Cyprinid-specific duplicated membrane TLR5 senses dsRNA as functional homodimeric receptors

Abstract

Membrane-embedded Toll-like receptor 5 (TLR5) functions as a homodimer to detect bacterial flagellin. Cyprinid grass carp (Ctenopharyngodon idella) encodes two TLR5 genes, CiTLR5a and CiTLR5b. Here, we show that cyprinid TLR5a and TLR5b homodimers unexpectedly bind the dsRNA analog poly(I:C) and regulate interferon (IFN) response in early endosomes and lysosomes. Although TLR5 homodimers also bind flagellin, an immune response to flagellin is only triggered by TLR5a/b heterodimer. Moreover, we demonstrate that two TLR5 paralogs have opposite effects on antiviral response: CiTLR5a slightly promotes and powerfully maintains, whereas CiTLR5b remarkably inhibits virus replication. We show that the ectodomain of CiTLR5 is required for dsRNA-induced IFN signaling, and we map the key poly(I:C) binding sites to G240 for CiTLR5a and to N547 for CiTLR5b. Furthermore, we reveal that differential N-glycosylation of CiTLR5a/b affects dsRNA-IFN signaling but has no role in flagellin-mediated NF-κB induction, with paralog-specific roles for CiTLR5a-T101 and corresponding CiTLR5b-I99. Moreover, we provide evidence that the ability to sense dsRNA represents a neofunctionalization specific for membrane-bound TLR5 in cyprinid, bridging viral and bacterial immune responses.

Keywords: N-glycosylation modification; TLR5; dsRNA; hetero- and homodimer; interferon.

Figure 1. CiTLR5a and CiTLR5b can form homodimers and heterodimer

1. A–C

   CFM on FHM cells transfected with CiTLR5a‐GFP and CiTLR5b‐RFP (A), CiTLR5a‐GFP and CiTLR5a‐RFP (B), or CiTLR5b‐GFP and CiTLR5b‐RFP (C). Green, red, blue, and yellow indicate CiTLR5a, CiTLR5b, nucleus, and co‐localization, respectively (scale bars: 10 μm).

2. D, E

   mRNA and protein expression of CiTLR5a‐HA (D), CiTLR5b‐HA (D), CiTLR5a‐GFP (E), and CiTLR5b‐GFP (E) plasmids in CIK cells were examined by qRT–PCR (upper panel) with EF1α gene as an internal control and by immunoblotting (IB) (lower panel) with β‐tubulin as a reference (n = 4).

3. F–H

   FHM cells in 10 cm² dishes were transfected with the indicated plasmids for 24 h. Co‐IP was performed with anti‐HA monoclonal antibody. Mouse IgG was used as control. Immunoblotting was performed with anti‐HA and anti‐GFP antibodies, respectively.

Data information: All data are presented as mean ± SD. The two‐tailed Mann–Whitney U‐test was used for (D) and (E). n represents biological replicates. ***P < 0.001.

Figure EV1. Endogenous PpTLR5a/5b does not affect the functions of CiTLR5a and CiTLR5b. RNA‐seq and qRT–PCR analysis of CIK cells co‐expressing CiTLR5a and CiTLR5b in different relative ratios

1. A

   mRNA expression of CiTLR5a and CiTLR5b in CIK cells was examined by qRT–PCR (n = 4). EF1α gene was employed as an internal control.

2. B–D

   Specific siRNAs of PpTLR5a/5b and negative control siRNA (si.C) were transfected into FHM cells and CIK cells for 12 h, respectively. The cells were analyzed by qRT–PCR to quantify the expression of PpTLR5a/5b (C), CiTLR5a (D), and CiTLR5b (E) (n = 4).

3. E, F

   FHM cells were transfected with s1, indicated plasmids for 12 h, and then stimulated with poly(I:C) (F) or FlaC (G) for 24 h. IFN1 (F) and NF‐κB2 (G) promoter activities were determined by DLR assay (n = 3).

4. G, H

   FHM cells were transfected with s2, indicated plasmids for 12 h, and then stimulated with poly(I:C) (H) or FlaC (I) for 24 h. IFN1 (H) and NF‐κB2 (I) promoter activities were examined by DLR assay (n = 3).

