexin domain containing proteins 1 and 2 (Plxdc1/2) that bind the RRV gH/gL glycoprotein complex. Preincubation of RRV with soluble Plxdc2 decoy receptor reduced infection by ~60%, while overexpression of Plxdc1 and 2 dramatically enhanced RRV susceptibility and cell-cell fusion of otherwise marginally permissive Raji cells. While the Plxdc2 interaction is conserved between two RRV strains, 26–95 and 17577, Plxdc1 specifically interacts with RRV 26–95 gH. The Plxdc interaction is mediated by a short motif at the N-terminus of RRV gH that is partially conserved between isolate 26–95 and isolate 17577, but absent in KSHV gH. Mutation of this motif abrogated the interaction with Plxdc1/2 and reduced RRV infection in a cell type-specific manner. Taken together, our findings characterize Plxdc1/2 as novel interaction partners and entry receptors for RRV and support the concept of the N-terminal domain of the gammaherpesviral gH/gL complex as a multifunctional receptor-binding domain. Further, Plxdc1/2 usage defines an important biological difference between KSHV and RRV.


Introduction
The rhesus monkey rhadinovirus (RRV), a member of the genus γ2-herpesvirus or rhadinovirus, is closely related to the only human pathogenic member of this genus, the Kaposi's sarcoma-associated herpesvirus (KSHV) [1,2]. Due to the high similarity in both genome organization and biology RRV is considered as an animal model virus for KSHV [3] and has been used as such in in vitro and in vivo studies. Two major RRV sequence groups have been identified [4], and each is represented by a cloned isolate, RRV 26-95 [5] and RRV 17577 [6]. Analogous to KSHV infection, primary RRV infection is asymptomatic in healthy hosts and leads to life-long persistence, most likely in the B cell compartment [7].
KSHV is associated with a solid tumor of endothelial origin, Kaposi's sarcoma (KS), and two B cell malignancies, primary effusion lymphoma (PEL) and the plasmablastic variant of multicentric Castleman's disease (MCD), most prominently in the context of human immunodeficiency virus (HIV) infection and in immunocompromised individuals. Similarly, simian immunodeficiency virus (SIV)-positive rhesus macaques developed B cell lymphomas upon experimental infection with RRV strain 17577 [8,9] and several studies correlated RRV infection with lymphomagenesis in SIV/SHIV-infected animals [10,11]. While RRV is not consistently associated with solid malignancies, RRV has been identified in retroperitoneal fibromatosis tissue [9,12], similar to retroperitoneal fibromatosis herpesvirus (RFHV) [11], and was recently isolated from hemangioma tissue [13].
Another shared characteristic of KSHV and RRV is the receptor usage on a range of cell types. Both viruses engage members of the Eph family of receptor tyrosine kinases (Ephs) through their glycoprotein (g)H/gL complex to facilitate entry into target cells. While KSHV preferentially interacts with A-type Ephs-specifically EphA2 as the high affinity receptor [14,15]-RRV can utilize both A-and B-type Ephs [15] for entry. These interactions have been characterized on different adherent cells types [16][17][18][19] and we could recently show that both viruses can utilize EphA7 as receptor on BJAB cells [20], a model B lymphocyte line. While for KSHV, in addition to Eph family receptors, several membrane proteins have been proposed as cellular receptors for different viral glycoproteins mediating either attachment or entry on a range of target cells (reviewed in [21]) the receptor usage of RRV is comparatively less well characterized. Nevertheless, studies using receptor knock-down or knockout [14,22], receptorand ligand-mediated blocking [15,23], and Eph de-targeted virus mutants [23,24] showed that both viruses and in particular RRV can infect various cells partially or completely independently of the Eph-interaction, which suggests that RRV engages at least one additional entry receptor that can functionally substitute for the Eph-interaction. This notion is also supported by a recent in vivo study that demonstrated that an RRV mutant deleted of gL and therefore unable to interact with Eph receptors still establishes persistent infection in RRV-naïve rhesus macaques upon intravenous inoculation [24].
We therefore aimed to identify additional rhadinovirus receptors that bind the gH/gL complex or gH and identified Plexin domain containing protein 2 (Plxdc2) as novel interaction partner of the gH/gL complex of RRV, but not KSHV. The closest homolog to Plxdc2, Plxdc1 was initially identified as overexpressed in blood vessels of solid human tumors [25], resulting in the original terminology tumor endothelial marker 7 (TEM7, Plxdc1) and tumor endothelial marker 7 related (TEM7R, Plxdc2) [26]. In general, the physiological functions of Plxdc1/2 are not well understood. Suggestive of a role in development, Plxdc2 has been described as mitogen for neural progenitor cells [27] and expression of both Plxdc1 and Plxdc2 in the developing nervous system has been demonstrated [28,29]. Cortactin, nidogen and the pigment epithelium-derived factor (PEDF) have been described as interactors for Plxdc1 and Plxdc2 [30][31][32]. However, the physiological relevance of these interactions is not fully understood. In this study we characterized the interaction of Plxdc1/2 with the gH/gL glycoprotein complex of RRV and establish Plxdcs as novel cellular RRV entry receptors.

