Molecular and functional characterization of Toll-like receptor (Tlr)1 and Tlr2 in common carp ( Cyprinus carpio ) Fish & Shell ﬁ sh Immunology

Toll-like receptors (TLRs) are fundamental components of innate immunity that play signi ﬁ cant roles in the defence against pathogen invasion. In this study, we present the molecular characterization of the full-length coding sequence of tlr1 , tlr2a and tlr2b from common carp ( Cyprinus carpio). Each is encoded within a single exon and contains a conserved number of leucine-rich repeats, a transmembrane region and an intracellular TIR domain for signalling. Indeed, sequence, phylogenetic and synteny analysis of carp tlr1 , tlr2a and tlr2b support that these genes are orthologues of mammalian TLR1 and TLR2. The tlr genes are expressed in various immune organs and cell types. Furthermore, the carp sequences exhibited a good three-dimensional ﬁ t with the heterodimer structure of human TLR1-TLR2, including the potential to bind to the ligand Pam 3 CSK 4 . This supports the possible formation of carp Tlr1-Tlr2 hetero- dimers. However, we were unable to demonstrate Tlr1/Tlr2-mediated ligand binding in transfected cell lines through NF- k B activation, despite showing the expression and co-localization of Tlr1 and Tlr2. We discuss possible limitations when studying ligand-speci ﬁ c activation of NF- k B after expression of Tlr1 and/or Tlr2 in human but also ﬁ sh cell lines and we propose alternative future strategies for studying ligand-binding properties of ﬁ sh Tlrs.


Introduction
Pattern recognition receptors recognize widely-conserved motifs of pathogens and are crucial for initiating immune responses against invading microorganisms. Toll-like receptors (TLRs) are a family of germline-encoded pattern recognition receptors and known to activate rapid inflammatory responses upon detection of their cognate ligands [1]. TLRs are type-I transmembrane proteins with numerous extracellular leucine-rich repeat (LRR) motifs which collectively form a horseshoe-shaped solenoid, which is responsible for ligand binding. The cytoplasmic part of each TLR contains a Toll/interleukin-1 receptor homology (TIR) domain which, upon dimerization of two TLRs, initiates a signalling cascade leading to activation of transcription factors such as NF-kB or AP-1, and subsequently to production of pro-inflammatory cytokines [2].
Most vertebrate genomes are recognized to have at least one gene representing each of the six major TLR families (TLR1, TLR3, TLR4, TLR5, TLR7 and TLR11) [3]. Also within the modern bony fish (Teleostei) the number of Tlr families generally is consistent with what is found for most (higher) vertebrates, although it is not unusual to find duplicated tlr genes due to several fish-specific wholegenome duplication events [4e8]. In addition to these duplications, some novel Tlrs seem to be "fish-specific", such as a soluble form of Tlr5, and Tlrs 18e27 [9,10], indicating that an expansion of Tlrs has occurred during the evolution of teleosts. To date, little is known about the ligand specificities of individual Tlrs in fish, since the LRR Abbreviations: BSA, bovine serum albumin; GFP, green fluorescent protein; IRAK1, interleukin-1 receptor-associated kinase 1; LRR, leucine-rich repeat; LTA, lipoteichoic acid; MYD88, myeloid differentiation primary response 88; NF-kB, nuclear factor kappa B; PAMP, pathogen-associated molecular pattern; Pam 3 CSK 4 , N-palmitoyl-S-(2,3-bis(palmitoyloxy)-(2RS)-propyl)-(R)-Cys-(S)-Ser-(S)-Lys 4 ; PBL, peripheral blood leukocytes; PBS, phosphate buffered saline; PE, phycoerythrin; PGN, peptidoglycan; RLU, relative light units; RT-qPCR, real-time quantitative polymerase chain reaction; TIR, toll/interleukin-1 receptor; TLR, Toll-like receptor; TRAF6, TNF receptor-associated factor 6; UTR, untranslated region; WGD, whole genome duplication. ectodomains of Tlrs are not always highly conserved and sequence information alone cannot infer functional properties [10]. In apparent contrast to the variation in the ectodomains, intracellular TIR domains of fish Tlrs appear highly conserved and downstream signalling via well-described molecules such as MyD88, Irak1 and Traf6 identified in several fish species, suggest a conserved mechanism of innate immune signalling could exist [11]. Yet, studies into ligand-binding properties of fish Tlrs are essential to characterize their exact function within the immune system of fish.
We previously identified and characterized common carp (Cyprinus carpio) Tlr2 [28,29]. Transfection of human HEK293 cells with carp tlr2 suggested the ability to bind the prototypical TLR2 ligands LTA, PGN and Pam 3 CSK 4 . Stimulation of carp macrophages with PGN induced tlr2 gene expression, MAPK-p38 phosphorylation and led to an increased production of nitrogen and oxygen radicals. Here, we present the identification of Tlr1 and molecular characterization of the mRNA and genomic structure of both tlr1 and tlr2 from common carp. We compare the gene expression of tlr1 and tlr2 in the same tissue samples and purified cell populations and describe our efforts to characterize the function of putative Tlr1/Tlr2 heterodimers by studying subcellular localization and ligand-binding properties. We discuss possible limitations when studying ligand-specific activation of NF-kB after overexpression of Tlr1 and/or Tlr2 in human but also fish cell lines and propose alternative future strategies for studying ligand-binding properties of fish Tlrs.

