Salmonella as a vaccine delivery vehicle

Abstract

Attenuated Salmonella vaccines can be administered orally to deliver recombinant antigens to mucosal surfaces inducing a protective immune response against a variety of targeted pathogens. A number of exciting new approaches and technologies for attenuated Salmonella vaccines have been developed recently. However, there remains a disconnect between results obtained with mice in preclinical studies and results obtained in human clinical trials. This is due to an incomplete understanding of S. Typhi interactions with human hosts and inadequate animal models available for study. In this review, we describe recent progress in identifying important differences underlying S. Typhi-host interactions, the development of novel approaches to vaccine design and six recent clinical trials evaluating Salmonella-vectored vaccines.

Introduction

Live recombinant attenuated Salmonella-vectored vaccines hold great promise as a means to improve human health by achieving long-lasting mucosal, humoral and cellular immunity against a variety of non-Salmonella pathogens at a low cost. The approach is fairly straightforward. Heterologous protective antigen genes from a pathogen are cloned and expressed in attenuated Salmonella enterica. The recombinant Salmonella strain expressing the heterologous gene can then be orally administered to elicit an immune response against the pathogen from which the heterologous gene was derived. While strains of other bacteria such as Escherichia coli, Listeria, Shigella and Vibrio have been and continue to be evaluated as vaccines, the invasive nature of Salmonella makes it adept at eliciting B and T cell memory responses and gives Salmonella the greatest potential to elicit long lasting immunity.

The Salmonella serovar of choice for human vaccines is the human-restricted Salmonella enterica serovar Typhi. As the causative agent of typhoid fever, S. Typhi is fully capable of invading human mucosal tissues and entering systemic sites, ultimately targeting the host immune system to stimulate strong mucosal, humoral and cellular responses. Although many other pathogenic serovars of Salmonella, such as Salmonella enterica serovar Typhimurium, cause disease in humans, they have not been pursued as actively as S. Typhi. This is because S. Typhimurium, though capable of causing acute gastroenteritis, is unable to invade beyond the gut mucosa in healthy, immunocompetent humans and is thus less equipped to stimulate long-lasting immunity [1]. Because S. Typhi is host-restricted, small animal models are of limited value. Fortunately, S. Typhimurium causes a disease in mice similar to typhoid fever in humans. Therefore, the behavior of S. Typhimurium in mice is used to evaluate new or novel attenuation and heterologous gene expression strategies. Results from hundreds of animal experiments have demonstrated the potential and utility of this approach. However, while quite useful in some respects, reliance on the S. Typhimurium mouse model to work out attenuation strategies has also been somewhat misleading. First, mutations that result in suitable attenuation and immunogenicity of S. Typhimurium are not necessarily adequate for S. Typhi vaccines. While there are mouse models to evaluate virulence and immunogenicity of S. Typhi-vectored vaccines, none utilize the oral route due to the inability of Typhi to establish an active infection in orally inoculated mice [2]. Second, the mouse model provides little insight into an S. Typhi vaccine’s reactogenicity. As a result, the few clinical trials evaluating attenuated S. Typhi vectoring a heterologous antigen (eight trials reported by Galen [3] and five additional trials reported here) have been hampered by strains that are either over attenuated and poorly immunogenic or too reactogenic [3].

What is the disconnect between successful S. Typhimurium animal studies and unsuccessful S. Typhi human trials? The answer is likely the combination of a number of factors. First, S. Typhimurium and S. Typhi interact differently with their respective hosts. S. Typhimurium has evolved to elicit strong inflammatory responses in the gut, characterized by a massive influx of neutrophils into the terminal ileum and colon. This is an important survival strategy as it leads to the production of alternative terminal electron acceptors, particularly tetrathionate and nitrate. S. Typhimurium is uniquely suited to utilize these molecules, and gains a growth advantage in the inflamed gut due to its ability to respire on tetrathionate [4] and to grow anaerobically on ethanolamine [5]. In contrast, S. Typhi does not cause a diarrheal disease. In the gut it elicits intestinal inflammatory infiltrates consisting primarily of mononuclear cells. S. Typhi also actively suppresses the development of pro-inflammatory cytokines by production of the Vi capsule. Second, S. Typhi produces colonization factors, i.e. fimbriae that are not present in S. Typhimurium, and vice versa [6–10]. While many of these have been studied, their in vivo functions are not always fully understood. Finally, many of the differences in lifestyle result from the expression of different sets of virulence genes. There are significant differences in the chromosomal structures between the two organisms, exemplified by the presence of >200 pseudo-genes in S. Typhi [11], a common occurrence in host-restricted organisms. S. Typhi also contains a number of unique pathogenicity island-associated virulence factors, including the Vi capsular antigen, bundle forming pilus and typhoid toxin (cdtB).

Thus it is likely that failures in the clinic are due, at least in part, to a lack of understanding of S. Typhi pathogenesis. The good news is that many laboratories have made significant progress in the last five years toward closing this gap in our understanding. In addition, there has been significant progress in developing new and novel strategies for vaccine design, some of which have made their way into the clinic. We note that there have been a number of recent reviews describing Salmonella vaccine technologies with regard to vaccine design and preclinical studies [12–17]. We refer our readers to those reviews for additional information. Our focus here is to describe important insights into typhoid pathogenesis, highlight new and novel strategies for vaccine design and review recent clinical trials.

