Subunit vaccine protects against a clinical isolate of Mycobacterium avium in wild type and immunocompromised mouse models

The nontuberculous mycobacteria (NTM) Mycobacterium avium is a clinically significant pathogen that can cause a wide range of maladies, including tuberculosis-like pulmonary disease. An immunocompromised host status, either genetically or acutely acquired, presents a large risk for progressive NTM infections. Due to this quietly emerging health threat, we evaluated the ability of a recombinant fusion protein ID91 combined with GLA-SE [glucopyranosyl lipid adjuvant, a toll like receptor 4 agonist formulated in an oil-in-water stable nano-emulsion] to confer protection in both C57BL/6 (wild type) and Beige (immunocompromised) mouse models. We optimized an aerosol challenge model using a clinical NTM isolate: M. avium 2-151 smt, observed bacterial growth kinetics, colony morphology, drug sensitivity and histopathology, characterized the influx of pulmonary immune cells, and confirmed the immunogenicity of ID91 in both mouse models. To determine prophylactic vaccine efficacy against this M. avium isolate, mice were immunized with either ID91 + GLA-SE or bacillus Calmette–Guérin (BCG). Immunocompromised Beige mice displayed a delayed influx of innate and adaptive immune cells resulting in a sustained and increased bacterial burden in the lungs and spleen compared to C57BL/6 mice. Importantly, both ID91 + GLA-SE and BCG vaccines significantly reduced pulmonary bacterial burden in both mouse strains. This work is a proof-of-concept study of subunit vaccine-induced protection against NTM.

tion and subsequent immune responses against NTM, we first established viable mouse models of disease. Since NTM infection and disease burden are more pronounced in immunocompromised populations, our investigation included the Beige mouse strain as the representative immunocompromised mouse model and the C57BL/6 mouse strain as a wild type (WT) resistant mouse strain. Indeed, Beige and C57BL/6 mice have been leveraged for comparative assessments of M. avium infections before, utilizing intranasal infections to evaluate the relative contributions of innate immune cells during early infection 29 . Other publications have used C57BL/6 mice with intratracheal installation 30 as well as intraperitoneal injection 31 to compare virulence and pathogenesis of M. avium isolates. A review of models for NTM therapies can be found here 32 . In this work we aim to expand these models using virulent M. avium isolates and aerosol infection routes, which we and others have used previously for mycobacterial challenge 33,34 , to evaluate vaccine-induced immunity. Interestingly, Beige mice demonstrate enhanced susceptibility to mycobacterial infections due to reduced bactericidal activity of granulocytes, a deficiency in NK cells, and a reported minor defect/delay in humoral responses [35][36][37] . We utilized the clinical isolate M. avium 2-151 smooth transparent (smt) as a challenge strain (generously provided by D. Ordway, Colorado State University) due to its enhanced virulence compared to the commonly used M. avium Chester laboratory-adapted strain, and its ability to establish a persistent pulmonary infection in the mouse model. M. avium 2-151 smt infection better represents human disease conditions from NTM infection and provides resolution for deciphering vaccine-induced protection. Indeed, while M. avium Chester bacterial burden significantly wanes over time following intranasal challenge, from the starting mean infectious challenge dose of 3.80 Log 10 colony forming units (CFU) down to a mean of 2.52 Log 10 CFU six weeks post infection (Fig. 1a), M. avium 2-151 smt establishes a sustained infection after intranasal challenge, with an average lung bacterial burden of 5.87 Log 10 CFU four weeks post challenge, up from the challenge dose (Fig. 1a). This sustained infection of clinical MAC isolates is in agreement with published studies 33,38,39 across several models as well as in our own data, whereby  Figure S1). Next, we optimized previous models using a virulent M. avium isolate and empirically developed the low dose aerosol (LDA) infection parameters for M. avium 2-151 smt (Fig. 1b), making it better aligned with the human aerosol route of infection.
