Recombinant gD1 and gD4 lacking the transmembrane region were produced in insect cells, purified by affinity and size exclusion chromatography and used for crystallization experiments (Figure 1). To evaluate the correct identity, sequence, and molecular mass of gD1, gD4, and equine MHC-I, mass spectrometry (MS) analysis was conducted (Supplementary Figure S1; Table S1). Recombinant equine MHC-I 3.1 (Eqca-1*00101) including a peptide (SDYVKVSNI) linked to the β2-microglobulin (β2m) region was produced in insect cells as well and purified in the same manner as gD1 and gD4.

A 2.25 Å resolution diffraction data set was collected for a gD1 crystal, and the gD1 structure (Figure 1A) was determined using the structure coordinates of HSV-1 gD (PDB-ID 2C36) for molecular replacement and refined to an R_work of 20.3% and R_free of 25.7% (Table 1, PDB-ID 6SQJ). The crystal structure of gD1 contains two gD molecules per asymmetric unit. In solution only the monomeric form was observed by size exclusion chromatography (SEC)-multi-angle static light scattering (MALS; Supplementary Figure S6). In the crystal structure, two ions interpreted as magnesium are trapped between the molecules forming the dimer which are coordinated by residues E242 and D261 of both protein chains together with water molecules. In chain A, the terminal residues E31 to R38, P281 to T348, and the loop region A N71 to N76 could not be modeled due to a lack of electron density. The same is true for the termini of chain B E31 to R40 and N277 to T348 (Supplementary Figure S7). N-linked glycosylations are visible at the predicted sites N53 and N61 (Flowers et al., 1991) which are conserved between gD1 and gD4 (Supplementary Figure S7) but not in gDs of other alphaherpesviruses (Supplementary Figure S8).

Glycoprotein D4 was crystallized with one protein per asymmetric unit. The structure was determined at a resolution of 1.9 Å (Table 1, PDB-ID 6TM8, Figure 1B) using the coordinates of gD1 structure for molecular replacement and refined to an R_work of 17.5% and R_free of 21.5% (Table 1). In total 244 residues could be modeled (R34 to R277, Supplementary Figure S7).

In the structures of gD1 and gD4, six cysteines were found to form three disulfide bonds at sites conserved in members of the gD plolypeptide family: C87/C209, C126/C223, and C138/C147 (Carfi et al., 2001; Li et al., 2017; Yue et al., 2020; Supplementary Figure S7). The overall folds of gD1 and gD4 are very similar with a root-mean-square deviation (rmsd) of 0.7 Å for 220 common Cα atoms (Figure 1C). The cores consist of a nine-stranded (A′, B, C, C′, C″, D, E, F, and G) β-barrel, arranged in a typical V-like Ig fold, flanked by N- and C-terminal extensions with loops, α-helices (α1, α2, α3’, and α3), and small β-strands (str2-4). The termini in both structures point in opposite directions (Figure 1C) and the unresolved C-termini are predicted to be unstructured by Foldindex (Prilusky et al., 2005).

The amino acid sequence identity between EHV-1 and EHV-4 gDs is 76%, much higher than compared to HSV-1 (25%, GenBank AAK19597), PrV (34%, GenBank AEM64108) or BoHV-1 (31%, GenBank NP045370; Supplementary Figure S8). While the global folds of gDs of these different viruses are very similar (Figure 2), the number and positions of α-helices differ. Compared to EHV-1 and EHV-4 gDs, there is an extra helix termed α1’ present in PrV and BoHV-1 gDs. PrV gD has an exclusive α2’ helix which cannot be observed in other gD structures elucidated so far. In HSV-1 and HSV-2 gDs, the α3’ helix is missing but is present in EHV-1, EHV-4, PrV, and BoHV-1 gDs. In HSV-1 and HSV-2 gDs, the α1 helix is split and the α3 helix is kinked in HSV-1 (Figure 2D and Supplementary Figure S9) which is not seen in the other known gD structures.

