Main

A major challenge for the international biomedical community is to develop vaccines for chronic diseases caused by intracellular pathogens. AIDS, malaria, tuberculosis and hepatitis are all established through chronic intracellular infections, and protection against the pathogens causing these diseases requires vaccines that elicit broad cellular immunity1. Vaccines based on either recombinant proteins or killed whole pathogens, although safe, typically induce weak cellular immunity. In contrast, vaccines based on live attenuated forms of a pathogen may have enhanced immunogenicity and elicit responses of increased breadth and durability, but they present safety risks, particularly among immunocompromised individuals. To address this dilemma in vaccine development, we sought to derive vaccine platforms based on whole organisms that elicited strong cellular immunity, yet retained the safety profile of killed or subunit vaccines.

Killed but metabolically active (KBMA) vaccines are based on bacterial nucleotide excision repair mutants, and are inactivated by photochemical treatment with a synthetic highly reactive psoralen known as S-59. Psoralens form covalent monoadducts and crosslinks with pyrimidine bases of DNA and RNA upon illumination with long-wavelength ultraviolet (UVA) light2. The primary repair pathway for psoralen crosslinks is nucleotide excision repair3, a process initiated by the ABC excinuclease, a nuclease encoded by the ultraviolet light response (uvr) genes. Deletion of any one of the three uvr genes renders bacteria exquisitely sensitive to ultraviolet light–induced DNA damage and to psoralen-induced crosslinks3. The KBMA vaccine approach is based on the hypothesis that the increased sensitivity of uvr mutant strains to photochemical inactivation is correlated with randomly distributed, infrequent psoralen crosslinks, leaving intact the ability of a population of inactivated bacteria to express its genes, while preventing productive growth and the ability to cause disease in the immunized host. Thus, the early events in the natural infection are unaffected, which facilitates the induction of an innate and adaptive immune response comparable to that induced by the live pathogen.

We used Listeria monocytogenes as a model intracellular pathogen to test this vaccine concept. L. monocytogenes is a Gram-positive facultative intracellular pathogen of animals and humans that has been studied for decades as a model to understand basic aspects of cellular immunology and microbial pathogenesis4,5,6. L. monocytogenes is a powerful inducer of innate immunity as it encounters cellular effectors, resulting in the release of T helper type 1 (TH1) polarizing cytokines6. L. monocytogenes taken up by macrophages and dendritic cells (DCs) escape the phagolysosome and access the cytosol in a process mediated by listeriolysin O (LLO)7,8. In response to entry of bacteria into the cytosol, infected host cells produce interferon (IFN)-β9, which, in contrast to viruses, may augment the pathogenicity of L. monocytogenes10,11,12. In the mouse listeriosis model, the liver and spleen are the primary targets of infection4. Within these target organs, bacteria spread intercellularly and remain largely sequestered from the extracellular milieu. Acquired immunity is entirely cell mediated, and antibodies have no role in protection6.

A well-known dichotomy in the mouse listeriosis model is that, whereas wild-type L. monocytogenes primes potent effector and memory T-cell responses that protect mice against bacterial challenge, vaccination with LLO-deleted L. monocytogenes (Lm Δhly) or with heat-killed L. monocytogenes does not elicit functional T cells or induce protective immunity13. This difference is evidence of the requirement for de novo gene expression and cytosolic access upon uptake of L. monocytogenes by host cells in vaccinated animals. Here, we show that psoralen-inactivated L. monocytogenes nucleotide excision repair mutants (Lm ΔuvrAB) retained essential properties of live bacteria, and as a result elicited functional T cells and long-term protective immunity in vaccinated mice that correlated with efficacy in mouse models of infectious disease and cancer.

