Engineered receptors for human cytomegalovirus that are orthogonal to normal human biology

A trimeric glycoprotein complex on the surface of human cytomegalovirus (HCMV) binds to platelet-derived growth factor (PDGF) receptor α (PDGFRα) to mediate host cell recognition and fusion of the viral and cellular membranes. Soluble PDGFRα potently neutralizes HCMV in tissue culture, and its potential use as an antiviral therapeutic has the benefit that any escape mutants will likely be attenuated. However, PDGFRα binds multiple PDGF ligands in the human body as part of developmental programs in embryogenesis and continuing through adulthood. Any therapies with soluble receptor therefore come with serious efficacy and safety concerns, especially for the treatment of congenital HCMV. Soluble virus receptors that are orthogonal to human biology might resolve these concerns. This engineering problem is solved by deep mutational scanning on the D2-D3 domains of PDGFRα to identify variants that maintain interactions with the HCMV glycoprotein trimer in the presence of competing PDGF ligands. Competition by PDGFs is conformation-dependent, whereas HCMV trimer binding is independent of proper D2-D3 conformation, and many mutations at the receptor-PDGF interface are suitable for functionally separating trimer from PDGF interactions. Purified soluble PDGFRα carrying a targeted mutation succeeded in displaying wild type affinity for HCMV trimer with a simultaneous loss of PDGF binding, and neutralizes trimer-only and trimer/pentamer-expressing HCMV strains infecting fibroblasts or epithelial cells. Overall, this work makes important progress in the realization of soluble HCMV receptors for clinical application.


Introduction
concentrations; and soluble PDGFRα blocks entry by multiple HCMV strains that express both trimer and pentamer complexes [38].
While appealing, the use of soluble PDGFRα to treat HCMV in humans comes with risk. Indeed, no protein-based viral receptors are currently approved as antiviral drugs by the FDA [39] (although at least one is in a phase II clinical trial for the treatment of COVID-19 by Apeiron Biologics), in part due to safety concerns over interactions with endogenous ligands. By comparison, there are multiple antibody drugs that are specific for viral targets used in the clinic [39]. PDGFRα is an important receptor for platelet-derived growth factors (PDGFs), and regulates cellular proliferation, differentiation and development of multiple tissues during embryogenesis and onwards through adulthood [40]. The PDGF ligands are small polypeptides that covalently associate in disulfide-bonded homo-and heterodimers, which display differing activities towards PDGFRα and its close relative PDGFRβ [41]. Four ligands interact with PDGFRα with high affinity (PDGF-AA, AB, BB, and CC) [42,43] and bind two receptor chains at opposing ends to form a receptor-(ligand) 2 -receptor signaling complex [44,45]. PDGFs compete with the HCMV trimer for PDGFRα binding and block infection [35,36], suggesting that PDGF binding to PDGFRα sterically hinders trimer binding or allosterically modulates the trimer binding interface in a nonproductive manner. Treatment with soluble PDGFRα will almost certainly disrupt physiological PDGF-signaling, and the ability of recombinant receptor to bind virions may be diminished by competing endogenous PDGFs. Ideally, mutations in the receptor can be identified that disrupt PDGF interactions while retaining the site or sites engaged by HCMV.
A high resolution crystal structure of PDGF-BB-bound PDGFRβ has illuminated atomic details of ligand-receptor interactions in the family, revealing a predominantly hydrophobic interface at the cleft between the second (D2) and third (D3) extracellular domains of the receptor [44]. However, the binding site for gO is ambiguous. Cryo-electron microscopy suffers from conformational diversity and low resolution [34]; PDGFRα peptide fragment analysis failed to discover any one peptide solely responsible for high affinity binding [38]; and while domain deletions have shown PDGFRα-D3 is important [37], such studies have coarse resolution. The absence of detailed structural information on the trimer-PDGFRα complex means the engineering of HCMV-specific receptors orthogonal to human biology (i.e. receptors that bind virus but do not interact with endogenous human ligands) remains an unsolved challenge.
This challenge is solved through the use of mutagenesis and in vitro selection. By tracking the enrichment or depletion of sequence variants in a diverse library using next generation sequencing, the relative phenotypes of thousands of mutations can be simultaneously assessed in a single experiment, referred to as deep mutagenesis [46]. Deep mutational scans have been used to address and engineer specificity in proteins that promiscuously bind multiple ligands [47][48][49], and methods for deep mutagenesis of membrane proteins expressed in human cells have been optimized [50][51][52]. By screening through all possible single amino acid substitutions in the D2-D3 domains, we discovered that, unlike PDGF binding, HCMV trimer interactions persist when PDGFRα conformation is disrupted by mutations within the folded core. Particularly relevant to therapeutic design, we also identified PDGFRα surface mutations that retain trimer binding in the presence of high PDGF concentrations. This study brings receptor-based inhibition of HCMV infection closer to realization.

