Evasion of CD8+ T Cells Is Critical for Superinfection by Cytomegalovirus

See allHide authors and affiliations

Science  02 Apr 2010:
Vol. 328, Issue 5974, pp. 102-106
DOI: 10.1126/science.1185350


Cytomegalovirus (CMV) can superinfect persistently infected hosts despite CMV-specific humoral and cellular immunity; however, how it does so remains undefined. We have demonstrated that superinfection of rhesus CMV–infected rhesus macaques (RM) requires evasion of CD8+ T cell immunity by virally encoded inhibitors of major histocompatibility complex class I (MHC-I) antigen presentation, particularly the homologs of human CMV US2, 3, 6, and 11. In contrast, MHC-I interference was dispensable for primary infection of RM, or for the establishment of a persistent secondary infection in CMV-infected RM transiently depleted of CD8+ lymphocytes. These findings demonstrate that US2-11 glycoproteins promote evasion of CD8+ T cells in vivo, thus supporting viral replication and dissemination during superinfection, a process that complicates the development of preventive CMV vaccines but that can be exploited for CMV-based vector development.

A general characteristic of the adaptive immune response to viruses is its ability to prevent or rapidly extinguish secondary infections by identical or closely related viruses. A notable exception is the herpesvirus family member cytomegalovirus (CMV), which can repeatedly establish persistent infection in immunocompetent hosts (13). Sequential infections are likely the reason for the presence of multiple human CMV (HCMV) genotypes in the human host (4). This ability to establish secondary persistent infections despite the pre-existence of persistent virus (referred to as “superinfection”) is particularly notable because healthy CMV-infected individuals develop high-titer neutralizing antibody responses and manifest very-high-frequency CD4+ and CD8+ CMV-specific T cell responses (>10% of circulating memory T cells can be CMV-specific) (5). This evasion of pre-existing immunity has frustrated attempts to develop preventive CMV vaccines (6, 7) but can be exploited for the development of CMV vectors capable of repeatedly initiating de novo T cell responses to heterologous pathogens in CMV-positive hosts (3).

The biologic importance of this superinfection capability has prompted our investigation of its extent and mechanism. We previously showed that inoculation of RhCMV+ rhesus macaques (RM) with 107 plaque-forming units (PFU) of genetically modified RhCMV (strain 68-1) expressing simian immunodeficiency virus (SIV) antigens resulted in superinfection manifested by the persistent shedding of the genetically modified CMV in the urine and saliva and by the induction and long-term maintenance of de novo CD4+ and CD8+ T cell responses specific for the SIV insert (3). To determine whether RhCMV would be able to overcome immunity at lower, more physiologic doses of infection, as reported for HCMV (7), a recombinant RhCMV containing a loxP-flanked expression cassette for SIVgag [RhCMV(gagL)] (fig. S1) was inoculated subcutaneously at doses of 104 or 102 PFU into four RM naturally infected by RhCMV, as manifested by the presence of robust RhCMV-specific T cell responses (table S1A). The SIVgag-specific T cell responses in peripheral blood mononuclear cells (PBMC) or in broncho-alveolar lavage lymphocytes (BAL) were monitored by flow cytometric analysis of intracellular cytokine staining (ICCS) (figs. S2 and S3) after stimulation with consecutive overlapping 15–amino acid peptides corresponding to SIVgag (8). Reduction of the inoculating dose had minimal impact on superinfection dynamics: All animals developed SIVgag-specific T cell responses within 2 weeks (Fig. 1A), and secretion of SIVgag-expressing virus in urine or buccal swabs was observed within 4 to 10 weeks of infection in both cohorts (Fig. 1B). The time to first detection of secreted virus in these low-dose-challenged RM was not materially different from that of eight RhCMV+ animals infected with 107 PFU of RhCMV(gagL) (Fig. 1B). Moreover, the SIVgag-specific T cell responses and RhCMV(gagL) secretion were stable for more than 3 years regardless of initial dose (Fig. 1, A and C). These data indicate that, consistent with HCMV in humans, RhCMV is able to overcome high levels of CMV-specific immunity and to establish secondary persistent infections, even with low doses of challenge virus.

