RNA sensing via the RIG‐I‐like receptor LGP2 is essential for the induction of a type I IFN response in ADAR1 deficiency

Introduction

Receptors of the innate immune system continuously sample the intra- and extracellular environment for signs of an ongoing infection. Viral infections can be detected through the presence of viral nucleic acids in the cytosol of infected cells (Goubau et al, 2013; Rehwinkel & Gack, 2020). Upon encountering viral DNA or RNA, cytosolic nucleic acid sensors, most notably cGAS or RIG-I-like receptors (RLRs), respectively, initiate an antiviral type I interferon (IFN) response (Ablasser & Hur, 2020; Rehwinkel & Gack, 2020). While a type I IFN response is important for defense against viral infections, its inadvertent activation by self-derived nucleic acids induces a sterile inflammatory response that causes immunopathology (Schlee & Hartmann, 2016). Cellular mechanisms that ensure the discrimination between foreign and endogenous nucleic acids are therefore critical to avoid autoinflammation (Schlee & Hartmann, 2016).

RNA modification by the enzyme Adenosine Deaminase Acting on RNA (ADAR1) constitutes an important mechanism by which cells ensure self/nonself RNA discrimination (Heraud-Farlow & Walkley, 2016; Uggenti et al, 2019; Quin et al, 2021). Through modification of endogenous RNA, ADAR1 prevents the activation of cytosolic RNA sensors, including RLRs, by cellular RNA molecules and the unwanted induction of an antiviral type I IFN response (Heraud-Farlow & Walkley, 2016; Uggenti et al, 2019; Quin et al, 2021). The importance of ADAR1 is highlighted by the severe consequences of ADAR1 mutations in patients with Aicardi-Goutières Syndrome (AGS) (Rice et al, 2012; Rodero & Crow, 2016). This rare genetic disorder belongs to the spectrum of type I interferonopathies, which are characterized by the constitutive induction of an antiviral type I IFN response in the absence of an infection (Rodero & Crow, 2016). The autoinflammatory condition that arises from inherited ADAR1 mutations leads to severe (neuro)pathological features (Livingston & Crow, 2016; Rice et al, 2017). Notably, ADAR1 has also emerged as an attractive target for novel immunotherapeutic approaches in cancer (Bhate et al, 2019). A subset of tumor cells is sensitive to growth arrest upon knockdown or knockout of ADAR1, both in vivo and in vitro (Gannon et al, 2018; Ishizuka et al, 2019; Liu et al, 2019). In addition, intratumoral loss of ADAR1 increases sensitivity to treatment with immune checkpoint inhibitors and overcomes resistance to such inhibitors in vivo (Ishizuka et al, 2019). Finally, depletion of ADAR1 in cancer cells potentiates the efficacy of epigenetic therapy and increases type I IFN induction (Mehdipour et al, 2020). Understanding the precise mechanism by which ADAR1 (dys)function impacts on innate immunity is therefore essential to better understand its disease-causing role in interferonopathies as well as its therapeutic potential in cancer.

ADAR1 exists as two isoforms. The nuclear p110 isoform is constitutively expressed, while the p150 isoform is induced by type I IFN receptor signaling and resides primarily in the cytoplasm (Heraud-Farlow & Walkley, 2016; Quin et al, 2021). Both isoforms act on base-paired RNA to deaminate adenosines and convert them to inosines. A-to-I editing is among the most widespread base modifications in mammals. Besides site-specific A-to-I editing, which can alter open reading frames, miRNA seed sequences or RNA splice sites, there is also highly promiscuous and abundant editing of base-paired RNAs with long regions of high complementarity such as transcripts spanning inverted repeat Alu (IR-Alu) elements (Eisenberg & Levanon, 2018). Without editing, such base-paired structures would resemble double-stranded RNAs (dsRNAs) that are abundantly found in cells infected with some viruses. Unedited self RNA molecules are therefore prone to activate antiviral innate immune mechanisms, such as protein kinase R (PKR) (Chung et al, 2018), OAS1/RNase L (Li et al, 2017), and the RLR pathway (Mannion et al, 2014; Liddicoat et al, 2015; Pestal et al, 2015). While activation of PKR and OAS/RNase L causes translational shutdown and cell death, RLR engagement initiates the type I IFN response.

