Transcriptomic Analysis and C-Terminal Epitope Tagging Reveal Differential Processing and Signaling of Endogenous TLR3 and TLR7

Toll-like receptor (TLR) signaling is critical for defense against pathogenic infection, as well as for modulating tissue development. Activation of different TLRs triggers common inflammatory responses such as cytokine induction. Here, we reveal differential impacts of TLR3 and TLR7 signaling on transcriptomic profiles in bone marrow-derived macrophages (BMDMs). Apart from self-regulation, TLR3, but not TLR7, induced expression of other TLRs, suggesting that TLR3 activation globally enhances innate immunity. Moreover, we observed diverse influences of TLR3 and TLR7 signaling on genes involved in methylation, caspase and autophagy pathways. We compared endogenous TLR3 and TLR7 by using CRISPR/Cas9 technology to knock in a dual Myc-HA tag at the 3’ ends of mouse Tlr3 and Tlr7. Using anti-HA antibodies to detect endogenous tagged TLR3 and TLR7, we found that both TLRs display differential tissue expression and posttranslational modifications. C-terminal tagging did not impair TLR3 activity. However, it disrupted the interaction between TLR7 and myeloid differentiation primary response 88 (MYD88), the Tir domain-containing adaptor of TLR7, which blocked its downstream signaling necessary to trigger cytokine and chemokine expression. Our study demonstrates different properties for TLR3 and TLR7, and also provides useful mouse models for further investigation of these two RNA-sensing TLRs.


INTRODUCTION
Innate immunity is the first line of defense for eliminating pathogens. To detect diverse pathogens, hosts have evolved various pattern recognition receptors (PRRs) to recognize foreign molecules derived from pathogens (1,2). Toll-like receptors (TLRs), some of the best studied PRRs, are critical for detecting bacterial and viral molecules and for triggering expression of downstream target genes, including cytokines and chemokines, to activate an inflammatory response and remove pathogens (3)(4)(5).
Of the 13 recognized TLRs, TLR3, TLR7, TLR8 and TLR9 are known as nucleic acid-sensing TLRs (4,6,7). TLR3 responds to double-stranded RNA (dsRNA), whereas TLR7 and TLR8 both recognize single-stranded RNA (ssRNA) (8)(9)(10). TLR9 binds to demethylated DNA, mainly derived from bacteria and viruses (11). In addition to detecting the DNA or RNA of pathogens, nucleic acid-sensing TLRs also recognize endogenous damage/ danger signals such as self mRNA and DNA derived from dead cells from injured tissues (4,12,13), as well as micro RNAs (miRNAs) delivered by exosomes (14). Nucleic acid-sensing TLRs are subjected to multiple regulatory mechanisms, such as intracellular localization, proteolytic processing, as well as ligand processing and recognition, which together control their activation by the partially digested nucleic acids from pathogens and other endogenous ligands (6,7,15).
After translation of nucleic acid-sensing Tlr mRNAs, the trafficking chaperone Unc93b1 controls their transport from the endoplasmic reticulum (ER) to endosomal compartments where they are proteolyzed into N-and C-terminal fragments (NTF and CTF) (6,16,17). The NTF and CTF of nucleic acidsensing TLRs are still associated with each other for ligand recognition and signal transduction (18)(19)(20). Downstream signaling of TLRs is mediated by interactions between the cytoplasmic Tir domain of CTF and other Tir domaincontaining adaptors (4,5). The two best-studied Tir domaincontaining adaptor proteins for TLRs are MYD88 and TIR domain-containing adapter inducing interferon-b (TRIF; also known as TICAM-1). Upon ligand binding, TLRs form a signalosome via interaction with MYD88 or TRIF to trigger cytokine expression (4,5). In general, the association between TLRs and their downstream adaptor protein is mediated by interactions between the Tir domains, except for the interaction between TLR3 and MYD88 (4,21). Instead of interacting with a Tir domain, TLR3 interacts with the death domain of MYD88 (21). Thus, when MYD88 binds TLR3, the Tir domain of MYD88 is still free to interact with other Tir domaincontaining molecules. The MYD88-containing TLR3 signalosome is likely to be different from other TLRs. Consistent with this speculation, TLR3 acts via MYD88 to control neuronal morphology, but TLR3 also interacts with TRIF to induce cytokine expression in cultured neurons (21). Consequently, compared with other TLRs, TLR3 likely displays distinct protein-protein interactions with downstream signaling molecules, although more structure analysis is required to conclude this point.
Our previous study indicated that TLR3, TLR7 and TLR8 cellautonomously control expression of distinct downstream target genes to regulate neuronal morphology (12,21,22). However, it is unclear whether these nucleic acid-sensing TLRs also have different target genes in non-neuronal cells, such as macrophages. Thus, we analyzed these nucleic acid-sensing TLRs by comparison their properties side-by-side in nonneuronal cells. Here, we first applied RNA-seq and transcriptomic profiling to confirm that downstream target genes of TLR3 and TLR7 are not identical, but partially overlapping in BMDMs. We then used CRISPR/Cas9 technology to dually tag Myc and HA cassettes at the Cterminal ends of endogenous TLR3 and TLR7 in mice. Anti-HA tag antibodies allowed us to analyze both TLR3 and TLR7 in immunoblotting, immunoprecipitation and immunostaining. Our comparison revealed the differences between TLR3 and TLR7 at multiple levels.

