ABSTRACT
Cytomegalovirus (CMV) entry into fibroblasts differs from entry into epithelial cells. CMV also spreads cell to cell and can induce syncytia. To gain insights into these processes, 27 antibodies targeting epitopes in CMV virion glycoprotein complexes, including glycoprotein B (gB), gH/gL, and the pentamer, were evaluated for their effects on viral entry and spread. No antibodies inhibited CMV spread in fibroblasts, including those with potent neutralizing activity against fibroblast entry, while all antibodies that neutralized epithelial cell entry also inhibited spread in epithelial cells and a correlation existed between the potencies of these two activities. This suggests that exposure of virions to the cell culture medium is obligatory during spread in epithelial cells but not in fibroblasts. In fibroblasts, the formation of syncytiumlike structures was impaired not only by antibodies to gB or gH/gL but also by antibodies to the pentamer, suggesting a potential role for the pentamer in promoting fibroblast fusion. Four antibodies reacted with linear epitopes near the N terminus of gH, exhibited strain specificity, and neutralized both epithelial cell and fibroblast entry. Five other antibodies recognized conformational epitopes in gH/gL and neutralized both fibroblast and epithelial cell entry. That these antibodies were strain specific for neutralizing fibroblast but not epithelial cell entry suggests that polymorphisms external to certain gH/gL epitopes may influence antibody neutralization during fibroblast but not epithelial cell entry. These findings may have implications for elucidating the mechanisms of CMV entry, spread, and antibody evasion and may assist in determining which antibodies may be most efficacious following active immunization or passive administration.
IMPORTANCE Cytomegalovirus (CMV) is a significant cause of birth defects among newborns infected in utero and morbidity and mortality in transplant and AIDS patients. Monoclonal antibodies and vaccines targeting humoral responses are under development for prophylactic or therapeutic use. The findings reported here (i) confirm that cell-to-cell spread of CMV is sensitive to antibody inhibition in epithelial cells but not fibroblasts, (ii) demonstrate that antibodies can restrict the formation in vitro of syncytiumlike structures that resemble syncytial cytomegalic cells that are associated with CMV disease in vivo, and (iii) reveal that neutralization of CMV by antibodies to certain epitopes in gH or gH/gL is both strain and cell type dependent and can be governed by polymorphisms in sequences external to the epitopes. These findings serve to elucidate the mechanisms of CMV entry, spread, and antibody evasion and may have important implications for the development of CMV vaccines and immunotherapeutics.
INTRODUCTION
Cytomegalovirus (CMV) infections cause birth defects among newborns infected in utero and morbidity and mortality in transplant and AIDS patients. Naturally acquired immunity to CMV is protective and beneficial (1–4). Cellular immunity is known to be important for controlling CMV disease in transplant and AIDS patients (5–7). Data from clinical studies also support a role for humoral immunity in protection against CMV infection and disease. For example, trials of the glycoprotein B (gB)/MF59 vaccine, believed to act primarily through induction of neutralizing antibodies (Abs), suggest a role for antibodies both in preventing primary CMV infections (8) and in reducing CMV disease in solid organ transplant patients (9). In addition, the use of CMV hyperimmunoglobulin (HIG; IgG isolated from the blood of CMV-seropositive donors) in certain transplant settings can ameliorate posttransplant CMV disease (recently reviewed in reference 10). Importantly, mounting evidence suggests that CMV HIG can be beneficial for the prevention and treatment of congenital CMV infections (4, 11–18). In a recent proof-of-concept study, a neutralizing monoclonal antibody (MAb) had efficacy in preventing fetal infection and loss in the guinea pig congenital infection model (19). Antibodies with robust neutralizing potency are in development for potential use as therapeutics or for passive prophylactic immunization (20, 21; reviewed in reference 22). Indeed, promising efficacy in high-risk renal transplant patients was recently reported for a bivalent CMV monoclonal therapeutic (23).
The mediators, mechanisms, and neutralizing targets of CMV entry are complex and cell type specific. Fibroblast viral entry requires gB and a trimeric complex of glycoproteins gH, gL, and gO, whereas viral entry into endothelial, epithelial, and dendritic cells requires an additional pentameric complex (pentamer) of gH, gL, UL128, UL130, and UL131 (24–34). Consequently, antibodies directed against gB, gH/gL, or gO impair viral entry into fibroblasts and endothelial and epithelial cells, whereas antibodies that specifically target the pentamer selectively block viral entry into epithelial and endothelial cells (28, 33, 35–37). Following natural infection, the latter activity is dominant, as the serum neutralizing titers measured with epithelial cells are significantly higher than those measured using fibroblasts (38–41).
That CMV persists in spite of robust humoral responses suggests that in vivo, CMV may evade neutralizing antibodies by spreading cell to cell. This may be especially relevant in transplant-associated CMV disease, where pathology is caused by viral spread within tissues of affected organs. Consistent with this, antibodies can slow but not completely prevent CMV spread in cultured fibroblasts (34, 42–46). However, just as viral entry is cell type specific, the mechanisms of cell-to-cell spread may also differ between cell types. Several studies suggest that cell-to-cell spread of CMV in endothelial cell cultures is sensitive to antibody inhibition (34, 39, 46, 47), while we and others have shown that antibodies can limit the spread of CMV in epithelial cell cultures (48–50).
