Research ArticleVaccines

A replication-defective human cytomegalovirus vaccine for prevention of congenital infection

See allHide authors and affiliations

Science Translational Medicine  26 Oct 2016:
Vol. 8, Issue 362, pp. 362ra145
DOI: 10.1126/scitranslmed.aaf9387

Progress toward a human cytomegalovirus vaccine

Most adults carry latent human cytomegalovirus, with no apparent symptoms. But if a pregnant woman becomes newly infected or an active infection is triggered, the fetus can be harmed and show severe neurodevelopmental deficits. Wang and colleagues report a key step toward an effective vaccine against this dangerous herpesvirus. They engineered a genetic off switch into the whole virus to render it harmless and show that it can elicit desirable immune responses such as neutralizing antibodies cell-mediated immune responses in several animal species, including nonhuman primates.


Congenital human cytomegalovirus (HCMV) infection occurs in ~0.64% of infants born each year in the United States and is the leading nongenetic cause of childhood neurodevelopmental disabilities. No licensed HCMV vaccine is currently available. Natural immunity to HCMV in women before pregnancy is associated with a reduced risk of fetal infection, suggesting that a vaccine is feasible if it can reproduce immune responses elicited by natural infection. On the basis of this premise, we developed a whole-virus vaccine candidate from the live attenuated AD169 strain, with genetic modifications to improve its immunogenicity and attenuation. We first restored the expression of the pentameric gH/gL/pUL128-131 protein complex, a major target for neutralizing antibodies in natural immunity. We then incorporated a chemically controlled protein stabilization switch in the virus, enabling us to regulate viral replication with a synthetic compound named Shield-1. The virus replicated as efficiently as its parental virus in the presence of Shield-1 but failed to produce progeny upon removal of the compound. The vaccine was immunogenic in multiple animal species and induced durable neutralizing antibodies, as well as CD4+ and CD8+ T cells, to multiple viral antigens in nonhuman primates.


Human cytomegalovirus (HCMV) is a ubiquitous herpesvirus that rarely causes symptomatic infection in healthy individuals (1). However, in immunocompromised patients, including those under immunosuppression after transplantation, HCMV infection may lead to life-threatening diseases (2, 3). Moreover, HCMV can cause congenital viral infections and is responsible for a wide range of neurodevelopmental abnormalities in children (4, 5). Development of a prophylactic vaccine is a public health priority (6).

Children born to HCMV-seropositive women are ~69% less likely to suffer congenital infection than those born to seronegative mothers who acquire HCMV infection during pregnancy (7, 8). Thus, our goal when designing a vaccine was to replicate the immune responses seen in healthy seropositive individuals. Natural HCMV infection elicits broad and robust responses involving both arms of adaptive immunity. HCMV-specific T cells in healthy adults can constitute as much as 10% of the total memory CD4+ and CD8+ T cells that recognize multiple viral proteins, notably, pp65, IE1, IE2, and gB (9). HCMV-specific antibodies potently neutralize viral infection of a variety of cell types (10, 11): gB antibodies primarily prevent infection in fibroblasts, whereas most antibodies that block infection of epithelial and endothelial cells, monocytes, and placenta cytotrophoblasts target the gH/gL/pUL128-131 pentameric complex (1214).

Although whole virus–based vaccines are more likely to produce immune responses resembling those of natural infection, live attenuated HCMV vaccines have yet to be successfully developed. The best-characterized live vaccines are fibroblast-adapted AD169 (15) and Towne (16). Both were safe and well tolerated in clinical studies (1517). Efficacy of the Towne strain has been tested in renal transplant patients (18, 19) and in seronegative women with children in day care (20). In both settings, it failed to prevent primary infection or viral reactivation. In a human challenge study, the Towne strain was less protective than natural immunity against wild-type HCMV infection (21). The reduced efficacy of Towne vaccine was at least partially attributed to its failure to elicit antibodies to the gH/gL/pUL128-131complex (22, 23). Towne virus carries a frameshift mutation in the UL130 open reading frame (ORF) (2325). As a result, the virus is deficient in production of the pentameric complex (23, 25). A frameshift mutation in the UL131 gene was found in AD169, which also affects the pentameric complex expression (24, 25). Because the pentameric complex determines viral tropism for epithelial and endothelial cells, in contrast to clinical isolates, neither Towne nor AD169 can infect these cells efficiently (23, 24, 26).

We recently reported that the restoration of the pentameric complex in AD169 virus significantly enhanced its immunogenicity (27). Here, we further improved the vaccine by rendering it conditionally replication-defective. The control of viral replication was achieved by fusing the destabilizing domain of FK506-binding protein 12 (ddFKBP) to viral proteins IE1/2 and pUL51 (28, 29). The fusion directed these essential proteins to proteasome degradation, which effectively blocked virus progeny production. The degradation of these proteins could be reversed by a synthetic molecule termed Shield-1 (Shld-1), which specifically binds to ddFKBP (28, 30). We tested the control of vaccine virus replication by Shld-1 and its ability to elicit humoral and cell-mediated immune responses, as measured by viral neutralization and interferon-γ (IFN-γ) enzyme-linked immunospot (ELISPOT) assays.