Animals
All mice were housed in the animal facility of the Institute of Molecular Biology, Academia Sinica, with a 12 h light/12 h dark cycle and controlled temperature and humidity. All animal experiments were performed with the approval of the Academia Sinica Institutional Animal Care and Utilization Committee and in strict accordance with its guidelines and those of the Council of Agriculture Guidebook for the Care and Use of Laboratory Animals (Protocol # 13-02-520). Both male and female mice were used randomly.
To generate Myc-HA-tagged Tlr7 mice (Tlr7 t/y and Tlr7 t/t are used hereafter to represent homozygous tagged knockin male and female mice, respectively), the Myc-HA tag sequences were inserted into exon 3 of Tlr7 before the stop codon by using CRISPR/Cas9-mediated knockin approach. The gRNA (5'ggcttatagtcaaatgttca) and ssODN (5'-cagaccttcatttatggaaacctttat gaaaacttcaggtaccaaggcatgtcctaggtggtgacattcttcagagagctaAGCG TAATCTGGAACATCGTATGGGTAaccgttCAGATCCTCTTC TGAGATGAGTTTTTGTTCTCTAGAgactgtttccttgaacatttga ctataagccacatgattgtctgtggtca) was used to incorporate an XbaI site (underlined capitals), as well as Myc (bold capitals) and HA (capitals) sequences.

RNA Sequencing (RNA-Seq)
BMDMs were treated with poly(I:C) or CL075 for 6 h. RNA-seq was performed as described previously (22,25). Briefly, total RNA was extracted using Trizol, followed by DNase I digestion as described above. RNA quality and quantifications were determined using an Agilent 2100 Bioanalyzer. The mRNA sequencing libraries were prepared using a Truseq Stranded mRNA kit (Illumina), and 75-76 cycle single-read sequencing was performed using the 500 High-output v2 (75 cycle) sequencing kit on an Illumina NextSeq500 instrument. After removing adaptor sequences, the quality of raw sequences was examined in FastQC software. Bioinformatics analysis was performed using two systems. The first analysis was conducted using CLC Genomics Workbench (http://www.clcbio.com). Raw sequencing reads were trimmed of low-quality sequences (Phred quality score of < 20) and of sequences with length < 25 bp. Sequencing reads were mapped to the mouse genome assembly (GRCm38.p6) from Ensembl with the following parameters: mismatches cost=2, insertion cost=3, deletion cost=3, m i n i m u m f r a c t i o n le n g t h = 0 . 9 , m i n i m u m fr a c t i o n similarity=0.9, and maximum hits per read=5. A Generalized Linear Model (GLM) and Wald Test were then used to analyze differential expression of genes and their statistical significance. The second system employed Hisat2 for sequence mapping (GRCm38.p6) with default settings, featureCounts to count reads, and edgeR (qlf model) for differential expression and statistical analysis. To select significantly altered genes, we first set a false discovery rate (FDR) of P < 0.05 and fold change > 1.5. To remove genes with background expression levels, the average transcripts per kilobase million (TPM) in one of the groups must be > 0.5. Genes meeting the criteria for both analysis pipelines (CLC and Hisat2-featureCounts-edgeR) were considered as significantly differentially-expressed genes. In addition, noncoding RNAs were removed before gene ontology (GO) analysis using Metascape (http://metascape.org/gp/index.html#/ main/step1). The raw datasets of the Tlr3 and Tlr7 target genes have been deposited online (NCBI GSE163347, https://www. ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE163347). Up-and down-regulated target genes of TLR3 and TLR7 are summarized in Supplementary Table 1.

Antibodies, Plasmids and Software
Detailed information on the antibodies used in this report is summarized in Supplementary Table 2. The mouse Tlr3 cDNA encoding TLR3 protein (a.a. 1-905) was amplified by using the primer pair Tlr3-1Fw 5' GGGATCCATGAAAGGGTGTTCC 3' and Tlr3-2715Rv 5' GCTGGTCGACATGTGCTGAATTCCG 3', and then subcloned into pGw1-cMyc vector at the BglII and SalI sites to generate a TLR3-Myc expression construct. The mouse Tlr7 cDNA encoding TLR7 protein (a.a. 1-1050) was amplified by using the primer pair Tlr7-1Fw 5' GGGATCCGCCACCA TGGTGTTTTCG 3' and Tlr7-3150Rv 5' GCTGGTCGACGACT GTTTCCTTGAAC 3', and then subcloned into pGw1-cMyc vector at the BglII and SalI sites to generate a TLR7-Myc expression construct. All software used in this report is listed in Supplementary Table 3.

Immunoprecipitation and Immunoblotting
For immunoprecipitation (IP), WT, Tlr3 t/t and Tlr7 t/y mouse brains and spleens were harvested and homogenized using a dounce homogenizer (20 passes) in ice-cold Tris-buffered saline (TBS; containing 20 mM Tris (pH7.4), 150 mM NaCl, 1 mM bglycerophosphate, 1 mM PMSF, 1 mg/ml Aprotinin, 1 mg/ml Leupeptin and 1 mg/ml pepstatin). Proteins were further extracted by adding 0.5% Triton X-100 and gentle mixing at 4°C for 30 min. After removing cell debris by centrifugation, 250 mg lysate was incubated with 1 mg TLR3 (PaT3), TLR7 (A94B10), or HA antibody and 10 ml protein A/G slurry (1:1) at 4°C for 2-3 h. The resulting protein A/G resin was washed three times with TBS containing 0.5% Triton X-100 and then twice with 50 mM Tris (pH7.4). The IP complex was eluted by adding 1X SDSsample buffer and boiling for 5 min. The samples were then subjected to immunoblotting analysis. To examine TLR7 signaling complex in BMDMs, WT and Tlr7 t/y BMDMs were treated with 0.5 mg/ml R848 for 30 and 60 min at 37°C. After rinsing with cold PBS, the BMDMs were lysed with cold TBS containing 0.5% Triton X-100 and kept on ice for 10 min. After centrifugation, the BMDM extract was then used to carry out IP using TLR7 antibody (A94B10), as described above. Immunoblotting was performed as described previously (27).
All full-size images of immunoblots are available in Supplementary Figures 8 and 9.