In the current study, HIG, 25 MAbs, and two antipeptide antisera targeting gB, gH, gH/gL, the pentamer, or pentamer subunits were evaluated for neutralizing and spread inhibition activities in fibroblast and epithelial cells. All antibodies failed to inhibit the spread of CMV within fibroblast cultures, regardless of their neutralizing activities. In contrast, spread within epithelial cell cultures was inhibited by antibodies specific to gB, gH, gH/gL, the pentamer, or pentamer subunits, and the neutralizing potencies correlated with the spread inhibition potencies. The formation of syncytiumlike structures in fibroblasts was impaired not only by antibodies to gB or gH/gL but also by antibodies to the pentamer, suggesting a potential role for the pentamer in promoting fusion of fibroblasts during infection. Finally, antibodies targeting epitopes in gH or gH/gL exhibited strain- and cell-type-dependent neutralizing activities that implicated a role for polymorphisms both within and outside gH/gL in governing sensitivity or resistance to neutralization. These findings may have implications for elucidating the mechanisms of CMV entry, spread, and antibody evasion and in determining which antibodies may be most efficacious following active immunization or passive administration.
RESULTS
Antibodies and their targets. Table 1 lists the antibodies used in these studies and their properties. Twenty-two MAbs were isolated from a rabbit that had been immunized with an experimental whole-virus vaccine comprised of a variant of CMV strain AD169 that expresses a functional pentamer (28). Eleven of these MAbs were previously shown to neutralize CMV, and another 11 were shown to bind CMV virions without neutralizing (28). The target specificities for five of the latter group are not known; the remaining rabbit MAbs have been shown to recognize gB, gH, gH/gL, or pentamer-specific epitopes (28, 51). Two polyclonal rabbit antisera were raised against 17- to 20-amino-acid synthetic peptides representing sequences within UL130 or UL131 (29) and were shown to selectively neutralize epithelial cell entry (36). The human MAbs 2-25 and TRL310 recognize the pentamer and selectively neutralize epithelial cell entry (35, 51), while human MAb TRL345 recognizes the AD-2 epitope in gB and neutralizes both fibroblast and epithelial cell entry (20, 52). HIG is a polyclonal mixture of human antibodies purified from CMV-seropositive donors and neutralizes both fibroblast and epithelial cell entry (38).
Antibodies used in this study
CMV spread in epithelial cells involves gB, gH, and the pentamer, and antibody inhibition of spread closely correlates with neutralizing activity.In a recent study, we showed that HIG or sera from naturally infected subjects can restrict CMV spread in epithelial cell cultures (48). To identify viral factors involved in epithelial cell spread and to determine whether any of our antibodies were discordant with respect to inhibition of epithelial cell entry versus spread, each was assayed for epithelial cell entry-neutralizing activity and spread inhibition. As shown by the results in Fig. 1, antibodies to epitopes within gB, gH, gH/gL, UL130, UL131, or the pentamer blocked CMV spread in epithelial cells. The most potent were MAbs to the pentamer, while antibodies to gB, gH, or gH/gL were less potent (Fig. 1).
Antibodies to gB, gH/gL, gH, the pentamer, or pentamer subunits inhibit CMV spread in epithelial cells. (A) ARPE-19 epithelial cell monolayers in 96-well plates were infected with 100 PFU/well of CMV virus ABV (strain Uxc). After incubation at 37°C for 24 h, the cultures were washed and medium (∅) or medium containing serial dilutions of the indicated antibodies was added. Representative micrographs were taken on day 16 postinfection. Dilution factors on the left indicate dilutions from initial stocks of HIG (5 mg/ml), rabbit anti-UL130 or anti-UL131 antiserum (diluted 1:5), or MAbs (100 μg/ml). (B) Five hundred PFU of CMV virus ABV (strain Uxc) was incubated with medium (∅) or medium containing serial dilutions of the indicated antibodies for 1 h at 37°C and then added to ARPE-19 monolayers in 96-well plates. Representative micrographs were taken on day 6 postinfection. Dilution factors on the left indicate dilutions from initial stocks of HIG (1.25 mg/ml), rabbit anti-UL130 or anti-UL131 antiserum (diluted 1:40), or MAbs (25 μg/ml, except for TRL310 and 57.4 [2 μg/ml] and 2-25 [1 μg/ml]).
Of the 27 antibodies tested, none were discordant between neutralizing and spread inhibition: all 11 rabbit MAbs that bound to CMV virions but did not neutralize (Table 1) also failed to inhibit viral spread (data not shown), and conversely, all 16 antibodies with epithelial cell entry-neutralizing activity inhibited epithelial cell viral spread (Fig. 1). To determine whether there was a relationship between potency of neutralizing activity and potency of spread inhibition, green fluorescent protein (GFP) intensity measurements were used to determine the 50% inhibitory concentrations (IC50s) (Table 2), and linear regression analysis of the log-transformed spread inhibition IC50 versus the log-transformed neutralizing IC50 for 14 MAbs revealed a strong correlation between how potently an antibody inhibits spread and how potently it inhibits entry (Fig. 2) (r = 0.94; P < 0.0001). These studies suggested that exposure of virions to the cell culture medium may be obligatory during spread in epithelial cells, and therefore, there may be little difference mechanistically between inhibition of entry and spread in epithelial cell cultures.
