ABSTRACT
Human papillomavirus (HPV) vaccines based on L1 virus-like particles (VLPs) can prevent HPV-induced genital neoplasias, the precursors of cervical cancer. However, most cervical cancers occur in developing countries, where the implementation of expensive vaccines requiring multiple injections will be difficult. A live Salmonella-based vaccine could be a lower-cost alternative. We previously demonstrated that high HPV type 16 (HPV16)-neutralizing titers are induced after a single oral immunization of mice with attenuated Salmonella enterica serovar Typhimurium strains expressing a codon-optimized version of HPV16 L1 (L1S). To allow the testing of this type of vaccine in women, we constructed a new L1-expressing plasmid, kanL1S, and tested kanL1S recombinants of three Salmonella enterica serovar Typhi vaccine strains shown to be safe in humans, i.e., Ty21a, the actual licensed typhoid vaccine, and two highly immunogenic typhoid vaccine candidates, Ty800 and CVD908-htrA. In an intranasal mouse model of Salmonella serovar Typhi infection, Ty21a kanL1S was unique in inducing HPV16-neutralizing antibodies in serum and genital secretions, while anti-Salmonella responses were similar to those against the parental Ty21a vaccine. Electron microscopy examination of Ty21a kanL1S lysates showed that L1 assembled in capsomers and capsomer aggregates but not well-ordered VLPs. Comparison to the neutralizing antibody response induced by purified HPV16 L1 VLP immunizations in mice suggests that Ty21a kanL1S may be an effective prophylactic HPV vaccine. Ty21a has been widely used against typhoid fever in humans with a remarkable safety record. These finds encourage clinical testing of Ty21a kanL1S as a combined typhoid fever/cervical cancer vaccine with the potential for worldwide application.
Human papillomavirus (HPV) infection, most often HPV type 16 (HPV16), is considered to be a necessary factor for the development of cervical cancer, with an estimated worldwide annual mortality of 250,000 (7). Given the high prevalence of HPV infection in women and the lack of antiviral agents against HPV, the development of a prophylactic HPV vaccine has been a long-sought strategy to prevent cervical cancer (28). It has been shown that the major papillomavirus capsid protein L1 has the intrinsic ability to self-assemble into virus-like particles (VLPs) that resemble the HPV virion but are noninfectious since they lack the viral genome. HPV vaccines based on these VLPs have proven to be well tolerated, highly immunogenic, and able to prevent the development of HPV16-induced cervical intraepithelial neoplasia in ongoing clinical trials (reviewed in reference 29). One VLP-based vaccine, Gardasil, has been approved for general use in young women in many countries. However, these expensive vaccines are administered in three intramuscular injections over 6 months, which will make access to these vaccines problematic in developing countries, where most cases of cervical cancer occur (38). It is thus of great importance to develop other strategies that have worldwide applicability.
Live attenuated Salmonella strains may be effective antigen delivery systems, as they are able to express foreign antigens and elicit mucosal as well as systemic immune responses against homologous and heterologous antigens after oral vaccination (reviewed in references 13, 26, and 45). We previously reported the induction of high HPV16-neutralizing titers after a single oral immunization of mice with attenuated Salmonella enterica serovar Typhimurium strains expressing a plasmid harboring an L1 codon optimized for expression in Salmonella, L1S (5). In this study, the Salmonella-based vaccine against HPV16 was further refined to make it appropriate for testing in women. First, the ampicillin selection marker used for plasmid maintenance was replaced by a kanamycin resistance gene, which is more suitable for use in humans given its superior biosafety record and given that it was approved by the FDA in 1994 for use in plants (14). The resulting plasmid was tested in three Salmonella enterica serovar Typhi vaccine strains that have been shown to be safe in humans, i.e., Ty21a (17), the actual licensed typhoid vaccine, as well as two highly immunogenic typhoid vaccine candidate strains, i.e., Ty800 (19) and CVD908-htrA (48). Salmonella enterica serovar Typhi vaccine strains have often been directly tested in human volunteers because they do not infect other hosts by the oral route. However, mice can be transiently infected if high doses of these bacteria are administered by the nasal route (15). In the present study, this unique animal model was used to compare the immune responses elicited by these three recombinant Salmonella enterica serovar Typhi vaccine strains against homologous and heterologous antigens.