Recent insights into S. Typhi pathogenesis

The past five years have seen some significant breakthroughs in our understanding of S. Typhi pathogenesis. This is important because, at present, the greatest hindrance to developing S. Typhi-based vaccines stems from the fact that S. Typhi is not S. Typhimurium, and mice are not humans. Recent research in the field has begun to break through these barriers to better understand the specific abilities and interactions of S. Typhi with its host. A great deal of work has been done to explore the properties of the Vi polysaccharide capsule. Not present in S. Typhimurium, the Vi capsule has a large impact on both the pathogenesis of S. Typhi and the development of the host immune response. Capsule is produced immediately following invasion into the intestinal epithelial mucosa [18]. Here, it suppresses the production of pro-inflammatory cytokines and IL-8 by masking critical toll-like receptor (TLR) ligands on the surface of the bacterium such as LPS and flagella and by down-regulating the expression of such ligands [19, 20]. However, Vi capsule production does not continue throughout the course of infection, most likely because the presence of Vi decreases the ability of the bacteria to attach to host cells [21]. Instead, it appears that Vi capsule is produced only after initial invasion into the host, and then is down regulated in the deeper lymphatic tissues [22]. This is in stark contrast to the behavior of S. Typhimurium, which promotes rapid pro-inflammatory cytokine release and neutrophil influx into the gastrointestinal mucosa immediately after infection.

Some interesting research has been done with the effectors of the Salmonella pathogenicity islands (SPIs) SPI-1, SPI-2 and SPI-3. All three pathogenicity islands are present in both S. Typhimurium and S. Typhi, but there are substantial differences in the ways the two serovars deploy these systems during infection. SPI-1 and SPI-2 are well-studied pathogenicity islands that encode type 3 secretion systems. In S. Typhimurium, SPI-1 is responsible for the initial invasion of the bacteria into the host, while SPI-2 is essential for survival inside macrophages [23]. The same does not appear to be true for S. Typhi, however, as SPI-2 is not essential for S. Typhi survival in the host [24]. There are also differences in the specific effector proteins secreted by these two systems. S. Typhi lacks some of the effector proteins that have been associated with gastric inflammation and fluid secretion (e.g. SopA and SopD2) [25, 26]. Another inflammation-associated effector, SopE2, is present in Ty2, but absent in CT18 [27]. The absence of these effectors contributes to the significantly lower level of inflammation associated with S. Typhi infections compared to S. Typhimurium infections. In addition, these missing proteins also appear to be detrimental to the natural course of typhoid pathogenesis, as introduction of the sopD2 gene from S. Typhimurium results in a decreased ability of S. Typhi to invade human epithelial cells [28]. Similar results are observed when the sseJ gene is reintroduced into S. Typhi [29].

The SPI-3 pathogenicity island is involved with the colonization and persistence of S. Typhimurium on the intestinal mucosa [30]. While the magnesium transport system of SPI-3 (mgtBC) is upregulated by S. Typhi inside macrophages [31], the remaining functions conferred by this system appear to interfere with the success of typhoid infection. Several genes in this system are pseudogenes in S. Typhi and re-introduction of a functional SPI-3 system into S. Typhi results in a dramatic decrease in the ability of the bacteria to survive in human macrophages [32]. These studies and continued research into the molecular basis for S. Typhi pathogenesis could provide crucial information leading to improved vaccine approaches.

Animal models for host-restricted serovars

Some of the most exciting research in the past five years has come in the area of typhoid host restriction. Recent studies have identified the phagosome of monocyte-derived cells (monocytes, macrophages and dendritic cells) as a key location driving host specificity. S. Typhi is able to manipulate the phagosomes of human (but not murine) cells to promote its own survival and to block antigen presentation, thereby initiating a successful infection [33, 34]. The SPI-1 effector protein GtgE, present in S. Typhimurium, is absent in S. Typhi. The lack of this protein has a profound effect on pathogenesis and appears to be at least partially responsible for the restriction of S. Typhi to human hosts [35, 36]. The GtgE protein of S. Typhimurium (and other broad host-range serovars) proteolytically inactivates the Rab29 GTPase recruited to the Salmonella-containing vacuole, allowing bacterial replication inside the cell [36]. The absence of this effector results in the rapid killing of S. Typhi inside non-human cells [35]. The identification of specific molecules that are responsible for the permissiveness of the host is an important breakthrough that will hopefully allow for the creation of accurate animal models for typhoid.

The lack of robust, reliable animal models has been one factor hampering the progress of vaccine development. Vaccine immunogenicity can be tested by intranasal administration of Typhi vaccine strains into BALB/c mice [37]. In a preclinical setting, this model is the current gold standard for assessing the immunogenicity of candidate strains. Despite its widespread use, this model suffers from the fact that it bypasses the oral compartment and provides no reliable measurement of strain reactogenicity or invasiveness in humans. However, several new animal models have been reported in recent years that may provide needed insight into developing vaccines from host-restricted Salmonella serovars such as Typhi and Paratyphi A.

Three humanized mouse models sensitive to S. Typhi infection have been reported [38–40]. In these models, immunodeficient mice (either Rag2⁻/⁻ γc⁻/⁻ or nonobese diabetic scid IL2rγ^null) are engrafted with human hematopoietic stem and progenitor cells (from fetal liver or cord blood). The resulting mice are susceptible to infection with wild-type S. Typhi and the resulting infection resembles the pathogenesis of typhoid in humans. These models will provide much-needed new avenues for increasing our understanding of important S. Typhi virulence factors, and they may also be useful for evaluating vaccines, since a phoPQ-attenuated S. Typhi stimulated serum antibody production against Salmonella in some of the infected mice, but was unable to cause disease [39]. The downside to these models is that they are expensive and labor-intensive, which will no doubt limit their use. In addition, there have thus far been no studies examining what happens when a host-restricted Salmonella serovar is administered orally. The establishment of a systemic infection after oral administration of S. Typhi would enhance the value of any small animal model.