A deep mutational scan of PDGFRα D2-D3 domains for trimer binding in the presence of PDGFs
was deleted to produce a soluble complex, and the gO subunit, which directly contacts PDGFRα [34,36,54], was fused at its C-terminus to superfolder GFP (sfGFP; [55]) for fluorescence detection. Medium from gH/gL/gO-sfGFP expressing cells was incubated with Expi293F cells transfected with an episomal plasmid encoding PDGFRα. Trimer binding was proportional to PDGFRα surface expression (determined by flow cytometry using antibody staining of an extracellular c-myc epitope tag at the PDGFRα N-terminus; Fig 1A and 1B), and was inhibited by co-incubation with an equimolar mixture of PDGF-AA, AB, BB, and CC ( Fig  1D). Medium from cells expressing gO-sfGFP alone, either the wild-type sequence or carrying the mutation C351S (C343S based on numbering of gO from the TB40/E strain) to remove the exposed cysteine that would ordinarily form a disulfide to gL-C144 [56], failed to display high binding signal to PDGFRα positive cells (S1 Fig). The binding signal is therefore dependent on co-expression with gH and gL, suggestive that it is trimer as opposed to monomeric gO that engages the receptor, although this is not explicitly demonstrated.
A single site-saturation mutagenesis (SSM) library of PDGFRα was constructed encompassing all single amino acid mutations in the D2-D3 domains (a.a. D123-E311), and transfected into Expi293F cells under conditions that typically yield no more than one sequence Expi293F cells expressing myc-tagged PDGFRα were incubated with dilutions of gH/gL/gO-sfGFP-containing medium, and binding (shown as ΔMean Fluorescence Units, ΔMFU, in which Mean Fluorescence Units of a negative control without trimer are subtracted as background) was measured by flow cytometry. Neat, no dilution. (B and C) Cells were transfected with wild-type (WT) PDGFRα (B) or the D2-D3 SSM library (C) under conditions where typically no more than a single sequence variant is expressed by any given cell. Under these conditions, most cells are negative for PDGFRα expression as determined by anti-myc staining. Cells were incubated with a subsaturating (12-fold dilution) of gH/gL/gO-sfGFP medium and analyzed by flow cytometry. (D) Cells were co-incubated with a 2-fold dilution of gH/gL/gO-sfGFP medium and different concentrations of PDGF (an equimolar mixture of PDGF-AA, AB, BB, and CC). The fluorescent signal of gH/gL/gO-sfGFP binding is compared to a sample without competing PDGF. (E and F) Cells expressing WT PDGFRα (E) or the D2-D3 SSM library (F) were incubated with a 3-fold dilution of gH/gL/gO-sfGFP medium and 100 nM PDGF (25 nM of each subtype). The choice of a PDGF concentration that does not completely inhibit trimer binding ensured (i) that there remained some positive binding signal for drawing appropriate sort gates, and (ii) that mutations with weak effects will be distinguished from the main population. During sorting of the D2-D3 library, the top 15% of PDGFRα-expressing cells that bind trimer were collected (UL74-High sort, green gate). Simultaneously, the bottom 15% of PDGFRα-expressing cells that bind trimer were also collected (UL74-Low sort, blue gate). For cells expressing WT PDGFRα, the equivalent gates correspond to 4% and 23% of PDGFRα-expressing cells, respectively. https://doi.org/10.1371/journal.ppat.1008647.g001