Results
To identify potential cellular receptors for RRV glycoprotein H, we performed immunoprecipitation using soluble RRV 26-95 gH, consisting of the extracellular part fused to the Fc part of human IgG (RRV gH-FcStrep) as bait and 293T whole cell lysate as prey ( Fig 1A). Bands present in the precipitation from 293T whole cell lysate, but not in control precipitation without 293T lysate were excised and analyzed by LC-MS/MS. The most abundant cell surface protein, identified in four of the five analyzed regions, was Plxdc2, also known as TEM7R, a cellular transmembrane protein. Plxdc1/ TEM7, the only homolog of Plxdc2 in humans and rhesus macaques, was not detected in our mass spectrometry analysis (see S1 Co-immunoprecipitation of V5-tagged expression constructs of gH from KSHV and from the two RRV isolates 26-95 and 17577, in the presence or absence of the corresponding Flagtagged gL proteins with myc-tagged human or rhesus Plxdc1 or Plxdc2 (hPlxdc1-myc/ hPlxdc2-myc/ mmPlxdc1-myc/ mmPlxdc2-myc) from transfected 293T cells confirmed the interaction of both RRV gH/gL complexes with Plxdc2 ( Fig 1B). While the predicted molecular weight of Plxdc1 and Plxdc2 is 56kDa and 60kDa, respectively, we and others detect Plxdc1/2 at an apparent molecular weight of~100kDa [31]. Neither KSHV gH/gL nor RRV 17577 gH/ gL interacted detectably with Plxdc1-myc, while RRV 26-95 prominently bound both Plxdc1 and Plxdc2 in the presence and absence of gL. Measurement of the binding of purified hPlxdc1 and hPlxdc2 ectodomain to immobilized RRV 26-95 gH-FcStrep/gL by enzyme-linked immunosorbent assay (Fig 1C and 1D) showed a half-maximal binding concentration of 22.2nM for Plxdc1 and 5.2nM for Plxdc2 as calculated for one site specific binding without pre-defined Hill slope.
To evaluate the effect of Plxdc-binding to RRV gH on the interaction with EphB3, the highaffinity Eph family receptor for RRV gH/gL [15], we used soluble human Plxdc1 or Plxdc2, consisting of the extracellular part of Plxdc1 or Plxdc2 fused to the Fc part of human IgG followed by a TwinStrep tag (hPlxdc1-FcStrep/ hPlxdc2-FcStrep) in immunoprecipitation experiments. Co-immunoprecipitation of hPlxdc1-FcStrep/ hPlxdc2-FcStrep with the gH-V5/ gL-Flag complexes of RRV isolates 26-95 and 17577 in the presence or absence of myc-tagged EphB3 from transfected 293T cells demonstrated the existence of a quaternary complex, indicating the ability of RRV gH/gL to interact with members of both receptor families simultaneously. As the complex was precipitated via Strep-tagged Plxdc1 or Plxdc2 -and Plxdc1/2 alone do not interact with EphB3 -EphB3 will only be co-precipitated when a complex between RRV gH/gL and both EphB3 and Plxdc1 or Plxdc2 is formed (Fig 1E).
While interaction of purified proteins and in transfected cell lysates is strongly suggestive of a functional interaction, the biologically relevant interaction for the entry process would occur with virion gH/gL. To evaluate the functionality of the gH/gL-Plxdc interaction on virus particles both for wildtype RRV and an Eph-binding-negative RRV mutant, we utilized RRV-YFP, an RRV 26-95 strain engineered for constitutive YFP expression upon infection, and RRV-YFP gH-AELAAN, an Eph-binding-negative RRV-YFP mutant that we had previously described (Fig 2A) [23]. To analyze the impact of competition with soluble Plxdc decoy receptor, RRV-YFP wt and RRV-YFP gH-AELAAN preparations were incubated with a concentration series of soluble hPlxdc2-FcStrep or an FcStrep control prior to infection of HaCaT cells (Fig 2B). According to RNA-Seq data of 36 cell lines (courtesy of Human Protein Atlas, www. proteinatlas.org [33], retrieved 2020), HaCaT cells exhibit the highest cell-line specific expression of Plxdc2 among the analyzed non-cancer cell lines and were therefore chosen for further . While preincubation with EphB3-Fc did not reduce RRV-YFP gH-AELAAN infection, preincubation with either hPlxdc2-FcStrep or a combination of hPlxdc2-FcStrep and hEphB3-Fc reduced infection by~50% as observed for RRV-YFP wt infection (Fig 2C, right column group). Infection of SLK cells and rhesus monkey fibroblasts (RF) was also slightly decreased by preincubation of the viral inoculum with hPlxdc2-FcStrep, infection of SLK by RRV-YFP wt to 63.5% ± 3.7% and infection of RF by RRV-YFP wt to 73.9% ± 11.3% relative to preincubation with FcStrep as control. However, this effect was less pronounced than on HaCaT and less pronounced than the effect of hPlxdc2-FcStrep on RRV-YFP gH-AELAAN infection of the same cell types (Fig 2D). Taken together, the immunoprecipitation and blocking experiments confirm the independence of the gH-Plxdc interaction of the previously described Eph interaction motif [23] and EphB3-binding.
To establish receptor function, we performed gain-of-function experiments using ectopic Plxdc1/2 overexpression (Fig 2E-2I). Raji cells were transduced with lentiviruses encoding TwinStrep-tagged human Plxdc1/2 constructs (hPlxdc1-Strep/ hPlxdc2-Strep). A TwinStreptagged human EphA7 construct (hEphA7-Strep) was used as control for Eph-mediated RRV infection ( Fig 2G). The EBV-positive, human lymphoblast cell line only allows for low-level RRV 26-95 infection even with amounts of input virus corresponding to high MOI on adherent cells like SLK, HaCaT or RF. Therefore, changes in susceptibility to infection mediated by Plxdc1/2 overexpression should be readily detectable and allow for a clear differentiation of the contribution of Plxdc1/2 to RRV infection over the very low intrinsic susceptibility to infection. Indeed, ectopic expression of both hPlxdc1-and hPlxdc2-Strep dramatically increased RRV-YFP 26-95 wt infection from~1.4% basal infection to 42.9% and 28.2% infection, respectively (Fig 2E and 2H). Similarly, RRV-YFP gH-AELAAN infection was increased from 0.04% on empty vector-transduced cells to 32.5% and 14.4% on cells transduced with constructs encoding hPlxdc1-Strep and hPlxdc2-Strep, respectively. EphA7, which had previously been described by our group to be critical for RRV infection of BJAB B lymphocytes enhanced RRV-YFP wt infection to 44.4% while infection with the Eph-binding-negative mutant RRV-YFP gH-AELAAN was not affected (Fig 2E and 2H). We did not observe  (Fig 2F and 2I). Taken together, the observed receptor usage on hPlxdc1/2-Strepoverexpressing Raji cells by RRV-YFP 26-95 and RRV-YFP 17577 reflects the Plxdc1/2 interaction patterns of RRV 26-95 and RRV 17577 gH/gL complexes observed in immunoprecipitation experiments (Fig 1B).
In a next step we characterized the Plxdc binding motif on RRV gH. As the RRV 17577 gH-Plxdc2 interaction depends on gL whereas the RRV 26-95 gH-Plxdc2 interaction does not, we focused on the N-terminal domain I of gH which, in analogy to EBV gH/gL [34], most likely constitutes the gL-binding interface. The differences in the Plxdc interaction of RRV isolates 26-95 and 17577 as well as the lack of an interaction between KSHV gH/gL and Plxdc receptors suggested a motif that is only partially conserved between the RRV isolates and missing in KSHV. Using sequence comparisons (Fig 3A) we identified a putative interaction motif spanning 7 or 6 amino acid motif in the N-terminal region of RRV 26-95 gH and RRV 17577 gH, respectively, that is not conserved in KSHV gH. The motif is located close to the Ephinteraction motif we described previously, facing in the opposite direction in a homology model of the RRV 26-95 gH/gL complex based on the EBV gH/gL crystal structure (3PHF) (Fig 3B). Deletion of this motif completely abrogated the interaction with Plxdcs of both RRV 26-95 gH ( Fig 3C) and RRV 17577 gH/gL ( Fig 3D). To further characterize the contribution of individual residues in the 'Tyr(Y)-Glu(E)-Tyr(Y)-Asn(N)-Glu(E)-Glu(E)-Lys(K)' (RRV 26-95) motif we performed single amino acid substitutions to alanine. The ability of mutant RRV 26-95 gH-V5 to bind myc-tagged Plxdc1/2 of human (hPlxdc1/2-myc) (Fig 3E) or rhesus macaque origin (mmPlxdc1/2-myc) (Fig 3F) was analyzed by immunoprecipitation of gH via the V5-tag and Western blot. While several single amino acid substitutions decreased the interaction of RRV gH with Plxdcs to some degree, residues Tyr23 and Glu25 that are conserved in isolates 26-95 and 17577 appear to be critical for the interaction of gH with human and rhesus macaque Plxdc1 and Plxdc2. Furthermore, substitution of glutamate with alanine at position 22, which is not conserved between isolates 26-95 and 17577, had a pronounced, albeit slightly weaker effect on the gH-Plxdc1/2 interaction (Fig 3E and 3F).
To further analyze the contribution of the gH/gL-Plxdc interaction in the context of infection we constructed virus mutants deleted in the seven amino acid interaction motif in the background of RRV-YFP 26-95 wildtype (RRV-YFP gHΔ21-27), and in the background of an RRV-YFP 26-95 strain mutated in the Eph-interaction motif described previously by our group (RRV-YFP gH-AELAAN, RRV-YFP gHΔ21-27-AELAAN) using a two-step, lambda red-mediated recombination system [35] (Fig 4A). Blocking experiments using soluble hPlxdc2-FcStrep decoy receptor on HaCaT cells confirmed that deletion of the seven amino acid motif was sufficient to abrogate the gH-Plxdc2 interaction on viral particles ( Fig 4B). While infection of RRV-YFP wt and RRV-YFP gH-AELAAN was inhibited by~60% and 70% respectively, infection of RRV-YFP gHΔ21-27 and RRV-YFP gHΔ21-27-AELAAN was not affected even by high concentrations of soluble hPlxdc2-FcStrep ( Fig 4B). All infections were carried out at~MOI 0.05. Analogously, the Eph-or Plxdc-receptor-binding-negative RRV mutants were no longer inhibited by preincubation with the respective soluble receptor (hEphB3-Fc or hPlxdc2-FcStrep) in single or double inhibition experiments on HaCaT cells ( Fig 4C). To analyze the direct effect of interference with Plxdc receptors on the cellular side for RRV infection we performed an siRNA mediated knock-down of Plxdc2 (S2 Fig). We choose 293T cells for their high transfection efficiency and confirmed Plxdc2 expression by qPCR, in accordance with our initial mass spectrometry experiment ( Fig 1A and S1 Fig). The