In vivo growth kinetics and immune responses in C57BL/6 and Beige mice. We next evaluated the in vivo growth kinetics of M. avium 2-151 smt in both the C57BL/6 and Beige models to determine the host-induced immune response to infection, histopathology as well as potential bacterial phenotypic differences between the WT and immunocompromised hosts. Lungs and spleens were harvested and homogenates were plated 24 h, 1, 2, 4, 6 and 8 weeks post infection to determine bacterial burden and colony phenotypes. Not surprisingly, despite receiving the same challenge dose, Beige mice were less able to control the infection with the clinical M. avium isolate, resulting in significantly higher bacterial counts in the lung and spleen over time, from one to eight weeks post challenge (Fig. 2a,b). This accelerated increase in bacterial burden in Beige mice corresponds with a two-week delay in neutrophil and mononuclear phagocyte pulmonary influx (Fig. 2c,d) and significantly decreased percentage of pulmonary lymphocytes (Fig. 2e, Supplemental Figure S2) 6-8 weeks post infection when compared to C57BL/6 mice. Indeed, previous reports have demonstrated that a decreased neutrophil recruitment in Beige mice is in part responsible for enhanced susceptibility to intranasal challenge with M. avium when compared to C57BL/6 mice 40,41 . Conversely, NK cell depletion in either mouse model has historically not influenced susceptibility to M. avium infection 40 . We confirmed in our model of M. avium 2-151 smt aerosol challenge that the kinetics and magnitude of pulmonary NK cells was not significantly different between Beige and C57BL/6 mice out to 8 weeks post infection (Fig. 2f). In addition to flow cytometric analysis of the kinetic influx of immune cells to the lung we also evaluated pulmonary immunopathology over the course of infection using Visiopharm software and a novel, custommade algorithm using decision forest training and classification (at Colorado State University) to detect areas of inflammation, or lesions, on hematoxylin and eosin (H&E) stained slides from accessory lung lobes (Fig. 3a,b). The high-throughput work flow presented here alleviates the heavy time burden that was previously required 34 and incorporates the entire accessory lobe in the unbiased analysis for a more progressive and reproducible scoring system. Interestingly, Beige mice are afflicted with a significantly greater proportion of lung identified as an immune lesion area as compared to C57BL/6 mice, 4-8 weeks post infection (Fig. 3b). Indeed, accessory lobes from C57BL/6 mice have less than 10% of the area identified as lesion at 8 weeks post infection, whereas Beige mice reach 60% lesion area by 6 weeks post infection (Fig. 3b).
In vitro growth kinetics and compound sensitivities of rough versus smooth colony morphotypes. Upon counting CFU, colony morphotypes were categorized as smooth or rough [loss or modification of glycopeptidolipid (GPL)] in appearance from ex vivo homogenate plating. While the starting infectious material was uniformly smooth transparent (smt), in both C57BL/6 and Beige mice we observed a stochastic emergence of a minor rough morphotype population (Fig. 4a-c). Interestingly, there are conflicting reports of the impact of colony morphotype on in vivo virulence. Within MAC and M. avium some studies suggest that rough variants (lacking GPL) induce a stronger host inflammatory response 42 whereas other studies observed no differences in induced host cytokine production or virulence between smooth and rough phenotypes of clinical M. avium isolates 43,44 . This includes observations that M. avium rough and smooth colony morphotypes do not differentially induce proinflammatory TNF-α from human PBMCs after 20 h of in vitro infection 43 . The impact of GPL is further complicated by NTM strain variation, as studies have demonstrated divergent mouse TNF-α responses induced by two smooth variants of M. avium 45 .
Infection stock of uniform smooth morphology was sub-cultured in vitro for 5 weeks where rapid emergence of rough phenotype was not observed. Indeed, rough colonies were not noted until 3 weeks of passaging  Table S1) in vitro whereas rough colonies were observed as early as 1 week directly ex vivo (Fig. 4). Three rough and three smooth colonies were isolated from ex vivo lung homogenate plating and were subsequently cultured in vitro. We observed equivalent growth kinetics between smooth and rough colonies. Furthermore, when rough and smooth colonies were co-cultured in vitro the total make-up of the population retained the relative ratios of morphotypes suggesting little interference or dominance between strains in vitro (Supplementary Table S2).