Neutralizing and spread inhibition activities of monoclonal antibodies
Antibody inhibition of CMV spread in epithelial cells correlates with neutralization of epithelial cell entry. GFP-based assays using CMV virus ABV (strain Uxc) were used to determine IC50s (μg/ml) for epithelial cell spread inhibition and epithelial cell entry neutralization for 14 MAbs (see Fig. 1 and Table 2). For each antibody, log-transformed IC50s for spread inhibition were plotted versus IC50s for neutralization, and linear regression was used to calculate the indicated best-fit line, correlation coefficient (r), and P value.
None of the antibodies inhibit CMV spread in fibroblasts.Similar assays were conducted using fibroblasts; however, neither HIG nor any of the 27 antibodies exhibited any significant effects on fibroblast spread (Fig. 3 and data not shown), even at the highest concentration tested (50 μg/ml). For the six MAbs having fibroblast entry-neutralizing activity, this concentration was 370- to 2,100-fold higher than their IC50s for neutralizing fibroblast entry. Thus, among 27 antibodies targeting CMV virion proteins, 15 of which neutralize and 11 of which do not, none were found to affect fibroblast spread. These results are consistent with those of previous studies (34, 42–46) and indicate that, in contrast to spread in epithelial cells, CMV spread in fibroblasts is highly resistant to antibody inhibition.
CMV spread in fibroblasts is insensitive to antibody inhibition. MRC-5 fibroblast monolayers in 96-well plates were infected with 100 PFU/well of CMV virus ABV (strain Uxc). After incubation at 37°C for 24 h, the cultures were washed and medium (∅) or medium containing dilutions of the indicated antibodies was added. Representative micrographs were taken on day 8 postinfection. Micrographs of neutralizing assays performed as described in the legend to Fig. 1B but using MRC-5 fibroblasts are included for comparison. Dilution factors on the left indicate dilutions from initial stocks of HIG (5 mg/ml), rabbit anti-UL130 or anti-UL131 antiserum (diluted 1:10), or MAbs prediluted to 100 μg/ml.
Antibodies to gB, gH/gL, and the pentamer limit the formation of syncytiumlike structures in infected fibroblasts.The spread assays whose results are shown in Fig. 3 began with a small number (∼100) of isolated GFP-positive (GFP+; infected) cells that spread to form GFP+ foci. By day 8 postinfection, untreated cultures contained large, rounded, GFP-bright structures that had the appearance of syncytia, while cultures treated with pentamer-specific MAbs lacked these large, rounded, syncytiumlike structures (Fig. 3). To evaluate this phenomenon more closely, spread assays were conducted with two modifications designed to enhance the formation of syncytiumlike structures: first, virus BADrUL131-Y4 (BADr) was used instead of virus ABV, as empirical observations suggested that BADr has a stronger propensity than ABV to form these syncytiumlike structures (data not shown), and second, the cultures were monitored until 13 days postinfection (dpi) to allow more extensive viral spread. In untreated cultures, small syncytiumlike structures began to form by day 7 postinfection, grew larger but remained distinct on day 10, and appeared to coalesce into large syncytiumlike structures on day 13 (Fig. 4A). The addition of HIG to the medium 24 h postinfection delayed this process but did not prevent the eventual formation of enlarged syncytiumlike structures, whereas the antibody to gB (TRL345) prevented the formation of large syncytiumlike structures. The antipentamer antibodies TRL310, 2-25, 57.4, and 276.1 had even more profound effects, allowing the majority of cells in the fibroblast culture to become infected by day 13 with minimal signs of syncytiumlike structures. These four antibodies have the highest epithelial cell entry-neutralizing potency but no neutralizing activity against fibroblast entry. Antibodies to gH or gH/gL also neutralized epithelial cell entry but with an order-of-magnitude-lower potency (Table 2). That the latter had a correspondingly less profound impact on the formation of syncytiumlike structures (Fig. 4A) suggests that a correlation may exist between the ability of these antibodies to inhibit the formation of syncytiumlike structures and their ability to neutralize epithelial cell entry.
Antibodies to gB, gH/gL, and the pentamer inhibit the formation of syncytiumlike structures in CMV-infected fibroblasts. (A) MRC-5 fibroblast monolayers in 96-well plates were infected with 50 PFU/well of virus BADr. After incubation at 37°C for 24 h, the cultures were washed and medium (∅) or medium containing HIG (5 mg/ml) or MAb (100 μg/ml for TRL345, 25 μg/ml for all others) was added. Representative micrographs were taken on day 7, 10, or 13 postinfection. (B) MRC-5 fibroblast monolayers in 96-well plates were infected at the indicated MOIs with a pentamer-negative (HB15) or pentamer-positive (BADr) variant of strain AD169 or a pentamer-negative (TS15) or pentamer-positive (TS15-rN) variant of strain Towne. Representative micrographs were taken on days 3 and 6 postinfection.
The above-described results suggested that the pentamer may play a role in the formation of syncytiumlike structures. Consistent with this, a tendency to form syncytiumlike structures during fibroblast infection had been anecdotally associated with repair of pentamer-disrupting mutations in strains AD169 and Towne (X. Cui and M. A. McVoy, unpublished data). To confirm this observation, MRC-5 cultures were infected with carefully matched inocula of pentamer-negative and pentamer-positive variants of CMV strains AD169 and Towne. For both strain backgrounds, the formation of syncytiumlike structures was enhanced in the pentamer-positive variant compared to that in the pentamer-negative variant (Fig. 4B).