The link between ADAR1 editing and RLR activation was first demonstrated in a series of mouse studies. In mice, genetic loss of ADAR1 p110 and p150, p150 alone, or knock-in of an editing-deficient ADAR1 mutant (AdarE861A/E861A) results in embryonic lethality, fetal liver disintegration, hematopoiesis defects, and an elevated type I IFN signature (Hartner et al, 2004; Wang et al, 2004; Ward et al, 2011; Liddicoat et al, 2015). The embryonic lethality of ADAR1 null or editing-deficient mice can be rescued by the concurrent deletion of the RLR family member melanoma differentiation-associated protein 5 (MDA5) or the downstream signaling hub MAVS (mitochondrial antiviral signaling, also known as VISA, Cardif, IPS-1), but not another RLR, retinoic-acid-inducible gene I (RIG-I) (Mannion et al, 2014; Liddicoat et al, 2015; Pestal et al, 2015; Heraud-Farlow & Walkley, 2016). In addition, loss of MDA5 or MAVS also eliminates the type I IFN signature in these mice. These observations indicate that unedited RNA mediates its immunostimulatory effects via MDA5 and MAVS and that the type I IFN response plays an important role in the immunopathology caused by loss of ADAR1.

MDA5 normally detects RNA from certain viral species, such as Picornaviridae (Dias Junior et al, 2019). It senses long stretches of dsRNA or base-paired single-stranded RNA, on which it oligomerizes to form filamentous structures (Dias Junior et al, 2019; Rehwinkel & Gack, 2020). In contrast, RIG-I is activated by 5′ di- or triphosphate moieties at the base-paired extremities of certain viral RNA species (Goubau et al, 2013; Rehwinkel & Gack, 2020). Activation of RIG-I or MDA5 by their respective RNA substrates leads to conformational changes that allow their N-terminal CARD domains to interact with the CARD domains of the adaptor MAVS (Sohn & Hur, 2016). This, in turn, leads to MAVS activation and subsequent phosphorylation and activation of the transcription factors IRF3 and NF-κB, which mediate the transcription of type I IFNs (most notably IFN-α subtypes and IFN-β), type III IFNs, and other pro-inflammatory cytokines (Goubau et al, 2013; Rehwinkel & Gack, 2020). Upon secretion, type I IFNs activate the IFN-α/β receptor (IFNAR) and induce JAK-STAT signaling, which results in the transcriptional upregulation of hundreds of IFN-stimulated genes (ISGs), which establish an antiviral state (Schoggins et al, 2011; Schneider et al, 2014). Laboratory of genetics and physiology 2 (LGP2) is the third and least well-understood member of the RLR family. LGP2 lacks the N-terminal CARD domains and is therefore not able to signal via MAVS (Rodriguez et al, 2014; Rehwinkel & Gack, 2020). Instead, LGP2 modulates the function of RIG-I and MDA5 during viral infection. While LGP2 suppresses RIG-I signaling, it synergizes with MDA5 to potentiate the sensing of certain RNA viruses (Rodriguez et al, 2014; Rehwinkel & Gack, 2020). Akin to MDA5-deficient mice, LGP2-knockout mice display increased sensitivity to infection with encephalomyocarditis virus (EMCV), a member of the Picornaviridae family (Venkataraman et al, 2007; Satoh et al, 2010). Mechanistically, LGP2 is incorporated into MDA5 filaments and enhances the interaction between MDA5 and RNA, thereby increasing the rate of MDA5 filament formation (Bruns et al, 2014; Duic et al, 2020). Simultaneously, LGP2 enhances the dissociation of MDA5 filaments in an ATP-dependent manner and generates shorter filaments that have greater agonistic activity than longer filaments (Bruns et al, 2014; Duic et al, 2020). Structural studies demonstrated that LGP2 primarily binds the ends of dsRNA, although it can also coat dsRNA in a similar fashion as MDA5 (Uchikawa et al, 2016). Thus, LGP2 promotes rapid MDA5-dsRNA filament formation yet yields shorter filaments, ultimately leading to enhanced downstream signaling and an increased type I IFN response.

LGP2 also impacts on type I IFN responses through alternative routes that are independent from its role as typical RNA sensor. Wild-type LGP2 and mutants that fail to hydrolyze ATP or bind RNA interact with MAVS at steady state and block the interaction between RIG-I and MAVS, thereby limiting RIG-I-mediated MAVS activation (Esser-Nobis et al, 2020). Upon stimulation with the dsRNA mimic poly(I:C), LGP2 releases MAVS for interaction with RIG-I (Esser-Nobis et al, 2020). LGP2 additionally limits RIG-I signaling and potentiates MDA5 signaling by a direct protein–protein interaction with the dsRNA-binding protein PACT (Sanchez David et al, 2019). Furthermore, LGP2 inhibits Dicer-mediated processing of dsRNA (Van der Veen et al, 2018), perhaps to preserve dsRNA substrates for the full-blown activation of the type I IFN response. Conversely, LGP2 may negatively regulate the antiviral type I IFN response by associating and interfering with the function of TRAF ubiquitin ligases, in a manner that is independent of ATP hydrolysis or RNA binding (Parisien et al, 2018). Finally, LGP2 controls CD8+ T cell survival and fitness during West Nile virus and lymphocytic choriomeningitis virus infection in mice, pointing to cell type-specific functions (Suthar et al, 2012).