MATERIALS AND METHODS
Plasmid construction and bacterial strains used.In plasmid pFS14nsd HPV16-L1S (5), the ampicillin resistance coding sequence was replaced by a kanamycin resistance coding sequence as follows. A SacII-XbaI fragment containing the kanamycin coding sequence and promoter was generated by PCR using pET-9a (Novagen) plasmid DNA as template. The primers used were a 25-mer primer located 54 nucleotides upstream from the first ATG of kanamycin and containing a SacII restriction site (underlined), 5′-GGGCCGCGGTGGTCATGAACAATAA-3′, and a 28-mer primer containing an XbaI restriction site (underlined), 5′-GGGTCTAGAAGCTGTCAAACATGAGAAT-3′. Another large SacII-XbaI fragment containing the entire pFS14nsd HPV16-L1S plasmid sequence, without the ampicillin resistance gene, was generated by inverse PCR with Expand High Fidelity PCR (Roche Molecular Biochemicals) with the following primers: a 28-mer primer located 92 nucleotides upstream from the ATG of ampicillin and containing a SacII site (underlined), 5′-GGGCCGCGGTTTGTAGAAACGCAAAAAG-3, and a 28-mer primer containing a XbaI site (underlined) and the stop codon of ampicillin (boldface type), 5′-GGGTCTAGATCCTAACTGTCAGACCAAG-3′. The two SacII-XbaI fragments were ligated together to generate plasmid pFS14nsd-kan3-HPV16-L1S. A correct nucleotide sequence was confirmed by sequencing the full plasmid. The new plasmid was introduced by electroporation (46) into the attenuated Salmonella enterica serovar Typhimurium PhoPc (CS022) (32), PhoP− (CS015) (31) (both kindly provided by John Mekalanos, Boston, MA), and ΔaroA (SL7207) (20) (kindly provided by Irene Corthésy-Theulaz, Lausanne, Switzerland) strains and into the attenuated Salmonella enterica serovar Typhi Ty800 (19) (kindly provided by the Virus Research Institute, Cambridge, MA), CVD908-htrA (48) (kindly provided by Myron Levine, Baltimore, MD), and Ty21a (17) (Berna Biotech, Switzerland) strains. The original pFS14nsd HPV16-L1S plasmid was also introduced in the three attenuated Salmonella enterica serovar Typhi vaccine strains for comparison purposes.
HPV16 L1 expression and analysis of assembly state.Expression of L1 in Salmonella lysates was analyzed by Western blotting as previously described (35) by using the anti-HPV16 L1 monoclonal antibody CAMVIR-1 (Anawa). The content in HPV16 VLP equivalents was measured by a sandwich enzyme-linked immunosorbent assay (ELISA) as previously described (6) by using two monoclonal antibodies (H16E70 or H161A and H16V5) (kindly provided by N. D. Christensen, Hershey, PA), which recognize conformational epitopes on HPV16 VLPs (11).
The L1 assembly state was analyzed by electron microcopy after OptiPrep separation as follows. Ty21a kanL1S bacteria grown to mid-log phase (optical density at 600 nm [OD600] of ≤0.7) were harvested by centrifuging at 3,300 × g. The bacteria were lysed in ice-cold B-Per II reagent (Pierce). Halt protease inhibitor (Pierce), EDTA, Plasmid Safe, Benzonase, and MgCl2 were added to final concentrations of 1%, 5 mM, 0.2%, 0.2%, and 9.5 mM, respectively. The lysates were incubated at room temperature for 10 min and centrifuged at 10,000 × g to obtain the pellet and supernatant fractions. For OptiPrep centrifugation, NaCl was added to the lysate to a final concentration of 0.8 M, and the lysate was clarified at 16,300 × g for 15 min at 4°C. Supernatants were loaded onto 27%, 33%, and 39% OptiPrep-Dulbecco's phosphate-buffered saline-0.8 M NaCl step gradients in polyallomer tubes (1.2 ml each; gradient diffused 1 to 2 h at room temperature before use). The OptiPreps were run at 50,000 rpm for 3 h 30 min at 16°C in a SW55Ti rotor. In some experiments, L1-positive fractions were further purified over 2% agarose columns (http://home.ccr.cancer.gov/lco/gelfiltration.htm ). L1-positive fractions were stained with uranyl acetate and examined by electron microscopy.
Immunization of mice, analysis of anti-HPV16 VLP antibodies, and recovery of Salmonella.Six-week-old female BALB/c mice (Iffa Credo, France) were used in all experiments according to ethical directives of Swiss veterinary authorities. Salmonella enterica serovar Typhi and Salmonella enterica serovar Typhimurium strains were grown in Luria-Bertani broth or agar, except for CVD908-htrA, which was grown in Aro broth or agar (21). Ampicillin (Sigma) and kanamycin (Fluka) were used at 50 μg/ml when indicated. For preparation of the Salmonella inoculum, the bacteria were grown to mid-log phase (OD600 of ≤0.7). For Ty21a kanL1S, 0.1% galactose was added at an OD600 of 0.3 when indicated. Twenty microliters of bacterial inoculum was administered orally (109 CFU) or intranasally (i.n.) (107 to 109 CFU) under anesthesia as previously described (22, 35). Purified baculovirus-expressed HPV16 VLPs were produced as described previously (18, 22a) and were administered subcutaneously (s.c.) (1-μg doses) or i.n. in anesthetized mice (5-μg doses) as previously described (4, 34). The latter protocol corresponds to an aerosol-like administration with inhalation of the VLP inoculum into the lung (3, 33). However, while the parenteral protocol is optimal, as three 5-μg s.c. VLP doses induced similarly high anti-HPV16 VLP titers (4, 34), the aerosol-like administration is not optimized, but higher VLPs doses cannot be used, and higher anti-HVP16 VLPs responses can be achieved only when mucosal adjuvants are added (4, 34, 42a). Sampling of blood and vaginal washes was performed before and after vaccination throughout a 2- to 10-month period depending on the experiments (see figure legends for detailed schedules).