Vaccine construction and in vitro characterization

A derivative from the AD169 vaccine manufactured at Merck (17), described hereafter as MAD169, was used as the parental virus to construct an infectious bacterial artificial chromosome (BAC) clone bMAD-GFP (green fluorescent protein) (fig. S1A). MAD169 carries attenuation markers as reported, such as the UL/b′ deletion (31, 32), UL36 substitution (33, 34), and a frameshift mutation in UL131. The UL131 mutation underlies AD169’s deficiency in infecting epithelial cells (2325), and the frameshift was repaired by deleting one adenine in the first exon of UL131 to create B(b)AC-derived, epithelial-tropic MAD169 GFP-expressing clone, designated as beMAD-GFP (fig. S1C). This clone was then modified by replacing the GFP with a cre recombinase (35) to create a self-excisable BAC, beMAD.

We further attenuated the virus by fusing ddFKBP to a panel of 12 essential genes individually to restrict its replication to a single cycle (table S1). All proteins selected were nonstructural (36), so the ddFKBP fusion proteins are unlikely to be packaged in mature viral particles. The viral constructs were named after the genes to which ddFKBP was fused. Because IE1 and IE2 are expressed from alternatively spliced, major immediate-early transcripts and share the first three exons, both were produced as ddFKBP fusion proteins.

Among the recombinant viruses, ddIE1/2, ddUL51, ddUL52, ddUL84, ddUL79, and ddUL87 were readily rescued. The ddUL37x1, ddUL77, and ddUL53 viruses produced small plaques (table S1), indicating impaired growth; increasing Shld-1 concentration to 10 μM did not improve their growth. The ddUL56 and ddUL105 mutants could not be recovered.

Efficient replication of all ddFKBP mutants depended on the Shld-1 concentration, albeit to varying degrees (fig. S2). In general, lower concentrations of Shld-1 reduced the titer of progeny virus. Among these ddFKBP viruses, only ddUL51 and ddUL52 absolutely required Shld-1 for replication. Other viruses, ddIE1/2, ddUL84, ddUL79, and ddUL87, could produce lower but detectable progeny in the absence of Shld-1 (table S1 and fig. S2), suggesting that addition of ddFKBP to these ORFs did not completely abrogate their functions.

Because the steps of HCMV replication occur in a specific temporal order, we hypothesized that the stringency of viral replication control could be improved if replication was blocked or attenuated at two different stages of the viral life cycle. To test this hypothesis, we constructed a double-tagged virus in which the ddFKBP was fused to the N termini of the IE1/2 and pUL51 proteins. As shown in Fig. 1A, at a multiplicity of infection (MOI) of 0.01 plaque-forming unit (PFU) per cell, no progeny of the ddIE1/2-ddUL51 virus could be detected if the Shld-1 concentration was below 0.1 μM at day 7 after infection. In contrast, after infection with the ddUL51 virus, low levels of progeny virus could be found in the presence of 0.05 μM Shld-1 (fig. S2). Because the ddIE1/2-ddUL51 virus showed the most stringent dependency on Shld-1, it was selected as the candidate and named hereafter as V160 (fig. S3).

Fig. 1. Conditional growth of V160 (ddIE1/2-ddUL51) in ARPE-19 cells.

(A) Production of V160 progeny virus at different concentrations of Shld-1. ARPE-19 cells were infected at an MOI of 0.01 PFU per cell with V160 for 1 hour and then incubated in a medium containing 0, 0.05, 0.1, 0.5, or 2 μM Shld-1. At day 7 after infection, the culture supernatant was collected, and the virus was quantified by median tissure culture infectious dose (TCID50) assay on ARPE-19 cells in the presence of 2 μM Shld-1. (B) Growth comparison of V160 with beMAD virus. ARPE-19 cells were infected with V160 at an MOI of 0.01 PFU per cell and incubated in the absence (○) or presence (●) of 2 μM Shld-1. Cells infected with beMAD at an MOI of 0.01 PFU per cell were included as a control (Δ). Progeny virus was collected at the indicated time points after infection and quantified by TCID50 assay on ARPE-19 cells supplemented with 2 μM Shld-1. The experiments were conducted twice, and the representative data are shown.

The growth kinetics of V160 in the presence of 2 μM Shld-1 were comparable to those of its parental virus beMAD in ARPE-19 cells, with both showing peak titers of ~1 × 107 PFU/ml around day 11. In the absence of Shld-1, no progeny virus could be detected in the supernatants (Fig. 1B). The stringency of V160 replication control by Shld-1 was also tested in different types of human cells including fibroblasts (MRC-5), umbilical vein endothelial cells, aortic smooth and skeletal muscle cells, and neuronal astrocytoma cells (Fig. 2). In all cell types tested, no V160 production could be detected in the absence of Shld-1, suggesting that the conditional replication of the vaccine is not cell type–specific.

