The lysosomal catabolism of nucleic acids—critical regulators of the innate immune system

Abstract

Lysosomes play a pivotal role in degrading and recycling cellular macromolecules, including nucleic acids. Notably, nucleic acids are critical modulators of the innate immunity sensed in endo-/lysosomes, highlighting the relevance of their rapid and tightly regulated turnover. This review explores the intricate processes governing the uptake routes of nucleic acids into lysosomes, the lysosomal catabolism of DNA and RNA to nucleosides and phosphate, and the export of the degradation products, emphasizing the key enzymes and proteins, regulatory mechanisms, and pathological implications of impaired degradation. We highlight open questions in this process and discuss controversies in the field. We discuss the interdependence of efficient nucleic acid degradation and endo-/lysosomal nucleic acid sensors (Toll-like receptors) and pathological implications in human diseases as a result of impaired nucleic acid degradation, e.g. in genetic deficiency disorders resulting in loss-of-function of critical enzymes. This review integrates current research findings, highlighting the significance of lysosomal nucleic acid catabolism in cellular physiology and its link to disease pathogenesis, and hopefully stimulates research in the field that will finally fully comprehend this complex interplay between lysosomal degradation and immunity.

Graphical Abstract

Graphical Abstract.

Main text

The intracellular catabolism of complex macromolecules like proteins, complex lipids, glycosaminoglycans, N- and O-glycans, or nucleic acids by lysosomal enzymes is a highly ordered process mediated by the concerted action of ∼50–60 enzymes and additional accessory proteins and is generally well-understood. However, there are still some gaps in our understanding of some hydrolytic processes in lysosomes and their resulting physiological and pathophysiological consequences, e.g. in the case of loss-of-function mutations.

Among the poorly defined degradative pathways is the complete hydrolysis of nucleic acids, which are first decomposed into their mononucleotide building blocks by nucleases before dephosphorylation to nucleosides and free phosphate by nucleotidases, specialized phosphateses (Fig. 1). Both RNA and DNA reach lysosomes by autophagy (e.g. cytosolic messenger (mRNA), transfer RNA (tRNA), ribosomes, mitochondria, or piecemeal autophagy of the nucleus) and, at least in a subset of cell types, by phagocytosis of foreign pathogens (bacteria and viruses) or self debris (e.g. due to the phagocytosis of apoptotic cells) [1]. The efficient degradation of these polymeric, high-molecular-weight nucleic acids to low-molecular-weight molecules that are exported from the lysosomal lumen to the cytosol is crucial for replenishing cellular fuels for nucleotide synthesis, energy production, and for preventing lysosomal accumulation and consequently lysosomal storage. The amount of nucleic acids that is turned over in lysosomes is remarkable: it was found that in rat liver under conditions of nutritional deprivation, 65% of total cytoplasmic RNA is degraded per day, with ~70%–85% occurring in lysosomes [2], highlighting the need for very efficient degradation.

Figure 1.

General catabolism of nucleic acids inside lysosomes. DNA or RNA polymers are degraded to nucleotide-3′-monophosphates by endo- and exonucleases through hydrolysis of the phosphodiester backbone at the ribose’s 5′-O-bond. The nucleotides are then further decomposed to nucleosides and free phosphate by nucleotidases before being exported from the lysosome.

Importantly, and in contrast to many other metabolites catabolized in lysosomes, the efficient turnover of nucleic acids has an additional critical physiological function: Late endosomes and lysosomes serve as organelles in the sensing of a subpopulation of receptors of the innate immune system [pathogen-associated molecular patterns (PAMPs)], namely intracellular Toll-like receptors (TLRs) [3]. Intracellular TLRs bind both DNA and RNA as specific ligands and transmit, upon dimerization, their signal to intracellular signaling cascades, finally activating nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) and, consequently, pro-inflammatory gene expression. Thus, TLR-mediated signaling renders the endolysosomal turnover of nucleic acids in a well-adjusted and well-controlled manner critical for a balanced immune response against pathogens and autoimmunity [3]. A complete understanding of the lysosomal degradation of nucleic acids is therefore critical for understanding and modulating endosomal TLR-mediated innate immunity.

The precise order and specificity of the lysosomal turnover and the responsible hydrolytic enzymes of both DNA and RNA are still incompletely understood. Experiments from the early 1970s revealed that the catabolic end-products of lysosomal degradation are inorganic phosphate and nucleosides [4]. Here, we review the degradation of nucleic acids in lysosomes, highlight open questions in this process, and discuss discrepancies in our current understanding of these degradative processes.

Mechanisms of nucleic acid delivery to lysosomes

Nucleic acids can enter lysosomes via different pathways (Fig. 2). Extracellular sources of nucleic acids include cell-free DNA and RNA, apoptotic cell debris, and microorganisms like bacteria, fungi, and protozoa. All of these can be taken up by immune cells, leading to the formation of phagolysosomes that digest cellular structures and liberate DNA and RNA into the lysosomal lumen [5]. Viruses can also infect host cells and end up in the endolysosomal system, where the genome of nonenveloped virions could come into contact with endolysosomal components as described, e.g. for parvo-, picorna-, or adenoviruses [6–9]. Under physiologic conditions, one of the most significant extracellular DNA sources is maturing erythroblasts, which release their nucleus to nearby macrophages during erythropoiesis. One nucleus contains a diploid human genome of ~6.3 gigabase pairs corresponding to 12.6 billion single nucleotides in its double-stranded DNA (dsDNA) [10]. A healthy adult with a 70 kg body weight produces ~2.2 × 10¹¹ new red blood cells per day [11], each of which bears its own nucleus, leading to a total turnover of 2.7 × 10²¹ DNA nucleotides (4.5 millimoles or 1.4 g) per day only through erythropoiesis [12].

Figure 2.

Schematic overview of the different routes for nucleic acids to enter the endolysosomal system. Exogenous nucleic acids derived from pathogens or cell debris are taken up, especially by phagocytes, via unselective phagocytosis into endosomes. Endogenous nucleic acids mainly enter the lysosome via autophagic processes, either unselective autophagy of cytoplasmic nucleic acids or regulated processes like mitophagy, ribophagy, or nucleophagy. A transport protein complex consisting of LAMP2C and SIDT2 has been described as a direct uptake mechanism for free cytosolic RNA in lysosomes, but SIDT2 lacks a channel for nucleic acids and its precise role in nucleic acid uptake is still controversial.

Apart from exogenous genetic material, endogenous DNA and RNA can end up in lysosomes through autophagy. Mitochondria, which contain both DNA and RNA, are taken up into lysosomes via mitophagy. Mitophagy contributes to the clearance of defective mitochondria and reduces heteroplasmy in embryos by selective allogenic degradation of paternal mitochondria, as shown in Caenorhabditis elegans [13–15]. Defects in mitophagy can lead—among others—to neurodegenerative, inflammatory, metabolic, and cardiovascular diseases [14, 16]. Cardiac hypertrophy, inflammation, and heart failure are especially characterized by defects in mitophagy and accumulation of mitochondrial DNA, activating TLRs [17–19]. Mitophagy and impaired degradation of mitochondrial nucleic acids are also implicated in the pathogenesis of neurodegenerative diseases [20].

