Human foreskin fibroblasts were cultured in minimum essential medium Eagle (MEM; Sigma–Aldrich, St. Louis, MO, United States) supplemented with 10% heat-inactivated fetal bovine serum (FBS; Gibco Thermo Fisher, Waltham, MA, United States) and 0.5% neomycin (Sigma–Aldrich). Human adult retinal pigmented epithelial cells (ARPE-19; ATCC, Manassas, VA, United States) were cultured in Dulbecco’s modified Eagle medium/nutrient mixture F12 (PAN-Biotech, Aidenbach, Germany) supplemented with 10% FBS and 1% penicillin-streptomycin (Thermo Fisher).

The gO genotype (GT) 4 mutants were generated using the HCMV strain TB40/E as parental strain that was cloned in a bacterial artificial chromosome (BAC) sequence [TB40-BAC4-luc; (Scrivano et al., 2011)]. By “en passant” mutagenesis in Escherichia coli GS1783 (Tischer et al., 2010), the glycoprotein O (gO) sequence of the parental strain (gO GT1c) was fully exchanged by gO GT4. For this, a marker cassette was amplified from pEP-Kan-S (kindly provided by N. Osterrieder) by PCR using primer pair 1 (see Supplementary Table 1). The cassette contained a kanamycin resistance gene, flanked on one site by an 18-bp I-Sce I restriction sequence and a gO GT-specific 50-bp sequence, and on both sites by a Sac I restriction site. A plasmid containing the gO GT4 sequence, interrupted by the previously mentioned 50-bp sequence and a Sac I restriction site was ordered from Eurofins Genomics (Luxembourg). The marker cassette and the plasmid were cut with Sac I and ligated. From the resulting transfer plasmid, a recombination cassette was generated by PCR using primer pair 2 (see Supplementary Table 1). The recombination cassette contained extensions of ∼50 bp sequences on each end, which were homologous to the site of insertion in the TB40-BAC-luc genome. The transfer cassette was electroporated into recombination-competent E. coli GS1783 carrying the TB40-BAC4-luc genome. For generation of cysteine mutants, single cysteines within gO of GT1c and GT4 were mutated to serine. For this, a recombination cassette was generated directly from pEP-Kan-S as described above. The cassette contained the sequence duplication necessary for red recombination and the gO GT-specific sequence harboring the cysteine mutation, and was electroporated into E. coli GS1783 carrying either the TB40-BAC4-luc genome or the TB40-BAC4-luc-gO GT4 genome. Single mutants were further used to generate double cysteine mutants by the described method. After electroporation, recombination-positive E. coli were subjected to kanamycin selection, and the introduced non-HCMV sequences were removed within E. coli by cleavage at the I-Sce I site and a second red recombination. Positive, i.e., kanamycin-sensitive, bacteria were selected, and overnight cultures were stored at -80°C until further use. Mutant BACs were confirmed by sequencing.

Bacterial artificial chromosome-DNA was purified from E. coli using the Nucleobond BAC100 kit (Macherey-Nagel, Düren, Germany). The pp71-encoding plasmid pCMV71 (kindly provided by Mark Stinski, University of Iowa) was transformed into NEB10-beta competent E. coli (New England Biolabs, Ipswich, MA, United States) by heat-shock, and subsequently purified from overnight cultures using Qiaprep Spin Miniprep kit (Qiagen, Hilden, Germany). For transfection, HFFs were seeded in 6-well plates (3 × 10⁵ cells/well). The following day, 2 μg of BAC DNA and 1 μg of pCMV71 DNA were mixed together with 9 μl of ViaFect reagent (Promega, Madison, WI, United States) and 100 μl of MEM. The mixture was incubated for 15 min at room temperature and then added to the cells, which were incubated further at 37°C. 24 h after transfection, cells were washed with PBS and fresh MEM was added. One week after transfection, cells were trypsinized and transferred into 75 cm² cell culture flasks. Cells were screened regularly by light microscopy for the appearance of cytopathic effect (CPE), and supernatants were tested for HCMV-DNA load by qPCR. When CPE was 90–100%, supernatants were centrifuged at room temperature for 20 min at 4000 × g. Of note, CPE of gO GT4-C336S remained restricted to single foci even after HFFs were detached by trypsin and reseeded two times during the course of reconstitution (at days 13 and 18 post transfection) to enhance viral spread. Thus, for gO GT4-C336S virus stocks were harvested when similarly high HCMV-DNA levels were reached. Cleared supernatants, containing virus particles, were stored as viral stocks in aliquots at -80°C, before further usage for infection and replication analyses. All stocks were subjected to whole genome sequencing.