Immunofluorescence Staining
Immunostaining of BMDMs and mixed glial cells was performed as described previously (27). Immunofluorescent images of cells were visualized at room temperature using a confocal microscope (LSM 700, Zeiss) equipped with a 63×/NA 1.4 objective lens (Plan-Apochromat) and Zen acquisition and analysis software (Zeiss). The images were processed using Photoshop (Adobe) with minimal adjustment of brightness or contrast applied to the entire images.

RNA Extraction and Quantitative Real Time-PCR (RT-PCR)
To detect gene expression in BMDMs and glial cells, primary cell cultures were treated with 10 mg/ml poly(I:C) or 4 mM CL075 for 6 h before harvesting using TRIzol reagent. To quantify gene expression in spleen, adult WT and Tlr3 t/t mice received an intraperitoneal injection of saline or 5 mg/kg poly(I:C). Six hours later, mouse spleens were harvested for total RNA extraction using TRIzol according to the manufacturer's instructions (Invitrogen), followed by DNase I digestion (New England BioLabs). Reverse transcription and quantitative RT-PCR analysis was performed using the Transcriptor First Strand cDNA Synthesis Kit (Roche) with an oligo(dT)18 primer and the Universal ProbeLibrary probes (UPL; Roche) system. Primer sets and probe numbers for selected genes from the results of RNA-seq and internal control Hprt are listed in Supplementary  Table 4. Relative expression levels of each gene were normalized to the levels of Hprt measured at the same time on the same reaction plate. Brain and spleen of adult WT and Tlr7 t/y mice were also harvested and subjected to RNA extraction to enable quantitative RT-PCR analysis of Tlr7 and Tlr8 expression levels.

Statistical Analysis
Statistical analyses were performed using GraphPad Prism 8.0 software. Experiments were performed blind by relabeling the samples with the assistance of other laboratory members. The majority of our data met the assumptions (normal distribution) for statistical tests. For two-group experiments, an unpaired ttest was used. For experiments with more than two groups, oneway ANOVA with post hoc Bonferroni correction was applied. For two-factor experiments, two-way ANOVA with post hoc Bonferroni correction was used. Data are presented as mean ± SEM. P values < 0.05 were considered significant.