Both MDA5 and RIG-I can bind and be activated by endogenous RNA in various contexts (Dias Junior et al, 2019; Streicher & Jouvenet, 2019; Stok et al, 2020). LGP2 has predominantly been studied upon viral infection or mimics thereof. A recent study demonstrated that mice bearing a mutation in the Zα domain of ADAR1 that is involved in binding to dsRNA in its unusual Z-conformation (Z-RNA) suffer from postnatal growth retardation and mortality and have a mild type I IFN signature, which can be reverted by crossing these mice with MDA5, MAVS, PKR, as well as LGP2-knockout mice (Maurano et al, 2021). The extent to which LGP2 is required for type I IFN induction in response to unedited RNA species more broadly (aside from Z-RNA), the molecular mechanism that is involved, and whether it is required in humans is unclear.

Here, we investigated the role of human LGP2 in induction of type I IFNs caused by ADAR1 deficiency. Using various genetic approaches and model systems, we demonstrate that LGP2 is essential for this induction, in a manner that involves its classical function as RNA sensor. Importantly, we further demonstrate that LGP2 is required both for sensing of unedited RNA and for reduced cell growth upon loss of ADAR1 in tumor cells. Finally, treatment of ADAR1-depleted tumor cells with epigenetic repressors, a promising strategy for cancer therapy, potentiates the type I IFN response in an LGP2-dependent manner. Our findings provide molecular insight into the effector mechanisms that are engaged upon dysregulation of ADAR1, with important clinical implications for the field of interferonopathies as well as cancer.

Results Human LGP2 is required for the induction of a type I IFN response upon depletion of ADAR1

To investigate the role of human RLRs in the induction of type I IFN caused by the absence of ADAR1, we first knocked out RIG-I, MDA5, or LGP2 in the human monocytic leukemia cell line THP-1 using CRISPR/Cas9-mediated genome engineering. Correct gene ablation was confirmed by immunoblotting cells treated with recombinant type I IFN to upregulate the expression of RIG-I, MDA5, and LGP2, which are encoded by ISGs themselves (Fig 1A). Intact type I IFN receptor signaling was verified by monitoring ISG60 upregulation (Fig 1A). For each RLR, two knockout clones were differentiated into macrophages and transfected with siRNAs targeting both isoforms of ADAR1. Despite the modest efficiency of the knockdown at the time point chosen for analysis (Figs 1B and EV1A), we observed a clear upregulation of transcripts encoding IFN-β and the ISG IFIT1 in wild-type cells, indicative of type I IFN induction (Fig 1B). Notably, loss of LGP2 completely abrogated type I IFN induction and signaling upon ADAR1 depletion (Fig 1B). Loss of MDA5, but not RIG-I, also interfered with the type I IFN response, consistent with published literature (Heraud-Farlow & Walkley, 2016). Note that throughout the manuscript ADAR1 knockdown efficiency is monitored through measurement of p110 expression levels, as analysis of the p150 isoform underestimates knockdown efficiency due to its IFN-inducible nature. Consistent with these observations, siRNA-mediated depletion of ADAR1 in primary human monocyte-derived macrophages induced a type I IFN response (monitored by IFN-β, IFIT1, and ISG15 transcript levels), which was markedly reduced upon co-depletion of LGP2 (Fig 1C). Together, these data indicate that, besides MDA5, expression of LGP2 is crucial for the induction of a type I IFN response in ADAR1 deficiency.

Details are in the caption following the image

Figure 1. Human LGP2 is essential for the induction of a type I IFN response upon depletion of ADAR1

