Listeria monocytogenes is a gram-positive bacterium that is commonly associated with gastrointestinal infections through the consumption of contaminated food. As a gastrointestinal infection, Lm infects epithelial cells in the small intestine through the function of a virulence factor known as Internalin A (InlA) (Gaillard et al., 1991). InlA specifically interacts with E-cadherin expressed on the basolateral surface of polarized epithelium to facilitate invasion into a cell (Pentecost et al., 2006). After invading the gut epithelium, Lm then migrates to other organs such as the spleen where it is phagocytized by antigen-presenting cells (APCs). Inside the cell, a prfA-mediated virulence gene program in Lm facilitates its survival and propagation (Freitag et al., 1993). PrfA is a Lm-specific transcriptional activator of many virulence factors such as the pore-forming toxin Listeriolysin O (LLO) and phospholipases that coordinate to compromise the integrity of phagosomal membranes and allow for Lm escape into the cytosol of the infected cell (Mengaud et al., 1987). Once it has escaped from the phagosome, Lm propels through the cytosol through the action of ActA, a protein that interacts with the Arp2/3 complex to activate the nucleation of actin filaments. It is this cytosolic motility of Lm that allows it to impact the cellular membrane with enough force to form a protrusion that is recognized and internalized by a proximal cell, therefore, disseminating the infection (Tilney and Portnoy, 1989).

The host response to infection by Lm is evident in both the strong innate response to the intracellular pathogen during an acute infection and the generation of strong adaptive responses to counter future challenges with Lm. The innate response to Lm is initially mediated by surface toll-like receptors (TLRs) that recognize bacterial pathogen-associated molecular patterns (PAMPs) from Lm such as lipoteichoic acid (Machata et al., 2008; Noor et al., 2008). Activation of TLRs by Lm results in downstream signaling through the MyD88 adaptor protein and production of proinflammatory cytokines (Edelson and Unanue, 2002). Once phagocytized, Lm may then be processed in the phagolysosomal compartment and peptides presented on MHC Class II for activation of Lm-specific CD4⁺ T cell responses. Alternatively, Lm can escape the phagosome and enter the cytosol where recognition of peptidoglycan by nuclear oligomerization domain-like receptors (NLRs) and Lm DNA by the DNA sensor, AIM2, activates inflammatory cascades (Kobayashi et al., 2005; Hasegawa et al., 2006; Park et al., 2007; Sauer et al., 2010; Warren et al., 2010). Once in the cytosol, Lm also secretes protein antigens that are processed and presented by MHC Class I molecules to CD8⁺ T cells for robust induction of Lm-specific CTL responses that provide protection against a subsequent exposure (Brunt et al., 1990). This combination of inflammatory responses and efficient delivery of antigens to the MHC I and MHC II pathways makes Lm a powerful vaccine vector.

The ability of Lm to induce strong cellular immune responses has led to its use for decades as a tool to dissect the mechanisms of immune memory formation and function (North, 1973; Stolpmann et al., 1985; Shedlock et al., 2003; Shen et al., 2003). However, it was not until several studies in the early 1990s that the enormous translational potential of Lm as an inducer of antigen-specific CTLs was realized (Schafer et al., 1992; Ikonomidis et al., 1994; Frankel et al., 1995; Pan et al., 1995a,b). In the initial study, Paterson and colleagues constructed recombinant strains of Lm that expressed a foreign antigen, E. coli β-galactosidase (β-gal) (Schafer et al., 1992). When mice were immunized either parenterally or orally with these constructs, β-gal-specific CTLs were induced and delayed type-hypersensitivity reactions to β-gal were observed in mice receiving the recombinant Lm strain but not a wild-type Lm strain. Building on this success, Paterson and colleagues further developed and refined the Lm vaccine platform with a construct that expressed a viral antigen, influenza nucleoprotein (NP) (Ikonomidis et al., 1994). However, this Lm vaccine sought to improve the processing and presentation of the foreign antigen by use of a strategy still used in Lm-based vaccines to this day. Unlike the β-gal Lm construct, the Lm-NP construct expressed the antigen as a fusion protein with the first 420 amino acids of Listeriolysin O. This LLO-NP fusion protein lacked the cholesterol-binding domain at the C-terminus of LLO necessary for pore formation and was, therefore, detoxified while still containing the signal sequence at the N-terminus allowing for efficient secretion (Michel et al., 1990). The rationale behind this approach was that the improved secretion of the foreign antigen into the cytosol of infected cells would increase the efficiency of processing and presentation to match the immunogenicity of endogenous Lm antigens. In fact, Lm-NP infected cells were able to efficiently process and present NP peptides for recognition by NP-specific CTLs. Consequently, Lm-NP constructs induced a highly functional CTL response against NP that was dependent on the ability of the constructs to deliver antigen to the cytosol, as LLO-deficient constructs were ineffective. The major advancement from this work of improving the processing and presentation of foreign antigens by Lm-based vaccines allowed for future application of this technology in CTL-mediated tumor immunotherapy.

After confirming the potential of Lm-based vectors to induce robust CTL responses against foreign antigens, Paterson and colleagues applied this technology to the then nascent field of tumor immunotherapy (Pan et al., 1995a). Much like infected cells, tumor cells can be recognized and lysed by antigen-specific CTLs (Freedman et al., 1972). Therefore, to assess the potential for Lm-based vaccines to induce effective anti-tumor CTL responses, the team utilized the previously constructed Lm-NP vaccine strain along with tumor cell lines that expressed NP as a model tumor antigen (Pan et al., 1995a). When mice were vaccinated with the Lm-NP vaccine, it provided prophylactic protection against tumor formation after challenge. In addition, after first allowing for tumors to form and establish, therapeutic intervention was effective at reducing tumor volume. This regression by Lm-NP correlated with infiltration of T cells into the tumor after vaccination and was abrogated after antibody-mediated depletion of T cells. A subsequent study by Jensen and colleagues also found Lm-based vaccines to be effective as a prophylactic against cottontail rabbit papilloma virus (CRPV) induced papillomas, a model for virus-induced malignant transformation (Brandsma, 1994). After previous success generating virus-specific CTL responses with recombinant Lm, they developed a novel Lm strain expressing the CRPV E1 antigen, E1-rLm, that induced cellular immune responses against E1 and protected the rabbits from papilloma formation (Shen et al., 1995; Slifka et al., 1996; Jensen et al., 1997). Therefore, the summation of these studies validated Lm as a viable platform for the delivery of TAAs in both a prophylactic and therapeutic setting. The promise demonstrated by these initial vaccines resulted in the development of numerous Lm-based immunotherapies, many outlined in this review, that continues to this day.