The six disulfide bonds C87/C209, C126/C223, and C138/C147 are conserved across EHV-1, EHV-4, PrV, HSV-1, HSV-2, and BoHV-1 gDs, while the predicted and resolved glycosylation sites in the crystal structure of gD1 and gD4 are only conserved between EHV-1 and EHV-4 (N52, N61, N297, N386; Supplementary Figures S7, S8). Between gD1 and gD4, also the magnesium-coordinating residues seen in the gD1 monomer-monomer interface are conserved.

gD binding affinities of different alphaherpesviruses to their receptors are known to differ greatly. For example, PrV gD binds human nectin-1 in the nanomolar range (Connolly et al., 2001). HSV-1 gD interacts more weakly, in a micromolar range, with nectin-1 and herpesvirus entry mediator (HVEM; Krummenacher et al., 2005) similarly to BoHV-1 gD with nectin-1 (Yue et al., 2020; Table 2). For HSV-1, HSV-2, and PrV gDs, it has been demonstrated that C-terminal truncation of the proteins increases the binding affinities up to 100-fold (Table 2).

To study the interaction of soluble gD1 and gD4 with recombinant equine MHC-I, a surface plasmon resonance (SPR) binding assay was conducted. α-chains together with β2m with linked peptide of equine MHC-I 3.1 were produced in insect cells and purified by immobilized metal ion affinity chromatography (IMAC) and SEC (Supplementary Figures S3D, S4D). Additionally, the receptor affinity of a C-terminally truncated EHV-4 gD, gD4_36-280, was tested. The truncated protein was produced in the same manner as gD1 and gD4 originally with the goal to crystallize it because the flexible C-terminus was suspected to hinder crystallization of gD1 and gD4. However, shortly after the production of the truncated gD4_36-280, crystal structures were obtained for both gD1 and gD4 proteins. Therefore, only for gD4_36-280 receptor binding kinetics were determined. Another truncated version, gD4_45-276, could not be produced in insect cells, suggesting that the protein failed to fold properly. Binding analyses for soluble gDs were conducted by using a protein dilution series in a range of 0 to approximately 10 μM.

Apparent dissociation constants ( K d app ) of 7,000 ± 2,000 nM and 5,800 ± 760 nM were calculated for gD1 and gD4, respectively (Figures 3A,B,D,E). The truncated gD4 version, gD4_36-280, exhibited a receptor binding affinity to MHC-I in the same order of magnitude (6,700 ± 750 nM; Figures 3C–E).

To test whether recombinant gD1 and gD4 can bind to cell surface MHC-I and subsequently inhibit virus entry, blocking assays were performed. Equine dermal (ED) cells were incubated with the recombinant proteins ranging from 0 to 150 μg/ml (0–3.5 μM) for 1 h on ice and subsequently infected with either EHV-1 or EHV-4 at a multiplicity of infection (MOI) of 0.1. Viruses expressing green fluorescent protein (GFP; Rudolph et al., 2002; Azab et al., 2009) during early infection were used to monitor and analyze the infection levels by flow cytometry. A dose-dependent reduction of infection of up to 50% and 33% on average with 150 μg/ml protein was observed for gD1 and gD4, respectively (Figures 4A,B).

Plaque reduction assays were performed by using a similar procedure. Here, 150 μg/ml of gD1, gD4, and gD4_36-280 were used and cells were infected with 100 plaque forming units (PFU) of EHV-1 or EHV-4. In the presence of soluble gD1, plaque numbers were decreased on average by 51% for EHV-1. For EHV-4, the infection was reduced by an average of 32% after blocking the cells with soluble gD4 recombinant protein. gD4 was also able to block the entry of EHV-1 by 40%. Likewise, gD1 reduced EHV-4 infection by 29%. In general, gD1 proved to be more efficient in blocking both virus infections. The gD4 variant lacking the C-terminal membrane-proximal residues, gD4_36-280, could inhibit EHV-4 infection more efficiently with an average of 46% and proved to be slightly more potent than untruncated gD4 (32%; Figure 4C).

Taken together, all recombinant gDs compete with viral native proteins. A dose-dependent reduction of infection was observed for gD1 and gD4. Notably, both recombinant gDs are able to efficiently block the entry of EHV-1 and EHV-4.