PLOS PATHOGENS
Engineered receptors for human cytomegalovirus variant per cell, providing a tight link between genotype and phenotype [51]. When incubated with gH/gL/gO-sfGFP medium at subsaturating dilutions, cells expressing the SSM library were surprisingly indistinguishable from cells expressing wild-type PDGFRα (Fig 1B and 1C). Normally, many mutations adversely impact protein activity through destabilization of folded structure or damage to functional sites, and loss-of-function variants tend to dominate naive libraries prior to any selection. That deleterious mutations are not prevalent in the naive SSM library immediately implied that HCMV trimer binding is resistant to most single non-synonymous mutations in the PDGFRα D2-D3 domains.
By comparison, many cells expressing the SSM library displayed higher trimer binding in the presence of PDGFs than cells expressing wild-type PDGFRα (Fig 1E and 1F). There thus appeared to be many PDGFRα mutations that selectively lost PDGF affinity. To identify these mutations, PDGFRα-expressing cells that had elevated HCMV trimer binding in the presence of competing PDGFs were collected by fluorescence-activated cell sorting (FACS). This is referred to as the UL74-High sort (see green gate in Fig 1F). Within the same experiment, cells expressing PDGFRα but displaying low trimer binding were also collected, referred to as the UL74-Low sort (see blue gate in Fig 1F). The UL74 gene encodes gO [31], and the names assigned to the sorted populations correspond to the raw and processed data files deposited in the GEO database [57]. PDGFRα mutants that lose affinity for trimer, or perhaps have enhanced affinity for competing PDGFs, will be preferentially enriched in the UL74-Low sort. PDGFRα mutants that fail to express will be depleted from both sorted populations. PDGF interactions are uniquely sensitive to mutational disruption of folded structure in the D2-D3 domains Enrichment ratios for nonsynonymous mutations are anticorrelated between the UL74-High and UL74-Low sorts, with few mutations other than premature stop codons being depleted in both selections. Hence substitution mutations in PDGFRα D2-D3 domains suffer little cost to surface localization, even though many mutations will almost certainly destabilize the domain fold. This differs markedly from a prior mutational scan of a multidomain membrane protein expressed in human cells, where nonconservative mutations buried within folded structure prevented protein trafficking to the plasma membrane due to intracellular retention by quality control machinery [50]. When mapping experimental conservation scores to the structure of PDGF-bound PDGFRα (modeled from the crystal structure of PDGF-BB-bound PDGFRβ; PDB ID 3MJG [44]), enriched mutations in the UL74-High sort are found to cluster to buried core positions and the PDGF interface (Fig 2). These include substitutions that create cavities, steric clashes, and/or introduce buried ionizable groups; these mutations are incompatible with biophysical principles governing protein folding and will destabilize folded structure. We hypothesize that PDGFRα mutants with disrupted structure have therefore preferentially lost affinity to PDGFs, thereby reducing competition and enhancing binding of the HCMV trimer. The epitope bound by the HCMV trimer is ambiguous from the data (S3 Fig), but must be at least partially independent of proper D2-D3 conformation, although partial structure may remain. This is consistent with either key contacts to the HCMV trimer residing on linear PDGFRα segments and loops, similar to how many antibodies recognize conformation-independent epitopes, or with favorable interactions to the HCMV trimer being distributed over a broad surface (possibly even beyond the D2-D3 domains) such that localized disruptions are tolerated. Libraries encompassing double or higher order PDGFRα mutations may further resolve details of the binding mechanism. While this illuminates aspects of PDGFRα/trimer recognition, using misfolded, destabilized PDGFRα variants to selectively neutralize HCMV is fraught with risks, including toxicity from non-specific interactions and protein aggregation due to the exposure of hydrophobic residues.