5. I

   CIK cells were transfected with the indicated combinations (Vector (2 μg), CiTLR5a (2 μg), CiTLR5b (2 μg), CiTLR5a (1 μg) + CiTLR5b (2 μg), CiTLR5a (2 μg) + CiTLR5b (2 μg), CiTLR5a (4 μg) + CiTLR5b (2 μg)) for 24 h, Samples were then stimulated with poly(I:C) (1 μg/ml) for 24 h. PBS was employed as a control. The samples were collected and RNA‐seq was carried out. A heatmap of mRNA expression of enriched antiviral immune genes (IFN1, IFN3, TBK1, Vig1, Gig1, Mx1, Mx2, ISG15, ISG56, and IL22) was made based on RNA sequencing.

6. J

   The samples prepared as the RNA‐seq were employed to verify mRNA expression of antiviral immune genes by qRT–PCR (n = 4). EF1α gene was used as an internal control.

Data information: All data are presented as mean ± SD. The two‐tailed Mann–Whitney U‐test was used for (A–H). n represents biological replicates. *P < 0.05, **P < 0.01.

Figure 2. CiTLR5a and CiTLR5b sense flagellin as a heterodimer and sense poly(I:C) as homodimers

1. A

   CIK cells were seeded in 12‐well plates for 24 h, stimulated with PBS (control), LPS, PGN, dsDNA, K. pneumonia lysate, poly(I:C), FlaA, FlaB, FlaC, A. hydrophila lysate, or GCRV for different time points (0, 3, 6, 12, 24, 48 h), and then mRNA expression of CiTLR5a and CiTLR5b was measured by qRT–PCR (n = 4).

2. B, C

   Effects of MAMPs and bacteria on NF‐κB2 (B) and IFN1 (C) promoter activities were measured by DLR assays in FHM cells. PBS serves as a control (n = 3).

3. D

   mRNA expression in 12 tissues, based on qRT–PCR analysis (n = 4). 18S rRNA gene was employed as an internal control.

4. E

   mRNA expression of CiTLR5a and CiTLR5b was examined by qRT–PCR in grass carp spleen and head kidney tissues at different time points post FlaC, A. hydrophila, poly(I:C), or GCRV challenge (n = 6).

5. F–H

   MAMPs (FlaA, FlaB, FlaC, dsDNA, poly(I:C), PGN, LPS, and negative control) were coated to 96‐well plates and then blocked with bovine serum albumin. The binding abilities between CiTLR5a‐ectodomain (E), CiTLR5b‐ectodomain (F), GST (G) and different MAMPs were assessed by ELISA (n = 4).

6. I–L

   Lysates of FHM cells overexpressing CiTLR5a and CiTLR5b were respectively incubated with FlaA, FlaB, FlaC, biotin‐poly(I:C), and biotin‐ssRNA (1 μg/ml) at pH 5.0 or pH 7.4 for 1 h at 4°C, mixed with HA Ab‐coupled beads or streptavidin agarose beads for 2 h at 4°C, and then analyzed by immunoblotting (IB) with anti‐HA or anti‐His antibodies.

7. M

   Amino acid sequence alignment of the EJM and putative transmembrane regions of CiTLR5a, CiTLR5b, DrTLR5a, DrTLR5b, and HsTLR5. The acidic amino acid residues in the EJM were highlighted in red. The transmembrane sequences were marked by shadow.

8. N, O

   CiUNC93B1‐Myc and CiTLR5a‐HA (M) or CiTLR5b‐HA (N) was transfected into FHM cells. Co‐IP and immunoblotting were conducted to test the interactions between CiUNC93B1‐Myc and CiTLR5a‐HA or CiTLR5b‐HA.

9. P, Q

   Effects of CiUNC93B1 on CiTLR5a‐ and CiTLR5b‐mediated IFN1 (O) and NF‐κB2 (P) promoter activities were examined in FHM cells by DLR assays (n = 3). Both poly(I:C) and FlaC are 1 μg/ml.

10. R, S

    FHM cells were transfected with CiTLR5a‐GFP (Q) or CiTLR5b‐GFP (R) and RAB5‐RFP, an early endosomes protein marker, RAB7‐RFP, a late endosome protein marker, or LAMP2‐RFP, a lysosome protein marker, seeded on microscope cover glasses in 12‐well plates for 24 h, and then fixed and examined by CFM. Green represents CiTLR5a‐GFP or CiTLR5b‐GFP, red represents RAB5‐RFP, RAB7‐RFP, or LAMP2‐RFP, and blue indicates nucleus. Yellow represents co‐localization between CiTLR5a or CiTLR5b and early endosomes or lysosomes (scale bars: 10 μm).