In order to remove non-encapsidated viral DNA and free cellular DNA, viral stocks and cell culture supernatants were treated with Turbo DNase (Thermo Fisher). For this, 100 μl of master mix (73 μl H₂O, 20 μl 10x DNase buffer, 5 μl 10x PBS, 2 μl Turbo DNase (2 μ/μl)) were added to 100 μl of sample and incubated for 1 h at 37°C in a thermoshaker at 1400 rpm. Thereafter, 2 μl of Protease (100 mg/ml, Qiagen) were added and incubated for 30 min at 37°C in a thermoshaker at 1400 rpm. Viral or cellular DNA was extracted from DNase-treated samples or from non-treated cell suspensions using the bead-based NucliSens EasyMag extractor (BioMérieux, Marcy-l’Étoile, France). DNA was eluted in 50 μl of nuclease-free H₂O.

For quantification of viral genomes, a region within US17 of the HCMV genome was amplified using the forward primer GCGTGCTTTTTAGCCTCTGCA (10 pM), the reverse primer AAAAGTTTGTGCCCCAACGGTA (10 pM), and the TaqMan probe FAM-TGATCGGCGTTATCGCGTTCTTGATC-TAMRA (2 pM). For quantification of cellular genomic DNA (gDNA), a region within the human beta-2-microglobulin gene was amplified using the forward primer TGAGTATGCCTGCCGTGTGA (3 pM), the reverse primer ACTCATACACAACTTTCAGCAGCTTAC (3 pM), and the TaqMan probe FAM-CCA TGTGACTTTGTCACAGCCCAAGATAGTT-TAMRA (1 pM). Primers and probe (all TIB Molbiol, Berlin, Germany) were mixed with 12.5 μl of TaqMan Universal PCR Master Mix (Applied Biosystems, Foster City, CA, United States) to a volume of 20 μl, which were added to 5 μl of extracted DNA for amplification. PCR amplicons were detected using the ABI 7300 real-time PCR system (Applied Biosystems). The results were compared to an HCMV standard (strain AD169; Advanced Biotechnologies Inc., Eldersburg, MD, United States) and calculated as genome copies/ml. A defined number of human cells was used as standard for calculation of cellular genomic DNA (gDNA).

For determination of infectious viral titer, limiting dilution assay was performed. HFFs were seeded in 96-well plates (1 × 10⁴ cells/well) and infected with serial dilutions of viral stock or cell culture supernatant in eight replicates per dilution. 14 days after infection, cells were screened by light microscopy for CPE, and viral titers were calculated as TCID₅₀/ml by the Reed–Muench method.

Human foreskin fibroblasts and ARPE-19 cells were seeded in white, clear, flat-bottom 96-well plates (Corning, Corning, NY, United States) at a density of 1 × 10⁴ cells/well. The following day, viral stocks were treated with DNase as described above, the number of encapsidated genomes of each aliquot was determined by qPCR, and the stocks diluted to the indicated concentrations in cell culture medium; 100 μl of viral dilution per well were used to infect the cells in triplicates for 2 h at 37°C. (MOI of 1 corresponds to 3.59 × 10⁸ encapsidated genomes/ml in gO GT1c). Cells were washed three times with PBS, supplied with 100 μl of medium or medium supplemented with 300 μg/ml of phosphonoacetic acid (PAA; Sigma–Aldrich) and incubated further at 37°C. 48 h post infection, luciferase assay of cell lysates was performed according to the manufacturer’s protocol (Steady-Glo Luciferase Assay System, Promega) and relative light units (RLU) were measured in a Victor Light 1420 plate reader (PerkinElmer, Waltham, MA, United States). Mean RLUs of triplicates were calculated for evaluation.