Listeriolysin O is a cholesterol-dependent cytolysin that, upon secretion by Lm, binds to cholesterol-containing membranes and forms a barrel resulting in a membrane pore (Shatursky et al., 1999). LLO-mediated pore formation results in reduced phagosome membrane integrity and aids the escape of Lm from the phagosomes of APCs (Geoffroy et al., 1987; Tilney and Portnoy, 1989; Beauregard et al., 1997). While the acidic environment of the phagosome increases the lytic activity of LLO for this purpose (Glomski et al., 2002), others have also found extracellular secretion of LLO by Lm that directly lyses T cells (Carrero et al., 2004). In a therapeutic setting, LLO has proven a powerful adjuvant but only after this lytic activity is reduced though “detoxification” with mutation or truncation of its cholesterol-binding domain (Michel et al., 1990). The first use of a detoxified LLO (dtLLO) in a tumor immunotherapy setting was in the context of a genetic fusion with a tumor-specific antigen (Pan et al., 1995a). Fusion of the antigen to dtLLO and secretion by Lm was found to significantly enhance anti-tumor immune responses in comparison to an Lm vector that secreted only the tumor antigen (Gunn et al., 2001). The likely mechanisms for this adjuvancy effect of dtLLO are several. The N-terminus of dtLLO contains an N-degron, a destabilizing residue that mediates processing by the Ub-proteasome machinery (Schnupf et al., 2006, 2007). Therefore, it was hypothesized that tumor antigens genetically fused to the C-terminus of dtLLO may have enhanced Ub-proteasome-mediated processing and presentation by APCs for activation of antigen-specific CTLs. More recently, several studies have inspected the potential for dtLLO to function as a PAMP and stimulate the necessary proinflammatory responses for effective anti-tumor adaptive immune responses (Peng et al., 2004, 2007; Wallecha et al., 2013b). A study by Wallecha et al found that purified dtLLO administered as a fusion protein with human papilloma virus type 16 E7 protein (HPV16 E7), a TAA expressed by HPV-transformed malignancies, or unfused in a mixture with HPV16 E7 can augment anti-E7 CD8⁺ T cell responses and therapeutically impact growth of murine tumors that express E7 (Wallecha et al., 2013b). To determine if the mechanism of adjuvancy was through PAMP-like activity, bone marrow-derived dendritic cells (BMDC) from wild-type mice were stimulated in vitro with dtLLO. dtLLO stimulation resulted in induction of proinflammatory cytokines such as IL-12 and increased expression of markers associated with BMDC maturation. While previous studies have suggested that PAMP-like activity of LLO and other cytolysins is dependent on TLR4 (Park et al., 2004), dtLLO treatment resulted in TLR4-independent induction of transcript for proinflammatory cytokines and costimulatory molecules suggesting additional mechanisms for this activity. These results suggest that the fusion or conjugation of dtLLO to antigen is likely not required to improve immunogenicity as previous studies with DNA vaccines suggested but this is yet to be determined in an Lm-based vaccine approach (Peng et al., 2007).

ActA production by Lm is essential to its cytoplasmic motility and cell-to-cell spread due to its ability to nucleate actin through the Arp2/3 complex (Welch et al., 1998). However, much like dtLLO, it is a powerful adjuvant as its fusion to Lm-secreted antigens results in enhanced immunogenicity (Sewell et al., 2004a, 2008; Souders et al., 2007). In work published by Sewell et al, secretion of HPV16 E7 fused to ActA resulted in significantly enhanced immunogenicity of E7 resulting in effective anti-tumor immune responses (Sewell et al., 2004a). The use of ActA as an adjuvant is not only restricted to Lm-based vaccines, as it also enhances antigen-specific anti-tumor immune responses in DNA and protein vaccine formulations (Wood et al., 2010). The promising potential of ActA in Lm-based vaccines was demonstrated by its ability to activate highly effective anti-tumor immune responses even to tolerized antigens (Souders et al., 2007; Sewell et al., 2008). Fusion of ActA to an antigen may increase immunogenicity through increased antigen processing and presentation due to the presence of several putative PEST domains in the N-terminus of ActA. However, ActA is a relatively stable protein (Moors et al., 1999). In support of an alternative mechanism, a subsequent study suggests the adjuvant properties of ActA are independent of fusion to the antigen (Sewell et al., 2004b; Wood et al., 2010). In fact, ActA protein appears to have PAMP-like properties due to its ability to induce proinflammatory cytokines and maturation of BMDCs (Wood et al., 2010).

The discovery of cyclic di-AMP as the Type I interferon (IFN)-inducing molecule by Lm was critical to our understanding of the host response to Lm (Woodward et al., 2010; Schwartz et al., 2012; Witte et al., 2013). Cyclic di-AMP is a small dinucleotide secreted by Lm once it escapes the phagosome of an infected cell and enters the cytosol. Within the cytosol, Lm-derived cyclic di-AMP activates a mitochondrial transmembrane receptor, STING, leading to Type I IFN production (Ishikawa and Barber, 2008). Type I IFN induced by Lm and other intracellular facultative bacteria can increase their acute pathogenicity and recent evidence suggests it may even hinder formation of adaptive immunity to Lm (Auerbuch et al., 2004; O'connell et al., 2004; Archer et al., 2014). However, there is a growing body of evidence emphasizing the importance of Type I IFN signaling in tumor immunosurveillance and the formation of effective anti-tumor adaptive immunity (Diamond et al., 2011; Fuertes et al., 2011). Therefore, methods that modulate cyclic di-AMP in Lm-based tumor immunotherapy may be promising but will require a delicate balance between its ability to impair or enhance anti-tumor immunity.