Fig. 2. Conditional replication of V160 in different cell types.

Human fibroblasts (MRC-5), endothelial cells [human umbilical cord endothelial cell (HUVEC)], muscle cells [aortic smooth muscle cell (AoSMC) and skeletal muscle cell (SKMC)], or neuronal cells (CCF-STTG1) were infected with V160 at an MOI of 0.1 PFU per cell, except for CCF-STTG1, which were infected at a MOI of 5 PFU per cell. After 1 hour, the cells were washed twice and incubated in the absence or presence of 2 μM Shld-1. Progeny virus was collected at the indicated time points after infection and quantified by TCID50 assay on ARPE-19 cells supplemented with 2 μM Shld-1. The experiments were conducted twice, and the representative data are shown.

Immunogenicity comparison in small animals

The immunogenicity of ddIE1/2, ddUL51, and V160 (ddIE1/2-ddUL51) was compared to that of beMAD. BALB/c mice were immunized with 2.5, 0.63, or 0.16 μg of purified virus at weeks 0 and 3; each microgram of vaccine contained 3 × 105 PFU of virus. Serum samples were collected after vaccination and evaluated for neutralizing activities in ARPE-19 cells. As shown in Fig. 3A, ddIE1/2, ddUL51, or V160 induced comparable neutralizing titers, calculated as reciprocal serum dilutions required to neutralize 50% input virus (NT50). The geometric mean titers (GMTs) for the 2.5 μg/dose groups were 1660, 1260, and 1160, and Tukey’s pairwise comparison revealed no difference among three viruses. The NT50 titers of the beMAD groups at the higher doses (2.5 and 0.63 μg) were slightly lower than those of other vaccines, with the GMT for the 2.5 μg/dose group calculated at 520 (P = 0.07, comparing to V160, Tukey’s pairwise analysis). Thus, the three replication-defective viruses were as immunogenic as beMAD in mice.

Fig. 3. Immunogenicity of V160 in mice and rabbits.

(A) Six-week-old female BALB/c mice (n = 10 per group) were immunized at weeks 0 and 3 with beMAD, ddIE1/2, ddUL51, and V160 (ddIE1/2-ddUL51) vaccines, at doses ranging from 0.16 to 2.5 μg. Serum samples were collected at week 6 and analyzed for HCMV neutralization in ARPE-19 cells. Lines indicate the GMT of NT50 in each group. Tukey’s method for multiple pairwise comparisons was conducted for the groups at the 2.5 μg/dose with n = 10 per group (P = 0.81 for ddIE1/2 versus ddUL51, P = 0.68 for ddIE1/2 versus V160, and P = 0.99 for ddUL51 versus V160). (B) Spleen cells were pooled from three mice in the 2.5 μg/dose groups at week 10. Cellular responses to pp65, IE1, IE2, and gB were determined by IFN-γ ELISPOT assay. (C) New Zealand White rabbits (n = 4 per group) were immunized at weeks 0, 3, and 8, with 10 μg of MAD169, beMAD, or V160 vaccines. Sera were collected at week 11 and analyzed for HCMV neutralization in ARPE-19 cells. Tukey’s method for multiple pairwise comparisons revealed a P value of 1.00 for beMAD versus V160 and a P value of <0.01 for both beMAD versus MAD169 and V160 versus MAD169.

T cell responses induced by beMAD and V160 in mice were determined by IFN-γ ELISPOT assay, using pools of overlapping peptides representing viral antigens pp65, IE1, IE2, and gB. The numbers of antigen-specific spot-forming cells (SFCs) elicited by beMAD and V160 were comparable (Fig. 3B).

Next, we compared the immunogenicity of MAD169, beMAD, and V160 vaccines in rabbits (Fig. 3C). Vaccination with beMAD or V160 viruses elicited comparable neutralizing antibody titers, with respective NT50 GMT values of 1920 and 1870 (P = 1.00). In contrast, the NT50 titer elicited by MAD169 was more than 30 times lower than that elicited by beMAD derivatives expressing the pentameric complex (P < 0.0001).

Vaccine immunogenicity in nonhuman primates

Next, we evaluated the V160 immunogenicity in rhesus macaques. Less than 5% of the rhesus macaques had detectable neutralizing titers to HCMV, even though the entire colony had been exposed to rhesus CMV (rhCMV) (27). The kinetics of vaccine induction of the neutralizing antibodies to HCMV in monkeys were similar to those of primary responses (27), suggesting that even with chronic rhCMV infection, most monkeys were naïve to the HCMV pentameric complex.

The vaccine candidate was tested in the following groups: 100 μg (3 × 107 PFU)/dose, 10 μg (3 × 106 PFU)/dose, and a formulation of 10 μg/dose with Iscomatrix, an adjuvant that has been evaluated in the clinic (37). As a control, we included a group of animals immunized with 30 μg of recombinant gB vaccine, formulated with an oil-in-water adjuvant with properties similar to those of MF59. Neutralizing activity became detectable after the first dose, and the titers peaked 4 weeks after the second and third vaccination (Fig. 4). The peak GMT for the 100 μg/dose group was 14,500, about three times higher than that for the 10 μg/dose group.