Fig. 1

Superinfection of RhCMV-positive animals is independent of viral dose. (A) At day 0, two cohorts of four RhCMV+ animals each were infected subcutaneously with 102 or 104 PFU of RhCMV(gagL). The SIVgag-specific T cell responses in PBMC or in BAL were monitored by flow cytometric analysis of ICCS for CD69 and tumor necrosis factor–α (TNF-α) (8) (figs. S2 and S3). (B) Day of first detection of SIVgag-expressing virus in the urine or buccal swabs collected at the indicated intervals from each animal in the two cohorts shown in (A). Also included are results from a third cohort of eight RhCMV+ animals inoculated with 107 PFU of RhCMV(gagL). Expression of SIVgag was determined by immunoblot using antibody to SIVgag from viral cocultures (8). Each circle represents an individual animal. (C) Long-term secretion of SIVgag-expressing virus. Urine was isolated at the indicated days post-infection (PID) from each of the RhCMV(gagL)-infected RM, and SIVgag expression was detected from cocultured virus by immunoblot. For control, a RhCMV-positive animal that did not receive RhCMV(gagL) was included.

We hypothesized that an essential step during CMV superinfection is the ability of the virus to clear an initial immunological checkpoint. A likely candidate for such an immunological barrier is CD8+ cytotoxic T cells (CTL), because they are crucial for controlling CMV-associated diseases (9). The importance of CTL control for CMV is also suggested by viral expression of multiple proteins that inhibit presentation of viral peptide antigens to CD8+ T cells via major histocompatibility complex class I (MHC-I) molecules (10). HCMV encodes at least four related glycoproteins, each with a unique mechanism to prevent antigen presentation: US2 and US11 mediate the retrograde translocation of MHC-I into the cytosol for proteasomal destruction (11), US3 retains MHC-I in the endoplasmic reticulum by interfering with chaperone-controlled peptide loading (12), and US6 inhibits the translocation of viral and host peptides across the endoplasmic reticulum membrane by the dedicated peptide transporter TAP (transporter associated with antigen processing) (13). RhCMV encodes sequence and functional homologs of these genes in a genomic region spanning Rh182 (US2) to Rh189 (US11) (fig. S1) (14). Furthermore, the Rh178 gene encodes the RhCMV-specific viral inhibitor of heavy chain expression (VIHCE), which prevents signal-sequence-dependent translation/translocation of MHC-I (15).

To determine whether MHC-I interference and CTL evasion played a role in the ability of CMV to superinfect CMV+ animals, we replaced the entire RhUS2-11 region with a SIVgag expression cassette using bacterial artificial chromosome (BAC) mutagenesis, resulting in virus ∆US2-11(gag) (8). We also deleted Rh178 to generate ∆VIHCE∆US2-11(gag) (fig. S1). We previously showed that MHC-I expression is partially restored upon US2-11 deletion, whereas additional deletion of Rh178 fully restores MHC-I expression in RhCMV-infected fibroblasts (15). In vitro analysis showed that all viruses were deleted for the targeted RhCMV open reading frames (ORFs), did not contain any unwanted mutations, and replicated comparably to wild-type RhCMV (figs. S4 and S5). First, we examined whether these viruses were able to infect animals that were CMV-naïve as shown by a lack of CMV-specific T cell responses (table S1B). Three groups of animals were challenged with 107 PFU of ∆US2-11(gag) (n = 2), ∆VIHCE∆US2-11(gag) (n = 2), or BAC-derived (wild-type) RhCMV(gag) (n = 2). T cell responses against both CMV and SIVgag in PBMC and against SIVgag in BAL were comparable between animals infected with the deletion mutants and the wild-type RhCMV(gag) control (Fig. 2A). Moreover, all animals secreted SIVgag-expressing virus from day 56 onward for the duration of the experiment (>700 days) (Fig. 2B). Polymerase chain reaction (PCR) analysis of DNA isolated from urine cocultured virus at day 428 confirmed that the secreted viruses lacked the respective gene regions and were able to persist in the host (Fig. 2C). Together these results show that viral MHC-I interference is dispensable for primary infection and the establishment and maintenance of persistent infection, despite the development of a substantial CMV-specific T cell response.