In order to accomplish stable attenuations of Lm vaccine strains, groups have employed strains containing deletions of genes involved in Lm virulence. The most common Lm vaccine strain, XFL-7, lacks the Lm transcriptional activator PrfA (Gunn et al., 2001). PrfA regulates the expression of numerous virulence factors such as LLO, therefore, strains lacking the prfA gene are highly attenuated due to their inability to escape into the cytosol and avoid phagolysosomal destruction (Freitag et al., 1993). However, in order to generate robust CTL responses, an effective Lm vaccine strain must gain access to the cytosol to deliver antigens. Therefore, XFL-7-based Lm vaccines are transformed with episomal plasmids encoding the prfA gene along with an antigen (Gunn et al., 2001). The introduction of this episomal plasmid complements the deficiency in prfA, however, the expression is not sufficient for XFL-7 to regain full virulence and is, therefore, attenuated. Subsequent studies have focused on the development of Lm strains with more targeted deletion of virulence factors that are not required for the development of robust immune responses. One such virulence gene, actA, is not necessary for the initial infection of Lm into a host cell or entry into the cytosol but is required for cell-to-cell spread of the bacterium (Brockstedt et al., 2004). While ActA-specific immunity is observable after an Lm infection, these responses are dispensable to the clearance of the bacterium, likely due to the membrane-bound nature of ActA hampering MHC Class I accessibility (Darji et al., 1998). Deletion of actA results in a highly attenuated Lm strain with an LD50 of 2 × 10⁸ CFUs as compared to 10⁵ CFUs by wild-type Lm while still retaining immunogenicity (Brundage et al., 1993; Starks et al., 2004). In fact, actA-deficient Lm expressing the model tumor antigen, chicken ovalbumin (OVA), is able to induce a more effective anti-tumor response than a wild-type strain of Lm expressing the same antigen (Starks et al., 2004). Therefore, attenuation of Lm may not just serve the purpose of improving its safety profile but it is also a rational strategy to increase immunogenicity to Lm-expressed TAAs. Further advancement in the development of attenuated Lm vaccine strains by Brockstedt and colleagues employed deletion of an additional virulence gene, internalin B (inlB) in an actA-deficient Lm strain (Brockstedt et al., 2004). The rationale for the deletion of inlB in this background was to construct a vaccine vector that exclusively infects antigen-presenting cells as InlB mediates infection of non-phagocytic cells (Gaillard et al., 1991). This was sought as a strategy to minimize toxicity from the infection, especially to the liver where toxicity is observed due to Lm infection of hepatocytes, while maximizing uptake of the vector to professional phagocytic antigen-presenting cells. Deletion of inlB in an actA deficient strain led to reduced capacity to infect both mouse and primary human hepatocytes and enhanced clearance from the liver of infected mice in comparison to a wild-type strain and the actA only deficient strain. Furthermore, deletion of inlB in the actA deletion strain that expressed the murine colon TAA, gp70, resulted in increased immunogenicity in comparison to actA deficiency alone as observed by significantly greater gp70-specific CTL responses, reduced metastatic burden with CT26 colon carcinoma tumors, and increased survival (Olino et al., 2012). The impressive immunogenicity and efficacy of the ΔplcB ΔactA Lm vaccine strain preclinically has led to its introduction into clinical trials for multiple cancers (Le et al., 2012). However, the deletion of a virulence gene such as inlB to limit infection of non-phagocytic cells may be eliminating additional mechanisms of anti-tumor efficacy by Lm-based vaccines (Kim et al., 2009). In addition to inlB, other studies have focused on deletion of plcB, another virulence gene of Lm encoding phospholipase C (PC-PLC), in the background of an actA deficient Lm vaccine strain (Angelakopoulos et al., 2002). PC-PLC is the broad-range phospholipase C of Lm that plays an important role in the breakdown of double-membrane vacuoles encountered during cell-to-cell spread by Lm (Vazquez-Boland et al., 1992; Smith et al., 1995). Deletion of plcB results in a highly attenuated Lm with reduced ability to spread cell-to-cell but also reduced toxicity as intracerebral infection is greatly reduced (Schluter et al., 1998). Deletion of both plcB and actA still allows for the induction of robust T cell responses in mouse models (Darji et al., 2003; Lenz et al., 2008) and a ΔplcB ΔactA strain has demonstrated safety in a human clinical trial (Angelakopoulos et al., 2002).

One of the earliest methods for safely inducing prophylactic immunity to pathogens was through the use of inactivated or killed forms of the same pathogen, commonly through heating (Smith and Little, 1926). While this method of inactivation is sufficient for induction of humoral responses to pathogen-associated immunogens, it is an ineffective approach to induce CTL-mediated immunity against Lm-derived antigens (Schafer et al., 1992). This lack of CTL-mediated immunity with heat-killed Lm (HKLM) is likely due in part to an inability to escape from the phagosome into the cytosol to deliver antigens to the MHC Class I pathway (Brunt et al., 1990). Additionally, a study by Bahjat et al. demonstrates that HKLM infection can also actively suppress cell-mediated immunity through induction of IL-10 in a MyD88-dependent manner (Bahjat et al., 2009). Therefore, a new method of inactivation was developed to provide the safety advantages of a killed vaccine but also maintain efficient delivery of antigens to the MHC Class I presentation machinery and induce the necessary inflammatory responses to provide immunogenicity. This new type of inactivation produced an Lm that was killed but metabolically active (KBMA) (Brockstedt et al., 2005; Skoberne et al., 2008). In order to achieve this method of inactivation, a recombinant vaccine strain of Lm that contains deletions of the uvrAB genes involved in nucleotide excision repair is treated with the DNA-intercalating agent psoralen and then DNA-psoralen crosslinking is induced with exposure to UV irradiation. The result is a bacterium that is unable to replicate and is functionally killed due to DNA crosslinking but still able to transcribe encoded genes and produce and secrete protein antigen. Importantly, KBMA Lm is also capable of escaping from phagosomes and delivering secreted antigens to the cytosol for MHC Class I processing and presentation (Dubensky et al., 2012). Vaccination with a KBMA Lm targeting a model TAA has also been shown provide comparable anti-tumor efficacy to a live-attenuated Lm strain but with slightly reduced antigen-specific CD8⁺ T cell responses (Skoberne et al., 2008).

