Live Attenuated Tularemia Vaccines for Protection Against Respiratory Challenge With Virulent F. tularensis subsp. tularensis

Francisella tularensis is the causative agent of tularemia and a Tier I bioterrorism agent. In the 1900s, several vaccines were developed against tularemia including the killed “Foshay” vaccine, subunit vaccines comprising F. tularensis protein(s) or lipoproteins(s) in an adjuvant formulation, and the F. tularensis Live Vaccine Strain (LVS); none were licensed in the U.S.A. or European Union. The LVS vaccine retains toxicity in humans and animals—especially mice—but has demonstrated efficacy in humans, and thus serves as the current gold standard for vaccine efficacy studies. The U.S.A. 2001 anthrax bioterrorism attack spawned renewed interest in vaccines against potential biowarfare agents including F. tularensis. Since live attenuated—but not killed or subunit—vaccines have shown promising efficacy and since vaccine efficacy against respiratory challenge with less virulent subspecies holarctica or F. novicida, or against non-respiratory challenge with virulent subsp. tularensis (Type A) does not reliably predict vaccine efficacy against respiratory challenge with virulent subsp. tularensis, the route of transmission and species of greatest concern in a bioterrorist attack, in this review, we focus on live attenuated tularemia vaccine candidates tested against respiratory challenge with virulent Type A strains, including homologous vaccines derived from mutants of subsp. holarctica, F. novicida, and subsp. tularensis, and heterologous vaccines developed using viral or bacterial vectors to express F. tularensis immunoprotective antigens. We compare the virulence and efficacy of these vaccine candidates with that of LVS and discuss factors that can significantly impact the development and evaluation of live attenuated tularemia vaccines. Several vaccines meet what we would consider the minimum criteria for vaccines to go forward into clinical development—safety greater than LVS and efficacy at least as great as LVS, and of these, several meet the higher standard of having efficacy ≥LVS in the demanding mouse model of tularemia. These latter include LVS with deletions in purMCD, sodBFt, capB or wzy; LVS ΔcapB that also overexpresses Type VI Secretion System (T6SS) proteins; FSC200 with a deletion in clpB; the single deletional purMCD mutant of F. tularensis SCHU S4, and a heterologous prime-boost vaccine comprising LVS ΔcapB and Listeria monocytogenes expressing T6SS proteins.

Francisella tularensis is the causative agent of tularemia and a Tier I bioterrorism agent. In the 1900s, several vaccines were developed against tularemia including the killed "Foshay" vaccine, subunit vaccines comprising F. tularensis protein(s) or lipoproteins(s) in an adjuvant formulation, and the F. tularensis Live Vaccine Strain (LVS); none were licensed in the U.S.A. or European Union. The LVS vaccine retains toxicity in humans and animals-especially mice-but has demonstrated efficacy in humans, and thus serves as the current gold standard for vaccine efficacy studies. The U.S.A. 2001 anthrax bioterrorism attack spawned renewed interest in vaccines against potential biowarfare agents including F. tularensis. Since live attenuated-but not killed or subunit-vaccines have shown promising efficacy and since vaccine efficacy against respiratory challenge with less virulent subspecies holarctica or F. novicida, or against non-respiratory challenge with virulent subsp. tularensis (Type A) does not reliably predict vaccine efficacy against respiratory challenge with virulent subsp. tularensis, the route of transmission and species of greatest concern in a bioterrorist attack, in this review, we focus on live attenuated tularemia vaccine candidates tested against respiratory challenge with virulent Type A strains, including homologous vaccines derived from mutants of subsp. holarctica, F. novicida, and subsp. tularensis, and heterologous vaccines developed using viral or bacterial vectors to express F. tularensis immunoprotective antigens. We compare the virulence and efficacy of these vaccine candidates with that of LVS and discuss factors that can significantly impact the development and evaluation of live attenuated tularemia vaccines. Several vaccines meet what we would consider the minimum criteria for vaccines to go forward into clinical development-safety greater than LVS and efficacy at least as great as LVS, and of these, several meet the higher standard of having efficacy ≥LVS in the demanding mouse model of tularemia. These latter include LVS with deletions in purMCD, sodB Ft , capB or wzy; LVS capB that also overexpresses Type VI Secretion System (T6SS) proteins; FSC200 with a deletion in clpB; the single deletional purMCD mutant of F. tularensis SCHU S4, and a heterologous prime-boost vaccine comprising LVS capB and Listeria monocytogenes expressing T6SS proteins.

INTRODUCTION
Francisella tularensis, the causative agent of tularemia, was originally named Bacterium tularense by McCoy and Chapin in 1911, who discovered it as the causative agent of a "plague like disease" prevalent among the ground squirrels in Tulare County, California (McCoy GW, 1912;Francis, 1925). The bacterium was designated as Pasteurella tularensis in 1920's and later renamed Francisella tularensis in honor of the contributions made by Dr. Edward Francis among others (Francis, 1928;Weinberg, 2004;Sjostedt, 2007). Several vaccine strategies were developed in the 1900s against virulent F. tularensis infection, including killed whole cell vaccines, subunit vaccines, and live attenuated homologous vaccines (Wayne Conlan and Oyston, 2007). A killed whole cell vaccine, the earliest vaccine developed against tularemia and referred to as the "Foshay" vaccine (Foshay, 1932;Foshay et al., 1942), was prepared from phenolized liquid culture and was not highly protective against subsequent systemic and aerosol challenge with highly virulent strains of F. tularensis in mice, guinea pigs, rabbits, and humans (Foshay, 1932;Foshay et al., 1942;Kadull et al., 1950;Vanmetre and Kadull, 1959;Eigelsbach and Downs, 1961;Saslaw et al., 1961a,b). Subunit vaccines comprising F. tularensis protein(s) or lipoproteins(s) in an adjuvant formulation also failed to demonstrate strong protective immunity against virulent non-Type A or Type A F. tularensis in animal models (Sjostedt et al., 1992b;Golovliov et al., 1995;Fulop et al., 2001;Conlan et al., 2002). Live attenuated homologous vaccines showed greater promise. The earliest live attenuated homologous tularemia vaccine, Live Vaccine Strain or LVS, derived from a virulent isolate of F. tularensis subsp. holarctica (Type B), was developed by U.S.A. and Russian scientists in the 1950s (Eigelsbach and Downs, 1961). However, LVS retained considerable virulence in animals and provided incomplete protection to humans challenged with F. tularensis subsp. tularensis (Type A) by aerosol, the route of transmission of greatest concern in a bioterrorist attack (Saslaw et al., 1961a;Fortier et al., 1991). The fundamental mechanism of LVS attenuation is not fully understood, although pilA and FTT0918 have been identified as contributing to LVS virulence (Salomonsson et al., 2009). The LVS vaccine has not been licensed in the U.S.A. 1,2 or in the European Union 3 .