Sampling of blood and vaginal washes as well as determination of specific antibody end-point titers were performed by ELISA as reported previously (22, 35). Briefly, for determinations of anti-HPV16 VLPs, anti-lipopolysaccharide (LPS), or anti-flagellin antibodies, ELISA plates were coated with 50 ng of HPV16 VLPs in phosphate-buffered saline with Salmonella serovar Typhi LPS (Sigma) coupled with methylated bovine serum albumin (Sigma) in carbonate buffer as previously described (37) or with flagellin purified from Ty21a (50 ng/well) in carbonate buffer, respectively. For total immunoglobulin G (IgG) or IgA determinations, plates were coated with 100 ng of sheep anti-mouse Ig (Boehringer) in carbonate buffer. Total or specific IgA or IgG antibodies were detected with biotinylated goat anti-mouse IgA (Kirkegaard & Perry Laboratories) or IgG (Amersham Pharmacia) as secondary antibodies, respectively. End-point dilutions of all samples were carried out. The specific IgA or IgG titers were expressed as the reciprocal of the highest dilutions that yielded an OD492 that was four times that of preimmune samples. These titers were normalized to the amount of total IgA or IgG in vaginal washes (22).
Recovery of Salmonella enterica serovar Typhimurium or Salmonella enterica serovar Typhi was determined using organs from euthanized mice by plating onto agar plates with and without antibiotics as previously described (35).
Neutralization assays.Neutralization assays were performed with secreted alkaline phosphatase (SEAP) HPV16 pseudoviruses as described in detail previously (41). Briefly, OptiPrep-purified SEAP HPV16 pseudoviruses diluted 3,000-fold were incubated on ice for 1 h with twofold serial serum or vaginal wash dilutions, and the pseudovirus-antibody mixtures were used to infect 293TT cells for 3 days. The SEAP content in 10 μl of clarified cell supernatant was determined using the Great ESCAPE SEAP chemiluminescence kit (BD Clontech). Neutralization titers were defined as the reciprocal of the highest serum or vaginal wash dilution that caused at least a 50% reduction in SEAP activity (100% SEAP activity ranging from 30 to 150 relative light units). The lowest serum and vaginal wash dilutions tested were 1:10 and 1:5, respectively, which correspond, when mixed with the pseudovirions, to final dilutions in the assay of 1:50 and 1:25, respectively. Preimmune vaginal washes at this dilution were not neutralizing. Vaginal washes of immunized mice that were not neutralizing at this dilution were attributed an arbitrary value of 5 (log10 = 0.7), which corresponded to the lower limit of detection in this assay.
Proliferation assays.Splenocytes were isolated by mechanical dissociation as previously described (3). CD4+ T cells were purified by magnetic antibody cell sorting using anti-CD4+-coated microbeads (Miltenyi Biotec, Gladbach, Germany) according to the manufacturer's instructions. CD4+ T cells from naïve or immunized mice were incubated for 3 days with HPV16 VLPs (2 μg/ml [GMP preparation from Novavax]) (36) and flagellin (10 μg/ml, purified from Ty21a). Incubations with medium alone or concanavalin A (2.5 μg/ml) were used as negative and positive controls, respectively. [3H]thymidine (0.5 μCi; Amersham) was added overnight, and incorporation was measured as counts per minute.
RESULTS
Expression of HPV16 L1S and plasmid stability in a kanamycin-resistant PhoPc strain.A kanamycin-resistant plasmid expressing HPV16 L1S was constructed by replacing the ampicillin-resistant gene in the original pFSnsdHPV16 L1S (5) by a kanamycin-selectable gene. An inverse PCR strategy was used to amplify the entire plasmid flanking the ampicillin-resistant gene sequence, and the resulting fragment was ligated into a kanamycin resistance-encoding sequence (see Materials and Methods for details). The resulting plasmid was designated pFS14nsd-kan3-HPV16 L1S and electroporated into PhoPc to yield PhoPc kanL1S (Table 1). In vitro expression of neutralizing antibody-reactive HPV16 L1 in this new strain was similar to that measured in the ampicillin-resistant PhoPc L1S strain (ca. 10 μg VLPs/1011 CFU) (5). Similar growth rates were also observed in the two strains, with about 7 h to reach the mid-log phase, and plasmid stability was very high, with almost 100% of the bacteria still harboring the plasmid after four consecutive cultures performed overnight in antibiotic-free medium. In addition, the stability of the kanL1S plasmid was increased in vivo compared to the ampicillin plasmid, with all bacteria harboring the kanL1S plasmid in all organs examined 2 weeks after oral immunization (Table 2).
Salmonella strains used in this study
Recovery of Salmonella PhoPc L1S carrying the ampicillin or kanamycin resistance plasmid 2 weeks after oral immunization
Anti-HPV16 VLP humoral responses after oral immunization with PhoPc, PhoP−, and AroA strains carrying the kanL1S plasmid.The kanL1S plasmid was then introduced into two attenuated Salmonella enterica serovar Typhimurium strains, the PhoP− and AroA strains (Table 1), carrying attenuating mutations in Salmonella enterica serovar Typhi vaccine strains that were previously shown to be safe for humans. Three groups of mice received a single oral dose of these three recombinant Salmonella enterica serovar Typhimurium strains, and HPV16 VLP-specific antibody responses in serum and vaginal washes were analyzed at 8 weeks postimmunization (Fig. 1). Interestingly, the serum anti-VLP antibody titers induced by the AroA kanL1S strain were as high as those induced by the PhoPc kanL1S strain (Fig. 1A) and by the former PhoPc L1S strain (5). In contrast, the anti-VLP antibody titers induced by PhoP− kanL1S were 2 logs lower, as previously reported for PhoP− L1S (5). HPV16 VLP-specific IgG and IgA were also induced in genital secretions of mice immunized with the PhoPc kanL1S and AroA kanL1S strains, while none were detected in mice receiving the PhoP− kanL1S strain (Fig. 1B). All serum titers remained stable for at least 6 months, while some decrease in genital titers were observed, as found previously with PhoPc L1S (5; data not shown). These data obtained in mice using oral immunization suggest that recombinant Salmonella enterica serovar Typhi harboring aro deletions may be more immunogenic than those harboring phoP/phoQ deletions.
