) 42–51www.elsevier.com/locate/gene

Gene 378 (2006

Full-length sequence and expression analysis of Toll-like receptor 9in the gilthead seabream (Sparus aurata L.)

Rafaella Franch, Barbara Cardazzo, Jenny Antonello, Massimo Castagnaro,Tomaso Patarnello ⁎, Luca Bargelloni

Department of Public Health, Comparative Pathology, and Veterinary Hygiene, University of Padova,viale dell'Università 16, 35020 Legnaro (Pd), Italy

Received 25 November 2005; received in revised form 24 March 2006; accepted 26 April 2006Available online 11 May 2006

Abstract

Toll-Like Receptors (TLRs) have recently emerged as key sensors of invading microbes, acting through recognition of pathogen-associatedmolecular patterns. It has been demonstrated that TLR9 is involved in the recognition of unmethylated CpGmotifs inmice, humans, and pigs.We reporthere the full-length sequence of TLR9 cDNA in the gilthead sea bream (Sparus aurata L.). The predicted protein (1063 amino acids) was similar tomammalian TLR9s, showing 21 leucine-rich repeats in the extracellular region and a typical Toll/IL-1R (TIR) domain in the intracellular region.Comparative analysis of TLR9 sequences indicated that critical residues for ligand-binding are conserved across vertebrate lineages, although evidenceof functional divergence was observed. Analysis of the genomic structure of sea bream TLR9 gene revealed the presence of two intervening sequences.Retention of the second intron produced an alternatively splicedmRNA (TLR9B) showing differential expression among tissues or developmental stagescompared to the wild-type isoform (TLR9A). RT-PCR analysis indicated a broad expression of TLR9A, especially in immune-related organs (spleen,head-kidney) andmucosal–epithelial barriers (gills, gut, skin). Using quantitative Real-Time RT-PCR, no statistically significant variation was observedforTLR9mRNAs expression in the spleen of experimentally infected animals compared to healthy controls. Comparing sequence and expression profileof sea bream TLR9with mammalian TLR9s suggested that the main function of TLR9might be conserved across vertebrates, although species-specificfeatures are present (modulation of ligand-binding specificity, alternative splicing).© 2006 Elsevier B.V. All rights reserved.

Keywords: Vertebrate TLR9; Teleost fish; Gene expression; Alternative splicing; Natural immunity

1. Introduction

Toll-Like Receptors (TLRs) have recently emerged as keysensors of invading microbes. TLRs recognize pathogen-as-sociated molecular patterns (PAMPs), which represent conservedmolecular features of a given microbial class, and initiate sig-nalling pathways that are critical for inducing the immune responsethrough the activation of anti-microbial genes and the production ofinflammatory cytokines (Takeda et al., 2003). TLRs are type Itransmembrane proteins with an extracellular domain containing

Abbreviations: TLRs, Toll-Like Receptors; TIR, Toll/IL-1R; PAMPs,pathogen-associated molecular patterns; CpG, CG dinucleotides; RACE,Rapid Amplification of cDNA Ends; UTR, Untranslated regions; LRRs,leucine-rich repeats; SNP, single nucleotide polymorphism; ODN, oligodeox-ynucleotides; MBD, methyl-CpG-DNA binding domain.⁎ Corresponding author. Tel.: +39 049 8272977; fax: +39 049 8272602.E-mail address: tomaso.patarnello@unipd.it (T. Patarnello).

0378-1119/$ - see front matter © 2006 Elsevier B.V. All rights reserved.doi:10.1016/j.gene.2006.04.025

interspersed leucine-rich repeat motifs that are involved in therecognition of various PAMPS (Aderem and Ulevitch, 2000).TLRs represent a central component of the innate immune responseas well as a link between natural and adaptive immunity.