There are notable differences in drug sensitivity between rough and smooth morphotypes documented for M. abscessus strains 26,46 , but there is little information to date for clinical M. avium isolates. Here we determined in vitro minimum inhibitory concentrations (MIC) by Optical Density (OD) following the addition of frontline antibiotics used to clinically treat MAC infections (Supplementary Table S3). We observed no significant difference in MIC between smooth and rough subcultures of M. avium 2-151 in vitro (based on a threshold of fourfold change or higher) to front line antibiotics (Supplementary Table S3 (Fig. 5a). These antigens were selected as candidates for Mtb vaccines based on their lack of human sequence homology and their ability to induce an ex vivo IFN-γ response in PPD + human peripheral blood mononuclear cells (PBMC), suggesting they are immunogenic in humans 47,48 . Despite being derived from Mtb, these antigens share  www.nature.com/scientificreports/ significant sequence homology amongst different mycobacterial species, including M. avium (Table 1). Indeed, when we executed a theoretical epitope prediction, using the IEDB database, we observed significant overlap of predicted MCHII binding peptides in the mouse H2-iAB allelic background (Supplemental Table S4-S7) as well as predicted immunogenic CD4 + T cell epitopes across human HLAs (Supplemental Table S8). Not surprisingly, the greatest overlap in these predicted epitopes were between Rv3619 and Rv1886 antigen homologs given they have the highest sequence identity between Mtb and M. avium. Therefore we leveraged this ID91-based vaccine as a proof-of-concept subunit candidate against M. avium 2-151 smt challenge in the preclinical model. Interestingly, T cell responses directed to Ag85B, an antigen in the vaccine candidate, have demonstrated some protective efficacy against M. avium in a P25 transgenic mouse model (the majority of TCR specific for Ag85B) akin to that observed in C57BL/6 wild type mice 49 . Moreover, we verified that ID91 antigens formulated in GLA-SE were immunogenic in both Beige and C57BL/6 mice. Vaccine candidate ID91 + GLA-SE was given twice intramuscularly two weeks apart and splenocytes were isolated one week after the boost immunization. Robust CD4 + T helper 1 responses (T H 1), including production of IFN-γ, IL-2 and TNF cytokines, were elicited from ex vivo ID91 restimulated splenocytes from ID91 + GLA-SE immunized mice but not from mice that received saline only (Fig. 5b, Supplemental Figure S3).
We next evaluated whether each mouse strain was equally sensitive and receptive to innate immune stimulation by the synthetic TLR4 agonist, GLA, by evaluating surface expression of activation and costimulatory markers on mouse bone marrow-derived dendritic cells (BMDC). As we have observed previously 50 , GLA induced expression of CD40 and CD86 activation markers on CD11c + cells in a dose-dependent manner (Fig. 6a,b). The percentage of CD40 and CD86 expression induced by LPS and GLA-aqueous formulation-(AF) was similar between C57BL/6 and Beige CD11c + BMDC (Fig. 6a,b). Although we detected no global deficiency in BMDC  www.nature.com/scientificreports/ GLA-induced activation between mouse strains we did observe dramatically lower expression of MHCII (Fig. 6c) in Beige mice across the concentrations of innate immune modulators tested compared to similar doses evaluated in C57BL/6 mice.