TULAREMIA AND F. TULARENSIS
Tularemia occurs in nature as a fatal bacteremia of various rodents and other animals, such as rabbits, and is transmitted to humans as an accidental infection (Francis, 1925). Depending primarily on the route of transmission, there are several clinical forms of tularemia, including ulceroglandular and glandular tularemia caused by an arthropod bite or skin contact with an infected animal; oculoglandular tularemia caused by direct infection of the eye; oropharyngeal tularemia caused by ingestion of water contaminated by infected rodents or other animals or consumption of under-cooked meat from an infected animal; typhoidal tularemia from any mode of transmission; and pneumonic tularemia caused by inhalation of aerosolized bacteria. Typhoidal and pneumonic tularemia are the most dangerous forms as they carry a high fatality rate−30-60% if untreated (Dennis et al., 2001;Matyas et al., 2007). Because F. tularensis is one of the most pathogenic human pathogens known and can be spread by aerosol transmission to cause highly fatal pneumonic tularemia, it is classified as a Tier 1 select agent of bioterrorism, i.e., among the most likely pathogens to be deliberately used in a bioterrorist attack. In fact, Japan, the U.S.A., and the U.S.S.R. have stockpiled F. tularensis as a bioweapon in the past (Harris, 1992;Christopher et al., 1997;Alibek, 1999). F. tularensis subsp. tularensis (Type A), prevalent in North America, is the most virulent subsp.; intracutaneous challenge or inhalation of as few as 10-50 colony forming units (CFU) (SCHU S4 strain) is able to cause clinical tularemia in humans (McCrumb, 1961;Saslaw et al., 1961a,b). F. tularensis subsp. holarctica (Type B), found in Europe, Asia and North America, is less virulent than subsp. tularensis. Subsp. mediasiatica, found in the Central Asia and the former USSR, is of similar virulence to subsp. holarctica (Sjostedt, 2007). F. novicida, genetically closely related to F. tularensis and frequently referred to as subsp. novicida (Rohmer et al., 2007;Huber et al., 2010), is currently classified as a separate species (Johansson et al., 2010;Kingry and Petersen, 2014). F. novicida shows low virulence in experimental models, occasionally causes disease in immunocompromised individuals, and has been isolated from patients with various clinical entities in the U.S.A., Canada, Australia, and Spain (Bernard et al., 1994;WHO, 2007).

BIOLOGY AND PATHOGENESIS OF EXPERIMENTAL TULAREMIA AND THE HOST RESPONSE
F. tularensis is a facultative intracellular pathogen that is capable of infecting multiple types of eukaryotic cells, including macrophages. Using a human macrophage-like cell line (THP-1), Clemens et al. showed that F. tularensis subsp. tularensis SCHU S4 and subsp. holarctica LVS enter human macrophages via a unique complement-dependent process termed looping phagocytosis; then enter a unique fibrillar-coated phagosome; and finally exit the phagosome to multiply freely in the cytoplasm using a Type VI Secretion System (T6SS) (Clemens et al., 2004(Clemens et al., , 2005(Clemens et al., , 2015Clemens and Horwitz, 2007). In addition to the macrophage, F. tularensis can infect alveolar type II epithelial cells, neutrophils, dendritic cells, and others (Metzger et al., 2007). Studies on the pathogenesis of experimental tularemia were mostly conducted in mice and to a lesser extent in Fisher rats and non-human primates (Jemski, 1981;Lyons and Wu, 2007;Hutt et al., 2017). Upon i.d. or aerosol infection, F. tularensis quickly disseminates systemically to spleen, liver, lung, lymph nodes, and bone marrow, and multiplies in these organs Fritz et al., 2014). As with many other intracellular pathogens, F. tularensis does not produce a toxin to cause tissue damage. Instead, F. tularensis damages tissues by invasion and destruction. Multiplication of F. tularensis, especially subsp. tularensis, in the tissues fails to trigger-but rather actively suppresses-innate immune responses in humans, which likely contributes to its enhanced pathogenicity Gillette et al., 2014). Although various immune responses to LVS vaccination and F. tularensis infection have been identified in animal models, correlates of protection are still not known. Both LVS vaccination and subsp. tularensis infection induce antibody responses. In the mouse, LVS vaccination induces F. tularensis antigen specific humoral (antibodies and B cells) and cellular (both CD4+ and CD8+ T cells) immune responses. While T-cell mediated immune responses are important for control of primary Francisella infection or vaccineinduced protection, antibodies may also provide prophylactic and therapeutic protection against pulmonary infection with low virulence F. tularensis, strains (Elkins et al., 2007;Metzger et al., 2007;Cole et al., 2009). Protective immunity induced by natural infection with F. tularensis or vaccination with LVS in humans depends primarily upon cell-mediated Th1type immune responses, including interferon-gamma (IFN-γ), tumor necrosis factor-alpha (TNF-α), and interleukin-12 (IL-12) (Karttunen et al., 1987;Tarnvik, 1989;Sjostedt et al., 1992a;Ericsson et al., 1994Ericsson et al., , 2001.