THP-1 monocytes were genetically engineered to knockout RIG-I, MDA5, or LGP2 using CRISPR/Cas9. Cells were differentiated toward macrophage-like cells using PMA and treated with recombinant type I IFN to upregulate RLR expression. Correct gene editing and intact type I IFN responsiveness were validated by SDS-PAGE and immunoblotting using the indicated antibodies (n = 3). *, nonspecific band. Cells generated in (A) were differentiated using PMA and transfected with a control siRNA (siCtrl) or an ADAR1-targeting siRNA (siADAR1). The type I IFN response was monitored 56 h post-transfection by RT-qPCR analysis to determine IFN-β and IFIT1 transcript expression, normalized to a housekeeping gene (ACTB). ADAR1 knockdown efficiency was monitored by ADAR1 p110 expression, normalized to ACTB, and displayed relative to siCtrl. Data are means ± s.d. from a representative of three biological replicate experiments. Primary human monocyte-derived macrophages were transfected with the indicated siRNAs. Cells were harvested 96 h post-transfection and RT-qPCR analysis was used to monitor the type I IFN response (IFN-β, IFIT1, and ISG15 transcripts) and knockdown efficiency of ADAR1 and LGP2 (DHX58). All transcripts were normalized to ACTB. Data from two independent donors (denoted with distinct symbols) are shown with mean ± s.d. HEK293 cells were genetically engineered to knockout RIG-I, MDA5, or both, and subsequently subjected to retroviral transduction to stably express FLAG-LGP2 or an empty vector (EV). Correct gene editing and intact type I IFN responsiveness were validated by SDS-PAGE and immunoblotting using the indicated antibodies (n = 2). *, nonspecific band. Cells generated in (D) were transfected with siCtrl or siADAR1. The type I IFN response and ADAR1 knockdown efficiency were monitored 78 h post-transfection as in (B). Data are means ± s.d. from a representative of four biological replicate experiments. Cells generated in (D) were transfected with siCtrl (siC) or siADAR1 (siA). Protein lysates were prepared 78 h post-transfection, followed by SDS-PAGE and immunoblotting using the indicated antibodies (n = 2). HEK293 WT cells and FLAG-LGP2-expressing HEK293 cells were transfected with siCtrl or siADAR1 and subsequently plated on coverslips for immunofluorescence microscopy. Cells were fixed, permeabilized, and stained 72 h post-transfection with anti-FLAG (red) and anti-IRF3 (green) antibodies. Nuclei were stained with DAPI (blue). Scale bar is 50 μm. Total nuclei (> 450 nuclei per experimental condition) and IRF3-positive nuclei were counted using semi-automated software analysis and plotted as percentage IRF3-positive nuclei of total nuclei per field of view (a representative of three biological replicate experiments is quantified). The boxplot indicates the interquartile range as a box, the median as a central line, and the whiskers extend from the minimum to the maximum value. Statistical analyses were performed using unpaired two-tailed Mann–Whitney U tests. ns, not significant; ****P < 0.0001.

Source data are available online for this figure.

Details are in the caption following the image

Figure EV1. LGP2 is required for the induction of a type I IFN response upon siRNA-mediated depletion of ADAR1 in HEK293, related to Fig 1

PMA-differentiated THP-1 WT, RIG-I-, MDA5-, and LGP2-knockout cells were transfected with siCtrl (siC) or siADAR1 (siA). Protein lysates were prepared 56 h post-transfection and ADAR1 knockdown efficiency was monitored by immunoblot analysis. Data correspond to the biological replicate shown in Fig 1B. Expression level and type I IFN inducibility of relevant proteins in HEK293 and THP-1. HEK293 and PMA-differentiated THP-1 cells were treated with or without recombinant type I IFN. Protein lysates were analyzed by SDS-PAGE followed by immunoblotting with the indicated antibodies (n = 3). *, nonspecific band. Kinetics of siRNA-mediated depletion of ADAR1 and induction of the type I IFN response in HEK293. HEK293 cells stably expressing FLAG-LGP2 were transfected with an siRNA targeting ADAR1 (siADAR1) and harvested at the indicated time points post-transfection. ADAR1 knockdown and IFIT1 upregulation were monitored by RT-qPCR analysis (using Taqman probes) and normalized to ACTB. Data are means ± s.d. from one experiment. HEK293 WT cells were transfected with siADAR1 or a control siRNA (siCtrl) and 8 h later with increasing amounts of a vector encoding FLAG-LGP2. As a control, cells were transfected with 250 ng of an empty vector (EV). Cells were harvested 80 h post siRNA transfection. RT-qPCR analysis was used to monitor IFN-β and IFIT1 expression, ADAR1 knockdown, and LGP2 (DHX58) expression. All transcripts were normalized to ACTB. Data are means ± s.d. from a representative of two biological replicate experiments. Cells were treated as in (D). Protein lysates were prepared and analyzed by SDS-PAGE followed by immunoblotting using the indicated antibodies.

Source data are available online for this figure.