A second murine typhoid model capitalizes on the differences between the murine and human innate immune systems to allow infection of mice by S. Typhi. Toll-like receptors (TLRs) play an important role in the innate immune system as sensors for pathogen associated molecular patterns as typified for gram-negative pathogens by lipid A (TLR4) and flagellin (TLR5) [41]. A recent report demonstrated that TLR11 is also a flagellin sensor [42]. TLR11 is produced in mice, but not in humans. Interestingly, mice deficient in TLR11 are susceptible to oral S. Typhi infection, leading to the hypothesis that TLR11 is a primary host factor responsible for the host-restricted phenotype of S. Typhi [42]. These TR11-deficient mice are otherwise immunocompetent and can develop protective immunity against S. Typhi infection by direct or passive immunization. In addition, S. Typhi pathogenesis was dependent on production of flagellin, consistent with the role of TLR11 as a flagellin sensor. One drawback of this model is that currently, TR11-deficient mice are not available commercially. While all of the mouse models we have mentioned have their weaknesses, this model, if commercialized, should be much less expensive than the humanized mouse models and is compatible with studies to evaluate orally administered Salmonella vaccines vectored by Typhi or Paratyphi A.

Finally, a rabbit model for evaluating immunogenicity and reactogenicity was described [43]. In that study, attenuated S. Paratyphi A strains were orally administered to New Zealand white rabbits. Interestingly, some of the strains elicited symptoms in the rabbit similar to those reported by human subjects in clinical trials, namely lethargy and loss of appetite. No fevers or diarrhea were reported for the rabbits. Rabbits inoculated with non-reactogenic strains such as Ty21a or the ΔphoPQ mutant Ty800, did not display any symptoms of reactogenicity. In addition, all strains elicited dose-dependent serum titers of anti-LPS IgG. Importantly, elevation of anti-LPS IgG serum titers required administration of live cells. While this model needs further development and characterization, it represents another useful tool for evaluating S. Typhi and S. Paratyphi A vaccine strains.

Recently described technologies for enhancing the immunogenicity and safety of Salmonella vaccines

In an effort to enhance the ability of Salmonella to survive within the host long enough to elicit an optimal immune response, Roy Curtiss and colleagues have developed several new approaches for vaccine development including regulated delayed attenuation, regulated delayed antigen synthesis and regulated delayed lysis (for a detailed review of these approaches, see [44]).

Regulated delayed attenuation

Vaccine strains with regulated delayed attenuation display wild-type characteristics when they are orally administered, providing them with a full complement of genes required to survive transit through the gastrointestinal tract and to carry out the initial stages of infection. Once inside host tissues, expression of specific virulence genes or attributes is turned off, resulting in a fully attenuated strain. Vaccine strains utilizing this strategy have been demonstrated to be safe and immunogenic in mice [45, 46]. One approach to this technology has been to utilize the arabinose-responsive araC P_BAD promoter to drive expression of Salmonella virulence genes, including crp, phoP, rpoS and fur [45]. When the vaccine is grown prior to administration, arabinose is added to the culture medium and the arabinose-regulated virulence gene(s) is expressed. After immunization, when the vaccine strain reaches host tissues where free arabinose is not present, the virulence genes cease to be expressed and their protein products are lost by dilution as the bacteria divides. Alternatively, deletion of the pmi or galE genes results in strains that are dependent on exogenous mannose or galactose, respectively, for the synthesis of O-antigen, a virulence factor critical for the survival of Salmonella in the host (reviewed in [13]). Cells are grown with the appropriate sugar prior to administration. Free mannose and free galactose are not present in host tissues and the cells gradually lose their O-antigen and become attenuated.

Regulated delayed antigen synthesis

A second technology, regulated delayed antigen synthesis, is designed to preclude or reduce the synthesis of the heterologous antigen until the vaccine has invaded host tissues. This is important because heterologous antigen synthesis consumes cellular resources, thus reducing the ability of the vaccine strain to grow and to cope with host defenses. There have been previous efforts by others in this area. Most approaches have placed the heterologous antigen under the control of a promoter that is typically turned on only in vivo, such as the ssaG or pagC promoters [47]. The strong, constitutive P_trc promoter was shown to be a good, but not ideal choice for driving antigen gene expression in vivo [48]. One interpretation of this result is that P_trc is suboptimal is because it is unregulated in Salmonella. To address this, Wang and colleagues introduced an arabinose-regulated lacI gene into the Salmonella chromosome, resulting in a strain that can reduce or eliminate antigen production when the strain is grown in the presence of arabinose [16]. When arabinose is absent (as in host tissues), synthesis of heterologous antigen begins and reaches maximum levels after approximately 9 cell divisions due to dilution of LacI [16]. This system can improve the host immune response to a heterologous antigen when the gene is expressed from the P_trc promoter. It provides the same level of immunogenicity as the in vivo-regulated promoter strategy, but is more flexible overall, since the P_trc promoter is compatible with a wider variety of attenuation strategies [47]. One drawback of this technology is that overproduction of LacI reduces Salmonella virulence [49]. Despite this, LacI does enhance the immunogenicity of an RASV carrying a P_trc-driven heterologous antigen gene, but reduces immunogenicity in strains carrying other promoters that do not bind LacI [47]. Thus, it is likely that the presence of the P_trc promoter on a multicopy plasmid titrates the LacI so that its negative effect on virulence and immunogenicity is suppressed.