Different Downstream Target Genes of TLR3 and TLR7 in BMDMs
Our previous study has indicated that TLR3 and TLR7 regulate expressions of distinct downstream genes to control cell morphology in neurons (21,22). Since TLR3 and TLR7 both are involved in anti-virus response (28,29), it is unclear whether TLR3 and TLR7 also trigger different transcriptomes in nonneuronal cells. To investigate this issue, we treated BMDMs with poly(I:C) or CL075 for 6 h to activate TLR3 and TLR7 respectively. The RNA samples were then subjected to RNAseq. We set a fold-change (FC) > 1.5 and a false discovery rate (FDR) < 0.05 as criteria to analyze the target genes of TLR3 and TLR7. Under these conditions, we identified 2778 and 3394 genes upregulated or downregulated, respectively, by TLR3 activation in BMDMs ( Figure 1A, Supplementary Table 1). Much fewer downstream target genes were identified for TLR7, with 1979 upregulated genes and 2372 downregulated genes ( Figure 1A and Supplementary Table 1). Notably, 1141 upregulated genes and 1629 downregulated genes were commonly shared between the TLR3 and TLR7 datasets ( Figure 1A and Supplementary Table 1). Gene ontology (GO) analysis using Metascape indicated that upregulated genes shared by the two proteins are highly relevant to innate immune responses, including inflammatory responses, cytokine production and signaling pathways, responses to viruses/ bacteria, IFNg and IFNb, and regulation of immune processes ( Figure 1B, upper panel, red bars; Supplementary Table 1). In terms of downregulated genes in common, the major GO terms related to DNA replication and damage responses and metabolic processes ( Figure 1B, upper panel, blue bars and Supplementary Table 1). Gene lists of these GOs are summarized in Supplementary Table 1. These findings are consistent with the functions of TLR3 and TLR7 in triggering innate immunity and the physiological consequence of attenuated metabolic activity during inflammation (30).
Next, we performed Ingenuity Pathway Analysis (IPA) to further dissect the involvement of common target genes in canonical signaling pathways (Supplementary Figure 1), which identified more than 400 such pathways (-log(p) > 1 and ratio > 0.125). We summarize each top 50 pathways for upand down-regulated target genes in Supplementary Figure 1. Details for five of those pathways are shown in Supplementary  Figures 2, 3, including the role of pattern recognition receptors in recognition of bacteria and viruses, TLR signaling, activation of IRF by cytosolic pattern recognition receptors, IL-6 signaling, and the coronavirus pathogenesis pathway. These results further suggest that TLR3 and TLR7 activation control some common transcriptional factors (such as NF-kB, IRF3, IRF7, C-JUN, HIF1a, NF-IL6 and ISGF3) to induce expression of inflammatory and anti-viral cytokines (including IL-1b, IL-6, IL-10, IL-12, IL-17, TNFa/b, type I IFN) and chemokines (such as CCL2 and CCL5). Expression of other pattern recognition receptors and their downstream signaling adaptors and kinases could also be increased upon TLR3 and TLR7 activation (Supplementary Figures 2, 3), suggesting positive feedback regulation of TLR3 and TLR7 activation. These analyses further suggest that TLR3 and TLR7 share some common downstream pathways to trigger innate immunity.
Apart from the shared target genes, TLR3 activation actually uniquely impacted the expression of 3402 genes (1637 upregulated and 1765 downregulated) ( Figure 1A). For TLR7, 838 genes were specifically upregulated and 743 were uniquely downregulated in BMDMs upon TLR7 activation ( Figure 1A). GO analysis of these uniquely impacted genes indicated that, in addition to common pathways, TLR3 and TLR7 activation differentially regulates diverse cellular responses and processes ( Figure 1B, middle and lower panels).
We further used a heat map to compare some of the common and uniquely impacted downstream genes of both TLR3 and TLR7, which included various cytokines and chemokines, other TLRs and their downstream adaptors and transcriptional factors, chromatin remodeling regulators, as well as caspase and autophagy pathways ( Figure 1C). Among the various cytokine and chemokine genes, all except Cxcl3 were upregulated by TLR3 or TLR7 activation, though to varying degrees ( Figure 1C). In terms of the different Tlr genes, apart from upregulating Tlr3 itself, TLR3 activation also increased expression levels of other Tlrs, apart from Tlr5 ( Figure 1C). In contrast, TLR7 activation only increased Tlr1, Tlr2 and Tlr6 expression, but it downregulated the expression of other Tlrs, including Tlr3, Tlr4, Tlr5, Tlr8, Tlr9, Tlr11, Tlr12, Tlr13 and Tlr7 itself ( Figure 1C). Moreover, TLR3 and TLR7 activations had opposing effects on Irf3, Fos, Stat4, Hdac2, Hdac7, Dnmt3a, Dnmt3l, Casp2, Casp3, Apaf1, Bid, Atg5 and Atg9a ( Figure 1C). Thus, our transcriptomic profiling has clearly indicated that activations of TLR3 and TLR7 in BMDMs  regulate the expression of shared and unique downstream genes. This outcome supports previous observations from neurons that endosomal TLRs, including TLR3 and TLR7, use differential downstream signaling pathways to control gene expression and cellular responses (22). We then performed quantitative RT-PCR to confirm the results of our RNA-seq analysis. We selected 14 target genes from the lists of TLR3 and TLR7 target genes ( Figure 1C) for further analysis. Overall, the results of quantitative RT-PCR were consistent with our RNA-seq data. For instance, the levels of Tlr2 and Myd88 were increased by both TLR3 and TLR7 activation. The levels of Hdac4 and Atg7 were reduced by both TLR3 and TLR7 activation. However, TLR3, but not TLR7, specifically increase the RNA levels of Tlr3, Tlr7, Tlr9, Irf7, Stat3, Hdac1 and Casp3 in BMDMs ( Figure 2). Thus, both RNA-seq and quantitative RT-PCR indicate that TLR3 and TLR7 activations differentially induce expression of downstream target genes in BMDMs.

Generation and Characterization of Myc-HA Knockin Tlr3 and Tlr7 Mice
We then tried to reanalyze the properties of TLR3 and TLR7 by comparing these two TLRs. We first deployed several commercially available TLR3 and TLR7 antibodies to detect the respective exogenous proteins in HEK293T cells, as well as the endogenous proteins in mouse spleen lysates, by immunoblotting analyses (Supplementary Figure 4). Among the three TLR3 antibodies we tested, only one antibody recognized overexpressed TLR3-Myc protein in HEK293T cells and none of them detected TLR3 in WT mouse spleen lysate (Supplementary Figure 4A). Only one of the tested TLR7 antibodies (eBioscience) recognized overexpressed TLR7-Myc, and it presented a stronger band at~60 kDa in WT mouse spleen lysates compared with Tlr7 knockout spleen (Supplementary Figure 4B). Unfortunately, this TLR7 antibody from eBioscience also detected several nonspecific bands and has been discontinued. It is difficult to examine expression profiles and the function of endogenous TLR3 and TLR7 proteins when validated antibodies are not available.
To overcome the antibody issue and to allow us to investigate expression and functions of the TLR3 and TLR7 proteins in vivo, we added epitope tags at the 3' end of the Tlr3 and Tlr7 genes to monitor expression of endogenous TLR3 and TLR7 proteins using commercial anti-tag antibodies. We applied a CRISPR/Cas9mediated knockin approach to insert a sequence containing both Myc and HA tags before the stop codons of the Tlr3 gene in C57BL/6 mice ( Figure 3A and Supplementary Figure 5A). Insertion of the dual Myc-HA tag was confirmed by genomic PCR and sequencing (Supplementary Figure 5A). Dual Myc-HA tagging did not seem to influence mouse growth or development as the appearance of knockin mice was comparable to their wild type (WT) littermates (Supplementary Figure 5B). We applied the same approach to generate and characterize knockin mice with a Myc-HA tag at the TLR7 C-terminal tail ( Figure 3A and Supplementary Figures 5C, D). Hereafter, for simplicity, we name these knockin models as Tlr3 t/t and Tlr7 t/y (male) or Tlr7 t/t (female) mice (where t indicates the dual Myc-HA tag).
Next, we examined the expression and distribution of TLR3 and TLR7 in our Myc-HA-tagged knockin mouse lines using anti-HA and anti-Myc tag antibodies. Hereafter, we denote these dual-tagged TLR3 and TLR7 proteins as TLR3-MH and TLR7-MH, respectively. FIGURE 2 | Quantitative RT-PCR confirms altered gene expression in response to TLR3 or TLR7 activation. Poly(I:C) and CL075 were used to activate TLR3 or TLR7, respectively, in WT BMDMs. The expression levels of indicated genes were determined by quantitative RT-PCR and normalized against internal control Hprt. Data are represented as mean ± SEM (error bars). Each dot indicates the data of one independent experiment. *P < 0.05; **P < 0.01; ***P < 0.001.  We tested commercially available anti-HA tag antibodies-namely rabbit monoclonal C29F4, rat monoclonal 3F10 and mouse monoclonal 16B12 antibodies-as well as one anti-Myc tag antibody (mouse monoclonal 9B11), to detect these recombinant proteins (Supplementary Table 2). Among them, anti-HA antibody 3F10 performed best in detecting TLR3-MH or TLR7-MH in immunoblotting and immunoprecipitation assays ( Figure 3 and Supplementary Table 2), despite some non-specific bands appearing in immunoblots of total lysates from brain and heart ( Figure 3B). Anti-HA antibody C29F4 performed well in immunofluorescence staining, immunoprecipitation, and immunoblotting analysis of BMDMs lysates, whereas anti-HA antibody 16B12 was effective in immunofluorescence staining (Supplementary Table 2). In contrast, anti-Myc antibody 9B11 proved ineffective at detecting endogenous TLR3-MH as it insufficiently recognized the tagged protein and displayed multiple nonspecific bands in mouse tissue lysates (Supplementary Figure 6). Thus, by using specific anti-HA antibodies, our Myc-HA knockin mice are potentially useful for investigating the expression, distribution and complex formation of endogenous TLR3 and TLR7 both in vivo and in vitro.