Fig. 2

Interference with MHC-I assembly is not required for primary infection of CMV-naïve animals. Three cohorts of two RM each were inoculated subcutaneously with 107 PFU of recombinant ∆US2-11(gag), ∆VIHCE∆US2-11(gag), or RhCMV(gag). (A) The RhCMV-specific T cell response in PBMC and the SIVgag-specific T cell response in PBMC and BAL were determined at the indicated days post-infection using overlapping peptides to RhCMV immediate early genes IE1 and IE2 or SIVgag by flow cytometric analysis of ICCS for CD69, TNF-α, and interferon-γ (IFN-γ) (8) (figs. S2 and S3). (B) Immunoblot of RhCMV-IE2 or SIVgag expressed in cocultures of urine samples obtained from animals infected with ∆US2-11(gag) or ∆VIHCE∆US2-11(gag). The IE2 blot confirms that the animals were negative for RhCMV before infection, consistent with results from T cell assays (table S1B). (C) PCR analysis of viral genomic DNA isolated from viral cocultures at 428 days post-infection. The presence or absence of indicated ORFs was determined by PCR using specific primers (8). One of the animals infected with RhCMV(gag) served as a control.

To examine whether viral MHC-I interference was required for superinfection of RhCMV+ RM, we challenged two cohorts of four naturally infected RM each with 107 PFU of ∆VIHCE∆US2-11(gag) or RhCMV(gag). All animals displayed immediate early gene (IE)–specific CD4+ and CD8+ T cell responses before challenge (Fig. 3A and table S1C). In keeping with previous results (3), RM inoculated with wild-type RhCMV(gag) displayed boosting of the RhCMV-specific T cell response and developed a SIVgag-specific immune response (Fig. 3, A and B, insets). They also secreted SIVgag-expressing virus (Fig. 3C). In contrast, we did not detect SIVgag-specific T cell responses in PBMC or BAL in RM inoculated with ∆VIHCE∆US2-11(gag), even after repeated inoculation (Fig. 3, A and B), and SIVgag-expressing virus was not detected in secretions (Fig. 3C). These results suggested that MHC-I interference was essential for superinfection. Inoculation of the same animals with ∆US2-11(gag) and, later, ∆VIHCE(gag) demonstrated that superinfection required the conserved US2-11 region but not the VIHCE region. The development of SIVgag-specific CD4+ and CD8+ T cell responses in blood and BAL (Fig. 3, A and B), as well as the boosting of preexisting RhCMV-specific CD4+ and CD8+ T cell responses in blood (Fig. 3A), or shedding of SIVgag-expressing RhCMV (Fig. 3D) were only detectable after challenge with ∆VIHCE(gag) but not with ∆US2-11(gag).

Fig. 3

US2-11-deleted RhCMV is unable to superinfect RhCMV+ rhesus macaques. (A) A cohort of four RhCMV+ RM was inoculated subcutaneously with 107 PFU of ∆VIHCE∆US2-11(gag) (∆V∆U) at days 0 and 91. The CD4+ and CD8+ T cell response to SIVgag or RhCMV-IE was monitored by flow cytometric analysis of ICCS for CD69, TNF-α, and IFN-γ in PBMC. The percentage of the responding, specific T cells within the overall memory subset is shown for each time point. At day 154 and again on day 224, the same cohort was inoculated with 107 PFU of ∆US2-11(gag) (∆U), and RhCMV-IE and SIVgag-specific T cell responses were monitored bi-weekly. At day 737, the cohort was inoculated with ∆VIHCE(gag) (∆V), and the T cell response was monitored as before. At day 989, the cohort was inoculated with ∆Rh186-8(retanef) (∆R). Besides SIVgag, a T cell response to SIVrev/nef/tat was detected by ICCS in all four animals (black lines) using corresponding overlapping peptides. (Inset) A separate cohort of four animals was infected with wild-type RhCMV(gag), and the RhCMV-IE and SIVgag-specific CD4+ and CD8+ T cell response was monitored as described above at the indicated time points for 133 days. (B) The CD4+ and CD8+ T cell response to SIVgag in BAL was measured in parallel to the PBMC T cell responses shown in (A). (C) RhCMV secreted in the urine collected from the cohort infected with RhCMV(gag), or deletion viruses ∆VIHCE∆US2-11(gag) or ∆US2-11(gag), labeled ∆CMV. Virus was isolated at the indicated days by coculture with telomerized rhesus fibroblasts (TRFs), and cell lysates were probed for expression of SIVgag by immunoblot. (D) Expression of RhCMV-IE2, SIVgag, and SIVretanef by virus secreted in urine collected at the indicated days. Note that all animals were IE2-positive at the onset of the experiment, confirming their RhCMV-positive T cell status (table S1D).