Surface mutations at the PDGF binding site favor PDGFRα specificity towards the HCMV trimer
The PDGFRα mutational landscape for high trimer binding in the presence of competing PDGF reveals key surface residues as hot spots for enriched mutations, especially PDGFRα residues L137, L208 and V242 that contact the protruding PDGF subunit, and Y206 which packs between the protruding and receding PDGF subunits (Fig 2D and 2E). Mutations to these residues will disrupt the PDGFRα-PDGF interface, especially by the addition of polar substitutions that invert the chemical properties of the highly hydrophobic binding site, while HCMV trimer interactions persist. Overall, enrichment of mutations for increased HCMV trimer binding in the presence of competing PDGFs is highly correlated to the mutated residue's connectivity within the D2-D3 core or across the PDGFRα-PDGF interface (S4 Fig), emphasizing the selective importance of PDGFRα conformation and surface contacts in the D2-D3 cleft for high affinity PDGF interactions.
Five mutations on the PDGFRα surface (L137K, L137Q and Y206S in D2, and V242K and V242T in D3, chosen based on their high enrichment in the UL74-High sort and covering multiple surface positions) were validated as having desirable binding properties by targeted mutagenesis. As predicted from the deep mutational scan, these PDGFRα variants maintain near wild-type levels of binding to soluble HCMV trimer from the TB40/E strain, yet have diminished sensitivity to the addition of competing PDGFs (Fig 3A and 3B), demonstrating successful engineering of specificity towards the viral target. Since the mechanism of trimer binding to receptor has not been resolved at atomic or even residue-level resolution, it remains uncertain whether the viral genome could mutate to distinguish wild-type from engineered PDGFRα. This concern is partially alleviated by the observation that the engineered PDGFRα mutants maintain high binding to soluble HCMV trimer from the Merlin clinical strain [58] ( Fig 3C), which has relatively high natural sequence variation compared to TB40/E (79% identity between glycoproteins O). Variation is particularly high in the N-terminus of glycoprotein O where important interactions to PDGFRα reside [54].
Fusion of soluble proteins to immunoglobulin Fc confers multiple desirable properties for a therapeutic [59]. Multimerized Fc chains impose avidity for enhanced apparent affinity, and Fc moieties are engaged by multiple receptors to enhance serum half-life or evoke effector functions, including complement activation and antibody-dependent cell-mediated cytotoxicity [59]. Soluble PDGFRα fused to the Fc region of human IgG1 was shown to bind HCMV trimer when first identified as its receptor [34], and the neutralization properties of Fc-fused soluble PDGFRα (sPDGFRα-Fc) have been explored since in greater detail [38]. These prior studies used commercially supplied sPDGFRα-Fc featuring a random linker of mixed amino acids that connects to the upper hinge of IgG1, upstream of Cys-220 that would ordinarily form a disulfide to the antibody light chain (S5A Fig). The nature of this Fc fusion may cause manufacturing liabilities if Cys-220 is exposed and free, and we refer to this construct as "legacy" sPDGFRα-Fc. We designed an alternative sPDGFRα-Fc construct that includes the PDGFRα signal peptide (a.a. 1-23), ectodomain (a.a. 24-524), a GGGS linker, and finally human IgG1 Fc beginning at Asp-221 (S5A Fig). Both sPDGFRα-Fc constructs bind membrane-localized HCMV trimer with equal affinity (Fig 3D and 3E), and subsequent studies focused exclusively on our redesigned fusion protein.
Soluble PDGFRα-Fc was expressed in Expi293F culture, and was isolated by affinity capture and size exclusion chromatography (SEC) to very high purity (S5B and S5C Fig). The purified protein is monodisperse by SEC, and the final yield is high (> 25 mg / L of transfected culture without any optimization). Soluble PDGFRα-Fc V242K was stressed for 7 days at 40˚C, pH 8.5, to promote asparagine deamidation and protein denaturation [60], yet remained mostly

PLOS PATHOGENS
Engineered receptors for human cytomegalovirus stable based on the SEC profile (S6 Fig). A harsher stress for 14 days at 40˚C, pH 5.5, to promote asparagine isomerization and protein denaturation [60] caused sPDGFRα-Fc V242K to form soluble aggregates (S6 Fig). We conclude that the protein has moderate stability and is suitable for further manufacturing development.
Two PDGFRα mutations, Y206S and V242K, were explored as Fc-fused soluble orthogonal receptors. Binding of the soluble receptors was measured towards HCMV trimer expressed at the cell surface, with gL and gO subunits tagged at their C-termini with short peptide epitopes, and gH expressed as full-length protein with its native transmembrane domain. The Y206S mutant had reduced binding to HCMV trimer compared to wild-type receptor, but sPDGFRα-Fc V242K bound HCMV trimer from Merlin and TB40/E strains with equal low nanomolar affinity (Fig 3D and 3E). Aspects of these experiments were replicated with independent protein preparations (S7A Fig). Furthermore, fusions of wild-type and V242K sPDGFRα with the Fc region of IgG3 also bound similarly to trimer from the TB40/E and Merlin strains (S7B and S7C Fig). Compared to IgG1, IgG3 has lower affinity for the neonatal Fc receptor (FcRn), which is associated with reduced serum half-life and placental transfer [61,62]. We speculate that these features may be advantageous in the treatment of pregnant women during acute HCMV infection to achieve a lower dose in the developing fetus, thereby further addressing safety concerns.

Engineered receptors for human cytomegalovirus
To further validate that the engineered receptors have lost growth factor interactions, we assessed their ability to inhibit PDGF signaling. Expi293F cells were transfected with a reporter plasmid encoding wild-type PDGFRα and a destabilized fluorescent protein (EGFP-PEST) [63] under the control of a serum response element (SRE) [64]. Treatment with PDGFs upregulates EGFP-PEST expression, which is inhibited by wild-type sPDGFRα-Fc acting as a decoy receptor (Fig 4). As intended, both sPDGFRα-Fc Y206S and V242K failed to inhibit PDGF signaling (Fig 4B). However, we also noticed that the engineered receptors, in particular sPDGFRα-Fc Y206S, unexpectedly enhanced the response to PDGF-A ligands. We suspect this is due to a generic 'carrier protein' effect (PDGF ligands and their receptors are highly hydrophobic [44] and are generally formulated with a carrier such as serum albumin), though we are unable to conclude this with certainty, and non-specific gamma globulin has no such effect (S8 Fig). This observation may require future investigation.
Soluble PDGFRα-Fc potently neutralized both HCMV strains at nanomolar concentrations (Fig 5). Neutralization by orthogonal sPDGFRα-Fc V242K was mostly indistinguishable from wild-type receptor and nearly complete at 1 to 10 nM (Fig 5), demonstrating that virus binding and neutralization can indeed be successfully separated from growth factor interactions by just a single amino acid substitution. Soluble PDGFRα-Fc Y206S also neutralized but with reduced potency, consistent with the biochemical binding studies. However, neutralization of trimer/ pentamer-expressing TB40/E remained incomplete even at high sPDGFRα-Fc concentrations. Hyperimmune globulin (sold under the trade name Cytogam by CSL Behring for prophylactic