To further delineate the contribution of LGP2 to the sensing of unedited self RNA, we knocked out RIG-I, MDA5, or both by CRISPR/Cas9-mediated gene editing in the human cell line HEK293. Correct gene editing was confirmed by immunoblotting and intact type I IFN receptor signaling in the selected clones was verified by monitoring ISG60 expression upon recombinant type I IFN treatment (Fig 1D). Unexpectedly, siRNA-mediated depletion of ADAR1 did not yield signs of a type I IFN response in parental HEK293 cells or its CRISPR/Cas9-engineered derivatives (Fig 1E). We noted that these HEK293 cells expressed nearly undetectable levels of LGP2, even after stimulation with recombinant type I IFN (Fig EV1B). Importantly, ectopic expression of LGP2 by means of retroviral transduction and stable integration of a FLAG-tagged LGP2-encoding vector (Fig 1D) enabled type I IFN induction upon siRNA-mediated ADAR1 depletion, as determined by the expression of IFN-β and IFIT1 transcripts (Fig 1E). This was evident in MDA5-sufficient but not in MDA5-deficient cells, confirming that LGP2 and MDA5 were both necessary. As expected, loss of RIG-I did not have a major impact on type I IFN induction upon ADAR1 knockdown (Fig 1E). Of note, increased MDA5 expression, through pretreatment with recombinant type I IFN, did not bypass the requirement for LGP2 (Fig EV2A and B), which suggests that the level of MDA5 is not the rate-limiting factor. The ISG signature reached its maximum around 78 h post-siRNA delivery in LGP2-overexpressing cells (Fig EV1C). These observations were confirmed at protein level: ADAR1 depletion upon siRNA treatment led to robust upregulation of the protein ISG60 exclusively in cells that express both MDA5 and LGP2 (Fig 1F). Moreover, the presence of both LGP2 and MDA5 was required for phosphorylation of IRF3 and STAT1, two key transcription factors that act downstream of MAVS and IFNAR to induce IFN-β and ISG transcription, respectively (Fig 1F). Finally, nuclear translocation of IRF3, a hallmark of type I IFN induction, only occurred upon expression of LGP2 in ADAR1-depleted cells (Fig 1G). These observations demonstrate that LGP2 is essential for type I IFN induction and signaling upon ADAR1 depletion.

Details are in the caption following the image

Figure EV2. LGP2-deficient cells fail to sense unedited self RNAs, yet maintain the ability to detect viral dsRNAs, related to Fig 1

A. WT and LGP2-knockout (clones 1 and 2) HEK293 cells were transfected with an siRNA targeting ADAR1 (siADAR1) or a control siRNA (siCtrl) and were treated 8 h later with recombinant type I IFN to upregulate RLR expression. Cells were harvested 80 h post siRNA transfection and RT-qPCR analysis was used to monitor IFN-β and IFIT1 expression and ADAR1 knockdown. All transcripts were normalized to ACTB. Data are means ± s.d. from a representative of three biological replicate experiments. B. Cells were treated as in (A). Protein lysates were prepared 80 h post siRNA transfection, followed by SDS-PAGE and immunoblotting using the indicated antibodies (n = 3). siC, siCtrl; siA, siADAR1. C. LGP2-knockout (clones 1 and 2) HEK293 cells were transfected with siADAR1 or siCtrl and 8 h later with a vector encoding FLAG-LGP2 or an empty vector (EV). Cells were harvested 80 h post siRNA transfection and RT-qPCR analysis was used to monitor IFN-β and IFIT1 expression and ADAR1 knockdown. All transcripts were normalized to ACTB. Data are means ± s.d. from a representative of two biological replicate experiments. D. Cells were treated as in (C). Protein lysates were prepared 80 h post siRNA transfection, followed by SDS-PAGE and immunoblotting using the indicated antibodies (n = 2). E, F. WT, LGP2-knockout (clones 1 and 2), and stably expressing FLAG-LGP2 HEK293 cells were transfected with transfection reagent only (LF2000), poly(I:C) (56, 112, 225, or 450 ng in (E)), or RNA isolated from HEK293 cells infected with EMCV in the presence of ribavirin (450 or 900 ng in (F)). Cells were harvested 16 h post-transfection and RT-qPCR analysis was used to monitor IFN-β and IFIT1 expression. All transcripts were normalized to ACTB. Data are means ± s.d. from a representative of four (E) or three (F) biological replicate experiments. G. MDA5-knockout HEK293 cells stably expressing FLAG-LGP2 or an empty vector (EV) were transfected with increasing amounts (5, 20, 40, 80, or 240 ng/ml) of a vector encoding FLAG-MDA5 WT or FLAG-MDA5 G495R. As a control, cells were transfected with 240 ng/ml control vector or left untreated. Cells were harvested 24 h post-transfection and RT-qPCR analysis was used to monitor IFN-β and IFIT1 expression. All transcripts were normalized to ACTB. Data are means ± s.d. from a representative of two biological replicate experiments.

Source data are available online for this figure.

Previous studies indicated that LGP2 can function as a concentration-dependent biphasic switch that favors MDA5 signaling in response to viral ligands at low concentrations while inhibiting MDA5-dependent responses at high concentrations (Rodriguez et al, 2014). However, in our experiments, increasing amounts of an LGP2-encoding plasmid led to a gradual increase in the type I IFN response upon siRNA-mediated ADAR1 depletion without any signs of an inhibitory effect (Fig EV1D and E) except at very high doses of LGP2, which negatively affect cell viability. The LGP2-dependent biphasic response previously reported in the context of viral dsRNA sensing is therefore not evident in self RNA sensing.