Live attenuated Salmonella vaccine strains have been constructed that utilize both regulated delayed attenuation and regulated delayed antigen synthesis. For example, S. Typhimurium strain χ9558 has been designed to include arabinose-regulated crp and fur genes in addition to a Δpmi mutation [46]. This strain also carries a ΔsopB mutation to enhance Th2 responses [50], an arabinose-regulated lacI gene to control heterologous antigen gene expression, a ΔasdA mutation to allow use of the AsdA-based balanced-lethal plasmid maintenance system, as well as some additional mutations designed to reduce biofilm formation to enhance biocontainment. Immunization of mice with strain χ9558 carrying a protective antigen from Streptococcus pneumoniae, resulted in a greater level of protection against lethal S. pneumoniae challenge than a traditionally attenuated S. Typhimurium strain carrying the same antigen [46]. Strain χ9558 derivatives carrying Yersinia antigens have also been shown protect mice against lethal challenge with Yersinia pestis. A set of S. Typhi strains with a similar constellation of mutations was constructed [51] for evaluation in a clinical trial (Table 1).

Table 1.

Clinical trials evaluating Salmonella–vectored vaccine strains.

Serovar	Strain	Attenuating mutations	Antigen	Plasmid stabilization	Antigen gene promoter	Study type	Reference
Typhi	TSB7	ΔaroC ΔssaV	LT-B	Not applicable	PssaG	Open label	[79]
Typhi	χ9639(pYA4088)	RpoS− and footnote 1	PspA	AsdA+	Ptrc regulated	Double blind	[102]
Typhi	χ9640(pYA4088)	RpoS+ and footnote 1	PspA	AsdA+	Ptrc regulated	Double blind	[102]
Typhi	χ9633(pYA4088)	RpoS+ and footnote 2	PspA	AsdA+	Ptrc regulated	Double blind	[102]
Typhi	CVD908-htrA	ΔaroC ΔaroD htrA thyA	OprF-OprI	ThyA+	Pyz	Not reported	[87]
Typhi	CVD908-htrA	ΔaroC ΔaroD htrA thyA	OprF-OprI	ThyA+	PpagC	Not reported	[87]
Typhi	Ty21a	galE thyA Vi− Multiple mutations	OprF-OprI	ThyA+	Pyz	Not reported	[87]
Typhi	Ty21a	galE thyA Vi− Multiple mutations	OprF-OprI	ThyA+	PpagC	Not reported	[87]
Typhi	Ty21a	galE thyA Vi− Multiple mutations	Urease	ThyA+	Undescribed constitutive promoter	Double blind	[86]
Typhi	Ty21a	galE thyA Vi− Multiple mutations	HP0231	ThyA+	Undescribed constitutive promoter	Double blind	[86]
Typhi	VXM01 (Ty21a)	galE thyA Vi− Multiple mutations	VEGFR-2	Not described	CMV	Double blind	[89]
Typhimurium	SalpIL2	Δcya Δcrp ΔasdA	Human IL-2	AsdA+	Ptrc constitutive	Open label	[103]

¹ΔP_crp527:TT araC P_BADcrp ΔP_fur81:TT araC P_BADfur Δpmi-2426 Δ (gmd-fcl)-26 ΔsopB1925 ΔrelA198:araC P_BADlacI TT ΔaraE25 ΔtviABCDE10 ΔagfBAC811 ΔasdA33

²ΔP_crp527:TT araC P_BADcrp ΔP_fur81:TT araC P_BADfur Δpmi-2426 Δ(gmd-fcl)-26 ΔsopB1925 ΔrelA198:araC P_BADlacI TT ΔaraE25 ΔaraBAD 23 ΔtviABCDE10 ΔagfBAC811 ΔasdA33

Regulated delayed lysis

Another recently described technology is the design of vaccine strains that undergo regulated delayed lysis [52]. This system was developed to address a number of goals, including biocontainment, release of non-secreted protein antigens and DNA vaccine delivery. In this system, the vaccine strain features strict arabinose-controlled expression of two genes required for peptidoglycan synthesis, asdA and murA. The two genes are effectively complemented by a plasmid that carries both genes, along with an antigen gene of interest. The P_trc promoter drives antigen gene expression, for delivery of protein antigens, while eukaryotic promoters such as P_CMV can be used for DNA vaccine delivery. For DNA vaccines, the plasmids have been further modified to include nuclear targeting sequences and to enhance stability inside host cells and a number of additional mutations are included in the Salmonella strain to facilitate efficient DNA delivery [53]. Results from preclinical trials demonstrate the utility of this system, but it has not yet been evaluated in a clinical setting.

Reducing reactogenicity by lipid A modification

Kong and associates have taken a unique approach to enhancing RASV safety my modifying the lipid A (endotoxin) in living cells to a less toxic form, while retaining its adjuvant properties [54]. The significance and impact of this strategy was recently reviewed [55].

