Johne's disease is caused by Mycobacterium avium subspecies paratuberculosis (hereafter referred to as MAP), an acid-fast bacillus that can be distinguished from other closely related mycobacteria by its unique requirement for the mycobactin J siderophore in artificial culture media (Merkal and Curran, 1974). This major veterinary pathogen can infect many species of animals (Whittington et al., 2011), but its impact is felt most profoundly on commercial ruminant livestock such as cattle, sheep, goats and deer. In the United States (US) alone, current economic loses to the dairy industry are unknown, but previously was estimated at between 200 million and 1.5 billion dollars annually (Stabel, 1998). Almost three decades ago, the prevalence of Johne's disease in US dairy cattle was estimated at between 3 and 18% based on slaughterhouse surveys (Chiodini and Van Kruiningen, 1986; Merkal et al., 1987). A more recent national level serological survey conducted by the National Animal Health Monitoring System (NAHMS) in 2007 suggested MAP prevalence on US dairy farms had risen above 30%. Unfortunately, this percentage continues to climb with the passage of time and implementation of more sensitive diagnostic tests. The most recent US dairy herd prevalence estimates are as high as 90% (Lombard et al., 2013) and New Zealand farmed deer herd estimates are at 59% (Stringer et al., 2013a). Collectively, these data suggest that MAP infection has long been endemic in the US and most likely across the world wherever dairy cows, sheep, goats and deer are intensively raised.

Vaccination against MAP infection has long been thought to be the best intervention strategy for this chronic and debilitating disease that is difficult to diagnose and slow to manifest. Animals actively shed large quantities of MAP before being diagnosed or exhibiting clinical signs—resulting in a transmission cycle that is very difficult to interrupt using traditional management strategies alone. Sub clinically infected animals transmit disease while appearing healthy and remaining undetectable by culture or PCR based approaches since these animals often shed MAP in a sporadic or intermittent manner. In stark contrast to the subversive “trickle and stealth” shedding pattern of MAP frequently observed in subclinical animals, there exist symptomatic, high shedder animals that excrete prodigious levels of organisms (up to 10⁸ CFU per gram) in their feces, which provide the source of significant on-farm contamination (Pradhan et al., 2011). Diagnostic tests for Johne's disease are improving, but accurate detection of all infected animals, especially those that are early in infection and transmitting the organism within a herd, is still not possible. This fact makes test and cull strategies ineffective, except when targeting only high-shedding animals (Lu et al., 2008; Bastida and Juste, 2011). Therefore, it is widely recognized that unless animals can be detected early during infection, vaccination remains the best hope for controlling and preventing Johne's disease.

The ideal vaccine would completely prevent infection and/or promote protective immunity thus blocking both horizontal and vertical transmission. The current vaccines for Johne's disease fall far short of this high standard. MAP vaccines have been shown to be effective at lowering fecal shedding levels (Kalis et al., 2001; Faisal et al., 2013a), tissue colonization (Sweeney et al., 2009) or clinical disease incidence (Stringer et al., 2013b), but do not completely eliminate all three. Subunit vaccines against MAP are likely to obviate some of the shortcomings of whole-cell vaccines, such as severe inflammation and granuloma formation at the injection site. However, subunit vaccines that have been tested thus far have yielded incomplete protection results in murine models of infection (Koets et al., 2006; Stabel et al., 2012) and even when combinations of proteins are used in calves and goats (Koets et al., 2006; Kathaperumal et al., 2008, 2009). For example, MAP was colonized in the lymph node and spleen at similar levels in control and vaccinate mice using a protein cocktail (Stabel et al., 2012). Similarly, there were no significant MAP burden differences in the liver and mesenteric lymph node at 8 weeks post-challenge when immunizing mice with a 74 kDa polyprotein, but there was a significant difference in the liver (Chen et al., 2008). However, additional studies using more animals, longer term to monitor protection (1–3 years) and field trials are needed. Unfortunately subunit vaccines are expensive to produce whereas the manufacture of attenuated mutants is low in cost and easy to produce. These considerations are even more important for food animal health. Thus, there are several considerations, but research in this area demonstrates the slow yet steady progress in vaccine development for Johne's disease.

