Background

Activation of microglia, the major innate immune cell in the brain, and the subsequent recruitment of immune cells occur in essentially all diseases of the central nervous system (CNS). Although inflammation occurs nearly universally in CNS injury, it is unclear how this response is evoked after brain damage. Immune receptors and their associated signaling pathways are important for crosstalk between immune cells and non-immune CNS cells (Heneka et al., 2014). In particular, specific pattern recognition receptors, such as the Toll-like receptors (TLRs), may play a major role in CNS disorders (Kielian, 2009). In principle, these receptors are assumed to have evolved in order to protect the host from invading pathogens including bacteria and viruses. Thus, they provide a first line of innate immune defense in the brain (Heneka et al., 2014). Over the past 20 years, a plethora of cell types, not only immune cells, have been found to express TLRs under both physiological and pathological conditions. Their activation not only results in an inflammatory response, but also impacts on cell survival, migration, plasticity, cell communication, etc. To date, 10 TLRs have been identified in human and 13 in mouse, of which TLR1 to TLR9 share a high degree of genetic and structural conservation. While the mouse genome encodes TLR1 to TLR13, among them TLR10 being considered non-functional, the human genome lacks functional TLR11, TLR12, and TLR13 (Hidmark et al., 2012; Kawai and Akira, 2010; Roach et al., 2005). A subgroup of TLRs, including TLR1, TLR2, TLR4, TLR5 and TLR6, is primarily expressed at the cell surface and recognizes lipidated and proteinaceous bacterial components. The nucleic acid-sensing receptors TLR3, TLR7, TLR8, TLR9, and TLR13 are located in endosomes (Hidmark et al., 2012; Kawai and Akira, 2010). In the CNS, TLRs are broadly expressed. While microglia, astrocytes and neurons express TLR1 to TLR9, oligodendrocytes were shown only to express TLR2 and TLR3 (Rietdijk et al., 2016; Hanisch et al., 2008; Lehmann et al., 2012; Xu et al., 2015).

In general, TLRs comprise monomeric proteins harboring an ectodomain composed of leucine-rich repeats (LRRs), which mediates ligand specificity, a transmembrane domain, and a cytoplasmic Toll/IL-1 receptor (TIR) domain that interacts with adaptor proteins within the cytosol. The canonical TLR signaling pathway consists of ligand binding, TLR dimerization or multimerization, docking of adaptor proteins such as myeloid differentiation primary response 88 (MyD88) to the TIR domain, followed by activation of downstream protein kinases, e. g. IRAK, NEMO and TRAF. This signaling cascade results in activation of transcription factors such as NF-κB and, consequently, in the production and release of chemokines and cytokines including tumor necrosis factor (TNF) as well as interferons (INF) (Kawai and Akira, 2010) (Figure 1).

Figure 1:

Signaling and regulatory pathway of ssRNA-sensing TLR7 and TLR8. LRR, leucine-rich repeats; TIR, Toll/IL-1 receptor domain; ssRNA, single-stranded RNA; siRNA, small interfering RNA; miRNA, microRNA.

TLR7 and TLR8 recognize ssRNA

In 2004, Heil and colleagues found that mouse TLR7 and human TLR8 in macrophages and dendritic cells (DCs) were activated by a single-stranded RNA (ssRNA) known as ssRNA40, derived from the human immunodeficiency virus-1 (HIV-1). Furthermore, they identified a specific RNA-activating consensus sequence composed of GUUGUGU repeats (G, guanine; U, uridine), which could be directly linked to the degree of TLR activation (Heil et al., 2004). Several studies linked GU-rich RNA motifs to species-specific TLR7/TLR8 recognition (Diebold et al., 2004; Heil et al., 2003; Heil et al., 2004; Jurk et al., 2002; Lund et al., 2004). Interestingly, not only viral, but also small interfering RNAs (siRNAs) containing such consensus variants can activate TLR7/TLR8 (Sioud, 2005, 2007; Sioud et al., 2007). Moreover, Forsbach et al. systematically evaluated nucleotide trimers and tetramers, finding those that were GU-rich or AU-rich to preferentially activate human TLR7 and TLR8, respectively. Indeed, diverse motifs were shown to have specific receptor preferences, resulting in the release of inflammatory factors including TNF and/or IFNs, indicating a specific RNA sequence-dependent mechanism triggering inflammation (Forsbach et al., 2008). Subsequently, Mancuso et al. found TLR7 to be important for the induction of type I INFs in bone marrow-derived DCs by streptococci (Mancuso et al., 2009). Thus, endosomal TLR sensing was suggested to be important for myeloid cell-mediated immunity against bacteria.