Results

Lm ΔuvrAB psoralen inactivation and metabolic activity

The ActA protein (encoded by actA) is required for cell-to-cell spread of L. monocytogenes8. Although live Lm ΔactA mutants are attenuated 1,000-fold in mouse virulence, the potency of vaccines derived from this strain is not affected14. We constructed nucleotide excision repair mutants by removing the uvrAB genes in attenuated L. monocytogenes–based vaccines in which actA was deleted and which encoded the chicken ovalbumin (OVA) model antigen (Lm ΔactA/ΔuvrAB-OVA). We determined the relative sensitivity to photochemical inactivation of Lm ΔactA/ΔuvrAB-OVA and its parental Lm ΔactA-OVA strain over a range of S-59 psoralen concentrations, using conditions established previously for Gram-positive bacteria2. The lowest S-59 concentration that resulted in full inactivation of the culture was established for each bacterial strain. Lm ΔactA/ΔuvrAB-OVA was ≥8 logs more sensitive at low S-59 concentration to photochemical inactivation than parental Lm ΔactA-OVA (Fig. 1a). We also constructed a nucleotide excision repair mutant strain of B. anthracis (Sterne) by deleting uvrAB genes through allelic exchange (Ba ΔuvrAB). Ba ΔuvrAB was ≥ 8 logs more sensitive to photochemical inactivation at 50 nM S-59 as compared to its parental strain (Supplementary Fig. 1 online). We measured the number of residual viable bacteria after treating 14 individual cultures (∼1011 colony-forming units (CFU)) of Lm ΔactA/ΔuvrAB-OVA and Lm ΔactA-OVA. On average, 1 colony (±3) grew from Lm ΔactA/ΔuvrAB-OVA cultures treated with 200 nM S-59, and 56 colonies (±61) grew from cultures of parental Lm ΔactA-OVA treated with 2,500 nM S-59, indicating incomplete killing of the DNA repair–competent parental strain despite using 12.5 times more psoralen. These S-59 concentrations were the standard photochemical inactivation conditions used for subsequent experiments, and therefore resulted in at least 10-log inactivation of Lm ΔactA/ΔuvrAB-OVA cultures, and approximately 9-log inactivation of cultures of the parental Lm ΔactA-OVA strain with intact nucleotide excision repair.

Figure 1: L. monocytogenes nucleotide excision repair mutants are exquisitely sensitive to photochemical inactivation and are metabolically active.
figure 1

(a) Viability of Lm ΔactA-OVA and Lm ΔactA/ΔuvrAB-OVA after treatment with the indicated S-59 concentration. A representative of at least five experiments is shown. (b) Immunofluorescence photomicrographs show that S-59–inactivated nucleotide excision repair mutant vaccine strains are metabolically active and form chains after incubation in broth culture; L. monocytogenes (green) and chromosomal DNA (blue). A Lm ΔsecA2 mutant strain is shown for comparison and shows productive L. monocytogenes vegetative growth with chromosome segregation (right). The Lm ΔsecA2 mutant strain grows in chains due to lack of secretion of peptidoglycan autolysins46. (c) Immunofluorescence photomicrographs show that psoralen-inactivated Lm ΔactA/ΔuvrAB-OVA escape from the phagolysosome of infected DC2.4 cells. Cytosolic L. monocytogenes is indicated by colocalization of bacteria (green) with actin (red); host-cell DNA (blue). Scale bar, 10 μm. Representative fields are shown.

Full size image

Next, we directly measured the psoralen adduct frequency in the DNA genomes of photochemically treated vaccines over a concentration range of 14C-labeled S-59 (Table 1). Under the standard S-59 psoralen/UVA light photochemical inactivation (S-59/UVA) conditions, Lm ΔactA/ΔuvrAB-OVA genomes had a significantly lower frequency of psoralen adducts compared to parental strain genomes, Lm ΔactA-OVA (28 ± 5 versus 178 ± 6 psoralen adducts/L. monocytogenes genome, respectively; P = 0.00075). Metabolic labeling showed substantial levels of new protein synthesis and secretion after photochemical inactivation from S-59/UVA Lm ΔactA/ΔuvrAB-OVA, but not from S-59/UVA Lm ΔactA-OVA (Supplementary Fig. 2 online). The observed differences in the level of protein synthesis between S-59/UVA Lm ΔactA/ΔuvrAB-OVA and live cultures may reflect a stress response15 to photochemical treatment. Consistent with the observation of continued metabolic activity after photochemical treatment, S-59/UVA Lm ΔactA/ΔuvrAB-OVA, but not S-59/UVA Lm ΔactA-OVA, formed filamentous chains upon incubation in brain-heart infusion broth after photochemical treatment, indicative of cell-wall synthesis without the ability to septate (Fig. 1b).