Animals
European common carp (Cyprinus carpio L.) were reared in the central fish facility Carus, at Wageningen University, Wageningen, The Netherlands. Fish were kept at 23 C in recirculating UV-treated tap water and fed pelleted dry food (Sniff, Soest, Germany) daily. R3ÂR8 carp are the hybrid offspring of a cross between fish of Polish origin (R3 strain) and Hungarian origin (R8 strain) [30]. Carp were between 9 and 11 months old at the start of the experiments. All studies were performed with approval from the local animal welfare committee (DEC) of Wageningen University.

Organ isolation
Carp were euthanized with 0.3 g/L tricaine methane sulfonate (TMS, Crescent Research Chemicals, Phoenix, AZ, USA) buffered with 0.6 g/L NaHCO 3 . Carp were bled from the caudal vein using a needle and syringe containing cRPMI medium (RPMI 1640 with 25 mM HEPES (Lonza, Basel, Switzerland) adjusted to an osmolality of 280 mOsm/kg with sterile water) containing 50 U/mL heparin (Leo Pharma, Ballerup, Denmark), 50 U/mL penicillin G (Sigma-Aldrich, St. Louis, MO, USA), and 50 mg/mL streptomycin sulphate (Sigma-Aldrich). For isolation of peripheral blood leukocytes (PBL), the heparinized blood was centrifuged at 100g for 5 min at 4 C and then another 5 min at 300g. The buffy coat was collected, carefully layered on Ficoll-Paque PLUS (GE Healthcare, Little Chalfont, UK) and centrifuged at 800g for 25 min at 4 C without brake. The leukocyte layer was collected and washed twice with cRPMI. The obtained PBL were stored at À80 C until used for RNA isolation. After bleeding the fish, the organs of interest were aseptically removed and immediately frozen in liquid nitrogen and stored at À80 C until used for RNA isolation.

Isolation of leukocyte subtypes
Carp leukocyte subtypes were isolated by density gradient separation and/or magnetic cell sorting using specific antibodies as described before for thrombocytes [31], granulocytes [32], B cells [59,60], and macrophages [33]. In short, PBL or single-cell suspensions derived from carp organs were incubated with primary mouse monoclonal antibody: WCL-6 for thrombocytes (from blood), TCL-BE8 for neutrophils (from mid kidney), WCI-12 for B cells (from blood), and WCL-15 for monocytes/macrophages (from spleen). After incubation and washing, cells were stained with phycoerythrin (PE)-conjugated goat anti-mouse secondary antibody. After washing and counting of cells, magnetic beads (anti-PE MicroBeads, Miltenyi Biotec GmbH, Bergisch Gladbach, Germany) were added and allowed to bind, before washing and magnetic separation on LS Midi Columns using a MidiMACS Separator (Miltenyi Biotec). Head kidney-derived macrophages were isolated and cultured as described by Joerink et al. [34].

RNA isolation
Total RNA from carp organs and leukocytes was extracted using the RNeasy Mini kit according to the manufacturer's protocol (Qiagen, Venlo, The Netherlands) including on-column DNase treatment with the RNase-free DNase set (Qiagen). Final elution was performed with 30 mL nuclease-free water. The integrity of the RNA was determined by agarose gel electrophoresis and the RNA quality and concentrations were assessed spectrophotometrically by measuring the absorbance at 260 nm and 280 nm (Nanodrop, Thermo Scientific, Waltham, MA, USA). RNA was stored at À80 C until use.

cDNA synthesis
Prior to cDNA synthesis, 500 ng-1 mg of total RNA was subjected to an additional DNase treatment by using DNase I Amplification Grade (Invitrogen, Carlsbad, CA, USA). Synthesis of cDNA was performed with Invitrogen's SuperScript III Reverse Transcriptase, according to the manufacturer's instructions. As control for genomic contamination, for each sample a reaction without SuperScript III Reverse Transcriptase was performed. cDNA samples were diluted 25 times in nuclease-free water before use as templates in real-time quantitative PCR experiments.

Real-time quantitative PCR
Real-time quantitative PCR (RT-qPCR) was performed in a Rotor-Gene 6000 with a 72-well rotor (Corbett Research, Qiagen) with the ABsolute QPCR SYBR Green Mix (Thermo Scientific) as detection chemistry as described previously [35]. All primers were from Eurogentec (Li ege, Belgium). RT-qPCR data were analysed by Rotor-Gene 6000 Series Software 1.7. The melting temperature and profile of the melting curves were used to validate the specificity of the amplification. The gene expression was analysed using the average amplification efficiency for each primer pair and the take-off value of each sample (as derived from the Comparative Quantitation Analysis of the Rotor-Gene 6000 Series Software 1.7). The relative gene expression was calculated as the average amplification efficiency to the power of the take-off value. The gene expression of the housekeeping gene 40s ribosomal protein s11 was used to normalize the data. The sequences of primers used in this study are given in Table 1.