11. T, U

    Effects of chloroquine (in 5 μg/ml) on CiTLR5a‐ and CiTLR5b‐dependent signaling were examined by DLR assays in FHM cells (n = 3). HsTLR5 and CiMDA5 (C. idella MDA5) were employed as positive controls.

Data information: All data are presented as mean ± SD. The two‐tailed Mann–Whitney U‐test was used for (B), (C), (P), (Q), (T), and (U). n represents biological replicates. **P < 0.01.

Figure EV2. mRNA distributions of CiTLR5a and CiTLR5b in blood cells of grass carp

mRNA expression in three types of grass carp blood cells was examined by qRT–PCR (n = 4). EF1α gene was employed as an internal control. Data information: All data are presented as mean ± SD. n represents biological replicates.

Figure EV3. Examine the interference efficiencies of CiTLR5a and CiTLR5b siRNAs

1. A, B

   Specific siRNAs of CiTLR5a and CiTLR5b along with negative control si.C were transfected into CIK cells for 12 h, respectively. The cells were harvested for qRT–PCR to quantify the expression levels of CiTLR5a (A) and CiTLR5b (B) (n = 4).

Data information: All data are presented as mean ± SD. The two‐tailed Mann–Whitney U‐test was used for (A) and (B). n represents biological replicates. **P < 0.01, ***P < 0.001.

Figure 3. CiTLR5a homodimer slightly promotes, whereas CiTLR5b homodimer potently restricts GCRV replication in CIK cells

1. A

   Effects of the indicated constructs on IFN1 and IFN3 promoter activities after GCRV challenge were measured by DLR assays in CIK cells (n = 3).

2. B

   Standard plaque assays in CIK cells with indicated overexpressing constructs or siRNAs and different dilution of GCRV.

3. C, D

   Relative plaque intensity ratios of (B) were analyzed by ImageJ (n = 3).

4. E, F

   CIK cells overexpressing (E) or interfering (F) were infected with GCRV at an MOI = 0.1, and supernatants were collected at 48 h for viral titer assays by TCID₅₀ (n = 3).

5. G, H

   CIK cells were treated with the supernatants for titer assays for 24 h, and mRNA expression of GCRV VP4, VP5, VP6, VP7, and NS38 genes was quantified by qRT–PCR (n = 4).

6. I, J

   Antiviral activities (cell viability) of the indicated constructs or siRNAs were measured by MTT assays in CIK cells at deferent time points post GCRV challenge (n = 4).

7. K, L

   mRNA expression of the antiviral regulator and effector molecules IFN1, Mx2, Gig1, and IL‐22 were quantified by qRT–PCR in overexpressing (K) or interfering (L) CIK cells post GCRV challenge (n = 4).

8. M, N

   mRNA expression of IFN1, IFN3, NF‐κB1, and NF‐κB2 in spleen tissue was investigated by qRT–PCR after RNAi in vivo for 24 h, and A. hydrophila or GCRV challenge for 24 h (n = 6).

9. O, P

   Protein expression of IFN1 (O) and NF‐κB1 (P) in spleen tissue was investigated by immunoblotting (IB) after RNAi in vivo for 24 h, and A. hydrophila or GCRV challenge for 24 h (n = 3).

10. Q

    Viral titers in spleen homogenates were examined with gene knockdown in vivo for 24 h and GCRV challenge for 24 h in CIK cells (n = 6).

Data information: All data are presented as mean ± SD. The two‐tailed Mann–Whitney U‐test was used for (A) and (C–Q). n represents biological replicates. *P < 0.05, **P < 0.01, ***P < 0.001.

Figure 4. Effects of different ectodomain and intracellular domain combinations of CiTLR5a and CiTLR5b on IFN1 and NF‐κB2 promoter activities

1. A

   Schematic representations of wild type (WT) and chimeric CiTLR5a and CiTLR5b constructs. ECD indicates ectodomain, TM represents the transmembrane region, and ICD stands for intracellular domain.

2. B, C

   FHM cells were transfected with the indicated combination receptors for 24 h, stimulated with poly(I:C) (B) or FlaC (C) for 24 h. IFN1 (B) and NF‐κB2 (C) promoter activities were examined by DLR assays (n = 3).

Data information: All data are presented as mean ± SD. The two‐tailed Mann–Whitney U‐test was used for (B) and (C). n represents biological replicates. **P < 0.01.