ID91 + GLA-SE and BCG immunizations induce T H 1 and humoral responses in both WT and immunocompromised mouse models. Bacillus Calmette-Guérin (BCG), the only licensed Mtb vac-
cine, has demonstrated cross-reactive immune responses against M. avium in both human PBMC and mouse lymphocytes 51 , and is protective against challenge with M. avium in mice [52][53][54] . Furthermore, retrospective epidemiological evidence from Finland suggests that cessation of universal BCG administration in infants across the nation has coincided with a significant increase of NTM infection and disease in children 55 . Route of administration of BCG is an active and promising area of research, including the exciting publication by P. A. Darrah and colleagues whereby intravenous (i.v.) and intradermal (i.d.) BCG immunizations in nonhuman primates were evaluated for efficacy against Mtb challenge 56 . In our mouse preclinical model, as well as those of our colleagues, BCG given i.d. has proven to be a moderately effective vaccine against Mtb 34,57-60 . Therefore, we utilized an i.d. administration of BCG as a positive control for our vaccine efficacy and immunogenicity evaluations. Six weeks post infection, vaccine immunogenicity (humoral and cellular) was evaluated in cohorts of mice that were immunized with adjuvant alone (GLA-SE), ID91 + GLA-SE or BCG and challenged with a LDA of M. avium 2-151 smt. C57BL/6 and Beige mice immunized with ID91 + GLA-SE demonstrated robust ID91 antigenspecific ( Fig. 7a-d) and Ag85B-specific (Fig. 7e-h) total IgG, IgG2c, IgG1, and IgA antibody responses, whereas GLA-SE or BCG immunized mice demonstrated much weaker ID91 and Ag85B antigen-specific humoral www.nature.com/scientificreports/ responses. Interestingly, C57BL/6 and Beige mice seemed to induce a similar degree of each antibody subtype within each immunization cohort. Lymphocytes were isolated from immunized cohorts six weeks post challenge from individual lung homogenates and restimulated ex vivo with ID91 antigens (Fig. 8, Supplemental Figure S3). These cells were then stained and evaluated by flow cytometry for CD4 + T H 1 responses. The magnitude and composition of antigen specific polyfunctional (expressing IFN-γ, IL-2, TNF or a combination of these cytokines) CD4 + T H 1 T cells was significantly higher in ID91 + GLA-SE immunized C57BL/6 mice than BCG or GLA-SE immunized C57BL/6 mice (Fig. 8). Interestingly this pattern was much less robust and not significant for Beige mouse cohorts.   www.nature.com/scientificreports/ phylactic protection as observed by a reduction of bacterial burden (consisting of both rough and smooth morphotypes, albeit a predominately higher proportion of smooth colonies), in the lungs and spleens of C57BL/6 and Beige mice (Fig. 9a,b). Excitingly, BCG and ID91 + GLA-SE reduced lung bacterial burden in the lungs of Beige mice by 1.76 and 0.77 Log 10 CFU, respectively, compared to adjuvant alone (Fig. 9a). Interestingly, there was no significant difference in pulmonary lesion area at 6-weeks post infection for any treatment group of C57BL/6 mice, with relatively low (< 10%) lesion scores (Fig. 9c,d). Conversely, Beige mice demonstrated significantly higher lesion area compared to C57BL/6 across all treatment arms. While no treatment was significant, prophylactic BCG immunization did trend with lower immunopathology in Beige mice (Fig. 9c,d).
A pattern of vaccine-induced protection was independently obtained from parallel studies performed by our collaborators, demonstrating significant ID91 + GLA-SE vaccine-induced protection against M. avium 2-151 smt in the C57BL/6 model 6 weeks post challenge (Supplementary Figure S4). These data suggest vaccination, even in an immunocompromised host setting, may be a viable approach for protection from NTM.

Conclusions
In this work, we observed a sustained infection in wild type (C57BL/6) and immunocompromised (Beige) mouse models using a virulent clinical isolate of M. avium. These representative mouse models will be useful for developing candidate vaccines targeting NTM in immunocompromised or immune competent individuals. Beige mice sustain a higher bacterial burden over time than C57BL/6 mice and demonstrate a differential pulmonary immune cell influx post challenge. C57BL/6 mice exhibit earlier and more robust monocyte and neutrophil pulmonary influx 24 h to 2 weeks post challenge, as well as an increase in lymphocytes 6 and 8 weeks post challenge compared to Beige mice, and considerably less immunopathology over time which may be directly related to their divergent ability to control infection early and late. These data represent an expansion on previous reports 40,41 , including the aerosol route of challenge and the specific concurrent quantification of lymphocytes, mononuclear phagocytes, NK cells, and neutrophils in the lungs of infected mice over time. While Beige mice do not represent a model that is defective in proinflammatory signaling axes, which can predispose humans to NTM infection, this mouse strain provides a unique opportunity as a model, where these mice can (1) sustain infection out to at least 6 weeks post challenge with NTM in spite of their slight immunodeficiencies, and can (2) still be further immunologically manipulated (e.g. cell depletion) to extrapolate vaccine-induced mechanisms of control/resistance. Interestingly, we observed a stochastic rough-colony phenotype emergence in vivo as early as 1 week following infection with M. avium 2-151smt, whereas a switch to rough morphology was less prominent in vitro suggesting that in vivo pressure(s) may be critical for this loss/modification to GPL. There was no evidence of interference between morphotypes in vitro or preferential growth kinetics. Furthermore, we did not identify differences in sensitivity to front line antibiotics in vitro between rough and smooth M. avium 2-151 morphologies.
