Background
The success of the prototype poxvirus vaccinia virus (VV) to eradicate naturally occurring smallpox worldwide stands as one of the most remarkable medical achievements in human history [
1]. This achievement came at a price, however. Inoculation with live VV can overwhelm the immune system of immunocompromised individuals, causing significant morbidity and mortality [
2,
3]. As a result of this realization, toward the end of the smallpox immunization campaign, a highly attenuated VV strain, modified vaccinia virus Ankara (MVA), was developed for use in individuals at high risk of VV-associated adverse events. MVA has since been safely administrated to over 120,000 individuals in the late stages of the smallpox eradication effort [
4]. MVA lost approximately 15% of the vaccinia genome in the course of over 500 passages in chicken embryonic fibroblasts
ex vivo, along with the ability to replicate in most primary mammalian cells [
4,
5]. Compared to replicating VV, MVA provides similar or higher levels of endogenous or recombinant gene expression even in non-permissive cell lines, higher but more transient Ag expression
in vivo, and comparable levels of immune responses in animal models [
6‐
12]. In light of this favorable degree of immunogenicity, combined with its highly attenuated phenotype and associated inherent safety properties, MVA is being actively explored as a promising vaccine vector in both preventative and therapeutic indications for a number of infectious diseases and malignancies. In addition, MVA has also been the subject of renewed interest and clinical evaluation as a safer vaccine to prevent smallpox in light of concerns about potential bioterrorist threats and the increased sensitivity to vaccine safety considerations especially at a time when far more individuals with acquired immunodeficiency (by virtue of either therapeutic immunosuppression or HIV infection) are present in the population than when the smallpox eradication campaign was mounted [
11,
13,
14].
Murine models represent an extensively utilized preclinical experimental system for vaccine development. In the case of MVA, the murine model has been employed, to date, to study both MVA-induced cellular and humoral immune responses, as well as the efficacy of protection following experimental pathogenic VV challenge [
15,
16]. However, little has been done to understand how this highly attenuated non-replicating virus activates both branches of the immune system. Of particular interest, few systematic studies have been carried out to investigate the interaction between MVA and murine dendritic cells (DCs), the most potent APCs to activate naïve CD4 and CD8 T cells [
17]. In contrast, there has been considerable interest and effort invested in the study of the interaction of VV and human DCs, and to a somewhat lesser extent, the interaction between MVA and human DCs [
12,
18]. VV has been shown to abortively infect human DCs, block their maturation, inhibit their phagocytosis and directional migration, and to induce extensive DCs apoptosis [
19‐
22]. These observations suggest that targeted infection of human DCs by VV may be an important viral strategy to circumvent host immune defenses. All of these studies employed human monocyte-derived DC (MoDC) cultures. It has not been assessed, in the very tractable murine model system, how and to what extent specific consequences of poxvirus infection of DC manifested in culture predict the ability of the host to generate antiviral immune responses
in vivo. Since MVA has lost multiple host range genes as well as immunomodulatory genes present in the parental VV genome [
5,
23], it is conceivable that the target cell tropism of MVA may differ from that of VV, and MVA might activate instead of suppress the maturation and function of DC, as suggested for VV.
In the current study, we sought to investigate the following questions in murine models: (1) the susceptibility of DCs to MVA infection; (2) the phenotypic and functional alteration of DCs upon MVA infection; (3) whether DCs activate T cells via direct or cross-presentation, and (4) whether DCs are required for in vivo T cell priming following MVA infection. Knowledge gained from these studies can provide important information concerning the mechanisms by which strong immune responses are elicited to MVA-encoded antigens.
Discussion
rVVs are established tools for the development of vaccines against a wide range of infectious and malignant diseases [
15,
16]. Concerns about the safety of VV have led to the use of replication-defective viral vector systems, of which MVA is considered one of the safest and most promising candidates. Currently, at least three types of approaches have been used in MVA-based vaccine development [
11]: (1) administration of MVA as primary vaccine or as a priming vector in the case of rMVAs expressing heterologous antigens; (2) administration of rMVA as booster vaccine following a primary immunization with an alternative vector encoding the same antigen of interest, such as delivered via DNA vaccination; and (3) immunization with DCs infected
ex vivo with rMVA. Immunogenicity studies have shown that, compared to the replicating conventional VV vaccines, MVA vaccines can achieve similar or even higher levels of cellular and humoral responses and protect animals from lethal poxvirus challenge [
6‐
12,
27]. Although the mechanisms by which this highly attenuated non-replicating virus induces strong immune responses is not well-understood, it is conceivable that all these approaches ultimately depend on the capacity of DCs to induce Ag-specific T cell responses – given the central role of DCs in generating innate and adaptive immune responses. This is supported by the recent report that CD11c
+ DCs isolated from Peyer's Patches and spleens of both MVA and VV-immunized mice induced substantial IFN-γ production from an antigen-specific CTL line [
27], although whether cell subsets other than DCs from the same lymphoid tissues could also be loaded with Ags and activate T cells was not tested in the study.