Cloning full-length tlr1 from common carp
Carp head kidney leukocytes were obtained by density gradient separation. RNA was isolated as described above and then used as template in reverse transcription-PCR with degenerate primers designed on the basis of tlr1 sequence information from other teleost fish. RACE (rapid amplification of cDNA ends) was then performed to obtain sequence at 5 0 and 3 0 ends. This allowed for design of carp primers in 5 0 and 3 0 untranslated regions (UTR), thus enabling the amplification of full-length carp tlr1 sequence. For this, RNA from head kidney leukocytes was used as template in RT-PCR using the LongRange 2Step RT-PCR kit (Qiagen). Reverse transcription was done with an oligo-dT primer and the subsequent PCR was done with gene-specific primers located in 5 0 and 3 0 UTR (CycaTLR1_FL_FW1 and CycaTLR1_FL_RV4, see Table 1). The PCR product was subsequently cloned in pGEM-T Easy (Promega, Madison, WI, USA) and transformed into Escherichia coli JM109 competent cells (Promega). Clones were sequenced and sequence data were analysed with Sequencher version 4.10 (Gene Codes, Ann Arbor, MI, USA) and aligned with Clustal Omega (http://www.ebi. ac.uk/Tools/msa/clustalo/). A clone with the consensus sequence was used for further studies.

Bioinformatics
We obtained the coding sequence of the unique carp tlr1 gene after cloning (our sequence is identical to the recently deposited GenBank accession number LHQP01021877, Contig21898, cypCar_00044472, from whole genome shotgun sequencing). Cloning of tlr2 was previously reported by our group [28]. Now referred to as tlr2a, the sequence has been updated in GenBank (accession number FJ858800) and the automatic annotation of the recently deposited LHQP01006764, Contig6769, cypCar_00005269 has been adjusted. A second tlr2 sequence was identified in this study, predicted as genome sequence data (Bioproject PRJNA73579) [36], and partially confirmed as RNAseq data (unpublished data) and will be referred to as tlr2b (GenBank accession number LHQP01045997, Contig46054, cypCar_00039549 has been adjusted to reflect our manual annotation). Exon-intron structure was studied by multiple alignments and open reading frame predictions (FGENESH at http://linux1.softberry.com/berry.phtml? topic¼fgenesh&group¼programs&subgroup¼gfind, and GEN-SCAN at http://genes.mit.edu/GENSCAN.html). Nucleotide sequences were translated into protein sequence using the EMBL-EBI tool (http://www.ebi.ac.uk/Tools/st/) and these were aligned with Clustal Omega (http://www.ebi.ac.uk/Tools/msa/clustalo/). Molecular weights were calculated with ProtParam (http://web.expasy. org/protparam/). The protein sequences were examined for the presence of a signal peptide using SignalP (http://www.cbs.dtu.dk/ services/SignalP/) and transmembrane regions were predicted using TMHMM Server v. 2.0 (http://www.cbs.dtu.dk/services/ TMHMM/). Individual LRRs were identified by LRRfinder (http:// www.lrrfinder.com/) and manually according to previous descriptions [37,38]. Synteny analysis was performed on the basis of sequence information from Ensembl release 82 (http://www. ensembl.org/). The multiple sequence alignment for the phylogenetic tree was made with ClustalX 2.1 [39] and the tree was constructed using the Neighbour Joining method with the number of bootstrap trials set to 10000. The phylogenetic tree was visualized with MEGA6.