Our proof of concept vaccine candidate ID91 + GLA-SE was readily immunogenic in both mouse models, including enhanced ID91-specific humoral and CD4 + T H 1 cellular responses. This is an important finding given that IFN-γ activates neutrophils and macrophages to phagocytose and/or kill intracellular pathogens, including mycobacteria such as NTM [61][62][63][64] . Furthermore, the positive feedback loop between IFN-γ and IL-12 is critical for the control of mycobacteria 65,66 . Cellular CD4 + T H 1 T cell responses are likely a critical component for mycobacterial infection susceptibility as persons living with HIV and waning CD4 + T cell counts (< 50 cells/µl) can present with disseminated extrapulmonary Mtb or NTM infections 67,68 . TNF, another prominent T H 1 cytokine, has been clinically shown to help control NTM infections, which if inhibited, as with other mycobacterial infections, can result in increased risk of infection 8,[69][70][71] . In cases where a genetic deficiency in the IL-12/IFN-γ signaling axis leads to enhanced risk of NTM infection, other T H 1 cytokines like TNF may be able to fill this void and reduce the overall susceptibility. The complete mechanisms of protection against NTM infection are not yet known 17 . Therefore, we hypothesized that a T H 1 skewing vaccine candidate may afford partial protection in WT and immunocompromized preclinical models.
Indeed, both ID91 + GLA-SE and BCG demonstrate prophylactic protection against M. avium 2-151 smt aerosol challenge. Importantly, BCG immunization resulted in greater protection in both Beige and C57BL/6 mice compared to the proof of concept ID91 + GLA-SE vaccine, emphasizing the need for more work in specific antigen design and optimization for future candidates. In these experiments, pulmonary lesion scores did not align with overall bacterial burden, suggesting the mechanism of protection may well involve immune cell kinetics and be less directly related to maintaining pulmonary architecture. Although BCG was leveraged in this study as a positive control, historically live-attenuated vaccines have been contraindicated for immune compromised individuals, the same population disproportionately affected by NTM disease. BCG is not recommended for individuals living with HIV/AIDS as BCG-induced disseminated disease has been reported 72,73 . Furthermore, while BCG vaccination before NTM exposure induces some cross-reactive immunity 51 , BCG vaccination in the preclinical model after infection or exposure to NTM does not seem to reduce NTM infection or disease 74 . The complex interplay of historical mycobacterial exposures, both repeatedly to environmental NTM and BCG vaccination, is an important consideration for early vaccine design and efficacy testing. We believe that through rational vaccine platform selection and design a vaccine against NTM and specifically M. avium, can be developed as a safe strategy against this opportunistic infection.
Although outside the scope of this proof-of-concept study, it is not clear whether BCG and a subunit vaccine could provide synergistic protection against NTM. Other vaccine platform approaches, including DNA-based vaccines encoding mycobacterial antigen 52,75 , have not yet demonstrated synergistic protection over that of BCG alone in the preclinical model, despite inducing significant antigen-specific IFN-γ + CD4 + T H 1 responses 52,53 . M. avium culture filtrate protein (CFP) has also been explored as a heterologous mixed protein vaccine candidate, akin to Mtb studies 76 , and when formulated with DDA adjuvant was able to afford 1.0 and 0.5 Log 10 CFU www.nature.com/scientificreports/ www.nature.com/scientificreports/ protection in the liver and spleen, respectively, in an intravenous M. avium BALB/c challenge model 66 . The work detailed here further demonstrates that subunit + adjuvant vaccine platforms are a viable approach for novel or complimentary interventions against M. avium.