PLOS PATHOGENS
Engineered receptors for human cytomegalovirus treatment of transplant recipients) completely neutralizes HCMV, albeit at high concentrations. It is unclear if residual pentamer-mediated infection in the presence of high sPDGFRα-Fc would in any way limit therapeutic or prophylactic efficacy in an in vivo setting. Depletion of HCMV-positive patient sera with either trimer or pentamer severely diminishes neutralization of epithelial cell infection [66], suggesting that targeting both glycoprotein complexes can give synergistic effects and is worth further exploration.

Discussion
Ideas surrounding the use of soluble virus receptors as decoys to treat or prevent infection have a long history [67][68][69][70]. Soluble PDGFRα as a prophylactic or treatment for HCMV is particularly promising due to exceptionally tight affinity for glycoprotein trimer and its ability to neutralize both trimer-and pentamer-mediated host cell entry. Here, we have shown how deep mutagenesis can inform the engineering of PDGFRα variants that maintain tight virus binding but have lost growth factor interactions, and are thereby orthogonal to normal human signaling pathways. This is anticipated to improve both efficacy and safety in vivo.
There are no simple, tractable animal models for investigating HCMV infection. Two highly relevant models are the study of rhCMV in rhesus macaques [71] or the engraftment of human tissues into immune-deficient mice [72][73][74], but these are non-trivial and require specialized expertise. However, it might be possible to test biocompatibility and pharmacokinetics of the orthogonal receptors independent of infection in small animal models. For instance, murine and human PDGF ligands and receptors share high identity (> 85%, with all key binding site residues of PDGFRα conserved), and one standard assessment of purified human PDGF ligands is to measure mitogenic activity towards mouse fibroblast lines [75,76]. This suggests the human and mouse PDGF ligands and receptors will cross-react, though this has not been quantitatively characterized. Mice, especially strains expressing human FcRn for human-like serum stability of Fc fused protein [77,78], may therefore be a suitable model for assessing safety during future development.
The binding site for HCMV gO on PDGFRα is uncertain from the deep mutational scan, however the data unambiguously shows that binding is independent of properly folded structure in the PDGFRα D2-D3 domains. Deletion studies have show that the D3 domain is essential for trimer-mediated HCMV entry [37], although no specific hot spot residues for mediating physical binding have been defined. It is possible that gO makes critical contacts to domains D4 or D5 that are outside the mutagenized region in the PDGFRα D2-D3 library, or simply that negative effects of single amino acid substitutions in the D2-D3 domains are too small to cause an observable decrease in binding to HCMV trimer by flow cytometry. Determining how gO binds PDGFRα at atomic resolution will be important for speculating on the likelihood of HCMV acquiring mutations to distinguish and escape from orthogonal receptors.
In our own prior studies, deep mutagenesis of human membrane proteins was used to understand how sequence relates to folding based on surface localization, as misfolded sequences were retained intracellularly due to quality control machinery [50,51,79]. However, PDGFRα is an exception, and nearly every missense mutation still reaches the plasma membrane, including those that are incompatible with the native folded structure (for example, mutations that disrupt disulfides, create cavities, impose steric clashes, or introduce buried charges). The structural organization of PDGFRα as a series of loosely connected β-sandwich domains may facilitate surface trafficking of folding mutants, whereas previous studies have focused on membrane proteins with complex tertiary and quaternary structures.
Finally, deep mutagenesis has been extensively applied to viral surface proteins based on virus replication and propagation in tissue culture as the selection, often in the presence of

PLOS PATHOGENS
Engineered receptors for human cytomegalovirus neutralizing antibodies to discover principles of immune recognition and escape [80,81]. However, it is by decoupling the sequence library from a virus genome [52] that we have been able to determine a mutational landscape of a virus receptor. The principles and approach outlined here can be easily adapted for the engineering of other viral receptors for improved efficacy and safety to treat or prevent infections. Indeed, the same methods we optimized for the development of a HCMV antiviral were rapidly repurposed for engineering soluble decoy receptors that bind with high affinity to the spike of SARS coronavirus 2, a zoonotic pathogen that recently spilled over to the human population and has profoundly impacted world events [82]. Receptor-based antiviral drugs may be especially desirable for emerging zoonotic viruses or viruses with high sequence diversity, as they potentially have broader specificity than monoclonal antibodies to target all virus strains that utilize a common receptor.