Sugar-inducible acid resistance

Another novel system is focused on increasing the acid resistance of S. Typhi vaccine strains. The basis of this approach is that, to be effective, a live attenuated Salmonella vaccine must first survive its encounter with the low pH environment of the stomach. While wild type salmonellae have evolved mechanisms to survive this environmental insult long enough to pass into the intestinal tract, many of the means used to attenuate Salmonella for virulence have a secondary effect of increasing sensitivity to acid. Additional problems arise because the model organism, S. Typhimurium, is inherently more acid resistant than S. Typhi [56]. Also, the gastric pH in a fasted mouse is around 4.0 [57], while the gastric pH in a fasted human is around 2.0 [58]. This difference could account, at least in part, for the discrepancies between results obtained in clinical studies and results obtained using the S. Typhimurium mouse model system. It may also help to explain why very high doses of S. Typhi vaccines are required to achieve an efficacious immune response, while lower doses of S. Typhimurium work well in mice. Because of the low gastric pH in humans, oral Salmonella vaccines are typically given with an agent designed to increase the gastric pH, such as bicarbonate. While this approach is helpful, it precludes the Salmonella vaccine from sensing an important environmental signal, low pH, signifying its entry into a host.

Enteric bacteria utilize a number of means to deal with low pH, including several amino acid decarboxylase systems [59]. However, these systems are often induced only by anaerobic growth at low pH. These growth conditions are not typically used for vaccine preparation, so S. Typhi cells are physiologically unprepared for passage through the stomach. To address this problem, we have constructed S. Typhi vaccine strains such that expression of adiA and adiC, encoding arginine decarboxylase and the arginine-agmatine antiporter, is driven by a sugar-inducible promoter [60]. Addition of the appropriate sugar results in expression of adiA and adiC and a concomitant increase in acid resistance. Our in vitro results demonstrated that this system results in significant increases in survival for ΔaroD and ΔphoPQ S. Typhi strains at both pH 3.0 and pH 2.5 [60]. We are currently examining the effects of other amino acid decarboxylases using a similar approach. Preliminary animal data indicates that inclusion of these systems into attenuated S. Typhimurium results in a significant enhancement of immunogenicity in mice (Brenneman and Roland, unpublished).

Plasmid stabilization systems

In most attenuated Salmonella vectored vaccines, heterologous antigens are carried on multicopy plasmids (but not all – see TSB7, Table 1). This is done to ensure that a high level of antigen synthesis is achieved to adequately stimulate the desired immune response. Stable maintenance of plasmids in vaccine strains is required to assure that every cell in the population is capable of producing and delivering the heterologous antigen. To obviate the need for antibiotic selection, a number of non-antibiotic stabilization systems were developed, beginning in 1990 with the AsdA⁺ balanced-lethal system [61], which has been extensively used in preclinical and clinical trials. Others include the ThyA⁺ system [62], used in some recent trials (Table 1), PurB⁺ [63], GlnA⁺ [64] and the ORT balanced-lethal system [65]. Although all of these systems have not been directly compared, in a recent preclinical study, the AsdA⁺ GlnA⁺ and PurB⁺ selection systems were evaluated in isogenic ΔphoPQ S. Typhimurium strains. The strains carried plasmids encoding the Yersinia pestis F1-V antigen that differed only in the gene used for plasmid selection [66]. Mice were given two doses of vaccine (days 0 and 14) and anti-F1-V serum IgG responses were determined. On days 28 and 42, the IgG titers elicited by immunization with the strain using the AsdA⁺ system were >3-fold higher than mice in the other two groups (ca. 2,000 vs. 650 and 380). These results indicate that the AsdA⁺ system led to more rapid serum responses than the other two. This is not necessarily surprising, since loss of a GlnA⁺ or PurB⁺ plasmids, as well as a ThyA⁺ plasmid, does not necessarily lead to cell death, only a cessation of growth. In fact, glnA mutants of S. Typhimurium are not attenuated in mice [67]. Thus it is not clear what impact the loss of a GlnA⁺ plasmid would have in vivo. It would be interesting to determine how the AsdA⁺ system compares to the ThyA⁺ system, since this system was used to stabilize plasmids in several recent clinical trials (Table 1).

A novel plasmid stabilization system was recently described by Galen and associates, which relies on complementation of a deletion in the gene ssb that encodes single strand binding protein [68]. Because ssb is an essential gene, loss of the antigen-encoding plasmid results in cell death. This system results in 100% plasmid retention. However, antigen load can still be a problem, resulting in strains that retain the plasmid, but lose the ability to synthesize the heterologous antigen. The authors demonstrated that antigen synthesis from lower copy number plasmids eliminated this problem, resulting in a strain with highly stable antigen expression. In their study, they evaluated the PA83 antigen from Bacillus anthracis. The antigen gene was fused to clyA to facilitate protein secretion [69] and expressed from ssb plasmids that also encoded a partition function to further enhance plasmid stability. The plasmids were introduced into S. Typhi strains derived from CVD908 and CVD908-htrA. Mice were intranasally immunized with two doses and serum IgG responses against PA83 were measured. Strains carrying plasmids with the lowest copy number (pSC101) elicited significantly higher anti-PA83 titers than strains carrying a higher copy number plasmid (p15a). For strains carrying the pSC101 plasmid, on day 41, prior to parenteral boosting with purified antigen, mice immunized with CVD908-htrA produced GMTs of 2,322 and mice immunized with CVD908 produced GMTs of 4,929. Six days after parenteral boosting on day 42, mice primed with CVD908 and CVD908-htrA produced remarkably high titers of 500,970 and 156,330, respectively, while the titers achieved with the higher copy number p15a plasmids were 59,247 and 13,806, respectively. These data support the testing of this system in a clinical trial.