The nucleus is the main organelle containing cellular DNA, which can be partially or wholly taken up into autophagosomes. These processes, summarized under the term “nucleophagy,” have mainly been studied in Saccharomyces cerevisiae and other fungi but can also be found in mammalian cells. Piecemeal microautophagy of the nucleus in S. cerevisiae, where the yeast vacuole at nucleus–vacuole junctions selectively engulfs small parts of the nucleus, has been shown to depend on the core autophagy genes, including atg7, atg8, and atg9, amongst others [21, 22]. Similarly, in mammalian cells, nucleophagy depends on the protein LC3, the mammalian homolog of yeast Atg8 [23]. Mammalian nucleophagy seems to occur physiologically during the terminal differentiation of skin keratinocytes and pathophysiologically in the context of nuclear envelopathies [23, 24], whereas, in yeast and other fungi, it also seems to be part of physiological functions like growth and gametogenesis [25–27].

In the case of ribosomal RNA, ribophagy refers to the selective autophagy of ribosomes. Nonfunctional ribosomes are usually degraded via cytosolic enzymes and the proteasome; however, lysosomal ribophagy seems to occur upon starvation or other cellular stress conditions [28, 29]. Upon starvation, ribosomes are delivered to the autophagosome by a complex of the nuclear fragile X mental retardation–interacting protein 1 (NUFIP1) and the zinc finger HIT domain-containing protein 3 (ZNHIT3), a process regulated by mTORC1 and LC3 [30]. In yeast, selective ribophagy requires the action of the ubiquitin protease complex Ubp3p/Bre5p, removing a ubiquitin molecule from the 60s ribosomal subunit [29]. Whether or not ubiquitination also plays a role in mammalian ribophagy remains unclear, but a specific ribosomal signal during starvation periods, either through chemical modification or protein binding, seems necessary for autophagic targeting [30]. Indeed, ribosomal RNA (rRNA) accumulates in lysosomes when RNA degradation is disrupted, as shown in the brain of rnaset2-deficient zebrafish, leading to neurodegenerative processes [31].

More recently, an uptake mechanism for cytosolic nucleic acids has been discovered and named DNautophagy or RNautophagy [32, 33]. Initially, the lysosome-associated membrane protein 2C (LAMP2C) has been described as a receptor for direct ATP-dependent DNA or RNA import into lysosomes via the transport protein SIDT2 [systemic RNA interference defective protein 1 (SID1) transmembrane family member 2] in analogy to chaperon-mediated autophagy of proteins [32–34]. The expression of both LAMP2C and SIDT2 genes is upregulated by the cytosolic RNA sensor protein MDA5 in response to foreign RNA [35].Overexpression of SIDT2 can lead to enhanced uptake of mRNA into lysosomes [36]. When located in the plasma membrane, SIDT2 furthermore acts as an import protein to take up cell-free single-stranded oligonucleotides into the cytosol of living cells, a process called gymnosis [37]. In Caenorhabditis elegans, the SIDT homolog SID1 is responsible for the uptake of small interfering RNA (siRNA) into cells, enabling RNA interference [38]. Interestingly, SIDT2 and its homolog SIDT1 also mediate the export of dsRNA from endo- and lysosomes into the cytosol for its recognition by cytosolic receptor proteins like RIG-I-like receptors (RLRs), leading to a type I interferon (IFN) response [39, 40]. Whether SIDT2 and its close relative SIDT1 indeed possess the ability to directly transport nucleic acids across the membrane is highly improbable: two recent studies independently solved the structure of C. elegans cSID1 bound to dsRNA by CryoEM, revealing that SID1 family proteins do not form channel pores that could enable the passage of double-stranded (dsRNA) across a membrane [41, 42]. Furthermore, dsRNA strands are coordinated by the extracellular domain of SID1 (corresponding to the luminal domain of SIDT1/2 in endolysosomes) in a way that the nucleic acid strand is oriented parallel to the membrane surface, further hindering its passage through the lipid bilayer. For longer dsRNA strands, this could even enable the binding to several SID1 proteins, gathering them in close proximity to each other. Thus, it was suggested that SID1 and its homologs in humans serve as dsRNA receptors leading to a clathrin-mediated endocytosis of nucleic acids [42]. In SIDT2, several tyrosine-based sorting motifs interact with the adapter protein complexes AP-1 and AP-2 to mediate its endocytosis and sorting into endo- and lysosomes [43]. Whether this process is used solely to sort SIDT2 into the endolysosomal system or is a receptor-mediated import mechanism for extracellular nucleic acids is still unclear. This, however, would still not explain how dsRNA actually crosses the membrane to reach the cytoplasmic site. Furthermore, the published CryoEM structures mainly confirm the binding of dsRNA to the extracellular or luminal domain of the SID1 family members. Although the addition of ssRNA to hSIDT2 yields an electron-dense area besides in its cytosolic domain, the exact binding process could not be mapped [44]. Functional studies revealed an arginine-rich motif in the cytosolic domain potentially responsible for nucleic acid binding on the cytosolic site [36]. A bidirectional transport mechanism, as supported by several functional studies, is thus inexplicable with our current knowledge and would likely involve two fundamentally different recognition and transport mechanisms. Moreover, for both SIDT1 and SIDT2, a phospholipase activity toward C18 ceramide has been described to be essential for nucleic acid transport without a further mechanistic explanation [44, 45]. Thus, it could be involved in both a lipid signaling process to initiate nucleic acid transport, as well as in the endolysosomal lipid metabolism. Both transport directions likely require further, yet unidentified proteins to translocate nucleic acids through the membrane, possibly also including LAMP2C, which is located in lysosomes and is most likely not involved in any SIDT1/2-mediated transport processes at the plasma membrane. It is currently plausible that SID1 family members serve as nucleic acid receptors in the process of receptor-mediated endocytosis/autophagy and are only indirectly implicated in the import into the lysosome. The physiological role of SID1 family members in mammals is subject to ongoing investigations. Sidt2-deficient mice are generally viable, but show signs of liver disease and, in higher ages, cerebellar ataxia, seizures, and further neurological defects [46, 47]. Furthermore, Sidt2^−/− mice show a reduced antiviral response against ssRNA viruses [40]. Recently, a human case study revealed a biallelic defect in the SIDT2 gene, leading to cerebellar atrophy, ataxia, and cognitive impairment [47]. In both human and murine cells, SIDT2 deficiency leads to autophagy defects and signs of lysosomal storage accumulation, although it is unclear to what extent cellular RNA is contributing to this phenotype [46, 47]. On the other hand, therapeutic interest in SIDT2 is growing as a delivery mechanism for oligonucleotide-based drugs into human cells [48]. For now, the mechanism of nucleic acid oligomer translocation across the membrane remains one of the most enigmatic transport processes in lysosomes and requires extensive further research to resolve the remaining questions.

