Specialization for Cell-Free or Cell-to-Cell Spread of BAC-Cloned Human Cytomegalovirus Strains Is Determined by Factors beyond the UL128-131 and RL13 Loci

Both cell-free and cell-to-cell spread are likely important for the natural biology of HCMV. In culture, strains clearly differ in their capacity for cell-free spread as a result of differences in the quantity and infectivity of extracellular released progeny. However, it has been unclear whether “cell-associated” phenotypes are simply the result of poor cell-free spread or are indicative of particularly efficient cell-to-cell spread mechanisms. By measuring the kinetics of spread at early time points, we were able to show that HCMV strains can be highly specialized to either cell-free or cell-to-cell mechanisms, and this was not strictly linked the efficiency of cell-free spread. Our results provide a conceptual approach to evaluating intervention strategies for their ability to limit cell-free or cell-to-cell spread as independent processes.

ABSTRACT

It is widely held that clinical isolates of human cytomegalovirus (HCMV) are highly cell associated, and mutations affecting the UL128-131 and RL13 loci that arise in culture lead to the appearance of a cell-free spread phenotype. The bacterial artificial chromosome (BAC) clone Merlin (ME) expresses abundant UL128-131, is RL13 impaired, and produces low infectivity virions in fibroblasts, whereas TB40/e (TB) and TR are low in UL128-131, are RL13 intact, and produce virions of much higher infectivity. Despite these differences, quantification of spread by flow cytometry revealed remarkably similar spread efficiencies in fibroblasts. In epithelial cells, ME spread more efficiently, consistent with robust UL128-131 expression. Strikingly, ME spread far better than did TB or TR in the presence of neutralizing antibodies on both cell types, indicating that ME is not simply deficient at cell-free spread but is particularly efficient at cell-to-cell spread, whereas TB and TR cell-to-cell spread is poor. Sonically disrupted ME-infected cells contained scant infectivity, suggesting that the efficient cell-to-cell spread mechanism of ME depends on features of the intact cells such as junctions or intracellular trafficking processes. Even when UL128-131 was transcriptionally repressed, cell-to-cell spread of ME was still more efficient than that of TB or TR. Moreover, RL13 expression comparably reduced both cell-free and cell-to-cell spread of all three strains, suggesting that it acts at a stage of assembly and/or egress common to both routes of spread. Thus, HCMV strains can be highly specialized for either for cell-free or cell-to-cell spread, and these phenotypes are determined by factors beyond the UL128-131 or RL13 loci.

IMPORTANCE Both cell-free and cell-to-cell spread are likely important for the natural biology of HCMV. In culture, strains clearly differ in their capacity for cell-free spread as a result of differences in the quantity and infectivity of extracellular released progeny. However, it has been unclear whether “cell-associated” phenotypes are simply the result of poor cell-free spread or are indicative of particularly efficient cell-to-cell spread mechanisms. By measuring the kinetics of spread at early time points, we were able to show that HCMV strains can be highly specialized to either cell-free or cell-to-cell mechanisms, and this was not strictly linked the efficiency of cell-free spread. Our results provide a conceptual approach to evaluating intervention strategies for their ability to limit cell-free or cell-to-cell spread as independent processes.

INTRODUCTION

Intervention approaches to control human cytomegalovirus (HCMV) infection, including DNA replication inhibitors, such as ganciclovir, and vaccines designed to elicit neutralizing antibodies have had limited success (1, 2), and this may be due in part to the complex genetic diversity of HCMV circulating within human populations (3–9). Basic annotations of the 235-kbp HCMV genome identify 165 canonical open reading frames (ORFs), although there is evidence of extensive transcription and translation beyond these loci (10, 11). While nucleotide polymorphisms are found throughout the genome, most sequences are well conserved. Twenty-one of the 165 canonical ORFs show high nucleotide diversity and are distributed as islands throughout the genome. Several groups have reported evidence of frequent recombination within the more conserved regions, which may essentially mix and match the more diverse loci into many different genotype combinations (3, 4, 7, 8). Adding to this complexity, some studies have shown evidence of gene-inactivating mutations (pseudogenes) and gene deletions (7, 8). While some infected individuals may harbor relatively pure populations of HCMV genotypes, complex, multiple-genotype infections and sequential infections by genetically distinct populations also occur. Cloning of HCMV from clinical samples on bacterial artificial chromosomes (BACs) has resulted in numerous genetically distinct strains for use in laboratory studies. It is not clear how the genetic and phenotypic differences among BAC-cloned HCMVs might reflect their natural diversity.

The glycoprotein H/L (gH/gL) complex is functionally conserved among the herpesvirus family and is involved in the initial receptor engagement and the regulation of the membrane fusion protein gB (reviewed in reference 12). HCMV gH/gL is present in the virion envelope as a gH/gL/gO complex (13, 14), a gH/gL/UL128-131 complex (15–18), and in complex with gB (19). Transient expression of HCMV gH/gL and gB is sufficient to drive cell-cell fusion, but it is unclear which forms of gH/gL contribute to membrane fusion during HCMV infection (20). The gH/gL/gO and gH/gL/UL128-131 complexes bind to various cell surface receptors through the gO and UL128-131 domains (21–23), and these interactions are important for entry into a broad range of cell types. gH/gL/gO is likely critical for infection of most, or all, cell types, with fibroblasts and epithelial and endothelial cells being the most extensively studied (24–27). Mutants of HCMV lacking gO are deficient at adsorption to cells, and soluble gH/gL/gO blocks such adsorption (25, 28). In contrast, gH/gL/UL128-131 is dispensable for the infection of fibroblasts and neuronal cells but critical for epithelial and endothelial cells, monocyte-macrophages, and dendritic cells (16, 18, 29–31). Soluble gH/gL/UL128-131 did not block virus adsorption, but there is evidence that engagement of receptors can elicit signal transduction pathways that influence the nature of the resulting infection pathway (21, 28, 30). Murine and guinea pig CMVs contain homologous gH/gL complexes that play similar roles in cell tropism, although there are some differences in the reported requirements for the complexes for infection of different cell types (32–37).

We and others have noted striking phenotypic variation among the HCMV strains TB40/e (TB), TR, and Merlin (ME) in terms of content of the gH/gL complexes in the virion envelope and the corollary effects on entry and tropism. TB and TR contain gH/gL mostly in the form of gH/gL/gO, and very little gH/gL/UL128-131, whereas ME virions contain overall smaller amounts of gH/gL, and this is mostly in the form of gH/gL/UL128-131 (27). The genetic polymorphism(s) responsible for these differences are unclear. Murrell et al. described a G→T mutation in the UL128 locus of TB that when engineered into ME reduced the assembly of gH/gL/UL128-131 through effects on mRNA splicing (38). However, this did not fully explain the observed strain differences, as TR is also low in gH/gL/UL128-131 but is congenic to ME at this locus (27). Zhang et al. showed that the expression of gO during replication is lower in ME-infected cells than in TR-infected cells, but again, the genetic correlates of this difference were not clearly identified (39). The cell-free infectivity of HCMV strains on both fibroblasts and epithelial cells correlates with the amounts of gH/gL/gO in the virion envelope; TB is by far the most infectious, followed by TR, while ME virions are poorly infectious (27). Repression of the UL131 promoter in ME resulted in virions with dramatically reduced amounts of gH/gL/UL128-131, somewhat higher levels of gH/gL/gO, and dramatically improved cell-free infectivity (27, 40).