Recent clinical trials - vectors

We have identified six clinical trials, published or in progress since 2007 that investigate twelve candidate Salmonella vaccines delivering heterologous antigens (Table 1). Before we discuss heterologous antigen delivery, we will first describe a few interesting studies focused on enhancing our understanding of host-restricted Salmonella delivery vectors. The group at the Center for Vaccine Development (CVD) in Maryland, USA has taken a methodical approach to Salmonella vector development, carefully assessing their attenuated strains in clinical trials and making minor, but rational strain modifications along the way. They began by evaluating strains with aro mutations and later added a mutation in htrA to achieve their current safe and immunogenic vector strain, CVD908-htrA (see ref [70] for a complete review of these strains).

In general, live S. Typhi vaccines do not elicit antibodies against the Vi antigen, most likely due to the tight regulation of Vi synthesis in S. Typhi (reviewed in [71]). However, high titers of anti-Vi IgG protect against infection. To overcome this problem, Levine and associates genetically modified CVD908-htrA to constitutively produce Vi antigen, resulting in strain CVD909 [72] (Table 2). While this strain is does not vector a heterologous antigen, we include it in our discussion here because the strain was manipulated to bypass natural Vi regulation so that capsule is constitutively synthesized in order to quickly stimulate the host immune system.

Table 2.

Clinical trials to evaluate live attenuated strains as paratyphoid or typhoid fever vaccines.

Serovar	Strain	Mutations	Comments	Reference
Paratyphi A	CVD1902	guaBA clpX		[103]
Typhi	CVD909	aroC aroD htrA Ptac tviA	prime boost	[72, 75]
Typhi	M01ZH09 (ZH9)	ΔaroC ΔssaV	Phase I, US volunteers	[91]
Typhi	M01ZH09	aroC ssaV	Phase I, US volunteers	[95]
Typhi	M01ZH09	aroC ssaV	Phase II, US volunteers	[96]
Typhi	M01ZH09	aroC ssaV	Phase II, Vietnamese children	[97]

When administered to volunteers, this strain elicited antibody-secreting cells against Vi, but only 2 of 32 volunteers produced anti-Vi serum IgG [73]. Importantly, the constitutive production of Vi did not diminish the ability of CVD909 to elicit T cell responses against Typhi-specific antigens, although these responses were not boosted by a second dose [74]. Building on a number of preclinical studies showing the priming capacity of Salmonella-vectored vaccine strains (e.g. [75, 76]), CVD909 was used to immunize volunteers who then received a parenteral boost with purified Vi [77]. While bacterially primed subjects produced higher levels of anti-Vi IgG and IgA, statistical significance was not achieved. However, priming did significantly enhance the production of memory B cells, indicating that this approach may lead to a longer-lived, more robust protection against typhoid fever than current vaccines.

As part of their ongoing work to develop a live attenuated Salmonella vaccine against enteric fever, the group at CVD has recently initiated a clinical trial to evaluate attenuated S. Paratyphi A strain CVD1902 [101] (Table 2). Although S. Paratyphi A is closely related to S. Typhi, immune responses to S. Typhi vaccine Ty21a are not cross protective [78]. Prior to this study, no clinical trials have been conducted using S. Paratyphi A, so it will be interesting to see how this serovar behaves in humans when attenuated.

Recent clinical trials – heterologous antigen delivery

S. Typhi strain M01ZH09 (previously ZH9) has been evaluated in Phase I and Phase II clinical trials for use as a typhoid vaccine and was found to be safe and immunogenic (Table 2). Khan and associates described an S. Typhi M01ZH09 derivative that carries a chromosomally integrated eltB gene, encoding the B subunit of heat-labile toxin from enterotoxigenic Escherichia coli [79]. The eltB gene was inserted into the aroC gene site, under transcriptional control of the ssaG promoter, so that eltB expression was upregulated in vivo [80]. In their study, 12 volunteers received two doses of 10⁸ CFU and 24 volunteers received two doses of 10⁹ CFU. The vaccine was generally well tolerated, although most volunteers experienced mild to severe intestinal symptoms that resolved within a few days. The vaccine was shed for up to 3 days, with a single volunteer shedding for 6 days. Importantly, 67% of volunteers produced an IgG or IgA immune response against heat labile toxin (LT) as determined by ELISA (IgG) or ELISPOT (IgA), with most seropositive volunteers producing anti-LT IgG. 97% of volunteers produced IgG and/or IgA responses to S. Typhi lipopolysaccharide (LPS), with most volunteers responding to both. In addition, there was evidence of first dose priming of LT responses. These data indicate that M01ZH109 is a strong candidate as both a live typhoid vaccine and as an antigen delivery vector.

As described above, the Curtiss group has developed a number of new technologies for enhancing the effectiveness and safety of Salmonella vaccines To assess the impact of some of these technologies in the clinic, they compared three S. Typhi strains that carry nine or ten mutations each to evaluate their ability to deliver the pneumococcal protein PspA [102](Table 1). Many of the attenuating mutations in these strains are based on their regulated, delayed attenuation strategy [45, 51] (see below) and pspA was carried on an AsdA⁺ plasmid. In their trial, volunteers were given escalating doses from 10⁷ to 10¹⁰ CFU. Their study was also designed to evaluate the impact of RpoS on immunogenicity [45, 51]. This is an important question, since most S. Typhi vaccine strains are derived from strain Ty2, an rpoS mutant and rpoS is a virulence factor for S. Typhimurium [81]. The outcome of this completed trial has not yet been published.