The N terminus of gH contains a neutralizing epitope specific for epithelial cell entry.Rabbit MAbs 15.1, 58.5, 223.4, and 347.3 were recently reported to react by Western blotting with an ∼125-kDa protein associated with purified CMV virions (51). When used to probe infected-cell lysates, these four MAbs reacted with a protein that migrated between 75 and 100 kDa, as well as a higher-molecular-mass protein (Fig. 5A). The predicted molecular mass of gH is 84 kDa, and the migration of gH in infected-cell lysates closely resembled that of the lower-molecular-mass protein detected by the four MAbs (Fig. 5A). Interestingly, all four MAbs failed to react with lysates of cells infected with the strain Towne virus TS15-rN (Fig. 5A). The remaining neutralizing MAbs failed to react in Western blots with virion- (51) or infected-cell-specific proteins (data not shown), suggesting that they likely recognize discontinuous epitopes.
Four rabbit MAbs recognize linear epitopes in CMV gH. (A) ARPE-19 cells were infected with viruses representing strain AD169 (BADr), Towne (TS15-rN), or Uxc (ABV), and infected-cell lysates were separated by SDS-PAGE, transferred to nitrocellulose membranes, and probed with the indicated rabbit MAbs or with a rabbit antipeptide antiserum specific for gH (anti-gH antiserum). Locations of molecular mass markers are indicated. (B) A fluorescence-based peptide array identified residues 27 to 48 in gH from AD169 as binding selectively to all four MAbs. Within this region, ClustalW alignment of 71 gH peptide sequences in the NCBI database identified five sequence variants that could be assigned to two groups represented by AD169 or Towne. Shown is a ClustalW alignment of the five variants assigned to the AD169 or Towne group. Frequencies within the database are indicated for each group and for individual variants. Signal intensities (fluorescence units) from the peptide array indicate binding of MAb 15.1 to each variant peptide. Similar results (not shown) were obtained for antibodies 58.5 and 223.4. Dashed box indicates the AP86 epitope (residues 34 to 43) that is reported to react with murine MAb AP86-SA4 (59). (C) Two synthetic peptides comprised of gH residues 27 to 48 from AD169 or 27 to 47 from Towne were used as substrates in ELISAs to confirm binding specificities of the four rabbit MAbs. The data shown are representative of two independent experiments.
To determine whether MAb 15.1, 58.5, 223.4, or 347.3 targeted gH and to more accurately map each of their epitopes, a peptide array was constructed that contained 1,612 15-amino-acid peptides with 14-amino-acid overlaps, representing 71 complete gH sequences from all CMV strains searchable from the NCBI database, including AD169 and Towne. Replicate arrays were incubated with each of the four rabbit MAbs, and binding to specific peptides was detected using a secondary antibody conjugated to a fluorescent tag. Antibody 347.3 did not bind specifically to any peptides on the array; however, MAbs 15.1, 58.5, and 223 reacted selectively with peptides comprising residues 27 to 48 of gH from strain AD169, which were identical in strain Uxc (Fig. 5B shows representative data for antibody 15.1). ClustalW alignment of this region obtained from 71 available CMV gH sequences identified a total of five variants that clearly differentiated into two major groups, represented by AD169 and Towne (Fig. 5B). The AD169 group sequences were present in 55% of available gH sequences and included a single variant with a change of A to V at position 38 (A38V) that was present in 1.4% of gH sequences. Towne group sequences were present in 4.5% of gH sequences and included an I30V variant found in 32.4% of gH sequences (e.g., the Merlin strain) and an E28K variant found in 4.2% of gH sequences. All three antibodies reacted with both peptides in the AD169 group (i.e., they tolerated the A38V change), while none reacted with the Towne group peptides (Fig. 5B and data not shown).
To verify the peptide array results, two synthetic peptides representing gH residues 27 to 48 from strain AD169 and 27 to 47 from strain Towne were used as substrates in enzyme-linked immunosorbent assays (ELISAs) to confirm the specificities of the four antibodies. In this experiment, all four antibodies (including MAb 347.3) bound to the AD169 peptide but not to the Towne peptide (Fig. 5C).
These results indicated that the neutralizing epitopes recognized by these four antibodies lie very near the N terminus of gH, within residues 27 to 48. Moreover, the failure of these antibodies to bind to Towne gH suggests that the epitopes at least partially overlap the polymorphic region of residues 28 to 37 (Fig. 5B).
Antibodies to gH are strain specific for neutralizing epithelial cell and fibroblast entry.To evaluate the effects of strain differences on the neutralizing activities of MAbs 15.1, 58.5, 223.4, and 347.3, fibroblast and epithelial cell entry-neutralizing assays were conducted using viruses derived from strains AD169 (BADr), Uxc (ABV), and Towne (TS15-rN). All four antibodies neutralized epithelial cell entry but not fibroblast entry; however, the epithelial cell entry-neutralizing activities of MAbs 15.1, 223.4, and 347.3 were specific for strains Uxc (Fig. 6A) and AD169 (not shown), while MAb 58.5 exhibited very weak activity against strain Towne (Fig. 6A). Thus, as AD169 and Uxc are identical within gH residues 27 to 48 but different from Towne (Fig. 5B), the strain specificities for epithelial cell entry neutralization by these four antibodies are fully concordant with the Western blot and peptide-binding data (Fig. 5).