HCMV can spread through monolayer cell cultures by the diffusion of cell-free virus in the culture supernatant or by a more direct cell-to-cell mode, but the mechanistic distinctions between these types of spread are not well characterized. Moreover, the pathway of cell-to-cell spread for HCMV has not been extensively studied. There have been suggestions of limited fusion between infected and uninfected cells allowing the transfer of subviral components, but the efficiency of these processes to facilitate the spread of HCMV infection is not clear (41–43). Syncytium formation has been observed in HCMV-infected cultures, but it is not clear whether these cell-cell fusions contribute to viral spread or merely represent the coalescence of late-stage-infected cells, long after progeny virus has exited and spread to adjacent cells (44, 45). Different HCMV strains can be characterized as cell free or cell associated based on the appearance of the foci formed in a culture monolayer. For example, ME is considered to be cell associated due to the formation of tightly localized foci, whereas TB forms more diffuse focal patterns characteristic of cell-free spread (40, 46). However, it is not clear whether the apparent cell-associated nature of ME simply reflects the lack of efficient cell-free spread or a specific mechanism to facilitate efficient cell-to-cell spread, nor is it clear how strains like ME and TB compare in their abilities to spread cell to cell. Finally, the genetic loci responsible for the apparent cell-free and cell-associated natures of strains are not clear. Some reports have monitored the size of foci in the presence of neutralizing antibodies and attributed the cell-associated nature of ME to the expression of the UL128-131 and RL13 loci (24, 47). It seems clear that either gH/gL/gO or gH/gL/UL128-131 is required for cell-to-cell spread on fibroblasts, whereas gH/gL/UL128-131 seems to be required for cell-to-cell spread in epithelial or endothelial cells (24, 26, 48, 49). Here, we report the use of flow cytometry to compare the spread of HCMV TB, TR, and ME in fibroblasts and epithelial cells at early time points. The results show that these strains differ not only in their ability to spread cell free but also cell to cell and that neither UL128-131 nor RL13 is the determinate locus for the mode of spread.

RESULTS

Representative HCMV strains differ not only in their efficiencies of cell-free spread but also in their ability to spread cell to cell.

Focal sizes and patterns can vary considerably among strains of HCMV and likely reflect differences in their proclivities for cell-free versus cell-to-cell modes of spread. Representative focal spread patterns of green fluorescent protein (GFP)-expressing HCMV strains TB, TR, and ME are depicted in Fig. 1. Cultures of fibroblasts and epithelial cells were infected at low multiplicities, and spread was monitored by fluorescence microscopy over 18 days. In fibroblasts, the spread of all three strains was localized to small, tight foci for the first 6 days. By day 12, TB and TR began to show signs of more diffuse spread that continued to increase through day 18, whereas ME foci remained generally smaller and more localized over the entire experiment (Fig. 1A). In contrast, foci in epithelial cell cultures for all three strains were smaller and more tightly localized throughout the experiment (Fig. 1B). The more diffuse focal patterns of TB and TR in fibroblasts were likely indicative of efficient cell-free spread, whereas the localized spread of ME was suggestive of less-efficient cell-free spread. This was consistent with our previous observations that progeny virus released to culture supernatants by TB and TR was far more infectious than was extracellular ME progeny (27). While similar focal spread characterizations have been reported by other groups (24, 38, 40, 50), such comparisons of focal size and patterns are not easily quantified and do not clearly distinguish whether smaller, more localized foci form as a passive result of inefficient cell-free spread or rather reflect an active preference for a specialized cell-to-cell mechanism. Moreover, while it is clear that these strains of HCMV can differ in their ability to spread via cell-free progeny, whether they also differ in their abilities to spread cell to cell has not been addressed.

FIG 1.

Comparison of focal spread patterns among distinct HCMV strains. (A and B) Confluent monolayers of fibroblasts (A) or epithelial cells (B) were infected at an MOI of 0.001 with GFP-expressing HCMV TB, TR, or ME. Foci were documented by fluorescence microscopy at 6, 12, and 18 days postinfection (d.p.i.). Four representative fields (×10 magnification) are shown for each.

For quantitative comparisons of spread among HCMV strains, experiments similar to those described for Fig. 1 were performed, but instead of microscopy, flow cytometry was used to measure the increasing number of infected (GFP-expressing) cells over the first 12 days of replication. In fibroblasts, the number of infected cells for both TB and TR increased exponentially over 12 days and fit well to a log-linear regression, indicating a constant rate of spread over the course of the experiment (Fig. 2A and B). In contrast, ME spread considerably faster between days 3 and 6 and then spread slower between days 9 and 12. The average spread rates (lymph node [LN] GFP⁺ cells/day, as indicated by the regression slopes in Fig. 2B) of all three strains were within 1.2-fold of each other, with only the difference between TB and TR reaching statistical significance (Fig. 2C). This was somewhat unexpected, as ME has been described to replicate poorly compared to other strains (38). Thus, the growth of the smaller, more tightly localized foci of ME was characterized by initially rapid expansion, followed by progressive slowing, whereas the more diffuse foci of TB and TR were characterized by a constant rate of expansion. These results highlight how visual analyses of focal size and pattern fail to detect the rapid, early focal expansion of ME. In epithelial cells, spread rates for all three strains were lower than those in in fibroblasts, and there were greater differences between strains. ME spread was approximately 2-fold faster than that of TR and 1.5-fold faster than that of TB, and all differences were statistically significant (Fig. 2D to F). The more efficient spread by ME on epithelial cells was consistent with the levels of gH/gL/UL128-131 being higher than those of TB and TR (27, 38, 51), but the tightly localized focal patterns suggested inefficient cell-free spread by all three strains on this cell type.

FIG 2.

Quantitation of HCMV spread by in fibroblasts and epithelial cells. (A and D) Confluent 9.5-cm³ monolayers of fibroblasts (A) or epithelial cells (D) were infected at an MOI of 0.001 with GFP-expressing HCMV strain TB, TR, or ME. The number of infected cells at 3, 6, 9, and 12 days postinfection was determined by flow cytometry. (B and E) The natural logarithm (LN) of the number of GFP⁺ cells from panels A and D, respectively. (C and F) The average spread rates (LN GFP⁺ cells per day) of the regression lines shown in panels B and E, respectively. All experiments were performed at least three times, and error bars represent the standard deviations of the results from all experiments (omitted if smaller than the data marker). P values were generated using ANOVA with Tukey’s multiple-comparison analysis with a 95% confidence interval (95% CI) (*, P < 0.05; **, P < 0.01; ***, P < 0.001).

Neutralizing antibodies were used to distinguish the contributions of cell-free and cell-to-cell mechanisms to the rate of spread for each strain. Antibodies chosen for these experiments were a mouse monoclonal antibody (MAb) that likely targets a discontinuous epitope at the membrane proximal region of gH (14-4b) (52, 53) and a mixture of rabbit anti-peptide sera that target the epithelial tropism factors UL130 and UL131 (17). The relative potencies of these antibodies to neutralize cell-free TB, TR, and ME were verified in neutralization experiments shown in Fig. 3. On fibroblasts, anti-gH was approximately 10-fold more potent against ME than against TB and TR, and there was a residual 20% TR infectivity that was resistant even at very high antibody concentrations (Fig. 3A). Consistent with previous studies, anti-UL130/131 sera did not neutralize any strain on fibroblasts (Fig. 3B) (27, 54). On epithelial cells, the potency of neutralization by anti-gH and anti-UL130/131 antibodies was more similar among the strains, and complete neutralization of each was achieved (Fig. 3C and D). In all cases, isotope controls showed no effect even at the maximum concentration (Fig. 3, bar graphs to the right of each neutralization curve). Note that experiments on fibroblasts used fibroblast-derived virus, while epithelium-derived virus was used on epithelial cells.

FIG 3.