Other than as part of nucleic acids, mononucleotides, as well as nucleosides and nucleotide derivatives like ADP-ribose, NAD⁺, FMN, or FAD, enter the lysosome as bycatch of phagocytotic processes, unselective autophagy, and mitophagy. Despite only accounting for a minority of nucleotides inside the lysosome, their recycling is especially important for long-living cell types and organisms to prevent an accumulation over time and to ensure replenishment of critical metabolites. Cellular levels of NAD⁺ are also regulated through mitophagy and thus major defects in autophagy promote cell death by depletion of cellular NAD⁺ levels [49, 50]. Conversely, the accumulation of glutamyl ribose 5-phosphate, a degradation product of ADP-ribosylated proteins, has been described in a rare lysosomal storage disorder with renal and neuronal defects [51], although the underlying genetic defect has not been discovered to date.

The lysosomal modification and degradation of DNA

After reaching acidic compartments, the degradation of DNA is initiated by DNase II, an endonuclease that cleaves dsDNA [52, 53] (Fig. 3 and Table 1). Mammals express two paralogs, coded by DNASE2 and DNASE2B (also called DNase II-like acid DNase = DLAD), which share about 40% sequence similarity on the amino acid level. Both proteins belong to the Phospholipase D (PLD) superfamily of hydrolases, which contains a variety of phosphoesterases acting as phospholipases, exo-, or endonucleases throughout all domains of life [54]. All superfamily members are characterized by a common domain architecture containing two PLD domains in antiparallel orientation around a central catalytic site, either formed by dimerization or contained in a single protein. Their catalytic site typically involves an HXKXXX(X)D amino acid motif (abbreviated as HKD), a common sequence motif responsible for the reaction mechanism, together with two more asparagine residues conserved in almost all family members. Both DNASE2 and DNASE2B genes encode N-terminal signal peptides, ensuring their translation into the ER lumen for further transport to the lysosome [55, 56]. Notably, DNASE2B can also be expressed as a shorter, N-terminally truncated isoform via an alternative transcription start site containing only one of the PLD domains. It is likely that this shorter isoform forms homodimers to complete the catalytic site, but to date, there is still no experimental evidence on the exact structure of both mammalian DNase II enzymes. For the short isoform, the alternative transcription starting site leads to the loss of its N-terminal signal peptide, which makes an involvement in the endolysosomal nucleic acid metabolism unlikely [57]. While DNASE2 is broadly expressed [55, 58], the expression of DNASE2B is restricted to specific tissues. In mice, the long isoform of DNase IIb is mainly found in the eye lens, where it is crucial for the degradation of nuclear DNA during lens fiber development, leading to a congenital cataract in case of Dnase2b deficiency, while for humans, DNASE2B expression could not be verified in lens fibers by RNA sequencing [57]. The short isoform of DNASE2B has been detected in human salivary glands and alveolar macrophages but seems to be absent in rodents; its exact role remains unclear. Structural models suggest the formation of an active site by dimerization, compensating for the fact that the first catalytic PLD hemidomain is lost to the alternative splicing [57]. DNase II is the major lysosomal DNase in macrophages, responsible for the degradation of extracellular sources of DNA, e.g., from apoptotic cells, in phagosomes [59–61]. The deficiency of DNase II in mice leads to perinatal lethality, autoimmune vasculitis, arthritis, severe anemia, and impaired thymus development [61–64]. The accumulation of nuclear DNA derived from developing erythroblasts in macrophages, highlighting its critical function already for initiating the turnover of dsDNA [62]. This phenotype is driven by the cytosolic stimulator of interferon genes (STING) pathway [65–67], which induces the production of type I IFNs, leading to anemia [62, 63, 68], as well as proinflammatory cytokines like interleukin (IL)-1β, IL-6, and tumor necrosis factor-α (TNF-α), promoting a chronic polyarthritis [64, 69]. Interestingly, TLRs seem to only play a minor role in this inflammatory reaction [65, 70]. In humans, the loss of DNase 2 activity leads to a type I interferonopathy comparable with Dnase2a^−/− mice, including severe anemia, hepatosplenomegaly, arthropathy, cerebral white matter lesions, and autoimmune glomerulonephritis requiring immunosuppressive therapy [71]. Apart from a complete loss of function, rare sequence variations in the DNASE2 gene have also been linked to autoinflammatory diseases, specifically rheumatoid arthritis and renal disease in patients with systemic lupus erythematosus [72–74].

Figure 3.

Lysosomal DNA degradation is initiated by endonucleolytic cleavage of dsDNA by DNase II into shorter fragments that finally dissociate into single strands, which are further degraded by the 5′-exonucleases PLD3 and PLD4 to deoxy-3′-mononucleotides (dNMPs). Short, CpG-motive containing ssDNA can activate the endolysosomal TLR9 as a pattern recognition receptor. In parallel, the adenosine deaminase ADA2 can convert deoxyadenosine nucleotides inside dsDNA strands into deoxyinosine nucleotides before further degradation.

Table 1.

Overview on lysosomal DNases and DNA-modifying enzymes, their encoding genes, tissue distribution, and enzymatic properties

Protein Name	Gene	Tissue distribution	Substrate	Product
DNase II	DNASE2	Ubiquitous	dsDNA	Shorter dsDNA and ssDNA fragments
DNase II-like acid DNase (long isoform)	DNASE2B	Liver and lens fibers (mouse)	dsDNA	Shorter dsDNA and ssDNA fragments
DNase II-like acid DNase (short isoform)		Salivary glands and alveolar macrophages (humans) Absent in mice
Phospholipase D3, 5'-3' exonuclease PLD3	PLD3	Ubiquitous, enriched in neurons	ssDNA ssRNA	Nucleoside-3′-phosphates
Phospholipase D4, 5'-3' exonuclease PLD4	PLD4	Enriched in immune cells and microglia	ssDNA ssRNA	Nucleoside-3′-phosphates
Adenosine deaminase 2	ADA2	Immune cells	Adenosine inside dsDNA	Inosine inside dsDNA

For all members of the DNase II family, the enzymatic reaction is catalyzed by the two conserved HKD motifs in the active site and an additionally conserved Asn residue [75]. Their endonucleolytic activity is independent of any metal ions and is even inhibited by divalent cations like Ca²⁺, Mg²⁺, Co²⁺, and Mn²⁺ [76]. Chelators like EDTA, on the other hand, seem to act stimulatory on DNase II, but not DNase IIb [74]. The pH optimum of human DNase II and murine DNase IIb has been determined to be around pH 5.2, corresponding to the luminal pH of late endo- and lysosomes [56, 76]. All DNase II family members generate DNA fragments with a 3′-terminal phosphate residue and a 5′-OH group [77, 78]. They cleave DNA with a low sequence specificity, showing a slight preference for G nucleotides [77, 79]. To what extent (i.e., length of DNA fragments) hydrolysis occurs remains to be elucidated. However, it is assumed that DNase II generates smaller single-stranded DNA (ssDNA) fragments that are subject to further ssDNA-specific exonuclease activity.

Indeed, an acidic ssDNA-specific 5′-exonuclease was already biochemically characterized in 1968 and named “spleen exonuclease,” [80] but the corresponding gene could not be assigned initially. In 2018, “spleen exonuclease” was identified as a mixture of two members of the PLD superfamily, Phospholipase D3 and D4 (PLD3/4), both of which do not exhibit classical phospholipase activity but rather exonuclease activity toward ssDNA [81]. PLD3 and PLD4 are also able to degrade ssRNA exonucleolytically (Table 1) [82, 83]. PLD3 and PLD4 are paralogs sharing redundant functions in vivo, although PLD4 displays a lower specific activity and more acidic pH optimum [81, 84, 85]. In contrast to PLD3, which is broadly expressed in various cell lines and tissues, the expression of PLD4 is restricted to immune cells, specifically microglia and plasmacytoid dendritic cells [85–87]. PLD4 expression increases in activated microglia cells in white matter lesions and upon stimulation with lipopolysaccharides [87].