Comparison of serum anti-HPV16 VLP antibody titers after oral vaccination with PhoPc kanL1S, PhoP− kanL1S, and ΔAroA kanL1S. Groups of five BALB/c mice were orally immunized with 109 CFU of PhoPc kanL1S (black bars), PhoP− kanL1S (white bars), and AroA kanL1S (striped bars). Anti-HPV16 VLP ELISA titers are shown 8 weeks after immunization in serum (A) and in vaginal washes (B). Data are expressed as the geometric means (log10) of the reciprocal dilutions of specific IgG from individual mice in serum and specific IgG and IgA per microgram of total IgG (IgA) in vaginal washes. Error bars indicate the standard errors of the means (SEM).
Humoral and cellular immune responses induced by Salmonella enterica serovar Typhi vaccine strains expressing HPV16 L1 in the i.n. murine model.We introduced the kanL1S plasmid into three Salmonella enterica serovar Typhi vaccine strains that were available to us, i.e., Ty21a, the licensed typhoid vaccine strain (ΔgalE with undefined attenuating mutations) (17); Ty800 (ΔphoP/phoQ) (19); and CVD908-htrA (ΔaroC ΔaroD htrA) (48) (Table 1). Salmonella enterica serovar Typhi strains are not invasive in mice by the oral route. However, mice can be transiently infected if bacteria are administered at high doses by the nasal route, as shown with recombinant CVD908-htrA (40, 42). The immunogenicity of the three recombinant Salmonella enterica serovar Typhi strains was therefore evaluated in BALB/c mice after administration of two i.n. doses (109 CFU for Ty21a and CVD908-htrA and 107 CFU for Ty800 recombinant strains), with the first dose given on day 1 and the second dose given 1 month later. The 109-CFU doses and schedule were chosen according to previously reported experiments performed by other groups using the nasal murine model with Ty21a (16, 23) and CVD908-htrA (27). Administration of Ty800 to mice was not previously reported. A lower dose (107 CFU) was used because higher inocula were lethal to the mice. Our observation is in agreement with a previously described higher mortality of such PhoPQ-deleted bacteria after intraperitoneal injection with hog gastric mucin in mice (2). The IgG titers against the heterologous HPV16 VLP antigen, as well as against two homologous antigens, LPS and flagellin, were measured in serum over an 8-week period (Fig. 2A). Interestingly, high anti-VLP IgG titers were induced only after immunization with Ty21a kanL1S, while low or barely detectable titers were induced by the two other strains. In contrast, Ty800 kanL1S and CVD908-htrA kanL1S induced higher anti-LPS and anti-flagellin IgG titers than Ty21a kanL1S, indicating that the two former recombinant strains had successfully infected mice and induced humoral responses against the bacterial antigens (Fig. 2B). However, the anti-LPS and anti-flagellin titers induced by Ty21a kan L1S and the parental Ty21a strain were indistinguishable (Fig. 2C). This result suggests that the parental and recombinant Ty21a strains may have a similar potential to protect against typhoid fever.
Comparison of serum anti-HPV16 VLP antibody titers after nasal vaccination with Ty21a kanL1S, Ty800 kanL1S, and CVD908-htrA kanL1S. Groups of five BALB/c mice were nasally vaccinated at weeks 0 and 4 with 109 CFU of Ty21a kanL1S (plain line) and CVD908-htrA kanL1S (pointed line) and with 107 CFU of Ty800 kanL1S (dashed line). Anti-HPV16 VLP (A), as well as anti-LPS and anti-flagellin (B) IgG ELISA titers in serum were determined every 2 weeks. Comparisons of serum IgG titers induced after immunization with Ty21a kan L1S (black bars) and Ty21a (white bars) are also indicated (C). Data are expressed as the geometric means (log10) of the reciprocal dilutions of specific IgG from individual mice. Error bars indicate the SEM.
CD4+ T-helper cells participate in the generation and maintenance of high antibody titers; therefore, cell-mediated immune responses against flagellin and HPV16 VLPs were also examined. Eight weeks after immunization, mice were sacrificed, and antigen-specific proliferation was measured using CD4+ T cells purified from the corresponding spleen (Fig. 3). HPV16 VLP stimulation of CD4+ T cells was strongly induced after i.n. vaccination with Ty21a kanL1S (P < 0.001) (Fig. 3A), in agreement with the induction of anti-HPV16 VLP antibodies. In contrast, weak CD4+ T-cell proliferation was observed in mice immunized with the other two strains (Ty800 kanL1S and CVD908-htrA kanL1S). Only immunization with Ty800 kanL1S induced significant flagellin-specific stimulation of CD4+ T cells (P < 0.001) (Fig. 3B).