Antibody neutralization of cell-free HCMV. (A to D) Equal numbers (genomes/ml) of fibroblast-derived (A and B) or epithelium-derived (C and D) HCMV TB, TR, or ME virions were incubated with multiple concentrations of anti-gH MAb 14-4b (A and C) or anti-UL130/131 rabbit sera (B and D) for 1 h at RT. Remaining infectivity was determined by titration on the matched producer cell type and plotted as the percentage of the no-antibody mock. Isotype controls were also tested (A to D, right) at doses of antibodies resulting in complete neutralization of cell-free HCMV. All experiments were performed in triplicate, and error bars represent the standard deviations.

In fibroblast cultures, anti-gH antibodies reduced the spread rates of TB and TR by 70% and 55%, respectively, whereas ME spread was reduced by only 25% (Fig. 4A to C). The apparent resistance of ME spread was especially noteworthy since cell-free ME was more sensitive to neutralization by this antibody than was either TB or TR (Fig. 3A). As expected, no effect on spread was observed with either a mouse MAb against major capsid protein (an isotype control for anti-gH) or the rabbit anti-UL130/131 serum. In all cases, the apparent anti-gH antibody resistant spread was greater than the spread in the presence of ganciclovir, even for TR, which is known to harbor resistance mutations in the UL97 kinase (55). This demonstrated that the observed antibody-resistant spread was not simply the transfer of cytoplasmic contents from infected to uninfected cells but rather represented bona fide viral spread dependent on the production of new viral genomes. These results support the hypothesis that spread of TB in fibroblasts predominately involves cell-free virions that were highly sensitive to neutralizing antibodies, whereas ME spread predominately by a distinct, antibody-resistant cell-to-cell mechanism. The intermediate inhibition of TR spread by anti-gH antibodies might indicate intermediate contributions of cell-free and cell-to-cell spread mechanisms. However, it may also reflect the lack of complete neutralization of cell-free TR as described in Fig. 3A.

FIG 4.

Effects of neutralizing antibodies on spread of HCMV strains in fibroblasts and epithelial cells. Confluent 9.5-cm³ monolayers of fibroblasts or epithelial cells were infected at an MOI of 0.001 with GFP-expressing HCMV strain TB, TR, or ME and allowed to spread in the presence of either anti-gH MAb 14-4b, anti-UL130/131 rabbit serum, isotype control antibody, or ganciclovir (Gan). (A to F) The average spread rates over 12 days (LN GFP⁺ cells per day; as in Fig. 2) for TB, TR, and ME on fibroblasts (A to C) and epithelial cells (D to F) are shown. n.d., conditions where no spread was detected. All experiments were performed three times, and the error bars represent the standard deviations. P values were generated by ANOVA with Tukey’s multiple-comparison analysis with 95% CI (*, P < 0.05; **, P < 0.01; ***, P < 0.001).

In epithelial cell cultures, the presence of anti-gH antibodies had no effect on the spread rates for any strain, whereas the anti-UL130/131 sera resulted in 20% to 30% reductions in spread rates for all three strains (Fig. 4D to F). It was unexpected that the anti-UL130/131 antibodies reduced spread while the anti-gH did not since both antibodies potently neutralized all three strains on epithelial cells (Fig. 3C and D). However, the preimmune control rabbit serum also reduced the spread by 20% to 30% for all three strains, suggesting nonspecific effects on the cells rather than inhibition of cell-free spread (Fig. 4D to F). Together, these data indicate that cell-to-cell spread is the predominant mode for all strains in epithelial cells and that ME spread is considerably more efficient than that of either TB or TR.

HCMV ME is uniquely efficient at cell-to-cell spread via a mechanism that depends on the intact cell monolayer and overcomes the low infectivity of progeny virus.

The efficient cell-free spread of TB in fibroblasts was likely a function of the quantity and infectivity of progeny virus release to culture supernatants. Similarly, it was expected that the highly efficient cell-to-cell spread of ME would correlate with the quantity and/or infectivity of progeny virus within the infected cells, or “cell-associated virus.” To test these hypotheses, infected cultures were separated into supernatants and cells, and the cells were further fractionated into nuclear and cytoplasmic portions. This fractionation allowed for more detailed comparisons of DNA replication and nuclear egress between strains. The quantity and infectivity of progeny virus within the infected cells and in the supernatants were then determined, and the results are shown in Fig. 5.

FIG 5.

Quantity and infectivity of cell-free and cell-associated HCMV progeny virus. Replicate cultures of fibroblasts or epithelial cells were infected at an MOI of 1 with TB, TR, or ME. Seven days later, infected cells were quantified by flow cytometry for GFP expression. (A and C). Cultures were fractionated into whole-cell (WC), nuclear (N), cytoplasmic (C), and supernatant (S) fractions, and viral genomes were determined by qPCR. The number of HCMV genomes in each fraction is plotted per GFP⁺ cell. (B and D) Infectious units were determined from WC sonicants and culture S by titration on the matched producer cell type. Total infectious units are plotted per 1,000 GFP⁺ cells. (A to D) Error bars represent the standard deviations of the results from three separate experiments, and P values were generated using unpaired, two-tailed t tests (*, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001).

In fibroblasts, the total intracellular viral genome accumulation was comparable for all three strains (Fig. 5A). The vast majority was in the nuclear fraction, likely in the form of noninfectious, unpackaged genomes and nontegumented, nonenveloped nucleocapsids. The number of viral genomes in cytoplasmic and supernatant fractions was slightly lower for TB-infected fibroblasts than for either TR or ME. At least a portion of these cytoplasmic and supernatant genomes would be expected to represent the infectious progeny that could contribute to either cell-to-cell or cell-free spread, respectively. Despite the lower numbers of progeny genomes for TB in cytoplasmic and supernatant fractions, these corresponded to log folds greater cell-associated (i.e., sonically disrupted cells) and cell-free (i.e., supernatant) infectivity per cell compared to TR or ME (Fig. 5B). The greater infectivity of TB, despite lower virion numbers, was in agreement with our previous measures of lower specific infectivity ratios (genomes/PFU) for TB than for TR and ME (27) and was positively correlated with the relative efficiencies of cell-free spread for these strains. Surprisingly, cell-to-cell spread was inversely correlated with the levels of cell-associated infectivity, with cell-to-cell spread of ME being highly efficient despite producing the lowest detectable cell-associated infectivity.

In epithelial cells, the accumulations of whole-cell, nuclear, and cytoplasmic progeny genomes were similar among all three strains (Fig. 5C). In contrast, ME released nearly 10-fold more progeny genomes to culture supernatants than did TB, and TR was by far the least efficient at releasing progeny to supernatants. These results suggest major differences among strains in the later assembly and release stages of the replication cycle in epithelial cells. As was observed in fibroblasts, ME had the lowest cell-associated infectivity in epithelial cells, despite the more efficient cell-to-cell spread in this cell type (Fig. 4F and 5D). Infectivity in epithelial cell culture supernatants mirrored the quantities of genomes, indicating that the epithelium-derived virions of all three strains are of comparable specific infectivity. This was in stark contrast to the drastic differences in specific infectivity among these strains when produced in fibroblasts (Fig. 5A and B) (27).

Taken together, these results confirmed that the robust cell-free spread of TB and TR in fibroblasts is associated with the release of highly infectious progeny virus to the culture supernatant and that the poor cell-free spread of ME was due to the low infectivity of the cell-free progeny. Surprisingly, ME was much more efficient at cell-to-cell spread than was TB, despite having far less intracellular infectious progeny than TB or TR. In epithelial cell cultures, the cell-free characteristics for all of the strains were more similar to one another than they were on fibroblasts, and this was consistent with a comparably small contribution of cell-free spread for all three strains in this cell type. The spread by all three strains in epithelial cells was predominantly cell to cell, and ME was by far the most efficient, whereas TR was notably inefficient. The cell-to-cell spread of ME thus depends on efficient trafficking of progeny to adjacent cells, and this seems to overcome the low infectivity of the progeny virions.