Human TLR8 was found to be a promiscuous receptor for uridine-rich 23S ribosomal (r) RNA from S. aureus, E.coli, S. pyogenes, and S. agalactiae, resulting in the induction of cytokines by mononuclear phagocytes (Ehrnstrom et al., 2017; Eigenbrod et al., 2015; Kruger et al., 2015). Furthermore, TLR8-mediated sensing of S. aureus RNA results in the formation of IFN-β and IL-12, which could be antagonized by concomitant TLR2 signaling (Bergstrom et al., 2015). Moreover, a TLR8 polymorphism is associated with immunity induced by a bacillus Calmette-Guérin (BCG) vaccine, by whole bacteria and by bacterial RNA, activating human and porcine antigen-presenting cells to propagate T follicular helper (TFH) cell differentiation (Ugolini et al., 2018). Finally, allergy-protective properties of the probiotic bacterium, Lactococcus lactis, have been linked to its RNA recognition by DC via TLR8 (humans) and TLR13 (mice).

TLR13 recognition of bacterial RNA

Whereas TLR7 was found to recognize bacterial RNA in mouse DCs, it appeared to be redundant in the activation of mouse macrophages (Mancuso et al., 2009). Still, the endoplasmic reticulum protein, UNC-93B, a signaling intermediate downstream of endosomal TLRs, contributed to the response by macrophages toward ssRNA from a broad range of bacteria, although all of the previously described nucleic acid-sensing TLRs (TLR3, 7, 8 and 9) were dispensable in this context (Deshmukh et al., 2011). The puzzle was solved with the identification in macrophages and DCs of TLR13 as an endosomal receptor for staphylococcal 23S ribosomal RNA (rRNA) (Hidmark et al., 2012; Oldenburg et al., 2012). TLR13 was found to be essential for the recognition of streptococci by phagocytes including microglia and intestinal macrophages. By contrast, TLR13 was largely dispensable for the induction of cytokines by streptococci in inflammatory monocytes. This indicates that the engagement of endosomal TLRs by streptococci and other bacteria depends on the macrophage differentiation status rather than on origin and tissue-specific cues (Kolter et al., 2016). Notably, the specific TLR13-binding sequence in bacterial rRNA overlaps with the binding site for macrolides, lincosamides and streptogramines (MLS antibiotics). Since 23S rRNA modifications resulted in circumvention of bacterial recognition by TLR13, antibiotic resistance could be linked to a lack of TLR13 signaling (Oldenburg et al., 2012). At a structural level, ssRNA induces TLR13 dimerization in a stem-loop-like structure, distinct from that in the bacterial ribosome. RNA-specific interactions with the concave surface of TLR13 enable the discrimination from DNA (Song et al., 2015). Whereas the principle role of TLR13 in the recognition of bacteria, in particular staphylococci and streptococci, seems now beyond doubt, its contribution to host defense and immunopathology in vivo has not been fully resolved (Hafner et al., 2019; Kolter et al., 2016). Moreover, although TLR13 expression was detected in several cell types of the brain, including neurons and immune cells, it remains enigmatic at this stage how endosomal TLRs contribute to bacterial meningitis and meningoencephalitis (Mishra et al., 2008).