Comparison of HPV16 VLP and flagellin-specific CD4+ T-cell proliferations. Groups of five BALB/c mice were nasally vaccinated at weeks 0 and 4 with 109 CFU of Ty21a kanL1S (black bars) and CVD908-htrA kanL1S (punctated bar) and with 107 CFU of Ty800 kanL1S (striped bars). HPV16 VLP (A)- and flagellin (B)-specific CD4+ T-cell proliferations are shown at week 8. Data are expressed as mean stimulation indices of triplicate cell cultures (cpm in presence of antigen/cpm in absence of antigen). CD4+ T-cell proliferation was also determined in naïve mice (white bars), and statistical analysis was performed with one-way analysis of variance and a Bonferroni posttest using GraphPad Prism software.
Stability and expression of the kanL1S plasmid in three Salmonella enterica serovar Typhi vaccine strains and L1 assembly status in Ty21a kanL1S.In order to gain insights into the higher immune responses against VLP observed with the Ty21a strain, plasmid stability and expression in these three recombinant Salmonella enterica serovar Typhi strains were compared. The stability of the kanL1S plasmid in the three vaccine strains in the absence of antibiotics was first examined in vitro (Fig. 4). In contrast to the above-described results for Salmonella enterica serovar Typhimurium strains, a lower stability of the kanL1S plasmid was observed in all Salmonella enterica serovar Typhi strains tested. A slow loss of plasmid occurred after the second culture performed overnight, but after five consecutive cultures performed overnight, the plasmid was retained in 10, 6, and 20% of the bacteria in Ty21a, Ty800, and CVD908-htrA, respectively. To examine whether this plasmid instability was linked to the Salmonella serovar Typhi bacterial host or to the kanamycin selection marker, the original L1S plasmid was introduced into the three Salmonella enterica serovar Typhi vaccines strains, and its stability was determined in vitro (Fig. 4). After five consecutive cultures performed overnight, the stability of the original L1S plasmid containing the ampicillin selection marker was rather similar to that of the kanL1S plasmid in Ty21a and CVD908-htrA (14% and 8% of bacteria retaining the plasmid, respectively). In contrast, the L1S plasmid was highly unstable in Ty800 (0.0002% of bacteria retaining the plasmid). Therefore, plasmid instability can be linked to either strain or resistance marker and remains unpredictable.
In vitro stability of the kanL1S and L1S plasmids in different Salmonella enterica serovar Typhi strains. The number of successive cultures at a 1/100 dilution performed overnight (ON) in medium without antibiotic is indicated on the horizontal axis. Each morning, bacteria (plain line, kanL1S; dashed line, L1S) were plated onto agar in the presence or absence of antibiotic. The vertical axis represents the percentage of bacteria that have retained the plasmid. Error bars indicate SEM.
Salmonella enterica serovar Typhi vaccine strains will undergo only limited rounds of replication in humans. To estimate how the plasmid instability observed in vitro may translate in vivo, plasmid stability in Salmonella cells recovered from the lung and spleen of mice was examined 1 week after i.n. immunization with 109 CFU of Ty21a kanL1S and CVD908-htrA kanL1S or 107 CFU of Ty800 kanL1S. Despite the use of a 2-log-lower dose of Ty800, the total number of bacteria recovered from the lung was higher with this strain, suggesting a higher invasiveness (Table 3). In agreement with this finding, bacteria were recovered only from the spleen of mice immunized with Ty800 recombinant strains (data not shown). Interestingly, all bacteria recovered from the lung after Ty21a kanL1S immunization harbored the plasmid, while after Ty800 kanL1S and CVD908-htrA kanL1S immunizations, few (15.8 and 1.7%, respectively) harbored the kanL1S plasmid (Table 3). Comparison after immunization with the strains harboring the original L1S plasmid also revealed a higher plasmid stability in Ty21a (Table 3). Our data thus indicate that the L1S-expressing plasmids are more stable in Ty21a than in the other Salmonella enterica serovar Typhi vaccine strains tested in the nasal murine model.
Recovery from the lung of mice of different Salmonella enterica serovar Typhi strains carrying the kanL1S plasmid 1 week after nasal immunization
In vitro expression levels of neutralizing antibody-reactive HPV16 L1 were similar in the three Salmonella serovar Typhi strains (20 to 30 μg VLP/1011 CFU), suggesting that anti-VLP immunogenicity did not correlate to in vitro expression levels. The sandwich ELISA used to quantify conformationally correct L1 does not discriminate between L1 capsomers, capsomer aggregates, or VLPs, since all are presumably recognized by H16 E70 and H16 V5, the antibodies used in this assay (11). We therefore further evaluated the L1 assembly state in Ty21a extracts. Exponentially growing cultures were lysed and subjected to centrifugation at 10,000 × g. L1, as assessed by Western blotting with an L1-specific monoclonal antibody, was distributed approximately equally between the pellet and supernatant (data not shown), suggesting that a considerable fraction of L1 was in a high-molecular-weight complex, higher than expected for individual VLPs. The supernatant was subjected to OptiPrep step gradient centrifugation under conditions that separate VLPs from capsomers (8). Western blots reveled L1 in both high- and low-molecular-weight fractions (Fig. 5A). Separate fractions or pools of L1-positive fractions were examined by electron microscopy after negative staining. Structures resembling pentamers and various-sized amorphous aggregates of pentamers were seen, but no well-ordered VLPs were visible in any of the fractions (Fig. 5B). Star-shaped objects (Fig. 5B) were detected in the intermediate-molecular-weight fractions. Examination of these fractions after the addition of purified HPV16 L1 VLPs or pentamers confirmed that these objects were intermediate in size (data not shown), with a diameter of 17 nm, compared to 12 nm and 60 nm for pentamers and VLPs, respectively. However, they are clearly not T = 1 capsids, which can spontaneously assemble in vitro from Escherichia coli-derived N-terminally-deleted HPV16 L1, but are reported to be 30 nm (10). The star-shaped objects remained after rabbit HPV16 L1 VLP antiserum-mediated immunodepletion of all Western blot-detectable L1 (data not shown). Consequently, it is unlikely that these structures are L1 derived. We were concerned that the lysis buffer used to disrupt the cells (B-Per II reagent) might have caused a disaggregation of preformed VLPs after lysis. However, incubation of purified insect cell-derived HPV16 L1 VLPs with B-Per II reagent did not lead to the disassembly of the VLPs, as assessed by electron microscopic examination of negatively stained preparations (data not shown).