No structures are available for gD1 or gD4 in complex with MHC-I. Therefore, protein–protein docking experiments were performed. Based on data from EHV-1 and EHV-4 mutational studies with diverse MHC-I genotypes it can be assumed that gD binds in close proximity to MHC-I A173 since genotypes with other residues at this position are highly resistant against infections (Sasaki et al., 2011; Azab et al., 2014). Available crystal structures of equine MHC-I Eqca-N*00602 (1.18.7-6) and Eqca-N*00601 (10.18; Yao et al., 2016) feature a glutamic acid and a threonine residue at position 173 in the α2 chain, respectively, and are known not to support EHV-1 and EHV-4 infection (Azab et al., 2014). Additionally, they contain mouse instead of equine β2m. Therefore, for in silico modeling of gD1 and gD4 binding MHC-I, a homology model of MHC-I genotype 3.1 was constructed to reproduce the experimental setup from the in vitro assays. The α-chain template (PDB ID: 4ZUU (Yao et al., 2016)) and target sequences showed 85% identity allowing the development of a high confidence model. In the next step, a homology model of equine β2m was build to achieve a physiological MHC-I state. The β2m template (PDB ID: 4ZUU) and target sequences showed 63% identity and were therefore also highly suitable for homology modeling. The final homology models of the α-chain and β2m contained no Ramachandran outliers (Ramachandran, 1963; Supplementary Figure S10). The calculated backbone RMSD to the template amounted to 0.4 Å in each structure suggesting the correct global fold of the model. All positions (101–164 and 203–259 in MHC-I and 25–80 in β2m) and geometries of disulfide bonds considered typical for MHC were correct, suggesting a high model quality. The final model structure was obtained after assembling both chains and relaxing the homology model with a molecular dynamics (MD) simulation and used directly for the gD docking (Supplementary Figure S12A).

In order to mimic the experimental setup more realistically, the peptide SDYVKVSNI was inserted into the MHC-I homology model for docking. This peptide binds in a cleft between α2 and α3 helices of MHC-I and was used in the cell-based assays. Since the peptide conformation is strongly dependent on the peptide length (Yao et al., 2016), the peptide CTSEEMNAF from MHC-I 1.18.7–6 (PDB-ID: 4ZUU) was used to build a plausible model. The modeled peptide SDYVKVSNI showed no steric clashes and exhibited reasonable bond geometries with a negligible deviation of backbone atom positions (calculated backbone RMSD of 0.8 Å to the template; Supplementary Figure S11).

To test whether additional bias emerging from peptide modeling influenced docking experiments, two docking rounds were performed. First, gD1 was docked to peptide-free MHC-I. Second, gD1 was docked to MHC-I containing the peptide SDYVKVSNI to check if docking provides comparable protein-protein interfaces (PPIs). In the first docking round, the 10 highest scored docking poses with the lowest Rosetta energy were selected (Chaudhury et al., 2011) as the initial filtering step for the peptide-free gD1—MHC-I docking. For further filtering, the rules described in the Methods section were applied (Table 3). Four out of 10 docking poses fulfilled all three rules. Subsequently, single MD simulations for each docking pose were performed to examine PPI stability. For all protein–protein complexes the backbone RMSD was calculated to obtain an overview of the amplitude of protein movements. Only one docking pose showed a nearly constant backbone RMSD value of 6 Å indicating low complex movement (Supplementary Figure S12B). In order to characterize the obtained PPI, selection criteria were applied and three residue patch classes identified in the binding surface as described in the Methods section (Table 4). An inspection of the PPI over an MD simulation trajectory with PyContact (Scheurer et al., 2018; Supplementary Table S4) revealed two gD hot spot residues: D261 (contacting binding pocket MHC-I residue R169 over the whole simulation time) and W257 (contacting binding pocket MHC-I residue I166 over the whole trajectory, Supplementary Table S4).

In the second round, gD1 and gD4 were docked to the MHC-I homology model with the modeled peptide under the same conditions as the peptide-free gD structures. Subsequently, a structure showing initially identified contacts between MHC-I and gD was searched. The EHV-1 gD complex was one of the 3% best scored results and the EHV-4 gD complex was in the 11% top results suggesting that both docking solutions represent low-energy protein complexes. Both structures showed similar gD-MHC-I orientations (Figures 5A,C) and recurring comparable contacts over the trajectories of MD simulations (Supplementary Table S3).

Two contacts frequently observed between gD1 and MHC-I were identified. The first hot spot residue is D261 contacting MHC-I binding pocket residue R169. The second hot spot residue is F213 contacting MHC-I binding pocket residue I166. Additionally, an extensive hydrogen bond network between residues R103—E242 and E113—R238 (Figures 5B,D) was detected (MHC-I—gD residues, respectively). All contacts and their frequencies over the trajectory of MD simulations are summarized in Supplementary Table S3. The gD1- and gD4—MHC-I PPIs over the course of MD simulations revealed minor movements measured as backbone RMSD of maximal 3.5 Å and 6.5 Å (gD1 and gD4, respectively; Supplementary Figures S12C,D).

