Activation of host nucleic acid sensors by Mycobacterium: Good for us or good for them?

Abstract

Although the importance of deoxyribonucleic acid (DNA) and ribonucleic acid (RNA) sensors in controlling viral infection is well established, their role in promoting an effective immune response to pathogens other than viruses is less clear. This is particularly true for infections with mycobacteria, as studies point to both protective and detrimental roles for activation of nucleic acid sensors in controlling a mycobacterial infection. Some of the contradiction likely stems from the use of different model systems and different mycobacterial species/strains as well as from which nucleic acid sensors were studied and what downstream effectors were evaluated. In this review we will describe the different nucleic acid sensors that have been studied in the context of mycobacterial infections, and how the different studies compare. We conclude with a section on how nucleic acid sensor agonists have been used therapeutically and what further information is needed to enhance their potential as therapeutic agents.

Introduction

The genus Mycobacterium

The genus Mycobacterium consists of over 190 species, most of which are saprophytic and live in various environmental reservoirs. However, some of these environmental species, collectively referred to as nontuberculous mycobacteria (NTM), can cause disease in humans as opportunistic pathogens [1]. Other species of mycobacteria are obligate pathogens, including Mycobacterium tuberculosis (Mtb) and several closely related species, which primarily cause pulmonary infections, and Mycobacterium leprae, which causes disease of the skin, eyes, nose, and muscles [2] [3] [4]. The obligate pathogens Mtb and M. leprae are transmitted from person-to-person via aerosolized droplets [2] [3] [4].

It is estimated that over 2 billion individuals have been infected with Mtb and that 5 to 10% of those infected will develop active tuberculosis (TB) sometime during their lifetime, resulting in approximately 10 million new cases of active TB infections annually. Although there are effective antibiotics to treat drug-sensitive TB, the disease still causes approximately 1.5 million deaths annually, making it the second leading cause of death by an infectious organism behind Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) (https://www.who.int/health-topics/tuberculosis#tab=tab_1). The only FDA approved vaccine for TB is M. bovis BCG, which has been shown to be effective against some forms of tuberculosis like TB meningitis but limited effectiveness against pulmonary TB [5]. Moreover, approximately 1/3 of TB cases go undiagnosed and this is even higher among children [6] [7]. Further complicating TB control is the increasing levels of drug-resistant TB. However, new classes of antibiotics have been recently approved providing some hope that we will soon have better outcomes for treating these drug-resistant TB patients [8].

Though most NTMs do not cause disease in humans, some, especially Mycobacterium avium, Mycobacterium intracellulare, Mycobacterium abscessus, and Mycobacterium kansasii, are known human pathogens. Indeed, the most common type of mycobacterial infections seen clinically in the United States, much of Europe, and Japan are pulmonary infections caused by NTMs, and evidence suggests that the incidence of these infections is on the rise globally [9] [10]. Disease-causing NTMs are transmitted from environmental sources including water, soil, and infected animals. There is little evidence that NTMs are transmitted person-to-person [11]. NTMs are opportunistic pathogens, causing disease in patients with some underlying lung disease, such as cystic fibrosis or acute respiratory distress syndrome (ARDS), those who are immune compromised due to human immunodeficiency virus (HIV) infection or drug-induced immune suppression, or elderly people with certain predisposing characteristics [11] [10]. M. avium, the most common cause of NTM infection in the US, is difficult to diagnose and treat. Once diagnosed, treatment with multiple antibiotics can extend for 18–24 months, with clinical guidelines dictating a continuous treatment for 12 months after the infecting bacteria are no longer detected in patient sputum. Even with this extensive antibiotic treatment course, recurrence of infection is common [9]. While pulmonary disease is the most common manifestation of NTM opportunistic disease, other environmentally acquired infections, such as those with Mycobacterium ulcerans or Mycobacterium marinum, affect the skin and soft tissue in immunocompetent individuals and are transmitted from contaminated water sources [12] [13].

Mycobacterial pathogenesis

The majority of disease-causing mycobacteria are intracellular pathogens, with macrophages serving as the primary host cells. However, dendritic cells, neutrophils, and perhaps non-immune cells, such as epithelial cells, may also serve as host cells during different stages of infection [14]. The involvement of epithelial cells in the pathogenesis of TB is supported by studies on M cells [15] and Type II alveolar epithelial cells [16] [17] [18]. Paradoxically, macrophages are both the primary niche for bacterial survival and replication and the major mediator of host protection through T cell-mediated activation of uninfected cells. During a pulmonary infection, mycobacteria are phagocytosed by alveolar macrophages in the lung and subsequently colonize the underlying epithelial layer. This triggers recruitment of mononuclear cells from neighboring blood vessels, and these become host cells for multiplying bacteria. Subsequent recruitment of activated T cells and initiation of granuloma formation are necessary for controlling a pulmonary mycobacterial infection. Interestingly, some individuals seem refectory to disease and are able to control an infection without granuloma formation suggesting that in some exposed individuals the innate response is sufficient to eliminate the bacteria [19]. Similar processes are at play during non-pulmonary infections, such as infections of the skin or other soft tissue, where initial infection of macrophages is followed by granuloma formation and recruitment and activation of T cells. Chemokines and cytokines which drive this immune response are induced upon exposure of mycobacterial pathogen associated molecular patterns (PAMPs) to their corresponding pattern recognition receptors (PRRs).