Current knowledge of the interaction of poxviruses with DCs is mostly derived from the studies of VV infection of human DC cultures. These studies have shown that VV abortively infects human MoDCs, inhibits their maturation, phagocytosis, and migration, and reduces the ability to stimulate allogeneic or autologous T cells
in vitro [
19‐
22,
24,
28]. The suggestion that the perturbation of DC physiology observed in tissue culture infections of human DCs reflect a fundamental strategy of viral immune evasion is difficult to reconcile with the substantial humoral and cellular immune responses generated in VV-vaccinated individuals [
29‐
31]. Whether the results from the
in vitro studies accurately depict the
in vivo VV-DC interaction is still an open question. Conflicting data have recently been reported regarding MVA infection of DCs. In one study, MVA was shown to induce immature human MoDC activation, based on the upregulation of co-stimulatory molecules and the secretion of pro-inflammatory cytokines [
18]. Another study showed that MVA-infected immature human MoDCs failed to undergo maturation, but nevertheless were able to present viral Ags to the Ag-specific human CTL line [
12]. Using a murine model, Belyakov et al showed that endogenous DCs isolated from MVA immunized mice stimulated murine CTL lines
in vitro [
27]. However, the CTL lines used in these studies may have less stringent requirement than primary naïve T cells for co-stimulatory signals provided by APCs. Priming of naïve CD8
+ T cells in mice immunized with MVA-infected immature BMDCs was demonstrated by Behboudi et al, even though the infected BMDCs downregulated MHC class I molecules and underwent apoptosis [
32,
33]. It was unclear in these studies whether MVA Ags were presented directly by MVA- infected DCs or indirectly via cross-presentation in these studies.
In the current report, we systematically characterized the interaction of MVA and host DCs using both
in vitro and
in vivo murine models to understand the role of DCs in MVA-induced immunogenicity. Using two different approaches,
in vitro infection of splenocytes and
in vivo infection of experimental mice, we first demonstrated that DCs are preferentially infected by MVA among various cellular components of the adaptive immune system (Fig.
1). This is consistent with our recent report of a similar DC tropism for VV infection in human PBMC [
28]. Thus, despite the extensive loss of parental vaccinia host range genes from the MVA genome, the unique DC tropism of VV is well-conserved in MVA and thus may be of great biological importance for the interactions between poxviruses and the host immune system. The search for the cellular receptor(s) required for poxvirus binding and entry has continued for decades with little success. Our data suggest that the identification of poxvirus receptor(s) on DCs may greatly facilitate the design of DC-targeting Ag delivery system to further improve vaccine immunogenicity.
Infection of DCs by viruses could be either beneficial or harmful for the host immune system, depending on the subsequent phenotypic and functional changes of the infected DCs. For example, influenza virus and dengue virus infections lead to DC maturation and efficient T cell activation [
34,
35]. Consequently, the viruses are quickly cleared with resolution of the acute infection period. On the other hand, three viruses known to induce immunosuppression in humans, CMV, measles, and HIV, have all been documented to infect DCs directly and induce adverse functional alterations [
36‐
39]. Our data support the contention that MVA infection efficiently induces murine DC maturation, manifested by the up-regulation of DC maturation markers (Fig.
3). We also demonstrated that following MVA infection, both BMDCs and splenic DCs secreted substantial amount of pro-inflammatory cytokine IFN-α (Fig.
4). These observations provide a rational explanation of MVA-induced immunogenicity, but are in direct contrast with the reported impairment of human DC maturation and function following VV infection [
19‐
22,
24,
28]. The discrepancy could be attributed to or explained by a number of possibilities. First, it has been shown that, when viral gene expression was prevented by UV light or heat treatment, VV induced rapid human DC activation [
18]. This suggests that the
de novo synthesis of VV-encoded gene products interferes with DC maturation induced by viral binding or internalization per se. It is therefore possible that MVA infection of DCs may lead to distinct immunological consequences due to the deletion of the inhibitory genes from MVA genome. Second, MVA but not VV infection was shown to activate NF-κB in human embryonic kidney cells [
40,
41]. Since NF-κB activation is essential for DC activation and Ag presentation [
40,
41], it is possible that MVA but not VV infection activates NF-κB-dependent pathways in DCs leading to their maturation. Third, in a concurrent study, we observed rapid maturation of murine BMDCs upon VV infection (unpublished data), suggesting possible species-specific and/or DC subtype-specific differences in the consequences of poxvirus infection on DCs.