Expression, Cleavage and Glycosylation of TLR3-MH Protein In Vivo
Since Tlr3 mRNA levels in mouse cells/tissues have been revealed previously (31-33), we first examined the expression of TLR3-MH proteins in different organs of Tlr3 t/t mice by immunoblotting, which generated four important results. First, unlike in WT mice, we detected in varying amounts a protein of~70 kDa using anti-HA antibody in all examined organs of Tlr3 t/t mice ( Figure 3B). In addition, much lower HA immunoreactivity was observed at~130 kDa for some Tlr3 t/t organs, such as brain, lung and spleen ( Figure 3B). Since full-length (FL) TLR3 is cleaved into the NTF and CTF fragments upon transport into endosomes (18,20), the 70-kDa and 130-kDa proteins present in our Tlr3 t/t mouse tissue lysates likely represent the CTF and FL of TLR3-MH, respectively. Given that amounts of CTF was much higher than for FL TLR3-MH, this result implies that the TLR3-MH proteins are ready to be processed and transported after synthesis in vivo. Second, considering the organs we examined, TLR3-MH displays highest amounts in spleen and lung ( Figure 3B), which is consistent with TLR3's function in triggering the innate immune response in immune organs (such as spleen) and organs most exposed to the environment (such as lung). Apart from spleen and lung, we also observed quite high levels of TLR3-MH in brain, liver, kidney and testis ( Figure 3B). This result evidences different expression levels of TLR3 in various organs.
Third, the ratio of FL to CTF for TLR3-MH varies in different organs. For instance, even though amounts of the CTF fragment of TLR3-MH were comparable between spleen and lung, we noted higher amounts of FL TLR3-MH in the former ( Figure 3B). Similarly, when we compared levels in brain, liver and kidney, although CTF levels were lower in brain compared to liver and kidney, amounts of FL TLR3-MH were higher in brain than detected in liver and kidney, no matter whether we assayed crude lysates or immunoprecipitated products ( Figure 3C). These results imply that proteolytic processing of TLR3 varies in different organs.
Finally, the mobility, and thus size, of FL TLR3-MH in SDS-PAGE also varied across tested tissues ( Figures 3B, C, E), likely due to differential glycosylation. Since glycosylation of the TLR ectodomain is known to be critical for TLR trafficking, complex assembly, responses to ligand stimulation, and signal transduction (34)(35)(36), varied glycosylation patterns may influence TLR function in different organs.
We had added the dual Myc-HA tag to the C-terminal end of TLR3, so anti-HA antibody could not be used to monitor expression of the NTF fragment of TLR3-MH. However, we searched the literature and found a published TLR3 antibody, PaT3, which recognizes the N-terminal of TLR3 in immunoprecipitation and immunostaining assays (37) ( Figure 3D). The PaT3 antibody indeed precipitated FL TLR3-MH from Tlr3 t/t mouse brain lysates ( Figure 3E). In addition to FL TLR3-MH, we also found that the PaT3 immunoprecipitates contained large amounts of the CTF fragment, in fact >10-fold that of FL TLR3-MH ( Figure 3E). Since the PaT3 antibody does not work in immunoblotting (37), we were unable to monitor levels of NTF in lysates and immunoprecipitates. Nevertheless, the cleaved NTF and CTF fragments of TLR3 have been shown to still associate with each other in cells for ligand binding and signal transduction (20,37), so the presence of CTF in PaT3 immunoprecipitates is likely due to interactions between NTF and CTF.
Thus, the analyses using our Tlr3 t/t mice reveal four properties of endogenous TLR3, which were not identified previously. When combined with specific HA antibodies, our Tlr3 t/t mice are powerful tools for investigating expression of endogenous TLR3 in various tissues.