PLD3 is a resident lysosomal protein and is synthesized as a highly N-glycosylated type II transmembrane protein, which undergoes proteolytic processing in acidic compartments, yielding a stable, soluble, domain [88]. In contrast to other lysosomal membrane proteins, the transport of PLD3 involves the ubiquitination of the cytosolic N-terminus and subsequent sorting into intraluminal vesicles, where proteolytic processing occurs [88]. PLD3 shows a slight preference for 5′-thymidine but is able to cleave any 5′-nucleotide with decreasing efficiency (T > G > A > C) [84]. Consistent with other nucleases of the PLD superfamily, the catalytic mechanism is independent of divalent metal ions as cofactors and is mediated by two catalytic triads of His, Lys, and Asn in antiparallel orientation [84, 89]. Similar to the lysosomal RNases discussed below, the products of both PLD3 and PLD4 are mononucleotides with a 3′-phosphate group. PLD3 has been initially shown to cleave phosphorothioate linkages present in many synthetic oligodeoxynucleotides in vitro, albeit later analyses showed that this activity is much slower, making its relevance in vivo questionable [81, 84]. It has also been shown to cleave 5′-phosphate groups from nucleic acids with much lower efficiency than for 5′-nucleotides [90]. In vivo, it most likely completes the degradation of short oligonucleotides generated by endonucleases to mononucleotides [85].

Genetic variants in PLD3 have previously been shown to double the risk of developing Alzheimer’s disease [91], and interestingly, the expression levels of PLD3 are highest in neurons. This genetic link, however, could not be reproduced by other studies [92–94]. Human SH-SY5Y neuroblastoma cells deficient for PLD3 show an accumulation of mitochondrial DNA inside their lysosomes, indicating a disruption in mitophagy [95]. Genetic variants in the human PLD4 gene have been associated with autoimmune diseases like systemic lupus erythematosus and systemic sclerosis [96, 97]. Biallelic loss-of-function mutations in the PLD4 gene can even manifest as systemic lupus erythematosus through the activation of TLR signaling and production of proinflammatory cytokines and interferons [98]. A rare nonsense mutation in the PLD4 gene of Fleckvieh cattle, abolishing its enzymatic activity, leads to an autoimmune disorder leading to skin and mucosal lesions with growth retardation resembling a hereditary zinc deficiency [99, 100]. Pld4^−/− mice show subtle signs of TLR9-dependent inflammation with a slight splenomegaly and increased levels of IFN-γ, while the single deficiency for Pld3 did not cause an obvious phenotype and only led to a stronger response to TLR9 agonists [81]. In contrast to these rather subtle phenotypes, the knockout of both Pld3 and Pld4 in mice leads to death soon after birth due to a strong autoimmune reaction mimicking a primary hemophagocytic lymphohistiocytosis. This is promoted by the endolysosomal TLRs 7 and 9, as well as the cytosolic cGAS–STING pathway due to the accumulation of their nucleic acid ligands [81, 95, 101].

Apart from nucleases, other enzymes act on DNA inside the lysosome. Specifically, the adenosine deaminase 2 (ADA2), although traditionally thought to be a secreted protein, has recently been shown to be involved in the lysosomal DNA metabolism. ADA2 converts deoxyadenosine inside dsDNA into deoxyinosine but has much lower activity on free adenosine [102]. This modification generates CpI motives from former CpA dinucleotides, structurally resembling CpG motives, and therefore are believed to activate TLR9. Deficiency of ADA2 in humans causes a hereditary autoinflammatory disease characterized by vascular inflammation, immunodeficiency, and dysregulated interferon signaling [103]. However, the underlying pathophysiological mechanisms and the exact contribution of ADA2 to pathogen recognition in lysosomes are still unclear, especially since ADA2 activity seems to act proinflammatory in cellulo, but its deficiency in vivo also leads to autoinflammation. Its pH optimum for the deamination of free adenosine lies ~6.5, corresponding to early endosomes [104]. However, its action on deoxyadenosine as part of a dsDNA strand proceeds efficiently from pH 5.0 to 6.5, indicating a role throughout the endolysosomal system [103, 104]. This makes an involvement in the editing of endolysosomal DNA substrates more likely than a role in the degradation of free adenosine to inosine.

The lysosomal degradation of RNA

The only well-characterized lysosomal RNase is RNase T2, a glycoprotein with relatively low sequence specificity but a preference for purine bases and uridine that shows high activity against different RNA substrates (Fig. 4 and Table 2) [105, 106]. An early discovery of a lysosomal acid RNase from HeLa cells most likely corresponds to RNase T2 by its pH optimum, substrate specificity, and susceptibility to inhibition by copper and zinc ions [107]. The crystal structure of human RNase T2 was solved, and like almost all other RNases of the T2 family, it is found in a monomeric state with one monomer in the asymmetric unit, which also comprises the active biological unit. The structure shows a typical fold for members of the RNase T2 family with seven α-helices and eight β-strands constituting a typical α/β hydrolase fold [108]. Although RNase T2 does not require divalent ions as a cofactor, it can be efficiently inhibited by copper and zinc ions, for which putative binding sites have been discovered in its crystal structure [108]. It preferably cleaves RNA endonucleolytically between a purine and a uridine nucleotide via a transesterification reaction using the purine’s 2′ hydroxyl group as a nucleophile to attack the 3′-phosphate group. This reaction yields a terminal 2′,3′-cyclic phosphate [105] and is the reason why RNase T2 has no activity on DNA.

Figure 4.

The lysosomal degradation of RNA proceeds in different steps: First, long single-stranded and double-stranded RNA molecules are cleaved endonucleolytically by RNase T2 [between a purine (X) and a uridine (U)], and the RNases 2 and 6 (preferentially between a uridine and an adenine (A). This step releases shorter RNA fragments with a 2′,3′-cyclic nucleotide at their neo-3′-terminus, as well as uridine-3′-monophosphate as a free nucleotide. These shorter fragments are further degraded by the 5′-exonucleases PLD3 and PLD4 to 3′-nucleotide monophosphates and a purine-2′,3′-cyclic phosphate, which is further hydrolyzed to a 3′-nucleotide by RNase T2.

Table 2.

