Increased attenuation but decreased immunogenicity by deletion of multiple vaccinia virus immunomodulators

Highlights • Vaccinia virus-derived vaccine vectors are being engineered to improve immunogenicity.• Deleting genes with immunomodulatory function can increase immunogenicity and decrease virulence.• Deletion of N1L, C6L or K7R individually improves immunogenicity, but not in combination.• A virus lacking all three genes induces poorer CD8+ T cell and neutralising antibody responses.


Introduction
The poxvirus family is a diverse group of dsDNA viruses that replicate in the cytoplasm of infected cells. The potential use of this family, particularly the well-characterised vaccinia virus (VACV) strains, as vaccine vectors has been under investigation since the 1980s [1]. VACV has many features that make it attractive for this use; a large capacity for foreign gene expression, relative ease of manipulation, and, importantly, induction of strong innate and memory immune responses, including both cellular and humoral arms [2]. Much research has focused on modified virus Ankara (MVA) and NY-VAC because of their severe attenuation, and, at least for MVA, the inability to replicate in most mammalian cells.
These vectors are safe even in immunocompromised hosts [3,4], which is of particular importance for vaccination against diseases such as HIV-1, malaria and tuberculosis [2]. There is still a need to improve the immunogenicity of these vectors, particularly because large viral doses, repeated vaccination or prime-boost regimes are required to achieve adequate correlates of protection [5].
The host antiviral innate immune response provides a strong selective pressure and, consequently, viruses have evolved a plethora of mechanisms to counteract its effects. Engagement of viral components with innate immune receptors activates transcription factors including nuclear factor kappa-lightchain-enhancer of activated B cells (NF-jB) and interferon (IFN) regulated factors (IRF)-3/7 that coordinate the production of pro-inflammatory cytokines, chemokines and type I IFNs (reviewed in [6,7]). Importantly this pro-inflammatory milieu attracts professional antigen presenting cells to the site of infection, providing an important link to the adaptive immune response and the subsequent establishment of immune memory.
VACV dedicates between one third and one half of its genome to dampening host innate responses (reviewed in [6]). For example, nine intracellular NF-jB inhibitors have been identified and there is evidence that more remain to be discovered [8]. In addition, VACV encodes numerous IRF-3/7 inhibitors and multiple mechanisms to counteract the actions of IFNs [6]. The mechanisms by which innate immunity impacts successful vaccine design remain incompletely understood. Importantly, recent data have demonstrated that the deletion of several VACV immunomodulatory genes individually enhances the immunogenicity of these vectors (reviewed in [6,9]). Examples of such genes include the NF-jB inhibitor N1 [10], the IRF-3/7 inhibitor C6 [11,12] and the dual NF-jB and IRF-3/7 inhibitor K7 ( [13] and Ren et al., unpublished data). Proteins N1 and K7 have Bcl-2 folds [14,15] and C6 is predicted to do so [16] and all are virulence factors [17][18][19].
Some studies have investigated the effect of deleting multiple immunomodulatory genes from VACV vectors but so far have not included an in-depth comparison of single gene deletions in isolation versus deletion of genes in combination [13,20,21]. These comparisons have also not always included challenge experiments, instead measuring aspects of immunological memory that may correlate with immune protection. This study therefore tested whether the immunogenicity of VACV could be improved further by deleting three intracellular innate immunomodulators in combination (encoded by genes N1L, C6L and K7R) from VACV strain Western Reserve (WR). These immunomodulators were selected because their function in innate immunity was known, and because deletion of each of these genes in isolation from VACV WR increased immunogenicity, but decreased virulence. This study shows that deletion of these three genes did not affect VACV growth in vitro, but led to sequential attenuation of the virus in two in vivo models. In challenge models, despite each individual gene deletion enhancing immunogenicity, a virus lacking all three genes was a poorer vaccine, accompanied by inferior memory T cell responses and lower neutralising antibody titres. This illustrates how the design of vaccines for optimal immunogenicity must consider how the degree of attenuation impacts on the induction of immunological memory.

Construction of Western Reserve recombinant viruses
Viruses were constructed using the transient dominant selection method [22] as described [18], using vDN1 [17] as a starting point and plasmids Z11DC6 [18] and pSJH7-DK7 [19] to delete C6L and K7R respectively. Z11 is a pCI-derived plasmid containing the Escherichia coli guanylphosphoribosyl transferase (Ecogpt) gene fused in-frame with the enhanced green fluorescent protein (EGFP) gene under the control of the VACV 7.5 K promoter. Revertant viruses were constructed by replacing the deleted genes in their natural loci using plasmids Z11C6Rev [18] and pSJH7-K7 [19].
The genotype of resolved viruses was analysed by PCR following proteinase K-treatment of infected BSC-1 cells using primers that anneal to the flanking regions of N1L, C6L and K7R [17][18][19]. Infectious virus titres were determined by plaque assay on BSC-1 cells.