Our results show that genes within the US2-11 region are essential for superinfection, which is consistent with the known function of US2, US3, US6, and US11 as inhibitors of MHC-I antigen presentation. There are, however, three genes of unknown function (Rh186 to Rh188) encoded between US6 and US11. Rh186 and Rh187 are most closely related to the HCMV glycoproteins US8 and US10, respectively (14), whereas Rh188 is an uncharacterized RhCMV-specific ORF. Although binding of HCMV-US8 and US10 to MHC-I has been reported, it is unclear whether this affects antigen presentation because MHC-I surface expression is not reduced by US8 or US10 from either HCMV or RhCMV (14, 16, 17). To determine whether Rh186, Rh187, or Rh188 are required for superinfection, we generated deletion virus ∆Rh186-8. To enable us to monitor superinfection by this recombinant in the same cohort of animals that had already been reinfected with ∆VIHCE(gag), we applied a distinct immunological marker, SIVretanef, a fusion-protein consisting of SIV proteins rev, tat, and nef (3). ∆Rh186-8(retanef) is deleted for Rh186-188 and contains the Retanef expression cassette between the ORFs Rh213 and Rh214 (fig. S1). We inoculated the same cohort with ∆Rh186-8(retanef) and monitored the T cell response to this fusion protein as well as to RhCMV-IE and SIVgag using corresponding peptides. As shown in Fig. 3, A and B, all four RM developed a SIVretanef-specific T cell response within 2 weeks post-challenge, indicating successful superinfection. Moreover, virus expressing SIVretanef was shed in the secretions of infected animals together with SIVgag-expressing ∆VIHCE(gag) (Fig. 3D). We thus conclude that the Rh186-8 region is dispensable for superinfection.

Together, our results suggested that RhCMV was unable to superinfect in the absence of the homologs of US2, US3, US6, and US11 because the virus was no longer able to avoid elimination by CTL. To further examine this hypothesis, a new group of RhCMV+ RM (table S1D) was depleted for CD8+ lymphocytes by treatment with cM-T807, a humanized monoclonal antibody to CD8, before superinfection with ∆US2-11(gag) or ∆VIHCE∆US2-11(gag). Flow cytometric analysis of total CD8+ T cells revealed that depletion was extensive, but transient, with detectable CD8+ T cell recovery beginning on day 21 after challenge (Fig. 4, A and B). Upon inoculation with ∆US2-11(gag) or ∆VIHCE∆US2-11(gag), SIVgag-specific CD4+ T cell responses were recorded as early as day 7 post-challenge, showing the ability of the deletion viruses to superinfect these animals (Fig. 4C). Moreover, SIVgag-specific CD8+ T cells were observed within the rebounding CD8+ T cells in blood and BAL at day 21 in two RM and at day 28 in a third; in the fourth RM, such responses were only observed in BAL after day 56. From these data, we conclude that CD8+ lymphocytes, most likely CD8+ T cells, were essential for preventing superinfection by ∆US2-11 virus, strongly indicating that the MHC-I inhibitory function of these molecules is necessary for superinfection of the CMV-positive host. Notably, CMV-specific CD8+ T cells were unable to eliminate RhCMV lacking MHC-I inhibitors once persistent infection had been established (Fig. 4D), providing additional evidence that persistent infection is insensitive to CD8+ T cell immunity, even when the ability of the virus to prevent MHC-I presentation is compromised.

Fig. 4

CD8+ T cells protect rhesus macaques from infection by RhCMV lacking MHC-I inhibitors. (A) Four CMV-positive RM were treated at the indicated days with CM-T807, an antibody to CD8, before and after inoculation with 107 PFU of ∆VIHCE∆US2-11(gag) (two animals, black lines) or ∆US2-11(gag) (two animals, red lines). The absolute counts of CD8+ T cells in the blood of each animal are shown over time. (B) The presence of CD4+ and CD8+ T cell populations in PBMC of one representative animal is shown for the indicated days. (C) SIVgag-specific CD4+ and CD8+ T cell responses in PBMC and BAL of CD8+ T cell–depleted animals were monitored by ICCS for CD69, TNF-α, and IFN-γ and are shown as a percentage of total memory CD4+ or CD8+ T cells. Note the delayed appearance of SIVgag-specific CD8+ T cells. (D) Expression of SIVgag or IE2 by RhCMV secreted in the urine of animals infected upon CD8+ depletion.