Table 1 Psoralen adduct frequency in S-59/UVA-treated L. monocytogenes strains
Full size table

We next sought to determine whether this differential metabolic activity could be observed during infection of cultured cells. Expression of LLO is essential for escape of L. monocytogenes from the phagolysosome of the infected host cell, and entry into the cytosol is a requisite step for induction of protective immunity in vaccinated mice6,8,16. We assessed the capacity for S-59/UVA Lm ΔuvrAB to escape from the phagolysosome of infected DC 2.4 cells, a mouse dendritic cell line17, by fluorescence microscopy. We measured the frequency of cytosolic L. monocytogenes by colocalization of bacteria with host-cell actin. A single L. monocytogenes protein, ActA, mediates the polymerization of host actin filaments on the bacterial cell surface, a necessary step for cytosolic motility8. All L. monocytogenes strains used for fluorescence microscopy contained native actA, in order to visualize cytosolic access by phytochemically treated bacteria through colocalization with polymerized host-cell actin. At 5 h after infection, S-59/UVA Lm ΔuvrAB, but not S-59/UVA wild-type L. monocytogenes, formed multiple short bacteria chains that were enshrouded by host-cell actin, indicating that the S-59/UVA nucleotide excision repair mutants accessed the host-cell cytosol (Fig. 1c). Unlike live Lm ΔuvrAB, S-59/UVA Lm ΔuvrAB were not able to septate and did not form the signature 'comet tail' indicative of cytosolic motility18. Quantitative fluorescence microscopy of cytosolic L. monocytogenes 5 h after infection showed that 61% of live bacteria (521 of 855) and 47% of S-59/UVA Lm ΔuvrAB (493 of 1,047), respectively, escaped the phagolysosome (combined data from five individual experiments; data not shown). In contrast, we detected less than 5% of S-59/UVA wild-type Lm or Lm Δhly (LLO−) in the cytosol (Fig. 1c).

The host cell recognizes bacterial products in the cytosol, leading to production of IFN-β9. Synthesis of IFN-β distinguishes between bacteria confined to the vacuole and bacteria that can propagate in the cytosol. Rapid induction of mRNA encoding IFN-β was observed in primary C57BL/6 mouse bone marrow–derived macrophages infected with both live L. monocytogenes strains and with S-59/UVA Lm ΔactA/ΔuvrAB-OVA. As shown previously9, we did not observe induction of mRNA encoding IFN-β in macrophages infected with live Lm Δhly. Together, these data show that nucleotide excision repair mutant vaccines are exquisitely sensitive to psoralen-mediated photochemical inactivation and contain a significantly (P = 0.00075) reduced frequency of adducts in the genome; thus, a larger proportion of the genome is available for transcription. S-59/UVA Lm ΔactA/ΔuvrAB-OVA, although unable to grow productively, retains sufficient metabolic activity to access the cytosol of infected cells, a necessary step for induction of CD8+ T cells and protective immunity.

S-59/UVA Lm ΔuvrAB MHC class I antigen presentation

Proteins secreted from viable cytosolic L. monocytogenes can be efficiently processed by the major histocompatibility complex (MHC) class I pathway19. Using OVA-specific CD8+ T-cell hybridoma cells (B3Z), we determined the relative capacity of psoralen-inactivated Lm ΔactA/ΔuvrAB-OVA to directly load the immunodominant MHC class I OVA epitope SIINFEKL on H-2Kb within infected DC2.4 cells. Notably, S-59/UVA Lm ΔactA/ΔuvrAB-OVA, but not parental S-59/UVA Lm ΔactA-OVA, maintained its capacity to load antigen into the MHC class I pathway independent of its ability to form colonies on agar media (Fig. 2a). There was a window defined by an S-59 psoralen concentration range between 100 and 400 nM in which activation of the T-cell hybridoma was not correlated with ability of S-59/UVA Lm ΔactA/ΔuvrAB-OVA to form colonies on agar media. In contrast, MHC class I presentation could not be unlinked from bacterial growth using the Lm ΔactA-OVA parental strain.

Figure 2: Photochemically inactivated L. monocytogenes nucleotide excision repair mutant vaccines induce class I reponses.
figure 2

(a) S-59/UVA Lm ΔactA/ΔuvrAB-OVA load antigen into the MHC class I pathway of antigen-presenting cells. DC2.4 cells were infected with L. monocytogenes vaccines treated with varying concentrations of S-59 psoralen as indicated and illuminated with 6 J/cm2 UVA light. H-2Kb–restricted presentation of SIINFEKL epitope measured by B3Z activation (filled diamond; represented as percent of live bacteria) as a function of L. monocytogenes viability (open diamond) is shown. A representative of at least three experiments is shown. (b) Bone marrow–derived DCs infected with S-59/UVA Lm ΔactAΔuvrAB-OVA prime functional CD8+ T cells in mice. Splenic OVA-specific CD8+ T cells were measured by TNF-α and IFN-γ cytokine flow cytometry and tetramer analysis. Data are presented as mean ± s.d. Representative dot plots for three mice per group are shown. This experiment was repeated twice with similar results.