Fibroblast and epithelial cell neutralizing activities of antibodies to linear epitopes in gH are strain specific. (A) MAbs recognizing linear epitopes in gH were assayed as described in the legends to Fig. 1B and 3 for neutralization of viruses representing CMV strain Uxc (ABV) or Towne (TS15-rN) using MRC-5 fibroblasts or ARPE-19 epithelial cells. Representative micrographs were taken on day 3 (TS15-rN/MRC-5), 4 (ABV/MRC-5 and TS15-rN/ARPE-19), or 6 (ABV/ARPE-19) postinfection. For ABV-based assays, the dilution factors on the left indicate dilutions from initial stocks of HIG (1.25 mg/ml) or MAbs (25 μg/ml). For TS15-rN-based assays, the dilution factors on the left indicate dilutions from initial stocks of HIG (2.5 mg/ml) or MAbs (50 μg/ml). (B) The indicated MAbs were assayed for neutralization of viruses representing CMV strains AD169 (BADr), Towne (TS15-rR), NR, and TB40/E. Neutralizing titers are inverse IC50s (μg/ml); error bars indicate 95% confidence limits.
Similar assays were performed on antibodies 15.1, 58.5, and 223.4 using GFP-tagged epithelial cell-tropic variants of strains AD169 (BADr), Towne (TS15-rR), NR, and TB40/E. The activities of antibodies 15.1 and 223.4 against strain NR paralleled those of Towne, while MAb 58.5 exhibited weak neutralizing activity against epithelial cell entry of Towne but not of NR (Fig. 6B). Given that the gH amino acid sequences from Towne and NR are identical (53), the latter result suggests that sequences outside gH (e.g., possibly gL) can influence the ability of 58.5 to bind gH and neutralize epithelial cell entry. Surprisingly, although all three antibodies were raised against strain AD169 (28), all three neutralized fibroblast entry of strain TB40/E but not of AD169 (Fig. 6B). This strain specificity could not be attributed to polymorphisms in gH residues 27 to 48, as the gH sequences of AD169 and TB40/E are identical in this region.
Antibodies to gH/gL are strain specific for neutralizing fibroblast but not epithelial cell entry.Antibodies 70.7, 124.4, 270.7, 316.2, and 324.4 recognize discontinuous epitopes formed by gH and gL (51) and neutralize both fibroblast and epithelial cell entry (Table 2). Remarkably, antibodies 70.7, 124.4, 270.7, and 316.2 exhibited no strain specificities for neutralizing epithelial cell entry of strain AD169, Uxc, or Towne, but clear strain specificity was evident for neutralizing fibroblast entry, wherein all four antibodies neutralized strains Uxc (Fig. 7A) and AD169 (not shown) but not strain Towne (Fig. 7A). In similar assays, antibodies 70.7, 124.4, and 324.4 neutralized epithelial cell entry of strains AD169, Towne, NR, and TB40/E but failed to neutralize fibroblast entry of strain Towne or NR (Fig. 7B).
Fibroblast but not epithelial cell neutralizing activities of MAbs to discontinuous epitopes in gH/gL are strain specific. (A) Antibodies recognizing discontinuous epitopes formed by gH/gL were assayed as described in the legend to Fig. 6A. (B) The indicated MAbs were assayed as described in the legend to Fig. 6B.
DISCUSSION
Antibodies that neutralize entry of cell-free virus or inhibit cell-to-cell viral spread in vitro have the potential to treat or prevent CMV infection in the clinic. For example, antibodies in mucosal secretions could neutralize inoculum virus and thereby block horizontal transmission, while circulating antibodies could limit dissemination and reduce viral replication at secondary or tertiary sites of infection, such as the placenta, thereby reducing the incidence or severity of fetal disease. However, due to the complexity and cell type specificity of CMV entry and spread, the potency and ability of antibodies to inhibit either entry or cell-to-cell spread depend greatly on the type of cells involved and the specific antigen/epitope that the antibody recognizes. The present studies yield a number of insights into CMV entry and spread within fibroblasts or epithelial cells.
Within the panel of antibodies, all antibodies that neutralize CMV entry into epithelial cells, including those to gB, gH, gH/gL, and the pentamer, also inhibit spread in epithelial cells. That potencies of spread inhibition strongly correlate with neutralizing potencies further suggests that the two assays may be functionally equivalent. For example, if spread to neighboring epithelial cells requires the release of cell-free virus, then spread inhibition by antibodies may simply reflect their neutralizing activity. However, it should be noted that the diversity of epitopes targeted by this panel of antibodies is limited, and it is possible that additional targets or epitopes exist that selectively function in entry or spread. In contrast to the results in epithelial cells, none of the antibodies that neutralized fibroblast entry had a significant impact on spread in fibroblasts, even at very high concentrations. This is consistent with previous reports that both polyclonal and monoclonal neutralizing antibodies have very little effect on CMV spread in fibroblasts (34, 42–46). In addition, 11 nonneutralizing MAbs recognizing known or unknown virion surface epitopes also failed to inhibit spread in fibroblasts. It is curious that CMV spread in epithelial cells apparently lacks this presumably valuable mechanism of antibody evasion. It may be that in vivo, it is more important for CMV to evade antibodies in the context of fibroblast replication than in epithelial cells. However, it is also possible that ARPE-19 cells grown in monolayers have lost the capacity to support an antibody-resistant mechanism of spread, whereas fibroblasts have not. More authentic three-dimensional or raft epithelial cell culture systems may be helpful in addressing this issue.