The highly efficient cell-to-cell spread mechanism of HCMV ME is not determined by the high expression of UL128-131.

The cell-associated nature of ME has been linked to the high expression levels of UL128-131 and the corollary poor infectivity of cell-free ME virions (24, 27, 40, 47). To address this in our quantitative spread assay, we made use of the previously described tetracycline (Tet) repression system developed by Stanton et al. (40). Briefly, the BAC clone of ME used in these studies contained tetracycline operator (TetO) sequences in the promoter of UL131. We previously showed that replication of this recombinant ME in human fetal foreskin fibroblasts expressing the tetracycline repressor (TetR) protein (HFFF-tet) produced extracellular virus with dramatically reduced gH/gL/UL128-131, slightly higher gH/gL/gO, and, as a result, greatly improved cell-free infectivity (27). We refer to ME propagated under conditions of UL131 repression as “Merlin-T” (MT).

Direct comparisons of ME spread rates in normal human dermal fibroblast (nHDF) and HFFF-tet cells were complicated by the fact that in addition to expressing TetR, HFFF-tet cells were also telomerase immortalized, and this could have independent effects on the HCMV replication cycle. Moreover, since we also wished to address the role of UL128-131 expression on spread in epithelial cells, two new TetR-expressing cell lines were generated, nHDF-tet and adult retinal pigment epithelial-tet (ARPE-tet). A luciferase reporter assay demonstrated efficient repression of TetO-containing promoters in each of these cell lines (Fig. 6A). Consistent with our previous reports, ME virions produced from nHDF cells contained high levels of gH/gL/UL128-131 and small amounts of gH/gL/gO, whereas MT virions produced in HFFF-tet cells contained far less gH/gL/UL128-131 but more gH/gL/gO (Fig. 6B) (27). MT produced in the new nHDF-tet cell line showed a similar reduction of gH/gL/UL128-131 and increase in gH/gL/gO, and MT virions derived from either TetR-expressing fibroblast cell line were dramatically more infectious than ME virions derived from nHDF cells (Fig. 6B and C). ME virions produced in control ARPE cells were gH/gL/UL128-131 rich, but gH/gL UL128-131 low-MT virions were produced by ARPE-tet cells (Fig. 6B). However, unlike fibroblast-derived MT, epithelium-derived MT did not contain increased gH/gL/gO, and the infectivity was actually reduced (Fig. 6B and C). It was notable that both fibroblast-derived ME and epithelium-derived ME were rich in gH/gL/UL128-131 and low in gH/gL/gO, suggesting that the different producer cell types had little or no influence on the assembly of gH/gL complexes. In contrast, epithelium-derived TB virions contained less gH/gL/gO than did fibroblast-derived TB virions, but this was not accompanied by a notable increase in gH/gL/UL128-131 (Fig. 4B, right). Furthermore, the smaller amounts of gH/gL/gO in epithelium-derived TB correlated with a 10-fold lower infectivity compared to that with fibroblast-derived TB (Fig. 5).

FIG 6.

Expression of gH/gL/UL128-131 in fibroblasts and epithelial cells. (A) Tetracycline repressor protein (TetR) expressing fibroblasts (HFFF-tet and nHDF-tet) or epithelial cells (ARPE-tet) were transfected with TetO-firefly luciferase reporter, and the relative activity was measured. HFFF-tet cells were treated with doxycycline (+) or left untreated (−). (B) Nonreducing (top) or reducing (bottom) immunoblot analyses of MT, or ME and TB virions derived from TetR-expressing or control cells. Blots were probed with anti-gL antibodies (top) or anti-MCP antibodies (bottom). Mass standards (kDa) are indicated to the left. (C) Specific infectivity (genomes/IU) of MT and ME virions derived from culture supernatants of TetR-expressing or control cells, respectively. Titrations of fibroblast-derived virus were performed of nHDFs and of epithelium-derived virus on ARPEs. All experiments were performed in triplicate, and error bars represent the standard deviations. P values were generated by unpaired, two-tailed t test (*, P < 0.05; **, P < 0.01; ***, P < 0.001).

In nHDF cells, the spread of MT was reproducibly faster than that of ME, but in the presence of neutralizing anti-gH antibodies, the spreads were indistinguishable (Fig. 7A). The faster spread of MT and the greater relative inhibition due to the neutralizing antibodies are consistent with enhanced cell-free spread as a result of improved virion infectivity. However, the comparable spreads of ME and MT in the presence of neutralizing antibodies suggest that the efficiency of cell-to-cell spread is independent of the efficiency of cell-free spread. Similar experiments were performed with HFFF-tet cells with or without doxycycline to block the TetR repression of UL131 for the generation of ME and MT, respectively (Fig. 6A). Spread rates were generally lower in HFFF-tet cells than in nHDF cells, but overall, the results were comparable to those obtained in the nHDF system (Fig. 7B). The reasons for the differences in the overall spread between the fibroblast lines are unclear but may be related to the telomerase immortalization of HFFF-tet or the inherent heterogeneity of primary cell lines obtained from different donor sources. Regardless, the results from both fibroblast systems indicate that the efficiency of ME cell-to-cell spread is not dependent on high levels of gH/gL/UL128-131 and that efficient cell-to-cell spread is not simply a default of inefficient cell-free spread.

FIG 7.

Effect of gH/gL/UL128-131 repression on HCMV ME spread in fibroblasts and epithelial cells. (A to C) Confluent 9.5-cm³ monolayers of nHDF or nHDF-tet, for ME and MT, respectively (A), HFFF-tet (+) or (−) doxycycline, for ME and MT, respectively (B), or ARPE or ARPE-tet, for ME and MT, respectively (C), were infected with at an MOI of 0.001 with GFP-expressing HCMV ME/MT. Spread rates in the presence of either anti-gH MAb 14-4b (A and B) or anti-UL130/131 rabbit (C) sera were determined as described in Fig. 2. Plotted are the averages and standard deviations of the results from three independent experiments. P values were generated by a unpaired, two-tailed t test (*, P < 0.05; **, P < 0.01).

In epithelial cells, the repression of gH/gL/UL128-131 resulted in a small decrease in average spread rate (from 0.38 to 0.29), consistent with the importance of gH/gL/UL128-131 in this cell type (Fig. 7C). As shown earlier in Fig. 4, anti-UL130/131 rabbit serum reduced ME spread in epithelial cells by about 25%, but this was not likely due to blocking of cell-free spread since nonspecific preimmune rabbit serum had a similar effect. However, anti-UL130/131 was able to potently neutralize cell-free, epithelium-derived virus (Fig. 3D). Thus, if the repression of gH/gL/UL128-131 in MT shifted the mode of spread toward cell free, the anti-UL130/131 serum would be expected to have a greater inhibition on that spread. On the contrary, the effect of the anti-UL130/131 serum was comparable for both ME and MT. Together with the observation that epithelium-derived MT was less infectious than epithelium-derived ME (Fig. 6C), these results indicate that spread in both cases was predominantly cell to cell. Thus, while the high-levels of gH/gL/UL128-131 of ME clearly enhance the efficiency of the spread in epithelial cells, this does not appear to determine the mechanism.

RL13 glycoprotein tempers both cell-free and cell-to-cell spread.

