Lysosomal endonuclease RNase T2 and PLD exonucleases cooperatively generate RNA ligands for TLR7 activation

Abstract

Toll-like receptor 7 (TLR7) is essential for recognition of RNA viruses and initiation of antiviral immunity. TLR7 contains two ligand-binding pockets that recognize different RNA degradation products: pocket 1 recognizes guanosine, while pocket 2 coordinates pyrimidine-rich RNA fragments. We found that the endonuclease RNase T2, along with 5' exonucleases PLD3 and PLD4, collaboratively generate the ligands for TLR7. Specifically, RNase T2 generated guanosine 2',3'-cyclic monophosphate-terminated RNA fragments. PLD exonuclease activity further released the terminal 2',3'-cyclic guanosine monophosphate (2',3'-cGMP) to engage pocket 1 and was also needed to generate RNA fragments for pocket 2. Loss-of-function studies in cell lines and primary cells confirmed the critical requirement for PLD activity. Biochemical and structural studies showed that PLD enzymes form homodimers with two ligand-binding sites important for activity. Previously identified disease-associated PLD mutants failed to form stable dimers. Together, our data provide a mechanistic basis for the detection of RNA fragments by TLR7.

Keywords: PLD3; PLD4; RNA sensing; RNase T2; TLR7; plasmacytoid dendritic cell.

Graphical abstract

Figure 1

RNase T2 acts upstream of TLR7 (A) Unmodified CAL-1 cells (WT) or two independent TLR7^−/y CAL-1 clones were unstimulated or stimulated with pR, RNA40^S, RNA40^O, RNA9.2s^O, R848, 2′,3′-cGMP (++ = 1 mM, + = 0.5 mM), CpG^O, or CpG^S. After 16 h, IFN-β release was measured by enzyme-linked immunosorbent assay (ELISA). Data are depicted as mean ± SEM of n = 3 independent experiments. Statistical analysis was conducted by two-way ANOVA with Dunnett’s multiple comparison tests. (B) TLR7 expression in CAL-1 WT and TLR7^−/y CAL-1 clones by immunoblot. One representative blot of three independent experiments is shown. (C) Isolated primary plasmacytoid dendritic cells were unstimulated or stimulated with pR, RNA40^S, RNA40^O, RNA9.2s^O, R848, 2′,3′-cGMP (++ = 1 mM, + = 0.5 mM), and CpG^O. After 24 h, IFN-α2 release was measured by ELISA. Data are depicted as mean ± SEM of n = 3 independent donors. (D) Unmodified CAL-1 cells (WT) or two independent RNASET2^−/− CAL-1 clones were stimulated as in (A) and IFN-β release was measured. Data are depicted as mean ± SEM of n = 3 independent experiments. Statistical analysis was conducted by two-way ANOVA with Dunnett’s multiple comparison tests. (E) Urea gel of indicated substrates digested with RNase T2 (0.37 nM). One representative gel of three independent experiments is shown.

Figure 2

PLD3 and PLD4 act upstream of TLR7 (A) mRNA expression levels of PLD3, PLD4, RNase T2, TLR7, and TLR8 in indicated cell types are plotted. (B) Single-cell RNA-seq data of human PBMCs from Hao et al. were visualized using the UCSC Cell Browser for the transcripts PLD3 and PLD4. Color coding represents the range of gene expression. The insert in the upper left corner specifically highlights the pDC population. Annotations for relevant cell populations were retrieved as annotated. (C) Quantification of PLD3 and PLD4 protein expression in CAL-1 cells, primary plasmacytoid dendritic cells, and primary monocytes. Data are presented as box and whiskers of n = 4–5 independent experiments, statistics indicate a paired two-tailed Student’s t test. (D) Unmodified CAL-1 cells (WT) or two independent PLD3^−/− × PLD4^−/− CAL-1 clones were unstimulated or stimulated with pR, RNA40^S, RNA40^O, RNA9.2s^O, CpG^O, or CpG^S. After 16 h, IFN-β release was measured by ELISA. Data are depicted as mean ± SEM of n = 3 independent experiments. Statistical analysis was conducted by two-way ANOVA with Dunnett’s multiple comparison tests. (E) Indicated knockouts of primary human monocytes were unstimulated or stimulated with ssRNA40^O in the presence of CU-CPT9a, R848, and LPS. After 16 h, IL-6 release was measured by ELISA. Each replicate of 3 independent donors is depicted. For statistical analysis, the data of each donor was normalized to WT^NTC and two-way ANOVA was conducted on log-transformed data with Dunnett’s multiple comparison test.