Expression and Distribution of TLR3-MH in Tlr3 t/t Primary Cultures of BMDMs and Microglial Cells
We examined the expression of TLR3-MH proteins in primary cultures of BMDMs and glial cells from Tlr3 t/t mice and their WT littermates. Similar to our results from mouse tissue lysates, we observed both FL and CTF of TLR3-MH in BMDMs and glial cells using HA antibody. The amounts of the cleaved CTF fragment were still much higher than for FL protein (Supplementary Figures 7A, B). We also monitored the expression pattern of TLR3-MH in these primary cultures by immunostaining. Using phalloidin to label Factin at the cell periphery, we found that TLR3-MH signal presented a punctate pattern in BMDMs (Supplementary Figure 7C). In microglia colabeled by IBA1 antibody, HA antibody also revealed a punctate signal of TLR3-MH proteins (Supplementary Figure 7D). These results indicate that TLR3-MH is located in intracellular compartments of primary cell cultures prepared from Tlr3 t/t mice.

Triggering Gene Expression by TLR3-MH Activation
As shown in Figure 1, TLR3 activation induces expression of many downstream genes, including Tlr3 itself, as well as cytokines and chemokines (21,22,38). To confirm that TLR3-MH is responsive to a TLR3 ligand, we applied polyinosinic: polycytidylic acid (poly(I:C)), a synthetic dsRNA, to cultured BMDMs and glial cells for 6 and 24 h and monitored TLR3 expression ( Figure 4). Compared with control groups, poly(I:C) treatment for 6 h noticeably enhanced FL TLR3-MH protein levels ( Figure 4A). Treatment for 24 h further increased the level of the CTF fragment ( Figure 4A). TLR3-MH glycosylation levels in BMDMs remarkably differed between the 6-and 24-h treatments ( Figure 4A). Immunostaining also revealed a robust increase in TLR3-MH levels in BMDMs after 24 h of poly(I:C) stimulation ( Figure 4B).
In glial cell cultures, we also observed clearly enhanced TLR3-MH expression upon poly(I:C) treatment, but the proteolysis and glycosylation patterns were different from that in poly(I:C)treated BMDMs ( Figure 4C). FL TLR3-MH proteins rapidly accumulated 6 h after poly(I:C) treatment and had been cleaved to generate the CTF fragment within 24 h of poly(I:C) stimulation ( Figure 4C). Fluorescence staining also indicated strong expression of TLR3-MH proteins in both microglia and astrocytes 24 h after poly(I:C) treatment ( Figure 4D). We noticed that some GFAP-positive astrocytes presented a strong response to poly(I:C), resulting in high-level expression of TLR3-MH ( Figure 4D), whereas others displayed a low level response, even after 24 h of treatment ( Figure 4D, white asterisk). This outcome indicates that there were at least two different types of astrocytes in our glial cell cultures in terms of responsiveness to poly(I:C) stimulation.
We further employed quantitative RT-PCR to investigate the response of TLR3-MH to poly(I:C) stimulation. First, we intraperitoneally injected poly(I:C) or vehicle control into Tlr3 t/t mice. Six hours later, spleens were harvested for analysis. Compared with vehicle control, poly(I:C) treatment significantly increased levels in Tlr3 t/t mice of Tlr3 RNA as well as of different cytokines, including Il-6, Tnfa and Ifnb ( Figure 5A). In cultured BMDMs and glial cells, poly(I:C) treatment for 6 h also enhanced transcript levels of Tlr3, Il-6, Tnfa and Ifnb (Figures 5B, C). We further investigated whether TLR3-MH activation also regulates expression of TLR3 target genes identified from transcriptomic profiling in WT cells (Figures 1, 2). Consistent with the upregulation of TLR3 and cytokines, TLR3-MH activation also increased expression of Myd88, Irf7, Hdac1 and Casp3 but reduced the levels of Mettl3 and Hdac4 ( Figure 5D). These results indicate that TLR3-MH is functional in terms of delivering its immune signal and inducing gene expression, including of Tlr3 itself and downstream cytokines. Thus, our Tlr3 t/t mice can be deployed to track TLR3 expression and function in vivo and in vitro.