Fig 3. An amino acid sequence motif in the N-terminal region of RRV 26-95 and RRV 17577 gH is essential for the Plxdc interaction. A)
Multiple sequence alignment of the N-terminal region of gH of KSHV and the two RRV isolates 26-95 and 17577. Boxes indicate the binding motives for Eph receptors (green) and Plxdc receptors (blue). Numbers above sequences represent positions in the alignment, asterisks indicate conserved amino acids. Individual amino acid positions are given to the left and right of the respective gH sequence. B) Homology-based structure prediction of the RRV 26-95 gH/gL complex based on the crystal structure of the EBV gH/gL complex (PDB number 3PHF) using the Iterative Threading ASSembly Refinement (I-TASSER) server and the CO-THreader (COTH) algorithms for protein-protein complex structure and multi-chain protein threading. The Eph receptor interaction motif is shown in green, the Plxdc interaction motif is shown in blue, gL is shown in red, gH is shown in grey. C) Deletion of the Plxdc interaction motif in RRV 26-95 gH (amino acid 21-27, "YEYNEEK") abrogates gH interaction with Plxdc1 and Plxdc2. V5-tagged gH wt or gHΔ21-27 were immunoprecipitated in the presence of full-length human or Macaca mulatta Plxdc1-myc or Plxdc2-myc using monoclonal antibody to the V5-tag. Precipitates were analyzed by Western blot. D) Deletion of the Plxdc interaction motif in RRV 17577 gH (amino acid 21-26, "YVYDEK") abrogates gH interaction with Plxdc2. V5-tagged gH wt or gHΔ21-26 was co-expressed with Flag-tagged RRV 17577 gL. gH-V5/gL-Flag complexes were immunoprecipitated in the presence of full-length human or Macaca mulatta Plxdc1-myc or Plxdc2-myc using monoclonal antibody to the V5-tag. Precipitates were analyzed by Western blot. E) Mutational scan of the Plxdc interaction motif (amino acid 21-27, "YEYNEEK") of RRV 26-95 gH identifies human Plxdc1/2-interacting residues. V5-tagged gH mutants were immunoprecipitated in the presence of full-length human Plxdc1-myc Similarly, serum SY8513 reduced RRV-YFP wt and RRV-YFP gH-AELAAN infection by 34.6% ± 1.6% and 40.0% ± 5.7% compared to pre-immune serum. While pre-treatment of 293T cells with serum SY8513 did not significantly affect infection with RRV-YFP gHΔ21-27, it lead to a 30.6% ± 6.9% reduction of infection with the Plxdcbinding negative RRV-YFP gHΔ21-27-AELAAN (S2E Fig). Overall, we observed a weak reduction of infection by RRV gH-AELAAN by siRNA mediated knockdown of Plxdc2 expression, and comparatively potent and specific inhibition by one out of two rabbit antisera to Plxdc2.
The receptor-specificity conveyed by the respective interaction motif was further analyzed in lentiviral vector-mediated Plxdc1/2-Strep overexpression experiments in Raji B lymphocytes. EphA7 was used as control for Eph-mediated infection. Expression of Plxdc1/2-Strep as well as EphA7-Strep dramatically enhanced susceptibility of Raji cells (Fig 4D and 4E). RRV-YFP wt infection increased from 0.14 ± 0.04% on empty vector transduced Raji cells tõ 8.5% upon EphA7 overexpression and to~17% upon Plxdc1/2 overexpression, without pronounced differences between the Plxdc family members. Mutation of the Eph-interaction motif in RRV-YFP gH-AELAAN completely abrogated the gain in susceptibility on EphA7 overexpressing cells while mutation of the Plxdc-interaction motif completely abrogated the gain in susceptibility on Plxdc1/2 overexpressing cells, confirming selective knockout of each individual receptor interaction in the respective mutant. Furthermore, deletion of the Ephinteraction motif in RRV-YFP gH-AELAAN did not impact the infection of Plxdc1/2 overexpressing cells in comparison to RRV-YFP wt infection. In contrast, we observed an~2-fold higher infection of RRV-YFP gHΔ21-27 on EphA7 overexpression cells, when compared to RRV-YFP wt. Together, the blocking and overexpression experiments indicate an independent rather than cooperative nature of Eph and Plxdc receptor function.
To quantitatively analyze the contribution of the Plxdc1/2-interaction to RRV infection of different cell types, RRV-YFP wt and RRV-YFP receptor binding mutant inocula were normalized to genome copies as determined by qPCR, and target cells were inoculated with the same number of encapsidated input virus genomes for wt and each mutant virus strain. At least three (four for SLK, HaCaT, Raji, MMB1845, five for MFB5487) independent sets of RRV wt and mutant stocks were used in infection experiments at various dilutions to compensate or Plxdc2-myc using monoclonal antibody to the V5-tag. Precipitates were analyzed by Western blot. RRV gHΔ21-27 serves as negative control. F) Mutational scan of the Plxdc interaction motif (amino acid 21-27, "YEYNEEK") of RRV 26-95 gH identifies rhesus macaque Plxdc1/2-interacting residues. V5-tagged gH mutants were immunoprecipitated in the presence of full-length Macaca mulatta Plxdc1-myc or Plxdc2-myc using monoclonal antibody to the V5-tag. Precipitates were analyzed by Western blot. RRV gHΔ21-27 serves as negative control. Abbreviations: IP: immunoprecipitation, IB: immunoblotting, h: human, mm: Macaca mulatta (rhesus macaque).