Fig. 4. Longitudinal neutralizing titers in rhesus macaques vaccinated with V160.

Groups (n = 5 per group) were vaccinated with V160 at 100 μg/dose, 10 μg/dose, or 10 μg/dose + Iscomatrix at weeks 0, 8, and 24, or recombinant gB vaccine at 30 μg/dose with an oil-in-water adjuvant at weeks 0, 4, and 24. The sera were collected at indicated time points and evaluated for viral neutralization. The GMT of NT50 is plotted longitudinally with the SE for each group. Statistical analysis using Tukey’s method of multiple pairwise comparisons with repeated-measures analysis of variance (ANOVA) revealed P < 0.01, P = 0.02, and P < 0.01 for the overall geometric means over time of V160 at 100 μg/dose, 10 μg/dose, and 10 μg/dose + Iscomatrix versus gB, respectively, and P = 0.89 for V160 at 100 μg/dose versus 10 μg/dose + Iscomatrix.

Iscomatrix enhanced neutralizing titers, with a peak GMT of 15,800 versus 4660 for the 10 μg/dose unadjuvanted groups. Low levels of neutralizing activity were detected in the gB group, with GMTs below 200 throughout the study. At almost 1 year after vaccination, the GMTs for the 100 μg/dose group and the 10 μg/dose + Iscomatrix groups were maintained at 1400 and 3000, whereas the GMT for the 10 μg/dose unadjuvanted group had drifted to about 200. The multiple pairwise comparisons showed significantly higher titers of all V160 groups over the gB group when we assessed vaccine treatment across all time points (P < 0.01, P = 0.02, and P < 0.01 for the 100 μg/dose, 10 μg/dose, and 10 μg/dose + Iscomatrix versus gB group, respectively). The 100 μg/dose and the 10 μg/dose + Iscomatrix groups were comparable (P = 0.89), indicating that the formulation with the Iscomatrix adjuvant could induce immune responses equivalent to those of a higher vaccine dose.

HCMV immune sera could neutralize virus in a variety of cells, but the titers required in epithelial cells are in general five to eight times higher than those in fibroblasts (22, 38). V160 immune sera demonstrated similar patterns (fig. S4), whereas the gB immune sera in our rhesus study showed low neutralizing activities in both cell types, consistent with previous observations (22, 39). To ensure that the gB vaccine was immunogenic, we tested the immune sera for binding activities to purified HCMV virions or recombinant gB, and the result confirmed that the gB vaccine was immunogenic in monkeys (fig. S5).

We tested peripheral blood mononuclear cells (PBMCs) collected before and after vaccination in the IFN-γ ELISPOT assay, using pools of peptides from pp65, IE1, IE2, and gB (Fig. 5). Before vaccination, the number of antigen-specific cells was comparable to that elicited by mock antigen, except in two monkeys that showed responses to gB (Fig. 5B). V160 vaccination elicited T cell responses to both structural (pp65) and regulatory antigens (IE1 and IE2). Iscomatrix adjuvant led to a significantly higher response to pp65 and IE1 than did the unadjuvanted 10 μg/dose treatment (Fig. 5B versus Fig. 5C; P = 0.03 and P = 0.04, repeated-measures ANOVA). The gB vaccine group showed modest responses to gB peptides after vaccination but not to other antigens (Fig. 5D).

Fig. 5. IFN-γ ELISPOT responses of rhesus macaques receiving V160.

Monkeys (n = 5 per group) were vaccinated at 100 μg/dose (A), 10 μg/dose (B), or 10 μg/dose + Iscomatrix (C) or recombinant gB (D) as in Fig. 4, and PBMCs were isolated at weeks 0, 16, 28, and 52. The cells were incubated overnight with the 15-mer peptides pools representing pp65, IE1, IE2, and gB. The spots corresponding to the IFN-γ–producing cells were enumerated and converted to SFC per 106 PBMCs. Mock antigen was the dimethyl sulfoxide (DMSO) solvent. Responses to each antigen in each subject are represented as symbols, and the bars represent GMT for the group. Statistical analysis with repeated-measures ANOVA revealed P = 0.03 and P = 0.04 for the comparison of V160 at 10 μg/dose versus 10 μg/dose + Iscomatrix for the overall geometric mean responses over time to pp65 and IE1, respectively.

PBMCs from the 100 μg/dose and the 10 μg/dose + Iscomatrix groups were further analyzed by intracellular IFN-γ cytokine staining (Fig. 6). One naïve monkey, included as a negative control, showed minimal responses to HCMV antigens but responded to superantigen staphylococcal enterotoxin B (SEB), as expected. All vaccinated monkeys from the 100 μg/dose and 10 μg/dose + Iscomatrix groups responded to viral antigens, with comparable CD4+ and CD8+ T cell responses when stimulated with peptides from pp65, IE1, and IE2; only CD4+ responses were observed after stimulation with purified virions.