The first study to apply recombinant Lm vaccine technology to a therapeutically relevant model of cancer in mice was performed by the Paterson laboratory and published in 2001 (Gunn et al., 2001). To develop their Lm-based therapeutics for cervical cancer, the group utilized a mouse model for HPV-induced malignancy, the TC.1 tumor model (Lin et al., 1996). The strategy for developing the first Lm-based immunotherapies was to engineer two separate recombinant Lm strains with each expressing HPV16 E7 due to its broad expression among all ano-genital malignancies (Gunn et al., 2001). However, one recombinant Lm expressed and secreted E7 alone, Lm-E7, while another strain was developed that expressed and secreted HPV16 E7 genetically fused to dtLLO, Lm-LLO-E7. After implanting TC.1 tumors in mice and allowing for establishment, mice were administered either vaccine and tumor growth monitored. While each E7-secreting vaccine successfully delayed tumor growth in comparison to untreated or control Lm treated mice, only Lm-LLO-E7 resulted in complete eradication of established tumors. Surprisingly, the increased anti-tumor efficacy of Lm-LLO-E7 in comparison to Lm-E7 was not the result of greater induction of E7-specific CTL responses in peripheral lymphoid organs as each vaccine was just as effective. Further studies determined that vaccination with Lm-E7 resulted in induction of greater numbers of CD4⁺ Tregs. In fact, it appeared that much of the reduced efficacy of Lm-E7 in comparison to Lm-LLO-E7 was due to Treg-mediated suppression of anti-tumor immune responses as Treg-depletion resulted in comparable anti-tumor responses. These results were confirmed in a subsequent study demonstrating that Lm-E7 preferentially induces Tregs that suppress effector cell function through production of IL-10 and TGF-β (Hussain and Paterson, 2004). Furthermore, work by Peng et al. found that Lm-LLO-E7 infection of BMDCs resulted in greater BMDC maturation, IL-2 production, and expansion of E7-specific CTLs than Lm-E7 infection (Peng et al., 2004). Lin et al provided additional evidence of the benefit of this approach in a study that reported significant anti-tumor protective immunity after vaccination with an Lm-based construct, rLm-E7, that secreted E7 fused to both a portion of LLO and E. coli phoA (Lin et al., 2002). These studies demonstrate the promising therapeutic potential of Lm-based vaccines against cervical cancer but provide caution to others with the evidence that therapeutic vaccines may also induce peripheral tolerance if constructed improperly.

Previous studies have found that expression of HPV16 E7 can result in loss of responsiveness to IFN-γ (Park et al., 2000; Zhou et al., 2011). While this lack of responsiveness may be a resistance mechanism to limit the direct anti-viral and anti-tumor effect of IFN-γ, studies have suggested it may also impact anti-tumor immune responses (Beatty and Paterson, 2000, 2001). In fact, the initial study with Lm-LLO-E7 by Gunn et al. determined that the anti-tumor efficacy of Lm-LLO-E7 was dependent on the effects of IFN-γ (Gunn et al., 2001). To determine the role of IFN-γ in the anti-tumor response mediated by Lm-LLO-E7, Dominiecki et al utilized a TC.1 variant cell line unresponsive to IFN-γ due to constitutive expression of a dominant negative form of the IFN-γ receptor, termed mugR (Dominiecki et al., 2005). While TC.1-mugR tumors grew slightly slower than TC.1 parental tumors in the absence of therapy, they were resistant to the anti-tumor response mediated by Lm-LLO-E7. The mechanism for this resistance to Lm-LLO-E7 therapy appeared to correspond to the inability of CTLs to migrate into the tumor and effect tumor regression. Paterson and colleagues studied this phenotype further and discovered that IFN-γ stimulation of TC.1 tumor cells results in the production of a chemokine, CXCL9, that mediates infiltration of CD8⁺ T cells into TC.1 tumors after Lm-LLO-E7 vaccination (Guirnalda et al., 2013). Therefore, the effective anti-tumor immune response induced by Lm-LLO-E7 is due to a number of factors such as reduced peripheral tolerance and increased maturation of dendritic cells but this response is only effective if tumors remain compliant through responsiveness to IFN-γ, a determinant that may be useful to predict patient outcomes with Lm-based immunotherapies.

Numerous studies have found that responses to tumor immunotherapy are greatly enhanced when mechanisms maintaining peripheral tolerance are inhibited (Gunn et al., 2001; Duraiswamy et al., 2013; Mkrtichyan et al., 2013). While there are several pathways regulating peripheral tolerance, the PD-1 receptor pathway is gaining particular interest as a target for improving immunotherapies (Topalian et al., 2012). Interestingly, BMDCs infected with Lm-LLO-E7 were found to upregulate expression of the PD-1 ligand, PD-L1 (Peng et al., 2004; Mkrtichyan et al., 2013). Therefore, Khleif and colleagues sought to determine if activation of the PD-1 immune inhibitory pathway was limiting the efficacy of Lm-LLO-E7 (Mkrtichyan et al., 2013). Antibody-mediated blockade of PD-1 along with Lm-LLO-E7 did result in increased anti-tumor efficacy as determined by TC.1 tumor load, survival, and E7-specific T cell responses in comparison to vaccine alone. However, the Lm-LLO-E7 dose in this study was approximately 20-fold lower (5 × 10⁶ CFUs) than in previous studies (10⁸ CFUs) so it is unclear if this enhanced efficacy is only evident when given with a suboptimal dosage of Lm-LLO-E7 (Gunn et al., 2001; Guirnalda et al., 2013). Nevertheless, the results obtained from this study could lead to a very effective synergistic clinical treatment regimen especially with the promising results of both Lm-LLO-E7 and anti-PD-1 treatment separately in clinical trials (Maciag et al., 2009; Petit and Basu, 2013; Topalian et al., 2012).