One day prior to infection, HFFs and ARPE-19 were seeded in 24-well plates at a density of 1 × 10⁵ cells/well. Viral stocks were diluted in the appropriate cell culture medium. In Figure 1 cells were infected with an MOI of 0.1. In Figure 5 initial infectivity was normalized for each strain to yield 500–1500 RLUs at 48 h post infection. Cells were infected with the RLU-adjusted viral dilutions, and 48 h post infection the RLUs were again determined to confirm normalized infectivity of all strains (see Supplementary Table 1). In Supplementary Figure 1 cells were infected with the same number of encapsidated genomes (3.59 × 10⁷/ml) corresponding to an MOI of 0.1 of gO GT1c. In each experiment cells were infected in triplicates for 2 h at 37°C, washed three times with PBS and supplied with 1 ml of medium. Supernatants and cells were harvested separately at indicated time points over a time course of 2 weeks. Viral load and human genomic DNA (gDNA) were determined by qPCR. For cell-associated viral DNA, viral genomes were normalized to the same copy number of human gDNA. Growth curves were generated by assessing viral titers (TCID₅₀/ml) of cell-free supernatants at the indicated time points.

DNA from BAC purification (as described above) and extracted DNA from DNase-treated viral stocks and cell culture supernatants were quantified using the Qubit 2.0 fluorometer (Thermo Fisher) according to the manufacturer’s instructions. One to 2 ng of DNA per sample were taken for library preparation using the Nextera XT DNA Library Preparation Kit and uniquely indexed samples using the Nextera XT Index Kit were pooled and sequenced together. Pooled libraries were sequenced with paired-end reads (2 × 150–250) on a MiSeq system (Illumina, San Diego, CA, United States). Data were analyzed by CLC genomics workbench 8 software (Qiagen). Average >Q30 score ranged between 66 and 92% of reads, with average 867-fold coverage. Low-quality reads were trimmed before alignment to the respective reference sequence.

For immunoblotting, supernatants from infected HFFs were harvested when cells displayed ∼90% CPE. Supernatants were precleared from cellular debris by centrifugation for 10 min at 4000 × g, before centrifugation for 80 min at 70000 × g at 10°C. The virion pellets were resuspended in TAN buffer (0.05 M triethanolamine, 0.1 M NaCl, pH 8) on ice, and stored in aliquots at -80°C.

For sample preparation, virus stocks were mixed undiluted or diluted in TAN buffer with an equal volume of reducing 2x sample buffer (125 mM Tris/Cl pH 6.8, 6% SDS, 10% glycerol, 10% 2-mercaptoethanol, 0.01% bromophenol blue) and incubated on ice for 10 min before boiling at 95°C for 10 min. Samples were separated on 10% SDS PAGE gels together with a high-range rainbow marker (Amersham ECL High-Range Rainbow Molecular Weight Marker, GE Healthcare, United Kingdom). Separated proteins were transferred to polyvinylidene difluoride (PVDF) membranes (Immun-Blot, Bio-Rad, Richmond, CA, United States) in blotting buffer (40 mM Tris, 39 mM glycine, 1.3 mM SDS, 20% methanol), which were than incubated overnight in blocking buffer (PBS, 1% BSA, 0.1% Tween-20) at 4°C. All antibodies were diluted in blocking buffer. Primary mouse anti-major capsid protein (MCP) mAb, anti-gH (AP86-SA4) mAb (both kindly shared by Michael Mach), and anti-gO.02 mAb (kindly provided by Barbara Adler) were incubated for 2 h at RT. Sheep, anti-mouse IgG-HRP (Amersham, GE Healthcare, United Kingdom) was used as secondary antibody and incubated for 1 h at RT. SuperSignal West Femto Maximum Sensitivity substrate (Thermo Fisher) was applied to the membranes according to the manufacturer’s instructions. Chemiluminescent signals were visualized and analyzed using the ChemiDoc Imager and the Image Lab 6.0 software (both Bio-Rad).