As a rich collection of species with expanding prevalence and global health burden, research focus on NTM should be prioritized. Vaccine strategies that (1) develop NTM-specific antigen candidates that are immunogenic in humans and (2) devise focused regimens with respect to different disease indications (e.g. Cystic Fibrosis (CF) or anti-TNF-therapy, scenarios with defects in mechanical or physical disruptions to the pulmonary space as well as those with disrupted proinflammatory signaling), have the capacity to significantly impact the burden of NTM disease. Moreover, the successful development of a vaccine against NTM species may help inform vaccines against other mycobacteria, such as Mtb.

Materials and methods
Preclinical animal model. Female C57BL/6 mice and Beige mice 4-6 weeks of age were purchased from Charles River Laboratory and Jackson Laboratory (Sacramento, CA, USA), respectively. All methods were carried out in accordance with relevant guidelines and regulations with respect to animal welfare. Mice were housed at the Infectious Disease Research Institute (IDRI) biosafety level 3 animal facility under pathogen-free conditions and were handled in accordance with experimental protocols that were approved by IDRIs Institutional Animal Care and Use Committee (IACUC). Mice were infected either by the intranasal (i.n.) route or aerosol challenge with 10 4 CFU of Mycobacterium avium Chester or Mycobacterium avium 2-151 smt 77  Bacterial burden/CFU. Bacterial counts were enumerated from 4 to 7 mice per group, 24 h to 8 weeks post infection. With methodology we have previously reported 34 , mice were euthanized with CO 2, and lung and spleen tissue from infected animals was isolated and homogenized in 5 mL of either RPMI + FBS (lung) or PBS + Tween-80 (Sigma-Aldrich, St. Louis, Missouri, USA) CFU buffer (spleen) using an Omni tissue homogenizer (Omni International, Kennesaw, GA, USA) or Precellys Tissue Homogenizer (Bertin Corp. Rockville, MD, USA). Serial dilutions of homogenate were made in CFU buffer and aliquots were plated on Middlebrook 7H10 agar plates and subsequently incubated at 37 °C and 5% CO 2 for 2-3 weeks before colonies were counted. Bacterial burden, as CFU/mL, was calculated per organ and is presented here as Log 10 values. Colony morphotype (rough versus smooth appearance) was also denoted at counting. Reduction in the bacterial burden was calculated as the difference in mean Log10 values between groups assessed.
Histopathology and image analysis. Lung accessory lobes were collected at different time points after M. avium 2-151 smt challenge and were immediately perfused and subsequently stored in 10% normal buffered formalin. Fixed lungs were embedded in paraffin, cut and slides were generated with hematoxylin and eosin (H&E) staining at the Benaroya Research Institute Histology Core facility (Seattle, WA, USA), with methodology previously reported 34 . Stained slides were blinded and analyzed by a veterinary pathologist (Dr. Brendan Podell, Colorado State University). H&E stained sections were scanned at 20X magnification using an Olympus VS120 microscope, Hamamatsu ORCA-R2 camera, and Olympus VS-ASW 2.9 software. Visiopharm software was used for image analysis. For each tissue section, a region of interest (ROI) was generated at a low magnification with a custom tissue detecting algorithm using decision forest training and classification to differentiate tissue versus background based on color and area. Lesions were identified within tissue ROI's at a high magnification with an additional custom-made algorithm using decision forest training and classification based on staining intensity, color normalization and deconvolution, area, and morphological features. Percent lesion calculations were integrated into the same algorithm and calculated from tissue area and lesion area as designated by the ROI and lesions detected. Lesion identification and quantification were then reviewed and edited by a pathologist as needed 79 . In vitro growth kinetics. M. avium 2-151 smooth and rough colony variants were isolated from ex vivo lung homogenate plating. These isolates were grown in Middlebrook 7H9 medium supplemented with 10% v/v OADC (oleic acid, albumin, dextrose, catalase) and 0.05% w/v Tween 80 (7H9-Tw-OADC) under aerobic conditions.   Antibody ELISAs. Serum was collected from mice immediately after euthanasia, six weeks post challenge.