Activation of TLRs by endogenous ligands in the CNS

TLRs play an essential role not only in regulating innate immunity against pathogens but also in cellular responses to endogenous stimuli. Widely accepted is their role in the inflammatory response as a consequence of tissue damage, e. g. in autoimmune and tumor diseases. Endogenous ligands for TLRs expressed in immune cells, both inside and outside the CNS include components of the extracellular matrix, such as hyaluronan and versican (activators of TLR4 and TLR2 signaling, respectively), heat shock proteins (HSP) HSP60 and HSP70 (activators of TLR2 and TLR4 signaling), and RNA molecules including mRNA and microRNA (miRNA) that are recognized by TLR3 and TLR7, respectively (Beg, 2002; Kariko et al., 2004; Lehnardt et al., 2008; Lehmann et al., 2012; Hu et al., 2015). In previous work we demonstrated that activation of microglial TLR4 and MyD88 by HSP60 released from injured neurons leads to further CNS injury (Lehnardt et al., 2008). Likewise, neuronal miRNAs such as let-7 are capable of inducing neurodegeneration through TLR7 expressed in both microglia and neurons (Lehmann et al., 2012). Thus, in the CNS, recognition of endogenous molecules by TLRs and their subsequent signaling cascades may contribute to chronic inflammatory diseases and further pathological processes in the context of brain disorders (Lehnardt, 2010; Lehnardt et al., 2008; Lehmann et al., 2012). Indeed, there is growing evidence that TLR signaling not only contributes to CNS infection, but also to brain diseases in which no obvious pathogen-derived molecules are detected. For example, the possibility that different TLRs contribute to neurodegenerative diseases are being intensively studied in mouse models of Alzheimer’s disease (AD), Parkinson’s disease, and amyotrophic lateral sclerosis (Heneka et al., 2014). It is well established that activation of microglia and chronic neuroinflammation occur at primary stages of these diseases. However, the molecular and cellular mechanisms of TLR activation by host-derived ligands, which ultimately lead to specific tissue injury in the respective brain disorder, remain unclear. It is conceivable that TLRs play different roles in the CNS under different pathological conditions. Endogenous TLR activators may activate multiple signaling pathways and induce transcription of distinct gene classes, as observed for pathogen-derived ligands. Furthermore, endogenous ligands, which are potentially released from injured CNS cells, may trigger further neuronal inflammation and damage, irrespective of the origin of the respective brain dysfunction. Thus, identification of TLRs and their endogenous ligands in the context of CNS injury is essential, both for understanding CNS disease mechanisms and development of therapeutic options.

miRNAs are new endogenous TLR ligands promoting CNS injury

MiRNAs are small noncoding RNAs of 19–24 nucleotides that typically bind to 3’ untranslated regions (UTR) of their target RNA. Thus, they are post-transcriptional regulators that in most cases inhibit translation. In humans, about 5000 sequences have so far been discovered that fulfill the classification of a miRNA – and the number is still growing. Moreover, the level of gene regulation by miRNAs is immensely complex since, based on their sequence, a single miRNA can bind many RNAs. In addition, one 3’ UTR can be occupied by several miRNAs. Notably, it was recently shown that miRNAs within exosomes and vesicles, as well as being part of protein complexes, can be secreted by cells into the extracellular space (Gaudet et al., 2018) (Figure 2). Due to as yet unidentified features, miRNAs are seemingly stable in body fluids including blood and cerebrospinal fluid (CSF). This enables them to mediate communication between cells and different tissues, and to serve as valid biomarkers for diverse diseases including inflammatory and neurodegenerative disorders (Bekris et al., 2013; Cogswell et al., 2008). So far, it remains speculative whether structural and/or sequence-based features of miRNAs, or their presence, e. g. in protein complexes, contribute to their firm stabilization in the extracellular space (Gaudet et al., 2018).