THE LVS VACCINE IN HUMAN STUDIES
The LVS vaccine was studied for safety and efficacy in small numbers of individuals in the 1960s. Saslaw et al. showed that individuals vaccinated with LVS intradermally (i.d., via multiple punctures) develop local lesions manifested by erythema (∼1 cm in diameter) followed by non-tender papules that fade rapidly with no fever or systemic reactions (Saslaw et al., 1961a). Hornick et al. showed that among persons exposed to aerosolized LVS, 90% of individuals that inhaled 10 8 LVS exhibited mild typhoidal tularemia and 80% had temperature elevations of >100 • F; symptoms and signs were milder when the inhalation dose was reduced to 10 6 . Saslaw et al. further showed that i.d. vaccination of humans with LVS induced incomplete protection (3 of 18 persons developed disease vs. 8 of 10 nonvaccinated controls) against aerosol challenge with 10-53 CFU of subsp. tularensis SCHU S4 strain; in contrast, in a separate part of the study, the Foshay vaccine was not protective as there was no significant difference in the incidence of disease between the vaccinated and unvaccinated groups after aerosol challenge with similar doses of subsp. tularensis SCHU S4 strain (Saslaw et al., 1961a). Hornick and Eigelsbach demonstrated that individuals vaccinated with 10 6 or 10 8 CFU LVS by aerosol were well protected (none required antibiotic treatment) against a high dose challenge with aerosolized SCHU S4 (25,000 CFU, 500-2,500 minimum infective doses) 4-6 months post-vaccination. Similarly vaccinated volunteers were better protected than individuals immunized with a lower dose of LVS (10 4 CFU) by aerosol or vaccinated i.d. by acupuncture; 46% of the latter group required antibiotic treatment vs. 89% of unvaccinated controls . Hornick et al. also demonstrated that oral vaccination with LVS provided incomplete protection against aerosol challenge with SCHU S4 . However, the i.d. route, but not the aerosol or oral route, has been regularly used for humans in the U.S.A. Saslaw and Carhart found that the greater efficacy of the live LVS vaccine than the Foshay killed vaccine was not related to antibody titers (Saslaw and Carhart, 1961). Recently, a new lot of LVS manufactured under modern GMP conditions has been studied in animals and in two human clinical trials to evaluate its safety, reactogenicity, and immunogenicity (Pasetti et al., 2008;El Sahly et al., 2009;Mulligan et al., 2017). In a Phase I study, Sahly et al. reported that individuals vaccinated with LVS by scarification or subcutaneously (s.c.) frequently experienced ≥grade 2 (interfering with activity) systemic complaints, especially headache and fatigue; many experienced erythema and induration at the injection site of which ∼one-third had ≥grade 2 erythema (≥30 mm) and induration (≥15 mm). Of note, some individuals, especially those vaccinated by scarification, developed transient lymphangitis and papular satellite lesions surrounding the infection site. High dose scarification vaccination induced serologic immune responses and scarification and to a lesser extent s.c. vaccination induced high IFN-γ responses (El Sahly et al., 2009). In a Phase II study, Mulligan et al. compared the new lot of LVS to the existing USAMRIID LVS vaccine (Mulligan et al., 2017). They found that both vaccines appeared safe. Injection site reactogenicity was deemed generally mild and similar for the two vaccines; severe vaccine site reactions (erythema or induration >5 cm) occurred in 4.4 and 3.5% of vaccinees administered the new lot and USAMRIID vaccines, respectively. Similar percentages of the vaccinees admimistered the new lot or USAMRIID vaccine experienced severe (2 and 3%, respectively) or moderate (24 and 22%, respectively) systemic reactions. Both vaccines resulted in similar (94%) rates of seroconversion.
Thus, while LVS retains residual toxicity, it has demonstrated substantial albeit incomplete protection in humans against aerosol challenge with virulent subsp. tularensis SCHU S4. Although LVS has not been licensed for use in the U.S.A. and European Union, it is the only vaccine that has been shown to be reasonably safe and efficacious in humans. There is a general consensus that any vaccine that warrants further consideration as a human vaccine needs to be safer than LVS while providing protection comparable to or greater than that provided by LVS against aerosolized fully virulent F. tularensis subsp. tularensis. Because LVS retains significant virulence in animals, has toxicity in humans, and provides incomplete protection against aerosol challenge with SCHU S4, there has been great interest since 2001 in developing alternative vaccines-these are reviewed below and summarized in Tables 1-5. We focus here on vaccines that have been tested against subsp. tularensis SCHU S4 respiratory challenge and compared with LVS for efficacy and virulence.

Nutrient Metabolic Mutant
Pechous et al. constructed an LVS mutant with a deletion in the purMCD purine biosynthetic locus, LVS purMCD, by allelic exchange (Pechous et al., 2006), and complemented the purMCD strain in trans with wild type purMCD. The authors showed that the LVS purMCD mutant is defective for growth in medium containing limiting concentrations of purines and is defective for intra-macrophage growth, as it escapes from the phagosome but fails to replicate in the cytosol. The purMCD mutant is significantly attenuated in BALB/c mice with an LD 50 >5 × 10 6 by the i.p. route (vs. < 10 1 CFU for LVS). 100% of BALB/c mice immunized i.p. with 5 × 10 4 -5 × 10 6 CFU of the LVS purMCD mutant and challenged 21 days later i.p. with parental LVS (5 × 10 1 -5 × 10 3 ) survived the challenge, while 40% of naïve mice survived the challenge (Pechous et al., 2006). The authors furthered tested the efficacy of LVS purMCD, with or without boosting, against respiratory challenge with SCHU S4 and compared it with LVS. In contrast to the result after challenge of the immunized mice with LVS i.p., mice immunized i.n. once with LVS purMCD were not protected (0% survival) against i.n. challenge with SCHU S4, similar to sham-immunized mice, whereas mice immunized i.n. once with LVS had 100% survival. However, the protection was increased to the levels induced by LVS following a homologous booster vaccination ( Table 1, Nutrient metabolic mutant) (Pechous et al., 2008). These studies underscore the fact that protection against challenge with subsp. holartica LVS is not predictive of protection against challenge with subsp. tularensis SCHU S4, especially when the vaccination and challenge routes are different, in this case i.p/i.p. vs. i.n./i.n.