Failure of Inducing Cytokine Expression by TLR7-MH Activation
We then applied the same approaches to characterize Tlr7 t/y (male) and/or Tlr7 t/t (female) mice. Similar to our observations for TLR3-MH, we detected FL and the cleaved CTF fragment of  (Figures 6A, B). However, unlike TLR3-MH, TLR7-MH was predominantly expressed in spleen rather than in the other organs we examined ( Figure 6A). Comparing TLR3-MH and TLR7-MH proteins in Tlr3 t/t and Tlr7 t/y spleen lysates, we noted that amounts of the CTF fragments of TLR3-MH and TLR7-MH were similar, whereas levels of FL TLR7-MH were much higher than for FL TLR3-MH in both lysates and immunoprecipitates ( Figure 6C). These results indicate that proteolytic processing of TLR3 and TLR7 could differ.
Immunofluorescence staining also revealed an intracellular punctate pattern of TLR7-MH in BMDMs and glial cells ( Figure 6D), which is similar to the staining pattern observed for TLR3-MH ( Figures 4B, D and Supplementary Figures 7C, D). However, unlike TLR3-MH, TLR7-MH did not respond to its agonist CL075, a thiazoloquinolone derivative (12,22). To activate mouse TLR7, we added CL075 into primary cultures of BMDMs or glial cells for 6 h, but found that it did not induce expression of Il-6, Tnfa, Ccl3, or Il-1b in Tlr7 t/y cells (Figures 6E, F). In contrast to Tlr7 t/y cells, CL075 effectively triggered expression of the Il-6, Tnfa, Il-1b and Ccl3 genes in WT cells (Figures 6E, F). These results indicate that although TLR7-MH can be proteolytically processed into the NTF and CTF fragments, it lacks the ability to trigger downstream signaling and to subsequently induce cytokine expression.
Notably, unlike TLR3 (Figures 1, 5), TLR7 activation did not elicit a positive feedback mechanism to induce its own expression in either BMDMs or glial cells prepared from WT mice ( Figures 6E, F), suggesting that the downstream signaling and transcriptional regulatory mechanisms of TLR3 and TLR7 differ. Our previous studies have shown that Tlr7 knockout upregulates Tlr8 expression in brain or cultured neurons (12,22). Although TLR7-MH appears not to be functional in Tlr7 t/y mice, we found that Tlr8 transcripts were not increased in primary cell cultures or in brain/spleen tissues (Figures 6G, H). Since TLR7-MH signaling is blocked and TLR8 is not upregulated in Tlr7 t/y mice, they may serve as a more precise model for functional study of TLR7.
In conclusion, our results have shown that applying the same tagged knockin approach results in different outcomes for TLR3 and TLR7, with TLR3-MH remaining functional whereas TLR7-MH loses the ability to induce an immune response.

TLR7-MH Does Not Form Signalosomes With MYD-88
Next, we investigated how our C-terminal Myc-HA tagging of TLR7 impaired its activity. As for other endosomal TLRs, the NTF and CTF fragments of TLR7 remain linked to each other after cleavage and are required for TLR7 responses (17,19). Accordingly, we investigated if the NTF and CTF of TLR7-MH still interact with each other. We employed monoclonal TLR7 antibody (A94B10) (19) and polyclonal TLR7 antibody (eBioscience) that both recognize the N-terminal region of TLR7 ( Figure 7A) for immunoprecipitation and immunoblotting, respectively, allowing us to monitor TLR7 or TLR7-MH protein complexes in vivo. Our data show that A94B19 antibody precipitated both the FL and NTF of TLR7 and TLR7-MH from WT and Tlr7 t/y mouse spleen lysates, respectively ( Figure 7B). Similar to our observations for TLR3, we detected the cleaved CTF fragment of TLR7-MH by anti-HA antibody in the immunoprecipitated complex ( Figure 7C), suggesting that the NTF and CTF of TLR7-MH likely interact with each other. MYD88 is a critical signaling adaptor molecule for most TLRs, including TLR7 (1,4,39). Upon ligand binding, the cytoplasmic Tir domain of TLRs associates with MYD88 to recruit other downstream molecules, allowing the resulting signalosome to trigger cytokine expression (1,4,40). To test if C-terminal Myc-HA tagging influenced MYD88 binding, we treated WT and Tlr7 t/y BMDM primary cultures with a TLR7 ligand, R848, for 30 and 60 min. The cell lysates were then subjected to immunoprecipitation and immunoblotting analyses using TLR7 (A94B10) and MYD88 antibodies. We found that only WT TLR7 coimmunoprecipitated with MYD88 ( Figure 7D middle panel). In contrast, MYD88 was not present in TLR7-MH complexes, even though amounts of FL and CTF of TLR7-MH were higher than for WT TLR7 ( Figure 7D). Similar to our observations of spleen and brain lysates, the NTF and CTF fragments of TLR7-MH still remained associated in BMDM primary cultures ( Figure 7D). Thus, adding the Myc-HA tag at the C-terminal end of TLR7 disrupts signalosome complex formation, thereby impairing TLR7 responses.