Fig. 6. CD8 and CD4 T cell responses in rhesus macaques receiving V160.

PBMCs were collected at week 28 from groups of 100 μg/dose or 10 μg/dose + Iscomatrix. One naïve monkey was included as a control. PBMCs were stimulated with antigens, followed by staining for intracellular IFN-γ and cell surface CD4/CD8 T cell markers. SEB was included as a positive control. Antigens included 15-mer peptide pools for pp65, IE1, IE2, and gB and purified beMAD particles. The data are presented as the number of CD4 (A) or CD8 (B) IFN-γ–positive cells per 106 PBMCs. Each symbol represents one monkey sample, and the bars represent the GMT of the groups.

Conditional replication of V160 vaccine virus

Three studies were conducted to assess the stringency of the ddFKBP/Shld-1 mechanism on control of productive V160 infection. First, we evaluated the MOI effect on virus growth in the absence of Shld-1. ARPE-19 cells were infected at MOIs ranging from 0.01 to 10 PFU per cell. After culturing in the absence of Shld-1, cell-free virus was collected at various time points after infection, and the titers were determined by TCID50 in the presence of Shld-1. No cell-free progeny virus could be detected at any time point after infection, regardless of the initial MOI tested (table S2).

Second, we determined the minimum concentration of Shld-1 required to rescue V160 progeny production. ARPE-19 cells were infected at MOIs ranging from 0.01 to 10 PFU per cell and then cultured in a medium containing various concentrations of Shld-1. The supernatants were collected after infection, and infectivity was measured in the presence of Shld-1. Rescued replication was defined by presence of detectable infectious progeny. An Shld-1 concentration of 50 nM or greater was necessary for viral replication, at an MOI above 0.01 PFU per cell, in ARPE-19 cells (table S3). At an MOI of 0.1 PFU/ml, greater than 25 nM Shld-1 was needed to rescue V160 replication in MRC-5 cells.

Last, we refined the viral culture scheme with two rounds of amplification, attempting to detect Shld-1–independent virions from an inoculum of ~1 × 109 PFUs of V160. The inoculum was plated onto ARPE-19 cells, at an MOI of 2 PFU per cell, statistically capable of infecting 86% of cells (fig. S6A). At day 21 after infection, the supernatants were harvested and added to fresh ARPE-19 monolayers, and these were incubated for an additional 21 days. No visible plaques could be detected in the V160 inoculum, although this expansion procedure allowed detection of Shld-1–independent virus in the control groups, such as beMAD, at levels as low as one infectious virion with no matrix interference and as low as 10 PFUs of beMAD when mixed with the V160. Under the microscope, we observed tiny clusters of two to five cells that were positive for gB immunostaining in V160-inoculated culture (fig. S6B, left), and these clusters of cells were morphologically distinct from the plaques caused by beMAD in which gB-positive staining could be observed in tens to hundreds of cells in each plaque (fig. S6B, right). In this inoculum, we detected 26 clusters with two to five gB-positive cells, suggesting that in the absence of Shld-1, productive V160 replication was reduced by at least six orders of magnitude in vitro.

Screening for pharmaceutical agents that mimic Shld-1 function

It is theoretically possible that a medicinal agent could mimic Shld-1 in function as a ligand for ddFKBP and potentially achieve a sufficient plasma concentration that inadvertently triggers viral replication in the host. We addressed this risk by screening a compound library that included all pharmaceutical agents approved for human use in the United States, Europe, and Japan, as well as some compounds in late-stage clinical development, in a cell-based assay.

Among the compounds tested, only tacrolimus (FK506), sirolimus (rapamycin), and deoxygedunin inhibited ddFKBP-mediated degradation at 25 to 65% of the activity observed with Shld-1 (fig. S7A). These compounds, along with the rapamycin derivatives of everolimus, pimecrolimus, temsirolimus, and deforolimus, were further evaluated (fig. S7B). Only tacrolimus stabilized ddFKBP but with a median effective concentration (EC50) of 0.88 μM, about 20% as potent as Shld-1 (EC50 of 0.18 μM). Sirolimus and other rapamycin derivatives were unable to stabilize ddFKBP in this assay. The concentration of tacrolimus required to support V160 progeny production in vitro was greater than 1 μM (fig. S8), which is more than 12 times higher than the peak plasma concentration in liver transplant patients receiving tacrolimus (40). Thus, we concluded that no currently approved pharmaceutical agents could rescue V160 when used at approved dose levels.