Full size image

Next, we tested whether autologous immature bone marrow–derived DCs infected ex vivo with S-59/UVA Lm ΔactA/ΔuvrAB-OVA could prime functional OVA-specific CD8+ T cells in vaccinated mice. We observed functional cytokine-secreting OVA-specific CD8+ T-cell responses measured 1 week after vaccination only in mice immunized with S-59/UVA Lm ΔactA/ΔuvrAB-OVA–infected DCs, resulting in about 2% OVA-specific CD8+ T cells (Fig. 2b). In contrast, we did not observe significant CD8+ T cell responses in mice vaccinated with DCs infected with either heat-killed Lm ΔactA/ΔuvrAB-OVA or parental S-59/UVA Lm ΔactA-OVA.

S-59/UVA Lm ΔuvrAB protects against viral challenge

Live attenuated recombinant L. monocytogenes vaccines induce robust CD8+ T-cell responses specific for a secreted heterologous antigen, conferring preventative or therapeutic immunity in animal models of infectious disease and cancer20,21. We immunized C57BL/6 mice with selected S-59/UVA Lm-OVA vaccines to determine whether abrogation of nucleotide excision repair was also required to prime CD8+ T cells in vivo (Fig. 3a). A single vaccination with S-59/UVA Lm ΔactA/ΔuvrAB-OVA induced 2% OVA-specific CD8+ T cells and three immunizations given on 3 consecutive days induced 3.8% OVA-specific CD8+ T cells. In contrast, vaccination with S-59/UVA Lm Δact-OVA or heat-killed Lm ΔactA/ΔuvrAB-OVA did not result in any significant induction of OVA-specific CD8+ T cells, regardless of immunization regimen. These CD8+ T-cell responses correlated with protection against challenge with recombinant vaccinia virus-OVA (rVV-OVA) in mice immunized with S-59/UVA Lm ΔactA/ΔuvrAB-OVA (Fig. 3b). We observed a 4-log reduction in rVV-OVA titer in the ovaries of challenged mice immunized with S-59/UVA Lm ΔactA/ΔuvrAB-OVA, as compared to challenged naive mice. In contrast, we observed only a 1-log reduction in virus titer in mice immunized with the S-59/UVA Lm ΔactA-OVA parent strain. The level of protective immunity against vaccinia challenge was not significantly different between mice immunized with S-59/UVA Lm ΔactA/ΔuvrAB-OVA or with live Lm ΔactA/ΔuvrAB-OVA.

Figure 3: S-59/UVA Lm ΔactAΔuvrAB-OVA elicit functional CD8+ T cells in vaccinated mice and protect against a subsequent viral challenge.
figure 3

(a) OVA-specific CD8+ T-cell responses determined 1 week after immunization. (b) Protection against rVV-OVA challenge. One-way ANOVA, P = 0.0273; Kruskal-Wallis nonparametric test with Dunn post test. HK, heat-killed. (c) Primary T-cell responses induced by S-59/UVA Lm ΔactA/ΔuvrAB-OVA can be boosted. Primary OVA-specific CD8+ T-cell responses were determined 7 d after immunization with the indicated strains. The secondary T-cell response was determined 6 d after boost immunization on day 14. Data are presented as mean ± s.e.m. All data represent at least three independent experiments.

Full size image

We then evaluated the comparative immunogenicity of S-59/UVA Lm-OVA vaccines after a conventional prime and boost immunization dosing regimen (Fig. 3c). S-59/UVA Lm ΔactA/ΔuvrAB-OVA, but not S-59/UVA Lm ΔactA-OVA, elicited a robust OVA-specific CD8+ T cells with a prime-boost immunization regimen separated by 14 d. The primary and secondary OVA-specific CD8+ T-cell responses of live Lm ΔactA/ΔuvrAB-OVA and live Lm ΔactA-OVA were comparable, showing that deletion of the uvrAB genes does not affect the potency of live L. monocytogenes vaccines. The data also show that the boosted OVA-specific CD8+ T-cell responses in mice immunized with live or S-59/UVA Lm ΔactA/ΔuvrAB-OVA vaccines were similar.