The human brain expresses several hundred miRNAs that are specific to a given lineage or cell type (Lagos-Quintana et al., 2002). They exert key roles in CNS development and physiological function, as well as in various CNS disorders including traumatic brain injury, stroke, brain tumors, and neurodegenerative diseases such as AD (Junn and Mouradian, 2012). Given that miRNAs are present extracellularly, and that neurodegenerative processes are associated both with TLR-mediated neuroinflammation and altered miRNA profiles in brain cells, we have raised the question whether miRNAs can also serve as signaling molecules for TLRs. Since the miRNA let-7b, which was first discovered in C. elegans and is also highly conserved in human tissues (Reinhart et al., 2000), is (i) highly abundant in the brain and (ii) contains the TLR recognition motif GUUGUGU (Lehmann et al., 2012), we hypothesized that in the brain this miRNA can function as an endogenous ligand for ssRNA-sensing TLRs. Indeed, both microglia and peripheral macrophages respond to extracellularly delivered let-7b by releasing TNF in a time- and dose-dependent fashion, and this response strictly requires TLR7. Furthermore, CNS injury is accompanied by neuronal release of let-7b, which can in turn induce further neuronal apoptosis, dependent on TLR7 expression in these cells (Lehmann et al 2012). The pathophysiological relevance of these findings was confirmed in vivo, when intrathecal injection of let-7b into mice resulted in loss of neurons and an increase in caspase-3-positive cells in the cerebral cortex. Importantly, the number of let-7b copies was elevated in CSF from AD patients compared to healthy individuals, implying an effect of this miRNA in a neurodegenerative disease context (Lehmann et al., 2012). In line with these findings, the high degree of sequence homology among the nine different let-7 miRNA family members (all of them contain 3’ GU-rich sequences), together with increased copy numbers of both let-7b and let-7e from extracellular vesicles, were detected specifically in the CSF of AD patients (Derkow et al., 2018). Recent studies in our lab revealed that injured and dying neurons release several miRNAs, not only let-7 miRNAs, which can serve as signaling molecules for TLR7 and/or TLR8. These miRNAs can then enter neighboring neurons and microglia that both express the receptors described above. Following TLR7/8 activation in the CNS, neurons undergo both cell-autonomous apoptosis and cell death mediated by pro-inflammatory and neurotoxic inflammatory molecules released from activated microglia (Lehmann et al., 2012; Lehnardt, unpublished data) (Figure 2). Principles governing the biological mechanism by which miRNAs serve as TLR signaling molecules were further characterized by the work of Fabbri and colleagues. These authors demonstrated that miRNA-21 and miRNA-29a are secreted by lung tumor cells via exosomes, followed by uptake into macrophages, which in turn leads to murine TLR7 and human TLR8 activation and subsequent secretion of TNF and IL-6 (Fabbri et al., 2012). In this lung tumor model, the miRNA-TLR7/8 interaction contributes to disease progression, as well as metastasis formation. Since then, numerous studies have identified several extracellular miRNAs, including miRNA-34a, miRNA-29b, and let-7c, as endogenous ligands for TLR7 expressed in e. g. cardiomyocytes, spleen cells, and neurons, which lead to chemokine and cytokine release after receptor activation (Feng et al., 2017; Salama et al., 2014; Yelamanchili et al., 2015). In addition, Salvi and colleagues showed that miRNAs, such as miR574, are present in exosomes and initiate INF-α release from human plasmacytoid dendritic cells (pDCs) through TLR7 signaling (Salvi et al., 2018). Interestingly, the role of miRNA-TLR signaling in neuroinflammatory processes was further characterized in the brain of rhesus macaques. Here, miRNA-21 copies were found to be increased in extracellular vesicles derived from simian immunodeficiency virus (SIV)-infected neurons, resulting in neurotoxicity through the TLR7 signaling pathway. In this context, IL-6 and TNF were released from microglia and macrophages upon miRNA-21 treatment (Yelamanchili et al., 2015). Furthermore, Liu and colleagues showed that murine neuronal TLR7 detects let-7c and miRNA-21, leading to changes in dendritic morphology (Liu et al., 2015). Taken together, extracellular miRNAs were recently uncovered as novel endogenous ligands for TLRs, indicating a role for these small RNA molecules in disease initiation and progress, as well as potential therapeutic targets (Gaudet et al., 2018; Iranifar et al., 2019).

Fig. 2:

Role of miRNAs as signaling molecules for TLRs in CNS inflammation. Neurons suffer an initial insult and undergo cell death, thereby releasing ssRNAs, such as miRNAs, into the extracellular space. These host-derived molecules bind to TLR7/8 expressed in (i) microglia that in turn release numerous proinflammatory mediators (e. g. TNF, yellow stars; IL-6, blue stars), or (ii) in neighboring neurons. In the case of microglial activation, this neuroinflammatory response causes injury of neighboring neurons leading to the release of endogenous TLR ligands. In the case of TLR7/8 stimulation in neighboring neurons, apoptosis and cell-autonomous neurodegeneration are induced, also leading to the release of endogenous TLR ligands and thereby closing the vicious cycle of neuronal injury.