Nucleic acids as mycobacterial PAMPs

Introduction to nucleic acid sensors

All pathogens use RNA or DNA as their genetic code, so it is not surprising that host organisms have evolved mechanisms to identify and respond to foreign nucleic acids. The nucleic acid sensors can be divided into two major categories, based on cellular localization: endosomal sensors and cytoplasmic sensors. Within both categories are sensors that bind to RNA or DNA, allowing detection of either or both nucleic acid types depending on the pathogen. Prime examples of endosomal nucleic acid sensors are four members of the toll-like receptor (TLR) family, which are preferentially expressed in professional phagocytes such as monocytes and macrophages, though they have also been detected in a variety of immune and other cell types. This allows these cells to detect nucleic acids of pathogens that have entered the cell via phagocytosis or endocytosis. These endosomal sensors include TLR3, which recognizes double-stranded RNA (dsRNA), TLR7, which recognizes single-stranded RNA (ssRNA), TLR8, which recognizes ssRNA, and TLR9, which recognizes unmethylated single-stranded DNA (ssDNA) containing the cytosine-phosphate-guanine (CpG) motif. TLR13 is a fifth member, not expressed in humans but expressed in mice, which recognizes the catalytic center of bacterial 23S ribosomal RNA.

While expression of nucleic acid sensing TLRs is somewhat specialized to phagocytic immune cells, expression of cytosolic nucleic acid sensors enables detection of foreign nucleic acids in a broad range of host cell types. Expression of cytosolic sensors is particularly important in the recognition of viral nucleic acids, since the viral genome must be present in the cytosol at some stage of a viral infection. Activation of these cytoplasmic sensors is essential for controlling many viral infections, and their expression in most cells reflects the fact that viral infections can have a broad host cell range. The cytosolic sensors can be classified into four categories: retinoid acid-inducible gene I (RIG-I)-like receptors (RLRs), protein kinase R (PKR), cyclic GMP-AMP synthase (cGAS), and absent in melanoma 2 (AIM2)-like proteins (ALRs).

Although most studies on nucleic acid sensors have been in the context of viral infections, there is growing evidence of their importance in the context of other infections, including those with mycobacteria. The following sections review current understanding of the role of endosomal and cytoplasmic sensors in the context of mycobacterial infections (Table 1), concluding with a discussion of how nucleic acid sensors could be targeted to improve vaccine efficacy or to enhance drug treatment during an active infection.

Table 1:

List of nucleic acid sensor receptors

Sensor		Location	Agonist(s)	References
Toll-like Receptors (TLRs)	TLR3	endosome	dsRNA; poly(I:C)	[100] [12] [13] [14]
	TLR7	endosome	ssRNA (murine); imidazoquinoline amine derivatives; guanine analogs	[15] [16] [17] [18] [19] [29]
	TLR8	endosome	ssRNA (human); TL8-506	[15] [16] [17] [18] [21] [22] [23] [25] [26] [27] [28] [29]
	TLR9	endosome	CpG DNA	[28] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42]
RIG-I-like Receptors (RLRs)	RIG-I	cytosol	short dsRNA	[43] [44] [45] [46] [48] [49] [51] [52] [53]
	MDA5	cytosol	long dsRNA	[43] [44] [45] [46] [48]
	LGP2	cytosol	dsRNA; other RNA sensors	[47] [48] [58]
PKR		cytosol	dsRNA	[43] [47] [59]
cGAS		cytosol	short and long DNA	[60] [62] [63] [64] [65] [66] [67] [68] [69] [70] [71]
AIM-2 like Receptors (ALRs)	AIM2	cytosol	DNA	[61] [72] [73] [74] [75] [78] [79] [80] [81] [82]
	IFI16/IFI205	cytosol	DNA	[61] [72] [73] [74] [75] [77]
	DAI/ZBP1	cytosol	DNA	[61] [72] [73] [74] [77]

Endosomal nucleic acid sensors

TLR3

TLR3 signaling is induced by binding to dsRNA. A unique feature of TLR3 is its utilization of the toll/interleukin-1 receptor (TIR) domain-containing adaptor protein inducing interferon-β (TRIF)-dependent pathway. This pathway leads to activation of tumor necrosis factor (TNF) receptor-associated factor 3 (TRAF3), which in turn phosphorylates and activates TANK-binding kinase 1 (TBK1). TBK1 then phosphorylates the transcription factor interferon-regulatory factor 3 (IRF3) causing IRF3 translocation into the nucleus, where it binds the interferon (IFN)-β promoter to activate IFN-β expression (Fig. 1). TLR3 binding to dsRNA also causes activation of the transcription factor nuclear factor (NF)-kB, leading to production of both pro- and anti-inflammatory cytokines [20].

Figure 1:

Depiction of the various signaling pathways initiated upon engagement of nucleic acid sensing TLRs.

Little is known about the role of TLR3 in response to mycobacterial infections. One study (Bai et al.) examined TLR3 knock-out (KO) mice and mouse peritoneal macrophages infected with Mycobacterium bovis bacillus Calmette-Guerin (BCG), an attenuated strain of Mycobacterium bovis [21]. They observed that BCG RNA can activate TLR3 in wild-type (WT) macrophages and mice to induce interleukin (IL)-10 production through a phoaphatidylinositol 3-kinase (PI3K)/ protein kinase B (AKT) signaling pathway. In TLR3 KO macrophages and mice, IL-10 was significantly reduced while IL-12p40 levels were elevated relative to WT controls. Mice lacking TLR3 showed higher numbers of splenic CD4⁺ T helper type 1 (T_H1) cells producing IFN-γ compared to WT infected mice. Taken together, these findings [21] suggest that TLR3 activation has a negative effect on control of a BCG infection. However, another study found that TLR3 engagement upon addition of poly(I:C), a synthetic analog of dsRNA, stimulated autophagy in BCG infected RAW264.7 murine macrophages and reduced mycobacterial survival [22]. Since this study relied on a synthetic dsRNA analog, it is unknown whether TLR3-mediated autophagy is activated in response to detection of natural mycobacterial components. For both studies the use of BCG makes it unclear whether similar findings would be observed in the context of infections with pathogenic mycobacteria. Further, it is not known if type I IFN production, as induced by TLR3 in the Bai et al. study, is required for increased IL-10 expression, as previous studies have shown a positive link between IFN-β and IL-10 production [23].