It can be concluded that PPIs of peptide-free and peptide-bound docking poses are formed with similar residue patches (Supplementary Tables S3, S4) suggesting that the presence of the peptide in MHC-I does not influence gD binding. We observed that peptide-bound docking poses exploit larger PPIs with more possible interactions than the peptide-free docking pose. We assume that more contacts between gD and MHC-I are favorable for the binding. Therefore, peptide-bound docking poses were chosen as the final ones. Based on the optimized docking models, two gD variants were designed for experimental validation in the next step: F213A and D261N. Both residue exchanges are predicted to disrupt gD—MHC-I contacts and lead to inhibition of viral replication in a cell-based assay.

The gD1/4-MHC-I binding hypotheses (Figure  5) were experimentally investigated by mutating the proposed key residues F213 to alanine and D261 to asparagine in EHV-1 and EHV-4 gDs. Two-step Red-mediated mutagenesis (Tischer et al., 2006) was performed on EHV-1 and EHV-4 bacterial artificial chromosomes (BACs) and multi-step growth kinetics with plaque reduction assays were used for virus characterization.

All mutant viruses were successfully reconstituted from mutated BAC and the modified gD gene sequences were confirmed by Sanger sequencing. EHV-1-gD_F213A displayed a significant 2-log reduction in growth kinetics and low titers in cell supernatant compared to the parental virus. Reverting the mutation rescued virus growth in cell culture (Figure 6B). Plaque sizes of wild type, mutant and revertant viruses were similar. The virus mutants EHV-1-gD_D261N, EHV-4-gD_D261N and EHV-4-gD_F213A did not grow in cells to the extent where growth kinetics could be analyzed. However, reverting the residue exchange in EHV-1-gD_D261N rescued virus growth (Figure 6A). Taken together, the gD_D261N and gD_F213A variants lead to replication-deficient viruses in EHV-1 and EHV-4.

EHV-1 strain RacL11 and EHV-4 strain TH20p are maintained as bacterial artificial chromosome infections clones (BAC). The viruses have GFP under the control of the HCMV major immediate-early (IE) promoter inserted into the Mini-F sequence to easily recognize infected cells. Clones were generated as described previously in Rudolph et al. (2002) and Azab et al. (2009, 2011). The viruses were grown on equine dermal (ED) cells (CCLV-RIE 1222, Federal Research Institute for Animal Health, Greifswald, Germany) at 37°C under a 5% CO₂ atmosphere.

ED cells were grown in Iscove’s modified Dulbecco’s medium (IMDM; Pan, Biotech, Aidenbach, Germany) containing 20% fetal bovine serum (FBS; Biochom GmBH, Berlin, Germany), 1 mM sodium pyruvate (Pan Biotech, Aidenbach, Germany), 1% nonessential amino acids (NEAA; Biochom GmBH, Berlin, Germany), and P-S solution (100 U/ml penicillin: Panreac, AppliChem GmBH, Darmstadt, Germany); 100 μg/ml streptomycin: Alfa Aesar, Thermo Fisher Scientific, Kandel, Germany (P-S) at 37°C under a 5% CO₂ atmosphere.

Human embryonic kidney (293 T, ATCC CRL-11268) cells were propagated in Dulbecco’s modified Eagle’s medium (DMEM; Biochom GmBH, Berlin, Germany), supplemented with 10% FBS and P-S. Sf9 cells (IPLB-Sf21-AE, Invitrogen, Germany) were propagated in serum free Sf-900 III medium (Gibco, Thermo Fisher Scientific, New York, USA) and High Five™ cells (BTI-TN-5B1-4, Invitrogen, Germany) in serum free Express Five medium (Gibco, Thermo Fisher Scientific, New York, USA) at 27°C on orbital shaker.