https://doi.org/10.1371/journal.ppat.1008979.g003 for variability in stock preparation. For adherent cells, dilutions, which produced an RRV-YFP wt infection in an MOI range of 0.05 to 1 on the respective cell line, were chosen for further analysis. Infection as determined by the percentage of YFP+ cells was normalized to RRV-YFP wt, which was set to 1. For suspension cell lines, experiments with RRV-YFP wt infection over 1% were included in the analysis. None of the analyzed adherent cell lines showed a preferential use of Plxdc receptors over Eph receptors based on the reduction of specific infectivity of Eph-binding and Plxdc-interaction-deficient mutants (Fig 5A). Compared to RRV-YFP wt infection on rhesus monkey fibroblasts (RF), RRV-YFP gHΔ21-27, RRV-YFP gH-AELAAN and RRV-YFP gHΔ21-27-AELAAN exhibited a decrease in infection of~30%, 50% and 70%, respectively. Similarly on HaCaT, infection with RRV-YFP gH-AELAAN and RRV-YFP gHΔ21-27-AELAAN was reduced by~60% and 75%, respectively, compared to RRV-YFP wt. While RRV-YFP gHΔ21-27 exhibited a defect comparable to RRV-YFP gH-AELAAN in three of the analyzed independent sets of RRV stocks, this reduction in infectivity did not reach significance due to one outlier. In agreement with reduced infection of Eph-and Plxdc-binding virus mutants on HaCaT cells, EPHB3 and PLXDC2 expression was detected in HaCaT cells in published datasets (GSE95080, GSE138800) and by confirmatory qPCR of our samples (S3A and S3B Fig). On SLK cells, mutation of the Eph-interaction motif led to an~65% decrease in infection, while RRV-YFP gHΔ21-27 infection was on average comparable to RRV-YFP wt infection. In contrast, we identified B lymphocyte lines of human and macaque origin that exhibit a preference in the receptor usage for either Eph or Plxdc family members (Fig 5B). When normalized to genome copies, RRV-YFP gHΔ21-27 infection on human Raji lymphoblasts was comparable to RRV-YFP wt infection while mutation of the Eph-interaction motif lead to an~90% decrease in infection, indicating preferential infection through Eph family receptors, albeit at low levels, as mentioned above. Correspondingly, with the exception of EPHB4, expression of all members of the Eph and Plxdc receptor families is low to absent in Raji cells, both in a published data set (GSE111880) and our subset analysis by qPCR (S3C and Mutation of both the Eph-and Plxdc-interaction motif (RRV-YFP gHΔ21-27-AELAAN) led to an even more pronounced defect of~90% on MFB5487.
To evaluate the contribution of potential attachment effects on the observed differences in specific infectivity, we analyzed the capacity of virions to bind Plxdc1/2 overexpressing Raji cells in comparison to empty vector transduced Raji cells (Fig 5C and 5D). It should be noted that transduction with an Plxdc1 encoding lentiviral vector invariably resulted in higher levels was measured by flow cytometry. Infection with FcStrep was set to 100% (MOI~0.1-0.2, triplicates, error bars represent SD). D) Raji cells were transduced with TwinStrep-tagged human EphA7, Plxdc1 or Plxdc2 (hEphA7-Strep/ hPlxdc1-Strep/ hPlxdc2-Strep) expression constructs or an empty vector control, briefly selected and infected with RRV-YFP wt, RRV-YFP gH-AELAAN, RRV-YFP gHΔ21-27 or RRV-YFP gHΔ21-27-AELAAN normalized to genome copies as determined by qPCR. YFP expression as indicator of infection was measured by flow cytometry. The mean across three independent sets of RRV stocks is indicated by columns. The means of individual triplicate infections for each set of RRV stocks are given as symbols within the respective columns. E) Micrographs show representative infection of one set of RRV stocks in (D).
Our interaction data (Fig 1A and 1B) suggests that at least RRV 26-95 gH can bind Plxdc1/ 2 in the absence of gL. We therefore aimed to analyze if infection with a RRV 26-95 gL deletion mutant (RRV-YFP ΔgL) is influenced by Plxdc1 and 2 expression to a similar degree as RRV-YFP wt and RRV-YFP gH-AELAAN infection. As shown before, overexpression of hEphA7-Strep or hPlxdc1/2-Strep in Raji cells dramatically increased RRV-YFP wt infection compared to empty vector transduced cells (22-fold for hEphA7-Strep, 23-fold for hPlxdc1-Strep and 26-fold for hPlxdc2-Strep) (Fig 6A and 6B and S4 Fig). RRV-YFP gH-AE-LAAN infection was increased from 0.25% ± 0.05% basal infection on empty vector transduced cells to 52.5% ± 2.06% and 65.5% ± 1.22%, respectively, upon hPlxdc1-Strep or hPlxdc2-Strep overexpression, while hEphA7-Strep overexpression had no pronounced effect on infection. We observed a pattern similar to RRV-YFP gH-AELAAN for the infection with two independent RRV-YFP ΔgL clones, RRV-YFP ΔgL 3-3 and RRV-YFP ΔgL 3-5 . While hEphA7-Strep overexpression did not influence RRV-YFP ΔgL infection, hPlxdc1/2-Strep overexpression enhanced infection from 0.36% ± 0.03% (0.48% ± 0.04%) on empty vector transduced Raji cells to 31.1% ± 0.95% (36.3% ± 1.42%) and 35.6% ± 0.94% (42.6% ± 0.57%) upon hPlxdc1 and hPlxdc2 overexpression for RRV-YFP ΔgL 3-3 (RRV-YFP ΔgL 3-5 ), respectively (Fig 6B). Even though the effect of Plxdc1/2 overexpression on RRV-YFP ΔgL infection was slightly less pronounced than the effect on RRV-YFP wt and RRV-YFP gH-AELAAN MMB1845) or five (MFB5487) independent sets of RRV stocks are given as symbols of different color. Sets with RRV-YFP wt infection exceeding 1% were used for analysis. The maximal achieved RRV-YFP wt infection was 6.1% for Raji, 5.6% for MMB1845 and 3.1% for MFB5487, respectively. The mean across the independent sets of RRV stocks is indicated by black lines.  Our data indicating that Plxdc-mediated RRV entry is gL-independent is interesting as the gH/gL/gB complex is widely regarded as the conserved herpesviral core fusion machinery [21]. Therefore, we sought to determine if cell-cell fusion mediated by RRV glycoproteins is similarly influenced by Plxdcs in an gL-independent manner. We established a quantitative fusion assay using 293T cells transfected with a VP16-Gal4 transcription factor construct as effector cells and Raji cells transduced with a lentivirus encoding a Gal4-response element-driven luciferase construct as target cells (Raji-Gal4-Luc). 293T cells were co-transfected with viral glycoproteins gH, gL and gB as indicated while stable Raji-Gal4-Luc cells were transduced with lentiviral vectors encoding TwinStrep-tagged human receptor constructs. In addition to Plxdc1 and Plxdc2, EphA7 and the high-affinity RRV receptor EphB3 were used as controls for gL-dependent receptor interaction of gH (Fig 6C and 6D). Luciferase activity as indicator of cell-cell fusion was measured after two days of effector-target cell co-culture ( Fig 6E).