The absolute requirement for LGP2 in the induction of a type I IFN response following ADAR1 depletion was surprising and distinct from its role in viral dsRNA sensing, where LGP2 evidently potentiates MDA5 signaling but is not strictly required. Indeed, while bona fide LGP2-knockout HEK293 cells failed to induce a type I IFN response upon ADAR1 depletion (Fig EV2A and B), they retained the ability to induce a modest, yet reduced, type I IFN response upon stimulation with the dsRNA mimic high molecular weight (HMW) poly(I:C) or RNA isolated from EMCV-infected cells, both of which activate MDA5 (Fig EV2E and F). As a control, the siADAR1-induced IFN response was restored in LGP2-knockout cells upon ectopic LGP2 expression (Fig EV2C and D). Whether the differential detection of unedited self RNA versus viral RNA by LGP2/MDA5 is caused by a qualitative or quantitative difference, or both, is not clear. Unedited self RNA may either be less abundant in cells or be a less suitable MDA5 ligand (e.g., because it contains only short stretches of base-paired regions as opposed to long dsRNA found in viral RNA) and therefore it may be more reliant on LGP2 for its detection. Either way, it is evident that the requirement for LGP2 becomes critical in the case of an “imperfect” MDA5 ligand. As a side note, in another setting of autoinflammation due to a gain-of-function mutation in MDA5 (MDA5 G495R) (Rice et al, 2014), LGP2 also enhanced but was not strictly required for type I IFN induction (Fig EV2G), which indicates that LGP2 is not necessarily essential for the detection of all types of self RNA. Altogether, these findings implicate human LGP2 as a key player in the response to unedited self RNA in ADAR1-depleted cells.

Sensing of unedited self RNA via LGP2 requires RNA binding and ATP hydrolysis

The limited expression of LGP2 and the absence of a type I IFN response upon ADAR1 depletion in wild-type HEK293 cells allowed us to create ADAR1 knockout cells through CRISPR/Cas9, without the activation of innate immune pathways that hinder cell proliferation. Two ADAR1-knockout clones were selected that completely lost expression of the ADAR1 p110 and p150 isoform yet remained responsive to type I IFNs, as determined by immunoblotting (Fig 2A). Genetic loss of ADAR1 did not reveal a type I IFN response until introduction of LGP2 (Figs 2B and C, and EV3A and B), in line with our earlier observations using ADAR1 siRNAs. As reported (Pestal et al, 2015), the IFN response was largely due to the loss of the p150 isoform, as reconstitution of p150 expression completely blocked type I IFN induction in LGP2-expressing ADAR1 knockout cells (Figs 2B and C, and EV3A and B). In contrast, overexpression of the p110 isoform reduced, but did not block, this type I IFN response. The reduction can most likely be explained by overexpression of this isoform, which is normally restricted to the nucleus but can “spill” into the cytosol in overexpressing cells. The ADAR1-deficient cells with a tunable, LGP2-dependent type I IFN response provide us therefore with a useful tool to dissect the features of LGP2 and its interaction partners that are required for unedited self RNA sensing.

Details are in the caption following the image

Figure 2. A type I IFN response is unleashed in ADAR1-knockout cells upon expression of LGP2

HEK293 cells were genetically engineered to knock out ADAR1 using CRISPR/Cas9. Cells were treated for 24 h with recombinant type I IFN to upregulate ADAR1 p150 and ISG60 to confirm correct gene editing and type I IFN responsiveness, respectively. Protein lysates were analyzed by SDS-PAGE followed by immunoblotting using the indicated antibodies (n = 3). ADAR1-knockout HEK293 cells (clone 1) were cotransfected with an empty vector (EV) or a FLAG-LGP2-encoding vector (LGP2) combined with a vector encoding GFP-tagged ADAR1 p110 or p150. Cells were harvested 72 h post-transfection and the type I IFN response was monitored by RT-qPCR analysis of IFN-β and IFIT1 expression, normalized to ACTB. Data are means ± s.d. from a representative of four biological replicate experiments. ADAR1-knockout cells (clone 1) were transfected as in (B). Protein lysates were analyzed by SDS-PAGE followed by immunoblotting using the indicated antibodies (n = 4).

Source data are available online for this figure.