Efforts to improve upon current vaccine formulations have long been hampered by the lack of standardized challenge models and the high costs associated with trials conducted in natural ovine, caprine, deer or bovine hosts. Progress is however being made, and standardized animal models (mouse, goat, cow) for Johne's disease have been developed by a group of international investigators under the auspices of the Johne's Disease Integrated Program (JDIP), a USDA-funded research consortium (Hines et al., 2007) to help facilitate comparisons between different vaccine trials. However, the costs and logistical challenges of MAP vaccine trials, particularly in the natural hosts, remain a major concern. And even with an efficacious vaccine, it is remotely possible that MAP could be maintained within a herd due to vertical transmission as noted in a recent vaccine modeling study (Lu et al., 2013). Finally, there are negative implications of JD vaccination on both Johne's disease and bovine TB diagnostics as will be discussed further below.

Consistent with what is known about other intracellular pathogens, a MAP vaccine that drives the immune response to a proinflammatory Th1 profile and prevents a shift to the humoral Th2 response, maybe more effective in delaying progression of the disease to a clinical state (Coussens, 2004; Stabel, 2010). However, it has recently been suggested that humoral immunity may also be important against M. tuberculosis infection (Achkar and Casadevall, 2013). Nonetheless, current dogma suggests a late humoral response is not a protective immune response against Johne's disease. The focus of the JDIP vaccine studies was on live attenuated mutants rather than subunit or DNA vaccines because they are more likely to generate the beneficial Th1 immune response, which is considered the most protective against mycobacterial infections (Stabel, 2000; Hostetter et al., 2002). The cause for the switch from Th1 to Th2 immune responses in Johne's disease is unknown, however mathematical modeling of the bovine immune response to MAP suggests it may be attributed to extracellular MAP that persists outside of macrophages (Magombedze et al., 2014). Although these inferences need to be carefully tested experimentally, it is possible a mutant strain deficient in this capability might be the critical factor to an efficacious vaccine for Johne's disease.

Despite significant recent advances in the ability to make defined mutants, there have been only a limited number of trials testing live attenuated vaccines against MAP. These have been largely limited to in vitro studies or trials in the mouse (Scandurra et al., 2009; Settles et al., 2014), and except in a few cases (Scandurra et al., 2010; Faisal et al., 2013b), reflect a striking absence of robust and cost effective strategies for evaluating vaccine candidates in the relevant natural ruminant host. To help address this unmet need, and establish a rational framework based for the evaluation of novel vaccine candidates against MAP, we adopted an inclusive and collaborative approach that may also serve as a guide for future vaccine evaluation trials for MAP and other major animal pathogens. We here present the state-of-the-field summary for vaccination against Johne's disease, describe current candidate live attenuated mutants of MAP, and present an overview of the JDIP three-phase vaccine trial and lessons learned.

It is paradoxical that vaccination against MAP has not been widely used in the US given the high prevalence of Johne's disease in this country. Currently, there are at least three commercial Johne's disease vaccines worldwide, but only one (Mycopar®) approved vaccine for use in the US (Bastida and Juste, 2011). Silirum® is produced by Zoetis and was used as the control vaccine in the JDIP mouse and goat trials discussed below. Mycopar® is manufactured by Boehringer Ingelheim Vetmedica Inc. whereas Gudair® is produced by CZ Veterinaria in Spain and is being used in countries with large sheep populations (Bastida and Juste, 2011). Importantly, all three consist of an inactivated whole-cell preparation of either M. avium subspecies avium or subspecies paratuberculosis. As summarized by Bull et al. live attenuated strains were used in the 1960s and 1970s and then concerns about their use in the 1980s forced a shift to using killed vaccine formulations (Bull et al., 2013). Live vaccine strains can now be revisited with the technology used to create defined mutations.