Electron microscopy analysis of Ty21a kanL1S lysate. (A) Clarified lysate was separated on a 27%, 33%, or 39% OptiPrep step gradient. Sixteen fractions were collected from the bottom of the tube. The first 11 fractions were analyzed by Western blotting with CAMVIR-1 monoclonal antibody, which recognizes denatured 16L1. Molecular mass markers (in kilodaltons) are depicted on the right. The location of full-length monomeric L1 is indicated by the arrow. (B) Fractions that were 16L1 positive in CAMVIR-1 Western blots were pooled and further purified on a 2% agarose column. Sixteen L1 Western blot-positive fractions were pooled and examined by transmission electron microscopy at a magnification of ×26,000. Capsomer aggregates (large arrows), 17-nm-diameter “stars” (small arrows), and 12-nm capsomers (arrowheads) are indicated.
Induction of anti-HPV16 VLPs and HPV16-neutralizing antibodies in serum and genital secretions of mice immunized with Ty21a kanL1S or the prototype HPV16 VLP vaccine.In order to evaluate the potential of Ty21a kanL1S as a cervical cancer prophylactic vaccine, we compared its immunogenicity to that of purified HPV16 VLPs. Groups of five mice were immunized with two i.n. doses of Ty21a kanL1S (ca. 109 CFU) at weeks 0 and 4 (Fig. 6, squares); three s.c. doses of 1 μg VLPs at weeks 0, 4, and 25 (Fig. 6, triangles); or three i.n. (aerosol-like) doses of 5 μg VLPs at weeks 0, 1, and 2 (Fig. 6, circles). The parenteral and aerosol-like protocols using purified VLPs were chosen according to our previous experience (4, 34) and to mimic vaccination of humans (18, 33). A combination of VLPs (two s.c. 1-μg doses at weeks 0 and 4) and Ty21a kanL1S (a single i.n. boost at week 8) was also tested (Fig. 6, diamonds). Anti-HPV16 VLP and HPV16-neutralizing antibody titers were determined in serum and vaginal washes by ELISA and the SEAP HPV16 pseudovirion neutralization assay, respectively, at short term (4 to 5 weeks after the last immunization) (Fig. 6A, C, and E) and at long term (18 to 24 weeks after the last immunization) (Fig. 6B, D, and F). Our data show that immunization with Ty21a kanL1S induced VLP-specific and HPV16-neutralizing antibodies not only in serum (Fig. 6A and B, squares) but also in genital secretions (Fig. 6C, D, E, and F, squares), which may be critical for protection against genital HPV infection. The immune responses at short term (Fig. 6A, C, and E, squares) are representative of three independent experiments.
Anti-HPV16 VLPs and HPV16-neutralizing antibodies in serum and genital secretions of mice immunized with Ty21a kanL1S or the prototype HPV16 VLP vaccine. Groups of five mice were immunized with two i.n. doses of Ty21a kanL1S (ca. 109 CFU) at weeks 0 and 4 (squares); three s.c. doses of 1 μg VLPs at weeks 0, 4, and 25 (triangles); three i.n.(aerosol-like [aer]) doses of 5 μg VLPs at weeks 0, 1, and 2 (circles); or a combination of VLPs (two s.c. 1-μg doses at weeks 0 and 4) and Ty21a kanL1S (a single i.n. boost at week 8) (diamonds). Anti-HPV16 VLP and HPV16-neutralizing antibody titers were determined in serum and genital secretions by ELISA and the SEAP HPV16 pseudovirion neutralization assay, respectively, at short term (4 to 5 weeks after the last immunization) (A, C, and E) and at long term (18 to 24 weeks after the last immunization) (B, D, and F). Data are expressed as the log10 of the reciprocal dilutions of specific IgG in serum and specific IgG and IgA per microgram of total IgG or IgA, respectively, in secretions for the ELISA titer or reciprocal dilutions yielding 50% SEAP inhibition for the neutralizing titers. Between-group differences were analyzed with a Student's t test (GraphPad Prism), and significant differences are indicated. *, P < 0.05; **, P < 0.01; ***, P < 0.001.