TLR7/TLR8

TLR7 and TLR8 mainly recognize uridine and GC-rich ssRNA during viral infections, but expression of these sensors has also been shown to be upregulated during Mtb infection [24] [25] [26]. While more abundant in dendritic cells (DCs), TLR7 and TLR8 are also expressed in macrophages. TLR8 is the primary ssRNA sensor in human DCs while TLR7 drives the response to ssRNA in murine DCs [27] These TLRs utilize a MyD88-dependent pathway to activate NF-κΒ, interferon regulatory factor (IRF)5, and IRF7 through IL-1 receptor-associated kinase (IRAK)2/IRAK4 and TRAF3/TRAF6 to promote expression of type I IFNs and various pro-inflammatory cytokines [27] (Fig. 1). TLR7 activation has also been linked to the induction of autophagy and major histocompatibility complex (MHC) II expression [28].

In Bao et al., TLR7 expression was upregulated in RAW264.7 cells during an Mtb infection. When treated with ssRNA to activate TLR7, infected cells showed lower bacterial burden relative to untreated infected cells [24]. This was due to TLR7 mediated activation of the autophagy pathway, which resulted in the formation of autophagosomes and killing of intracellular Mtb. In contrast, downregulation of TLR7 prior to the Mtb infection resulted in reduced autophagosome formation. Taken together, these findings suggest that TLR7 is activated during a natural infection and promotes a protective response [24].

Along with recognition of ssRNA, TLR7 also recognizes imidazoquinoline amine derivatives and guanine analogs, compounds that can be used as therapeutic agonists. When the imidazoquinoline compound imiquimod (IMQ) was used to stimulate TLR7 activation in macrophages, decreased macrophage viability was observed in a dose-dependent manner (4). IMQ triggered autophagic cell death by inducing high levels of reactive oxygen species (ROS). Studies (Lee et al.) revealed that IMQ promoted mitophagy, a form of selective autophagy targeting mitochondria, which is also known to contribute to elimination of intracellular pathogens. IMQ-mediated TLR7 activation resulted in increased nitric oxide. (NO) production, autophagosome formation, and decreased bacterial burden in Mtb infected macrophages [28]. Since IMQ is already a drug used to treat human papillomavirus (HPV), herpes simplex virus (HSV), and cancer, there may be use for IMQ as a supportive therapy for treatment of TB [29]. However, in vivo infection studies are needed to determine whether IMQ is beneficial to the control of Mtb infection.

The involvement of TLR8 in the immune response against Mtb infection has been difficult to access from mouse models because homology between mouse and human TLR8 is not sufficient to do mechanistic studies. However, transfer RNA (tRNA) from Mtb was found to induce production of IL-18 in human peripheral blood mononuclear cells (PBMCs) through a TLR8-mediated pathway. This TLR8-dependent activation increased Th1 differentiation in the PBMCs, leading to elevated production of IFNγ, IL-12p70, and Type I IFNs in the tRNA treated cells [30]. To further assess TLR8 in the context of Mtb infection, Tang et al. expressed human TLR8 in a C57Bl/6 background and found these mice more susceptible to Mtb infection. In contrast, mice vaccinated with a formulation containing the early secreted antigenic target 6 (ESAT-6) antigen and the TLR8 agonist TL8–506, a benzodiazepine compound, displayed protection against Mtb infection, suggesting that TLR8 agonists may be useful adjuvants in a TB vaccine [31] [32]. This adjuvant activity was mediated mainly by type I IFN signaling. These results complement BCG studies that showed activation of the RNA sensor pathway and production of Type I IFN enhances the vaccine effectiveness of BCG [33]. The ability of RNA sensor activation to enhance vaccine effectiveness is discussed in more detail below.

Although functional studies with TLR8 are limited, there are a number of genomic studies which have suggested a role for TLR8 in TB susceptibility. Most studies, including those involving Russian, Indonesian, and Moldavian males [34], Turkish children, and Indian adults [35] found that the single nucleotide polymorphism (SNP) A1G in the TLR8 gene was linked to protection against Mtb infection. This A to G SNP results in a N-terminal truncation of 3 amino acids in the TLR8 protein. In contrast, linkage studies from Pakistan [36] and among Han Chinese [37] found the G-allele of the A1G TLR8 SNP to be linked to increased susceptibility to TB. Interestingly, in a Han Chinese cohort, another SNP at position −129 in the promoter region (rs3764879: C > G) was in complete linkage disequilibrium with TLR8 A1G. This allele, which confers enhanced promoter activity, showed an elevated response to RNA in a human embryonic kidney (HEK)-cell-model, as well as inactivation of Mtb in PBMCs, suggesting that a stronger activation of the TLR8 pathway may protect against TB.

The importance of endosomal RNA sensing is not limited to Mtb infections, as knockdown of TLR7 and TLR8 in primary human macrophages also resulted in decreased production of TNF-α, IL-6, and IL-10 during an M. avium infection [38]. This activation of TLR7 and TLR8 appears to occur within a LAMP1-positive compartment, indicating that some M. avium must be transported to a late phagosome for activation of the MyD88 dependent pathway to occur [38].