tlr1 and tlr2 expression plasmids
Each of the genes encoding the fluorescent proteins GFP and mCherry were cloned into pcDNA3 using the EcoRI and XbaI sites. An XhoI site and a short linker sequence of 15 nucleotides encoding the amino acids GGSGG was placed upstream of the GFP and mCherry sequence. Subsequently, the full-length carp tlr1 sequence Table 1 List of primers used in this study. Restriction sites are italicized, Kozak sequence is shown in bold, sequence encoding a FLAG-tag is single underlined, and His-tag is bolded and italicized. was modified from the above-mentioned pGEM-T Easy construct to include sequence to encode a FLAG-tag at the N-terminus (after the leader peptide), and to remove the stop codon, using primers TLR1_pcDNA3_FW1, TLR1_pcDNA3_FW2, TLR1_ pcDNA3_FLAG and TLR1_pcDNA3_RV1 (see Table 1). The modified tlr1 was then cloned between the BamHI site and the newly created XhoI site in pcDNA3, thus creating a fusion of tlr1-GFP and tlr1-mCherry each in pcDNA3. Given the high sequence similarity between tlr2a and tlr2b, for functional analysis, we proceeded with the previously characterized tlr2a construct [28]. This gene was sub-cloned into mCherry-pcDNA3 using primers TLR2_pcDNA3_FW1, TLR2_pcDNA3_FW2, TLR2_pcDNA3_His and TLR2_pcDNA3_RV1 (see Table 1) thereby creating a His-tagged fusion of tlr2a-mCherry. Furthermore, tlr1 was sub-cloned into pBI-CMV1, a plasmid with two multiple cloning sites (MCS), where tlr1 was cloned in MCS1 between BamHI and PvuII sites, and the GFP was no longer fused to the tlr sequence but subcloned in the MCS2 between EcoRI and XbaI sites. A similar construct was made for tlr2a. Isolation of transfection-grade plasmid DNA was performed with the S.N.A.P. MidiPrep Kit (Invitrogen) according to the manufacturer's instructions.
The Epithelioma papulosum cyprini (EPC) cell line, which is an adherent cell type derived from fathead minnow, was cultured at 27 C at 5% CO 2 in RPMI 1640 (Lonza) supplemented with 10% foetal bovine serum, 2 mM L-glutamine, 50 U/mL penicillin G, and 50 mg/mL streptomycin sulphate. EPC cells were seeded in six-well plates, 8 Â 10 5 cells/well, and allowed to adhere. EPC cells were transfected the following day with 2 mg tlr1-GFP-pcDNA3 using FuGENE HD Transfection Reagent (Promega) according to the manufacturer's instructions at a 1:3.5 ratio of plasmid:transfection reagent. Three days after transfection, EPC cells were detached with cold medium, fixed with 4% paraformaldehyde and washed with 1% BSA in PBS. Some samples were permeabilized with 0.1% Triton X-100 in 1%BSA/PBS while others were not permeabilized. Antibody staining was performed with mouse anti-FLAG primary antibody (1:200 dilution, Sigma-Aldrich) followed by Cy3-labelled donkey anti-mouse secondary antibody (1:100 dilution, Jackson Immu-noResearch, West Grove, PA, USA). Cells were mounted with Vectashield.
Cells were visualized with a Zeiss LSM-510 (Zeiss, Oberkochen, Germany) confocal laser scanning microscope with a Plan-Apochromat 63Â/1.4 oil immersion objective. Green fluorescent signal (GFP or Alexa Fluor 488) was excited with a 488 nm argon laser and detected using a band-pass filter (505e550 nm). Red fluorescence (mCherry protein or Cy3) was excited with a 543 nm helium-neon laser and detected using a long-pass filter (585 nm). Image processing was performed with ImageJ (http://imagej.nih. gov/ij/).

Structural model of heterodimer formation
The sequences of carp Tlr1, carp Tlr2a/b, human TLR1 and human TLR2 were used for sequence alignment using ClustalX [40]. Using these alignments and the available structure of the human TLR1/TLR2 complex, including a tri-acylated lipopeptide moiety (Pam 3 CSK 4 ) and all sugar chains and water molecules (PDB-id: 2z7x) as templates, structural models were obtained for the carp Tlr1-Tlr2a/b complexes using the Modeller program (version 9.12) [41]. Thirty comparative models were generated, after which the model with lowest corresponding DOPE score [41] was selected for image generation.

Ligand binding
HeLa-57a cells (stably transfected with NF-kB luciferase reporter [42]) were cultured in DMEM with 5% foetal bovine serum (Bodinco, Alkmaar, The Netherlands) at 37 C in 10% CO 2 atmosphere. Cells were transfected in 6-well plates with a combination of carp tlr1 and tlr2 plasmids, or human tlr1 and tlr2 [43] as positive control. All plasmids were transfected using FuGENE HD (Roche, Basel, Switzerland) in a DNA:FuGENE ratio of 1:3. A total of 2 mg plasmid DNA was transfected; when multiple plasmids were combined, equal amounts of each plasmid were used. To all combinations (except empty vector) a plasmid encoding human CD14 [43] was added. Twenty-four hours after transfection cells were redistributed from 6-well plates to 48-well plates and left to attach for 24 h. Cells were then stimulated with the following ligands: Pam 3 CSK 4 (100 ng/mL), peptidoglycan (PGN, 10 mg/mL), and lipoteichoic acid (LTA, 1 mg/mL) which were purchased from Invivogen, San Diego, CA, USA. Stimulation was performed for 5 h, after which the cells were washed twice with PBS and then lysed with 100 mL reporter lysis buffer (Promega) and frozen at À80 C. After 1 h cells were thawed at room temperature and 20 mL cell lysate was mixed with 50 mL Luciferase-6-reagent (Promega). Luminescence was measured with a Turner Designs TD20/20 luminometer. Results are expressed as relative light units (RLU).