Figure 3

RNase T2 and PLD enzymes release 2′,3′-cyclic GMP (A and B) Fluorescence intensity signal of FAM-RNA40-BMN-Q530 over time, incubated with indicated concentrations of (A) PLD3 or (B) PLD4. Data are depicted as mean of n = 3 independent experiments. (C) Overlay of extracted ion chromatograms (EIC) of 5′-GMP, 3′-GMP, and 2′,3′-cyclic GMP. Top: standards used as reference. Middle: in vitro digests of RNA9.2s with PLD3 (250 nM). Bottom: PLD4 (250 nM). (D) Normalized signal areas of EIC from released 5′-NMPs, 3′-NMPs, and 2′,3′-cNMPs by PLD3 (250 nM) and PLD4 (250 nM) after degradation of RNA9.2s^O in %. (E and F) CAL-1 WT cells (300,000 cells/well) were stimulated with increasing concentrations of R848 and 2′,3′-cGMP. After 16 h, IFN-β release was determined by ELISA. Each replicate of n = 2 (E) or n = 3 (F) independent experiments is depicted. A four-parameter dose-response curve was fitted to calculate half-maximal effective concentration (EC₅₀). (G) CAL-1 cells (300,000 cells/well) of indicated genotypes were unstimulated or stimulated with indicated concentrations of 2′,3′-cGMP and 3′-GMP. Data are depicted as mean ± SEM of three independent experiments. (H) Schematic view of the in vitro digestion assay. (I) RNA40^O was digested with RNase T2 (370 nM), PLD3 (250 nM), and PLD4 (250 nM) or in combinations of RNase T2 (370 nM) with either PLD3 (250 nM) or PLD4 (250 nM) for 20 min, and the release of 2′,3′-cGMP was analyzed by LC-MS. (J and K) RNA40^O was digested with RNase T2 (370 nM) in combination with either PLD3 (+ = 2.5 nM), PLD4 (+ = 50 nM), or with PLD4 (++ = 195 nM) for 20 min, and the release of 2′,3′-cGMP was analyzed by LC-MS. (L) RNA40^O was digested with RNase 1 (5.7 nM), RNase 2 (27 nM), RNase 6 (29 nM), or RNase T2 (37 nM) only or in combination with PLD3 (25 nM) for 30 min, and the release of 2′,3′-cGMP was analyzed by LC-MS. For (I), (J), (K), and (L), data are depicted as mean ± SEM of n = 3 independent experiments. (M) Detection of 2′,3′-cGMP in cell lysates of RNA40^O-stimulated WT, RNase T2, or PLD3^−/− × PLD4^−/− CAL-1 cells by LC-HRMS. Data are depicted as mean ± SEM of n = 3 independent experiments. Statistical analysis was conducted by one-way ANOVA with Dunnett’s multiple comparison tests.