Ty21a – new uses for an old vaccine

Recently, there has been interest in expanding the uses of the licensed vaccine strain, Ty21a. Despite the fact that the immunogenicity of Ty21a is less than ideal, a number of groups have been looking more closely at it for a number of different purposes. While there are a number of demonstrably more immunogenic live strains (e.g. CVD908-htrA), we suspect that the focus on Ty21a is driven by market forces (time to licensure, cost of licensing, Phase III trials, etc. for a new vaccine vs. potential profits, availability to target populations). Thus, finding new uses for Ty21a may represent the most rapid, least expensive path toward reaping potential health benefits. Several reports have examined the potential of Ty21a to elicit cross protective immune responses against S. Paratyphi A, S. Paratyphi B [78, 82] and non-typhoid salmonellae [83]. In addition, several groups have modified Ty21a to deliver heterologous antigens in clinical trials that are described below (see Table 1). A preclinical study demonstrated the ability of Ty21a to secrete and deliver the Bacillus anthracis protective antigen, PA [84]. A low copy number plasmid was constructed in which the gene encoding PA was transcribed from an in vivo-inducible promoter and secreted via the HlyA type I secretion system. Mice immunized intranasally produced high serum IgG titers against PA and were protected from an aerosol spore challenge.

Several groups have used the clinic to compare multiple vaccine candidates in a single study. Aebeischer and associates utilized a thyA derivative of Ty21a to deliver two different Helicobacter pylori antigens, UreAB (urease) and HP0231, a protective antigen related to E. coli DsbG [85], whose genes were carried on plasmids utilizing a ThyA⁺ selection system [86]. The study was set up in two parts, a preliminary trial with 20 subjects and a secondary study using 47 subjects. Volunteers received three doses (first trial) or four doses (second trial), with two days between doses. In the first trial, volunteers were challenged with H. pylori five months later. By seven days post-challenge, all volunteers were positive for H. pylori as determined using the [¹⁴C] urea breath test (UBT). By week six, three of nine vaccinated volunteers were H. pylori negative (33%) as judged by the UBT and three other diagnostic tests including histological analysis of biopsy samples. All four control individuals were positive by one or more test, indicating that vaccination provided some degree of protection. In the second trial, volunteers were challenged 42 days after the primary immunization. Although H. pylori clearance by all four diagnostic tests was not demonstrated in any volunteer, some evidence of protection using the UBT was achieved with both vaccine candidates. With regard to immune responses, on day 5 post-vaccination, urease-specific CD4⁺ T cells were detected in 25% of volunteers in the first study and 8% of volunteers in the second study. No antigen-specific CD8⁺ cells were detected. The authors also reported that PBMC isolated from some volunteers could be stimulated with purified antigen to produce IFNγ. However, no evidence of antigen-specific B cell responses was detected, although anti-Salmonella responses were detected in some volunteers. Taken together, these data indicate that the antigens were detected and processed by the host immune system, though efficacy remains in question.

Bumann and associates tested the potential of Salmonella–based vaccines to protect against airway infections by Pseudomonas aeruginosa (Table 1). They compared the ability of two antigen delivery strains, ThyA⁻ derivatives of CVD908-htrA and Ty21a, to deliver an OprF-OprI fusion protein [87] shown to protective against P. aeruginosa infection in animal studies [88]. The genes for the OprF-OprI antigen were carried on a ThyA⁺ plasmid. There were two plasmid variants that differed only in the promoter driving transcription of the Pseudomonas gene. The antigen gene was transcribed from either an artificial, constitutive promoter (designated Pyz), or from the in vivo-inducible pagC promoter. Volunteers were inoculated with either a single dose of the CVD908-htrA-vectored vaccines (10⁸ CFU/dose) or 3 doses of the Ty21a-vectored vaccines (10¹⁰ CFU/dose), following the standard schedule for Ty21a, days 0, 2 and 4. The study also included control groups that received the purified OprF-OprI antigen, delivered either by injection with an aluminum adjuvant or intranasally in a gel without adjuvant. Twenty-eight days after the primary inoculation, all groups received an injection of OprF-OprI in adjuvant. The groups given the Salmonella–vectored vaccines failed to mount a serum antibody response against the P. aeruginosa antigen after the initial immunization, while both groups receiving purified antigen developed a 1000-fold increase in IgG and a 100-fold increase in IgA titers to OprF-OprI on day 28. After the booster injection, all groups produced similar high titers of serum IgG and IgA against the antigen. Interestingly, both Salmonella–vectored vaccines elicited similar, significant anti-OprF-OprI IgG and IgA responses in induced sputum samples, a measure of the mucosal responses in the lower airways. Similar mucosal responses were observed in subjects immunized intranasally, but not in those immunized systemically by injection. Systemic boosting did not boost the mucosal responses in any group. Surprisingly, all four Salmonella vaccines elicited similar immune responses, regardless of differences in attenuation strategies and promoters driving antigen gene expression. However, volunteers in the CVD908-htrA group received only a single dose compared to 3 doses administered in the Ty21a group, indicating that CVD908-htrA was more immunogenic. These results are consistent with previous animal studies indicating that Salmonella–vectored vaccines induce strong mucosal responses, but weak systemic responses. While cellular responses were not directly measured, no differences were observed in IgG1/IgG2a ratios. These results are encouraging as they demonstrate that it is possible to generate a mucosal antibody response in the respiratory tract after oral administration of a Salmonella vaccine.