Overview on putative and confirmed lysosomal RNases, their encoding genes, tissue distribution, and enzymatic properties

Protein name	Gene	Tissue distribution	Substrate	Product
RNase 1	RNASE1	Ubiquitous Exocrine pancreas acini	dsRNA, ssRNA	Shorter dsRNA/ssRNA and Nucleoside-3′-phosphates
RNase 2	RNASE2	Immune cells (eosinophils, monocytes)	dsRNA, ssRNA	Shorter dsRNA/ssRNA and Nucleoside-3′-phosphates
RNase 4	RNASE4	Urinary tract, colon epithelium	dsRNA, ssRNA	Shorter dsRNA/ssRNA and Nucleoside-3′-phosphates
RNase 6	RNASE6	Immune cells (Macrophages)	dsRNA, ssRNA	Shorter dsRNA/ssRNA and Nucleoside-3′-phosphates
RNase T2	RNASET2	Ubiquitous	dsRNA, ssRNA	Shorter dsRNA/ssRNA and Nucleoside-3′-phosphates
Phospholipase D3, 5'-3' exonuclease PLD3	PLD3	Ubiquitous,enriched in neurons	ssDNA, ssRNA	Nucleoside-3′-phosphates
Phospholipase D4, 5'-3' exonuclease PLD4	PLD4	Enriched in immune cells and microglia	ssDNA, ssRNA	Nucleoside-3′-phosphates

RNase T2 deficiency leads to a severe childhood-onset cystic leukoencephalopathy affecting patients in their first year of life, mimicking a cytomegalovirus brain infection with multifocal white matter lesions and microcephaly [109]. Consistently, mice with a knockout in both RNase T2 genes Rnaset2a and Rnaset2b show signs of a type I interferon-driven neuroinflammation and additional hepatosplenomegaly due to immune cell infiltration, leading to overall reduced survival after 10 weeks of age [110]. The autoinflammation thereby seems to depend on the overactivation of TLR13 by bacterial rRNA or another endogenous ligand [111, 112]. A study of a zebrafish rnaseT2 knockout model also found a role in rRNA degradation, which accumulates in the lysosomes of rnaset2 deficient zebrafish embryos [31]. Later work suggested an additional role in the degradation of apoptotic nuclei in microglia during embryonal development [113, 114]. RNase T2-deficient monocytes fail to respond to foreign RNA via TLR8, most likely due to the defective processing to shorter oligoribonucleotides as suitable ligands for TLR8. Interestingly, the immune response in RNASET2^−/− cells is skewed toward TLR3, typically recognizing double-stranded RNA, and—at least in rodents—toward TLR13, detecting bacterial rRNA [111, 112, 115]. This indicates that RNase T2 is mainly involved in breaking down longer RNA molecules into smaller oligo- and mononucleotides, which can then be processed further. Recent works also highlighted that the concerted action of RNase T2 and the 5′-exonucleases PLD3 and PLD4 releases guanosine 2′,3′-cyclic phosphate, and short oligoribonucleotides, which are both ligands for TLR7 [85]. The cyclic nucleotides resulting from the concerted degradation by RNAse T2 and PLD3/4 exonucleases are hydrolyzed to the respective 3′-mononucleotides also by RNase T2. An excess of free nucleotide 2′- or 3′-monophosphates can, however, inhibit this phosphodiesterase activity of RNase T2 in the sense of product inhibition [116].

In addition to RNase T2, lysosomes likely contain other, not yet fully characterized RNases [117]. Especially members of the RNase A family have recently been shown to play a role in the lysosomal turnover of immunogenic nucleic acids [118–120]. In humans, the RNase A family consists of 13 genes (RNASE1 to RNASE13), which are homologs to the bovine pancreatic RNase A and likely evolved by tandem gene duplication as all of them are localized inside a single gene cluster on chromosome 14 [121, 122]. Although mainly known as secreted proteins with antimicrobial activity, recent studies revealed that RNase 2 and RNase 6 also localize to lysosomes by immunofluorescence microscopy following ectopic expression [119, 123]. Notably, in independent studies, RNase 6 was shown to contain mannose 6-phosphate, a typical modification found on soluble lysosomal proteins [123, 124]. Members of the RNase A family typically act as endonucleases, cleaving preferably after pyrimidine nucleotides via a nucleotide-2′,3′-cyclic phosphate intermediate [75]. The RNase A family is typically related to host-defense mechanisms, either directly reducing the infectivity of viruses or bacteria or modulating the immune response against pathogens [125–127]. Their reaction mechanism is similar to RNase T2 and proceeds via a transesterification reaction and a 2′,3′-cyclic phosphate as intermediate, which is further hydrolyzed to a 3′-monophosphate and a 5′-OH group at the newly generated ends [128, 129]. Consequently, chemical modifications of the 2′-OH group, like 2′-O-methylation, typically render oligoribonucleotides resistant to both cleavage by RNase T2, as well as by members of the RNase A family, and seem useful in designing synthetic agonists to endolysosomal TLRs [118–120].

Recently, the role of RNase 2 in the degradation of pathogen-derived RNA inside the lysosome has been highlighted (Fig. 4) [119]. Like other members of the RNase A family, RNase 2 cleaves nucleic acids preferentially after a pyrimidine followed by a purine, with UpA being the most efficient ligand [75, 130, 131, 132]. RNase 2 and RNase T2 synergistically degrade single-stranded RNA to shorter oligoribonucleotides, which are then able to stimulate endolysosomal TLRs. While RNase 2 preferably cleaves at the 3′-position of uridine nucleotides, RNase T2 hydrolyzes the 5′-phosphodiester bond before pyrimidines, leading to the release of uridine mononucleotides by their combined action [105, 119], which is a known ligand for TLR8 in human cells [133]. Along with pathogen RNA, RNase 2 is responsible for the cleavage of endogenous tRNAs between U/CpA nucleotides in the anticodon loop [132]. Outside the lysosome, RNase 2 is known as eosinophil-derived neurotoxin, a protein secreted from eosinophil granulocytes with direct antiviral activity against RNA viruses like the human immunodeficiency virus (HIV) [134], the respiratory syncytial virus (RSV) or parainfluenza viruses [135]. This effect seems to depend on its RNase activity since inactivating mutations also reduce the antiviral activity [136]. It is currently hypothesized that a C-terminal sequence of cationic amino acids helps in disrupting the viral envelope, enabling the contact between RNase 2 and the viral genome inside [135]. Alternatively, it is conceivable that RNase 2 released by eosinophils is internalized into neighboring cells, degrading RNA that is exposed by other hydrolases, degrading the capsid proteins and envelope [125, 126].

Although the first discoveries of RNase 6 in the lysosomal proteome date back to the 2000s [123, 124], its role in lysosomal nucleic acid degradation has been unclear until recently. RNase 6 assists in maintaining an efficient TLR8 response to bacterial RNA in human monocytes by generating immunostimulatory RNA fragments, similar to RNase 2 [118]. Human monocytes deficient for RNase 6 show reduced cytokine production upon contact with bacterial RNA, and Rnase6^−/− knockout mice are more susceptible to urinary tract infections with uropathogenic Escherichia coli [118, 137]. This raises the question of whether the antimicrobial activity of RNase 6 is a direct effect of its enzymatic action, as hypothesized for RNase 2, or rather indirectly mediated by the induced immune response. A combination of both mechanisms seems likely, especially since an antimicrobial effect of recombinant RNase 6 on different bacterial strains has also been shown in vitro [138]. Additional nonenzymatic effects might also contribute to its bactericidal activity, as positive charges around the N-terminal part of RNase 6 have been shown to destabilize the bacterial plasma membrane and promote aggregation of bacterial cells [139]. A common sequence variation reducing the cationic surface charge of RNase 6 also lowers its bactericidal activity, probably due to a reduced affinity to lipopolysaccharides on the bacterial membrane [140]. Similar to RNase 2, RNase 6 has also been shown to reduce the infectivity of HIV-1 virions toward T lymphocytes in vitro [134].