The cytoplasmic domain is characterized by a Toll/IL-1receptor (TIR) motif that is involved in signal transduction(Aderem and Ulevitch, 2000). To date, 10 human and 12 murineTLR proteins have been characterized (Takeda et al., 2003, 2004;Zarember and Godowski, 2002). TLRs have also been found inother vertebrate groups (Oshiumi et al., 2003; Meijer et al., 2004;Jault et al., 2004). Functional studies using gene-disrupted miceand TLR-overexpressing cells indicate that each TLR recognizesspecific PAMPs (Akira and Hemmi, 2003).

Bacterial DNA acts as a PAMP by virtue of a 20-fold greaterfrequency of unmethylated CG dinucleotides (CpG) found inmicrobial DNA compared to vertebrate DNA. It has recentlybeen demonstrated that TLR9 is involved in the recognition of

43R. Franch et al. / Gene 378 (2006) 42–51

unmethylated CpG motifs in mice (Hemmi et al., 2000), hu-mans (Bauer et al., 2001), and pigs (Shimosato et al., 2003).Mammalian TLR9 appears to recognize its ligand through directbinding of CpG DNA in the lysosome (Latz et al., 2004).

We report here the full-length sequence of TLR9 cDNA in thegilthead sea bream (Sparus aurata L.), a comparative analysis ofvertebrate TLR9 sequences, and RT-PCR analysis of TLR9 ex-pression in larval and adult samples of sea bream. S. aurata is amarine teleost fish that belongs to the Perciform order. Teleostsrepresent the largest group of vertebrates, and several teleostspecies, including the gilthead sea bream, are of great importancefor fisheries and aquaculture. Lower vertebrates, especially teleosts,rely heavily on innate immunity, while their adaptive immuneresponses mature later, and humoral responses lack class switchingand efficient affinity maturation (Trede et al., 2004). Therefore,since TLRs appear to be crucial in the natural immune response,isolation and characterization of these proteins in the teleost fishmight be helpful toward a better understanding of innate immunityin fish. In addition, comparing TLR9 sequence and expressionprofile between sea bream and mammals might shed more light onthe evolution of this important class of genes across vertebrates.

2. Materials and methods

2.1. Animals collection and RNA extraction

Sea breams were collected at the fish farm ‘Impianto diAcquacoltura Ca' Zuliani’ (Monfalcone, Italy). Samples con-sisted of different developmental stages (whole larvae of 4, 15,30, 48 days), as well as different tissues/organs dissected fromjuvenile fishes of 110 days and from young adults (150 g): brain,intestine, heart, skeletal muscle, gill filaments, gonads, skin,spleen, liver and head-kidney (only from adult fish). All tissuesamples were dissected from anaesthetized animals and stored inRNA later (Ambion, Austin, TX) at −80 °C; subsequently tissuesamples were homogenized in Trizol Reagent (Invitrogen,Carlsbad CA) and total RNA isolation was completed followingthe manufacturer's protocol. Extracted RNAs were treatedwith RQ1 RNAse-free DNAse (Promega, Madison, WI) andpurified with a standard phenol-chloroform extraction or usingthe RNeasy MinElute Clean Up kit (Qiagen GmBH, Hilden,Germany).

Spleen tissue samples were dissected from sea breams of 0.5 gbody weight experimentally exposed to a bacterial fish pathogen.Details of the experimental infection are reported elsewhere(Zappulli et al., 2005). Briefly, sea breams were divided in twogroups. Individuals of group 1 were challenged by immersionwith Photobacterium damselae subspecies piscicida (3×105

bacteria l−1). Group 1 was challenged with a low virulence strain(DI21, Zappulli et al., 2005) to obtain sub-lethal infection. Group2 was mock-infected by immersion in clean aerated seawater.Forty-eight hours after challenge, five individuals were removedfrom each group. Tissue samples from these fishes were collectedand stored as described above. A pathogen-specific test wasperformed on DNA extracted from infected tissues (Zappulliet al., 2005). All group 1 individuals were strongly positive for thepresence of the pathogen.