Structural aspects of ssRNA-TLR interaction and immune function

Over the last years, several groups have succeeded in crystallizing nucleic acid-sensing TLR fragments from different species (Shimizu, 2017). With respect to ssRNA-sensing TLRs, crystallographic studies on monkey TLR7 and human TLR8 revealed structures at atomic resolution. Studies by Tanji et al. and Zhang et al. showed that two ligand binding sites in these receptors co-exist, which can synergistically lead to receptor activation. In monkey TLR7, one site can be bound by a free guanosine, and this interaction alone can lead to receptor homodimerization. The other binding site preferentially binds uridine-containing ssRNA and enhances the binding of guanosine to the first site. Therefore, TLR7 is a dual sensor for guanosine and uridine-harboring ssRNAs. Within human TLR8, the first binding site at the dimerization interface prefers uridine to guanosine, while the second site binds ssRNA fragments and leads to receptor activation (Tanji et al., 2015; Zhang et al., 2016). In contrast, the synthetic guanosine analogs loxoribine and resiquimode are able to activate TLR7 signaling without additional ssRNA interactions in the second binding site (Majer et al., 2017). With respect to the binding of miRNAs to TLR7 and TLR8, it is of interest that these small oligoribonucleotides hold different sequence features that can potentially bind to both sites within the respective receptor, resulting in an immune response (Lehnardt, unpublished observation). Thus, it is tempting to hypothesize that miRNAs can potentially function as activating chimeras by presenting uridine- and guanosine-containing sequences to the respective ssRNA-sensing TLRs. So far, it is not clear how a single ssRNA or miRNA molecule is connected to the binding sites, and how this structure may be composed. In crystallization experiments, only short uridine- and guanosine-containing RNA fragments were co-crystallized with TLRs, leading to the assumption that RNA could be fragmented by regulatory RNases in endosomes to be presented to the TLRs (Tanji et al., 2015) (see Figure 1). It is also likely that a combinatorial effect of sequences from different miRNAs could trigger the TLR response. Therefore, analyzing the TLR-ssRNA structure at atomic resolution may be an essential step in the design and identification of potentially therapeutic substances that might reduce collateral inflammation damage in a neuroinflammatory context.

Future perspectives

It is now well established that the interaction of exogenous and endogenous RNA species with endosomal TLRs contributes to early recognition of pathogens, immune regulation, and end organ damage. Over the last five years, miRNAs have been identified as new endogenous TLR ligands, thereby unraveling a new mechanism for regulating inflammation not only in the brain, but also in several other organs in the context of human disease. GU-/AU-rich sequences are needed for the activation of distinct binding sites within TLR7/8. However, it is not exactly clear how and where different ssRNAs including miRNAs are presented to such interfaces within and/or outside of the CNS. Moreover, it is not yet fully understood what miRNA structural features are essential determinants of TLR activation (and the subsequent peripheral and neuroimmune responses that follow), nor whether a given TLR activation requires the concerted binding of different miRNAs. Thus, further investigation of the RNA-TLR interaction is required to uncover potential structural features of RNA species that govern binding to TLRs and the conformational changes that result therein. For example, it is possible that particular secondary structure motifs stabilize a respective miRNA, especially since miRNAs seem to have a role in both cell-to-cell and long-distance communication within the organism. Further, for a better understanding of miRNAs’ role as TLR ligands, their binding kinetics should be characterized using biophysical and imaging approaches. Although some miRNAs have been shown to be secreted within exosomes and/or microvesicles, the temporal and spatial kinetics of the release, uptake into cells and transport to endosomes are unknown. Finally, the complex regulation of an immune response within and outside of the brain is most certainly not mediated only by a single ligand, such as a specific ssRNA molecule, or a single immune receptor class. Analysis of the combinatorial release of different ssRNAs/miRNAs, with e. g. different sequential features, and investigation of parallel immune receptor interaction/activation may be the key to understanding pathological processes and to defining novel biomarkers for numerous human diseases including CNS disorders.