That antibodies to gB, gH/gL, or the pentamer inhibit the formation of syncytiumlike structures in fibroblasts suggests that each of these glycoprotein complexes plays a role in their formation. While these structures, observed here by fluorescence microscopy, have the appearance of syncytia, additional studies are needed to determine whether they result from fusion of infected cells. However, given that gB and gH/gL form the core CMV fusion machinery (54) and that both gB- and gH-specific antibodies can inhibit their cell fusion-promoting activities (42, 55, 56), it is plausible that the inhibitory effects of gB or gH/gL antibodies on the formation of syncytiumlike structures reported here result from blocking the fusion of infected cells. With respect to the pentamer having a role distinct from that of gH/gL in promoting fusion, there are conflicting data. Vanarsdall et al. found that the pentamer is neither more nor less effective than gH/gL in cooperating with gB to mediate fusion of ARPE-19 cells (54), while Ciferri et al. observed that gB coexpressed with the pentamer resulted in fusion of ARPE-19 cells, while gB coexpressed with gH/gL did not (57). Recent reports that pentamer-specific antibodies inhibit fusion of CMV-infected ARPE-19 cells (50, 58) are consistent with the latter observation, while our data suggest that the pentamer may also enhance fusion during fibroblast infection, since the formation of syncytiumlike structures was enhanced by genetic restoration of pentamer expression and inhibited by pentamer-specific antibodies (Fig. 4A). While the clinical significance of these in vitro findings is uncertain, the occurrence of multinucleated giant cells is a hallmark of CMV infections in vivo and may play a role in pathogenesis.
Rabbit MAbs 15.1, 58.5, 223.4, and 347.3 react with linear epitopes that map within gH residues 27 to 48. As the binding of all four antibodies is strain specific, their epitopes must at least partially overlap the polymorphic residues (28–37) within this region (Fig. 5B). The epithelial cell entry-neutralizing activities measured against AD169, Uxc, and Towne strain viruses were concordant with both Western blotting and peptide binding results in that all four antibodies bound to gH sequences from AD169 and Uxc but not Towne (Fig. 5) and three of the four antibodies neutralized strains AD169 and Uxc but not Towne while the fourth (58.5) neutralized Towne weakly (Fig. 6A and B). These results are similar to findings previously reported for MAb AP86-SA, which reacts with an epitope designated AP86 that is comprised of gH residues 34 to 43 from strain AD169 (Fig. 1B, dashed box) (59). A recent report implicates this region in gH/gL enhancement of gB-mediated fusion of adenovirus-transduced ARPE-19 cells, in that fusion can be impaired both by MAb AP86-SA and by mutations in the AP86 epitope (56). Also recently, Chiuppesi et al. showed that AP86 and a newly isolated MAb, 18F10, both recognize the AP86 epitope and neutralize epithelial cell entry of strains TB40/E and TR (50), which are identical to AD169 in the residue 27-to-48 region. Thus, with respect to epithelial cell neutralization, it appears that the ability to bind to gH residues 27 to 48 determines the strain specificities of MAbs AP86-SA, 18F10, 15.1, 58.5, 223.4, and 347.3.
Antibodies 15.1, 58.5, and 223.4 also neutralized fibroblast entry of TB40/E, but in contrast to epithelial cell neutralization, the strain specificity of their fibroblast neutralization was discordant with their binding activities. Although gH residues 27 to 48 are identical in strains TB40/E and AD169, antibodies 15.1, 58.5, and 223.4 neutralized fibroblast entry of TB40/E but not of AD169 (Fig. 6B). Similarly, AP86-SA was reported to neutralize fibroblast entry of strain AD169 (59), and yet, Chiuppesi et al. found that AP86 and 18F10 both lacked the ability to neutralize fibroblast entry of strains TB40/E and TR (50).
Taken together, these findings suggest that sequence polymorphisms within the residue 27-to-48 region govern both binding and neutralization of virus entry into epithelial cells and fibroblasts; however, fibroblast neutralization is uniquely susceptible to an additional level of restriction by strain polymorphisms that must reside outside the residue 27-to-48 antibody binding region. Similarly, a second group of rabbit MAbs (antibodies 70.7, 124.4, 270.7, 316.2, and 324.4) that neutralize by binding to discontinuous epitopes formed by gH/gL exhibit strain specificity only for fibroblast and not for epithelial cell neutralization (Fig. 7).