Identification and characterization of carp tlr1 and tlr2 genes
Information on tlr1 and tlr2 genes is available for some fish species but tlr1 orthologues had not yet been identified in carp. In the process of identifying new genes in carp, it is often useful to compare with the closely related zebrafish, a species which has a very well annotated genome. Usually, the presence of two genes in carp versus a single gene in the genome of zebrafish is expected based on an additional whole genome duplication (WGD) event that has taken place in carp [36]. Thus, although we previously reported on the presence of a tlr2 orthologue in carp, given the additional WGD event in the carp lineage, a second copy of the tlr2 gene was expected.
Making use of a conventional cloning approach combined with information from the recently annotated carp genome we identified a single complete tlr1 gene (Fig. 1) and an additional, but truncated, tlr1 sequence in the carp genome (not shown). The fulllength carp tlr1 is composed of a single exon, similar to channel catfish, Tetraodon and fugu tlr1 genes. Human and mouse TLR1 are composed of four exons, but the entire coding region is contained in a single exon. Also for zebrafish, tlr1 is divided over multiple (two) exons with the coding region contained in a single exon. The fulllength coding sequence of carp tlr1 is 2394 bp and translates into a Tlr1 protein of 797 aa with predicted molecular weight of 91 kDa. Tlr1 has 20 leucine-rich repeats (LRRs) and a C-terminal LRR (LRRCT), a transmembrane region and a highly conserved Toll/IL-1R (TIR) domain. Similar to other Tlr1 molecules [19], there is no Nterminal LRR (LRRNT).
We previously described a carp tlr2 gene [28]. Upon further investigation of the carp genome we now identified an additional tlr2 gene in the carp genome (Fig. 2). Both carp tlr2 genes are composed of a single exon, similar to the human, zebrafish and channel catfish tlr2 genes, whereas in some other fish species tlr2 genes are encoded by multiple exons, e.g. Tetraodon and fugu tlr2 each are comprised of 11 exons, confirming that the distribution and number of introns/exons among TLRs do not seem conserved across species [19]. The full-length sequences of carp tlr2a and tlr2b are 2367 and 2358 bp encoding for Tlr2 proteins of 788 and 785 aa with predicted molecular weights of 91 and 90 kDa, respectively. They share 88% amino acid identity. Both Tlr2 molecules have 20 LRRs, and in addition an N-terminal (LRRNT) and a C-terminal LRR (LRRCT), a transmembrane region and a highly conserved TIR domain.
Phylogenetic analyses on amino acid sequences of multiple Tlr1 and Tlr2 sequences (Fig. 3), using carp Tlr3 as an outgroup to root the phylogenetic tree, showed an overall topology indicating clusters of Tlr sequences consistent with evolutionary distance between different fish families. Sequences for both Tlr1 and Tlr2, from Cyprinids (carp, zebrafish) and Siluriforms (catfish), which are their closest living relatives, clustered together with very high bootstrap values, away from other fish species belonging to the Salmonids, Tetraodontiforms and Perciforms. All fish Tlr1 as well as Tlr2 sequences form common clusters separate from mammalian TLR1 and TLR2 sequences.
In the human genome, TLR1, TLR6 and TLR10 are organised as a conserved scaffold of genes. Clearly, in all fish genomes investigated so far, both tlr6 and tlr10 are missing, indicating a recent duplication of TLR1 in the mammalian but not the teleost lineage. Conservation of synteny of teleost tlr1 and tlr2 was investigated by comparing the genomic regions immediately up-and down-stream of human TLR1 (Fig. 4) and human TLR2 (Fig. 5), with the genomic regions up-and down-stream of the corresponding homologue in the annotated genome of several teleost fish species, including zebrafish and common carp. For tlr1, the genomes of fish appear to have in common a block of 10 genes fairly conserved in the investigated fish species. For carp, although the scaffolds are limited in length, the analysis does confirm conservation of synteny with ints10 downstream of tlr1, and klb upstream of the putative tlr1 pseudogene. The fact that the carp tlr1 pseudogene shows conservation of synteny, supports the whole genome duplication event from zebrafish to common carp, with loss of function associated with the second tlr1 gene in carp. For TLR2, in the human genome a block of 8 genes fairly conserved in the investigated fish species could be identified. However, there is a gap in zebrafish between the tlr2 region and the rnf175/trim2 gene cluster further downstream. For carp, although the scaffolds are limited in length, the analysis does confirm conservation of synteny with zebrafish.
Altogether, sequence, phylogenetic and synteny analysis of carp tlr1 and tlr2 support that these genes are indeed orthologues of mammalian TLR1 and TLR2.

Differential expression of tlr1 and tlr2 genes
To investigate the relative gene expression of the newly identified carp tlr1 and to determine whether the newly identified tlr2b would present an expression pattern similar to the previously reported tlr2a sequence, we measured the constitutive gene expression of tlr1, tlr2a and tlr2b, in organs of four healthy carp and in sorted leukocytes. The results from 15 organs are shown in Fig. 6A. Significant differences were found between organs for all tlrs (oneway ANOVA, P < 0.05). Typically, highest tlr gene expression was found in head kidney and mid kidney, peripheral blood leukocytes, spleen and gut. The relatively high gene expression in immune organs, and low to non-detectable expression in other tissues was confirmed by transcriptome analysis of the same organs from carp (unpublished data).
Detection of tlr1-and tlr2-specific gene expression in various immune cell types from carp (Fig. 6B), showed significant differences between cell types (one-way ANOVA, P < 0.05). In general, among immune cell types B cells, in particular, express high levels of tlr genes. Of interest, granulocytes express tlr1>tlr2, whereas head kidney-derived macrophages express tlr2a>tlr1, suggesting high expression of tlr1 and tlr2 is not always restricted to the same immune cell type.

Structural model of potential Tlr1/Tlr2 heterodimer
Mammalian TLR1 and TLR2 have been shown to heterodimerize and the crystal structure of the extracellular portion of human TLR1-TLR2 heterodimer with Pam 3 CSK 4 as a ligand is known (PDBid: 2z7x). Using the known human model, we set to investigate whether the identified carp Tlr1 and Tlr2 could potentially interact. In the human heterodimeric complex of TLR1-TLR2, the receptor multimer is arranged in a symmetrical manner. Three-dimensional modelling shows the typical horseshoe shape of each TLR with a well-positioned pocket to accommodate the tri-acylated lipopeptide ligand. Having noted that the number of LRRs is conserved between the human and carp molecules it was considered appropriate to model the carp Tlrs on their human counterparts. Thus, carp Tlr1 was modelled together with either carp Tlr2a (Fig. 7) or Tlr2b (data not shown). Both Tlr1-Tlr2 combinations exhibited a good three-dimensional fit with the heterodimer structure of human TLR1-TLR2, including the potential to bind to Pam 3 CSK 4 . This three-dimensional fit supports the possible formation of carp Tlr1-Tlr2 heterodimers.