Details are in the caption following the image

Figure EV3. A type I IFN response is unleashed in ADAR1-knockout cells upon expression of LGP2, related to Figs 2 and 3

ADAR1-knockout HEK293 cells (clone 2) were cotransfected with an empty vector (EV) or a FLAG-LGP2-encoding vector (LGP2) combined with a vector encoding GFP-tagged ADAR1 p110 or p150. Cells were harvested 48 h post-transfection and the type I IFN response was monitored by measuring IFN-β and IFIT1 transcript expression, relative to ACTB expression, by RT-qPCR. Data are means ± s.d. from a representative of three biological replicate experiments. ADAR1-knockout HEK293 cells (clone 2) were transfected as in (A). Protein lysates were analyzed by SDS-PAGE followed by immunoblotting using the indicated antibodies (n = 3). MDA5-knockout HEK293 cells, generated in Fig 1D, were transfected with an ADAR1-targeting siRNA (siADAR1) or a control siRNA (siCtrl) and 8 h later with an empty vector (EV) or a vector encoding the indicated WT, truncation, or point mutant(s) of MDA5. Cells were harvested 72 h post-siRNA transfection and RT-qPCR analysis was used to monitor ADAR1 knockdown and MDA5 (IFIH1) expression. All transcripts were normalized to ACTB. Data are means ± s.d. from a representative of two biological replicate experiments.

Source data are available online for this figure.

The canonical function of LGP2 as an RNA sensor involves RNA binding and ATP hydrolysis while other roles, such as interaction with MAVS and TRAFs, do not (Parisien et al, 2018; Esser-Nobis et al, 2020). We introduced, by means of lentiviral transduction, a doxycycline-inducible system to stably express FLAG-LGP2 WT or a mutant that completely fails to bind RNA (FLAG-LGP2 K138E/R490E/K634E, denoted as “LGP2 KRK” in figures) in ADAR1 KO cells (Fig 3A and B). Doxycycline-induced expression of LGP2 WT in ADAR1 KO cells led to robust ISG60 protein (Fig 3A) and IFN-β and IFIT1 transcript induction (Fig 3B). In contrast, expression of the LGP2 RNA-binding mutant did not induce a type I IFN response. Consistent with these findings, induction of LGP2 WT, but not the RNA-binding mutant, allowed nuclear translocation of IRF3 (Fig 3C). These observations indicate that binding to RNA substrates is required for LGP2-dependent type I IFN induction in ADAR1-deficient cells.

Details are in the caption following the image

Figure 3. RNA binding by LGP2 is required for receptor oligomerization and type I IFN induction in ADAR1-knockout cells

ADAR1-knockout HEK293 cells (clone 1) were modified with a lentiviral-based inducible system to express FLAG-LGP2 WT or a FLAG-LGP2 RNA binding mutant (K138E/R490E/K634E, denoted as “KRK mutant”) in a doxycycline-regulated manner. Cells were treated 72 h with doxycycline (dox). Protein lysates were analyzed by SDS-PAGE and immunoblotting using the indicated antibodies (n = 3). iEV, inducible empty vector; iLGP2, inducible LGP2. Cells generated in (A) were treated with doxycycline for 72 h to induce LGP2 WT or KRK mutant gene expression. The type I IFN response (IFN-β and IFIT1 transcripts) and LGP2 (DHX58) expression were monitored by RT-qPCR analysis. All transcripts were normalized to ACTB. Data are means ± s.d. from a representative of three biological replicate experiments. Cells generated in (A) were plated on coverslips and treated with or without doxycycline for 72 h. Cells were fixed, permeabilized, and stained with anti-FLAG (red) and anti-IRF3 (green) antibodies. Nuclei were stained with DAPI (blue). Scale bar is 50 μm. Total nuclei (> 500 nuclei per experimental condition) and IRF3-positive nuclei were counted using semi-automated software analysis and plotted as percentage IRF3-positive nuclei of total nuclei per field of view (a representative of two biological replicate experiments is quantified). The boxplot indicates the interquartile range as a box, the median as a central line, and the whiskers extend from the minimum to the maximum value. Statistical analysis was performed using a Kruskal–Wallis test with a Dunn's post hoc test for multiple comparisons. ns, not significant; ****P < 0.0001. Cells generated in (A) were treated with doxycycline for 72 h. During the last 24 h, recombinant type I IFN was added to upregulate endogenous MDA5 protein expression. Protein lysates were analyzed by SDD-AGE and SDS-PAGE using the indicated antibodies to determine protein oligomerization and total expression levels, respectively (n = 3). MDA5-knockout HEK293 cells, generated in Fig 1D, were transfected with an ADAR1-targeting siRNA (siADAR1) or a control siRNA (siCtrl) and 8 h later with an empty vector (EV) or a vector encoding the indicated WT, truncation, or point mutant(s) of MDA5. Cells were harvested 72 h post-siRNA transfection and the type I IFN response was monitored by RT-qPCR analysis of IFIT1 and ISG15 transcript expression, normalized to ACTB. Data are means ± s.d. from a representative of two biological replicate experiments.