At first these results seem perplexing, since neutralization in the context of epithelial cell entry clearly demonstrates that a given antibody can bind gH/gL of a given strain, yet for some strains, the same antibody interacting with the same gH/gL fails to neutralize fibroblast entry. This implies that polymorphisms in sequences external to gH/gL can modulate antibody recognition of certain gH/gL epitopes, presumably by occluding the epitopes or altering their conformation. For example, antibody access to either the linear or discontinuous epitopes in gH/gL could be restricted by polymorphisms in gO. No strain specificity would be observed during neutralization of epithelial cell entry because antibodies can bind to gH/gL in the context of the pentamer, which lacks gO. However, in some strains, polymorphisms in gO either modify the conformation of gH or directly occlude access of the same antibodies to their epitopes in the context of gH/gL/gO, thus blocking the ability of these MAbs to neutralize fibroblast (but not epithelial cell) entry. A similar phenomenon has been described for inhibition of spread by a gH-specific MAb in which strain differences in gO but not gH were proposed to confer resistance (60). While additional studies are needed to confirm such a role for gO, this intriguing hypothesis suggests that gO may play a role in viral escape from antibody neutralization. While presumably this mechanism could be important for reinfection by heterologous strains, the fact that MAbs 15.1, 58.5, and 223.4 were raised against strain AD169 (28) and can neutralize fibroblast entry of TB40/E but not strain AD169 (Fig. 6B) further demonstrates that this mechanism can also protect against antibodies raised against the same strain.
It remains uncertain which cell types are most important in vivo for viral transmission, dissemination, or pathogenesis. A better understanding of the mechanisms and mediators involved in entry, spread, and antibody evasion in different cell types may help to determine what type(s) of antibodies will be most efficacious following active immunization or passive administration.
MATERIALS AND METHODS
Antibodies.The antibodies used in these studies are described in Tables 1 and 2. Rabbit MAbs isolated from a rabbit immunized with a CMV whole-virion experimental vaccine have been described previously (28). Polyclonal rabbit antipeptide antisera to UL130, UL131, and gH are described elsewhere (29) and were a gift from David Johnson. TRL310 (Trellis Bioscience) is a human MAb reconstructed (20) from published sequences of antibody 1F11 specific for a discontinuous epitope formed by UL130 and UL131 (35). Human MAb 2-25 was isolated and cloned from cultured memory B cells of a healthy CMV-seropositive donor (51). TRL345 (Trellis Bioscience) is a human MAb specific for the AD-2 (site I) epitope of gB (20, 52). HIG (Cytogam, CSL Behring, King of Prussia, PA) was purchased from the manufacturer.
Cells and viruses.Human MRC-5 fibroblasts (ATCC CCL-171) and ARPE-19 epithelial cells (ATCC CRL-2302) were obtained from ATCC and propagated in high-glucose Dulbecco's modified Eagle's medium (Gibco BRL) supplemented with 10% fetal calf serum (HyClone Laboratories), 10,000 IU/liter penicillin, and 10 mg/liter streptomycin (Gibco BRL) (medium). Virus BADrUL131-Y4 (BADr) is a variant of CMV strain AD169 in which a mutation in UL131 has been repaired to express a functional UL131 protein (61). Virus HB15-t178b is a nonepithelial cell-tropic pentamer-negative variant of strain AD169 (36). Viruses TS15-rN and TS15-rR are epithelial cell-tropic variants of CMV strain Towne in which a mutation in UL130 has been repaired (48). Virus ABV is an epithelial cell-adapted variant of CMV clinical isolate Uxc (48) (X. Cui and M. A. McVoy, unpublished data). Virus TB40/E has been described previously (62). Virus NR was isolated from a kidney transplant patient (H. Zhu, unpublished data). All viral genomes were cloned as bacmids and have been modified to express green fluorescent protein (GFP) (63, 64). Infectious viruses were reconstituted by transfection of bacmid DNA into ARPE-19 cells and amplified in ARPE-19 cultures to produce working stocks, and the titer was determined as described previously (64).
Accession numbers.Genome sequences associated with the indicated GenBank accession numbers are available for strains AD169 (X17403.1 ), Towne (GQ121041.1 ), and Uxc (KX544840 ). Sequences for the TB40/E (KX544839 ) and NR (KX544831 ) strain variants that were used in these studies were recently published (53).
Western blotting.ARPE-19 cells were infected with BADr, ABV, or TS15-rN at a multiplicity of infection (MOI) of 0.5. After 5 to 7 days, lysates were prepared in Laemmli sample buffer (Bio-Rad) and clarified by centrifugation at 2,000 rpm for 5 min. Clarified lysates were mixed 20:1 with 2-mercaptoethanol, heated to 95°C for 5 min, and separated on precast 10 to 20% SDS-PAGE gels (Bio-Rad) using Tris-glycine-SDS buffer (Bio-Rad). Proteins were transferred electrophoretically for 2 h at 420 mA to 0.2-μm nitrocellulose membranes (Bio-Rad) using Tris-glycine buffer (Bio-Rad) containing 20% methanol. Membranes were blocked for 1 h at room temperature with 5% nonfat dry milk in phosphate-buffered saline (PBS) and then incubated overnight at 4°C with rabbit MAbs in PBS (333 ng/ml) or rabbit anti-gH antiserum diluted 1:2,000 in PBS. Membranes were washed in PBS with 0.2% Tween 20 and then incubated with horseradish peroxidase (HRP)-conjugated goat anti-rabbit IgG (Thermo) diluted 1:5,000 in PBS for 1 h at room temperature. Membranes were again washed with PBS with 0.2% Tween 20, and bound HRP-conjugated antibody was detected using SuperSignal West Pico chemiluminescent substrate (Pierce) and exposure of X-ray film.