Relative epithelial cell tropism between parental and mutant gO strains (Figure 2), and fibroblast and epithelial cell infectivity of cysteine mutants compared to their parental strains (Figure 3) were analyzed by unpaired two-tailed t-test. To compare infectivity between the two GT1c clones and two GT4 clones (Figure 3), differences in mean RLUs were analyzed by one-way ANOVA test. Mean values from three to eight independently repeated experiments were used for statistical analyses. P-values < 0.05 were considered significant. GraphPad Prism version 7.01 was used for statistical analyses.

To investigate whether the exclusive exchange of the full-length open reading frame (ORF) of gO with a different genotype sequence influences the growth properties of HCMV, the gO GT4 mutant was generated. We used the BAC clone TB40-BAC4-luc as parental strain to seamlessly replace its resident gO GT1c sequence by gO GT4. As previously reported, TB40-BAC4-luc (from now on termed gO GT1c) carries a luciferase expression cassette which allows a sensitive and quantitative measurement of infection over a wide range of multiplicities of infection (MOI) (Scrivano et al., 2011). After “en passant” mutagenesis the genomic integrity of two resulting viral BAC-mutant gO GT4 clones (Table 1 and Supplementary Table 1) which showed complete replacement of GT1c ORF (464 aa) by GT4 ORF (457 aa) was confirmed by whole genome deep sequencing. Reconstitution of gO GT1c and gO GT4 mutant BAC clones in HFFs resulted in infectious viruses with similarly high TCID₅₀ and similar copy numbers of encapsidated viral genomes (Table 1). Both, gO GT1c and gO GT4 induced the characteristic HCMV CPE, with even distribution over the entire HFF monolayer. Hence, full-length replacement of gO GT1c by GT4 sequence allowed the generation of infectious virus particles upon reconstitution in HFFs. Whole HCMV genome sequencing of both gO GT1c (GT1c_1 and GT1c_2) and gO GT4 mutant virus stocks (GT4_1 and GT4_2) confirmed that no additional mutations emerged during reconstitution.

TB40E-derived strain TB40-BAC4-luc is able to infect and replicate in a broad range of different cell types including fibroblasts and epithelial cells (Scrivano et al., 2011). To compare the replication efficiencies of the two HFF-derived gO GT4 mutants, GT4_1 and GT4_2, with that of two independently generated HFF-derived gO GT1c virus stocks (GT1c_1 and GT1c_2), HFFs and ARPE-19 cells were infected at an MOI of 0.1. Over a time course, the production of cell-free infectious virus was assayed by limiting dilution assay (TCID₅₀) and the amount of HCMV-DNA in supernatant was determined by HCMV-specific qPCR. As shown in Figure 1, in HFFs both gO GT4 mutants displayed similar release of cell-free infectious virus and HCMV-DNA load in comparison to the two gO GT1c viruses. Also, the endpoint titers (Figure 1A) and endpoint HCMV-DNA levels were similarly high in gO GT1c and gO GT4 strains (Figure 1B). In ARPE-19 cells, however, both gO GT4 clones produced 0.89–1.36 log₁₀ TCID₅₀/ml higher endpoint titers than the gO GT1c clones (Figure 1C). In overall, the HCMV-DNA load kinetics in supernatant paralleled the kinetics of cell-free virus production (Figure 1D).

These findings demonstrate that despite replacement of the entire resident gO gene sequence of TB40-BAC4-luc by a different gO gene sequence, the resulting viruses retained the growth capacity in different cell types.