Discussion
In this study we identified the Plexin domain containing proteins 1 and 2 as a novel family of entry receptors for RRV. Plxdc1/2 interact with the gH/gL complex of RRV in a region close to the previously characterized binding motif for Eph receptors. While the Eph-interaction as well as the critical Eph binding motif in domain I of gH is conserved between RRV and the closely related human pathogenic KSHV [23], the interaction with Plxdc1/2 is exclusive to RRV and even exhibits differences between isolates 26-95 and 17577 as prototypic members of the two described RRV sequence clades. of transduced Raji target cells (D) was analyzed by Western blot. Expression controls were harvested directly prior to the start of coculture. Luciferase activity as indicator of cell-cell fusion was measured after 48h of effector and target cell co-culture (E). Luciferase activity is normalized to empty vector transfected 293T effector cells (n = 3, error bars represent SD). https://doi.org/10.1371/journal.ppat.1008979.g006

PLOS PATHOGENS
According to our results, differences between Plxdc1 and Plxdc2 may also exist in terms of function. While overexpression of both Plxdc1 and Plxdc2 in Raji cells, that were virtually "non-susceptible" under the conditions used, lead to robust RRV 26-95 infection (Fig 4D and  4E), only overexpression of Plxdc1 enhanced attachment of RRV wt and RRV gH-AELAAN in comparison to either non-transduced Raji cells or Plxdc-binding deficient RRV mutants ( Fig  5D). Whether this is primarily due to differences in expression levels, consistently observed between Plxdc1 and Plxdc2 upon lentiviral overexpression or due to underlying functional differences in the gH interaction with these molecules remains to be determined.
Mutation of residues Tyr23 and Glu25 in gH domain I, which are conserved between RRV isolate 26-95 and 17577, almost abolished the interaction with Plxdc2, which is bound by gH/gL of both isolates. Although the same residues are critical for the interaction of RRV gH 26-95 with Plxdc1, the partial conservation of the region appears to be insufficient to confer binding of 17577 gH to Plxdc1. Interestingly, the interaction of 17577 gH with Plxdc2 is dependent on the presence of gL in the gH/gL complex, whereas 26-95 gH is able to bind Plxdc1/2 independently of gL. This interaction data was corroborated by functional analysis of Plxdc-mediated RRV 26-95 and RRV 17577 infection and fusion. As shown above, RRV 26-95 infection was enhanced by both Plxcd1 and Plxdc2 expression (Fig 2E and 2H) and similarly RRV 26-95 gH/gL or gH alone facilitated fusion with Plxdc1 or Plxdc2 overexpressing Raji cells (Fig 6E). In contrast, recombinant overexpression of Plxdc2, but not of Plxdc1 enhanced infection with RRV 17577 (Fig 2F and 2I), which was paralleled in our fusion assay: RRV 17577 gH/gB/gL effector cells fused readily with Plxdc2 overexpressing target cells, but not in the absence of gL or with Plxcd1 overexpressing target cells (Fig 6E). These features of Plxdc1/2 binding specificities appear similar to the interaction of the gH/gL complex with Eph receptors, wherein mutation of the strictly conserved Eph interaction motif is sufficient to abrogate receptor binding, but differences in KSHV and RRV affinities for A-and B-type Eph RTKs indicate the existence of additional regions in gH or gL that contribute to or modulate the interaction, which is corroborated by a recent crystal structure of KSHV gH/gL in complex with EphA2 [36]. Whether these preferences for different members of conserved receptor families also influence e.g. cell or tissue tropism, viral spread or pathogenicity has not been determined. Sequence comparisons of over twenty RRV isolates identified dramatic differences in the extracellular domains of gH (58.7% amino acid identity between phylogenetic groups) as well as in gL (54.4% amino acid identity between phylogenetic groups) between isolates that fell in two discrete groupings either similar to 26-95 or 17577, while variation in other glycoproteins R1, gM, gN, orf68 was minimal between clades [4]. However, if these clade-specific glycoprotein variations influence the observed differences in pathogenicity between RRV strains 26-95 and 17577 remains to be seen, but such differences have been reported for the representative isolates (summarized in [37]). An interesting coincidence is that the primary sequence of Japanese macaque rhadinovirus (JMRV) [38] gH is highly similar to that of RRV 26-95 with 96% amino acid identity, highly suggestive of conservation of the Plxdc interaction. This virus was found associated with encephalomyelitis in macaques [39], and Plexin domain containing proteins are expressed in brain ( [28] and https://www. proteinatlas.org/), supporting the notion of a certain neurotropism of RV2 rhadinoviruses. Another interesting question would be if the receptor-binding function of the N-terminal region of gH is conserved between RRV and KSHV, and probably EBV, and whether KSHV and EBV can bind another receptor through this region. So far, we did not identify corresponding interactions for KSHV, but in principle the region of gH encompassing the Plxdc2 binding site is present in both viruses and it is tempting to assume some functional conservation.
Along those lines, the evolutionary factors that drive interaction with different receptor families and the resulting multitude of herpesvirus-receptor interactions is highly interesting. For e.g. EBV and HCMV, a clear correlation between receptor usage, dependent on viral interaction partners of the gH/gL complex, and cell tropism has been demonstrated in various studies ( [40], reviewed in [41]). However, for rhadinoviruses, the picture is less clear. We demonstrated that RRV infection of select cell lines exhibits a dependence on specific receptors, e.g. Raji infection-to the modest extent that was possible without recombinant receptor overexpression-was dependent on the gH/gL-Eph interaction, while MFB5487 infection was more dependent on the Plxdc-interaction (Fig 5B). Additionally, mutation of the Eph-interaction motif did not impact infection on all cell types equivalently [23]. Although the complexity and partial redundancy of the RRV interaction with different members of the Eph and Plxdc receptor family complicates the inference of direct correlations between receptor expression and infection, expression data of select Eph receptors in comparison to Plxdcs is in line with the observed receptor preferences on different cell types ( S3 Fig). Yet, a definite correlation between exclusive receptor usage and infection of specific cell types is not obviously apparent. While the notion of a role of different receptor interactions in KSHV and RRV cell and tissue tropism is tempting and an established concept for related viruses [41][42][43], the possibility of a redundant function should not be discarded. Redundancy could be driven by the need to escape antibodies e.g. to one receptor binding site. For instance, in vitro infection of human keratinocytes or rhesus monkey fibroblasts seemed to be impacted to a similar degree by either deletion of the Eph-or of the Plxdc-interaction motif (Fig 5). To ultimately address the correlation between receptor-and tissue-tropism, in vivo studies using receptor-de-targeted mutants to analyze cell and tissue tropisms will be required. The importance of in vivo studies is also supported by a recent report that showed that a gL-null RRV mutant still established persistent infection in the B cell compartment upon intravenous inoculation, while infection of B cells in vitro was drastically reduced [24], a finding that could be explained by the gL-independent usage of Plxdc1/2 for B cell infection in vivo. At a minimum, our data on RRV-YFP ΔgL infection of Plxdc1 and Plxdc2 overexpressing Raji cells (Fig 6B) confirms that RRV can efficiently use Plxdc1 and Plxdc2 as entry receptors in the absence of gL.
Finally, the fact that a mutant that was deleted in both the Eph and the Plxdc1/2 interaction motif, RRV-YFP gHΔ21-27-AELAAN, was still able to replicate on rhesus monkey fibroblasts, as evidenced by the fact that we were able to grow a virus stock on these cells, and was still infectious to a certain degree on a number of cell lines (Figs 4 and 5) indicates a surprisingly high degree of redundancy in the entry pathways that RRV can use and hints at the existence of additional receptors or host factors besides Ephs and Plexin domain containing proteins.
On a more speculative note, the apparent overlap between virus receptors and tumor-associated membrane proteins may represent an interesting research subject. Eph receptors were first identified in an attempt to characterize tyrosine kinases involved in cancer [44] and altered expression in various cancer types has been demonstrated for several Eph family members (reviewed in [45]). Similarly, Plxdc1 was first described in a screen for novel tumor endothelial members [25] and expression of both Plxdc1 and Plxdc2 is elevated in the endothelium of solid tumors [25,26,46,47]. Plxdc1 expression has been described as prognostic marker and modulating factor for various human cancers [46][47][48][49][50]. Given the overlap between the required changes e.g. in metabolism, transcription, and signaling for cancer growth and virus replication it seems not unlikely that either elevated expression or signaling of these molecules is favorable for both virus infection and cancer progression.