TLR9

Like TLR7 and TLR8, TLR9 activation stimulates the myeloid differention primary response 88 (MyD88) pathway. TLR9 is activated in response to infection after it recognizes unmethylated CpG DNA motifs found commonly in bacteria and viruses but rarely in mammalian cells. The use of TLR9 agonists as adjuvants in vaccines have been shown to increase levels of IFN-γ, TNF, and IL-2, and they have been components of multiple candidate vaccines against Mtb [39] [40]. TLR9 plays a role in mycobacterial infection control, and some studies indicate that TLR9 genetic variants have increased or decreased susceptibility to pulmonary TB [41].

Early studies using MyD88 and TLR KO mice suggest that MyD88 is essential for controlling an aerosol Mtb infection but that loss of TLR2, TLR4, or TLR9, alone or in combination, did not diminish bacterial control [42] [43]. Triple knockout (TLR2/4/9) mice infected with Mtb also expressed similar levels of pro-inflammatory cytokines, including IFN-ɣ, as Mtb infected WT mice. In contrast, a separate study suggested that the loss of TLR2 and TLR9 results in a marked decrease in the ability of mice to control an Mtb infection. TLR2/9 KO mice were highly susceptible to Mtb infection, dying around 80 days earlier than Mtb infected WT mice [44]. The TLR2/9 KO mice also showed decreased production of pro-inflammatory cytokines and altered lung pathology [44]. The reason for the difference between these studies is unclear but may be due to differences in animal housing conditions, which may influence mouse sensitivity to infection, and/or to differences in Mtb strains used for infections. Later studies showed that TLR9 engagement by the administration of tuberculin purified protein derivative (PPD)-coupled beads to BCG-immunized mice was necessary for inducing the expression of the Notch ligand Delta-like 4 (dII4) in both lung and bone marrow DCs, which was required for Th17 cell activation. Loss of dll4 resulted in enlarged granuloma formation [45]. In a separate study, activation of TLR9 by M. bovis or TLR9-specific ligands led to the production of a number of chemokines by plasmacytoid DCs. Loss of this TLR9 activation resulted in a significant alteration in monocyte, macrophage, neutrophil, and natural killer (NK)-cell recruitment to the site of infection [46]. TLR9 activation was also required for controlling infections with M. avium strains 2447 and 2151, but it was not required for M. avium induced production of IL-12 and TNF-α. This may be due to a redundant role for other TLRs leading to activation of NF-κB and other transcription factors during M. avium infection. One possible mechanism for TLR9-mediated control of an M. avium infection could be through limited type I IFN production in TLR9 KO mice and a potential protective role for IFN-β in controlling M. avium infection [47]. Interestingly, both TLR7 and TLR9 induction of type I IFN is inhibited by activation of TLR2, and studies found that activation of TLR2 during Mtb infection inhibits TLR9 mediated type I IFN production and type I IFN dependent MHC class I antigen cross priming [48]. This was due to downregulation of IRAK1 expression upon TLR2 activation, and, therefore, it is not expected to affect MyD88-independent signaling by TLR3 or the cytosolic RNA sensors.

To evaluate the importance of TLR9 in human disease, a study examined how SNPs in the TLR9 gene affect susceptibility to pulmonary and meningeal TB [49]. Using data from 339 TB cases, two SNPs were identified to be associated with susceptibility to, but not increased mortality from, TB. These SNPs are found outside the TLR9 coding region within the twinfillin-2 (TWF2) gene [49]. TWF2 has been shown to be dispensable in mammals, but these mutations could cause a functional or regulatory change in TLR9 that increases susceptibility to TB. Additional studies identified SNPs in TLR9 that were associated with increased or decreased susceptibility to TB [37] [50] [51]. Since mouse models suggest that engagement of both TLR9 and TLR2 may provide some protection against TB, future work could also address TLR9 and TLR2 SNPs in combination.

Cytosolic nucleic acid sensors

RLR family and protein kinase R

Several cytosolic RNA sensors have been shown to be crucial for detecting Mtb RNA secreted into the cytosol [52] [53]. RIG-I and melanoma differentiation-associated protein 5 (MDA5) are two prominent RLR sensors that detect short and long dsRNA, respectively [54]. These proteins associate closely with Mitochondrial antiviral-signaling protein (MAVS) to initiate a signaling cascade that facilitates phosphorylation of the transcription factors IRF3 and IRF7, which translocate to the nucleus to drive type-I IFN production (Fig. 2) [55]. Laboratory of genetics and physiology 2 (LGP2), another RLR, has a complex role in which it interacts with MDA5 and RIG-I to either promote or inhibit type-I IFNs and IFN precursors [56]. The RLR protein family is defined structurally to have a DEXD/H box RNA helicase domain. These domains function in RNA binding and exhibit ATPase activity, enabling conformational changes to catalyze their reactions [57]. RIG-I and MDA5 contain two caspase activation and recruitment domains (CARD) at their N-termini, which bind to MAVS to activate downstream IRFs [57]. RIG-I and laboratory of genetics and physiology 2 (LGP2) have autoinhibitory capabilities via their repressor domains, which inactivate them when pathogenic cytosolic RNA is absent [57]. More recently, PKR, a serine/threonine, dsRNA-detecting kinase that is encoded by the human eukaryotic translation initiation factor 2 alpha kinase 2 (EIF2AK2) gene, has been implicated in recognizing Mtb RNA during infection [56].