Constructs were amplified from insect cell codon-optimized DNA fragments (Bio Basic Inc., Canada) for protein production in High Five insect cells. Synthetic truncated genes contained the gene of interest (gD1 residues 32–249, gD4 residues 32–249, equine MHC-I 3.1,), a C-terminal TEV protease cleavage site (ENLYFQG), and a hexa-histidine tag (His₆), all flanked by EcoRI-and ScaI-restriction sites (Figure 7). Sequences of codon optimized genes can be found in the supplementary data. A further truncated form of gD4 containing the residues 36–280 was amplified from gD4 synthetic gene with the primer pair VK50/VK56 (Supplementary Table S2).

The Autographa californica nuclear polyhedrosis virus (AcNPV) baculovirus gp64 signal sequence under control of the very late polyhedrin promoter was inserted into the insect cell vector plasmid pACEBac1 (Addgene, LGC Standards Teddington, UK) by using another synthetic gene (VK18, LGC Genomics, Berlin, Germany) and the primer pairs VK7/VK7 (Supplementary Table S2). Subsequently, plasmids containing synthetic genes (gD1, gD4, MHC-I) were amplified in Escherichia coli (E. coli), purified, and digested with EcoRI-and ScaI-restriction enzymes for insertion into the transfer vectors which was digested with the same restriction enzymes. After ligation, these plasmids were transformed into DH10MultiBac electrocompetent cells and recombinant baculoviruses produced according to manufacturer’s instructions (Bac-to-Bac expression system, Invitrogen). All constructs were verified by sequencing (VK8 or VK10/WA2, VK35/38; Supplementary Table S2). Recombinant BACs were isolated and used for virus production in Sf9 cells as described in Santos et al. (2012).

Equine MHC-I, gD1_32-349, gD4_32-349, and gD4_36-280 were expressed in HighFive cells. Cell supernatant was harvested after 72 h post infection by centrifugation. The pH was adjusted to 7 with 1 M tris (hydroxymethyl) aminomethan (Tris)-HCl buffer at pH 9 on ice and incubated for at least 1 h with washed Ni²⁺-NTA beads for affinity chromatography. Beads with bound recombinant protein were collected by a gravity flow column and the proteins were eluted with a buffer containing 20 mM Tris–HCl at pH 7.5 or 2-(N-morpholino) ethanesulfonic acid (MES) at pH 6 for gDs and MHC-I, respectively, and 200 mM NaCl, 5% (v/v) glycerol, and 200 mM imidazole. Concentrated protein was loaded onto a 16/600 Superdex 200 gel filtration column (GE Healthcare Piscataway, NJ). The buffer conditions were the same as in Ni²⁺-NTA affinity chromatography but with 20 mM NaCl and no imidazole. Proteins collected from size-exclusion chromatography were concentrated (Concentrators, Amicon Ultra, Millipore, Darmstadt, Germany), aliquoted and directly used for crystallization or stored at −80°C.

Crystals of EHV-1 gD were obtained by the sitting-drop vapor-diffusion method at 18°C with a reservoir solution composed of 0.1 M Tris/HCl buffer at pH 8.5, 0.2 M MgCl₂, and 30% (w/v) polyethylene glycol (PEG) 4,000. Crystals were cryo-protected with a solution composed of 75% mother liquor and 25% (v/v) glycerol and subsequently flash-cooled in liquid nitrogen. Synchrotron diffraction data were collected at the beamline P11 at DESY (Hamburg, Germany) and at the beamline 14–2 of the MX beamline of the BESSY II (Berlin, Germany) and processed with X-ray detector software (XDS; Kabsch, 2010). The structure was solved by molecular replacement with PHASER (Bunkóczi et al., 2013) using the coordinates of PDB-ID 2C36 as search model for gD1 which was then used as search model for gD4. A unique solution with two molecules in the asymmetric unit for gD1 and molecule for gD4 were subjected to the program AUTOBUILD in PHENIX (Adams et al., 2010) and manually adjusted in COOT (Emsley et al., 2010). The structures were refined by maximum-likelihood restrained refinement using PHENIX (Adams et al., 2010; Afonine et al., 2012). Model quality was evaluated with MolProbity (Williams et al., 2018) and the JCSG validation server (Yang et al., 2004). Secondary structure elements were assigned with DSSP (Kabsch and Sander, 1983) and for displaying sequence alignments generated by ClustalOmega (Sievers et al., 2011) ALSCRIPT (Barton, 1993) was used. Structure figures were prepared using PyMOL (DeLano, 2002). Coordinates and structure factors have been deposited in the PDB for gD1 with PDB-ID 6SQJ as well as for gD4 with PDB-ID 6TM8. Diffraction images have been deposited at proteindiffraction.org (gD1: DOI 10.18430/m36sqj and gD4 DOI 10.18430/m36tm8).