No immune responses to FKBP detected in nonhuman primates

IE1/2 and UL51 are not viral structural proteins, and their ddFKBP fusion variants are not expected to be in the vaccine composition (36). However, T cell responses to IE1 and IE2 indicated that they were expressed in the vaccinated animals (Figs. 5 and 6). Thus, we investigated the responses to FKBP after vaccination in PBMCs and serum samples from 49 V160-vaccinated macaques. T cell responses were assessed by IFN-γ ELISPOT assay after stimulation with overlapping peptides corresponding to the FKBP amino acid sequence. Responses to gB were included as a control. An animal was scored as a positive vaccine responder if the antigen-specific response was ≥35 SFC per 1 × 106 PBMCs and ≥3.5 times than the mock control value or was ≥3.5 times than prevaccination response. Statistical analysis showed that the responder rates were 0% [95% confidence interval (CI), 0 to 7.3] for FKBP and 49.0% (95% CI, 34.4 to 63.7) for gB (table S4; P < 0.01, Fisher’s exact test).

Paired serum samples from each monkey before and after vaccination were analyzed for end point titers to FKBP and gB proteins by enzyme-linked immunosorbent assay (ELISA). Only 1 of the 49 vaccinated animals showed a modest rise in titers to FKBP (from 1:40 to 1:640), whereas the vast majority of monkeys showed elevated titers to recombinant gB after vaccination. The difference in antibody response was statistically significant with a 10-fold (95% CI, 7.2 to 14.3) rise in titers to gB versus to FKBP after vaccination (P < 0.01, paired t test).

The GMT values from ELISPOT and the antibody titers before and after vaccination for each monkey are summarized in Table 1. The paired t test showed no significant change of ELISPOT responses to FKBP peptides after vaccination (P = 0.42), comparing to an about threefold rise in responses to gB peptides (P < 0.01). The fold rise in antibody titers to FKBP was estimated to be 1.1 (P = 0.10) compared to gB at 11.4 (P < 0.01). Thus, de novo expression of ddFKBP in the context of V160 vaccination did not elicit measurable T cells or antibodies to FKBP in monkeys.

Table 1. Geometric mean fold difference in ELISPOT and antibody responses to FKBP and gB.

Fold differences refer to the ratios of number of spots or serum dilution titers before and after vaccination or between ELISPOT or antibody responses to gB control and FKBP from the same individual monkeys (n = 49) and were estimated with paired t tests of log-scaled number of spots or serum titers.

View this table:


Preconception maternal immunity to HCMV is associated with protection against maternal-fetal transmission (7, 8), and both humoral and cellular immunity may be required for preventing congenital transmission (41, 42). Using naturally acquired immunity as a benchmark, we developed a conditionally replication-defective virus, V160, with an antigenic composition closely matching that of wild-type HCMV but without the safety concerns associated with live viruses.

Although both the AD169 and Towne vaccines have been shown to be safe in previous studies (1517), the attenuation profile of a live virus with restored epithelial tropism would be difficult to define without additional clinical studies. The most commonly used strategies to ensure vaccine safety are to inactivate the vaccine through chemical treatment or genetic modification of the virus. Chemical inactivation often compromises the antigenic integrity of the vaccine (43). Traditionally, replication-defective viruses are constructed by viral gene deletion; this strategy has been explored to develop vaccines or vaccine platforms such as RepliVax, alphavirus, and adenovirus vectors (44). However, this technology often requires complementing cell lines to supply the necessary functions that have been deleted from the virus. This strategy is challenging for HCMV because the virus replicates primarily in nontransformed cells of limited life span. The ddFKBP/Shld-1 technology provides an alternative approach to create replication-defective viruses.

Although Shld-1 is more than 1000 times more specific to ddFKBP than original FKBP (30), the technology carries theoretical or undercharacterized risks. First, the control by ddFKBP/Shld-1 on viral replication is conditional, not absolute. Therefore, residual levels of Shld-1 in the vaccine must be monitored to control the risk of V160 replication after vaccination. In addition, pharmaceutical agents have to be monitored to make sure that they do not have any Shld-1–mimicking effects.

V160 replication is severely restricted at two stages of the virus life cycle. However, fusion of ddFKBP with IE1/2 was unable to completely abolish their expression, and low levels of DNA replication and late gene expression continued to occur in ddIE1/2-infected cells (fig. S2). As a result, IE1/2-tagged virus could produce progeny virions in the absence of Shld-1 but at two orders of magnitude lower titer. Similar results were observed in a previous study with another AD169 variant in which IE1/2 was fused to ddFKBP when the progeny virus was cultured in the absence of Shld-1 (29). In contrast to IE1/2, targeting pUL51 alone for degradation was sufficient to block progeny virus production. However, because viral gene expression can continue in the absence of IE1/2 and pUL51, low levels of virus protein production should be expected. Although no infectious virions in culture supernatants were produced as measured by TCID50 assay, we detected gB expression in cells infected with V160 in the absence of Shld-1 (fig. S6B). De novo viral gene expression may be beneficial for V160 immunogenicity because it could provide antigens for processing through the major histocompatibility complex I–restricted endogenous pathway, leading to induction of CD8+ T cells, as shown in our study (Fig. 6).