Figure 2:

Depiction of the various signaling pathways initiated upon engagement of the cytoplasmic nucleic acid sensors.

RIG-I and MDA5 have emerged as prominent RNA sensors in the host response to Mtb infection. In Ranjbar et al, human monocyte-derived macrophages infected with live Mtb or treated with Mtb extracts showed elevated transcript levels for RIG-I, MDA5, PKR, and IFN-β [52]. They also found that macrophages treated with nitazoxanide (NTZ) prior to Mtb infection showed increased bacterial burden relative to untreated controls [52]. NTZ is known to enhance RLR activity in human cells, and the authors hypothesized that the limited ability to control Mtb infection in NTZ-treated cells was due to increased RLR activity, although the role of RLRs was not tested directly in this study. In a separate study using THP-1 cells and human monocyte-derived macrophages, it was found that Mtb and BCG infections stimulated RIG-I activation, leading to increased production of IFN-β and IL-1α/β, decreased autophagy, and increased Mtb survival in macrophages [58].

Similar to the finding in human macrophages mentioned above, Mtb infection stimulates IFN-β production in mouse bone marrow-derived macrophages (BMMs) in a RIG-I/MAVS dependent manner [53]. Infection of BMMs isolated from MAVS KO mice resulted in an ~80% reduction in IFN-β production relative to Mtb infection of WT BMMs [53]. In both WT and MAVS KO cells, IFN-β secretion was dependent on Mtb expressing both the auxillary protein secretion system (SecA2) and early secretory antigenic target (ESAT-6) specialized secretion system (ESX-1) secretion systems. The lack of these secretion systems resulted in a loss of detectable mycobacterial RNA in the cytosol of the infected BMMs, suggesting that the presence of mycobacterial RNA causes RIG-I/MAVS activation during infection [53]. This was supported by the ability to pull down mycobacterial RNA using an antibody to RIG-I in an immunoprecipitation assay, demonstrating physical association of RIG-I and the foreign RNA [53]. Interestingly, the activation of the MAVS pathway was dependent on the initial activation of the DNA sensor cyclic guanosine monophosphate (GMP)- adenosine monophosphate (AMP) synthase (cGAS)/ stimulator interferon genes (STING), likely due to the importance of cGAS/STING in driving the initial IFN-β production. This initial burst of IFN-β was necessary for activation of the janus kinase/signal transducers and activators of transcription (JAK/STAT) pathway through its autocrine engagement of the interferon alpha and beta receptor subunit 1 (IFNAR1) receptor, which results in the elevated expression of IRF7 (Fig. 2). IRF7 is then activated via the RIG-I/MAVS pathway, further enhancing IFN-β production [53]. Mtb infected MAVS KO mice showed decreased bacterial load, particularly later in infection, compared to Mtb infected WT mice. In agreement with the in vitro data, IFN-β expression was significantly decreased in Mtb infected MAVS KO mice relative to Mtb infected WT mice [53]. How the loss of MAVS expression, and therefore loss of the cytosolic RNA sensor pathway, results in better control of Mtb infection is unclear, but it may be due to a markedly low production of IFN-β in the MAVS KO mice, since IFN-β is known to play a negative role in the immune response to Mtb infection [23] [59]. In contrast to Mtb, infection of MAVS KO mice with M. avium resulted in a higher bacterial load, and this was associated with increased lung pathology [60]. The explanation for this difference in host response between Mtb and M. avium infection when using MAVS knockout mice is not known. Similar to observations with Mtb, production of IFN-β was also significantly decreased in M. avium infected MAVS KO mice relative to infected WT mice [60], suggesting that IFN-β may play a different role in the host response to an M. avium infection compared to an Mtb infection.

Interestingly, activation of the RIG-I/MAVS pathway is not limited to infected cells, as extracellular vesicles (EVs) released from Mtb infected BMMs can activate this RNA sensor pathway in EV recipient macrophages [61]. EVs, which include exosomes, microvesicles, and apoptotic bodies, are important in intercellular communication and are known carriers of all the different macromolecules including nucleic acids, proteins, lipids, and carbohydrates [62]. Studies have shown that macrophages infected with mycobacteria release EVs carrying mycobacterial components including RNA [63] [64] [65] [66]. Pre-treatment of naive BMMs with EVs released from Mtb-infected BMMs resulted in a population of BMMs that can partially control an Mtb infection [62]. This was due to increased maturation of the Mtb-containing phagosome through a noncanonical pathway involving microtubule-associated protein light chain 3 (LC3), a marker of autophagy and autophagosome formation [62]. The ability of these EVs to enhance the macrophage response to an Mtb infection was dependent on the expression of RIG-I and MAVS by the recipient BMMs, suggesting that activation of the cytosolic RNA sensor pathway was necessary for the BMM response. However, it is likely that there are other signaling pathways activated by the EVs which are important in promoting the macrophage response to an Mtb infection, but these pathways remain to be defined. These results also suggest that EVs released from Mtb infected macrophages may be useful as immuno-therapeutic agents in combination with antibiotics, a topic that will be discussed later in this review.

LGP2 is an RNA-sensing RLR with a unique role as an activator and inhibitor of innate immune signaling in viral infections, stimulating interest in its mechanism of action. While its role in mycobacterial infections has not been characterized, it offers an intriguing target for future study, as LGP2 dramatically reduces IRF3 and NFκB transcription and, in turn, limits IFN-β and TNF-α secretion [67]. LGP2 affects this response by functioning downstream and independently of MAVS, specifically by targeting TRAF protein activation, TRAF-NFκB activation, and IFN-β production [67]. LGP2 could hold insight into the interplay between RNA/DNA sensing pathways and pro-inflammatory responses, especially in inflammatory and autoimmune diseases where TRAF pathways are abnormally active. However, how it preferentially mediates its functions in such a diverse array of immune cascades is unclear.