Oxidative Stress Response Mutants
Several oxidant mutants of LVS have been reported (Lenco et al., 2005;Bakshi et al., 2006Bakshi et al., , 2008Buchan et al., 2009;Melillo et al., 2009;Honn et al., 2012Honn et al., , 2017Ma et al., 2014;Suresh et al., 2015;Saha et al., 2017) among which LVS mutants with a point mutation in sodB (FTL_1791), encoding an iron superoxide dismutase (sodB Ft ), with a transposon insertion in a putative gene for the EmrA1 (FTL_0687) secretion protein (emrA1), or with a deletion in dsbA, encoding a disulfide oxidoreductases protein family homolog ( dsbA/FSC200) , have been tested in mice for their efficacy against respiratory challenge with SCHU S4 (Bakshi et al., 2006Straskova et al., 2015;Suresh et al., 2015).  Both sodB Ft and emrA1 LVS mutants are sensitive to oxidant stress and are more attenuated than the parental LVS in mice; C57BL/6 and BALB/c mice infected i.n. with 1 × 10 4 CFU of the sodB Ft mutant of LVS had 83 and 60% survival, respectively, while C57BL/6 and BALB/c mice infected i.n. with 1 × 10 4 CFU parental LVS had 8.3 and 0% survival, respectively (Bakshi et al., 2006). C57BL/6 mice immunized i.n. with 5 × 10 3 sodB Ft mutant and 21 days later challenged i.n. with 14 CFU SCHU S4 had 40% survival, significantly greater than that of naïve mice (0% survival) and mice immunized i.n. with 5 × 10 3 LVS (0% survival). The protection was similar following administration of a homologous booster vaccination (42% survival) against i.n. challenge with 103 CFU of SCHU S4 ( Table 1, Oxidative stress response mutants) . It is noted that the i.n. vaccination doses for the sodB Ft mutant and LVS were both close to the LD 50 i.n. and the interval between the last immunization and challenge was only 21 days. The emrA1 mutant of LVS is more attenuated than the sodB Ft mutant of LVS in mice, with an LD 50 > 10 6 i.n. in C57BL/6 mice (Ma et al., 2014). However, mice immunized with 1 x10 6 emrA1 and challenged 21 days later i.n. with 32 CFU of the heterologous subsp. tularensis SCHU S4 strain had 0% survival; mice homologously primed-boosted with 1 × 10 6 emrA1 via the i.n. or the i.d. route or both routes alternately (i.d./i.n,. or i.n./i.d.) and challenged 21 days later had 0-20% survival, similar to naïve mice ( Table 1, Oxidative stress response mutants). It is noted that mice immunized with emrA1 had 100% survival after i.n. challenge with the homologous parental LVS strain at doses up to 1 × 10 8 CFU LVS . Thus, the vaccine was poorly protective against SCHU S4 but highly protective against LVS. These results suggest that over attenuation of LVS resulted in a significant reduction in protective immunity. Importantly, these results once again show that efficacy against respiratory challenge with subsp. holartica LVS does not predict efficacy against respiratory challenge with subsp. tularensis SCHU S4 (see Summary and Discussion).

Putative Capsular and Membrane Mutants
F. tularensis Type B mutants defective in genes involved in putative capsular and membrane synthesis are highly attenuated in mice, including LVS mutants defective in capBCA (FTT0806, FTT0805, and FTT0798, respectively), wzy, wbtA, FTL_0057,

Other subsp. holarctica Mutants
Subsp. holarctica LVS mutants deficient in FTL_0552 (a transcriptional response regulation gene, pmrA), FTL_0291, or FTL_0304 also have shown attenuated phenotypes and protective efficacy against i.n. challenge with SCHU S4 in murine models. However, these mutants were not compared for efficacy with the parental LVS vaccine (Table 1, Other mutants) (Sammons-Jackson et al., 2008;Mahawar et al., 2013).

LIVE ATTENUATED F. NOVICIDA VACCINE CANDIDATES
F. novicida vaccine candidates have been constructed by deleting genes involved in the purine biosynthesis pathway (purA and purF), the T6SS (iglB, iglC, pdpB, and iglD), the response regulator (pmrA), or in other activities. A complete list of the F. novicida mutants can be found in a review article written by Pechous et al. (2009). However, only a few of the F. novicida mutants, including mutants with a deletion in purF, purA, iglB, iglD, or pmrA, have been tested against challenge with virulent SCHU S4, and even fewer against respiratory challenge with SCHU S4. In one study, BALB/c mice immunized i.p., but not s.c., with F. novicida U112 purF::cm were partially protected against i.p. challenge with the homologous U112 strain, but not against challenge with the heterologous SCHU S4 strain; a similar mutant, U112 purA::cm, was not protective against either U112 or SCHU S4 challenge (Quarry et al., 2007). Two F. novicida U112 iglB deletional mutants, U112 iglB and U112 iglB::fljB (expressing one domain D of the Salmonella typhimurium FljB flagellin, reportedly a TLR5 agonist), administered orally, have been shown to partially protect against i.n. or intratracheal (i.t.) challenge with SCHU S4 in mice and Fisher rats; however, efficacy was not compared with LVS ( Table 2; Cong et al., 2009;Signarovitz et al., 2012;Cunningham et al., 2014). An F. novicida mutant with a deletion in iglD (Fn IglD), although not protective in a murine model of tularemia, has demonstrated protection against i.t. challenge with SCHU S4 in the more resistant Fisher rat model and in the non-human primate model of tularemia (see below for details) (Chu et al., 2014). Another F. novicida mutant with a deletion in an orphan response regulator gene pmfA was protective against homologous U112 but not against heterologous SCHU S4 challenge ( Table 2; Mohapatra et al., 2007). Other studies also have shown that F. novicida and its derivative mutants can induce protective immunity against homologous respiratory challenge with wild-type F. novicida; however, these mutants have not been shown capable of inducing protection against respiratory challenge with subsp. tularensis in murine models of pneumonic tularemia by traditional i.d. or i.n. routes (Pammit et al., 2006;Sanapala et al., 2012).