S-59/UVA LM ΔuvrAB therapeutic antihumor efficacy

We evaluated the potency of the S-59/UVA Lm ΔuvrAB platform in a rigorous setting of CT-26 tumor-bearing mice. Therapeutic efficacy in this model requires breaking of self-tolerance to the H-2Ld immunodominant epitope AH1 from the gp70 endogenous tumor antigen22,23. We constructed isogenic vaccines that differed only in nucleotide excision repair capacity and that expressed and secreted the altered CD8+ T-cell epitope AH1-A5 (ref. 22). Therapeutic vaccination of mice with S-59/UVA Lm ΔactA/ΔuvrAB-AH1-A5 given on 3 consecutive days, 3 d after intravenous implantation of CT-26 tumor cells, resulted in reduction of lung tumor burden (Fig. 4a) that correlated with a significant prolongation of life (Fig. 4b). Therapeutic antitumor efficacy resulted from breaking tumor antigen self-tolerance as shown by the induction of cytolytic AH1 epitope–specific CD8+ T cells in vivo (Supplementary Fig. 3 and Supplementary Methods online). In contrast, immunization of tumor-bearing mice with S-59/UVA parental Lm ΔactA-AH1-A5 did not elicit significant AHI-specific CD8+ T cells or provide therapeutic benefit.

Figure 4: Nucleotide excision repair mutation confers therapeutic efficacy of S-59/UVA Lm ΔactAΔuvrAB-AH1-A5 vaccines in tumor-bearing animals.
figure 4

(a) Lung tumor metastases 20 d after CT26 tumor cell seeding. (b) Survival; mice were killed when moribund. n = 10 per group (P = 0.0041; one-way ANOVA).

Full size image

S-59/UVA LM ΔuvrAB protects against bacterial challenge

To test the possibility of using S-59/UVA nucleotide excision repair mutants of pathogens as a vaccine against the cognate wild-type virulent organism, we assessed the protective immunity elicited in mice vaccinated with S-59/UVA Lm ΔactA/ΔuvrAB. S-59/UVA Lm ΔactA/ΔuvrAB, but not heat-killed Lm ΔuvrAB or S-59/UVA Lm ΔactA, induced a potent LLO190–201-specific CD4+ T-cell response (Fig. 5a). We evaluated protective immunity by challenging C57BL/6 mice with twice the lethal dose in 50% of subjects (LD50) dose of wild-type L. monocytogenes 90 d after a prime and boost immunization regimen. We enumerated bacteria in the spleen 3 d after wild-type L. monocytogenes challenge. We observed comparable protective immunity—as shown by an approximate 6-log reduction in spleen colonization after wild-type L. monocytogenes challenge relative to control vaccinated mice—in mice vaccinated with live or S-59/UVA Lm ΔactA/ΔuvrAB (Fig. 5b). Notably, Lm ΔactA/ΔuvrAB inactivated with 2,500 nM S-59, the psoralen concentration required to inactivate the parental Lm ΔactA strain, did not confer protection in vaccinated mice. As expected16,24, protective immunity was not induced when mice were vaccinated with either heat-killed Lm ΔactA/ΔuvrAB or Lm Δhly (Fig. 5b). Together, these data show that S-59/UVA Lm ΔactA/ΔuvrAB–based vaccines retain the biologic properties of live Lm ΔactA/ΔuvrAB vaccines at a level sufficient to elicit protection against lethal wild-type bacterial challenge.

Figure 5: S-59/UVA Lm ΔactA/ΔuvrAB vaccines protect immunized mice against wild-type L. monocytogenes challenge.
figure 5

(a) Splenic LLO190–201-specific CD4+ T-cell responses from C57BL/6 mice 7 d after a single immunization with the indicated vaccines. (b) CFU per spleen from mice challenged with two times LD50 wild-type L. monocytogenes 90 d after a prime-boost immunization regimen with the indicated vaccines. HK, heat-killed. Bacterial levels for three individual mice and the mean for each experiment are shown (one-way ANOVA, P = 0.0273; Kruskal-Wallis nonparametric test with Dunn post test).

Full size image

Discussion

We describe proof-of-concept studies for a new class of vaccines, termed 'killed but metabolically active', or KBMA, that is based on photochemically inactivated nucleotide excision repair mutant microbes. KBMA vaccines have two separate potential applications: (i) recombinant L. monocytogenes expressing heterologous antigens relevant to infectious disease or cancer; and (ii) vaccines for bacterial diseases based on a modified form of the pathogen itself. This later case is theoretically advantageous in instances where the antigenic correlates of immune protection are unknown or poorly defined, such as for Mycobacterium tuberculosis. KBMA vaccines retain the biologic properties of the parental organism, and thus stimulate relevant innate and adaptive immune responses. KBMA vaccines cannot grow productively and do not present the infectious risk inherent in live vaccines.