Ligand stimulation of Tlr1 and Tlr2 does not lead to NF-kB activation
To study the biological activity and ligand binding properties of possible Tlr1-Tlr2 heterodimers we first created human HEK and cyprinid EPC cells stably transfected with a NF-kB luciferase reporter [44]. Both cell lines were subsequently transiently transfected with carp Tlr1 and Tlr2a, individually and combined. Each Tlr was cloned as a fusion to GFP/mCherry to monitor transfection efficiency. Stimulation with classical Tlr1-Tlr2 ligands including Pam 3 CSK 4 , LTA and PGN, however, did not lead to luciferase reporter activity (data not shown). Alternatively, we used HeLa-57A cells transfected with various tlr1/tlr2 expression plasmids, which however gave a similar outcome (Fig. 8): none of the tested ligands led to a significant increase in NF-kB activation via overexpressed carp Tlr1-Tlr2. To rule out any inhibitory effect of fusing GFP to the C-terminal end of the Tlr molecules, we also used pBI-CMV1 constructs in which GFP was placed in a cloning site separate from the Tlr. This still allowed for detection of transfection levels by GFP, whilst leaving the structure and function of the Tlr unaffected, but did not alter the negative outcome. Furthermore, in order to rule out an inhibitory effect of GFP itself (whether fused or not to the Tlrs), mCherry constructs were used instead of GFP. Again, no increase in NF-kB activity after ligand stimulation could be measured. Since the positive control; HeLa-57A cells transfected with human TLR1 and TLR2, did lead to high responses to Pam 3 CSK 4 , LTA and PGN (see Fig. 8), we conclude that both the reporter system itself and the ligands used, were functional.

Carp Tlr1 and Tlr2 are expressed and co-localize
Since we could not confirm NF-kB activation by over-expressed carp Tlr1 and Tlr2 stimulated with ligands, we next investigated whether this could be due to erroneous expression of these molecules. In a previous study, carp Tlr2 was successfully transfected into human HEK cells, so we continued with first transfecting this cell line with constructs expressing each carp Tlr fused to the fluorescent protein mCherry. The cell membrane was visualized by staining with wheat germ agglutinin-Alexa 488 conjugate which exhibits green fluorescence. Confocal microscopy confirmed expression of Tlr1, and Tlr2a, at least in the cytoplasm of the cells mostly concentrated in discrete speckles (Fig. 9). For comparison, mCherry protein alone was expressed without being fused to a Tlr and red fluorescence was evenly distributed throughout the cells, indicating that the staining pattern observed when fused to a Tlr is determined by the Tlr itself. Upon further examination of subcellular co-localization of Tlr1 and Tlr2, using carp Tlr1 fused to GFP and carp Tlr2a fused to mCherry, it became clear that both carp Tlr1 and Tlr2a co-localized in the cytoplasm of HEK cells (Fig. 10). To confirm the suspected cytoplasmic localization, carp Tlr1 fused to GFP was detected using anti-FLAG-specific antibody. Only in permeabilized cells, carp Tlr1 could be detected with antibody staining, confirming that Tlr1 was not expressed on the surface of HEK cells overexpressing Tlr1 (Fig. 11). To exclude that the observed expression pattern was not an artefact due to the use of mammalian cells grown at 37 C, we performed the same experiment using various fish cell lines grown at lower temperature. Similar results were obtained with EPC and CLC cells transfected with the same Tlr constructs (data not shown). Altogether, these data demonstrate that carp Tlr1 and Tlr2 can be both successfully (over)expressed in vitro in homologous and heterologous cell lines, and they appear to be present in the same subcellular compartments of these cells.