What does the scientific literature report about the efficacy of these commercial MAP vaccines? In three dairy herd operations with Johne's disease, investigators showed that fewer Mycopar® vaccinated cattle had detectable fecal culture positive results compared with controls (Knust et al., 2013). This same vaccine also showed partial protection in a goat challenge model (Faisal et al., 2013a). However, the inactivated strain used in the Mycopar® formulation is not MAP, but an isolate called strain 18 that is nonetheless a member of the M. avium species. The manufacturer data sheet still lists Mycopar® as an “inactivated M. paratuberculosis bacterium suspended in oil.” Interestingly, this parallelism occurs with the suboptimal response of BCG, derived from the related species M. bovis, as a vaccine against tuberculosis (Mangtani et al., 2014). Gudair consists of heat-killed MAP 316F and is designed for use in sheep and goats according to the manufacturer. Gudair® has been shown to be an effective vaccine in Merino sheep (Reddacliff et al., 2006). Vaccination not only delayed the onset of fecal shedding, but reduced mortality attributed to Johne's disease by 90% in the 200 vaccinated sheep. Silirum® is a killed strain of MAP, also related to 316F, that has been tested in Australian cattle. This vaccine has recently been shown to reduce prevalence of clinical disease, as measured by lymph node pathology and fecal culture at slaughter, in farmed deer in New Zealand (Stringer et al., 2013b). Currently, there are no subunit, live or DNA vaccines commercially available against Johne's disease.

Although attenuated MAP, generated by repeated subculture of the strain, had been used as a live vaccine long ago (Saxegaard and Fodstad, 1985), no defined mutants of MAP had been tested as a vaccine simply because no method was available for their construction. In 1995, the first concerted effort to develop the genetics of MAP was performed in Raúl Barletta's laboratory at the University of Nebraska (Foley-Thomas et al., 1995). Studies in his laboratory demonstrated for the first time that MAP could be transformed with foreign DNA as well as transfected with bacteriophage DNA. Furthermore, they found that MAP could be productively infected with the mycobacteriophage TM4 and later demonstrated the use of a thermosensitive derivative of TM4 to generate the first transposon mutant library (Harris et al., 1999). This methodological advance was adapted by other laboratories to create their own transposon mutant banks (Cavaignac et al., 2000; Shin et al., 2006). Phage-mediated techniques are now being applied to construct directed knockout mutations in MAP using allelic exchange. This technique worked well in M. tuberculosis, but the allelic exchange efficiency was low in MAP until investigators at Washington State University tweaked the protocol to increase efficiencies to generate ΔrelA, Δlsr2, and ΔpknG knockout strains (Park et al., 2008a). Now robust enough to be performed in other laboratories, this method enabled a more strategic approach to obtaining the precise knockout of a specific virulence gene rather than depend on random transposon mutagenesis and screening approaches to find a suitable knockout to try as a vaccine. The ppiA (Scandurra et al., 2010), relA, lsr2, pknG (Park et al., 2008a), leuD, mpt64, and secA2 genes (Chen et al., 2012) have all been successfully knocked out using this method. Each of these mutants has also been evaluated as a vaccine candidate as discussed in the next section.

MAP vaccination studies have been thoroughly summarized for cattle, sheep and goats by examining pathological, epidemiological and production effects of vaccination (Bastida and Juste, 2011). This meta-analysis revealed that while vaccination is useful in limiting microbial contamination and production losses, most of the vaccines are similar in content and preparation. These findings have recently been reinforced by the reduction of fecal contamination in three dairy herds using a killed vaccine (Knust et al., 2013). The possibility exists that new generations of vaccine formulations, including live attenuated, may further improve on the killed vaccine preparations currently being used. This section summarizes the pertinent literature on MAP vaccination as it relates to the JDIP trials.

Subunit vaccines have been delivered using heterologous hosts such as Salmonella (Chandra et al., 2012), or Lactobacillus (Johnston et al., 2014). Only two DNA vaccines against MAP have been tested in mice. One used expression library DNA immunization of Balb/cJ mice and challenged with a recently isolated bovine MAP strain 6112 (Huntley et al., 2005). Four pools of 108 clones each protected mice from colonization of spleen and mesenteric lymph nodes. The second study tested a cocktail of five genes encoding antigen 85A, B, and C along with superoxide dismutase and 35-kDa protein by intramuscular injection of C57/BL6 (Park et al., 2008b). There was a significant reduction in colonization of the liver and spleen of mice immunized with this cocktail. The subunit vaccines from both studies induced a Th1 immune response as measured by interferon gamma; however, none of these DNA formulations were tested in cattle or other ruminant hosts.