PKR has recently emerged as an RNA sensor that functions mechanistically independent of the described RLR/MAVS pathway. PKR activation triggers the stimulation of NF-κB and kinase pathways, including MAPK-mediated pathways [68]. Mtb infection increases the expression of both PKR and its active, phosphorylated form. Mtb infection stimulates PKR-pathway activity, evidenced by the phosphorylation of its downstream target, the alpha subunit of protein synthesis initiation factor (eIF2α) [52]. Despite its activation being associated with IFN-β production and restriction of intracellular growth [52], the mechanism of PKR activation during Mtb infection remains undefined.

cGAS and AIM2-like receptors

DNA cytosolic sensors, which include cGAS and ALRs, activate upon binding to DNA, leading to IFN-β and pro-inflammatory cytokine production (Fig. 2). cGAS is the primary DNA sensor molecule that activates the STING to drive transcription of IFN stimulated genes (ISGs). cGAS can detect short and long DNA, forming dimers on lengthy DNA molecules to initiate a more robust signaling response [69]. ALRs are essential in inflammasome assembly and include absent in melanoma 2 (AIM2), interferon-gamma inducible protein 16 (IFI16/IFI205), and DNA-dependent activator of IFN-regulatory factors (DAI/ZBP1) [70].

A role for the cGAS/STING pathway in response to Mtb infection was originally described in four landmark papers published in 2015 [71] [72] [73] [74]. Presently, cGAS/STING is the most highly studied nucleic acid sensor pathway in the context of mycobacterial infections. The initial four publications all showed a requirement for cGAS/STING in driving optimum production of IFN-β upon Mtb infection, but they differed on what additional downstream effects this pathway has on the macrophage response to Mtb. Work by Wassermann et al. showed that Mtb DNA, released in an ESX-1 dependent manner, bound cGAS in the cytosol to drive inflammasome activation and IL-1β production [71]. Mtb DNA binding to cGAS is not the only driver of STING activation, as Dey et al. found that c-di-AMP produced by Mtb during macrophage infection can bind STING and stimulate IFN-β production and autophagy [73]. Other studies also found that a loss of STING or cGAS significantly reduced the association of LC3 with Mtb, indicating that this pathway is required for peak autophagosome formation during an Mtb macrophage infection [72] [74]. This limited recruitment of LC3 likely explains the lower ability of BMMs lacking cGAS or STING expression to control an Mtb infection [72] [74]. Interestingly, there is a more substantial reduction in autophagy markers in STING knockout cells than in cGAS knockouts, supporting a potential role for Mtb c-di-AMP in driving autophagy by binding STING independent of cGAS.

Interestingly, a recent paper by Liu et al. indicates that formation of lipid droplets in Mtb infected macrophages, which is a hallmark of Mtb-infected macrophages in vivo, is linked to the activation of cGAS/STING pathway [75]. They found that Mtb inhibits the DNA repair responses resulting in an increased number of micronuclei, triggering the cGAS/STING pathway leading to increased IFN-β production. The results suggest a novel pathway by which this nucleic acid detection pathway can regulate Mtb growth in macrophages.

Despite their observed role in autophagy, a limited role of cGAS and STING was observed during an in vivo infection. Mice deficient in these proteins showed similar colony forming unit (CFU) in the lung and spleen during the infection period [72] [74] [76] as well as similar pathology and levels of inflammatory cytokines such as IFN-γ, IL-1β, IL-12, and TNF-α [72] [76]. There was also no statistical difference in IFN-β production between Mtb infected WT C57Bl/6 and cGAS KO mice in one study [72] and a limited but significant difference in a separate study [74]. This is in contrast to Mtb infected MAVS KO mice, which showed markedly lower bacterial loads, decreased lung pathology, and diminished IFN-β production relative to infected WT mice [53]. This suggests that the RNA sensor pathway may play a more prominent role in driving IFN-β production during Mtb infection of C57Bl/6 mice. However, it is unclear if this extends to other strains of Mtb/host mice or whether a comparable role for these cytosolic sensors is observed during the course of a human infection. The importance of strain differences is supported by studies by Wiens and Ernst who found significant differences in IFN-β induction by Mtb DNA from different strains [77]. They also observed differences in mitochondrial stress induced by the different Mtb strains, leading to variations in both mtDNA released and cGAS/STING pathway activation. A similar role was observed for oxidative stress and release of mitochondrial DNA (mtDNA) in driving inflammasome-mediated IL-1β and cGAS-mediated IFN-β production in M. abscessus-infected macrophages [78]. Differences were also observed between Mycobacterium smegmatis and M. avium ssp. paratuberculosis in the release of extracellular mycobacterial DNA and activation of TBK1 within infected macrophages [79]. Interestingly, addition of the type I IFN inducer poly (I:C) (a synthetic dsRNA analog) or recombinant IFN-β impaired control of the M. avium ssp. paratuberculosis infection, while the use of IFNAR1 KO mice did not affect paratuberculosis survival [79]. This suggests that the timing and level of type I IFN induction, and, therefore, the timing and extent of nucleic acid sensor activation, can lead to significantly different host responses to mycobacteria. It is clear that our understanding of how type I IFN influences the immune response to mycobacterial infections remains limited. What we know about Type I IFNs during mycobacterial infections has recently been reviewed [80].