Deletion of nucleotide excision repair should, in principle, make a single psoralen crosslink a lethal event to a bacterium by forming an absolute block to chromosome replication. Quantitative assessment of psoralen adduct frequency supports this hypothesis. According to the Poisson distribution, an average frequency of 23 crosslinks per genome (3 × 106 bp), or one crosslink per 130,000 bp, would be required to kill 10 logs of L. monocytogenes, assuming that a single psoralen crosslink is a lethal event. The average number of psoralen adducts per genome required to inactivate 10 logs of the nucleotide excision repair mutant Lm ΔuvrAB was 28, approximately one psoralen adduct for every 107,000 bp, or about every 100 genes, considerably close to the frequency predicted by Poisson distribution. In contrast, the number of adducts per genome to inactivate 9 logs of its parental strain was 178, corresponding to a psoralen adduct every 17,000 bp, or about every 17 genes. Moreover, our psoralen assay measures both crosslinks and monoadducts, and thus the number of crosslinks may be less than 28 per genome. These data also suggest that deletion of nucleotide excision repair by itself is sufficient, and that deletion of recA or other DNA repair genes would not augment the immunogenicity of the KBMA vaccine by increasing psoralen sensitivity. Notably, higher psoralen concentrations were required to inactivate nucleotide excision repair mutant bacteria when at mid-log phase of growth (data not shown). We hypothesize that this observation resulted from an average chromosome number per bacteria in the mid-log phase cultures that exceeded one, thus facilitating recA-mediated homologous recombination repair. In turn, vaccines prepared under these conditions were poorly immunogenic (data not shown).

Formation of bacterial chains by S-59/UVA Lm ΔactA/ΔuvrAB but not its parental S-59/UVA Lm ΔactA strain after incubation in culture broth provided evidence of the capacity to synthesize macromolecules, including proteins and cell-wall constituents, and strongly distinguished the potential metabolic activity of the two S-59/UVA–treated bacterial populations. Chaining of S-59/UVA Lm ΔactA/ΔuvrAB is similar to the phenotype observed with certain topoisomerase mutants, in which chromosomes are not resolved after replication, producing chains of anucleate cells25, and it suggests that chromosomes after photochemical treatment are physically constrained by psoralen crosslinks, preventing cell division of nucleotide excision repair mutants.

Photochemically inactivated Lm ΔactA/ΔuvrAB–based vaccines may prove to be a useful tool for ex vivo DC-based therapies. Immature DCs efficiently take up L. monocytogenes, and in response undergo activation and maturation26,27. Here, we used S-59/UVA Lm ΔactA/ΔuvrAB-OVA vaccines to activate, mature and load a heterologous encoded antigen onto DCs, which resulted in direct MHC class I presentation in vitro, and primed a potent CD8+ T-cell response in vaccinated mice. This approach avoids the complications related to using live bacteria in mammalian cell culture, limits the use of expensive cytokines and DC maturation reagents, and obviates the need for peptide loading. Although we have not yet compared the relative potency between recombinant S-59/UVA Lm ΔactA/ΔuvrAB-based and peptide-loaded DC vaccines, whole antigens can be expressed by recombinant L. monocytogenes–based vaccines, offering the theoretical advantage of application to individuals with diverse MHC haplotypes and without requiring knowledge of specific immunogenic epitopes.

Psoralen-inactivated nucleotide excision repair mutant analogues of commonly studied biodefense agents, such as Bacillus anthracis (the causative agent of anthrax), Yersinia pestis (the causative agent of plague) and Francisella tularensis (the causative agent of tularemia) could be derived from attenuated organisms in which toxins are genetically inactivated. We constructed a B. anthracis (Sterne) nucleotide excision repair mutant and showed its increased sensitivity to psoralen inactivation that was similar to Lm ΔuvrAB (data not shown). This work is based on the hypothesis that combining psoralen inactivation and metabolic activity within the context of an appropriately attenuated whole B. anthracis organism will result in a vaccine with greater potency that induces immune responses targeted at multiple bacterial antigens, as compared to the licensed anthrax vaccine adsorbed28. Continued studies will be a valuable test of whether this vaccine approach can be applied to understanding infection, like with B. anthracis, a circumstance under which humoral immunity is an essential component of immunological protection29.