Discussion
Ever since TLRs were first described as important pattern recognition receptors for mice and humans, they have been studied extensively in an ever increasing number of animal species, including invertebrates such as sea urchin, where a great expansion of tlr genes has occurred [45]. The abundance of Tlrs in sea urchin suggests this class of receptors plays an important role in the innate immune defence also in lower animals. Often, the conserved nature of Tlrs is emphasized, with particular emphasis on the extracellular leucine rich repeat regions and the intracellular TIR domain. In this study, we present the molecular characterization of tlr1 and tlr2 from common carp and discuss our attempts to determine the ligand-binding properties of putative Tlr1/Tlr2 heterodimers.
Sometimes, after a WGD, gene duplication allows for partitioning of function or acquiring of new function, whereas in other cases the duplicated gene can lose its function. A single full-length carp tlr1 gene could be identified; a second tlr1 hit was truncated with a partial TIR domain only. Many of the orthologous relationships of TLRs can be confirmed by observations of conserved synteny, i.e. preservation between species of the order and orientation of orthologous genes [3]. Based on the conserved order of the klb gene neighbouring the second tlr1 hit we consider the truncated tlr1 a true orthologue of TLR1, but one which has lost its function. In rainbow trout a similar situation exists; a single full-length tlr1 gene is present on chromosome 14 and a tlr1 pseudogene is located on chromosome 25 [20]. Thus, it appears that only a single fulllength tlr1 gene remained in the genomes of rainbow trout and common carp. In contrast, two full-length genes for carp tlr2 could be identified, both of which give rise to detectable transcripts. Also rainbow trout express two tlr2 genes, one of which is contained within a single exon but the other segmented into multiple exons [21]. Of interest, a similar situation has recently been reported for yellow croaker [26], suggesting more fish species might have tlr2 genes with quite different intron-exon organisations. Zebrafish tlr2 and also channel catfish [19], similar to human TLR2, is intronless. Likewise, we have not found evidence for the presence of a segmented tlr2 gene in carp.
High expression of tlr1 and both tlr2 genes was found in immune organs such as head kidney and mid kidney, peripheral blood leukocytes, spleen and gut. Of the two tlr2 genes present in carp, tlr2a was always higher expressed than tlr2b in all the investigated organs and cell types. Leukocytes of both myeloid and lymphoid origin expressed tlr1 and tlr2; B cells, granulocytes, and head kidney-derived macrophages had the highest expression levels. Also rainbow trout tlr2 is expressed in mononuclear cells/macrophages at levels similar to that in B cells [21], tlr1 gene expression was not studied in rainbow trout. In mammals, B cells as well as cells of myeloid origin express TLRs [46,47], suggesting a conserved expression pattern for carp tlr1 and tlr2 on phagocytes and B cells.
Our three-dimensional modelling of the carp Tlr1 and Tlr2   [48,49], would be needed to unequivocally prove molecular interaction between the two Tlr proteins. The work by Sandor et al. [50] demonstrates that human TLR1 and TLR2 co-localize in double-transfected (TLR1-YFP þ TLR2-CFP) HEK cells in a ligandindependent manner, similar to our observations for carp Tlr1 and Tlr2. The pre-assembled human TLR1-TLR2 heterodimers are present inside as well as on the surface of cells.
Possibly because over-expression affects the natural molecular distribution, or because only minute and therefore difficult to detect amounts of Tlrs are required at the cell surface, we were unable to pinpoint sub-cellular localization to the cell surface. This may also be true for human TLR1 when over-expressed in HEK cells, because TLR1 is hard to detect at the cell membrane and is found mainly localized inside the cells with a diffuse pattern of distribution [50]. Further, the endogenous cytoplasmic pool of TLR1 in HEK Fig. 4. Synteny analysis of tlr1. Comparative gene organization map of the regions where human, stickleback, medaka, Tetraodon, zebrafish, rainbow trout and carp tlr1 genes are found. Orthologous genes are indicated by the same colour, except unique genes which are all light grey. Gene direction is indicated with arrows. Rainbow trout and carp have a pseudogene (j) in addition to the full-length tlr1 gene. The human genome assembly version GRCh38.p3, stickleback genome assembly BROAD S1, medaka genome assembly HdrR, Tetraodon genome assembly TETRAODON 8.0, zebrafish genome assembly GRCz10, and the carp genome [36] were used for this analysis. Information on rainbow trout genomic organization is from Palti et al. [20]. cells can be detected in permeabilized cells only, and human TLR2 is not found exclusively at the cell membrane either [50]. Also chicken TLRs cannot easily be detected on the cell surface after transfection (unpublished data). Yet, both human and chicken TLRs, when overexpressed, can lead to NF-kB activation upon stimulation with appropriate ligands. These collective findings suggest that, although it is not easy to detect TLR molecules at the cell surface after transfection, intracellular localization of carp Tlr1 and Tlr2 may not necessarily explain the absence of NF-kB activation in our experiments.
In chicken, both TLR1 and TLR2 exist as duplicated copies; each pair of genes is found as tandem array genes, suggesting that they arose from local gene duplication events [51], rather than from a whole-genome duplication, as is the case for carp Tlrs. Chicken TLR1 and TLR2 are not functional as homodimers whereas they do recognize bacterial lipoproteins (di-as well as tri-acylated) as heterodimers in certain combinations [43,52]. Although the threedimensional model of carp Tlr1-Tlr2 does not suggest this would be the case, it is difficult to fully exclude the possibility that Tlr2b rather than Tlr2a could be the natural partner of a heterodimeric Tlr1-Tlr2, an option we did not explore experimentally.
