1. Is M. abscessus molecular pathogenesis just TB with a different mycobacterial species?

Mycobacterium tuberculosis is the most successful human pathogen having claimed more lives than any other species across history and remains the number one microbial species responsible for infectious disease deaths globally. The genus level similarities between M. abscessus and M. tuberculosis span genomic similarity demonstrated by both genetic measures such as sequence similarity and essential gene analyses, and pathogenic manifestations such as chronic pulmonary infection and residency within macrophages [1,2]. However, the primary M. tuberculosis virulence factor is the early secretory antigenic target secretion system-1 (ESX-1) while M. abscessus lacks ESX-1 but has two analogous type VII secretion systems ESX-3 and ESX-4 which are necessary for in vivo virulence [3]. Even with relatively few isolates of M. abscessus sequenced, it is clear that the M. abscessus genome is highly diverse with signs of active horizontal gene transfer compared to M. tuberculosis [4].

On the host side, shared susceptibility factors include Mendelian susceptibility to mycobacterial disease (MSMD) genes and autoantibodies impairing IFN-γ signaling, and immunosuppression by inhaled corticosteroids [5–8]. Interestingly, divergent host susceptibility factors also provide evidence of unique molecular pathogenesis. Cystic Fibrosis (CF) and Hyper Immunoglobulin E Syndrome are two examples where increased susceptibility to mycobacterial infection is seen for M. abscessus but not M. tuberculosis [9–11]. While HIV/AIDS is a well-known risk factor for M. tuberculosis infection, there is limited evidence for increased susceptibility to M. abscessus in this population [12], although we must note that the availability of effective anti-retroviral therapy over the last 20+ years may mask any association. These distinctions, along with the observation that M. abscessus infection more commonly occurs in individuals with underlying structural lung diseases such as non-CF bronchiectasis, highlights the divergent host factors influencing susceptibility [13].

2. What do we know about the host–pathogen interface of M. abscessus infection?

Unfortunately, the answer is not much. Clinical specimens are mainly microbiological isolates and host peripheral blood rather than in situ biopsy samples of M. abscessus at the host-pathogen interface. The paucity of M. abscessus biopsy and autopsy samples continues to limit our understanding of M. abscessus within-host molecular diversity and adaptation, and host immune responses at the sites of M. abscessus infection.

Previous research has proposed that M. abscessus nodular disease initiates from patients being colonized by a smooth colony morphotype from the environment, which accumulates mutations in the genotoxic host environment from antibiotic and immune sources, causing an irreversible transition to a rough colony morphotype lacking glycopeptidolipids. It is then the exposed surface of rough colony morphotype bacteria that drives excessive host inflammation and tissue destruction that manifests as cavitary disease [13,14]. This chain of events raises the questions of why nodular and cavitary disease manifest in different parts of the lung and if animal models will contribute to an explanation for the spatial distribution of colony morphotypes?

There is strong evidence for the increased experimental virulence and clinical severity of rough morphotype M. abscessus from clinical isolates and isogenic strains [14–18]. However, recent large-scale phenotypic profiling suggests that there is more to M. abscessus virulence than smooth to rough colony morphotype transition with Boeck and colleagues finding no association between colony morphotype and experimental virulence, clinical clearance or lung function decline across 199 diverse clinical isolates [19]. One limitation of this study was the lack of link between morphotype and radiologic diagnosis to allow correlation of experimental data to disease endotype.

Indirect observation of the host-pathogen interface through peripheral blood analysis identifies core immune responses to mycoacterial infection, such as the production of the ubiquitous cytokine interferon gamma. A potentially important differential is higher levels of IL-17 in M. abscessus infection compared to M. tuberculosis. This may reflect the relative lack of immune subversion carried out by M. abscessus to shape a permissive immune response away from a “default” lung surface damage immune response through active subversion by ESX-1 and Hip1 activities, two factors that are absent in M. abscessus [20–22].

3. Acute non-resolving infection models uncover cell biology and predict antibiotic activity

Immunocompromised mouse models allow high levels of M. abscessus growth in the lungs (recently reviewed in [23]). Despite shortcomings in modeling disease pathobiology, these models provide sufficient M. abscessus growth for studying the effect of therapeutics in the context of mammalian pharmacodynamics. The GM-CSF-deficient mouse model is of particular note as it produces strong histopathology and is responsive to chemotherapy.

Systemic infection of zebrafish embryos with M. abscessus by intravenous infection is the simplest vertebrate model of M. abscessus infection. This model pits M. abscessus against the developing innate immune system of the zebrafish embryo and can be titrated, by inoculum and embryo age at the time of infection, to deliver an outcome between overwhelming infection and elimination of the bacteria. This model has utility in examining antibiotic and phage therapeutics [24,25], and the cell biology of host susceptibility to M. abscessus infection [10]. However, this model produces little histopathology for correlation to pulmonary disease.