Plaque size analysis
BSC-1 cells were inoculated at approximately 50 plaque forming units (p.f.u.) per well of a 6-well plate and stained with crystal violet 3 days later [18]. The sizes of 20 plaques per well were measured using Axiovision acquisition software and a Zeiss AxioVert. A1 inverted microscope as described [25].

Murine intranasal and intradermal models of infection
Female BALB/c mice (n = 5, 6-8 weeks old) were infected intranasally (i.n.) with 5 Â 10 3 p.f.u. of purified VACV strains. VACV was purified from cytoplasmic extracts of infected cells by two rounds of sedimentation through 36% (w/v) sucrose at 32,900g for 80 min. Virus was resuspended in 10 mM Tris-HCl pH 9. Virus used for infections was diluted in phosphate-buffered saline containing 1% bovine serum albumin and the titre of the diluted virus that was used to infect mice was determined by plaque assay on the day of infection. Mice were monitored daily to record body weight and signs of illness as described [26,27]. Female C57BL/6 mice (n = 5, 6-8 weeks old) were inoculated intradermally (i.d.) in both ear pinnae with 10 4 p.f.u. and the resulting lesions were measured daily as described [28,29]. For the challenge experiments, mice that had been inoculated i.n. were challenged 6 weeks later and mice that had been inoculated i.d. were challenged 4 weeks later, i.n., with 5 Â 10 6 p.f.u. of wild-type VACV WR.

Intracellular cytokine staining
Splenocytes were prepared as described [10] and incubated for 4 h with a C57BL/6-specific CD8 + VACV peptide B8 20-27 [30] or a negative control CD8 + VACV peptide specific for BALB/c mice, E3 140-148 [31] at a final concentration of 0.1 lg/ml. After 1 h Golgi stop (BD Biosciences) was added and the cells were incubated for a further 3 h. Cells were then stained for CD8 and either IFNc or TNFa and analysed by flow cytometry [10].

51 Cr release cytotoxic assay
Cytotoxic T lymphocyte (CTL) activity was assayed with a standard 51 Cr-release assay using VACV-infected EL-4 cells as targets, as described [32]. The percentage of specific 51 Cr-release was calculated as specific lysis = [(experimental release À spontaneous release)/(total detergent release À spontaneous release)] Â 100. The spontaneous release values were always <5% of total lysis.

IMV plaque reduction neutralisation assay
The neutralising titre of anti-VACV antibodies was calculated by plaque assay on BSC-1 cells as described [10]. Neutralisation dose 50 (ND 50 ) values represent the reciprocal of the serum dilution giving 50% reduction in plaque number compared with virus incubated without serum.

Serum antibody titration by ELISA
The binding of serum antibodies to VACV-specific epitopes was measured by enzyme-linked immunosorbent assay (ELISA) using plates coated with lysates of VACV strain WR-infected cells that had been treated with ultraviolet light and psoralen to inactivate VACV infectivity as described [33,34]. Plates coated with bovine serum albumin were used as a negative control. IgG end-point titres were defined as the reciprocal serum dilutions giving twice the average optical density values obtained with bovine serum albumin.

Statistical analysis
Data were analysed using an un-paired Student's t-test, with Welch's correction where appropriate, or a Mann-Whitney test as indicated. Statistical significance is expressed as follows: * P < 0.05, ** P < 0.01, *** P < 0.001.

Ethics statement
This work was carried out in accordance with regulations of The Animals (Scientific Procedures) Act 1986. All procedures were approved by the United Kingdom Home Office and carried out under the Home Office project licence PPL 70/7116.

Construction of a VACV strain WR lacking the innate immunomodulatory genes N1L, C6L and K7R
To construct a virus lacking N1L, C6L and K7R, C6L was deleted from a virus already lacking N1L (vDN1, [17]) by transient dominant selection (see methods) yielding vDD, followed by the removal of K7R yielding vDDD. As controls, a revertant virus where the deleted gene was re-inserted back into its natural locus was constructed at each stage (Fig. 1a). Deletion of C6L and K7R was confirmed by PCR analysis of proteinase K-treated lysates of infected BSC-1 cells using primers specific for these genes in addition to N1L and A49R as a control (Fig. 1b). The phenotype of the resulting recombinant viruses was confirmed at the protein level by immunoblotting of lysates from infected BSC-1 cells using antisera against N1, C6 and K7, as well as a monoclonal antibody against VACV protein D8 as an infection control (Fig. 1c). Proteins N1, K7 and C6 are each non-essential for virus replication or spread in cell culture [17][18][19] and measurement of the plaque size of mutants lacking 2 or 3 of these genes confirmed that deletion of these genes in combination did not affect viral spread in BSC-1 cells (Fig. 1d).