The human ortholog of bovine pancreatic RNase A is RNase 1, originally described as a pancreatic enzyme secreted from the exocrine acinus cells to degrade ingested RNA and an additional potential candidate for a lysosomal RNase [141]. RNase 1 was found, similar to RNase 6, in mannose 6-phosphate affinity chromatography purification experiments of human urine and brain [142, 143], making it a prime candidate for a resident lysosomal protein. Apart from the pancreatic acinus cells, it is also expressed and secreted by endothelial cells and involved in the clearance of RNA and DNA-RNA hybrids in the bloodstream and in the differentiation of immature dendritic cells [126, 144]. It also showed direct antiviral activity against HIV-1, reducing virus release from infected T cells in vitro [145]. Extracellular RNase 1 has been shown to enter cells via endocytosis, ending up in endo- and finally lysosomes, although the neutral pH optimum of human RNase 1 makes a role in these rather acidic compartments unlikely [146]. Murine RNase 1, however, shows a slightly more acidic pH optimum around pH 6.4 [146], theoretically enabling the degradation of RNA inside early endosomes, although cellular evidence on this activity is lacking so far. More work is needed to clarify the possible physiological functions of RNase 1 in lysosomes. Finally, we identified RNase 4 in liver lysosomes as an additional possible candidate (unpublished). RNase 4 is expressed in the kidney, the urinary tract, monocytes, and intestinal epithelial cells [147–149] and shows a strong cleavage preference for uridine nucleotides [150]. All these RNases are found as secreted proteins in the blood and body fluids, and it is often impossible to assign the localization of a soluble protein of the secretory pathways as a “secreted protein” or a “resident lysosomal protein.” If they have, in addition to the extracellular function(s), a physiologically relevant function in lysosomes remains to be determined. In summary, obviously, more work is needed to comprehend the function and substrate spectrum of lysosomal RNases fully.

To what extent both RNase T2 and RNase A family members degrade RNA to small oligoribonucleotides or even completely to mononucleotides in vivo remains unknown. There is, however, evidence that the 5′-exonucleases PLD3 and PLD4 already mentioned above also play a role in the complete decomposition of shorter ssRNA fragments released by endonucleases to 3′-mononucleotides (Table 2). Interestingly, mice defective for both Pld3 and Pld4 genes accumulate short ssRNA fragments activating TLR7 and the STING pathway [101], while they are at the same time required to generate both guanosine derivatives and short RNA ligands recognized by TLR7 [85].

The dephosphorylation of nucleotides and the export of catabolic end products

Both the endolysosomal RNase(s) and DNases generate 3′-phosphate-mononucleotides as catabolic products (Fig. 1). Given that nucleotides cannot be exported from the lysosome, they must be enzymatically dephosphorylated by one or more lysosomal phosphatase(s). Surprisingly, the dephosphorylation of nucleotides in mammalian cells is mostly unexplored.

Experimental data on the physiologically relevant phosphatase(s) mediating the lysosomal dephosphorylation of nucleotides and their specificity toward different nucleotides (e.g., containing different bases, 3′ versus 5′ phosphate groups, Fig. 5) are scarce. In the yeast vacuole, the non-specific alkaline phosphatase Pho8 has been shown to dephosphorylate, among others, 3′-nucleotides generated from RNA by the T2-type RNase Rny1, especially under nitrogen starvation conditions, which lead to upregulation of both Pho8 and Rny1 expression [151]. Deletion of the pho8 gene results in an increase in cellular 3′- and 5′-mononucleotide levels upon nitrogen starvation. Mammalian lysosomes, however, contain at least two well-defined phosphatases: “lysosomal acid phosphatase,” coded by ACP2, and the “tartrate-resistant acid phosphatase” (“TRAP,” coded by ACP5) (Table 3). Even though early in vitro experiments showed that lysosomal acid phosphatases can dephosphorylate nucleotides [152], the contributions of each phosphatase gene product remained mainly unclear. The knockout of Acp2 in mice leads to a lysosomal storage phenotype that is restricted to specific cell types. However, the nature of the storage material is unknown [153]. Acp5 knockout mice show no signs of lysosomal storage but have disordered macrophage inflammatory responses [154]. The cause of this inflammatory condition remained unclear. Acp2/Acp5 double knockout mice show lysosomal storage in many cell types, including macrophages and hepatocytes, suggesting at least partially redundant substrates. Interestingly, Acp2/Acp5 knockout mice show severe splenomegaly suggestive of immune dysregulation [155]. In humans, mutations in the ACP5 gene have been found causative for Spondyloenchondrodysplasia (SPENCD) with or without immune dysregulation, a rare hereditary disease following biallelic missense mutations resulting in a complete loss of ACP5 protein [156, 157]. The disease typically manifests in dysplastic bone lesions and a type I interferonopathy resulting in autoimmune syndromes like autoimmune thrombocytopenia, hemolytic anemia, systemic lupus erythematosus, or juvenile idiopathic arthritis, and, at the same time, immunodeficiency with recurrent infections [158]. Some cases also show neurological features, including spasticity or developmental delay, renal disease, or endocrine abnormalities. While the skeletal abnormalities have been linked to ACP5’s contribution to the degradation of osteopontin, a highly phosphorylated bone matrix protein [157, 159], the immunologic phenotype has not been explained in mechanistic details, although osteopontin has also been described as cytokine in immune reactions against pathogens [160]. Given the remarkable expression of ACP5 in antigen-presenting phagocytes like dendritic cells and alveolar or splenic macrophages, an involvement of the TLR axis also seems conceivable [161].

Figure 5.

The lysosomal degradation of both DNA and RNA converges in the dephosphorylation of the 3′-mononucleotide building blocks (3′-NMPs and 3′-dNMPs) by an acid phosphatase, of which there are two known in lysosomes: ACP2 and ACP5. The resulting nucleosides, potential ligands for TLR8, are then exported via the transport protein ENT3, while a lysosomal phosphate transporter has only been classified functionally, and the coding gene is not known to date. Cyclic nucleotides (2′,3′-cNMPs) generated as byproducts by RNase T2 and the RNase A family members are degraded into mononucleotides by RNase T2.

Table 3.