Much literature exists showing the potency of L. monocytogenes–based vaccines in animal models for infectious disease or cancer, in both prophylactic and therapeutic settings30,31,32,33,34,35,36. But its role as a food-borne pathogen has hindered development and testing in humans. We have reported previously on a unique live-attenuated L. monocytogenes vaccine platform deleted of two virulence determinants, actA and inlB, which retains the immunogenicity of wild-type L. monocytogenes but is attenuated 1,000-fold in mouse virulence30. Photochemical treatment of L. monocytogenes nucleotide excision repair mutants represents an additional vaccine strategy for L. monocytogenes–based platforms, and may have application for treatment or prevention of diseases having a risk-benefit profile not appropriate for live attenuated vaccines.

Methods

Bacterial strains and virus.

All L. monocytogenes strains were derived from the wild-type L. monocytogenes strain DP-L4056 (ref. 36). We generated L. monocytogenes strains and B. anthracis (Sterne) with uvrAB in-frame deletion by splice-overlap extension PCR and allelic exchange using established methods38,39. We used the pPL2 integration vector to construct recombinant L. monocytogenes strains containing a single copy of OVA and AH1-A5/OVA integrated in the bacterial genome30,37. Vaccine stocks were prepared from mid-log phase L. monocytogenes formulated in a mixture of 8% DMSO and PBS and stored at −80 °C. The parent and recombinant (rVV-OVA) vaccinia viruses were provided by N. Restifo (National Cancer Institute), and were grown and titered on Vero cells.

S-59 psoralen/UVA inactivation of bacteria and metabolic activity.

We grew L. monocytogenes or B. anthracis (Sterne) in brain-heart infusion broth (Difco) at 37 °C to an OD600 of 0.5 or 0.3, respectively, and we added S-59 psoralen directly to the cultures for 1 h. We transferred bacterial cultures to culture plates and UVA irradiated them at a dose of 6 or 6.5 J/cm2 (FX1019 irradiation device, Baxter Fenwal), respectively. We formulated psoralen-inactivated bacteria and stored them as described above for live bacteria. We assessed metabolic activity of live or S-59/UVA L. monocytogenes by autoradiography of secreted bacterial proteins separated by electrophoresis and labeled with 35S-methionine for 30 min after incubation for 30 min in methionine-depleted medium.

Psoralen adduct frequency.

We determined psoralen adduct frequency by using 14C labeled S-59 psoralen for photochemical inactivation, according to the concentrations listed in Table 1. After photochemical treatment, we isolated genomic DNA as described40. We determined the adduct frequency (defined as the base-pair interval between psoralen adducts) by dividing the calculated DNA genomes by the S-59 psoralen molecules, determined from the DNA concentration, the radioactivity of each sample, and specific activity of 14C S-59 psoralen. Table 1 is representative of three independent experiments.

Fluorescence microscopy.

We infected DC2.4 cells grown on coverslips with L. monocytogenes for 30 min at 37 °C at an multiplicity of infection of 10. We removed extracellular bacteria by washing with PBS and incubated infected cells for 5 h at 37 °C with 50 μg/ml gentamicin to prevent growth of extracellular bacteria. We fixed cells with 3.5% formaldehyde, stained them with L. monocytogenes–specific rabbit antibody (Difco), and visualized with FITC-labeled rabbit-specific goat antibody (Vector Laboratories). We detected actin using phalloidin-rhodamine (Molecular Probes) and mounted coverslips using ProLong Gold Anti-Fade containing DAPI (Molecular Probes). For assessment of bacterial morphology, we added psoralen-inactivated L. monocytogenes vaccines to brain-heart infusion broth media, incubated them at 37° C for the indicated period and spotted aliquots of cultures onto poly-L-lysine coverslips; we fixed bacteria, stained them with antibody and mounted them as described above.

Mice and immunizations.

We treated 6–8-week-old female Balb/c or C57Bl/6 mice (Charles River Laboratories) according to US National Institutes of Health guidelines. All protocols requiring animal experimentation received prior approval from the Cerus Animal Care and Use Committee. All vaccinations for immunogenicity, protection and tumor studies were by intravenous injection using the following doses (three mice per group, unless otherwise indicated): live L. monocytogenes (all strains except Lm Δhly), 1 × 107 CFU; live Lm Δhly, 1 × 108 CFU; heat-killed L. monocytogenes, 1 × 108 infectious units; and S-59/UVA L. monocytogenes, 1 × 108 IU.