Deletion of N1L, C6L and K7R in combination attenuates the virus sequentially in murine intradermal and intranasal models of infection
The N1L [17,35], C6L [18] and K7R [19] single deletion viruses are attenuated in both the i.d. and i.n. models of murine infection, highlighting their importance as virulence factors. When compared side-by-side in the i.d. model, the level of attenuation was found to be similar amongst the three single deletion viruses (Suppl. Fig. 1a). In contrast, a virus lacking both N1 and C6 (vDD) was significantly more attenuated than vDN1, and a virus lacking all three immunodulators (vDDD) was attenuated further still (Fig. 2a), indicating that the roles of these innate inhibitors in vivo is nonredundant. The double deletion revertant (vDDR) and the triple gene deletion revertant (vDDDR) control viruses behaved as expected, demonstrating that the observed attenuation phenotypes were due to specific genes deletions and not mutations elsewhere in the viral genome (Suppl. Fig. 1b). Similar results were obtained in the i.n. model of infection where deletion of C6 and K7 again led to sequential attenuation of the virus, as indicated by reduced weight loss (Fig. 2b) and fewer signs of illness (Fig. 2c).

Deletion of N1L, C6L and K7R in combination does not enhance vaccine potency
To determine whether the double and triple gene deletion viruses have improved vaccine potency compared with the N1 single deletion virus, mice were vaccinated i.d. and challenged i.n. one month later with a lethal dose of wild-type VACV WR. As reported [10], single deletion of N1 provided the mice with better protection against challenge, indicated by significantly less weight loss over a period of 10 days (Fig. 3a). However, vDD provided less protection than both vDN1 and vWT (although this did not reach statistical significance when compared to the WT virus on any one day) and vDDD provided significantly less protection again (Fig. 3a). Similar results were observed following challenge of mice that were vaccinated i.n. (Fig. 3b). When compared head-to-head each of the single deletion viruses enhanced vaccine potency to a similar degree (Suppl. Fig. 2a) and the revertant control viruses behaved as expected in challenge experiments (Suppl. Fig. 2b).

The double and triple gene deletion viruses induce lower CTL responses
To understand why vaccination with vDD or vDDD afforded less protection than vDN1, CD8 + T cells responses one month post-i.d. vaccination were analysed. To measure the cytolytic activity of VACV-specific T cells a chromium release cytotoxicity assay was performed. The specific cytolytic activity of T cells from vDN1-vaccinated mice was significantly higher than that from vWT-vaccinated animals (Fig. 4a), corroborating recently published findings [10]. Conversely, significantly lower VACV-specific cytolytic activity of T cells was measured in vDDor vDDD-vacci nated mice. Interestingly, no significant difference in cytolytic activity was observed between vDD and vDDD.
The release of cytokines by splenic CD8 + T cells that were stimulated ex vivo with VACV peptides was also analysed by intracellular cytokine staining. In agreement with data published recently [10], splenic CD8 + T cells from vDN1-vaccinated mice secreted enhanced levels of IFNc (Fig. 4b) and TNFa (Fig. 4c) following stimulation than cells from vWT-vaccinated mice. In contrast, splenic CD8 + T cells from both vDD-and vDDD-vaccinated mice secreted significantly less cytokines. Again, no significant difference was observed between the double and triple gene deletion viruses.

Double and triple gene deletion viruses induce lower neutralising antibody titres
The poorer immunogenicity of the double and triple gene deletion viruses might also be due to altered antibody responses. To investigate this possibility, sera were collected from mice one month post-i.d. vaccination and the VACV-specific antibody titres and VACV-specific neutralising titres were measured by ELISA and plaque reduction neutralisation respectively. The titre of VACV-specific antibodies measured by ELISA was not different between the groups of vaccinated mice and good antibody levels were induced in all cases (Fig. 5a). The same was not true, however, when the titre of neutralising serum antibodies was measured by plaque reduction neutralisation assay. In this case the ND 50 of sera from both the vDD-and vDDD-vaccinated mice was significantly lower than that from both vWT and vDN1-vaccinated mice, although there was no significant difference between the double and triple gene deletion viruses (Fig. 5b).