The efficiency of ddFKBP/Shld-1 to control replication could be affected by the expression level or degradation kinetics of the targeted protein. High-level expression of the ddFKBP-tagged protein may overload the cellular degradation machinery. We routinely detected IE1 expression in V160-infected cells by immunofluorescence staining, consistent with a report showing incomplete degradation of ddFKBP-tagged IE1 in the absence of Shld-1 by immunoblot (29). In cells infected with V160, especially at high MOI, the IE1/2 protein production may overwhelm the cellular degradation capacity, resulting in the accumulation of pUL51 protein, which could lead to low-level production of progeny virions. This constrained infection in culture might be, in part, responsible for the clusters of the gB-positive cells observed by immunostaining (fig. S6B).

Ensuring the genetic stability of the V160 is essential for manufacturing and clinical development. Mutations introduced in ddFKBP under various conditions, such as selection under a suboptimal concentration of Shld-1, could restore viral replication. Although the risk of reversion has been mitigated by targeting two essential genes IE1/2 and UL51, loss of function of ddFKBP fused to pUL51 alone can partially restore growth of V160. Therefore, the absence of replication-competent virus during the vaccine production has to be verified as part of the product quality assurance, and the phenotypic and genetic stability of V160 also needs to be demonstrated during clinical development.

The attenuation markers in AD169, including the UL/b′ deletion and mutations in RL5A, RL13, and UL36, were preserved in V160 (33). Similar mutations have been documented in the Towne strain, which caused no shedding or latent infection in vaccinees (45). The mechanism of attenuation has not yet been identified, but one of the key markers has been mapped to the UL/b′ deletion, an ~15-kb fragment containing 19 genes, many of which modulate host immune responses (1) and promote latency (46). The UL/b′ region is necessary for HCMV to replicate in human tissues engrafted in severe combined immunodeficient mice (26, 47); repairing viral deficiency in epithelial tropism in Towne and AD169 strains while maintaining the UL/b′ deletion is not sufficient to restore the viral growth in this mouse model (48). Last, monkeys infected with an rhCMV strain with the UL/b′ deletion also fail to transmit the virus to cohoused naïve monkeys (49). Thus, maintaining the previously known attenuation markers in V160 provides an additional level of vaccine safety in addition to the ddFKBP/Shld-1 control of viral replication.

V160 elicited cellular responses to viral nonstructural proteins IE1 and IE2 in mice and rhesus monkeys (Figs. 3B and 6). However, there is no evidence that HCMV productively infects mouse or rhesus cells in culture. We were unable to detect V160 shedding or viremia in the vaccinated animals. Viral gene expression after V160 vaccination could be a result of viral entry through an unknown mechanism but was unlikely to be by productive cross-species infection. These animals therefore cannot serve as a challenge model for HCMV. However, rhesus macaques could potentially be used to test the replication-defective vaccine strategy against rhCMV primary infection or reinfection if a rhesus version of V160 is constructed. One complication is that rhCMV-positive monkeys shed virus more frequently (greater than 30% at any time) and at higher viral loads (105 copy number/ml saliva) (49). Thus, it would be technically difficult to assess whether monkeys living in a social environment and being constantly exposed to inoculums of high infectivity can be challenged for reinfection. In addition, there is limited literature on reinfection in rhCMV-positive colonies, and the transmission patterns and reinfection rate are poorly defined.

With built-in attenuation and replication regulation features, V160 presents a low risk for persistent or latent infection in human. V160 growth was unimpeded when supplied with sufficient Shld-1, and escape mutation is unlikely to be favorably selected during manufacturing. The antigenic properties of V160 were indistinguishable from replication-competent HCMV, and the de novo viral gene expression in the absence of active viral replication was desirable for cell-mediated responses to HCMV. Addition of Iscomatrix adjuvant could promote T cell responses and produce similar levels of immune responses with lower vaccine doses. Preliminary data from our ongoing clinical evaluation suggest that V160’s safety and immunogenicity profiles in human are consistent with those described in this preclinical study. V160 is therefore a promising vaccine candidate against congenital HCMV transmission.


Study design

We exploited the ddFKBP/Shld-1 technology to construct a conditionally replication-defective HCMV vaccine. The Shld-1–dependent replication was demonstrated by the deficiency of progeny production in cell cultures in the absence of Shld-1. The stringency of the regulation was determined by the minimum concentration of Shld-1 required to rescue the vaccine virus replication. The immunogenicity of the virus was evaluated in mice, rabbits, and rhesus macaques. The animals were randomly allocated and vaccinated with V160. Vaccine immunogenicity was determined by analysis of HCMV-specific antibodies and T cells. The animal studies were not blinded, and the group sizes varied from 5 to 10 in the different studies.