As gO GT4 displayed a subtle enhancement for growth in ARPE-19 cells compared to GT1c the relative epithelial cell tropism was explored. Therefore, HFFs and ARPE-19 cells were synchronously infected at a low MOI (0.1–0.4) by gO GT1c and gO GT4, and after 48 h cell lysates were assayed for expression of luciferase as indicator for viral infection. A number of independent experiments was performed either with or without the replication inhibitor PAA. In each experiment RLUs were determined in triplicates and the ARPE-19 to HFF infection ratio in percent was determined. As shown in Figure 2, each of the two gO GT4 mutant viruses, GT4_1 and GT4_2, showed a higher epithelial cell tropism compared to gO GT1c viruses GT1c_1 and GT1c_2. In the presence of PAA (Figure 2A) the epithelial cell tropism was significantly increased for GT4_1 when compared to GT1c_1 (p = 0.0292), and for both GT4_1 and GT4_2 when compared to GT1c_2 (p = 0.0148 and 0.0169, respectively). Similarly, in the absence of PAA (Figure 2C), the ARPE/HFF infection ratio of GT4_1 was significantly enhanced when compared to GT1c_1 (p = 0.007), and both of GT4_1 and GT4_2 when compared to GT1c_2 (p = 0.0005 and 0.0318), respectively. Similarly as reported by Scrivano et al. (2011), we also observed that the values obtained with this assay are influenced by varying quality of cells and passage number. Nevertheless, when plotting single assays (Figures 2B,D), the pattern of enhanced epithelial tropism of both GT4 clones is always maintained.

These data show that replacement of gO GT1c by gO GT4 results in a higher epithelial cell tropism and this most likely leads to the higher release of infectious gO GT4 viruses than gO GT1c viruses when the same MOI is used for epithelial cell infection.

Next, to gain more insight into gO genotype-dependent variances we aimed to compare the phenotype of both gO genotypes upon introduction of point mutations. Since mass spectrometry and mutagenesis analysis of the trimeric gH/gL/gO complex recently revealed that the amino acid residues gO-Cys-351 and alternatively gO-Cys-226 (numbering according to gO GT5 of Merlin strain) play important roles in trimer formation and/or stability (Ciferri et al., 2015; Stegmann et al., 2016), we mutated these highly conserved cysteines to serine in gO GT1c (C218S and C343S) and gO GT4 (C216S and C336S), either individually or in combination (Table 1 and Supplementary Figure 1). During the course of reconstitution in HFFs from all newly generated BAC mutant clones, the CPE was visually inspected by light microscopy and the release of HCMV-DNA in supernatant was measured by qPCR (Table 1). The genome sequence correctness of all clones was confirmed before and after reconstitution.

Among all single cysteine mutants only the two gO GT4-C336S clones displayed a CPE that differed from the parental strain-like phenotype (even distribution over the whole monolayer) in that only few scattered clusters of infected cells were seen (Table 1). Virus stocks were harvested from the supernatant 27–28 days post transfection when all cysteine mutants reached comparable levels of HCMV-DNA. While also the genome copy numbers of encapsidated HCMV-DNA were similar among all single cysteine mutants, the TCID₅₀/ml values which were titrated on HFFs revealed a 2 log₁₀ TCID₅₀/ml lower virus titer for the two gO GT4-C336S mutant clones compared to parental strain gO GT4 (Table 1). Furthermore, the calculated difference in mean log₁₀ values between encapsidated genomes and TCID₅₀/ml was 5.10 and 5.15 for the GT4-C336S mutants, while for all other strains it ranged from 2.45 to 3.49.

When both cysteine residues, C218 and C343 in GT1c or C216 and C336 in GT4, were mutated, no infectious particles could be detected in supernatant upon reconstitution in HFFs. However, at day 48 post transfection and after several rounds of cell reseeding, the gO GT1c-C218/343S double mutant displayed single foci of infected cells and this phenotype resembled the phenotype of the gO deletion mutant (Jiang et al., 2008; Wille et al., 2010) (Table 1). No viral titer was determinable and, hence, no further experiments were performed with the double mutants.

Since the gO GT4-C336S mutants were severely reduced in HFF-assessed viral titer despite normal amounts of encapsidated genomes when compared to all other viral strains, this is indicative that mutation of C336 in GT4 substantially affects HFF infectivity. Thus, in a next step, we aimed to determine the fibroblast and epithelial cell infection capacities of all cysteine mutant viruses in comparison to their parental strains gO GT1c and gO GT4, respectively. To this end, HFFs and ARPE-19 cells were infected with the same amount of encapsidated virus genomes, corresponding to an MOI of 1 in gO GT1c, to ensure that the same numbers of virus particles were used for the parental and cysteine mutant strains. Infectivity was quantified by monitoring luciferase expression in cell lysates 48 h post infection as mentioned above and three independent experiments were performed. As shown in Figures 3A,B and Table 2, while both gO GT1c cysteine mutants, gO GT1c-C218S and gO GT1c-C343S, displayed a similar infectivity for HFFs as their parental strains GT1c_1 and GT1c_2, the mutation of C336S in gO GT4 led to reduction in infectivity, which was significant when compared to GT4_1 (p = 0.0025) and highly significant when compared to GT4_2 (p < 0.0001). The markedly reduced infectivity of gO GT4-C336S compared to parental strain GT4 corresponds well to the above mentioned reduced virus titer despite normal number of encapsidated genomes upon reconstitution in HFFs (Table 1).