Plasmids
Human PDGFRα isoform 1 (GenBank NM_006206.4) was cloned in to the NheI-XhoI sites of pCEP4 (Invitrogen) with a N-terminal HA leader (MKTIIALSYIFCLVFA), myc-tag, linker (GSPGGASG), and followed by the mature polypeptide (a.a. 24-1,089). For measuring signaling activity, a minimal promoter under the control of tandem serum response elements was subcloned from SRE reporter vector_559 (Addgene # 82686) [64], a GFP-PEST fluorescent reporter [63] was inserted at the AscI site, and the entire reporter cassette was inserted at the NruI site of pCEP4-myc-PDGFRα by Gibson assembly. As a negative control, human PDGFRβ isoform 1 (GenBank NM_002609.3) was similarly cloned between the NheI-XhoI sites of pCEP4 with a N-terminal HA leader, myc-tag, and linker that connects to the mature protein (a.a. 106). No interactions between PDGFRβ and HCMV trimer were observed. Soluble PDGFRα-Fc was cloned in to the NheI-XhoI sites of pcDNA3.1(+) (Invitrogen), and encompassed PDGFRα a.a. 1-524, a short connecting linker as outlined in S5A Fig, and the Cterminus of human IgG1 (GenBank KY432415.1) beginning at either C220 or D221 as described in Results. Alternatively, PDGFRα a.a. 1-524 were fused via linker GGGS to D221-K447 of human IgG3 (GenBank P01860.2; this is an R435 allele with reduced binding to FcRn). Synthetic human codon-optimized gene fragments (Integrated DNA Technologies) for HCMV gO (GenBank ABV71596.1 for TB40/E strain, AJY56739.1 for Merlin strain) were genetically fused at the C-terminus to superfolder GFP [55] (for detection of soluble HCMV trimer binding to PDGFRα-positive cells) or to an 8-histidine tag (for expression of membrane-tethered HCMV trimer), and were ligated in to the NheI-XhoI sites of pcDNA3.1(+). Codon-optimized gene fragments for full-length HCMV gH (GenBank ABV71597.1 for TB40/ E strain, YP_081523.1 for Merlin Strain) were cloned in to the NheI-XhoI sites of pcDNA3.1 (+) for expression of membrane-tethered HCMV trimer. For production of soluble trimer, the extracellular region of gH with the leader peptide (a.a. 1-717 for TB40/E, a.a. 1-716 for Merlin) was subcloned with a 8-histidine tag. Codon-optimized synthetic genes for gL (GenBank ABV71629.1 for TB40/E strain, YP_081555.1 for Merlin) were cloned with C-terminal FLAG tags in to the NheI-XhoI sites of pcDNA3.1(+). Targeted mutations were made by overlap extension PCR. All plasmids were sequence verified (ACGT, Inc) and are deposited with Addgene.

Cells and viruses
Expi293F cells (ThermoFisher) are a suspension culture derivative of HEK293, and were cultured in Expi293 Expression Medium (ThermoFisher) at 125 rpm, 8% CO2, 37˚C. MRC-5 embryonic lung fibroblasts and ARPE-19 adult retinal pigment epithelial cells were from the

Flow cytometry analysis of soluble HCMV trimer binding to receptor
Soluble HCMV trimer was prepared by transfecting Expi293F cells at a density of 2 × 10 6 / ml with 400 ng pcDNA3-gO-sfGFP, 400 ng pcDNA3-gH-8his and 400 ng pcDNA3-gL-FLAG per ml culture using Expifectamine (ThermoFisher). Transfection Enhancers (ThermoFisher) were added 18 h later, and 4 days post-transfection the culture was centrifuged (1,000 × g, 10 minutes, followed by a second spin of the supernatant at 21,000 × g, 5 minutes). The medium supernatant was stored at -20˚C and used directly in binding experiments. Expi293F cells at 2 × 10 6 / ml were transfected with plasmids encoding receptors (500 ng DNA per ml of culture) using Expifectamine. 24 h post-transfection the cells were washed with PBS supplemented with 0.2% bovine serum albumin (PBS-BSA), incubated for 30 minutes on ice with anti-myc Alexa 647 (clone 9B11, 1/250 dilution; Cell Signaling Technology) and the indicated dilutions of gH/gL/gO-sfGFP-containing medium, washed twice with PBS-BSA, then analyzed on a BD LSR II flow cytometer. For competition assays, washed cells were pre-incubated with PDGF ligands (R&D Systems) for 2 minutes on ice prior to addition of gH/gL/gO-sfGFP-containing medium.