After finding preclinical success treating models of virus-induced cancer, development progressed on Lm-based vaccines for other challenging indications such as breast cancer where the only identified targets are self-TAAs, such as Her2/neu. Her2 is a epidermal growth factor receptor family protein that is overexpressed in roughly 25% of human breast tumors and correlates with poor prognosis (Slamon et al., 1987). While most vaccines targeting Her2 were primarily used in prophylactic settings, Singh et al rationalized that the use of Lm as a vaccine vector may induce a robust CTL response that is therapeutically effective (Singh et al., 2005). However, in order to construct Lm-based vaccines against a large membrane-bound receptor like Her2, several constructs were required to be developed. Each construct contained a portion of the intracellular (Lm-LLO-IC1 and Lm-LLO-IC2) or the extracellular (Lm-LLO-EC1, Lm-LLO-EC2 and Lm-LLO-EC3) portion of the full-length rat Her2/neu protein to allow for efficient production and secretion by Lm. Therapeutic vaccination with each of the constructs individually resulted in reduced or static growth of NT-2 breast tumors, a tumor cell that constitutively expresses rat Her2/neu under the mouse mammary tumor virus (MMTV) promoter (Singh et al., 2005). In comparison to DNA vaccines targeting Her2, Lm-based vaccines demonstrated significantly greater therapeutic efficacy. Interestingly, the efficacy of Lm-based Her2 vaccines was not due solely to the bacterium as a DNA vaccine study suggested the fusion of the Her2 antigenic regions to dtLLO is also an important factor (Singh and Paterson, 2006b). The mechanism for this improved efficacy of Lm or LLO-based Her2 vaccines appears to be in their ability to induce a greater repertoire of Her2-specific CTLs. In fact, vaccination with Lm-based Her2 vaccines also revealed more epitopes than had been previously identified with a DNA vaccination strategy (Singh et al., 2005). This suggests that vaccination against Her2 with Lm or LLO-based vaccines generated both a quantitatively and qualitatively different response than by other methods and with greater therapeutic efficacy.

While initial studies with the Lm-Her2 vaccines demonstrated promising efficacy against a mouse tumor cell line that expresses rat Her2/neu, they were performed in a mouse model that was not fully tolerant to the antigen. Although rat and mouse Her2/neu are roughly 94% homologous, rat Her2/neu is immunogenic in mice (Nagata et al., 1997). Therefore, future studies were performed in an MMTV-Her2/neu-transgenic mouse model that constitutively expresses rat Her2/neu under control of the MMTV promoter and shows profound tolerance to rat Her2/neu (Guy et al., 1992; Reilly et al., 2000). Although fewer mice responded to Lm-Her2 therapy, complete tumor regressions were observed. Further studies elucidated several mechanisms for the vaccines not reaching their full therapeutic potential. First, after vaccination with Lm-Her2 vaccines, tumors would undergo immunoediting, a process wherein immunological pressure against a TAA will result in selective outgrowth of tumor cells that have mutations in the targeted antigen (Singh and Paterson, 2007a). Second, central tolerance to self-TAAs, in particular for these studies with Her2, is detrimental to the induction of high avidity tumor-specific CTLs (Singh and Paterson, 2007b). When FVB/N Her2/neu transgenic mice bearing Her2-positive tumors were vaccinated with Lm-based Her2 vaccines, CTLs were generated against previously revealed epitopes but the avidity of those CTLs was reduced by at least one log in comparison to CTLs generated in a wild-type mouse. Finally, peripheral tolerance mediated by suppressive cell types such as Tregs also limits the efficacy of Lm-based tumor immunotherapies (Singh and Paterson, 2007b). Antibody-mediated depletion of Tregs concomitant with Lm-based vaccination against Her2 resulted in enhanced anti-tumor efficacy, therefore, providing a possible therapeutic avenue for improved efficacy in the clinic (Singh and Paterson, 2007b).

To translate the success of Lm-based immunotherapies targeting rat Her2/neu into the clinic, development of several vaccines was undertaken that expressed regions or defined epitopes within human Her2 (Seavey et al., 2009b; Shahabi et al., 2011). As in the previous studies with rat Her2/neu, Lm-based vaccines were developed that expressed separate regions of the intracellular (Lm-LLO-hIC1) and extracellular (Lm-LLO-hEC1 and Lm-LLO-hEC2) portions of human Her2. Due to the high degree of homology between human and rodent Her2/neu and the presence of defined H-2^q epitopes within the portions from human Her2 (Seavey et al., 2009b), mouse breast tumor models such as NT-2, 4T1, and the MMTV-Her2/neu transgenic could be used to assess efficacy. Each of these vaccines was found to be equally as effective at inducing Her2/neu-specific CTL responses and controlling tumor growth in each of the tumor models as the previous rat Her2/neu-based constructs. However, to improve the breadth of the immune response against Her2, an additional vaccine was developed that contained most of the HLA epitopes found within Lm-LLO-hIC1, Lm-LLO-hEC1, and Lm-LLO-hEC2 but expressed as a single polyvalent antigen (Seavey et al., 2009b). This chimeric Her2 vaccine, named Lm-hHer2/neu chimera, demonstrated significant anti-tumor efficacy and was able to delay autochthonous tumor formation in a tolerized Her2/neu transgenic mouse. Additionally, the Lm-hHer2/neu chimera was able to prevent lung metastasis formation from intravenous challenge with the highly aggressive 4T1-Luc breast tumor model. This impressive efficacy of Lm-hHer2/new chimera at eliciting a potent anti-tumor response was highlighted by the relatively low expression of Her2/neu by 4T1-Luc tumor cells (Seavey et al., 2009b). The success of the Lm-hHer2/neu chimera vaccine resulted in its further development into a new Lm-based vector that is absent of any antibiotic resistance genes, and is therefore more easily translatable to the clinic (Shahabi et al., 2011). This new vaccine, ADXS-cHER2, demonstrated improved efficacy over Lm-hHer2/neu chimera in autochthonous Her2/neu breast tumor studies but also an impressive ability to protect mice from intracranial challenge from the EMT6-Luc breast tumor cell line. The preclinical development of Lm-based chimeric Her2 vaccines predicts a promising future that will be determined in clinical trials that are ongoing in canines and currently planned for humans (Figure 3).