BAC mutagenesis and virus production
RRV recombinants (RRV-YFP gHΔ21-27, RRV-YFP ΔgL and RRV-YFP gHΔ21-27-AELAAN) were generated based on BAC35-8 [51] and RRV gH-AELAAN [23] respectively, using a two-step, markerless λ-red-mediated BAC recombination strategy as described by Tischer et al. [35]. RRV-YFP ΔgL harbors a 128 bp deletion, which introduces a frameshift after amino acid 26 and a stop codon after amino acid 37, leaving only six amino acids of the original gL sequence after the putative signal peptide cleavage site. RRV-YFP gHΔ21-27 revertants were generated based on RRV-YFP gHΔ21-27 following the same protocol described by Tischer et al. [35]. In short, recombination cassettes were generated from the pEPKanS template by polymerase chain reaction (PCR) with Phusion High Fidelity DNA polymerase (Thermo Fisher Scientific) using long oligonucleotides (Ultramers; purchased from Integrated DNA Technologies (IDT)) (see S2 Table for a complete list of primers). Recombination cassettes were transformed into RRV-YFP-carrying GS1783 followed by kanamycin selection, and subsequent second recombination under 1% L(+)arabinose (Sigma-Aldrich)-induced I-SceI expression. RRV-YFP 17577 was generated by shutteling the RRV 26-95 CMV-YFP expression cassette flanked by overhangs identical in RRV 26-95 and RRV 17577 into an RRV 17577 wt bacmid (a kind gift from Scott W. Wong). Recombination cassettes were transformed into RRV 17577-carrying GS1783 and recombination was performed as described above. Colonies were verified by PCR of the mutated region followed by sequence analysis (Macrogen), pulsed-field gel electrophoresis and restriction fragment length polymorphism. For this purpose, bacmid DNA was isolated by standard alkaline lysis from 5ml liquid cultures. Subsequently, the integrity of bacmid DNA was analyzed by digestion with restriction enzyme XhoI and separation in 1% PFGE agarose (Bio-Rad) gels and 0.5×TBE buffer by pulsed-field gel electrophoresis at 6 V/cm, 120-degree field angle, switch time linearly ramped from 1s to 5s over 16 h (CHEF DR III, Bio-Rad). Infectious RRV-YFP recombinants were generated as described previously [23]. In short, bacmid DNA (NucleoBond Xtra Midi) was transfected into 293T cells using GenJet Ver. II (Signagen) according to manufacturer's instructions. Transfected 293T cells were transferred onto a confluent rhesus monkey fibroblasts monolayer two days after transfection and co-cultivated until a visible cytopathic effect (CPE) was observed. For virus stock preparations, confluent primary rhesus monkey fibroblasts were inoculated with infectious supernatant of 293T/rhesus monkey fibroblast co-cultures. After multiple rounds of replication, virus-containing RF supernatant was clarified by centrifugation (4750g, 10min), concentrated by overnight centrifugation (4200rpm, 4˚C) and careful aspiration of approximately 95% of the supernatant. The pellet was resuspended overnight in the remaining liquid. Stocks of wt and recombinant viruses were aliquoted and stored at -80˚C.
Mutations were verified by PCR amplification of the respective region followed by sequence analysis (Macrogen). (See S2 Table for a complete list of primers and constructs).