Discussion
Inducing a robust innate immune response is an important step to designing an immunogenic vaccine and this is often achieved by the addition of a vaccine adjuvant. However, a full understanding of how the innate immune system impacts immune memory is lacking and would greatly enhance our ability to rationally design vaccines with enhanced immunogenicity profiles. VACV-based vectors are popular candidates, however their genomes still encode proteins with a known role in dampening the host innate immune response, which may negatively impact their potential use as vaccine vectors. Indeed, the deletion of numerous VACV immunomodulatory genes has been shown to enhance immunogenicity including the chemokine-binding protein A41 [21,32], the IL-1bbinding protein B15 [21,36], the inhibitor of MHC class II antigen presentation A35 [37], the IL-18-binding protein C12 [38], the type I (B18) and type II (B8) IFN-binding proteins [39], the IRF3/7 inhibitor C6 [11,12], the NF-kB inhibitor and anti-apoptotic protein N1 [10], the dual NF-kB and IRF3/7 inhibitor K7 ( [13] and Ren et al., unpublished results) and the TLR signalling inhibitor A46 [40]. Data presented here demonstrate that deletion of three of these genes (N1L, C6L and K7R) in combination from VACV WR did not further enhance the immunogenicity and in fact provided poorer protection than the single gene deletion viruses. These data highlight that in the context of a replicating vaccine vector there is a fine balance between viral attenuation and immunogenic potential.
Given that vDD and vDDD are sequentially more attenuated than vDN1 (Fig. 2) they may have generated lower antigen levels and/or been cleared more quickly by the host immune system and hence induced a weaker adaptive response. The importance of antigen availability for the formation of CD8 + T cell-dendritic cell interaction kinetics and the ensuing memory response was demonstrated recently [41]. Of further interest, the inferior CD8 + T cell responses and neutralising antibody titres observed with vDD and vDDD were not significantly different between these two viruses, however vDDD provided worse protection than vDD in the challenge studies (Fig. 3). These data may indicate that other aspects of immune memory such as CD4 + T cells, or the recently identified memory NK cells [42] may play an important role in determining vaccine efficacy of VACV. Whether equivalent protection to WT or vDN1 vaccination could be achieved by increasing the vaccination dose of vDD and vDDD warrants further investigation.
A recent study using an MVA vector expressing HIV-1 antigens (MVA-B) in the context of a DNA prime/MVA boost regime found that the deletion of C6L and K7R in combination enhanced the magnitude and quality of HIV-1-specific CD4 + and CD8 + T cell responses in mice, as well as Env antibody levels compared to the parental MVA-B vector and MVA-B lacking C6L alone [13]. These data demonstrate that in the context of a non-replicating strain of VACV, the deletion of more than one immunomodulatory gene may be beneficial. What is not clear from this study, however, is whether these memory responses would have been achieved with the deletion of K7R alone, because an MVA-B K7R single deletion was not included. Furthermore, whether these correlates of protection will translate to protection in an in vivo challenge model remains to be determined. Another recent study, also based on an MVA vector expressing HIV-1 antigens, found enhanced HIVspecific CD4 + and CD8 + T cell responses as well as Env-specific antibody responses in rhesus macaques with a vector lacking 4 immunomodulatory proteins [20]. However again this vector was not compared to single gene deletions and protection was not determined by a challenge experiment.
Mounting evidence indicates an important role for type I IFN [43,44] in vaccine immunogenicity and each of the proteins selected for this study inhibits type I IFN production. Furthermore, the importance of NF-jB in generating robust memory immune responses was demonstrated recently for N1, an inhibitor of NF-jB and apoptosis. By vaccinating mice with viruses encoding N1 mutants that were competent for only one of these functions it was demonstrated that the anti-NF-jB activity of N1 was important for the enhanced protection observed with vDN1, with no apparent contribution of its anti-apoptotic function [35]. These types of studies demonstrate how viruses lacking innate immunomodulators can be utilised as tools to further our understanding of the relationship between innate immunity and immune memory, which will be important for future vaccine design.  The neutralisation capacity of antibodies in the serum was assessed by plaque-reduction neutralisation against VACV strain WR intracellular mature virus. The median value for each population is represented by a horizontal black bar. Statistical analyses were performed using the Student's t-test. ** P < 0.01, ns: not significant. Experiments were conducted at least 3 times and representative data are shown.