In a unique application, Ty21a is being investigated as a DNA vaccine delivery vector. In this scheme, Ty21a carries a plasmid encoding a humanized, full length VEGFR-2, which encodes vascular endothelial growth factor receptor 2 [89] for use as an anti-angiogenic cancer therapeutic vaccine (Table 1). In preclinical studies using an S. Typhimurium aroA mutant as the delivery vector, a similar construct, using a mouse-derived gene, was effective at protecting mice against challenge with melanoma, colon carcinoma and lung carcinoma cells [90]. Importantly, it reduced the growth of established metastases. In the clinical trial, the vaccine was administered to patients with stage IV and advanced pancreatic cancer in a Phase I trial. The results of this study are not yet available, but it represents the first human trial to evaluate DNA delivery by Salmonella.

S. Typhimurium vaccines

Saltzman’s group has been evaluating an attenuated S. Typhimurium strain expressing a fragment of human IL-2 (SalpIL2) as a therapeutic cancer vaccine [91, 92]. The gene for IL-2 is carried on a low copy number plasmid stabilized using the AsdA⁺ balanced-lethal system. In preclinical trials, this strain has been shown to home to tumor cells, to be highly immunogenic and to reduce tumor burden and tumor-associated systemic inflammation, while increasing proliferation of tumor-targeting cytotoxic NK cells [91]. Due to the success of this novel approach, SalpIL2 is currently being evaluated in a dose escalation study in humans. The outcome of this trial will be of interest for several reasons. This strain may provide a new tool to be used for treating a particularly devastating form of cancer. However, as mentioned before, S. Typhimurium is not well adapted to achieve systemic spread, so it is not clear whether the vaccine will be able to reach the tumor in a human host. In addition, only two trials using a Typhimurium vector have been reported. The strains in these trials were attenuated by different means, ΔphoPQ [63], ΔaroC ΔssaV [93] and, in the current trial, Δcya Δcrp [101]. While the first two attenuation strategies were shown to be attenuating in S. Typhi, a Δcya Δcrp S. Typhi strain was too reactogenic to be considered for further development [94]. Thus, it will be of interesting to see how effective the Δcya Δcrp strategy is for attenuating S. Typhimurium in human volunteers.

Key Issues.

1. S. Typhi vaccine vectors do not produce strong immune responses in humans. This is true not only for typhoid antigens, but especially for vectored heterologous antigens.

2. A broader understanding of S. Typhi pathogenesis allows researchers to better understand how to attenuate vaccine strains.

3. Predictive correlates for human immunity and reactogenicity must be established to provide a basis for interpreting preclinical data.

4. Development of alternative, relevant animal models will lead to more immunogenic, safer Salmonella-vectored vaccines. Use of these new models will allow more effective screening of vaccine candidates, which should improve the chances for successful clinical trials and development of effective vaccines.

5. Recent research has introduced a variety of new vaccine approaches, including attenuation strategies, antigen synthesis and antigen delivery mechanisms.

6. Recent research has also elucidated some of the key differences between S. Typhi and S. Typhimurium, allowing researchers to design vaccines that take these differences into account.

Appendix

Expert Commentary

The need for inexpensive, mucosally applied vaccines to serve the underdeveloped world has been a driving force in our laboratory. Salmonella-vectored vaccines are ideally suited for this task. Many of the vaccines for the underdeveloped world must be suitable for administration to children. Thus, efforts toward development of a protein-antigen-based Salmonella–vectored pneumococcal vaccine has led the Curtiss group to develop a number of new strategies to enhance safety and immunogenicity. The need to understand the safety and immunogenicity of RASVs in infants and children has led a number of groups to carefully examine the impact of their vaccines in infant and baby mice. An expansion of our knowledge of mucosal immunology in the young, malnourished and immunocompromised has supported efforts to develop vaccines for the developing world.

While there are still crucial problems yet to be solved before RASVs are effective enough to be commercially viable, recent clinical trials have served to expand our knowledge base of S. Typhi vectored vaccines and to demonstrate their many potential applications. As outlined in Table 1, RASV technology has been applied to vaccines against traveller’s diarrhea, infant pneumonia, Pseudomonas aeruginosa infections in cystic fibrosis patients, Helicobacter pylori infections and cancer. These trials demonstrate the ability of RASVs to elicit mucosal responses and to serve as effective immunological priming agents in human volunteers. Careful analyses of mucosal and cellular responses in vaccinees will drive our understanding of how these vaccines may best be used to unlock their enormous potential.

5-year view

The biggest advances in the field will come as the new animal models are applied to the existing pool of potential Salmonella-based vaccines. Right now, the biggest restriction on the field is the amount of time and effort it takes to evaluate a single candidate vaccine. Researchers spend years to gather pre-clinical data, apply to the FDA and then perform a relatively straightforward experiment, the clinical trial, to see if their original hypothesis about vaccine design was correct. This is the main reason that there are so few clinical Salmonella vaccine trials. The use of humanized mouse or rabbit models will allow researchers to rapidly generate data on the actual vaccine strain, as opposed to a S. Typhimurium surrogate, to serve as a relevant indicator of what will happen in the clinic. That will allow the entire field to move much more quickly and have the potential to explore other long-standing problems in the field, such as how to direct the attention of the host immune response to the heterologous antigen and ignore Salmonella. Research conducted in the next five years should provide us with long-awaited solutions to many of the fundamental problems facing the field.