Figure 4

PLD enzymes create TLR7 second binding pocket fragments (A) CAL-1 cells of indicated genotypes were unstimulated or stimulated with pR, RNA40^S, RNA40^O, RNA9.2s^O, R848, 2′,3′-cGMP (0.5 mM), CpG^O, and CpG^S. After 16 h, IFN-β release was determined by enzyme-linked immunosorbent assay (ELISA). Data are presented as mean ± SEM of n = 3 independent experiments. Statistical analysis was conducted by two-way ANOVA with Dunnett’s multiple comparison tests. Note that the WT data are identical with the ones shown in Figure 1A. (B) Urea gels of RNA9.2s^O digested with PLD4 (+ = 25 nM, ++ = 250 nM) over time. One representative gel of three independent experiments is shown. (C) Urea gels of RNA9.2s^O digested with PLD3 (+ = 0.39 nM, ++ = 1.56 nM) over time. One representative gel of two independent experiments is shown. (D) Urea gel of RNA9.2s^O (1 μg) incubated with PLD4 (++ = 250 nM, + = 25 nM) for 2 h. One out of three independent experiments is shown. (E) LC-HRMS total ion current (TIC) chromatogram of RNA9.2s^O-derived ORN fragments after digestion with PLD4 (250 nM) for 2 h. (F) Calculated and found masses (m/z) of RNA9.2s-derived ORN fragments after digestion with PLD4 (250 nM) for 2 h. (G) Urea gel of indicated substrates digested with PLD4 (++ = 250 nM, + =25 nM). One out of two independent experiments is shown. (H) LC-MS/MS analysis of depicted substrates digested with PLD4 (250 nM) for 20 min. Data were normalized to the amount of the different nucleosides present in the sequence and are depicted as mean ± SEM of n = 3 independent experiments. (I) CAL-1 cells of indicated genotypes were unstimulated or stimulated with 2′,3′-cGMP (0.5 mM), short ORNs, or ORNs in combination with 2′,3′-cGMP (0.5 mM). Data are depicted as mean + SEM of n = 3 independent experiments. Statistical analysis was conducted by two-way ANOVA (left panel) or one-way ANOVA (right panel) with Dunnett’s multiple comparison tests.

Figure 5

PLD3 and PLD4 form homodimers (A) Size exclusion chromatography (SEC) run of PLD3 superimposed with the SEC run of PLD4. (B) Mass distribution of PLD3 and PLD4 observed by mass photometry. (C) Ribbon representation of the PLD3 dimer shown together with the cryo-EM density map. The inset shows a detailed view of the dimer interface, with the involved residues shown in stick representations. The mutated residue R340 and active site HxK motifs are highlighted in pink. (D) Structural comparison between the PLD3 cryo-EM structure, colored in turquoise and dark blue, and the PLD4 model predicted by Alphafold, colored in beige. The inset illustrates the comparison of the PLD3/PLD4 dimer interface, with involved residues highlighted using stick representations. (E) Representative 2D classes of PLD4 particles illustrating the dimeric conformation. (F) Size exclusion chromatography (SEC) run of PLD3 superimposed with the SEC run of PLD3(R340D). Note that the SEC control run of PLD3 is identical to Figure 5A. (G) Mass distribution of PLD3 and PLD3(R340D) observed by mass photometry. Note that the mass distribution control of PLD3 is identical to Figure 5B. (H and I) Urea gels of RNA9.2s^O and CpG^O-DNA digested with indicated concentrations of PLD3 and PLD3(R340D). One out of two independent experiments is shown.

Figure 6

PLD3 shows high-affinity binding to long oligonucleotides (A) Urea gel of RNA9.2s^O incubated with PLD3 (250 nM) and PLD3(H201N, H416N) (250 nM). One out of three independent experiments is shown. (B and C) Fluorescence anisotropy assays assessing the binding of PLD3(H201N, H416N) at increasing concentrations to indicated substrates. Data are shown as mean ± SEM of n = 3 independent experiments. Note that the 19-nt RNA9.2s binding data are identical in (B) and (C). (D) The electrostatic surface of the PLD3 dimer, calculated with APBS at pH 5, illustrates the presence of two positively charged areas. The two ssRNA segments, colored in gold, have been modeled using Hdock. (E) Ribbon representation of PLD3 with highlighted active site HxK motif colored in pink and residues presumably involved in substrate recognition shown as sticks and color coded according to the corresponding protomer, with turquoise or dark blue, respectively. The inset shows a detailed view of the comparison between PLD3 (turquoise) and PLD4 (orange) at the active site cleft, illustrating the conservation of the highlighted residues. (F) Cryo-EM density map for PLD3 bound to ssRNA highlighting the additional density appearing in gold. (G) Representative 2D classes of PLD3 particles or PLD3 ssRNA complex particles. (H–J) Urea gels of RNA9.2s^O digested with decreasing concentrations of PLD3 or with decreasing concentrations of depicted PLD3 mutants. For all urea gels, one out of two independent experiments is shown.