Although most studies have focused on infected macrophages, there is evidence that the cGAS/STING pathway can also play a role in DC activation. Mtb DNA transfected into bone-marrow derived DCs (BMDCs) stimulated IFN-β and IL-12 secretion and increased surface expression of CD40 and CD86 in a cGAS/STING dependent manner [76]. However, only IFN-β production was cGAS- and STING-dependent following an Mtb infection of BMDCs, suggesting that multiple PRRs engaged during an Mtb infection can stimulate DC activation, with the exception of IRF3 activation which is dependent on STING. In separate studies, activation of the cGAS/STING pathway was also observed in M. bovis and Mycobacterium paragordonae infected BMDCs [81] [82].

The mechanism by which mycobacterial DNA enters into the cytosol to stimulate the cGAS/STING pathway remains an open question. It is possible that DNA enters the cytosol via 1) lysosomal breakdown of mycobacteria, 2) transport of DNA into the cytosol from live mycobacteria via secretion systems, or 3) release of extracellular DNA present on the surface of mycobacteria [54]. It is clear that the ESX-1 secretion system is required for release of the Mtb DNA into the cytosol and activation of the cGAS/STING pathway. However, determining whether ESX-1 plays a direct role in release of the Mtb DNA or functions indirectly by affecting mycobacterial trafficking or by promoting mitochondrial damage remains to be elucidated. The latter possibility is supported by studies by Wiens and Ernst, who observed differences in mitochondrial stress induced by different Mtb strains, leading to variations in both quantity of mtDNA released and activation of the cGAS/STING pathway [77].

ALRs have complex, specialized roles in binding DNA and initiating or inhibiting specific signaling pathways [83] [84]. Both cGAS and ALRs associate with STING and are antagonized by their interactions with AIM2, hence, many studies have highlighted how AIM2 antagonizes the type-I IFN response in macrophages by binding to ALRs and STING [84]. For example, AIM2 has been shown to bind IFI205 and STING and block their interaction, resulting in attenuated type I IFN responses downstream [84]. In contrast, it has been argued that ALRs are not essential in the STING-dependent pathways nor for IFN responses [85]. Hence, the complex roles of ALRs warrant further exploration.

ALRs are encoded by PYHIN gene clusters, most of which have been structurally identified to have N-terminal pyridinoline (PYD) domains for facilitating protein-protein interactions and hematopoietic expression, interferon-inducible nature, and nuclear localization (HIN) domains at their C-termini for DNA binding [86]. There are four known ALRs in humans and 14 identified in mice, and the ALRs common to both species include IFI16/IFI204 and AIM2 [86]. ALRs have been noted for their diverse roles in modulating innate immunity by affecting STING-dependent type-I IFN production and inducing inflammasome formation upon detection of endogenous DNA [83]; [87]. IFI204 has been identified as a predominant DNA sensor in BMMs during Mtb infection. In one study, knocking down IFI204, the mouse homolog to IFI16, but not DAI, significantly reduced IFN-β expression in response to Mtb infection [88]. These data suggest a crucial role for ALRs in driving type-I IFN responses via STING/TBK1/IRF3 signaling while highlighting how these cytosolic DNA sensors are activated by extracellular Mtb DNA. IRF3 KO mice infected with Mtb saw significantly higher survival rates and lower CFU loads in the lung and spleen during the chronic stage of infection. Additionally, IRF3 KO mice had significantly lower levels of IFN-β, Regulated upon Activation, Normal T Cell Expressed and Presumably Secreted (RANTES), and monocyte chemoattractant protein-1 (MCP1) while exhibiting higher levels of IL-12 and IL-1β [88]. This suggests that activation of nucleic acid sensor pathways during Mtb infection promotes mycobacterial survival, at least in the context of a mouse infection. However, defining which cytosolic sensors (i.e. TLRs, RIG-I, cGAS, ALRs, others) are activated to drive IRF3 activation during the mouse infection remains an open question, and appears to vary between studies [74] [53] [72] [88].

AIM2 is an important cytosolic receptor that binds double-stranded DNA (dsDNA) and forms an inflammasome complex with apoptosis-associated speck-like protein containing a CARD (ASC), which activates caspase-1 and NF-κB [89]. Upon detecting DNA, AIM2 can increase caspase-1 cleavage and subsequent IL-1β and IL-18 activation to promote pro-inflammatory responses [90]. In one study, AIM2 KO mice infected with Mtb were significantly more susceptible to disease as evidenced by decreased survival and increased bacterial load in the lung and spleen [91]. Mice lacking AIM2 also had diminished levels of IL-10, IL-18, and IFN-γ as well as markedly reduced levels of caspase-1 cleavage following Mtb infection. Mtb DNA induced the p10 form of caspase-1 in the cytosol of WT macrophages but not in macrophages lacking AIM2, indicating that Mtb DNA can activate caspase-1 and stimulate production of IL-1β, IL-18, and IFN-γ in an AIM2-dependent manner [91]. In contrast, in a separate study, WT and AIM2 KO BMMs were infected with various mycobacterial species ranging from M. smegmatis to virulent Mtb H37Rv. There was an inverse relationship between IL-1β production/secretion and species virulence, with no significant difference in IL-1β between WT and AIM2 KO BMMs when infected with virulent Mtb [92]. Interestingly, a pre-infection of BMMs with Mtb H37Rv resulted in macrophages being refractory to M. smegmatis induced, AIM2-dependent IL-1β activation, and this was dependent on Mtb having an intact ESX-1 secretion system [92]. The reason for the contrasting results between these AIM2 studies and the later studies showing a dependence on cGAS for IFN-β production and inflammasome activation is unclear, but, as indicated above, likely depends on multiple factors including the use of different Mtb strains. AIM2 has also been studied in other mycobacteria species. In M. bovis infected BMMs, activation of AIM2 resulted in decreased autophagosome formation [93]. The data indicate that activation of AIM2 inhibits STING-dependent signaling pathways that are potentially induced by M. bovis DNA activating the cGAS pathway [93]. This suggests that M. bovis DNA can bind to multiple DNA sensors, and the degree to which it binds these cytosolic DNA sensors can modulate the level of inflammasome activation and IFN-β production.