T6SS Mutant
The Francisella pathogenicity island (FPI) proteins, including IglD, are required for bacterial phagosome escape, intracellular replication, and virulence, and are components of a T6SS apparatus (Clemens et al., 2015). Chu et al. reported on the protective immunity induced by iglD deletion mutants of subsp. tularensis (Ftt iglD) and F. novicida (Fn iglD) in murine, Fischer rat, and non-human primate models of pneumonic tularemia (Chu et al., 2014). Ftt iglD and Fn iglD are defective for intramacrophage replication and attenuated in mice (the LD 50 for Fn iglD is 9.7 × 10 8 CFU i.n. in BALB/c mice vs. ∼10 3 CFU for LVS). Mice immunized i.n. with Fn iglD were fully protected against subsequent homologous challenge with 10 3 CFU of the parental F. novicida U112 strain; however, mice immunized i.n. with Fn iglD or Ftt iglD were not protected against respiratory challenge with the heterologous subsp. tularensis SCHU S4 strain. Arguing that the mouse may be too sensitive to tularemia, Chu et al. evaluated the Fn iglD vaccine in the Fischer rat, an animal model that has been shown to be more resistant to various Francisella subspecies than mice and nonhuman primates. Fischer rats vaccinated with Fn iglD orally or i.t. and challenged i.t. 30 days later with subsp. tularensis SCHU S4 had 83-100% survival post-challenge vs. 17-25% survival postchallenge for naïve rats. Of note, Fn iglD was not compared with LVS for efficacy in the murine and rat models. Cynomolgus macaques vaccinated with 10 8 CFU Fn iglD via bronchoscopy and challenged via head-only aerosol inhalation 30 days later with ∼2,500-5,000 CFU SCHU S4 had 83% survival, somewhat less than those immunized s.c with 10 8 CFU LVS (100% survival); mock vaccinated animals died 7-13 days post-challenge ( Table 2; Chu et al., 2014).

LIVE ATTENUATED SUBSP. TULARENSIS VACCINE CANDIDATES
Because subsp. tularensis and subsp. holarctica differ in genetic organization, antigen expression, and disease pathogenesis, it was hypothesized that attenuated mutants on the background of subsp. tularensis may offer better protection against respiratory challenge with the parental subsp. tularensis than mutants derived from subsp. holarctica Wu et al., 2005). Thus, a series of mutants have been generated in the SCHU S4 background with mutations previously shown to attenuate LVS and tested in murine, rabbit, and non-human primate models of pneumonic tularemia (Twine et al., 2005(Twine et al., , 2012Pechous et al., 2008;Qin et al., 2008Qin et al., , 2009Conlan et al., 2010;Michell et al., 2010;Shen et al., 2010;Rockx-Brouwer et al., 2012;Reed et al., 2014;Santiago et al., 2015). Summarized below are some SCHU S4 mutants that have been tested against respiratory SCHU S4 challenge.

Nutrient Metabolic Mutants
Based on the capacity of the attenuated LVS purMCD mutant to induce protective immunity, Pechous et al. generated a defined subsp. tularensis SCHU S4 purMCD mutant and showed that it is significantly more attenuated in mice when delivered by the i.d. (LD 50 > 10 6 CFU) and i.n. (LD 50 > 10 6 CFU) routes than its parental SCHU S4 strain (LD 50 < 10 CFU i.n. or i.d.) (Pechous et al., 2008). However, mice immunized i.n. with SCHU S4 purMCD had tissue damage in the lung and were protected no better than mice vaccinated with LVS or with an analogous LVS purMCD mutant. BALB/c mice immunized once i.n. with 10 4 CFU SCHU S4 purMCD, 10 6 CFU LVS purMCD or 10 2 CFU LVS and challenged 42 days later i.n. with 100 CFU SCHU S4 had 14, 0, and 100% survival, respectively; mice immunized twice i.n. with the same vaccines had 71, 100, and 100% survival after i.n. challenge with 100 CFU SCHU S4 (Results for SCHU S4 mutant and LVS shown in Table 3, Nutrient metabolic mutants) (Pechous et al., 2008); thus homologous i.n. boosting substantially improved the efficacy of the SCHU S4 purMCD vaccine. Several other SCHU S4 mutants that are deficient in nutrient metabolic enzymes have also been tested against respiratory challenge with SCHU S4 in a murine or rat model ( Table 3, Nutrient metabolic mutants) (Reed et al., 2014;Santiago et al., 2015).