Overview on putative and confirmed lysosomal phosphatases and phosphodiesterases, their encoding genes, tissue distribution, and enzymatic properties

Protein name	Gene	Tissue distribution	Substrate	Product
Lysosomal acid phosphatase	ACP2	Ubiquitous	Phosphoesters	Free phosphate
Prostatic acid phosphatase (Transmembrane isoform)	ACP3	Ubiquitous	Phosphoesters	Free phosphate
Tartrate-resistant acid phosphatase	ACP5	Immune cells, osteoclasts	Phosphoesters (shown for ATP)	Free phosphate
RNase T2	RNASET2	Ubiquitous	Nucleoside-2′,3′-cyclic phosphates	Nucleoside-3′-phosphates
Ectonucleotide pyrophosphatase/phosphodiesterase 4	ENPP4	Endothelium	Dinucleotide polyphosphates (Ap3A, Ap4A)	Nucleotides
Ectonucleotide pyrophosphatase/phosphodiesterase 5	ENPP5	Unknown	Dinucleotides (NAD, ADP-ribose), UDP-glucose	Nucleotides

Furthermore, a transcript variant of “prostatic acid phosphatase” (coded by ACP3), yielding a membrane-bound protein, has recently been discovered as being expressed ubiquitously and localized to lysosomes. This isoform was shown to be identical to the thiamine monophosphatase (TMPase), an ectonucleotidase highly expressed on the cell surface of dorsal root ganglion cells. It was shown that ACP3 could dephosphorylate 5′-AMP but barely 5′-ADP or 5′-ATP, thereby regulating extracellular adenosine levels to modulate neuronal activity [162–164]. However, its role in lysosomal phosphate metabolism is largely unknown to date, and whether it influences nucleotide catabolism still needs to be addressed (Fig. 5).

While 3′-mononucleotides are the main product of lysosomal nucleases, the exact composition of the lysosomal nucleotide metabolome has never been thoroughly described. Especially, the concentration and degradation of 2′,3′-cyclic nucleotides, resulting from the combined action of RNase T2 and PLD exonucleases, is unclear, as no specific lysosomal phosphodiesterases are described. In vitro experiments have shown 2′,3′-phosphodiesterase activity for RNase T2 [116]. Other nucleotide derivatives like NAD⁺, FAD, and FMN will most likely also reach lysosomes, especially during the autophagic degradation of mitochondria. Whether they are fully degraded to nucleosides, as is the case for nucleic acids, or exported by specialized transport systems is largely unknown. Two members of the ectonucleotide pyrophosphatase/phosphodiesterase (ENPP) family, ENPP4 and ENPP5, have been identified in proteome-wide subcellular proteomics approaches in lysosomal fractions [165, 166]. Both are type-I-transmembrane proteins with their catalytically active, zinc-dependent phosphodiesterase domain facing the luminal side [167]. ENPP4’s principal substrates in vivo seem to be diadenosine polyphosphates, while the release of pyrophosphate and 5′-AMP from ATP has only been shown in vitro and seems to be negligible due to a low affinity compared to other members of the ENPP family [168]. ENPP5 shows a unique amino acid substitution from phenylalanine to tyrosine in its active site, sterically hindering the binding of nucleotides to the phosphodiesterase domain [169], but it has been shown to cleave the pyrophosphate bonds present in NAD, ADP-ribose, or UDP-glucose [167]. Furthermore, its luminal domain can be released as a soluble enzyme upon cleavage by the β-amyloid cleaving enzyme BACE1 [170]. Both enzymes, however, have mainly been described as extracellularly acting hydrolases and are not known to release free nucleosides for potential export from lysosomes (Table 3).

After dephosphorylation, nucleosides and phosphate are exported by specific transporters. While the phosphate transporter still remains enigmatic and was only characterized biochemically in human fibroblasts [171], nucleosides are exported via the Equilibrative Nucleotide Transporter 3 (ENT3, coded by the SLC29A3 gene) [172, 173]. The deficiency of ENT3 causes an expanding spectrum of human genetic disorders [e.g., H syndrome, Pigmented Hypertrichosis with Insulin-dependent Diabetes Mellitus (PHID) syndrome, and sinus histiocytosis with massive lymphadenopathy (SHML) syndrome] in human patients [174–176] and histiocytosis in mice accompanied by spontaneous splenomegaly, accumulation of nucleosides in lysosomes, and congestion of apoptotic nuclei in macrophages [173]. This phenotype has been shown to rely on the overstimulation of TLR7 and TLR8 by accumulating nucleosides, which stimulated the proliferation of monocytes, ultimately leading to histiocytosis [177]. The highest transport activity can be observed for ENT3 at pH 5.5, highlighting its role in the lysosomal export of nucleosides. ENT3 itself does not show a preference for any specific nucleoside and is even able to transport free nucleobases and nucleoside analogs used as antiviral or cytostatic drugs [178]. Adenosine efflux by ENT3 seems to play a role in regulating autophagy via the AMP-activated protein kinase (AMPK) and mechanistic target of rapamycin (mTOR) [179].

Endolysosomal nucleic acid degradation as a critical regulator of innate immune response receptors

The endolysosomal system is equipped with several TLRs, a well-studied family of pattern recognition receptors for sensing nucleic acids and their degradation products. In humans, the four TLRs 3, 7, 8, and 9 are located inside the endolysosomal membrane, with their ligand binding domains facing the lumen [3, 180]. Mice additionally express TLR13, which is not present in humans [181]. The proper function of all endolysosomal TLRs relies on the transport protein UNC93B, which associates with their transmembrane domain, promoting their direct transport from the Golgi apparatus to endosomes [182].

Specifically, TLR3 detects long dsRNA molecules, as shown for the synthetic polyinosinic-polycytidinic acid, mimicking double-stranded RNA [183]. dsRNA molecules need to be ≥40 bp in length in order to be detectable by TLR3, with longer strands being bound more efficiently [184]. Thereby, TLR3 does not only form homodimers as it is common among the TLR family; several of these homodimers furthermore cluster along longer dsRNA molecules, binding them in a cooperative manner [185, 186]. RNase T2 is likely responsible for the degradation of TLR3 ligands, as RNase T2-deficient macrophages show an activation of the TLR3 signaling pathways instead of TLR7, most likely due to the accumulation of long dsRNA molecules [115]. Whether other lysosomal RNases contribute to the processing or degradation of TLR3 ligands remains to be demonstrated. Although dsRNA is typically considered a form of viral genome, TLR3’s exact role in the response to viral infections is unclear since the absence of TLR3 does not increase the susceptibility toward RNA virus infections [187, 188]. Recent studies rather highlight its role in the differentiation and activation of osteoclasts, as dsRNA endocytosed by osteoblasts induces a TLR3-dependent expression and secretion of the receptor activator of NF-κB ligand (RANKL) and prostaglandin E₂, known to be potent stimulators for osteoclast differentiation via the NF-κB and MAPK pathways [189, 190]. A recent study on breast tumor cells demonstrated that TLR3 stimulation can trigger the production of type III interferon by conventional dendritic cells, activating T lymphocytes, which leads to a T_H1-polarized antitumor immune response [191].

TLR7 and TLR8 both detect short ssRNA molecules as degradation products of lysosomal RNases, especially RNase T2 [85, 105, 119]. Both receptors have two binding pockets for ligands: site 1 recognizes short oligoribonucleotides, while site 2 binds to mononucleotides and nucleosides, both of which are products of lysosomal RNases and PLD exonucleases [85]. The exact ligands, however, differ between both TLRs and in different species: TLR8 seems to detect uridine nucleosides, whereas TLR7 shows a strong activation upon binding to guanosine analogs, especially 2′,3′-cyclic GMP [133, 192, 193]. Regarding site one ligands, TLR7 prefers oligoribonucleotides carrying subsequent uridine nucleotides, while TLR8 slightly prefers short, GU-motive-containing oligonucleotides [133, 193–195]. The second binding pocket is also engaged by synthetic imidazoquinoline compounds, which act as strong agonists of TLR7 and 8 [193, 196, 197]. A genetic variation in site 2 of the TLR7 protein, enhancing the binding toward its 2′,3′-cyclic GMP ligand, has been shown to cause systemic lupus erythematosus in humans, highlighting the need for balanced control of TLR activation in order to prevent autoimmune dysregulation [198].