Our data imply that T cell evasion is not required for establishment of primary CMV infection or once the sites of persistence (e.g., kidney and salivary gland epithelial cells) have been occupied, but rather it is essential to enable CMV to reach these sites of persistence from the peripheral site of inoculation in the CMV-immune host. One possible scenario is that viral infection of circulating cells, for example, monocytes, can succeed only if the virus prevents elimination of these cells by virus-specific CTLs. More work, however, will be required to identify the cell type supporting superinfection.

Although the biochemical and cell biological functions of US2, US3, US6, and US11 have been studied extensively (18), their role in viral pathogenesis had remained enigmatic. Analogous gene functions in murine CMV (MCMV) had been similarly found to be dispensable for both primary and persistent infection (10), although reduced viral titers have been reported for MCMV deleted for these genes (19). Thus, the reason all known CMVs dedicate multiple gene products to MHC-I down-regulation had remained elusive. Our current results now identify a critical role for these immunomodulators to enable superinfection of the CMV-positive host. Furthermore, these results suggest that the ability to superinfect is an evolutionary conserved function among CMVs and therefore might play an important role in the biology of these viruses. Superinfection could promote the maintenance of genetic diversity of CMV strains in a highly infected host population, which could provide an evolutionary advantage. However, there is another possibility. CMV is a large virus with thousands of potential T cell epitopes and therefore a high potential for CD8+ T cell cross-reactivity (20). Indeed, in a study of pan-proteome HCMV T cell responses, 40% of HCMV seronegative subsets manifested one or more cross-reactive CD8+ T cell responses to HCMV-encoded epitopes (5). As CMV recognition by cytotoxic T cells appears to effectively block primary CMV infection, individuals with cross-reactive CD8+ T cell immunity might be resistant to CMV. Thus, US2-11 function may be necessary to evade such responses and establish infection in this large population of individuals that might otherwise be CMV-resistant.

Our results also may explain why, so far, it has not been possible to develop a vaccine that efficiently protects humans from HCMV infection. Although antibody-mediated mucosal immunity might reduce the rate of superinfection (7, 21), once this layer of defense is breached, CMV-specific CTLs seem to be unable to prevent viral dissemination, due to MHC-I down-regulation by US2-11. Thus, although CMV-vaccines might be able to limit CMV viremia and associated morbidity, this MHC-I interference renders it unlikely that sterilizing protection against CMV infection is an achievable goal.

Supporting Online Material

Materials and Methods

Figs. S1 to S5

Table S1


1 December 2009; accepted 5 February 2010

  • * These authors contributed equally to this work.

  • Present address: Salk Institute for Biological Studies, 10010 North Torrey Pines Road, La Jolla, CA 92037, USA.

References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. We are grateful to P. Barry for providing the RhCMV-GFP and RhCMV-BAC and M. Messerle for plasmid ori6k-F5. The NIH Nonhuman Primate Reagent Resource Program provided the CD8α-specific antibody cM-T807 used in this work (contracts AI040101 and RR016001), which was originally obtained from Centocor, Inc. We thank A. Townsend for help with the graphics. We thank D. Drummond, L. Coyne-Johnson, M. Spooner Lewis, C. Hughes, N. Whizin, M. Giday, J. Clock, J. Cook, J. Edgar, J. Dewane, and A. Legasse for technical assistance. This research was supported by the National Institutes of Health (RO1 AI059457 to K.F. and RO1 AI060392 to L.J.P.), the International AIDS Vaccine Initiative (to L.J.P.), the National Center for Research Resources (RR016025 and RR18107 to M.K.A.; RR00163 supporting the Oregon National Primate Research Center), the Ruth L. Kirschstein National Research Service Awards T32 AI007472 (to C.J.P.) and T32 HL007781 (to R.R.), the OHSU Tartar Trust fellowship (to C.J.P.), and the Achievement Reward for College Scientists (to R.R.). The CMV vector technology has been patented by Oregon Health and Science University (US2008/0199493A1) with L.J.P., J.A.N., and M.A.J. as coinventors. This patent has been licensed to the International AIDS Vaccine Initiative. The comparative genome sequencing data shown in fig. S5 were deposited in the Gene Expression Omnibus database (accession number GSE20308).
View Abstract

Navigate This Article