While Her2 is a promising target for breast cancer tumor immunotherapy, its wide application is limited since only about 25% of breast cancer patients have elevated expression of Her2 in their tumors (Slamon et al., 1987). This is especially problematic for breast cancer patients with tumors that lack expression of Her2/neu, progesterone receptor, and the estrogen receptor, denoted as triple-negative, and have few therapeutic options available (Hudis and Gianni, 2011). Therefore, the discovery of novel TAAs for use in breast tumor immunotherapy is a desired goal. Progress in this area of research was recently reported in a study that targeted a small ubiquitin-like protein elevated in even triple-negative breast tumors, ISG15, with an Lm-based immunotherapy (Bektas et al., 2008; Wood et al., 2012). After validating overexpression of ISG15 in mouse models of breast cancer and developing an Lm-based vaccine that expresses ISG15 fused to dtLLO, Wood et al found that vaccination of mice with Lm-LLO-ISG15 resulted in induction of ISG15-specific CTL responses. In a therapeutic setting, treatment with Lm-LLO-ISG15 significantly reduced the growth of breast tumors and the number of metastatic lesions in the lungs of tumor-bearing mice. When administered prophylactically in MMTV-Her2/neu transgenic mice, autochthonous mammary tumor formation was significantly delayed as compared to a control Lm. These results suggest that ISG15 is not only a novel immunotherapeutic target for breast cancer but also that Lm-based vaccines are powerful platforms to validate TAA discovery.

MAGE are a type of cancer/testis antigens commonly targeted in tumor immunotherapy due to their limited expression in normal tissues (Simpson et al., 2005). MAGE-b, in particular, is highly expressed in over 90% of breast tumors (De Backer et al., 1995; Park et al., 2002). For this reason, Kim et al constructed several Lm-based vaccines containing regions of MAGE-b fused to dtLLO (Kim et al., 2008). In a mixed prophylactic and therapeutic vaccination regimen, mice were administered twice with one of four Lm-Mage-b vaccines prior to tumor challenge with 4T1 mammary carcinoma cells and then subsequent administration of a third vaccination. Surprisingly, vaccination with one vaccine, Lm-LLO-Mage-b/2nd resulted in significantly reduced 4T1 lung metastases as compared to control vaccination but no impact on primary tumor growth. Further examination discovered that elevated expression of IL-6 by the primary tumor was the likely culprit for this anti-tumor effect being limited to only metastatic disease. An additional mechanism of Lm-LLO-Mage-b/2nd anti-tumor efficacy was subsequently discovered involving direct killing of infected tumor cells though elevation of cytosolic ROS, as discussed above (Kim et al., 2009).

The tumor suppressor p53 is commonly mutated in tumors and correlates with decreased survival in breast cancer (Harris and Hollstein, 1993). However, mutations in p53 can also result in tumor cell-specific overexpression and render them susceptible to killing by p53-specific CTLs (Zambetti and Levine, 1993; Nikitina et al., 2001). Therefore, Ishizaki et al. constructed an Lm-based vaccine targeting human p53, LmddA-LLO-p53, and assessed its anti-tumor efficacy against 4T1 tumor cells stably expressing mutant human p53 (4T1p53) in tolerant human p53 knock-in (Hupki) mice (Ishizaki et al., 2010). While LmddA-LLO-p53 administration alone provided some prophylactic and therapeutic efficacy against 4T1p53 tumors, p53-specific immunity and prophylactic efficacy were dramatically improved when given with a modified vaccinia Ankara vaccine expressing human p53 (MVA-p53) in a heterologous prime/boost approach. Therapeutic efficacy and survival were also further improved when adjuvants (Poly I:C and CpG-ODN) were incorporated into the heterologous prime/boost regimen. Therefore, this study provides additional evidence that maximal anti-tumor efficacy of Lm-based vaccines may sometimes be achieved only when given in a combinatorial approach (Hannan et al., 2012; Mkrtichyan et al., 2013).

HMW-MAA was initially discovered as an antigen highly expressed in melanomas and other malignancies derived from the neural crest and a previous target of tumor immunotherapy (Campoli et al., 2004; Chang et al., 2004). In an effort to determine if an Lm-based immunotherapy targeting HMW-MAA would be effective against mouse models of melanoma, initial studies were performed against the B16F10 melanoma cell line stably expressing HMW-MAA along with an Lm vaccine expressing a region of HMW-MAA, Lm-LLO-HMWMAA-C (Maciag et al., 2008). While Lm-LLO-HMWMAA-C successfully reduced growth of a melanoma tumor expressing HMW-MAA, unexpected results were found in experiments targeting tumor cells that did not express HMW-MAA. Most strikingly, in two breast cancer models, Lm-LLO-HMWMAA-C, reduced breast tumor burden and increased time to progression in a Her2/neu transgenic autochthonous breast tumor model. Further examination of this unexpected efficacy determined that expression of the murine HMW-MAA homolog, AN2, was highly expressed by pericytes in the tumor vasculature. Pericytes are mesenchymal cells that express AN2 and are associated with, and stabilize, vasculature in tumors and the periphery (Schneider et al., 2001; Baluk et al., 2005; von Tell et al., 2006). After Lm-LLO-HMWMAA-C vaccination, CD8⁺ T cells colocalized with the tumor vasculature and the number of pericytes associated with the tumor vasculature was significantly decreased. Importantly, targeting of pericytes with Lm-LLO-HMWMAA-C did not impact normal processes associated with neovascularization such as wound healing and pregnancy. These observations may suggest that the specific susceptibility of tumors to pericyte-targeting therapy is due to the already decreased numbers of pericytes associated with tumor vasculature (Eberhard et al., 2000).