Several MAP vaccine candidates have been analyzed in isolation making it difficult to directly compare candidates between studies. Some have been bench marked against a commercially available vaccine (Faisal et al., 2013a; Hines et al., 2014) while other studies lack even that as a reference. As mentioned, all commercial vaccine formulations consist of a killed whole-cell preparation of MAP in an oil emulsion, but there are many other types of vaccines. Still other vaccine formulations included live M. avium bacteria, but it is not clear if or how they were attenuated (Begg and Griffin, 2005). In the current JDIP vaccine trial, we took a different approach in identifying a viable, yet attenuated vaccine strain, reasoning that a live strain will maintain a proinflammatory Th1 response (Stabel, 2000; Coussens, 2004). Th1-associated cytokines including interferon gamma, interleukin-2 (IL-2), IL-12 and tumor necrosis factor alpha were measured after antigen stimulation of cultured peripheral blood monocytes in these studies.

While there are no published bovine vaccine trials in Johne's disease, there are a number of caprine and ovine vaccine trials. Furthermore, a number of MAP vaccine studies in mice have been reported, but they are too numerous to adequately be summarized here. Importantly, only a few studies that began using the mouse model were further evaluated in a ruminant host (Scandurra et al., 2010; Park et al., 2011) and none using the number of mutant strains in the JDIP trials. This was a key component in the design of the JDIP vaccine trial. Using the data described in this special topics issue, we can now assess the predictive value of the mouse trial in obtaining good candidates for the ruminant host.

It was against this background that the idea for establishing independent testing labs to evaluate all the available vaccine candidates was conceived. This multi-institutional study was based on the following principles. Through the formation of a research consortium on Johne's disease, which was funded by the USDA-NIFA from 2002 to 2011, this governing body was successful in setting up a world-wide search for the best performing vaccine candidates. The idea behind this project was to combine the resource and intellect of JDIP with the financial resource of USDA-APHIS-VS to identify the best possible attenuated vaccine available. On January 12, 2008 a meeting was held in Chicago, Illinois to devise a coordinated effort to test attenuated vaccine candidates submitted by Johne's disease researchers. Although a few vaccine trials have tested individual mutants in macrophages, mice and goats, no trial of this scope using 5 or more mutants tested in parallel has ever been performed.

Vaccination against Johne's disease has considerable potential as a key management tool to control disease and transmission in ruminant livestock. Specifically, the use of attenuated mutants provide several advantages that include stimulation of a broad cellular immune response, ease of delivery, built in adjuvant characteristics and comparative cost. Several lessons were learned from the three-phase JDIP vaccine project as discussed below, and provide a rational framework for testing future vaccines against Johne's disease.

First, the results suggest that single gene knockout strains may not represent optimal vaccine candidates for MAP given the need to satisfy the “Goldilocks rule”—where the level of attenuation has to be just right in order to survive or replicate in target cells and stimulate a protective immune response but without causing disease. Consistent with this, while MAP vaccine candidates evaluated in the current trials were single gene knockouts, recent studies showed that a double mutant constructed in M. tuberculosis (ΔfbpAΔsapM was highly attenuated in human macrophages and, intriguingly, more immunogenic than the single knockout (ΔfbpA) (Saikolappan et al., 2012). These data also suggest the importance of careful targeting of genes/proteins for deletion based on prior knowledge of their role in pathogenesis, immunogenicity, or pathogen survival in vivo. For instance, the rationale for construction of the M. tuberculosis (ΔfbpAΔsapM strain was that ΔfbpA, which lacks Ag85, had already demonstrated immunogenicity and protection in mice, but the additional deletion of the acid phosphatase gene (ΔsapM) enabled the host phagosome to mature (Saikolappan et al., 2012). A considered approach for MAP vaccine candidate development that target orthologs in M. tuberculosis that have shown promise in vaccine efficacy trials or are known to be highly immunogenic targets seems reasonable, but has its limitations. For instance, there is no corresponding ortholog to sapM in MAP (Li et al., 2005), and hence, construction of a double knockout involving that locus would not be possible.