TLR9 is the only human TLR recognizing DNA ligands, specifically ssDNA carrying unmethylated CpG dinucleotide motifs, ideally ~10 nucleotides in length [199]. TLR9 ligands have been sorted into three categories differing in their structure and the resulting signaling pathways: Class A ODNs, featuring a palindromic sequence around a central CpG motive, leading to the formation of a hairpin loop as a secondary structure. They typically lead to a strong induction of type I interferon signaling but only weakly stimulate the NF-κB pathway. Class B ODNs, on the other hand, are linear ODNs including one or more CpG motives and stimulate mainly NF-κB signaling with only a weak interferon response. Finally, class C ODNs combine the structural elements of both previous types and lead to an intermediate effect on both NF-κB and IFN signaling [200]. Short ssODNs are typically generated from longer dsDNA upon endonucleolytic digest by DNase II [199] and degraded by the PLD exonucleases PLD3 and PLD4 [81]. CpG motifs in nuclear DNA are typically methylated, thereby preventing activation by endogenous nucleic acids. Mitochondrial DNA, however, does not contain 5-methylcytosine and can activate TLR9 in the context of mitophagy [201]. Another—less thoroughly studied—TLR9 ligand is deoxyinosine-containing CpI motives, naturally occurring after the deamination of adenosine bases by ADA2 [102]. Although CpI motives are weaker TLR9 agonists, they have been shown to introduce similar cellular responses in B cells but seem much less active in pDCs [202]. Their exact relevance and contribution to lysosomal ODN sensing in vivo have not been determined to date.

TLR13 is absent in humans but expressed in rodents, especially in macrophages and conventional dendritic cells [181]. It specifically detects the decanucleotide sequence “CGGAAAGACC” from the bacterial 23s ribosomal RNA, contributing to the detection of both Gram-positive and Gram-negative bacteria [203, 204]. Early studies have already demonstrated that the in vitro degradation of bacterial 23s RNA with RNases prevents its detection by TLR13 [204], suggesting that RNase activity inside the lysosome can limit its activation by degrading the specific ligand. In fact, recent studies revealed that the autoinflammatory phenotype in Rnaset2^−/− mice depends on the activation of TLR13 by the accumulation of long rRNA molecules [112]. Its stimulation thereby leads to the proliferation of macrophages, especially in the spleen and liver, and the production of pro-inflammatory cytokines. Interestingly, another study reported that this phenotype was consistent in germ-free mice, hypothesizing that TLR13 might additionally recognize a so-far unknown endogenous ligand [111]. However, in the case of human RNase T2 deficiency, also displaying an interferon-driven leukoencephalopathy [110, 205], the underlying mechanism must be different since TLR13 has no known paralog in humans.

Different sets of nucleic acid innate immune response receptors sense DNA in the cytoplasm or endolysosomal lumen. While TLRs sense nucleic acids in the lumen of endolysosomes, several pattern recognition receptors are localized directly in the cytosol. Upon binding of cytosolic DNA, the cyclic GAMP synthase (cGAS) produces the second messenger cGAMP, which further activates the “stimulator of interferon genes” (STING), inducing a type I interferon-driven immune response [206]. A contribution of cytosolic pathways to a pro-inflammatory reaction has been described for several defects in lysosomal nucleic acid degradation, as the accumulation of nucleic acids inside the lysosome can lead to a leakage of these nucleic acids into the cytosol. The exact nature of this leakage, i.e., whether it is mediated by a transport system like the SIDT family or occurring due to physical defects in the lysosomal membrane, needs further clarification. At least for dsRNA, a translocation by SIDT1 and SIDT2 to the cytosol has been described where it can activate RLRs [39, 40]. Defects in the DNase II–PLD degradation pathway also result in DNA leakage from lysosomes into the cytosol, where it can activate the cGAS–STING system, leading to a proinflammatory response [65, 66, 67, 95]. In Dnase2^−/− mice, the contribution of the “absent in melanoma 2” (AIM2) inflammasome to the cytosolic recognition of leaked DNA from lysosomes has been shown [70]. Interestingly, STING can be activated independently of cGAS, and this pathway is directly associated with lysosomal function: STING interacts with the lysosomal membrane protein Niemann–Pick type C1 (NPC1) as a trafficking cofactor [207]. NPC1 interacts with STING and recruits it to the lysosome for degradation. How far this direct link of STING and lysosomal function is relevant to DNA leakage-mediated activation needs further investigation. In summary, while physically separated, cytosolic and luminal nucleic acid sensing are closely interconnected.

Concluding remarks

Although the lysosome is classically regarded as the central organelle for the breakdown of macromolecules, our understanding of many pathways remains incomplete, and lysosomal proteins discovered more than a century ago still harbor unexplored functions in metabolic or signaling pathways. Especially during the past two decades, numerous advances have been made to help us understand the lysosomal breakdown and recycling of nucleic acids of both endo- and exogenous origin, as well as their significance for host immune defence and autoimmune diseases. Perturbations in the lysosomal nucleic acid catabolism are not only responsible for rare genetic disorders but are also involved in the pathogenesis of common autoimmune diseases. Understanding the underlying mechanisms is therefore critical for the understanding of these diseases and the potential development of new therapeutic strategies. While the signaling pathways for the recognition of nucleic acids are complex, their degradation follows fairly simple principles: long nucleic acid strands are cleaved into shorter oligonucleotides by endonucleases before being completely degraded into mononucleotides by exonucleases. An unidentified nucleotidase dephosphorylates the nucleotides to nucleosides prior to their export to the cytoplasm for further recycling or complete catabolic degradation. Several questions are, however ,still open for future research. Especially, the transport mechanisms of nucleic acids between the cytosol and the lysosomal lumen need further investigation to determine the involved transport proteins. This question is not only of academic interest but might prove valuable for the development of oligonucleotide-based drugs and vaccines, which are increasingly applied in clinical trials and even approved medications. Understanding their transport and degradation within cells could help optimize delivery methods, as well as drug stability, and minimize off-target effects. Another promising point to be addressed in future studies is the contribution of the different nucleases and nucleotidases and their potential as drug targets in immunomodulatory therapies. The pharmacological inhibition of these processes could strengthen the physiological immune response against pathogens or even cancer cells, promoting their clearance by the host immune system. Conversely, an induction of nucleic acid degradation could alleviate excessive immune reactions and provide novel targets for the therapy of autoimmune diseases triggered by various endolysosomal and cytosolic signalling pathways. Closing these knowledge gaps is therefore of major importance to enhance our understanding of immune reactions, autoimmune diseases, and the development of novel therapies.

Conflict of interest

The authors declare no conflict of interest.

Funding

This work was in part supported by the Deutsche Forschungsgemeinschaft (DFG) to M.D. (DA 1785/1-1 and DA 1785/2-1; FOR2625).

Data availability

No new data were generated or analyzed in support of this research.