The RL13 ORF encodes an envelope glycoprotein that has been described as a selective, or preferential, inhibitor of cell-free spread over cell-to-cell spread inasmuch as genotypes with inactivating mutations in RL13 arise during serial propagation of some HCMV strains in both fibroblasts and epithelial cells, and this has been correlated with the appearance of more cell-free spread characteristics (40, 50, 56). While the ME recombinant used in the present studies harbors such an inactivating RL13 mutation (a frameshift that predicts a truncated, non-membrane-anchored protein), the RL13 expression status of the TB and TR BAC clones is unclear. DNA sequencing confirmed that the RL13 ORF of both TB and TR was fully intact (data not shown), but a lack of quality antibodies precluded confirmation of RL13 glycoprotein expression. Thus, to directly and fairly compare the effects of RL13 on the spread of all three strains, and to avoid potential selection of RL13 mutants in TB and TR during spread experiments, fibroblast and epithelial cell lines that express RL13 were engineered. Immunoblot analysis demonstrated that these cell lines expressed both the mature and immature proteoforms, and flow cytometry confirmed similar expression levels (Fig. 8).

FIG 8.

RL13-expressing fibroblasts and epithelial cell lines. (A and B) nHDF and ARPE cells transduced with lentivectors encoding a C-terminal 6×His tagged RL13 were analyzed by immunoblot (mass standards [kDa] are indicated to the left) (A) or flow cytometry (B) using anti-6×His antibodies. Median fluorescence intensities for RL13-expressing nHDFs and ARPEs, and control cells are shown for comparison.

In fibroblasts, RL13 expression resulted in a comparable 40 to 50% reduction in spread rates for all three strains (Fig. 9A, compare black bars). This suggested that the inhibitory effect of RL13 can manifest regardless of the predominant mode of spread used by a given strain. To more directly test the hypothesis that RL13 was not specifically restricting cell-free spread, the effect of neutralizing antibodies was measured. As shown previously in Fig. 4, in the absence of RL13-expressing anti-gH antibodies dramatically reduced the spread of TB and TR but had little effect on the spread of ME (Fig. 9A, compare black bars and white bars in under RL13⁻ conditions). In the RL13-expressing cells, spread by all three strains was moderately inhibited by the mouse isotype control antibody, but TB and TR were more impaired by the specific anti-gH antibodies, whereas ME was not (Fig. 9A, compare black bars to the striped and white bars under RL13⁺ conditions). These results indicate that the spread of TB and TR in the RL13-expressing cells was still predominately cell free, whereas the spread of ME was still cell to cell. The reason for the moderate inhibition due to the isotype antibodies was unclear but may reflect the previously described Fc-binding activity attributed to RL13 (57).

FIG 9.

Effects of RL13 expression on spread of HCMV in fibroblasts and epithelial cells. (A and B) Spread rates of TB, TR, or ME were determined as in Fig. 2 in RL13-expressing fibroblasts (A) or epithelial cells (B) and respective control cells either in the absence of neutralizing antibodies (black bars) or in the presence of neutralizing antibodies (white bars) anti-gH MAb 14-4b (A) or anti-UL130/131 rabbit sera (B), or corresponding isotype control antibodies (striped bars). Plotted are average spread rates and standard deviations of the results from three independent experiments. P values were generated by ANOVA with Tukey’s multiple-comparison analysis with 95% CI (*, P < 0.05; **, P < 0.01; ***, P < 0.001).

In epithelial cells, RL13 expression had no effect on the spread of TR but resulted in a modest 20% reduction for both TB and ME (although this was only statistically reproducible for ME) (Fig. 9B, black bars). As before, the neutralizing anti-UL130/131 serum was no more inhibitory on spread than was the isotype control, except in the case of ME in the RL13-expressing cells (Fig. 9B, compare black bars to the striped and white bars under RL13⁺ conditions). The notion that RL13 selectively or preferentially restricts cell-free spread would have predicted that a greater fraction of the observed spread in RL13-expressing cells would be cell to cell and therefore less sensitive to neutralizing antibodies. On the contrary, these results showed that the sensitivity of spread to neutralizing antibodies was either unaffected or, in the case of ME, enhanced in RL13-expressing cells compared to the control cells.

RL13 exerts strain-dependent and cell type-dependent effects on the quantity and infectivity of progeny virus released to culture supernatants.

Two plausible mechanisms by which RL13 expression could negatively affect the cell-free mode of spread include (i) reducing the numbers of cell-free progeny released to culture supernatants, and (ii) reducing the infectivity of the cell-free progeny released. Both possibilities were tested using the RL13-expressing fibroblasts and epithelial cells. In fibroblasts, there were modest reductions in the numbers of progeny released to culture supernatants for TB and ME (Fig. 10A). There were also modest reductions in detectable infectivity in culture supernatants of the RL13-expressing cells (Fig. 10B) Taken together, the correlation between reduced numbers and reduced infectivity suggest that the specific infectivity of cell-free virions was not substantially affected by RL13 expression. For the efficient cell-free spreaders TB and TR, these reductions in cell-free progeny quantity and/or infectivity may help explain the reductions in spread rates described above. In contrast, the reduction in cell-to-cell spread for ME could not be attributed to reduced cell-free progeny quantity or infectivity.

FIG 10.

Quantity and infectivity of cell-free HCMV progeny virus from RL13-expressing fibroblasts and epithelial cells. (A to D) RL13-expressing fibroblasts (A and B) and epithelial cells (C and D) or control cells were infected with HCMV strain TB, TR, or ME at an MOI of 1. At 7 days postinfection, the number of infected cells was determined by flow cytometry, and culture supernatants were analyzed by qPCR for viral genomes (A and C) or by titration on a matched producer cell type to quantify infectious units (B and D). Plotted are averages and standard deviations of the results from three independent experiments. P values were generated by unpaired two-tailed t tests comparing RL13-expressing and control cells (*, P < 0.05; **, P < 0.01; ***, P < 0.001).

In epithelial cells, there were no effects on the numbers of supernatant progeny released for TB or TR, but there were actually more cell-free ME progeny produced from RL13-expressing cells (Fig. 10C). Despite the greater numbers of ME progeny in RL13-expressing epithelial cell supernatants, there were drastically lower levels of infectivity, indicating a substantial reduction in the per-virion infectivity (Fig. 10D). However, as was noted above for fibroblasts, since the spread of ME in epithelial cells was determined to be predominately cell to cell, these effects on cell-free virus may not fully explain how RL13 tempers spread in this cell type.

The drastic reduction in ME cell-free infectivity due to RL13 expression in epithelial cells suggested effects on the envelope glycoproteins. In particular, gH/gL/UL128-131 is the only envelope glycoprotein currently known to be specifically required for efficient infection of epithelial cells. Thus, ME virions produced by RL13-expressing or control fibroblasts and epithelial cells were analyzed by immunoblotting for the levels of gH/gL complexes. As shown before (Fig. 6B), ME virions produced by either fibroblast of epithelial cells had gH/gL predominately in the form gH/gL/UL128-131, and RL13-expression did not change this (Fig. 11). These results indicated that effect of RL13 on the infectivity of epithelium-derived ME virions is independent of gH/gL complexes.

FIG 11.

Effect of RL13 expression on levels of gH/gL complexes in ME virions. (A and B) Equal numbers (genomes/ml) of ME virions derived from either RL13-expressing or control fibroblasts or epithelial cells were separated by SDS-PAGE under nonreducing (A) or reducing (B) conditions and analyzed by immunoblot with antibodies directed against gL (A) or major capsid protein (B). Mass standards (kDa) are indicated to the left in both panels.