Real-time PCR (qPCR) analysis of EPH and PLXDC gene expression
For analysis of EPHA7, EPHB3, PLXDC1 and PLXDC2 expression by qPCR, 1x10 6 cells were harvested in RNAzol (Thermo Fisher) and RNA was isolated using the Direct-zol RNA Miniprep Plus Kit (Zymo) according to the manufacturer's instructions. 1μg total RNA was used for cDNA synthesis using the SensiFAST cDNA Synthesis Kit (Bioline) according to manufacturer's instructions. qPCR was performed on a StepOne Plus cycler (Thermo Fisher Scientific) in 20μl reactions using the SensiFAST SYBR Hi-ROX Kit (Bioline) (cycling conditions: 2min initial denaturation at 95˚C, 40 cycles: 95˚C for 5s and 65˚C for 25s). Relative expression was calculated based on ΔCt to GAPDH. Samples were analyzed in technical triplicates, only samples with amplification in all triplicates were scored as positive. See S2 Table for a complete list of primers.

Real-time PCR (qPCR)-based viral genome copy number analysis and virus attachment assay
Concentrated virus samples were treated with DNAseI (0.1 units/μl) to remove any nonencapsidated DNA (37˚C, overnight). Subsequently, DNAseI was inactivated and viral capsids were disrupted by heating the samples to 95˚C for 30min. qPCR was performed on a StepOne Plus cycler in 20μl reactions using the SensiFAST Probe Hi-ROX Kit (Bioline) (cycling conditions: 3min initial denaturation at 95˚C, 40 cycles 95˚C for 10s and 60˚C for 35s). All primerprobe sets were purchased from IDT as complete PrimeTime qPCR Assays (primer:probe ratio = 4:1). Samples were analyzed in technical triplicates. A series of five 10-fold dilutions of bacmid DNA was used as standard for absolute quantification of viral genome copies based on qPCR of RRV ORF73. For virus attachment assays transduced Raji cells were incubated with ice-cold virus dilutions at the indicated concentrations, normalized to genomes per cell, at 4˚C for 30min. After three washes with ice-cold PBS genomic DNA was isolated using the ISO-LATE II Genomic DNA Kit (Bioline) according to manufacturer's instructions. Genome copies of used input virus preparations were determined after overnight DNAseI digest as described above. qPCR was performed as described above using primer-probe sets specific for RRV ORF73 and a cellular genomic locus (CCR5). Viral copy numbers were determined based on a 10-fold dilution series of bacmid DNA, while cell numbers were calculated based on a 2-fold dilution series of Raji gDNA. For attachment assays all samples were analyzed in technical duplicates. See S2 Table for a complete list of primers.

siRNA-mediated knock-down
For siRNA-mediated knock-down of Plxdc2, 293T cells were plated one day prior to transfection in 6well plates. 10μl of a 5μM siRNA solution (set of 4x siCtrl or set of 4x siPlxdc2) purchased as ON-TARGETplus siRNAs from Dharmacon were transfected using 2μl fo Dhar-maFECT 1 transfection reagent (Dharmacon) according to manufacturer's instructions in serum-free DMEM without antibiotics. 6h post transfection fresh DMEM supplemented with 20% FCS was added to achieve a final concentration of 10%FCS. 48h post transfection 293T siCtrl cells and 293T siPlxdc2 cells were harvested, cell numbers were determined and cells were plated at cells were plated at 50 000 cells/cm 2 for infection assays (see below). For determination of knock-down efficiency, cells were harvested in RNAzol 96h post transfection and RNA was isolated using the Direct-zol RNA Miniprep Plus Kit according to the manufacturer's instructions. qPCR was performed on a StepOne Plus cycler (Thermo Fisher Scientific) in 20μl reactions with 40ng RNA/reaction using the SensiFAST Probe Hi-ROX One-Step Kit (Bioline) (cycling conditions: 10min 45˚Cmin, 2min initial denaturation at 95˚C, 40 cycles 95˚C for 5s and 60˚C for 20s). Primer-probe sets specific for Plxdc2 and GAPDH were used. Samples were analyzed in technical triplicates. See S2 Table for a complete list of primers.

Infection assays, blocking experiments and flow cytometry
For infection assays cells were plated at 50 000 cells/cm 2 (SLK, HaCaT, 293T), 25 000 cells/cm 2 (RF) or 200 000 cells/ml (Raji, MFB5487, MMB1845) respectively. One day after plating (for adherent cells lines) or directly after plating (for suspension cell lines), cells were infected with the indicated amounts of virus. Adherent cells lines were harvested 24h post infection by brief trypsinization, followed by addition of 5% FCS in PBS to inhibit trypsin activity. Suspension cell lines were harvested 48h post infection (24h post infection for transduced Raji cells) by pipetting. Subsequently, cells were pelleted by centrifugation (1200rpm, 10min), washed once with PBS, re-pelleted and fixed in PBS supplemented with 4% formaldehyde (Carl Roth). Block of RRV infection with soluble decoy receptor was assayed by infection with virus inocula that were pre-incubated with the indicated concentrations of soluble EphB3-Fc, hPlxdc1-FcStrep, hPlxdc2-FcStrep or FcStrep alone at room temperature for 30min. Calculation of molarity was based on dimeric proteins. For blocking assays using anti-Plxdc2 antibodies, custom rabbit immune sera raised against recombinant Plxdc2-FcStrep (Eurogentec) were used.
293T cells, were pre-treated with an 1:5 dilution of Plxdc2 immune serum or pre-immune serum of the respective animal in DMEM supplemented with 10% FCS for 30min at 37˚C. After pre-incubation virus was added in 1/5 th of the total incubation volume per well. Cell harvest and preparation for flow cytometry analyses was performed as described above. A minimum of 5 000-10 000 cells was analyzed per sample for YFP expression on a LSRII flow cytometer (BD Biosciences). Data was analyzed using Flowing Software (Version 2.5).