Deep mutagenesis
The D2-D3 domains (a.a. D123-E311) within plasmid pCEP4-myc-PDGFRα were mutagenized by overlap extension PCR [84] using primers with degenerate NNK codons. The plasmid library was transfected in to Expi293F cells using Expifectamine under conditions previously shown to typically give no more than a single coding variant per cell; 1 ng coding plasmid was diluted with 1,500 ng pCEP4-ΔCMV carrier plasmid per ml of cell culture at 2 × 10 6 / ml, and the medium was replaced 2 h post-transfection. The cells were collected after 24 h, washed with ice-cold PBS-BSA, and incubated for 2 minutes on ice with PDGF-AA, -AB, -BB, and-CC (25 nM of each after addition of soluble trimer-containing medium) and anti-myc Alexa 647 (clone 9B11, 1/250 dilution; Cell Signaling Technology). Medium from cells expressing gH/gL/gO-sfGFP was then added to a final dilution of 1/3, and the cells were incubated on ice for 20 minutes. Cells were washed twice with PBS-BSA, and sorted on a BD FACS Aria II at the Roy J. Carver Biotechnology Center. The main cell population was gated by forward/side scattering to remove debris and doublets, and propidium iodide (1 μg/ml final) was added to the sample to exclude dead cells. Of the myc-positive (Alexa 647) population, the 15% of cells with the highest and lowest GFP fluorescence were collected (Fig 1F) in tubes coated overnight with fetal bovine serum and containing Expi293 Expression Medium. Total RNA was extracted from the collected cells using a GeneJET RNA purification kit (Thermo Scientific),

PLOS PATHOGENS
Engineered receptors for human cytomegalovirus and cDNA was reverse transcribed with high fidelity Accuscript primed with a gene-specific oligonucleotide (RevTrans_PDGFRA_992R, TCATGCAGGTTGACAGCTTC). The diversified region of PDGFRα was PCR amplified as 3 fragments to provide full coverage of the D2-D3 domains during Illumina sequencing. During PCR, flanking sequences on the primers added adapters to the ends of the products for annealing to Illumina sequencing primers, unique barcoding, and for binding the flow cell. Amplicons were sequenced on an Illumina NovaSeq 6000 using a 2×250 nt paired end protocol. Data were analyzed using Enrich [85], and commands are provided in the GEO deposit. Briefly, the frequencies of PDGFRα variants in the transcripts of the sorted populations were compared to their frequencies in the naive plasmid library to calculate an enrichment ratio.

Production of soluble PDGFRα-Fc
Per ml of 2 × 10 6 Expi293F cells, 500 ng of pcDNA3-sPDGFRα-Fc (IgG1) and 3 μg of polyethylenimine (MW 25,000; Polysciences) were mixed in 100 μl of OptiMEM (Gibco) and incubated for 20 minutes at room temperature prior to adding to cells. Transfection Enhancers were added 18 h post-transfection, and cells were cultured for six to seven days. Cells were removed by centrifugation at 600 × g for 20 minutes at 4˚C. Cell debris and precipitates were removed by centrifugation at 18,000 × g for 25 minutes at 4˚C. Supernatant was loaded on to KANEKA KanCapA 3G Affinity sorbent (Pall), and the resin was washed with PBS. Soluble PDGFRα-Fc was eluted with 60 mM sodium acetate pH 3.7, and 1 M Tris pH 8.0 was added to the eluate to neutralize the pH. Eluted sPDGFRα-Fc was concentrated with a centrifugal device (MWCO 100 kDa; Sartorius) and NaCl was added to 150 mM final. The protein was separated on a Superdex 200 Increase 10/300 GL (GE Healthcare Life Sciences) size exclusion chromatography column equilibrated with PBS. Peak fractions were pooled, concentrated, and stored at -80˚C after snap freezing in liquid nitrogen. The proteins were quantified by absorbance at 280 nm using calculated molar extinction coefficients for the monomeric mature polypeptides.
Fusions of sPDGFRα to the Fc region of IgG3 were expressed as described above, and the expression medium was dialyzed against water. Cell debris was removed by centrifugation at 18,000 × g for 25 minutes at 4˚C. Supernatant was loaded on to protein G HTC agarose beads (GoldBio) equilibrated with PBS. Protein was eluted with 100 mM glycine pH 2.5 and the eluate was neutralized by addition of 1 M Tris pH 9.0. Protein was further purified as described for the IgG1 fusions. PDGFRα signaling assay 500 ng of PDGFRα reporter plasmid was transfected into 1 ml Expi293F cells at 2 × 10 6 / ml using Expifectamine. 7.5 μl of 6 μM sPDGFRα-Fc (concentration based on monomer) and 7.5 μl of 2 μM PDGF were mixed, incubated at room temperature for 40 minutes, and then added to 1 ml cells at transfection. Human gamma globulin was from Jackson Immuno Research Labs, and PDGFs were from R&D Systems. Cells were collected 24 h post-transfection and stained with anti-myc-Alexa 647 to detect PDGFRα expression. GFP-PEST reporter expression was measured by flow-cytometry.