Also, the relative ARPE-19 cell infectivity was not significantly affected by the single mutations, C218S and C343S, in gO GT1c (Figure 3C). Notably, subtle differences might be masked due to the rather high variability among distinct virus stocks of GT1c and GT1c-C343S, respectively. In contrast, in the background of gO GT4, both corresponding single mutants GT4-C336S and GT4-C216S, displayed significantly reduced infectivity (Figure 3D) compared to the parental strains gO GT4_1 and GT4_2 with a reduction of 22–23% of mean log₁₀ RLU caused by gO GT4-C216S (p < 0.005) and an even higher reduction of 37% (GT4-C336S_1) to 44% (GT4-C336S_2) attributed to the single point mutation C336S in GT4 (p < 0.005).

As shown in Table 2, the resulting ARPE-19/HFF ratios were similar between the GT1c cysteine mutants and the two parental gO GT1c clones. This contrasts to GT4 where the mutation C216S led to a decrease in the ARPE-19/HFF ratio, and C336S even caused an increase in epithelial cell tropism when compared to their respective parental gO GT4 strains. This indicates that despite an overall reduction of infectivity, the gO GT4-C336S mutants retained an enhanced epithelial cell tropism. Taken together, these data demonstrate that the impact of gO cysteine mutations on cell-free infectivity for fibroblasts and epithelial cells significantly differs between the two gO genotypic backgrounds, GT1c and GT4.

Of note, the cell-free infectivity of the GT4 clones compared to the GT1c clones was similar in fibroblasts (Figures 3A,B; p = 0.17; ANOVA) showing a mean fold increase of 1.2 (range: 0.7–2.3) for GT4_1 and 1.8 (range: 1.0 – 2.6) for GT4_2. The cell-free infection capacity for epithelial cells, however, was significantly different between the GT4 and GT1c clones (p = 0.0046; ANOVA) (Figures 3C,D). The mean fold increase in epithelial cell infectivity was 7.0 (range: 3.2–12.2) for GT4_1 and 6.3 (range: 4.4–9.1) for GT4_2. These data are well in line with the findings that gO GT4 strains displayed an enhanced epithelial cell tropism as shown in Figure 2.

Glycoprotein O appears to be present in the virions only in a complex with gH/gL and it is suggested that the abundance of gH/gL/gO correlates with the infectivity of the respective virions (Murrell et al., 2013; Zhou et al., 2013, 2015). Thus, we next determined whether the amount of gO and gH in extracellular virions of parental and mutant gO GT1c and GT4 strains differed according to their varying infection capacities as shown above. The gel loads for Western blot analyses under reducing conditions were normalized to MCP which allowed to compare the gO and gH content in the parental und mutant HCMV virions. As shown in Figure 4A, the amounts of gO and gH were similar between GT1c and GT4 with a trend toward slightly higher gH in GT4 virions. Quantitative comparison of gO levels among the cysteine mutant virions showed that GT4-C336S contained the lowest amount of gO, followed by GT1c-C343S, GT4-C216S, and GT1c-C218S (Figures 4B,C). The severely reduced gO level in GT4-C336S virions corresponds well with the pronounced infection defect as shown above. Consistent with recent findings by Stegmann et al. (2016), the rather low amounts of gO in GT1c-C343S appear to be sufficient to maintain its proper infection capacities for fibroblasts and epithelial cells. Also, the gH contents varied among the gO cysteine mutants, yet the lowest gH levels were observed in GT1c-C343S and not in GT4-C336S virions (Figure 4B). These data indicate that mutation of the highly conserved gO cysteine C343 in GT1c or the corresponding C336 in GT4 resulted in a substantial change in the gO and gH content, the extent of reduction, however, depends on the gO genotypic background.