4. Self-limiting mucosal infection models reflect aspects of CF and nodular infection

Embedding M. abscessus within agar beads and instillation into the lungs of mice causes granulomatous inflammation in the bronchi and small airways [26–28]. The reproduction of non-necrotic granulomatous inflammation and M. abscessus replication holds promise for replicating CF pathologies in combination with CFTR-deficient mice [29]. Importantly, bacterial burden and pulmonary pathology are self-limiting with bacterial clearance around 2–3 months post-implantation and pulmonary recovery soon after resorption of the agarose beads in most animals providing a significant window for experimentation.

The role of a pre-existing inflammatory condition in concert with structural lung damage is clear from the predilection of bronchiectasis and COPD patients to M. abscessus infection. The βENaC overexpression mouse has pronounced airway inflammation and is susceptible to mucosal surface growth of M. abscessus without significant penetration of tissue or progression down to alveoli [30]. Like other immunocompetent mice, infection is self-limiting with bacterial clearance seen by 1 month providing a shorter window for experimentation.

A ciliopathy bronchiectasis mouse model has been described utilizing knockout of IFT88 KO to drive pulmonary inflammation prior to M. abscessus infection in agarose beads [31]. This model generates granulomas and maintains bacterial burden beyond lifespan of the beads demonstrating M. abscessus engraftment. Notably, the authors found a possible direct role for cilia function in supporting immunity to M. abscessus, contributing to the understanding of immunopathology in diseases characterized by impaired ciliary function, such as CF and COPD. Use of this model is limited by a relatively complicated scheme for timed deletion of IFT88 by Cre/loxP induction and, like the preceding CFTR and βENaC genetic models allows use of high-resolution mouse immunology tools to study the host-pathogen interface.

Multiple instillations of immunocompetent rats with M. abscessus agarose beads twice facilitates expansion of M. abscessus load after 1 month and granuloma formation without central necrosis [32]. While the distribution of histopathology was not disclosed in the published report, we want to highlight the similarities between these mouse and rat models using agar bead instillation and the nodular bronchiectatic form where M. abscessus attacks the upper part of the bronchial tree leading to the tree-in-bud radiologic pattern of infection. We hypothesize that these sites of infection are likely to consist of cellular (non-necrotic) granulomas formed where M. abscessus has breached the mucus layer and remains in prolonged contact with mucosal tissue. The development of the rat model is particularly exciting in the field as CF rats have been reported to be spontaneously colonized by the CF-associated pathogen Pseudomonas aeruginosa raising the possibility that their respiratory system could be colonized by M. abscessus [33].

5. Do chronic granulomatous infection models recreate fibrocavitary lesion pathology?

An accessible mouse model of dexamethasone immunosuppression both sides of the infection inoculation yields chronic M. abscessus engraftment and a range of titratable granuloma phenotypes following removal of immunosuppression [34]. This is potentially a powerful model of prior corticosteroid usage which places patients at risk of M. abscessus for years after treatment cessation [6]. Paradoxically, the damage caused by unchecked M. abscessus growth during the immunosuppressive phase may fulfill the requirements of a creating a hyperinflammatory environment and structural lung damage that are missing from the βENaC overexpression mouse.

Adult zebrafish are susceptible to M. abscessus infection and produce necrotic granulomas within weeks to months of intraperitoneal infection [35]. The adult zebrafish platform recapitulates the increased pathology associated with rough compared to smooth morphotype M. abscessus. Strikingly, knockdown of genes involved in TNF signaling reduced the growth of rough morphology M. abscessus and associated pathology suggesting a key divergence in immune containment of rough colony morphotype M. abscessus compared to all other mycobacteria, but fits with the clinical observation that tissue damage is immune-mediated.

The fulminant growth of M. abscessus in these models and extensive tissue destruction in the face of a competent immune system lead us to hypothesize that these models will most closely resemble cavitary M. abscessus infection. We believe cavitary M. abscessus infection is preceded by rapid growth of M. abscessus either in cavitary pockets or host tissue which forces the formation of granulomas around existing extracellular M. abscessus.

Concluding remarks: The need for comparative host models

Comparative infection biology allows the models discussed here to be hypothetically linked to range of radiological disease classifications. Unlike the extensive molecular and cellular characterization on human TB granulomas, there are almost no characterizations of M. abscessus-infected human tissue in the literature. Rather than settling on a single experimental platform, we propose that multiple M. abscessus infection models will be required to answer the key outstanding questions in M. abscessus pathogenesis (Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Key challenges.

https://doi.org/10.1371/journal.ppat.1013414.t001

Acknowledgments

We thank Drs Albert Lim and Dorothy Ng for discussion of their clinical practice; Drs Amit Singhal, Pablo Bifani, Kevin Pethe, Ria Sorayah, and Jun Hao Liew for debates on preclinical mycobacterial infection models; Drs Kaiwen Chen and Yongliang Zhang for supervision.