Folkman's hypothesis that limiting tumor growth could be achieved by targeting tumor angiogenesis in order to limit the necessary oxygen and nutrient supply has been realized with the introduction of bevacuzimab, an antibody against VEGF, into the clinic (Hurwitz et al., 2004). VEGF receptors are found on activated vascular endothelium and the murine homolog of VEGFR2, Flk-1, plays an important role in tumor neovascularization and growth (Kim et al., 1993; Millauer et al., 1993, 1994). However, successful passive immunotherapies such as bevacuzimab have been met with therapeutic resistance (Ellis and Hicklin, 2008). In order to determine the efficacy of an active CTL-mediated immune responses against Flk-1 in tumor immunotherapy, three separate Lm-based vaccines were constructed based on regions of Flk-1 with vaccines targeting extracellular portions of the protein, Lm-Flk-E1 and Lm-Flk-E2, and intracellular, Lm-Flk-I1 (Seavey et al., 2009a). While each vaccine was able to induce specific IFN-γ responses against an MHC Class I epitope within their respective regions, only two of the vaccines, Lm-FlkE1 and Lm-FlkI1, were able to significantly inhibit Her2/neu+ breast tumor growth, reduce tumor vascularization, and induce secondary T cell responses against the non-targeted TAAs, a process known as epitope spreading (Vanderlugt and Miller, 2002). Interestingly, the anti-tumor efficacy of the Flk-1 targeting vaccines was completely abrogated when examined in a tolerant Her2/neu transgenic mouse. These results support a hypothesis that CTL-mediated targeting of antigens on activated vascular endothelium, such as VEGFR2, can result in anti-tumor efficacy but epitope spreading to additional TAAs is a requirement for this efficacy. Support for this theory was found in a follow-up study demonstrating that autochthonous tumor formation in Her2/neu transgenic mouse after Lm-Flk-E1 or Lm-Flk-I1 vaccination correlated with the occurrence of mutations in Her2/neu CTL epitopes within those tumors (Seavey and Paterson, 2009). Therefore, these studies suggest that therapeutic vaccines targeting antigens associated with tumor vasculature alone may not reach their full potential unless tolerance to TAAs is overcome. This predicts a possible promising therapeutic strategy of pairing one Lm-based vaccine targeting tumor vasculature with another vaccine that is able to break central tolerance to a TAA such as Lm-cHer2 to maximize therapeutic efficacy.

While previous studies proved Flk-1 to be an effective target for impacting tumor vasculature, further studies sought to validate additional promising tumor vasculature targets in an Lm-based immunotherapy approach. One subsequent study focused on a member of the TGF-β receptor complex, endoglin (CD105), which is highly expressed on the vascular endothelium of tumors and believed to play a role in tumor neovascularization (Wang et al., 1993; Perez-Gomez et al., 2005). Previous studies targeting CD105 with an antibody had already met with some success at controlling tumor burden in a clinical trial and preclinical models of cancer (Matsuno et al., 1999; Rosen et al., 2012). To determine the efficacy of targeting CD105 with an active immunotherapy approach, two Lm-based vaccines were engineered to secrete different regions of CD105 fused to dtLLO (Wood et al., 2011). Each vaccine, Lm-LLO-CD105A and Lm-LLO-CD105B, were found to effectively reduce the growth of Her2/neu+ breast tumors and delay progression of autochthonous tumor progression in a Her2/neu transgenic mouse. In addition, vaccination against CD105 reduced lung metastases in the 4T1 mouse model of metastatic breast cancer and led to epitope spreading to TAAs as seen previously with the Flk-1 vaccines. Tumor-associated expression of CD31 and hemoglobin were also reduced after CD105 vaccination demonstrating its promise as a target for CTL-mediated approaches to limit tumor angiogenesis.

The current standard of care for prostate cancer patients involving radiation and chemotherapy still does not prevent roughly 10–30% of men from having recurrent malignancy (Katz and McKiernan, 2007). Provenge administration in these patients results in successful but limited extension of lifespan but it carries with it the difficulties and high cost associated with autologous cell-based vaccines (Simoens, 2012). While Provenge targets PAP, improved efficacy may be found with another promising target for prostate tumor immunotherapy, prostate-specific antigen (PSA), a protein widely expressed in prostate adenocarcinomas (Cunha et al., 2006). PSA-specific CTL-mediated approaches developed to therapeutically treat prostate cancer have shown efficacy in the clinic and provided the rationale for development of an Lm-based vaccine targeting PSA, Lm-LLO-PSA (Heiser et al., 2002; Shahabi et al., 2008). Lm-LLO-PSA vaccination in mice was successful at generating PSA-specific CTL responses. Administration of Lm-LLO-PSA to mice bearing T-PSA23 prostate tumors, TRAMP-C murine prostate adenocarcinoma tumors stably expressing PSA, resulted in eradication of over 80% of tumors along with reduced infiltration of Tregs into tumors as compared to a control vaccine (Shahabi et al., 2008). In a comparison of vaccine vectors, Lm-LLO-PSA was also found to be the most effective at reducing prostate tumor burden in mice as compared to DNA and vaccinia virus-based vaccines expressing the same antigen.