Immunization with the prototype s.c. VLP vaccine (Fig. 6, triangles) induced significantly higher VLP-specific and HPV16-neutralizing titers in serum than the mucosal vaccines, i.e., aerosol vaccination with Ty21a kanL1S or VLPs (Fig. 6A and B, triangles versus squares and circles). However, this was not exactly reflected in genital secretions, where the differential induction of VLP-specific IgG and IgA by parenteral versus mucosal protocols of vaccination complicated the situation. While at short term, there was no statistical differences in the normalized titers of VLP-specific IgG induced in the four groups (Fig. 6C), at long term, the mice vaccinated with the prototype s.c. VLP vaccine exhibited significantly higher VLP-specific IgG titers than the mucosal vaccines (Fig. 6D). However, the opposite was true when the IgA tiers were examined: both at short term and long term, the prototype s.c. VLP vaccine induced significantly lower VLP-specific IgA titers than the mucosal vaccines (Fig. 6C and D). To have more definitive insights into the protective potential of Ty21a kanL1S compared to the prototype VLP vaccine, HPV16-neutralizing titers of vaginal washes were examined. Our data show that the protective potential of Ty21a kanL1S is not lower than those of VLP vaccines at either short or long term (Fig. 6E and F), despite the fact that only two doses were administered, suggesting that this is a promising vaccine to test in women. Evaluation of the combined VLP-Ty21a kanL1S protocol raised the possibility that the third dose of the s.c. prototype VLP vaccine may be replaced by Ty21a kanL1S, as no statistical differences were observed in serum or genital secretions with the complete s.c. VLP protocols. However, the VLP/Ty21a kanL1s prime-boost strategy generated a lower geometric mean neutralizing titer than the two-dose Ty21a kanL1S vaccination, at least at short term (Fig. 6E).
DISCUSSION
The generation of a prophylactic vaccine against cervical cancer that has improved worldwide applicability has been our major focus for several years. Here, we report on final improvements in the development of a Salmonella-based HPV vaccine for clinical testing, i.e., an acceptable selection marker and identification of a safe and immunogenic typhoid vaccine strain. The development of recombinant bacterial vaccines has usually required the use of selectable markers. Unfortunately, our attempt to use the aspartate β-semialdehyde dehydrogenase-balanced lethal vector-host system (12) to stabilize L1S expression in PhoPc has failed to induce VLP-specific immune responses (data not shown). We have therefore chosen kanamycin as an antibiotic-selectable marker, since it has an established biosafety record and its use has been approved for recombinant plant applications by the FDA (14). Although this does not guarantee that it may be approved for use in a recombinant vaccine, it should be noted that the kanamycin resistance marker will not give the organisms a selectable advantage outside the laboratory, and this phenotype is already quite ubiquitous in nature. In fact, the widespread bacterial resistance to kanamycin has limited the use of this antibiotic in human medicine, and the high substrate specificity of the inactivating enzyme argues against the development of resistance or interference against modern antibiotic therapies. It is probable that regulatory authorities in different countries will have to consider the balance between theoretical safety issues and the disease burden that may be prevented by such a recombinant vaccine. Interestingly, the presence of the kanamycin resistance gene has further improved the stability of the L1S-expressing plasmid in PhoPc in vivo, but the immune responses induced by oral immunization with the three attenuated Salmonella enterica serovar Typhimurium strains harboring the kanL1S plasmid were similar to those obtained with bacteria harboring the ampicillin L1S plasmid (5).
With the exception of the Salmonella enterica serovar Typhimurium PhoP− strain that was tested as a recombinant vaccine in humans (1), Salmonella-based vaccines are all obtained from Salmonella serovar Typhi, and they were first developed as vaccines against typhoid fever (reviewed in reference 25). As potential vaccine carriers for HPV16 VLP, we have tested here the licensed typhoid vaccine strain Ty21a, for which no Salmonella serovar Typhimurium counterpart exists because the attenuating mutations are unknown. In addition, two of the candidate typhoid vaccines, Ty800 (19) and CVD908-htrA (48), which harbor attenuations in the same pathways (phoPQ and aro, respectively) as the Salmonella serovar Typhimurium strains (PhoP− and AroA) were also examined. In vitro analysis of the three recombinant strains revealed a similar VLP-equivalent expression level and an intermediate plasmid stability in the sense that the kanL1S plasmid is less stable in Salmonella serovar Typhi strains than in Salmonella serovar Typhimurium strains, but it is more stable than the ampicillin resistance L1S plasmid. Plasmid instability in Salmonella serovar Typhi but not in Salmonella serovar Typhimurium strains was also observed when a Helicobacter pylori urease-encoding plasmid was tested (1).
Attenuated Salmonella enterica serovar Typhi strains administered i.n. at high doses in mice elicit an array of immune responses similar to those observed in volunteers given the attenuated strains orally (reviewed in reference 39). This finding supports the use of this small-animal model in the preclinical evaluation of the immunogenicity and efficacy of live-strain Salmonella vaccine candidates. This model was also used to investigate the in vivo stability of the kanL1S plasmid. Relatively high numbers of bacteria were recovered from the lung 1 week after immunization, reflecting lung targeting achieved with nasal vaccination under deep anesthesia (4, 27), while no bacteria, with the exception of Ty800, were recovered from spleen at this late time point, in agreement with previous studies (23, 40, 42). Although the kanL1S plasmid was harbored by all Ty21a bacteria recovered, it was present in only a few CVD908-htrA and Ty800 bacteria. To our knowledge, experiments with recombinant Ty800 in the nasal murine model were not previously reported, while plasmid stability was observed in both Ty21a (16) and CVD908-htrA (27). The instability of L1S plasmids in CVD908-htrA may be linked to the presence of a constitutive promoter in our plasmids instead of the in vivo-inducible nirB promoter, which was generally used in other recombinant CVD908-htrA strains (27, 44, 50).