Modulating nucleic acid sensors: Potential prophylactic and therapeutic use

BCG is the only Food and Drug Administration (FDA) approved vaccine against TB. This live, whole cell vaccine is most effective for prevention of meningeal and miliary TB and, to a lesser extent, pulmonary TB in young children, while its protection against pulmonary TB in adolescents and adults is highly variable [94]. While there are clearly multiple reasons for the limited efficacy of BCG against pulmonary TB, one may stem from limited MHC expression and antigen presentation in BCG infected macrophages. Interestingly, Bakhru et al. found that activation of TLR7, TLR9, and, to lesser extents, TLR4 and TLR2 stimulated antigen presenting cell (APC) MHC II expression and presentation of BCG antigens, suggesting that activation of nucleic acid sensing pathways could enhance the efficacy of the BCG vaccine [95]. Targeting of nucleic acid sensors has also been explored in the development of other vaccine approaches. Studies found that liposome encapsulation of CpG oligodeoxynucleotides with the Mtb antigenic secreted protein ESAT-6 provided significant protection against Mtb infection which was dependent on type I IFN production, likely indicating TLR9 recognition of the CpG motifs [96]. Another study found that STING-activating cyclic dinucleotides (CDNs) increased immune protection against Mtb infection when used as an adjuvant to a TB subunit vaccine [97]. Nucleic acid sensor agonists have also shown promise in certain therapeutic approaches to active TB infection. Studies have shown that the oral FDA-approved drug NTZ significantly inhibits intracellular Mtb growth and amplifies Mtb-stimulated RNA sensor gene expression and activity [52]. Activation of nucleic sensors may also be used in immunotherapy, as liposomes containing Mtb RNA, when combined with antibiotic treatment, were found to significantly reduce the bacterial load in Mtb-infected mice relative to antibiotic or liposome treatment alone [61].

The potential for using agonists to nucleic acid sensors for treatment of TB and other mycobacterial diseases would benefit from a better understanding of what cellular process are activated/inhibited downstream of their activation. Most studies have focused on inflammasome activation, autophagy, and cytokine production. However, a broader understanding of the effects associated with activation of the different nucleic acid PRRs is critical, not only for defining their therapeutic potential but also to define their role in the context of a mycobacterial infection. For example, we know that MAVS activation is associated with metabolic changes in cells [98], but these changes have not been evaluated in the context of mycobacterial infections. Moreover, how the different nucleic acid receptors work in combination during a mycobacterial infection has not been addressed. Therefore, it is not known if a combination of agonists may provide increased therapeutic benefit or be antagonistic.

Conclusions

The sometimes-conflicting results seen with the various RNA and DNA sensor knockout mice in the context of a mycobacterial infection may, in part, be due to redundancy between different receptors. Many of the receptors feed into the same transcriptional responses (figure 1 and 2) leading to similar protein expression profiles, although, as indicated above, there are likely changes specific to each receptor that have not been defined. The combination of nucleic acid receptors that are engaged during a mycobacterial infection may vary based on several factors including cellular location of the mycobacteria and the activation state of the infected cell. More studies are needed to define when a particular nucleic acid sensor receptor is engaged during a Mtb or NTM infection in vivo and how this may change during the course of the infection. Additionally, gaining insight into how experimental factors, such as mouse and bacterial strains, affect which receptors play a more dominant role is critical. The field would benefit by having some standard methods and strains with regards to mice and bacteria for the studies or, at minimum, having different sensor pathways tested in parallel in the same laboratory.

A role for nucleic acid sensors in human mycobacterial diseases has mostly been defined through genetic studies such as the linkage studies described for TLR8. However, a functional role for nucleic acid surveillance in TB patients was suggested in a recent study by Sousa et al.. In this study, they determined that Mtb strains isolated from mild disease patients consistently induced a robust cytokine response when added to human PBMCs compared to Mtb strains isolated from severe TB patients [99]. This was particularly evident for secretion of IL-1β, which was due to higher inflammasome activation in PBMCs infected with isolates from mild TB patients. The limited activation of the inflammasome by Mtb isolates from severe TB patients was due to their ability to evade the macrophage cytosolic nucleic acid surveillance systems [99], suggesting a linkage between severity of disease, inflammasome activation by Mtb, and cytosolic nucleic acid sensors. Although this study and some of the genetic studies are suggestive that activation of the nucleic sensors plays a beneficial role during an Mtb infection, we are far from having a definitive answer to this important question. Clearly, additional genetic and functional studies using both animal infection models and human patient populations are needed.

Biography

Biographical note: Joshua Ongalo is a graduate student in the Department of Biological Sciences, University of Notre Dame, William McManus, Joseph Vecchio and Kylie Webber are graduate students in the Intergraded Biomedical Sciences (IBMS) graduate program, University of Notre Dame and Jeff Schorey is a Professor of Biological Sciences, and Director of the IBMS program, University of Notre Dame.

Data Availability Statement:

No Data was included in this manuscript.