Putative Capsular Mutants
Defined subsp. tularensis SCHU S4 mutants with a single deletion in capB ( capB::Cam) or double deletions in FTT0918 and capB ( FTT0918 capB) have been developed by Michell and Conlan et al. and evaluated as vaccine candidates Michell et al., 2010). Michell et al. showed that the s.c. median lethal dose for SCHU S4 capB::Cam, a single capB deletion mutant with a chloramphenicol resistance cassette (Cam) inserted at the deleted capB locus, is >1.6 × 10 6 CFU in BALB/c mice. The level of protection afforded by s.c. immunization with 10 4 SCHU S4 capB::Cam is comparable to that of s.c. immunization with 10 4 CFU LVS against systemic (s.c.) challenge with 10 3 SCHU S4 56 days later. The SCHU S4 capB::Cam vaccine was not tested against respiratory challenge (Michell et al., 2010). Conlan et al. showed that SCHU S4 FTT0918 capB administered i.d. is as attenuated as LVS in BALB/c mice (LD 50 ≥ 10 7 CFU) but more virulent than LVS in C3H/HeN mice (LD 50 < 10 5 CFU vs. LD 50 > 10 5 CFU for LVS). In a virulence study, BALB/c and C3H/HeN mice infected orally with 10 8 CFU SCHU S4 FTT0918 capB had 80% (12/15) and 60% (9/15) survival, respectively . In a study summarized in Table 3, Capsular mutant, BALB/c or C3H/HeN mice were not immunized, or immunized i.d. with 10 3 CFU or orally with 10 8 CFU FTT0918 capB or LVS, homologously boosted or not boosted orally, and challenged by aerosol 6 weeks later with 20 CFU SCHU S4. BALB/c mice immunized i.d. or orally with FTT0918 capB had 40 and 0% survival, respectively, after SCHU S4 aerosol challenge whereas none of the C3H/HeN mice immunized i.d. or orally with FTT0918 capB survived the challenge; nor did the naïve mice or mice immunized i.d. or orally with LVS. Homologous boosting orally did not significantly improve the protection against aerosol challenge with SCHU S4 except that C3H/HeN mice immunized i.d. and boosted orally with FTT0918 capB had improved survival (50% vs. 0% without boosting) . In an earlier study, Twine et al. showed that the SCHU S4 FTT0918 single deletional mutant (SCHU S4 FTT0918) had an i.d. LD 50 of ∼10 5 CFU and was thus 10-fold more virulent than LVS in BALB/c mice. BALB/c mice immunized i.d. with 10 5 FTT0918 and challenged i.n. 9 weeks later with 10 CFU FSC 033 had 33% survival, while LVS-immunized mice had 0% survival after aerosol FSC033 challenge ( Table 3, Other mutants) (Twine et al., 2005). The results of these studies suggest that single deletional mutants in the background of SCHU S4, although more attenuated than the parental SCHU S4 strain, are generally still more virulent than LVS. Introducing a second major attenuating deletion attenuates the mutant further but diminishes its capacity to induce protective immunity such that it is comparable to or less protective than LVS vaccination against virulent subsp. tularensis challenge.

SUMMARY AND DISCUSSION
The vaccines summarized here include deletional mutants of subsp. holarctica LVS or FSC200 (Table 1), F. novicida U112 ( Table 2), and subsp. tularensis SCHU S4 (Table 3), and homologous LVS mutants overexpressing F. tularensis antigens and heterologous vectors expressing F. tularensis antigens ( Table 4). The unlicensed LVS vaccine, the only vaccine available in the U.S.A., has shown substantial albeit incomplete efficacy in humans but retains residual toxicity. As noted, it would seem reasonable to expect that any vaccine warranting further consideration as a human vaccine satisfy the following two criteria: 1) the vaccine is safer than LVS; and 2) the vaccine provides protection equivalent to or greater than LVS against respiratory challenge with subsp. tularensis SCHU S4 in appropriate animal models. The data summarized here show that it is relatively easy to generate genetically defined mutants that are safer than LVS so as to meet the first criterion; however, it is difficult to control the balance between attenuation and protection so as to meet the second criterion simultaneously. Practically speaking, among the vaccine candidates summarized here, only a few meet both criteria. Often, the LVS-derived vaccine candidates need either prime-boost vaccination or intranasal administration in order to provide very high-level protective immunity.
With respect to the relative efficacies of vaccine candidates, in many cases, vaccines and challenge strains have been tested under different conditions, e.g., different preparations of vaccine and challenge strains; different routes of vaccine administration (mucosal vs. systemic); different immunizationchallenge intervals (typically ranging from 3 to 6 weeks); different routes and doses of challenge strain (less virulent subsp. holarctica or F. novicida vs. highly virulent subsp. tularensis SCHU S4); and different animal models (mice, rats, rabbits, or non-human primates), making it difficult to compare their relative efficacies. Of course, head-to-head comparisons are the most reliable way to compare vaccine candidates, but except for comparisons with LVS-the only vaccine tested and shown protective in humans and effectively the current gold standardthis is rarely done. We summarize vaccines that were tested against respiratory challenge with SCHU S4 and compared with LVS for efficacy in Table 5 and discuss factors that affect vaccine efficacy below.

Vaccine and Challenge Strain Stock Preparation
As summarized in Tables 1-4, some studies used LVS as a positive control to evaluate the efficacy of various vaccines against respiratory challenge with SCHU S4 strains. It is interesting that while the median/mean survival time for unvaccinated or sham-immunized mice fell between 4 and 6 days post-respiratory SCHU S4 challenge in most studies, the immunity induced by LVS varied significantly in these studies. Of note, some vaccine strains were produced on solid agar while others in broth medium. Eigelsbach et al. conducted a study on virulence and immunogenicity of live vaccine strains. The study showed that LVS harvested from modified casein partial hydrolysate medium (MCPH) appeared more virulent and immunogenic in mice than LVS harvested from glucose cysteine hemin agar (GCHA); however, LVS prepared from GCHA and MCPH induced comparable protection in guinea pigs (Eigelsbach and Downs, 1961). While it is impossible to compare vaccines among different studies, it would be helpful to include LVS prepared using a standardized method as a positive control in efficacy studies.