Analysis of the Ty21a kanL1S lysates by sandwich VLP ELISA suggested that most of the bacterial L1 displays conformation-dependent neutralizing epitopes and therefore is overall correctly folded. However, electron microscopy examination of Ty21a kanL1S lysates showed L1 assembled in capsomers and various-sized capsomer aggregates but no VLPs. Previous studies found that purified E. coli-derived L1 capsomers could induce neutralizing antibody responses after parenteral injection (43). Therefore, it is not surprising that high titers of neutralizing antibodies were induced by the L1 recombinant Salmonella strain, despite our inability to detect well-formed VLPs in the lysates. The rare VLPs observed in the original PhoPc L1 lysate (35) may be linked to the higher L1 expression level (50 μg VLP/1011 CFU) in this strain.
After i.n. inoculation, the recombinant Ty21a kanL1S strain was able to induce high anti-HPV16 VLP antibodies in both serum and genital secretions. Indeed, the HPV16-neutralizing potential of vaginal washes of mice vaccinated twice with Ty21a kanL1S was similar to that achieved by three vaccinations with the prototype s.c. VLP vaccine, as determined using a standard SEAP-pseudovirion neutralization assay. However, in serum, the anti-HPV16 VLP IgG and HPV16-neutralizing titers induced by Ty21 kanL1S were lower (ca. l log) than those induced by s.c. vaccination with VLPs. This was accompanied by lower VLP-specific, probably serum-transudating, IgG in genital secretions but was compensated for by the induction of more VLP-specific IgA by mucosal vaccination with Ty21a kanL1S. In our samples, HPV16 VLP ELISA titers showed a high correlation with HPV16-neutralizing titers in serum (Pearson r = 0.763; P < 0.0001) but also in genital secretions (Pearson r = 0.847; P < 0.0001 [IgA plus IgG]), suggesting that the neutralization assay is also valid for these types of samples. Interestingly, neutralization correlated better with the VLP-specific IgA content (Pearson r = 0.731; P < 0.0001) than with the VLP-specific IgG content (Pearson r = 0.385; P = 0.02) in genital secretions, emphasizing the importance of IgA induction in the Ty21a kanL1 immune response.
The kanL1S instability in CVD908-htrA and in Ty800 may explain the lower immune responses against HPV16 VLP generated by these bacteria. In contrast, CVD908-htrA and Ty800 induced higher anti-LPS and anti-flagellin antibodies as well as higher flagellin-specific CD4+ T cells responses for Ty800 than Ty21a. This finding is in agreement with previous studies where CVD908-htrA and/or Ty800 and Ty21a were compared using the murine nasal model (50) or in human oral delivery trials. Ty21a was moderately immunogenic, requiring three oral doses, administered daily, to confer protection against typhoid fever (24), while CVD908-htrA and Ty800 were more immunogenic and/or protective with a single oral dose (19, 48, 49). However, the potential of Ty21a kanL1S as a typhoid vaccine was seemingly not impaired by the expression of HPV16 VLPs, as similar titers of anti-LPS and anti-flagellin antibodies were induced by the parental and L1 recombinant strains in the nasal murine model. There have been limited investigations of the use of Ty21a as a carrier for heterologous proteins. The induction of specific antibody responses by Ty21a expressing surface-displayed viral antigens (23) or hemolysin (16) was reported in the mouse nasal model, while no humoral response but some specific T-cell responses were reported after oral vaccination of human volunteers with Ty21a expressing H. pylori urease (9, 30).
In conclusion, in this study, we show that Ty21a kanL1S induces high HPV16-neutralizing titers and VLP-specific CD4+ T-cell responses in the nasal murine model, including HPV16-neutralizing antibodies in genital secretions. The titers were similar to those induced by the prototype s.c. VLP vaccine, which is promising for the potential of Ty21a kanL1S to prevent HPV16 infection. To date, CVD908-htrA expressing tetanus toxin fragment C is the only recombinant Salmonella serovar Typhi strain that has been tested both in the nasal murine model (40) and after oral administration in humans (47). The induction of protective levels of antitoxin antibodies in mice was confirmed with human volunteers (47), supporting the hypothesis that data obtained using the nasal murine model may be predictive of immune responses in humans. Further support for the premise that Ty21a kanL1S is an attractive candidate for a combined vaccine against both typhoid fever and cervical cancer will require an evaluation of its safety and immunogenicity in women volunteers.
ACKNOWLEDGMENTS
We thank Hakim Echchannaoui for critical reading of the manuscript.
This work was supported by the Swiss National Science Foundation (grants 631-057969.99 and PP00A-104318 to D.N.-H. and 32-63021.00 to D.B.).
FOOTNOTES
- Received 17 April 2007.
- Returned for modification 6 June 2007.
- Accepted 29 July 2007.
- Copyright © 2007 American Society for Microbiology