Epitope mapping by peptide array and ELISA.High-density peptide arrays were custom-printed at PEPperPRINT (Cambridge, MA). A total of 13,923 CMV sequences were downloaded from GenBank release 204. Pairwise ClustalW alignments were made for each sequence versus sequences from strain Merlin. Amino acid sequences at each position of the alignment were extracted and stored in a relational database. Peptide design and selection were focused on the following seven CMV proteins: UL32 (pp150), UL55 (gB), UL75 (gH), UL132, UL141, UL146, and UL148. For each protein, a library of 15-amino-acid peptides with 14-amino-acid overlaps was designed, based on strain Merlin as a template but extended to include additional unique peptides representing variants present in the GenBank CMV sequences that contain polymorphisms relative to the Merlin sequence. In total, 9,255 unique peptides were identified, of which 1,612 peptides cover the gH (UL75) protein.
Peptide arrays were incubated with rabbit MAbs at 100 ng/ml, and binding was detected using DyLight 680-conjugated goat anti-rabbit IgG(H+L) at a dilution of 1:5,000. The arrays were scanned at an intensity of 7 (red). Hemagglutinin (HA) and Flag control peptides framing the peptide arrays were stained for internal quality control to confirm the assay quality and peptide microarray integrity (using red/green scanning intensities of 7/7). Fluorescence intensities above the background level were used to identify antibody-reactive peptides.
ELISAs were used to confirm the binding specificities of four rabbit MAbs. Synthetic peptides were immobilized at 2 μg/ml in PBS on 96-well FluoroNunc MaxiSorp plates at 4°C overnight. The plates were blocked with 3% nonfat milk in PBS–0.05% Tween 20 and then incubated for 1.5 h with serially diluted rabbit MAbs. The plates were washed and then incubated with HRP-conjugated goat anti-rabbit IgG (Southern Biotech) for 60 min. A fluorogenic HRP substrate, 10-acetyl-3,7-dihroxyphenoxazine (Virolabs), was added at 100 μl per well for 3 to 5 min to generate resorufin, with concentrations proportional to the HRP concentration. Fluorescence signals with excitation at 531 nm were measured with emission at 595 nm using a Victor III plate reader (PerkinElmer). The results were expressed as arbitrary fluorescence units.
GFP-based spread inhibition and neutralization assays.Neutralization and spread inhibition assays were conducted as described previously (48). For neutralizing assays, antibodies were serially diluted in medium, mixed with an equal volume of medium containing 500 PFU of virus, incubated for 1 h at 37°C, and then transferred in triplicate to wells of 96-well plates containing confluent MRC-5 or ARPE-19 cells. For spread inhibition assays, monolayers in 96-well plates were infected with 50 or 100 PFU/well and incubated without antibodies for 24 h. The culture medium was then removed and replaced with medium containing serial dilutions of antibodies. For assays using viruses BADr, TS15-rN, and ABV, representative images were taken using a Nikon Diaphot 300 UV microscope and relative fluorescence units of GFP were measured for each well using a BioTek Synergy HT multimode microplate reader 3 to 6 days postinfection (dpi) for neutralizing assays or 8 to 16 dpi for spread assays. The 50% inhibitory concentration (IC50) values were determined as the inflection points of four-parameter curves fitted to plots of mean relative fluorescence units (from triplicate wells) versus Log (antibody concentration) as described previously (36), using Prism 5 (GraphPad Software, Inc.). Correlation of spread inhibition with neutralizing activity was determined by linear regression analysis of Log-transformed IC50s using Prism 5.
For neutralization assays using viruses BADr, TS15-rR, NR, and TB40/E, approximately 300 PFU of each virus was mixed with each antibody dilution and the mixtures incubated at 37°C for 1 h and then transferred to 96-well plates seeded with 15,000 cells/well of ARPE-19 cells or MRC-5 cells. The plates were imaged 36 to 48 h later on a high-content imager (acumen Cellista laser scanning fluorescence microplate cytometer; TTP Labtech Ltd., Melbourn, United Kingdom). GFP fluorescence was excited with a 488-nm laser, and emission signals with two standard deviations above the background signal were regarded as positive. A positive event was defined as a cell with an isolated fluorescence area between 5 and 10 μm in diameter. The percentage of neutralization was calculated for each well by dividing the number of events in test wells by the number of events in control wells with no antibody. The IC50s were then calculated by four-parameter curve fitting using Prism 5 software.
ACKNOWLEDGMENTS
We thank Thomas Shenk for bacmid clone BADrUL131-Y4, Eain Murphy for providing GFP-tagged TB40/E, Hua Zhu for providing GFP-tagged NR, and David Johnson for rabbit antisera to UL130, UL131, and gH.
This work was supported by grants R01AI088750 and R21AI073615 from the National Institutes of Health (to M.A.M) and by research grants from Merck & Co., Inc., Kenilworth, NJ, USA (to M.A.M.). The funders had no role in study design, data collection, or interpretation.
FOOTNOTES
- Received 19 August 2016.
- Accepted 27 March 2017.
- Accepted manuscript posted online 5 April 2017.
- Copyright © 2017 American Society for Microbiology.