Figure 5. CiTLR5a and CiTLR5b share conserved flagellin‐binding regions and sites

1. A

   Comparison of the structures and relevant sequence regions in CiTLR5a and CiTLR5b with counterparts in DrTLR5a and DrTLR5b. The two representative conserved amino acids in flagellin‐binding regions for the following studies were denoted with triangles. Identical amino acid residues were shown by asterisks (*), and high and low similarities were indicated with colons (:) and points (.), respectively.

2. B, C

   Effects of the two selected flagellin‐binding sites on NF‐κB2 promoter activity were investigated by site‐directed mutagenesis and DLR assays in FHM cells (n = 3).

3. D, E

   Lysates of FHM cells overexpressing mutants were incubated with FlaC for pull‐down assays and analyzed by immunoblotting (IB). Vector and wild type (WT) were used as controls. The images of (D) and (E) were one of the three independent biological experiments in Fig EV4A and B. Relative band intensity ratios were analyzed by ImageJ (n = 3).

Data information: All data are presented as mean ± SD. The two‐tailed Mann–Whitney U‐test was used for (B–E). n represents biological replicates. *P < 0.05, **P < 0.01.

Figure EV4. CiTLR5a and CiTLR5b mutants reduce the binding capacity to FlaC

1. A, B

   Lysates of FHM cells transfected with mutants of CiTLR5a (A) and CiTLR5b (B) were incubated with FlaC and analyzed by immunoblotting (IB). Vector and wild type (WT) were used as controls. Images were from three independent experiments. The images of Fig 5D and E were one of the three independent biological experiments in (A) and (B).

Figure 6. CiTLR5a‐G240 and CiTLR5b‐N547 are key poly(I:C) binding sites and play crucial roles in binding and signal transduction

1. A

   SWISS‐MODEL, AutoDock Vina, and PyMOL software were employed to predict and show the interaction patches and sites of CiTLR5a or CiTLR5b with poly(I:C).

2. B

   Schematic motif representations of CiTLR5a, CiTLR5b, and their region deletants.

3. C

   FHM cells were transfected with the indicated constructs; IFN1 promoter activity was detected by DLR assays (n = 3).

4. D

   Lysates of FHM cells overexpressing CiTLR5a and CiTLR5b (including region deletants) were incubated with biotin‐poly(I:C). Vector was used as a control. Pull down and immunoblotting (IB) were performed.

5. E

   FHM cells were transfected with the indicated constructs; NF‐κB2 promoter activity was detected by DLR assays (n = 3).

6. F, G

   FHM cells were transfected with site mutants of CiTLR5a and CiTLR5b and then stimulated with poly(I:C) or FlaC. Vector and wild type (WT) were used as controls. G272A and G270A for key flagellin‐binding sites in CiTLR5a and CiTLR5b were employed as specificity controls. IFN1 (F) and NF‐κB2 (G) promoter activities were examined by DLR assays (n = 3).

7. H–K

   Lysates of FHM cells overexpressing CiTLR5a and CiTLR5b site mutants were incubated with biotin‐poly(I:C). Vector and wild type were used as controls. Pull down and immunoblotting were carried out (H, J). The images of (H) and (J) were one of the three independent biological experiments in Fig EV5A and B. (I, K) Relative band intensity ratios of (H) and (J) were analyzed by ImageJ, respectively (n = 3).

Data information: All data are presented as mean ± SD. The two‐tailed Mann–Whitney U‐test was used for (C), (E–G), (I), and (K). n represents biological replicates. *P < 0.05, **P < 0.01.

Figure EV5. CiTLR5a‐G240A and CiTLR5b‐N547A reduce the poly(I:C)‐binding capacity (A), but not vice versa (B)

1. A, B

   Lysates of FHM cells transfected with mutants of CiTLR5a and CiTLR5b were incubated with biotin‐poly(I:C) and analyzed by immunoblotting (IB). Vector and wild type (WT) were used as controls. Images were from three independent experiments. The images of Fig 6H and J were one of the three independent biological experiments in (A) and (B).

Figure 7. N‐glycosylation is necessary for CiTLR5a and CiTLR5b‐mediated IFN1 and NF‐κB2 signal transduction, but not ligand binding

1. A

   Ribbon diagrams of the structure models of CiTLR5a and CiTLR5b ectodomains. The predicted N‐glycosylation sites were labeled. The conserved Asn (N) residues between CiTLR5a and CiTLR5b are shown in pink, the specific Asn residues in CiTLR5a and CiTLR5b are shown in blue and orange, respectively.

2. B

   Comparison of predicted N‐glycosylation sites between CiTLR5a and CiTLR5b. The conserved Asn residues were marked by triangles. The differentiated residues were denoted with gray background. “‐” represents non‐N‐glycosylation site in the corresponding sequence.