DISCUSSION

Distinctions between cell-free and cell-associated phenotypes of different HCMV strains and isolates have been noted, but the importance of different genetic loci involved in these spread modes remains to be understood. Sinzger et al. showed that during the initial rounds of cell-associated subculturing in fibroblasts, some clinical HCMV isolates formed small and tightly packed foci but tended toward larger and more diffuse foci during later rounds of passage (58). The change in focal morphology coincided with the appearance of infectious virus in culture supernatants, and the basic interpretation was that the adaptation of clinical HCMV isolates to the cell culture environment involved a shift from predominately cell-to-cell spread to cell-free spread. A more recent report by Galitska et al. noted considerable variation among clinical HCMV samples, with some displaying the characteristic small foci but others showing larger diffuse foci even in early rounds of subculturing (59). This variation fits well with the growing evidence of genotypic diversity of HCMV populations within natural populations of HCMV (3, 4, 6, 8, 9) and with the earlier suggestion by Sinzger et al. that the observed culture adaptation of HCMV clinical isolates might represent random sampling or purifying selection of preexisting genotypic variants rather than the acquisition of de novo mutations (58). In line with this view, Subramanian et al. used a flow cytometry method to determine the increase in HCMV-infected cells over time to better quantitate spread kinetics and found that some low-passage-number clinical isolates spread more rapidly than do others (60). They interpreted slow kinetics to represent a cell-associated-only phenotype and fast kinetics to represent a phenotype of both cell-associated and cell-free spread. Indeed, the faster-spreading isolates were more sensitive to the presence of neutralizing antibodies. Together, these and other studies suggest considerable variation in the modes of spread exhibited by genetically distinct HCMV strains, but the exact genetic variations that determine these phenotypes remain unclear.

Here, we applied a flow cytometry-based approach to compare the spread characteristics of three commonly studied and genetically distinct BAC clones of HCMV (TB, TR, and ME) in both fibroblast and epithelial cell cultures. It has been widely appreciated that TB and TR exhibit characteristics of cell-free spread, whereas ME is essentially restricted to cell-to-cell spread. Previous studies suggested that the distinction between these strains in the ability to spread cell free in fibroblast cultures is due largely, if not entirely, to the fact that cell-free TB and TR virions are log folds more infectious than are ME virions and that cell-free spread of ME in fibroblast cultures can be enhanced by transcriptional repression of epithelial/endothelial cell tropism factor gH/gL/UL128-131, which in turn leads to improved infectivity of cell-free ME virions (27, 38, 40). However, it has remained unclear whether the apparent cell-associated nature of ME is due simply to the poor cell-free infectivity or reflects a specific ability to facilitate more efficient cell-to-cell spread than viruses like TB and TR that can spread efficiently cell free. An important corollary consideration for our studies was that the UL128-131 and RL13 loci have been implicated as important for determining the cell-free or cell-associated nature of HCMV strains (24, 27, 40, 47, 50, 56). The results presented here demonstrate that the cell-associated characteristic of ME is not simply due to the poor cell-free infectivity. Rather, ME is capable of a specifically efficient cell-to-cell mechanism of spread, TB is specifically inefficient at cell-to-cell spread, and TR may have intermediate cell-to-cell spread capacity. Moreover, we demonstrate that cell-free and cell-associated spread characteristics are not determined by either the expression levels of UL128-131 or by RL13 expression, although these loci do influence the efficiency in strain-dependent and cell type-dependent manners. Thus, we suggest the involvement of other variable loci in distinct mechanisms of cell-free and cell-to-cell spread, and that epistatic relationships among these loci govern the efficiency of these processes.

One candidate for a variable factor that might influence cell-free and cell-to-cell spread in an epistatic manner is gO, which is one of the more diverse proteins encoded by HCMV. There are 8 alleles of the UL74 (gO) ORF that vary between 10 and 30% of predicted amino acids (51, 61–63). Mutational inactivation of UL74 in TR and TB dramatically reduced cell-free infectivity but allowed cell-to-cell spread, albeit at a reduced level (26, 64). In contrast, the spread of a gO-null ME mutant was indistinguishable from that of the parent (24). Together, these observations suggested that gH/gL/gO is critical for cell-free spread, but cell-to-cell spread can be mediated by gH/gL/UL128-131 alone. However, it has more recently been reported that viruses lacking gH/gL/UL128-131 can spread cell to cell, but this is dependent on the gH/gL/gO receptor PDGFRa (48). Furthermore, a recent study by our group demonstrated that switching the UL74 allele in both TR and ME backgrounds could influence the efficiency of cell-to-cell spread, demonstrating that there were epistatic phenomena involved inasmuch as the influence of some UL74 alleles was observed in only one of the two genetic backgrounds (65).

Our finding that the level of gH/gL/UL128-131 expression did not affect the spread rate of ME in fibroblasts conflicts with a report by Stanton et al., who showed that UL128-null ME plaques were significantly larger than UL128-intact ME plaques (40). The apparent contradiction may be explained by the fact that our flow cytometry assays measured the spread rates averaged over the first 12 days, compared to the measurements of 21-day plaque sizes by Stanton et al. Log-linear regression analyses of the data collected by flow cytometry indicated that ME spread at a higher rate between days 3 and 6 and then began to slow after day 9, unlike the constant rate of cell-free spread demonstrated by TB and TR (Fig. 2). The slowdown displayed by ME may simply reflect a limiting number of potential uninfected target cells available via cell-to-cell spread during focal expansion. Alternatively, it could be that the cells at the periphery of an expanding ME focus rapidly become less permissive, for example, via activation of the antiviral state. In any case, the increase in 21-day plaque size of a UL128-null ME is consistent with enhanced cell-free infectivity, but our analysis of spread rates over 12 days indicates this does not come at the expense of a distinct and comparably efficient mode of cell-to-cell spread. This interpretation is supported by the observation that the 12-day spread rate of ME was highly resistant to neutralizing antibodies regardless of the gH/gL/UL128-131 expression level (Fig. 7); this is consistent with the findings of Laib Sampaio et al., who showed much greater cell-free dispersion of UL128-null ME foci but more similar antibody resistant, cell-associated spread with or without UL128 (24).

In epithelial cells, the repression of gH/gL/UL128-131 did reduce the 12-day spread rate for ME (Fig. 7). This was consistent with reduced plaque sizes of UL128-null ME mutants documented by Murrell et al. and likely reflects the important role of gH/gL/UL128-UL131 in epithelial cells (38). However, as observed in fibroblasts, spread in epithelial cells was still highly resistant to neutralizing antibodies when gH/gL/UL128-131 was repressed. This was different than findings by Murrell et al., who concluded that spread of ME in epithelial cells was more sensitive to antibodies when gH/gL/UL128-131 was reduced (47). Again, the discrepancy may indicate that analyses of 21-day plaque size likely accentuate the influence of cell-free infectivity and mask the contribution of cell-to-cell spread. Thus, by measuring average spread rates over 12 days, we find that gH/gL/UL128-131 levels are not the sole determinants of the cell-free versus cell-associated spread phenotype for ME. Rather, the increase in cell-free infectivity due to a reduction in gH/gL/UL128-131 and the concomitant increase in gH/gL/gO are independent of the highly efficient cell-to-cell mode of spread.

The RL13 glycoprotein was another candidate for a factor influencing the cell-free and cell-associated phenotypes. Previous reports have suggested that the expression of RL13 limits HCMV replication in fibroblast, epithelial, and endothelial cell cultures through preferential effects on cell-free spread (24, 40, 50, 56). On the contrary, we find that RL13 tempers HCMV spread by either cell-free or cell-to-cell mechanisms and that this temperance was more pronounced in fibroblasts than in epithelial cells (Fig. 9). Interestingly, RL13 expression dramatically reduced the infectivity of cell-free virions released by ME from epithelial cells, and this was consistent with the previous report showing low cell-free titers for RL13-intact ME in epithelial cell cultures compared to those of RL13 mutant ME (40). The physiological basis of this infectivity loss remains unclear, as our analysis showed that virions released from RL13-expressing epithelial cells had comparable amounts of gH/gL complexes, but it is conceivable that RL13 could affect the processing and incorporation of other important envelope glycoproteins.