Source data are available online for this figure.

To determine whether LGP2 is required for MDA5 oligomerization in ADAR1-deficient cells, we utilized semi-denaturing detergent agarose gel electrophoresis (SDD-AGE) to monitor MDA5 aggregation. To circumvent discrepancies in MDA5 protein levels across samples (due to its increased expression as an ISG in LGP2-expressing ADAR1 KO cells), we treated cells with recombinant type I IFN to equalize MDA5 expression (Fig 3D, SDS-PAGE). Doxycycline-inducible expression of LGP2 WT, but not the RNA-binding mutant, revealed MDA5 aggregation in ADAR1-knockout cells (Fig 3D, SDD-AGE). SDD-AGE further revealed that RNA-binding competent LGP2 oligomerizes in ADAR1-knockout cells (Fig 3D), consistent with previous studies showing that human and chicken LGP2 itself can form filaments (Bruns et al, 2014; Uchikawa et al, 2016). We further tested what features of MDA5 are important for LGP2-dependent type I IFN induction in ADAR1-depleted cells. We transiently expressed various MDA5 mutants in MDA5 KO HEK293 cells that were stably transduced with FLAG-LGP2 or an empty vector (as a control) and depleted ADAR1. In contrast to WT MDA5, MDA5 mutants that have impaired capacity to form filaments (M570R/D572R, I841R/E842R) or to bind RNA (R728A, H927A, R728A/H927A) (Wu et al, 2013) or that lack the CARD domains (MDA5ΔCARD), all fail to induce a type I IFN response in ADAR1-depleted cells (Figs 3E and EV3B). Altogether, the above findings place LGP2 at the level of MDA5 oligomerization and activation in the ADAR1-induced type I IFN response.

We transiently expressed various LGP2 truncation mutants and point mutants (Fig 4A) in ADAR1-knockout cells and observed that, besides RNA binding, full-length LGP2 and its ability to hydrolyze ATP are strictly required to sustain a type I IFN response. Expression of the LGP2 N-terminal domain (NTD), C-terminal domain (CTD), or mutation of LGP2 residues that are critical for ATPase activity (K30A) or RNA binding via the LGP2 NTD (K138E/R490E) or CTD (K634E) (Pippig et al, 2009; Bruns et al, 2013; Uchikawa et al, 2016), all abolished type I IFN induction in ADAR1-deficient cells (Fig 4B and C).

Details are in the caption following the image

Figure 4. The function of LGP2 in sensing unedited self RNA involves its canonical role as dsRNA sensor that requires RNA binding and ATP hydrolysis

Schematic illustration of the domain structure of LGP2 and various point mutants and truncation mutants that are used in this study. The N-terminal domain (NTD) of LGP2 is composed of a conserved DExH/D helicase domain, subdivided into the helicase 1 (Hel1), helicase 2 (Hel2) and helicase insertion (Hel2i) domain, and a pincer motif (P). The NTD is followed by a C-terminal domain (CTD), involved in RNA binding. HEK293 WT or ADAR1-knockout cells (clone 1) were transiently transfected with an empty vector (EV) or a vector encoding the indicated WT, truncation, or point mutant(s) of LGP2. Cells were harvested 72 h post-transfection and the type I IFN response was monitored by RT-qPCR analysis of IFN-β and IFIT1 transcript expression, normalized to ACTB. Data are means ± s.d. from a representative of three biological replicate experiments. Cells were treated as in (B). Protein lysates were prepared and analyzed by SDS-PAGE followed by immunoblotting using the indicated antibodies (n = 3). The dotted line indicates the juxtaposition of two nonadjacent lanes.

Source data are available online for this figure.

Mutation of a cysteine residue in the C-terminal domain of LGP2 crucial for binding to the dsRNA-binding protein PACT (C615A) (Sanchez David et al, 2019) also prevented type I IFN induction, suggesting that PACT, via LGP2, may participate in the IFN response to unedited self RNA (Fig 4B and C). Of note, C615 is also important for the correct orientation of a Zn2+ ion in LGP2 (Pippig et al, 2009); hence, the role of PACT and/or Zn2+ binding will need to be evaluated in further studies.

A recent study identified a biochemical interaction between LGP2 filamentous structures and TRIM14, an unusual member of the TRIM family that lacks a RING domain and does not function as a ubiquitin E3 ligase (Kato et al, 2021). Mutations in the α3 helix of the Hel2 domain of LGP2 (Q390R/T395R or Q390A/Q394A) strongly decreased the interaction between the Hel2i-Hel2 domain of LGP2 and TRIM14 (Kato et al, 2021). We found that the Q390A/Q394A mutation did not impact on the ab

留言 (0)

沒有登入
gif