We note that the MAP vaccines evaluated herein represented the most comprehensive screen of MAP attenuated mutants to date, and included vaccine candidates with prior evidence of attenuation with in vitro or in vivo model systems that were enrolled on a volunteer basis (i.e., investigators were broadly invited and those that volunteered to participate and submit candidate vaccines were enrolled). This appears a reasonable strategy to adopt since it provides a facile and transparent mechanism for benchmarking candidate vaccines to each other to help identify the most promising candidates. However, not all described or available candidate strains of MAP were enrolled in this trial. Notably, the recently described ΔleuD, which provides protection in a goat model of infection (Faisal et al., 2013a) was not submitted for evaluation in the current trials, and neither were a set of invasion mutants (Alonso-Hearn et al., 2008) which should be included in future investigations.

Our studies have highlighted several major lacuna in our understanding of in vitro correlates of vaccine immunity and strain attenuation in MAP. For instance, how long should a strain of MAP take to be killed by phagocytes to be considered attenuated? And what are the physiologic (e.g., doubling time) or immunological (e.g., cytokine profiles elicited) correlates of protective immunity for MAP that should be considered prior to advancing a vaccine candidate? These and other similar questions regarding in vitro correlates of protection remain and need to be the focus of future investigations.

Second, and perhaps more importantly, the rationale for the use of the macrophage-to-mouse-to-native host pipeline for MAP vaccine development needs to be reconsidered. The survival of MAP strains in cultured macrophages has been previously used as a criterion for the selection of candidate vaccines for further testing in animal models. For instance, transposon mutants of MAP that were attenuated in macrophages were also shown to be attenuated in mice (Scandurra et al., 2009). While this strategy has been applied for selection of vaccine candidates for numerous other pathogens, particularly those that impact humans, the lack of validated or well rationalized in vitro correlates of protection does not enable an accurate assessment of either attenuation or immunogenicity of a given candidate in vivo. This shortcoming is further exacerbated by the lack of context for the host immune system and the natural milieu, and the fact that only short-term attenuation can be measured in macrophages, even though these are long-lived cells in vivo, can remain viable for only 48-h or so in laboratory culture. And can only measure degrees of attenuation in a somewhat artificial system that does not predict protection (Lamont et al., 2014).

The results of our investigations raise important questions on whether mice are relevant or appropriate models for assessment of anti-MAP vaccines. It has been long recognized that the mouse is not a natural host for MAP, but rather employed as a model for cost and practical (e.g., availability of immunologic reagents) considerations that may not be of relevance to infection or immunity against MAP in the natural host. By not being a natural host, challenge and vaccine studies are made difficult since it is often not possible to consistently infect mice. Furthermore, culturing of MAP from intestinal tissues (the predominant site of infection in the natural ruminant host) is rarely achieved in mice, and successfully infected animals do not exhibit any of the typical clinical signs (e.g., diarrhea) that are associated with MAP infection in ruminants. In hindsight, this is not altogether surprising since MAP infection is primarily a chronic granulomatous infection of the intestinal tract, and given that recent studies suggest that the mouse is a poor model for inflammatory conditions in humans (Seok et al., 2013), mice are not very likely to represent good models for chromic inflammatory diseases such as JD in cattle either. This is also consistent with recent observations (Scandurra et al., 2010; Park et al., 2011) that suggest attenuation in macrophages and marginal protection in the mouse model may not be a good predictor for the ruminant host. In addition, it was previously noted that a lack of survival in bovine macrophages does not predict survival in goats. For instance, the pknG mutant was attenuated in macrophages and yet showed persistence in goats and the immune response elicited by the mutant did not affect colonization by the MAP challenge (Park et al., 2011).

Thankfully, unlike the situation in humans, Johne's disease investigators have the luxury working with the natural (ruminant) host despite the apparent high costs. We suggest that the ability to quickly and relatively inexpensively use macrophage/mouse models, though tempting, may represent a situation of being “penny wise and pound foolish” and in the absence of clearly articulated and validated correlates of immune protection in these models, may result in perfectly good candidates being discarded during the triage process or, as was observed in the studies described herein, apparently promising targets move forward at considerable expense, only to fail to perform as expected when evaluated in the natural host.

In sum, the results of our studies strongly suggest that future investigations of MAP vaccine candidates are best conducted in the natural host, and considerable efficiencies can be realized by using a coordinated approach and standardized protocols for comparative benchmarking and evaluation of MAP vaccine candidates.