In mammals, TLR2 does not only form heterodimers with TLR1, but also with TLR6 (and TLR10 in humans). In fish genomes, tlr6 or tlr10 genes are absent but instead a number of fish-specific Tlrs have been recognized as members of the Tlr1 family. It is interesting to speculate that Tlr2 in fish might heterodimerize with some of these members, including Tlr14, Tlr18, Tlr25 and Tlr27 [9,19,53]. So far we could not identify tlr14, tlr25, or tlr27 in the common carp genome. However, a screen of the carp genome did confirm the presence of two tlr18 genes (data not shown), as could be expected based on the presence of tlr18 in the zebrafish genome [16]. Molecular and functional characterization of new, fishspecific, Tlr1 family members should provide insight in the number of different heterodimer combinations possible with Tlr2.
In previous work, we have described ligand-specific activation of carp Tlr2 via measurement of increased phosphorylation levels of the MAP kinase p38 in human HEK cells by Western blot [28]. Instead of this semi-quantitative method, we now used a quantitative read-out system based on NF-kB activation and subsequent luminescence measurements. We could not confirm our initial ligand binding studies based on phosphorylation of p38, neither using the human HEK or HeLa-57A cell lines, nor using a cyprinid fish cell line (EPC), from fathead minnow [54]. Of interest, the HeLa-57A cells have successfully been used to study ligand binding of several chicken TLRs [43] and even reptilian TLR5 [55], confirming that human cells can be used to study ligand binding by Tlrs from non-mammalian species. There are several possible explanations for the inability of carp Trl1 and/or Tlr2 to induce activation in our NF-kB reporter assay; i) carp Tlrs may not have been functioning optimally because they did not display the correct sub-cellular localization on the cell surface, e.g. due to over-expression, ii) carp Tlrs were not properly folded e.g. due to high temperature conditions for human cells, although this would not be the case for the fish cell line grown at 27 C, iii) the carp TIR domains may not Orthologous genes are indicated by the same colour, except unique genes which are all light grey. Gene direction is indicated with arrows. The human genome assembly version GRCh38.p3, stickleback genome assembly BROAD S1, medaka genome assembly HdrR, Tetraodon genome assembly TETRAODON 8.0, zebrafish genome assembly GRCz10, and the carp genome [36] were used for this analysis.  6. Basal gene expression levels of tlr1, tlr2a and tlr2b in carp organs and immune cell types. Real-time quantitative PCR was performed using primers specific for carp tlr1, tlr2a and tlr2b on cDNA from (A) organs or (B) immune cell types isolated from healthy fish. Gene expression is normalized to the expression of the 40s ribosomal protein s11 housekeeping gene. Bars indicate average gene expression of n ¼ 4 adult fish and error bars indicate standard deviations. Asterisks denote significant differences in expression between tlrs within a cell type, P < 0.05. PBL ¼ peripheral blood leukocytes. HK ¼ head kidney. Fig. 7. Heterodimer of carp Tlr1-Tlr2(a), modelled on human TLR1-TLR2 with Pam 3 CSK 4 as ligand. The extracellular part of the human TLR1-TLR2 heterodimer, crystallized with Pam 3 CSK 4 as a ligand (PDB-id: 2z7x) was used to create models of carp Tlr1 and Tlr2a using the Modeller program (version 9.12). Thirty comparative models were generated, after which the model with lowest corresponding DOPE score was selected for image generation. Carp Tlr1 is shown in green, Tlr2a in blue, Pam 3 CSK 4 in red. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) have been able to interact with human MYD88 (at least in HEK and HeLa-57A cells), thus initiating no signalling cascade, iv) carp Tlrs may need a co-receptor or other co-factor [56], absent from the cell lines we tested, v) different ligands are recognized by the carp Tlrs, vi) the present NF-kB read-out system (although appropriate for mammalian, chicken and amphibian TLRs) is not suitable for Fig. 9. Carp Tlr1 and Tlr2 localize intracellularly in HEK cells. HEK cells were transfected with (A) tlr1-mCherry-pcDNA3, (B) tlr2-mCherry-pcDNA3, or (C) mCherry-pcDNA3 (red) and analysed after 3 days by confocal laser scanning microscopy. Wheat-germ agglutinin conjugated to Alexa488 (green) was used to stain the cell surface. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) Fig. 10. Carp Tlr1 and Tlr2 co-localize intracellularly in HEK cells. HEK cells were co-transfected with tlr1-GFP-pcDNA3 (green) and tlr2-mCherry-pcDNA3 (red) and analysed after 3 days by confocal laser scanning microscopy. The overlay shows extensive co-localization of the green and red signal. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) studying Tlrs of fish. Indeed, ligand-dependent NF-kB activation mediated through fish-encoded Tlrs is not easily shown in vitro (as discussed in Ref. [21]). We do believe that bacterial PAMPs including lipoproteins could be potential activators of Tlr signalling in carp, as several of the components of the signalling pathway including MYD88 and TRAF6 [57], IRAK1 [58], and several accessory molecules for Tlrs [56] are all present. Indeed, carp macrophages respond to classical TLR2 ligands such as Pam 3 CSK 4 and MALP-2 by altering gene expression of pro-inflammatory cytokines [28]. The latter study on carp Tlr2 using phosphorylation of p38 as a read-out is not in contradiction with the present study, the two studies rather suggest that several TLR signalling pathways and transcription factors should be scrutinized. Future studies could include measuring activation of transcription factors other than NF-kB, such as AP-1 or IRF5, measuring caspase activity, or measuring downstream effects such as the production of pro-inflammatory cytokines. In other words, given the present difficulties detecting NF-kB activation downstream of Tlr1-Tlr2 heterodimerization, it may be worthwhile in future work to refocus on signalling pathways other than NF-kB as important activators of inflammatory cytokines.