Mucosal vs. Systemic Route of Vaccine Administration
As noted above, several live attenuated vaccine candidates induce more potent protective immunity against respiratory challenge with subsp. tularensis SCHU S4 when administered by the mucosal respiratory route (e.g., i.n., aerosol, or intratrachea), or by the mucosal oral route than by a traditional systemic route (i.d., s.c., or i.p.) Conlan et al., 2005;Wu et al., 2005;Pechous et al., 2008;Jia et al., 2010Jia et al., , 2013, with a few exceptions (Rockx-Brouwer et al., 2012). In one study, an alternative mucosal (i.e., oral) route for delivery of F. novicidaderived vaccine candidates showed greater protective immunity than the i.n. route against respiratory challenge with the virulent SCHU S4 strain (Cong et al., 2009). The i.n. route raises some additional safety concerns; that said, the current live attenuated flu vaccine administered i.n. has an excellent safety record (Pavot et al., 2012).

Interval Between Vaccination and Challenge
As summarized in Tables 1-4, intervals between the last or the only vaccination and challenge ranged between 3 and 6 weeks in different studies. Eigelsbach and Downs investigated the effect of the vaccination-challenge interval on the immunity of LVS in mice and guinea pigs (Eigelsbach and Downs, 1961). Albino mice (Webster) vaccinated with LVS s.c. 15-30 days prior to challenge with 10 3 CFU of SCHU S4 s.c. had a higher survival rate (68-88%) than mice vaccinated 60 days prior to challenge (49-56%). This difference was appreciably greater in guinea pigs. Guinea pigs (Harley) vaccinated 15 days prior to challenge had 45% survival vs. 5% survival for those vaccinated 30 days prior to challenge. LVS administered i.d. or i.n. to mice was cleared from all the infected organs by 3 weeks (Jia et al., 2010). Thus, it would be helpful to challenge animals at a longer interval, i.e., more than 3 weeks, to minimize the effect of non-specific immunity and to allow evaluation of various vaccines under more rigorous conditions.

Vaccine Genetic Background
Vaccines have been generated in the background of subsp. holarctica, subsp. tularensis, F. novicida, and heterologous vectors. The subsp. holarctica LVS strain is the only vaccine tested and shown efficacious in humans against virulent SCHU S4 challenge; however, its residual toxicity in humans and unknown attenuation mechanism may have presented obstacles to its licensure. LVS-derived vaccines are safer than LVS and retain the large antigen pool that might be required for protection against heterologous challenge with the virulent subsp. tularensis strain. However, a booster or intranasal vaccination route are generally required for enhanced vaccine-induced protection ( Table 1). F. novicida is less virulent than LVS in mice, guinea pigs, rabbits and Fisher rats, and rarely infects humans (so far only 12 cases have been documented) (Kingry and Petersen, 2014). However, in contrast to LVS, F. novicida differs from F. tularensis in the mechanism of pathogenicity; they differ in cell surface structure, means of cellular entry, types of cells infected in vivo, and ability to evade host immune responses (Kingry and Petersen, 2014). The F. novicida parental strain and its derivatives have not shown efficacy against respiratory challenge with SCHU S4 in mice and guinea pigs by i.d. or i.n. route; the protective immunity induced by the i.t. or oral route against i.t. challenge with SCHU S4 in Fisher rats was not compared with that of LVS ( Table 2). The subsp. tularensis-derived vaccines are generally more efficacious than LVS against homologous challenge with SCHU S4. However, this may be true only for SCHU S4-derived vaccines with a single deletion. A second major attenuating deletion significantly reduces vaccine efficacy ( Table 3). The vaccines constructed using a heterologous vector (i.e., Listeria monocytogenes) have a much more limited antigen pool, need multiple vaccinations, and are less efficacious than LVS ( Table 4).
The impact of vaccine genetic background on vaccine efficacy is also evident in comparisons of vaccines derived from different subspecies but comprising the same deletional gene mutation. For example, some gene deletional mutants in the LVS background are both highly attenuated and able to induce protective immunity against SCHU S4 challenge, but this may not be true of the same deletion in the F. novicida or SCHU S4 background. An LVS mutant with a deletion in purMCD, LVS purMCD, is both highly attenuated and able to induce protection against virulent SCHU S4 i.n. challenge; protective immunity is similar to that induced by LVS i.n. vaccination. However, an attenuated SCHU S4 with the same deletion, SCHU S4 purMCD, provides limited protection against SCHU S4 challenge, less than that induced by LVS vaccination (Pechous et al., 2006(Pechous et al., , 2008. Attenuated F. novicida mutants with similar deletions, purCD and purM, are not able to induce protection against the homologous wild-type strain challenge (Tempel et al., 2006;Quarry et al., 2007). The same findings have been reported for guaA, guaB, and guaBA mutants (Santiago et al., 2009(Santiago et al., , 2015. In contrast, clpB deletional mutants in the subsp. tularensis (SCHU S4) background induced full protection against respiratory challenge with the parent SCHU S4; a mutant with the same deletion in LVS showed no protection (Golovliov et al., 2013). Hence, the vaccine's genetic background is an important determinant of the vaccine's attenuation and protective efficacy.