3. C

   Lysates of FHM cells overexpressing CiTLR5a or CiTLR5b were, respectively, treated with Endo H and analyzed by immunoblotting (IB) with anti‐HA antibody.

4. D, E

   Effects of tunicamycin (5 μg/ml) on CiTLR5a‐ and CiTLR5b‐mediated NF‐κB2 (D) and IFN1 (E) signaling were investigated by DLR assays in FHM cells (n = 3). HsTLR5 and CiMDA5 were employed as positive controls, respectively.

5. F

   Mutants with the predicted N‐glycosylation sites were transfected into FHM cells and examined via immunoblotting to show the change of molecular weight.

6. G, H

   FHM cells were, respectively, transfected with mutants and the indicated receptor combinations and then stimulated with poly(I:C) or FlaC. Vector and wild type (WT) were used as controls. IFN1 (G) and NF‐κB2 (H) promoter activities were examined by DLR assays (n = 3).

7. I–L

   Lysates of FHM cells overexpressing CiTLR5a or CiTLR5b mutants were, respectively, incubated with biotin‐poly(I:C) (I) or FlaC (K). Vector and wild type were used as controls. Pull‐down and immunoblotting assays were conducted (I, K). The images of (I) and (K) were one of the three independent biological experiments in Appendix Fig S4A and B. (J, L) Relative band intensity ratios of (I) and (K) were analyzed by ImageJ (n = 3).

Data information: All data are presented as mean ± SD. The two‐tailed Mann–Whitney U‐test was used for (D), (E), (G), (H), (J), and (L). n represents biological replicates. *P < 0.05, **P < 0.01, ***P < 0.001.

Figure 8. M/M type TLR5 senses poly(I:C) but not other types

1. A–E

   Lysates of FHM cells overexpressing LoTLR5 (A), DrTLR5a (B), DrTLR5b (B), IpTLR5‐1 (C), IpTLR5‐2 (C), IpTLR5S (C), OmTLR5M (D), OmTLR5S (D), and HsTLR5 (E) were, respectively, incubated with FlaC‐His (1 μg/ml) at pH 5.0 or pH 7.4. Complexes were pulled down by anti‐GFP antibody coupling beads and analyzed by immunoblotting (IB) with anti‐His antibody.

2. F–J

   Lysates of FHM cells overexpressing LoTLR5 (F), DrTLR5a (G), DrTLR5b (G), IpTLR5‐1 (H), IpTLR5‐2 (H), IpTLR5S (H), OmTLR5M (I), OmTLR5S (I), and HsTLR5 (J) were, respectively, incubated with biotin‐poly(I:C) (1 μg/ml) at pH 5.0 or pH 7.4. Complexes were pulled down by streptavidin beads and analyzed via immunoblotting with anti‐GFP antibody.

3. K–M

   Effects of DrTLR5a and DrTLR5b constructs on IFN1 (K), IFN3 (L), and NF‐κB2 (M) promoter activities were measured by DLR assays in FHM cells (n = 3).

4. N, O

   Lysates of FHM cells overexpressing DrTLR5a and DrTLR5b mutants were incubated with biotin‐poly(I:C) and analyzed by immunoblotting (N), and relative band intensity ratios were analyzed by ImageJ (O) (n = 3). Vector and wild type (WT) were used as controls. The images of (N) were one of the three independent biological experiments in Appendix Fig S6A and B.

5. P

   Summarized models of TLR5‐ligand interaction and immune responses. Upper panel: The solid lines represent that ligand can be recognized by the corresponding types of TLR5, and the dotted lines represent that ligand cannot bind to the corresponding types of TLR5. Lower panel: The arrows indicate that different combinations of duplicated membrane TLR5 promote signal transduction after stimulation by the corresponding ligand; the dotted lines indicate that different combinations bind the corresponding ligand (flagellin) but do not transmit signals. The “⊥” symbol indicates that different combinations slightly suppress signal transduction after stimulation by poly(I:C).

6. Q

   Representative TLR5 protein sequences were employed to construct the phylogenetic tree by neighbor‐joining method within the MEGA11 program. The tree was bootstrapped 10,000 times.

7. R

   Summarized model of TLR5 evolution in vertebrates.

Data information: All data are presented as mean ± SD. The two‐tailed Mann–Whitney U‐test was used for (K–M) and (O). n represents biological replicates. *P < 0.05, **P < 0.01.