Immunoprecipitation and Western blot analysis
Affinity-purification of gH-interacting proteins using gH-FcStrep was performed as described previously [15]. Protein bands were visualized after PAGE on 8-16% gradient gels by silver staining using the SilverQuest Silver Staining Kit (Thermo Fisher Scientific), cut out using a scalpel, destained, and sent to the Taplin Mass Spectrometry Facility, Harvard Medical School, for analysis. For interaction analysis of gH-V5/gL-Flag complexes with Plxdc1/2, 293T cells were transfected using PEI as described before. Lysates of 293T cells transfected with the respective expression constructs for gH-V5/gL-Flag complexes were prepared in NP40 lysis buffer (1% Nonidet P40 Substitute (Sigma-Aldrich), 150mM NaCl (Sigma-Aldrich), 50mM HEPES (VWR), 1mM EDTA (Amresco) with freshly added Protease Inhibitor Cocktail, General Use (Amresco)) and protein content was determined by Bradford assay using Roti-Quant (Roth) according to manufacturer's instructions. 20μg total protein was denatured in 1x SDS sample buffer (Morris formulation) at 95˚C for 5min, separated by polyacrylamide gel electrophoresis (PAGE) using 8-16% Tris-Glycine polyacrylamide gradient gels (Thermo Fisher Scientific) with Tris-Glycine SDS running buffer (25mM Tris, 192mM glycine, 0.1% SDS) and transferred to 0.45μm (0.22μm for blots containing gL) Polyvinylidendifluorid (PVDF) membranes (200mA/Gel, max 30V, 1h in Towbin buffer (25mM Tris, 192mM glycine) with 20% methanol) in a wet tank system (Mini Blot Module, Thermo Fisher). The membranes were blocked in 5% dry milk powder in TBS-T (5mM Tris, 15mM NaCl, 0.05% Tween20) for 1h, at room temperature, washed once in TBS-T and incubated with the respective antibodies for 2h at room temperature or overnight at 4˚C (see S2 Table for a complete list of antibodies). After three washes with TBS-T, the membranes were incubated with the respective HRP-conjugated secondary antibody in 5% dry milk powder in TBS-T for 1h at room temperature washed three times in TBS-T and imaged on an ECL ChemoCam 3.2 Imager (Intas) using Immobilon Forte Western HRP substrate (Merck Millipore). For pulldown of gH-V5 or gH-V5/gL-Flag complexes, the amount of input lysate between wt and mutant gH constructs was normalized to gH expression as determined by Western blot and diluted to equal volume with cell lysate from non-transfected 293T cells prior to immunoprecipitation. Subsequently, lysates were incubated with 0.5μg V5-tag antibody (Bio-Rad) and ProteinG sepharose (GenScript) overnight at 4˚C with agitation. After three washes in NP40 lysis buffer, ProteinG beads with precoupled complexes were incubated overnight at 4˚C with agitation with lysate of full-length human or Macaca mulatta Plxdc1-myc or Plxdc2-myc expression plasmid transfected 293T cells normalized to Plxdc expression. Volumes were adjusted with lysate from untransfected 293T cells. ProteinG beads were collected by brief centrifugation and washed 3 times in NP40 lysis buffer. Precipitates were heated in 2x SDS sample buffer (95˚C, 5min) and analyzed by Western blot as described above. For co-immunoprecipitation of soluble Plxdc1/2-Strep constructs with gH-V5/gL-Flag and EphB3-myc, supernatant of Plxdc1/2-FcStrep transfected 293T cells was incubated with StrepTactinXT beads (IBA) overnight at 4˚C with agitation. After three washes in NP40 lysis buffer, StrepTactinXT beads with pre-coupled Plxdc1/ 2-FcStrep were incubated overnight at 4˚C with agitation with equal amounts of lysate of fulllength human EphB3-myc, RRV 26-95 gH-V5/gL-Flag or 17577 gH-V5/gL-Flag expression plasmid transfected 293T cells or the indicated combinations. Volumes were adjusted with lysate of untransfected 293T cells.

Enzyme-linked immunosorbent assay (ELISA)
F96 Maxisorp Nunc-Immuno Plates (Thermo Fisher Scientific) were coated with recombinant RRV 26-95 gH-FcStrep/gL (described previously [15]) at 1μg/ml in PBS overnight. After three washes with PBS-T, the wells were blocked with 10% FBS in PBS for 2h. Incubation with Plxdc1 ectodomain or Plxdc2 ectodomain was performed for 2h at room temperature in 10% FBS in PBS. The plates were washed three times with PBS-T. Bound protein was detected via the C-terminal 6XHis Tag using 6XHis Antibody MA1-135 (Invitrogen) followed by three washes in TBS-T and incubation with donkey anti-mouse horseradish peroxidase (HRP)-coupled secondary antibody (Dianova). After three washes, 3,3 0 ,5,5 0 -Tetramethylbenzidin (TMB) substrate (Thermo Fisher Scientific) was added and the reaction was stopped by adding 1M HCl. The plates were imaged on a Biotek Synergy 2 plate reader.

Fusion assay
On day1, Raji effector cells stably transduced with a lentiviral construct encoding a Gal4 response element driven TurboGFP-luciferase reporter (Raji-Gal4-Luc) were transduced with lentiviruses encoding TwinStrep-tagged human receptor constructs or an empty vector control. 293T cells were seeded in 96-well plates at 30 000 cells/well. On day2, 293T target cells were transfected with a plasmid encoding the Gal4 DNA binding domain fused to the VP16 transactivaton (VP16-Gal4) and the indicated viral glycoprotein combinations or a pcDNA control (VP16-Gal4: 31.25ng/well, gH: 12.5ng/well, gL (pcDNA in gH/gB only combinations): 62.5ng/well, gB: 18.75ng/well, pcDNA only: 93.75ng/well) using PEI as described before. 24h after transfection, medium on 293T target cells was completely removed and exchanged to 100μl fresh DMEM supplemented with 10% FCS and 50μg/ml gentamycin. Transduced Raji cells were counted, cells were pelleted, resuspended in fresh DMEM supplemented with 10% FCS and 50μg/ml gentamycin and 40 000 Raji cells transduced with receptor constructs or an empty vector control were added to 293T target cells in 50μl full medium. Triplicate wells were used for all target-effector combinations. After 48h, cells were washed once in PBS and lysed in 35μl 1x Luciferase Cell culture lysis buffer (E1531, Promega) for 15min at room temperature and centrifuged for 10min at 4˚C. 20μl of each cell lysate were used to measure luciferase activity using the Beetle-Juice Luciferase Assay according to manufacturer's instructions on a Biotek Synergy 2 plate reader.

Structure prediction and analysis
Homology based structure prediction was performed using the Iterative Threading ASSembly Refinement (I-TASSER) server on standard settings for structure prediction of RRV 26-95 gH and gL based on the crystal structure of the EBV gH/gL complex (3PHF). Modeling of the RRV 26-95 gH/gL complex was additionally performed using both the SPRING and CO-THreader algorithms for protein-protein complex structure and multi-chain protein threading with no differences between determined structures. Resulting I-TASSER structures were aligned to the gH/gL CO-THreader model with the VMD 1.9.3 OpenGL RMSD Trajectory Tool based on amino acids 25 to 62 of gH (RMSD of 0.344Å) and amino acids 2 to 100 of gL (RMSD of 2.697Å) to generate the depicted model. All further analyses and visualizations were performed using VMD 1.9.3 OpenGL.

Mathematical and statistical analysis
Statistical difference between groups was determined by unpaired Student's t-tests followed by Bonferroni correction for multiple comparisons or by ordinary one-way (Fig 2I) or two-way (Fig 2H and S2C and S2D Fig) analysis of variance (ANOVA) followed by Dunnett's correction for multiple comparisons. Statistical analysis for Fig 6E, presented on a logarithmical scale, was performed on log-transformed normalized data using a repeated measures two-way ANOVA followed by Dunnett's correction for multiple comparisons. All Statistical analyses were performed with GraphPad Prism version 6. For all statistics, � : p-value < 0.05, �� : pvalue < 0.01, ��� : p-value < 0.001, ns: not significant.  Table. List of primers and antibodies used in this study. (XLSX)