MAD169 genome was cloned as an infectious bacmid (35) and transformed into Escherichia coli SW105. The ddFKBP ORF was introduced by galK positive-negative selections (50). The growth kinetics was characterized as reported (25). Viruses were produced in MRC-5 or ARPE-19 cells, and viral particles were harvested and purified twice by centrifugation at 55,000g for 90 min through 20% sorbitol cushion in 50 mM tris (pH 7.2) and 1.0 mM MgCl2. The pellet was resuspended and stored at −70°C. Recombinant gB protein was purchased from Sino Biological Inc. The gB sequence was derived from Towne strain, with deletion of the transmembrane region and mutations in the furin cleavage site (51). The protein was produced by transient transfection and purified using affinity chromatography. The oil-in-water adjuvant was made on the basis of the composition of MF59 (U.S. patent 6299884). The adjuvant was characterized for both chemical and physical properties before use. Iscomatrix adjuvant was provided by CSL Ltd. Formulating V160 with Iscomatrix showed no effect on virus infectivity in culture.

Animal studies

All studies were conducted in accordance with the Guide for the Care and Use of Laboratory Animals, and the protocols were approved by the Institutional Animal Care and Use Committee. Specific pathogen–free female BALB/c mice 4 to 8 weeks old were immunized by intramuscular injection of both quadriceps in 50 μl of volume per site without anesthesia. Female New Zealand White rabbits 3 to 4 months of age were immunized by intramuscular injection of quadriceps in 0.5 ml without anesthesia. All monkeys were exposed to rhCMV before this study and were screened for inclusion based on their neutralizing titers and ELISPOT responses to HCMV. Rhesus macaques were anesthetized before vaccines were delivered intramuscularly in 0.5-ml volume into deltoid muscles.

Immunological assays

The neutralization antibodies were titrated by an immunostaining-based assay (52). Sera were heat-inactivated at 56°C for 30 min before the assay. No complement was added. HCMV-specific binding antibody titers were determined by ELISA (38). The IFN-γ ELISPOT assay was conducted as described (27). Antigens were peptide pools of 15-mer peptides overlapping by 11 amino acids representing proteins pp65, gB, IE1 and IE2, or FKBP. For IFN-γ cytokine staining, PBMCs were incubated with peptide pools, purified HCMV virions or SEB at 2 μg/ml, or medium control with 0.5% DMSO diluent, along with anti-CD28 and anti-CD49d costimulatory antibodies. Brefeldin A was then added, and cells were incubated for 6 hours. After stimulation, the cells were washed and stained for surface markers and intracellular IFN-γ (anti-human IFN-γ–fluorescein isothiocyanate), and then analyzed on FACSCalibur. Responses were normalized to the number of CD4/CD8 IFN-γ–positive cells per 1 × 106 PBMCs. The statistical analysis of the responses can be found in the Supplementary Materials and Methods.


Materials and Methods

Fig. S1. Schematic diagram on the construction of epithelial-tropic AD169 virus.

Fig. S2. Effect of Shld-1 concentration on progeny production of ddFKBP fusion mutants.

Fig. S3. Genetic map of V160, MAD169, and wild-type HCMV.

Fig. S4. Comparison of epithelial and fibroblast neutralization titers by V160 or gB vaccines in rhesus macaques.

Fig. S5. Kinetics of anti-beMAD or anti-gB antibodies after V160 vaccination.

Fig. S6. V160 infectivity and progeny production in the absence of Shld-1.

Fig. S7. High-throughput screening of medicinal compound collection.

Fig. S8. Potency of Shld-1 and tacrolimus (FK506) in rescuing V160.

Table S1. Construction and characterization of ddFKBP fusion mutants.

Table S2. Effect of MOI on Shld-1 dependent V160 replication.

Table S3. Minimum Shld-1 and tacrolimus (FK506) concentration required to rescue V160.

Table S4. ELISPOT responder rates to FKBP versus gB antigens.


  1. Acknowledgments: We thank M. Salnikova, C. Barr, and D. Krah of Merck for providing the adjuvants and MAD169. We thank the Merck Laboratory Animal Services and the animal care staff at the New Iberia Research Center for their excellent technical support. We appreciate the critical reviews from M. Feinberg, L. Musey, and R. Peluso of Merck, and E. Maraskovsky of CSL Ltd. Funding: All research activities were funded by the Merck Research Laboratories. Author contributions: The project was conceptualized and designed by D.W. and T.-M.F. Statistical analysis was performed by L.S. Vaccine viruses were constructed by X.H. and F.L. Virological properties were characterized by X.H., F.L., D.C.F., A.T., M.B.M., and J.X. Vaccine immunogenicity was evaluated by D.C.F., F.L., A.T., K.S.C., S.A.D., and S.C. Screening for pharmaceutical agent was performed by Y.L. The manuscript was written by D.W. and T.-M.F. and edited by D.C.F., Z.-Q.Z., A.C.F., and A.S.E. Data were analyzed and interpreted by all authors. Competing interests: All authors are currently or previously employed by Merck and Co. Inc. Merck has filed a patent application pertaining to the antigenic composition of this vaccine candidate. Data and materials availability: All data and materials are properties of Merck and are available upon request through a material transfer agreement.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article