Comparative studies of the MVA and VV genomes have revealed that many VV immune evasion genes that target host cytokine and chemokine functions have been lost from the MVA genome [
5,
42]. This has been proposed to at least partially compensate for the inability of MVA to replicate and sustain Ag production within hosts. Here, we show that a substantial amount of IFN-α was detected in the supernatant of MVA- infected DCs (both BMDCs and splenic DCs). Interestingly, when VV was used to infect BMDCs and splenic DCs, we failed to detect any IFN-α in the supernatant [see Additional figure 1A]. Furthermore, when known amount of recombinant IFN-α was added into the supernatant samples prior to ELISA, the recombinant IFN-α was no longer detected when added into VV-infected, but not MVA-infected, DC supernatant [see Additional figure 1B]. These data are consistent with the facts that vaccinia B18R gene encodes soluble high-affinity IFN-α/β receptors and the B18R gene has been deleted from the MVA genome [
43‐
45]. The importance of type I IFNs in the protective immunity against poxvirus infection is manifested by the attenuated virulence of B18R-knockout VV in both intranasally and intracranially infected mice [
43,
44]. Since type I IFN promote DC maturation and support CD8
+ T cell and Th1 responses [
46,
47], IFN-α released from MVA-infected cells may contribute to the maturation of DCs following MVA infection and the subsequent induction of innate as well as cellular immune responses.
We observed substantially reduced DC viability following MVA infection (Fig
5A). By 24 h after the infection, over 50% of DCs died. Similar to VV-induced cytopathy of human MoDC [
20], we found that mature murine DCs are more resistant than immature murine DCs to MVA-induced death (Fig.
5B), and that MVA infection induced more DC apoptosis than VV infection (data not shown). The relatively higher susceptibility of DCs for MVA-induced cell death is consistent with our recent report that, compared with VV, MVA infection of human MoDCs leads to accelerated decline of intracellular anti-apoptosis genes Bcl-2 and Bcl-
L [
24]. Deletion of the anti-apoptosis gene SPI-2 from the MVA genome may also contribute to decreased DC viability following infection [
24,
48]. Our
in vivo experiments provide further evidence for rapid DC apoptosis upon rMVA infection. Following intravenous infection of mice with rMVA-GFP, the number of GFP
+ cells in the spleen peaked at 9 h post infection and dropped sharply to background levels by 24 h post infection (data not shown). This result is in agreement with a report by Norbury et al which showed that the numbers of VV-infected cells in the draining lymph nodes following footpad injection decreased by 80% between 6 and 24 h post infection [
49]. Although it is possible that the infected DCs were killed by a virus-specific immune response, as hypothesized by the authors, we think it is unlikely that naïve T cells could be called into action to kill in such a short period of time. Thus the disappearance of infected cells is more likely due to the direct cytopathic effect of the virus infection, as seen in our
in vitro experiments.
It has been controversial whether T cell activation following poxvirus immunization is mainly due to direct Ag presentation by virus-infected DCs, rather than cross-presentation of viral Ags by uninfected DCs that have picked up the infected apoptotic cells [
49]. Rapid maturation of MVA-infected DCs suggests that these DCs can directly activate T cells before they undergo apoptosis. On the other hand, substantial DC apoptosis
in vitro and
in vivo following MVA infection indicates that direct Ag presentation by the infected DCs is likely to diminish quickly and that cross-presentation of viral Ag by uninfected bystander DCs might be involved after the majority of the infected DCs have undergone apoptosis. In support of this hypothesis, we observed efficient phagocytosis of MVA-infected DCs by uninfected DCs (Fig.
6). Furthermore, when MVA-infected β2m
-/- DCs were injected into mice, Ag-specific CTL activity was detected in the spleens, although at a lower level than in mice that received MVA-infected WT DCs. Similarly, Norbury et al. has shown that β2m
-/- kidney cells infected with VV induced CD69 upregulation of Ag-specific CD8
+ T cells
in vivo [
50]. These results strongly suggest that both direct and cross-priming of CD8 T cells occur following poxvirus infection. CD8
+ T cell cross-priming is further supported by the identification of late poxvirus gene-derived CTL epitopes in both human and mice [
51], since the inability of DCs infected by either VV or MVA to express late viral genes prevents the direct presentation of late viral Ags.
We also observed MVA infection of macrophages and B cells, although to a lesser degree comparing to DCs. It remains unclear whether these cells can also present viral Ags and activate T cells in vivo following MVA infection. Using the CD11c-DTR mouse model that allows the conditional ablation of DCs in vivo, we unambiguously demonstrated that DC depletion abrogated the generation of Ag-specific effector T cells (IFN-γ-producing). Therefore, DCs (but not other APCs) play an essential role in the generation of MVA-specific T cells in vivo. The preferential infection of DCs by poxviruses could serve as an efficient Ag delivery system to the most crucial APCs for the induction of strong immune responses.
Authors' contributions
LL designed, performed and analyzed all the experiments, wrote the manuscript. RC created the rMVA-GFP and rMVA-NP, and generated data in Fig.
2B. MBF participated in the design of the experiment and interpretation of data, revised the manuscript critically for important intellectual content. All authors read and approved the final manuscript.