A follow-up study with an updated Lm-based vaccine targeting PSA, but devoid of antibiotic selection markers, ADXS31-142 (Wallecha et al., 2009) was recently undertaken to determine whether its anti-tumor efficacy could synergize with the radiation therapy that is commonly administered to prostate cancer patients (Wallecha et al., 2009; Hannan et al., 2012). After implantation of T-PSA23 tumors, mice were treated with either radiation followed by three doses of ADXS31-142, radiation alone, or ADXS31-142 alone. Each treatment resulted in significantly lower tumor burden as compared to control Lm vaccination but radiation followed by ADXS31-142 vaccination resulted in synergistic therapeutic anti-tumor efficacy. The mechanism for this synergistic application of radiation and Lm vaccine appeared to be due to a robust four-fold increase in PSA-specific CTLs in the spleens of treated mice. These results are in line with previous studies that documented synergism of irradiation along with tumor immunotherapy and provide a rational approach for administration of Lm-based vaccines in prostate cancer patients (Kantoff et al., 2010b).

In order to mitigate the emergence of tumor immune escape due to mutation of a single targeted TAA epitope, some have developed chimeric Lm-based vaccines that incorporate multiple epitopes from regions spanning an entire TAA (Seavey et al., 2009b; Shahabi et al., 2011). While these studies have met with success, loss of complete TAA expression due to immune pressure, as previously observed with Her-2/neu, may render even this approach ineffective (Kmieciak et al., 2007). Therefore, to further impede the emergence of tumor immune escape, others have developed polyvalent therapeutic tumor vaccines that incorporate epitopes from several TAAs (Tanaka et al., 2002). Early Lm-based polyvalent vaccines have either contained model tumor antigens (Bruhn et al., 2005) or were not validated for in vivo efficacy (Sinnathamby et al., 2009) but a recent study by Chen et al. documents the validation of an Lm-based polyvalent vaccine targeting clinically relevant TAAs in a model of hepatocellular carcinoma (HCC) (Tanaka et al., 2002; Chen et al., 2012). HCC is a prevalent cause of cancer-related deaths worldwide, especially in countries such as China where oncogenic Hepatitis B virus (HBV) infection is endemic (Chen et al., 2012). In order to develop an effective Lm-based therapeutic vaccine to treat HCC that overcomes the challenges of immune escape of previous immunotherapies, the authors constructed a vaccine that expressed the entire HBV core protein (HBc) along with validated HLA-A2 epitopes for additional TAAs including HBV-X protein, alpha-fetoprotein, and MAGE-A. The resulting fusion protein was named multiple peptide fusing genes (MPFG) and the Lm-based vaccine that expressed this fusion protein, Lm-MPFG. Lm-MPGF administration to HLA-A2 transgenic mice induced robust IFN-γ and cytolytic responses against the MPFG antigens. When mice were vaccinated prophylactically or therapeutically after inoculation with Hepa1-6 tumor cells expressing MPFG and HLA-A2, Lm-MPFG treatment led to reduced tumor burden and increased tumor-free survival in comparison to a control Lm. The anti-tumor efficacy of Lm-MPFG correlated with increased infiltrations of MPFG-specific CD8⁺ T cells and increased cytolytic potential of CTLs. Additionally, prophylactic and therapeutic administration of Lm-MPFG resulted in significantly reduced number of Tregs in tumors and reduced production of anti-inflammatory cytokines such as TGF-β and IL-10 by the tumor-infiltrating Tregs. This study provides the first evidence that Lm-based immunotherapies could be effective in the treatment of HCC but it also demonstrates the validation of a polyvalent Lm-based vaccine approach that will prove useful in limiting tumor immune escape.

Early studies with Lm-based tumor immunotherapies targeting a mouse model of melanoma, B16F10-NP, demonstrated their potency against this malignancy. However, these studies were targeting the model tumor antigen influenza nucleoprotein, NP, and not a melanoma-specific TAA that is clinically relevant (Pan et al., 1999). One melanoma-specific TAA that previously was found to be immunogenic and a validated target for tumor immunotherapy is Tyrosinase-related protein-2 (TRP-2) (Bloom et al., 1997). TRP-2 is a melanogenic enzyme expressed specifically by melanocytes and melanoma cells (Wang et al., 1996). To determine the therapeutic efficacy of targeting TRP-2 with Lm-based immunotherapy, Bruhn et al. constructed several Lm-based vaccines expressing either TRP-2 alone or fused to an MHC class I H-2D^b-restricted epitope from LCMV NP (a.a. 396–404) (Bruhn et al., 2005). While either vaccine was capable of inducing TRP-2-specific CD8+ T cell responses, studies determining anti-tumor efficacy were subsequently carried out with only the Lm-TRP2-NP construct. After two prophylactic vaccinations with Lm-TRP2-NP and subsequent challenge with B16 tumors, mice receiving the TRP-2 vaccine had significantly reduced tumor burden as compared to mice that received the control Lm, Lm-NP, or empty Lm vector. The Lm-TRP2-NP vaccine was also found to be therapeutically effective against B16-MO5 metastatic lesions, a variant of the B16 tumor cell line stably expressing chicken ovalbumin (Falo et al., 1995). A follow-up study also demonstrated the ability of Lm-TRP2-NP to provide protective immunity in a B16 brain metastasis model (Prins et al., 2006). To further improve the anti-tumor efficacy of Lm-TRP2-NP, it was administered in combination with the TLR7 agonist, imiquimod (Craft et al., 2005). Imiquimod was previously found to be a strong adjuvant and lead to reduced tumor growth alone. When administered along with Lm-TRP2-NP, it enhanced the prophylactic anti-tumor efficacy of Lm-TRP2-NP in both primary and metastatic B16 tumor challenge models. The findings from these studies validate Lm-based immunotherapies as effective inducers of anti-tumor immune responses to sites of metastatic spread that are normally considered immune-privileged, a finding later confirmed in a mouse model for breast cancer by Shahabi et al. (2011).