Our results may help explain inconsistencies regarding the selection of UL128-131 and RL13 mutations during laboratory culturing of HCMV isolates, a phenomenon that has been universally ascribed to HCMV, despite clear strain variability. While ME is highly sensitive to the selection of de novo UL128-131 mutations in fibroblast cultures, TB and TR are remarkably stable in this respect (40, 50, 58). This disparity might simply reflect that many subculturing methods that include supernatant transfers, cell sonication, and high split-ratio transfer of intact infected cells would be expected to heavily favor cell-free spread. Thus, given the highly efficient cell-to-cell spread of ME reported in this study, propagation by low split-ratio passage of intact, infected cells might reduce the selective bottleneck on the loss of UL128-131 function to favor cell-free spread. Likewise, RL13-inactivating mutations have also been reported for many strains and clinical isolates in studies involving fibroblast, epithelial, and endothelial cells (10, 56). The BAC clones of TB and TR have intact RL13 ORFs, but since the expression of RL13 from transduced cells reduced replication by these strains, it may be that both strains carry mutations or polymorphisms in promoter or other regulatory sequences that impact RL13 expression. Moreover, our results suggest that the selection against RL13 may be more extensive in fibroblasts but less pronounced in epithelial cells.

Notably, a recent report showed that the presence of neutralizing antibodies during subculturing of clinical urine-derived HCMV in fibroblasts stabilized the UL128-131 and RL13 loci in the resultant virus populations (66). If these clinical isolates were similar to ME in their proclivity for cell-to-cell spread, it would seem logical that neutralizing antibodies would reduce the selective advantage of cell-free infectivity conferred by a UL128-131 mutation. Indeed, it is interesting that ME was also isolated from a clinical urine sample (10). However, the preservation of RL13 seems less intuitive since RL13 impacts both cell-free and cell-to-cell modes of spread. Given the varied effects we observed of RL13 on different strains, these observations again seem to point to epistasis phenomena, where the relative effect of any given perturbation at one variable locus is influenced by variable physiology related to other polymorphic loci.

Our data also shed new light on the notion that the cell type might influence phenotypic properties in a nongenetic manner through changes to the protein composition of the progeny virus. Among the herpesviruses, these so called “producer cell effects” have been particularly well characterized for Epstein-Barr virus (EBV), where the mechanism alternates the tropism of progeny virus between epithelial and B cells (reviewed in reference 67). Scrivano et al. suggested a similar phenomenon for HCMV based in part on data showing that progeny TB produced from endothelial cells had less UL128 protein in the envelope than did fibroblast-derived progeny (35). In contrast, Wu et al. showed that fibroblast-derived TB virions produced from a parental TB stock that was first passed on epithelial cells had less gH/gL/gO and more gH/gL/UL128-131 than TB virions that had only been passed on fibroblasts. (48). However, since both populations of TB virions analyzed were ultimately fibroblast derived, it seems likely that the apparent difference in gH/gL complexes was related to genetic changes that occurred during the passages in epithelial cells rather than a producer cell effect as defined above. In our analyses, ME virions produced from either fibroblast or epithelial cells were rich in gH/gL/UL128-131 and low in gH/gL/gO, suggesting little or no producer-cell effect on gH/gL complexes. However, TB virions produced from epithelial cells had less gH/gL/gO than did fibroblast-derived TB and still had low gH/gL/UL128-131 (Fig. 6). The discrepancies between our observations and those of Scrivano et al. might suggest differences between endothelial and epithelial cells. Moreover, their analysis normalized virion immunoblots to gB, whereas we used major capsid protein (MCP). It is possible that the incorporation of gB into progeny virus is also affected by glycoprotein processing/secretory pathway differences between cell types, whereas plasticity in the number of subunits per capsid subunits seems less likely. Thus, the incorporation of gH/gL/gO and gH/gL/UL128-131 into HCMV virions is clearly a complex process that can be impacted by both viral genetics and the physiology differences among cell types.

In conclusion, we have characterized very different cell-free and cell-associated spread phenotypes among three commonly studied BAC clones of HCMV. Our data suggest that these phenotypic variances are manifest through the combined influenced of diversity at multiple genetic loci other than UL128-131 and RL13. How this phenotypic variation is represented in vivo remains unclear. There is little clear evidence to support the notion that HCMV is predominately cell associated in vivo. The endemic nature and the pleomorphic disseminated disease presentations suggest that HCMV is able to thrive in many different bodily environments as it spreads within and among individual human hosts. Copious amounts of cell-free virus are shed in urine, saliva, and breast milk, and this is likely a major route of transmission between individuals, whereas HCMV in the blood is likely highly associated with leukocytes (68–70). Thus, it seems likely that both cell-free and cell-associated modes of spread are important to the natural history of HCMV.

MATERIALS AND METHODS

Cell lines.

Primary neonatal human dermal fibroblasts (nHDF; Thermo Fisher Scientific), RL13-nHDFs (nHDFs transduced with lentiviral vectors encoding RL13 of HCMV strain Merlin, selected with puromycin resistance), nHDF-tet (nHDFs transduced with retroviral vectors encoding the tetracycline repressor protein, selected with puromycin resistance), MRC-5 fibroblasts (CCL-171; American Type Culture Collection), and HFFF-tet cells (40) (provided by Richard Stanton, Cardiff University, United Kingdom) were grown in Dulbecco’s modified Eagle’s medium (DMEM; Sigma) supplemented with 6% heat-inactivated fetal bovine serum (FBS; Rocky Mountain Biologicals, Missoula, MT, USA) and 6% Fetalgro (Rocky Mountain Biologicals). Retinal pigment epithelial cells (ARPE19; American Type Culture Collection, Manassas, VA, USA), ARPE-tet cells (ARPE19 cells transduced with retroviral vectors encoding the tetracycline repressor protein, and selected by puromycin resistance), and RL13-ARPE cells (transduced with lentiviral vectors encoding RL13 of HCMV strain Merlin, selected with puromycin resistance) were grown in a 1:1 mixture of DMEM and Ham’s F-12 medium (DMEM–F-12; Gibco) and supplemented with 10% FBS.

Retroviral vectors.

The Tet repressor protein bearing a nuclear localization signal was extracted by PCR from the integrated sequence in the HFFF-tet chromosomal DNA and cloned into the same pMXs retrovirus vector used to construct HFFF-tet (40). The pMXs retrovirus vector plasmid was a gift from Toshio Kitamura at the Institute of Medical Science, University of Tokyo (71). The tet-containing vector plasmid was transformed in 293T cells together with the pUMVC and pMD2.G helper plasmids. The pUMVC helper plasmid was a gift from Bob Weinberg (Addgene plasmid no. 8449) (72). Two days after transformation, the retroviral particles in the supernatant were purified from cell debris through syringe filtration and centrifugation. After titration, the particles were used to transduce low-passage-number nHDF or ARPE-19 cells. After a week of puromycin selection, cell aliquots were tested for TetR expression after transfection of a firefly luciferase tetR reporter system, and aliquots were stored in liquid nitrogen until further usage. The codon-optimized RL13 with an intact ORF from Merlin HCMV strain (NCBI RefSeq accession no. YP_081461) was constructed by Gibson Assembly and used to replace the enhanced GFP (eGFP) ORF in the pLJM1-EGFP lentiviral transfer vector plasmid. The pLJM1-EGFP plasmid was a gift from David Sabatini (Addgene plasmid no. 19319) (73). The RL13-containing vector plasmid was transformed in 293T cells together with three lentiviral helper plasmids. The pMDLg/pRRE, pRSV-Rev, and pMD2.G helper plasmids were a gift from Didier Trono (Addgene plasmids 12251, 12253, and 12259, respectively) (74). Two days after transformation, the lentiviral particles in the supernatant were purified from cell debris through syringe filtration and centrifugation. After titration, the particles were used to transduce either low-passage-number nHDF or ARPE-19 cells. After a week of puromycin selection, cells were tested for RL13 expression, and aliquots were stored in liquid nitrogen until further usage.