Animal Model
Various animal models-mice, rats, rabbits, guinea pigs, and nonhuman primates-have been used to study tularemia vaccine efficacy as reviewed recently by Elkins et al. (2016); however, they differ in their sensitivity to the highly virulent SCHU S4 strain and in vaccine-induced protective immunity against SCHU S4 challenge. Mice, guinea pigs, rabbits, and primates are more sensitive to SCHU S4 than rats (White rats and Fisher rats). The LD 50 for SCHU S4 by a systemic route (s.c. or i.d.) is 1 CFU in mice, 1 CFU in guinea pigs, 1 CFU in rabbits, and >10 8 CFU (s.c.) in White rats; the LD 50 of SCHU S4 by the respiratory route is 1-3 CFU in mice (i.n.), 1 CFU in cynomolgus macaques (aerosol), and >5 × 10 2 CFU in Fisher rats (intratreacheal); in humans, as few as 10-50 CFU SCHU S4 can cause clinical tularemia (Eigelsbach et al., 1951;McCrumb, 1961;Saslaw et al., 1961a,b;Schricker et al., 1972;Kostiala et al., 1975;Jemski, 1981;Conlan et al., 2003;Barker and Klose, 2007;Lyons and Wu, 2007;Raymond and Conlan, 2009;Wu et al., 2009;Ray et al., 2010;Chu et al., 2014;Kingry and Petersen, 2014;Hutt et al., 2017;Nguyen et al., 2017). Mice, rats, guinea pigs, and primates also differ in the degree of vaccine-induced protective immunity against virulent SCHU S4 challenge. Mice, guinea pigs, and humans vaccinated with the Foshay-killed vaccine were not protected against aerosol challenge with SCHU S4, while White rats were protected (Lyons and Wu, 2007). Furthermore, different strains within species show substantial variability in their susceptibility to challenge and the degree of vaccine-induced protection. LVS-immunized BALB/c mice are more resistant than LVSimmunized C57BL/6 mice to low dose aerosol or high dose i.d. challenge with virulent subsp. tularensis FSC 033 strain ; however, C57BL/6 mice clear subsp. tularensis SCHU S4 infection more rapidly (Fritz et al., 2014). White rats and Fisher rats are more resistant to SCHU S4 challenge than Sprague-Dawley rats (Lyons and Wu, 2007;Raymond and Conlan, 2009). African green monkeys and cynomolgus monkeys are more sensitive to SCHU S4 aerosol challenge than Rhesus monkeys, with an aerosol lethal dose of 40, 32, and 2.8 × 10 5 CFU, respectively (Glynn et al., 2015). In addition, recent studies show that vaccines inducing strong protective immunity in one animal model may not do so in another model. For example, guaBA, aroD, and fipB (FTT1103) mutants have been tested in both mice and rabbits (Qin et al., 2009;Santiago et al., 2009Santiago et al., , 2015Reed et al., 2014). While the SCHU S4 guaB and guaA mutants, administered via scarification, induce partial protective immunity against respiratory SCHU S4 challenge in the New Zealand rabbit model, they do not do so in C57BL mice (Reed et al., 2014;Santiago et al., 2015). In contrast, the SCHU S4 fibB (FTT1103) mutant, while fully protective in C57BL mice and partially protective in BALB/c mice, showed no protection against SCHU S4 challenge in the rabbit model (Qin et al., 2009;Reed et al., 2014). Other reports have reported differences in protective immunity conferred by the same vaccine in the murine and Fisher rat models (Cong et al., 2009;Signarovitz et al., 2012;Chu et al., 2014). These findings highlight the impact of animal model on the outcomes of preclinical vaccine efficacy studies.
The low natural incidence of tularemia renders field trials of efficacy unfeasible, and ethical considerations make it unlikely that tularemia vaccines will ever again be tested for efficacy in human challenge studies. In such situations, the FDA has promulgated the Animal Rule, whereby a drug or vaccine may be licensed on the basis of efficacy in relevant animal models. That raises the question as to which animal models are most relevant. If a vaccine is ineffective in the highly sensitive mouse model, are efficacy studies in the relatively resistant Fisher rat model an acceptable substitute? Proponents of this rat model would argue that the susceptibility of the Fisher rat to various F. tularensis strains more closely approximates that of humans than the mouse. Be it as it may, from the standpoint of efficacy, a vaccine that is efficacious in the most sensitive animal models would seem to be a more reliable one for humans, who likely have varying susceptibility to infection depending upon numerous host variables, than a vaccine efficacious only in relatively resistant models.

CONCLUSION
Since 2001, several promising live attenuated vaccine candidates have been developed that meet what would seem to be minimal criteria for a new human vaccine-safety greater than LVS and protective efficacy equivalent to or greater than LVS against respiratory challenge with subsp. tularensis SCHU S4 in appropriate animal models ( Table 5). Some of these vaccines additionally demonstrate efficacy comparable to or greater than that of LVS against SCHU S4 respiratory challenge in the demanding mouse model. Vaccines meeting this higher standard include LVS mutants (LVS purMCD, LVS sodB Ft , LVS capB, and LVS::wzy), another F. tularensis type B mutant (FSC200 clpB), LVS capB overexpressing F. tularensis T6SS proteins (LVS capB/IglA, IglC or IglABC), a LVS capB-rLm/IglC heterologous prime-boost vaccine, and a single deletional SCHU S4 mutant ( purMCD). Another single deletional SCHU S4 mutant, SCHU S4 clpB, and a double deletional SCHU S4 mutant, SCHU S4 clpB capB demonstrates efficacy greater than LVS but its attenuation is equivalent to LVS. That these SCHU S4 mutants contain only one major attenuating deletion raises a safety concern, namely reversion to virulence. A second major attenuating deletion would alleviate this concern, but retaining efficacy upon the introduction of a second major attenuating mutation has been a major challenge for single deletional SCHU S4 mutant vaccines. The efficacy of the new vaccines has typically been greatest when administered by the intranasal or another respiratory route, but administering vaccines by this route raises additional safety issues. Nevertheless, vaccines that are safer than LVS and at least as efficacious, including by the i.d. route, have been developed, and they are promising candidates for going forward into more advanced animal studies such as in the non-human primate model and human safety trials.

AUTHOR CONTRIBUTIONS
QJ and MAH wrote the article. The comments in this article represent the opinions of both QJ and MAH.