Dendritic cells.

We prepared DCs from whole bone marrow of C57BL/6 mice using high GM-CSF concentrations (20 ng/ml mouse GM-CSF; R&D Systems)41. On day 10 after plating, we verified nonadherent cells phenotypically to be myeloid DCs (CD11chi). For immunization, bone marrow–derived DCs (CD11chi) were incubated for 1 h with L. monocytogenes vaccines, washed and then 4 × 106 infected bone marrow–derived DCs were given intravenously to C57BL/6 mice.

B3Z T-cell hybridoma activation.

B3Z is a lacZ-inducible CD8+ T-cell hybridoma specific for the OVA257–264 (SIINFEKL) epitope presented on the mouse H-2Kb class I molecule42. We coincubated 1 × 105 DC2.4 cells infected with L. monocytogenes as described for fluorescent microscopy studies with B3Z cells at a ratio of 1:1 in flat-bottom 96-well plates overnight in the presence of 50 μg/ml gentamicin. We added CPRG substrate and we determined absorbance at λ = 450 nm after 2 h. B3Z activation of DC2.4 cells infected with heat-killed Lm-OVA was not observed (data not shown).

Peptides.

Peptides for in vitro assays were synthesized commercially (SynPep). We measured OVA-specific CD8+ T cells with the H-2Kb–restricted epitope SIINFEKL14. We detected L. monocytogenes–specific CD4+ T-cell immunity with the LLO190–201 epitope (NEKYAQAYPNVS)43. We assessed Gp70-specific immunity using the endodenous H-2Ld–restricted epitope AH1 (SPSYVYHQF) and the altered peptide ligand AH1-A5 (SPSYAYHQF)22.

Cytokine flow cytometry.

We determined the frequency of IFN-γ– and tumor necrosis factor (TNF)-α–secreting CD4+ and CD8+ T lymphocytes specific for OVA, gp70, LLO from the pooled spleens of three mice per experimental group by cytokine flow cytometry as described previously30. We stained cells for cell-surface markers using antibody to CD4-fluorescein isothiocyanate (RM4-5; eBioscience) and antibody to CD8α-peridinin chlorophyll protein (53-6.7; BD Pharmingen). We fixed cells in 2% paraformaldehyde, permeabilized them with Perm/Wash buffer (BD Pharmingen), and incubated them with antibody to IFN-γ–allophycocyanin (XMG1.2; eBioscience) and antibody to TNF-α–phycoerythrin (MP6-XT22; eBioscience). Samples were acquired on a FACSCalibur flow cytometer and data analyzed using FloJo software (Treestar).

Vaccinia virus protection studies.

We gave C57BL/6 mice prime and boost vaccinations separated by 14 d with L. monocytogenes strains expressing OVA, and 30 d later we challenged mice intraperitoneally with 5 × 106 plaque-forming units (PFU) of rVV-OVA. We evaluated protection by measuring viral titer in the ovaries 5 d after virus challenge as described44. Mice immunized with S-59/UVA Lm ΔactA/ΔuvrAB-OVA were not protected against challenge with parental vaccinia virus (data not shown).

Tumor studies.

We intravenously implanted Balb/c mice with 2 × 105 CT26 cells. We randomized mice 3 d later (ten animals per group) and vaccinated them. For enumeration of tumor nodules, we harvested lungs from mice 20 d after tumor-cell seeding, fixed them in Bouin fluid and photographed them. For survival studies, mice were killed when they started to show any signs of stress or labored breathing.

L. monocytogenes protection studies.

To assess protective immunity, we vaccinated C57Bl/6 mice on days 0 and 14 with the indicated strains, and challenged them 90 d after vaccination with twice the LD50 of wild-type L. monocytogenes (1 × 105 CFU), and we measured CFU in spleen 3 d later in organ homogenates as described previously45.

Statistical analysis.

Differences in protection against rVV-OVA or L. monocytogenes challenges were determined by the Kruskal-Wallis nonparametric test with Dunn post test. Kaplan-Meier tumor survival curves were compared by the Mantel-Haenszel log-rank test. A P value of ≤0.05 was considered to be statistically significant. Unless otherwise indicated, all experiments were conducted at least twice.

Note: Supplementary information is available on the Nature Medicine website.