HCMV.

All human cytomegalovirus (HCMV) strains were derived from bacterial artificial chromosome (BAC) clones. BAC clone TB40/e (BAC4) was provided by Christian Sinzger (University of Ulm, Germany) (46). BAC clone TR was provided by Jay Nelson (Oregon Health and Sciences University, Portland, OR, USA) (75). BAC clone Merlin (pAL1393), which contains tetracycline operator sequences within the transcriptional promoter of UL130 and UL131, was provided by Richard Stanton (Cardiff University, Cardiff, United Kingdom) (40). All BAC clones were modified to express GFP with en passant recombineering (76) by replacing US11 with the eGFP gene. The constitutive expression of eGFP allows the monitoring of HCMV infection early and is strain independent. Infectious HCMV was recovered by electroporation of BAC-DNA into HFFFs, as previously described (26). For infectious unit (IU) determination, viruses were serially diluted, and infectivity was determined on fibroblasts or epithelial cells using flow cytometry at 48 h postinfection. Multiplicities of infection (MOIs) were determined as IU/cell, matched for the cell types used in each experiment (fibroblast IU for fibroblast experiments and APRE IU for ARPE experiments).

Antibodies.

Monoclonal antibodies specific to HCMV major capsid protein (MCP), gH (14-4b), and gB (27-156) were provided by Bill Britt (University of Alabama, Birmingham, AL) (52, 77, 78). 14-4b and MCP were purified by fast-performance liquid chromatography (FPLC) prior to use. Rabbit polyclonal sera against HCMV gL, UL130, and UL131 were provided by David Johnson (Oregon Health and Sciences University, Portland, OR) (17).

Viral spread assays.

Approximately 1 × 10⁵ (3 × 10⁵) nHDFs (ARPEs) were seeded into 6-well culture plates and allowed to grow to confluence. Confluent monolayers of nHDFs or ARPE cells were inoculated with 100 to 1,000 IUs of strain TB, TR, or ME at for 4 h at 37°C. Cells were then washed extensively with phosphate-buffered saline (PBS) and cultured in the appropriate growth medium supplemented with 2% FBS. Viral spread in the presence or absence of neutralizing antibodies, isotype controls, or ganciclovir was monitored over 12 days by flow cytometry. For fibroblasts, 50 μg/ml anti-gH 14-4b was used; for epithelial cells, a 1:1,000 dilution of anti-UL130/131 rabbit serum was used. Ganciclovir (50 μM) was used on both cell types. All experiments were performed in triplicate, and a minimum of 3 experiments were conducted for each condition. Spread rates were determined by plotting LN GFP⁺ cells over time and fitting the data to the log-linear rate expression LN(I)_t = m(t) + LN(I)₀ where (I) is the number of GFP⁺ cells, (t) is the time in days, and m is the spread rate.

Flow cytometry.

Recombinant GFP-expressing HCMV-infected cells were washed twice with PBS and lifted with trypsin. Trypsin was quenched with DMEM containing 10% FBS, and cells were spun at 500 × g for 5 min at RT. Cells were fixed in PBS containing 2% paraformaldehyde for 10 min at RT and then washed and resuspended in PBS. Samples were analyzed using an AttuneNxT flow cytometer. Cells were identified using forward scatter area (FSC-A) and side scatter area (SSC-A), and single cells were gated using forward scatter width (FSC-W) and forward scatter height (FSC-H). A BL-1 laser (488 nm) was used to identify GFP⁺ cells, and only cells with median GFP intensities 10-fold above background were considered positive. RL13 expression was measured using an intracellular staining kit (Thermo) and an anti-6×His antibody conjugated to Alexa Fluor-647 (Thermo) using the RL-2 laser (647 nm).

qPCR.

The real-time quantitative PCR (qPCR) assay used to quantify viral or cellular DNA molecules was performed as previously described (27). Briefly, HCMV DNA inside cell-free particles was purified using the PureLink viral RNA/DNA minikit (Thermo Scientific). A region within UL83 conserved among ME, TR, and TB was chosen as the HCMV-specific amplicon, and viral genomes were quantified by SYBR green qPCR, as previously described. Standard curves were performed using serial dilutions of a single PCR DNA band containing the sequences of the viral UL83 and cellular beta2-microglobulin (see below) amplicons and the corresponding set of primers. Finally, the concentration of HCMV DNA genomes in the supernatant was extrapolated from the UL83 standard curve and expressed as genome molecules per ml of medium. Total intracellular or cytoplasmic HCMV DNA was quantified with the UL83-based qPCR after extracting viral and cellular DNA from infected cells using the PureLink genomic DNA minikit (Thermo Scientific). We also measured the chromosomal DNA molecules present in these samples with an amplicon located in the human beta-2 microglobulin gene (79). These chromosome numbers were used to correct for the number of cells when measuring the whole-cell, nuclear, and cytoplasmic genomes from different samples. To obtain these fractions, infected cells were subjected to a subcellular fractionation kit for cultured cells (Thermo Scientific) to purify the cytoplasmic HCMV DNA from the excess of nuclear HCMV DNA. Last, the genomes present in the cytoplasmic fraction were purified using the Genomic DNA minikit, and their number was determined using the UL83-based qPCR assay.

Particle production.

The production of intracellular and extracellular HCMV genomes was determined over a time course of 4 days. Briefly, nHDFs and ARPEs were infected with HCMV strain TB, TR, or ME at an MOI of 1, and the cells were extensively washed at 3 days postinfection (dpi). At 3, 5, and 7 dpi, culture supernatants were collected and spun at 500 × g for 10 min at room temperature (RT), while cells were lifted with trypsin and then quenched with culture medium containing 10% FBS. Cells were spun at 500 × g and resuspended into aliquots. Cells were either resuspended in PBS for whole-cell analysis or fractionated for nuclear and cytoplasmic analysis using the subcellular fractionation kit for cultured cells (Thermo). Cells were also sonicated to release cell-associated virus for infectivity analysis or processed for flow cytometry to determine the number of GFP⁺ cells. The numbers of HCMV genomes in the whole-cell, nuclear, cytoplasmic, and supernatant fractions were measured by qPCR, normalized to the load, and divided by the number of GFP⁺ cells. The total infectivity of supernatant and sonicated fractions was determined by flow cytometry 48 h after inoculation of the producer cell type. The accumulation at day 7 from nHDFs and ARPEs is shown in Fig. 5.

Immunoblot analyses.

Cell-free virions were solubilized in a buffer containing 20 mM Tris-buffered saline (TBS) (pH 6.8) and 2% SDS. Protein samples were separated by SDS-polyacrylamide gel electrophoreses (SDS-PAGE) and electrophoretically transferred to polyvinylidene difluoride membranes in a buffer containing 10 mM NaHCO₃ and 3 mM Na₂CO₃ (pH 9.9) and 10% methanol. Transferred proteins were first probed with MAbs or rabbit serum, anti-mouse or anti-rabbit secondary antibodies conjugated to horseradish peroxidase (Sigma), and Pierce ECL Western blotting substrate (Sigma). Chemiluminescence was detected using a Bio-Rad ChemiDoc MP imaging system.

Statistical analysis.

All experiments were performed a minimum of three times. All statistical analyses were performed using the GraphPad Prism 6 software. Experiments comparing multiple strains were analyzed by analysis of variance (ANOVA) with Tukey’s multiple-comparison test using a 95% confidence interval. Standard two-tailed t tests were used for direct comparisons. Error bars represent standard deviations between experiments, and P values are represented as follows: *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.