Intact protein mass of gD1, gD4, and MHC-I was determined by matrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF-MS) using an Ultraflex-II TOF/TOF instrument (Bruker Daltonics, Bremen, Germany) equipped with a 200 Hz solidstate Smart beam™ laser. Samples were spotted using the dried-droplet technique on sinapinic acid (SA) or 2,5-dihydroxybenzoic acid (DHB) matrix (saturated solution in 33% acetonitrile/0,1% trifluoroacetic acid). The mass spectrometer was operated in the positive linear mode, and spectra were acquired over an m/z range of 3,000-60,000. Data was analyzed using FlexAnalysis 2.4. software provided with the instrument.

Protein identity was determined by tandem mass spectrometry (MS/MS) of in-gel digested Coomassie stained protein with 12_,5 μg _/ ml Glu-C and trypsin, and 10 μg _/ ml Asp-N in 25 nm ammonium bicarbonate.

N-terminal c and C-terminal (z + 2) sequence ion series were generated by in-source decay (ISD) with 1,5-diaminonaphthalene (1,5-DAN) as matrix (20 mg/ml 1,5-DAN in 50% acetonitrile/0,1% trifluoroacetic acid). Spectra were recorded in the positive reflector mode (RP PepMix) in the mass range 800–4,000.

For molecular mass determination of soluble, recombinant gD1, SEC-MALS (Tarazona and Saiz, 2003) was performed. Protein solution was run at room temperature on a Superdex 75 10/300 Gl (GE Healthcare, Piscataway, NJ) column with 2 mg/ml gD1 and a mobile phase composed of Tris–HCl at pH 7.5, 200 mM NaCl, 5% glycerol, and 0.02% sodium azide, attached to a high-performance liquid chromatography (HPLC) system (Agilent Technologies, Santa Barbara, CA, USA) with a mini DAWN TREOS detector (Wyatt Technology Corp., Santa Barbara, CA, USA). Data was acquired and analyzed with ASTRA for Windows software package (version 6.1.2).

Binding kinetics of soluble gD1, gD4, and gD4_36-280 binding to amine-coupled recombinant, equine MHC-I 3.1 were measured at 25°C on a surface plasmon resonance (SPR) GE Biacore J Biomolecular Interaction Analyser instrument (Uppsala, Sweden) using a polycarboxylate hydrogel sensor chip HC200M (XanTec bioanalytics GmbH). The second channel was coated with poly-L-lysine and positive nanogels (size 214 nm; Dey et al., 2018) that were shown to interact only weakly with gDs and used as negative control. The control sensorgrams were subtracted from reaction sensorgrams and normalized. The surfaces were regenerated with buffer containing 200 mM NaCl and 10 mM NaOH after each cycle. Serial dilutions of gDs ranging from 0 to approximately 10 μM were injected at medium flow and the interaction with MHC-I was monitored for 15 min to quantify the equilibrium response R eq . In the cases in which a clear flattening of the sensorgramm could not be observed within the experimental time, R eq was extrapolated using a simple exponential function (see Supplementary Figure S13 for an example). The response curves of gDs binding to the MHC-I were fitted to the binding model R eq = R ∞ A K d app + A (Schasfoort, 2017) using Sigma plot 12.0 software. R ∞ is the maximum signal corresponding to saturation, A is the protein concentration and K d app is the apparent dissociation constant.