Virus neutralization assays
Neutralization assays were performed similarly to Stegmann et al. [38] by serially diluting Fcfused PDGFRα proteins or anti-HCMV immunoglobulin (Cytogam, CSL Behring) in PBS (vehicle) and incubating diluted proteins with HCMV virions for 1 h at 37˚C. Virion-protein mixtures were adsorbed onto target cells for 2 h at 37˚C / 5% CO 2 , washed once with PBS, and infected cells were allowed to recover for 18 h. The ability of each treatment to neutralize HCMV infection was measured using indirect-immunofluorescence by determining the number of cells expressing viral IE1, using anti-IE1 clone 1B12 [83] (generously provided by Thomas Shenk, Princeton University) and counterstaining nuclei with DAPI. A multiplicity of infection of 1 was used for AD169 infections, and a multiplicity of infection of 0.5 was used for TB40/E infections (based on stock titers acquired on MRC-5 fibroblasts).

Structural modeling
The sequence of human PDGFRα was threaded onto the structure of PDGF-BB-bound PDGFRβ (PDB ID 3MJG [44]), with one of the two receptor chains and the D1 domain removed. The model was minimized with ROSETTA using FastRelax [86]. Connectivity was determined using the AverageDegree filter in ROSETTA [87]. Images were rendered with PyMOL (Schrödinger, LLC).

Reagent and data availability
Plasmids are deposited with Addgene. Raw and processed deep sequencing data are deposited in NCBI's Gene Expression Omnibus (GEO) under series accession number GSE138169.

PLOS PATHOGENS
Engineered receptors for human cytomegalovirus Nonsense mutations are red. (C) Agreement between log 2 enrichment ratios from replicates of the UL74-Low sort (negative selection). (D) Log 2 enrichment ratios for nonsynonymous mutations (black) are anticorrelated between the negative and positive selections. Nonsense mutations (red) are depleted from both sorts due to lost surface expression. (E-G) High correlation between conservation scores (calculated by averaging the log 2 enrichment ratios for all nonsynonymous mutations at a given amino acid position) from independent replicates of the UL74-High (E) and UL74-Low (F) sorts. Conservation scores are anticorrelated between the two sorted populations (G). The extracellular domain of PDGFRα (Gln24-Glu524; grey) was fused via a short linker (purple) to the Fc region of IgG1 (green). The "Legacy" sequence corresponds to the commercially supplied protein (R&D Systems) used in prior publications. The sequence was redesigned for this study. (B) Coomassiestained SDS gel (run under denaturing and reducing conditions) of wild-type sPDGFRα-Fc eluted from a protein A column. The monomeric protein MW is predicted to be 82 kD. Additional weight may come from glycosylation and/or anomalous electrophoretic mobility. has no effect on PDGF signaling in culture. PDGF signaling was assessed in Expi293F cells transiently transfected with a PDGFRα reporter plasmid as outlined in Fig 4A. A 3-fold molar excess of wild-type sPDGFRα-Fc (black) blocks 5 nM PDGF-AB signaling. Orthogonal engineered sPDGFRα-Fc V242K (white) does not block signaling and the mutation shows no interactions with PDGFs in competitive binding experiments (Fig 3). Furthermore, wild-type or mutant sPDGFRα-Fc up to 100 nM shows no binding to PDGFRαpositive cells, and we exclude unanticipated interactions between receptor chains. However, sPDGFRα-Fc V242K does cause an increase in signaling of PDGF-A ligands (AA and AB) for unknown reasons. A non-specific carrier effect in which the orthogonal receptor stabilizes the hydrophobic ligand in solution is suspected, especially because the more hydrophobic variant sPDGFRα-Fc Y206S promotes an even bigger increase in PDGF-A signaling (Fig 4B). Human gamma globulin (violet) at the same concentration has no effect in this assay. Data are mean ± SD, n = 2 independent replicates. (TIF)