DISCUSSION
In this report, we demonstrated the different properties of TLR3 and TLR7 in mouse tissues and primary cultured cells. First of all, transcriptomic profiling indicates that TLR3 and TLR7 regulate expression of distinct gene sets in BMDMs. We found that TLR3 activation generally increased the expression of all Tlrs (except Tlr5), Tir domain-containing adaptors (except Sarm1), and inflammation-related transcriptional factors (except Irf3) ( Figure 1C). In contrast, TLR7 activation generally downregulated the expression of Tlr genes, except for Tlr1, Tlr2 and Tlr6 ( Figure 1C). Thus, TLR3 activation globally upregulates the innate immune responses of BMDMs. However, TLR7 activation likely attenuates or restricts innate immunity mediated by other TLRs. It remains unclear how TLR7 activation specifically upregulates the expression of Tlr1, Tlr2 and Tlr6. These three Tlr genes all belong to the Tlr1 subfamily and are believed to have evolved via a series of independent gene duplications and subsequently diverged by positive selection (4,41). Thus, it is likely that the promoter and enhancer regions of these genes are all conserved and controlled by similar transcription factors. Since the TLR1 subfamily recognizes lipopeptides derived from microbes, it would also be interesting to explore the physiological relevance of enhanced lipopeptide recognition upon TLR7 activation.
In addition to TLRs, we also noticed that TLR3 activation greatly increased expression of Myd88 (~4-fold) but not Trif. For TLR7 activation, both Myd88 and Trif were increased around 1 to 2 fold (Figure 2). These results were relatively unexpected as TLR3 was auto-upregulated by TLR3 activation and TRIF is known to be the critical downstream adaptor of TLR3 to induce cytokine expression. It is unclear whether the existing TRIF is sufficient for overexpressed TLR3 to deliver the signaling. For TLR7 activation, it did not induce Tlr3 expression but increased Trif expression (Figure 2). It is interesting to further investigate whether TRIF has TLR3-independent function, such as regulating other TLRs' signaling (4).
Furthermore, we generated two knockin mouse lines displaying dual Myc-HA tagging at the C-terminal ends of the Tlr3 and Tlr7 genes. Using commercially-available anti-HA antibodies, we were able to establish specific expression profiles for endogenous TLR3 and TLR7 in vivo and also to visualize the subcellular distributions of TLR3 and TLR7 in different cell types. Moreover, we used anti-HA antibody to compare endogenous TLR3 and TLR7 proteins. Our data shows that TLR7 proteins are predominantly expressed in spleen, though lower amounts of that protein are still present in other organs. Apart from noticeably high levels of TLR3 being expressed in spleen and lung, considerable amounts were detected in most of the organs we examined. Although levels of cleaved fragments of both TLR3 and TLR7 were higher relative to uncleaved FL proteins, relative amounts of FL TLR7 are still greater than those of FL TLR3. In addition, proteolytic processing efficiency and posttranslational modification of TLR7 are comparable among different organs. In contrast, we found that TLR3 displayed differential posttranslational modification and proteolytic processing efficiency in different tissues. Our Tlr3 t/t and Tlr7 t/y mice enable investigation of the differences between TLR3 and TLR7 and facilitate assessments of endogenous TLR3 and TLR7 under more physiologically relevant conditions. Our C-terminal Myc-HA tagging had different consequences for TLR3 and TLR7. The TLR3-MH protein still responded to poly(I:C), which triggered subsequent expression of cytokines and chemokines. Thus, TLR3-MH is a functional protein and our Tlr3 t/t mice can be used for all kinds of TLR3-related studies. Unexpectedly, the same C-terminal Myc-HA tag disrupted the association of TLR7-MH with MYD88 and, upon ligand stimulation, downstream signaling was impaired. The Myc-HA-tagged Tlr7 mice can be considered a TLR7 loss-offunction animal model in which gene expression of nearby genes, such as Tlr8, is unaffected. It will be interesting to compare if there is any difference between Tlr7-knockout and Myc-HA tagged mice. It would also be intriguing to investigate how the extreme C-terminal tail of TLR7 contributes to MYD88 interaction.
Previous studies have generated reporter mice for Tlr2, Tlr4, Tlr5, Tlr7 and Tlr9 with a view to assessing TLR expression (42,43). In those cases, either a HA or Flag tag was placed at the 3' end of the respective Tlr genes, followed by an IRES sequence and various fluorescence proteins. The fluorescence proteins labeled TLR-expressing cells and the C-terminal HA or Flag tag could be used to detect the TLR proteins (42,43). In our study, we found that C-terminal tagging with a Myc-HA cassette impaired signaling by TLR7 but not by TLR3. Thus, we feel it is worth reinvestigating the aforementioned published reporter mouse lines to establish if the tagged TLRs respond normally to their ligands. It will also be interesting to compare the sensitivity of different TLRs to C-terminal tagging. If as for TLR7, other TLRs are sensitive to C-terminal tagging and lose their ability to deliver downstream signaling, it suggests that they may all form similar signalosomes for signal transduction. If contrasting results arise from such an experiment, it would suggest that the extreme C-terminal ends of different TLRs may contribute differentially to signal transduction.
Both TLR7 and TLR8 recognize ssRNA and regulate neuronal morphology and function (12,22). Our previous study showed that Tlr7 knockout upregulates Tlr8 expression in brain (12,22). This compensatory effect is brain-specific, and is not observed in spleen (12,22). Here, we report that TLR7-MH lacks the ability to deliver downstream signaling. Thus, our Tlr7 t/y mice still represent Tlr7 knockout mice even though they still express TLR7-MH protein. Interestingly, we found that Tlr8 transcripts were not increased in Tlr7 t/y mice, no matter whether we assessed BMDMs, glial cells or tissues. Although Tlr7 -/y knockout mice exhibit some behavioral abnormalities in terms of olfaction, anxiety, aggression and contextual fear memory (25), we cannot be sure if Tlr8 upregulation also influences the phenotype of Tlr7 knockout mice. Compared with Tlr7 -/y mice, the Tlr7 t/y mouse line we report herein may represent a good or even better genetic model for Tlr7 because it minimizes the potential interference contributed by Tlr8 upregulation. Investigations of the behavioral characteristics of Tlr7 t/y mice are warranted.
The proteolytic processing of TLRs has been studied for both endogenous proteins and in overexpression systems. For endogenous proteins, the proteolytic products of endogenous TLR3, TLR7, and TLR9 are much more abundant than the respective full length proteins in human retinal epithelial cell line RPE1, primary human monocyte-derived dendritic cells (MDDCs), and mouse BMDMs (18,37,43,44), which is consistent with our findings presented herein. In contrast, amounts of proteolyzed TLRs were found to be < 5% relative to FL TLRs in TLR-overexpressing HEK293 cells (19,20,45). Using our dual Myc-HA-tagged mouse lines, it is now relatively easy to prepare primary cell cultures for the study of TLR3 and TLR7 under more physiologically relevant conditions, greatly enabling investigations of TLR3 and TLR7 in the future.

ETHICS STATEMENT
The animal study was reviewed and approved by Academia Sinica Institutional Animal Care and Utilization Committee.