Finally, we investigated how the cysteine mutations GT1c-C343S and GT4-C336S affect the viral growth behavior in HFFs and ARPE-19 cells in comparison to the corresponding parental strains GT1c and GT4. Since in the previous experiments differences in infectivity among our viral strains were detected (Figures 2, 3), a preliminary dose finding experiment was performed to normalize infectivity of GT1c, GT1c-C343S, GT4, and GT4-C336S to 500–1500 RLUs at 48 h post infection in both cell types individually. This allowed for a matched initial infectivity for the growth curves of all four strains. Then, both cell types were infected with the previously determined virus dilutions (Supplementary Table 2), and cultured for 13 (HFFs) or 15 (ARPE-19) days. The kinetics of cell-free and cell-associated HCMV-DNA were monitored at multiple time-points post infection by qPCR (Figures 5A–D).

In HFFs, GT1c, GT1c-C343S, and GT4 displayed similar levels of cell-free and cell-associated HCMV-DNA over the entire time span, and comparable endpoint HCMV-DNA levels (Figures 5A,B). The GT4-C336S mutant, on the other hand, showed distinct HCMV-DNA kinetics. In the cell-free fraction, the levels of HCMV-DNA of this mutant were elevated over the first 6 days. This, however, might rather be explained by remnants of the higher amounts of virions used for initial infection (Supplementary Table 2) than by a new production of cell-free HCMV-DNA, since no differences were observed when equal amounts of encapsidated virions were used for initial infection (see Supplementary Figure 1). Also, cell-associated HCMV-DNA kinetics of GT4-C336S until day 10 were rather similar to that of the other strains, independently of whether same amounts of encapsidated virions or normalized infectivity was used for initial infection, which further corroborates that the replication efficiency in HFFs is comparable among all strains.

However, the GT4-C336S mutant showed substantially reduced endpoint HCMV-DNA levels, both in the cell-free (1.0 log₁₀ copies/ml difference) and in the cell-associated (0.6 log₁₀ copies/ml difference) fraction (Figures 5A,B). Since only the initial infectivity could be normalized, it is most likely that in subsequent rounds of infection the reduced fibroblast infection capacity of GT4-C336S virions caused reduced endpoint HCMV-DNA levels. Also, the CPE at day 13 post infection was clearly distinct in that the infected cells remained restricted to single foci (Figure 5E). All other strains displayed the characteristic CPE with swollen cells and cell syncytia that were spread over the entire cell monolayer.

In ARPE-19 cells, cell-free HCMV-DNA loads in the early phase post infection were again not uniform between the viral strains due to variety in applied virus amounts (Figure 5C and Supplementary Table 2). Of note, no differences in early viral loads were seen when the initial infection was performed with equal amounts of encapsidated virions (Supplementary Figure 1). At day 9 and later, similar amounts of cell-free HCMV-DNA were detected for all strains. GT1c, GT1c-C343S, and GT4-C336S yielded comparable endpoint DNA loads, while GT4 was slightly elevated, corresponding to the result seen in Figure 2. Further, no differences between the strains could be observed in the cell-associated HCMV-DNA kinetics (Figure 5D) indicating similar replication dynamics with all four virus strains. Further, all four strains induced a similar CPE at day 15 post infection (Figure 5E).

Taken together, these data indicate that the marked impairment in cell-free fibroblast infectivity of the gO GT4-C336S mutant leads to an attenuated growth phenotype with lower endpoint HCMV-DNA levels and the observed CPE pattern. On the other hand, the retained epithelial cell tropism (Table 2) allows the GT4-C336S mutant to grow in a similar fashion as its parental strain GT4, indicating that the cell-to-cell spread is not affected. The corresponding cysteine mutation in the background of gO GT1c, yet shows a GT1c-like growth phenotype in both cell types. This further underlines that these single cysteine mutations in gO GT1c do not substantially impact the function of gO despite the substantial reduction of the gO content in GT1c-C343S virions as shown above.