The point mutations F213A and D261N in EHV-1 and EHV-4 gDs were introduced via a two-step Red recombination (Tischer et al., 2006). In brief, polymerase chain reaction (PCR) primers (Supplementary Table S2) were designed in a way that the 50 nucleotide recombinantion arms include the point mutation and sequence to amplify the kan^R gene. For construction of EHV-1-gD_D261N, EHV-1gD_F213A, EHV-4-gD_D261N, and EHV-4-gD_F213A the primer pairs VK61/VK62, VK63/VK64, VK65/VK66, and VK67/VK68 were used for PCR amplification, respectively. After Dpn-1 digest of PCR products, fragments were electroporated into GS1783 containing EHV-1 or EHV-4 BACs. DNA from Kanamycin resistant colonies was extracted and correct mutants were selected based on Restriction fragment length polymorphism (RFLP) using the restriction enzyme Pst-I. Correct clones were subjected to another round of Red recombination to remove the kan^R gene. Final clones were further analyzed by RFLP and sequencing, BAC extracted, purified and transfected into 293 T cells. Cells and supernatant were harvested 3 days post transfection and used to infect ED cells. Revertants were produced from mutant clones using the same procedure with primer pairs VK69/VK70, VK71/VK72, VK73/VK74, and VK75/VK76 for producing EHV-1R-gD_D261N, EHV-1R-gD_F213A, EHV-4R-gD_D261N, and EHV-4-RgD_F213A, respectively. All genotypes were confirmed by PCR, RFLP, and Sanger sequencing using the primer pair WA2/VK8 and WA2/VK10 (Supplementary Table S2) for EHV-1 and EHV-4 mutants, respectively.

Western blot analysis was performed with soluble proteins: 50 μg/ml MHC-I, 5 μg/ml gD1, and 5 μg/ml gD4. Proteins were separated by 12% SDS-PAGE, transferred to a polyvinylidene difluoride (PVDF) membrane (Roth, Karlsruhe, Germany), detected with 1:1000 dilution rabbit anti-His₆ (Sigma-Aldrich, St Louis, USA) antibody and 1:10000 dilution goat anti-rabbit-HRP antibody (Sigma-Aldrich, St Louis, USA) and visualized by enhanced chemiluminescence (ECL Plus; Amersham).

To block cell surface MHC-I, 1,5×10⁵ ED cells were seeded in 24-well plates. The next day, cells were incubated with 20, 50, 100 or 150 μg/ml recombinant gD1, gD4 or gD4_36-280 for 1 h on ice. Subsequently cells were infected with either EHV-1 or EHV-4 at MOI = 0.1 and incubated for 1 h at 37°C. To remove un-penetrated viruses, cells were washed with citrate buffer, pH 3, containing 40 mM citric acid, 10 mM potassium chloride and 135 mM sodium chloride, then washed twice with phosphate buffered saline (PBS) and infection allowed to proceed for 24 h (EHV-1) or 48 h (EHV-4). For measurement of fluorescence intensity 10,000 cells were analyzed with a FACSCalibur flow cytometer (BD Biosciences) and the software CytExpert (Beckman Coulter, Krefeld). The experiment was repeated three independent times for each protein.

For plaque reduction assay the same protocol was applied for blocking surface MHC-I with minor changes. Cells were initially incubated with 150 μg/ml gD1 or gD4, infected with 100 PFU, and overlaid with 1.5% methylcellulose (Sigma-Aldrich, Taufkirchen, Germany) in Iscove’s Modified Dulbecco’s Medium (IMDM) after citrate treatment and washes with PBS. GFP plaques were counted after 48 h with a Zeiss Axiovert.A1 fluorescent microscope (Carl Zeiss AG, Jena, Germany). The experiment was repeated three independent times for each protein.

Virus replication was tested using multi-step growth kinetics and plaque areas were obtained as described before (Azab et al., 2010). ED cells were grown to confluency in 24-well plates, infected with an MOI of 0.1 virus and incubated for 1 h at 37°C. Viruses on the cell surface were removed by washing with citrate buffer. After neutralization with IMDM, cells were washed twice with PBS and finally overlaid with 500 μl IMDM. At indicated times after the citrate treatment cells and supernatant were collected separately for EHV-1 and together for EHV-4 and stored at-80°C. Titers were determined by plating dilution series onto ED cells and counting plaque numbers after 2 days under a methylcellulose overlay. All plates were fixed for 10 min with 4% paraformaldehyde, washed with PBS and stained for 10 min with 0.1% crystal violet solution in PBS which was washed away with tab water. Viral titers are expressed as PFU per milliliter from three independent and blinded experiments.

For blocking assays, plaque numbers were normalized to infection levels without recombinant proteins. Statistical analysis was done using GraphPad Prism 5 software (San Diego, CA, USA) and one-way ANOVA Bonferroni’s multiple comparison test, * indicates p ≤ 0.05, ** indicates p ≤ 0.01, *** indicates p ≤ 0.001. Statistical analysis was done using an unpaired, one-tailed test. p < 0.05 was considered significant.