2.2. Full-length cDNA isolation

Onemicrogram of total RNA from sea bream head-kidneywasreverse transcribed using a reverse transcriptase (Superscript II,Invitrogen) with random hexamers to obtain first-strand cDNA.The cDNAwas then used as a template for the subsequent PCRreactions. All PCR reactions were performed using a GeneAmpPCR System 9700 (Applied Biosystems, Forster City, CA) underthe following conditions: 3 min at 94 °C (initial denaturation), 40cycles with 20 s at 94 °C, 30 s at the annealing temperature (seebelow), 1 min at 72 °C and 10 min at 72 °C (final extension). Theannealing temperature was decreased from 60 °C to 50 °C in thefirst 20 cycles (−0,5 °C per cycle), then set at 50 °C for theremaining 15 cycles. One fifth of each PCR reaction was elec-trophoresed on a 1.8% agarose gel, stainedwith ethidiumbromide,and visualised under UV light. Initially, a partial fragment wasamplified using the following primers: TLR9F and TLR9R (seeTable 1). These primers were designed on the basis of the multiplealignment of the nucleotide sequences of human, murine, andpufferfish TLR9 genes. The PCR product was cloned into aplasmid vector using the TOPO-TA cloning kit (Invitrogen). Atleast three independent clones were sequenced. Based on thesequenced fragment, sea bream-specific oligos (Table 1) weredesigned to isolate the complete cDNA sequence by means of theRapid Amplification of cDNA Ends (RACE) technique. TheSMART RACE cDNA Amplification Kit (BD Biosciences, PaloAlto, CA) was used according to the manufacturer's instructions,in order to obtain 5′ and 3′ unknown cDNA regions. PCRproductsof 5′ and 3′ RACE reactions were cloned and sequenced asdescribed above. Three rounds of 5′ RACE were performed usinggene-specific primers that were located each time further uptoward the 5′ end of the gene [TLR95race-TLR95raceN (nested),5raTLR9-5ranTLR9n (nested), TLR9-5ra-TLR9-5ran (nested].To confirm the full length cDNA sequence, additional sea bream-specific primers (Table 1) were designed to amplify four partiallyoverlapping fragments (TLR9-6F-TLR9-5ra, TLR9-5F-TLR9-3R, TLR9-4F-TLR9-2R, TLR93raceN-TLR9-1R). PCR condi-tions were: 94 °C 20 s, 30 s at the specific annealing temperature,72 °C 1 min for 35 cycles. The exact annealing temperature foreach primer pair is listed in Table 1. The obtained PCR productswere directly sequenced without prior cloning.

2.3. Sequence analysis

The predicted amino acid sequence of sea bream TLR9 wasexamined as follows. The presence of a signal peptide waspredicted using the hidden Markov model approach implemen-ted in SignalP 3.0 (http://www.cbs.dtu.dk/services/SignalP-3.0/)under default settings. Known protein domains were definedusing Interpro (http://www.ebi.ac.uk/interpro/) and SMART(http://smart.embl-heidelberg.de/). Putative transmembraneregions were determined with Toppred (http://bioweb.pasteur.fr/seqanal/interfaces/toppred.html).

To obtain a suitable set of vertebrate TLR sequences, two fishgenomes were searched as follows. The latest version of genomesequence assemblies was downloaded for F. rubripes (ver. 3.0,August 26, 2002, http://genome.jgi-psf.org/fugu6/fugu6.home.

Table 1Sequence and experimental conditions for primers used in the present study

Primer name Primer sequence (5′–3′) Annealing temperature

TLR9F CTCTACGGCTGGGATCTTTGGTA 50 °CTLR9R GCCGCTGCTGGACCATGAAGAA 50 °CTLR95race TCCTGTGATCTGAGTTCTCCAGGTTGACC 68 °CTLR95raceN GACCCAATCCCTCACAGCATTGTTGC 68 °C5raTLR9 GACCCGGCGACCTCCAACAATTCC 68 °C5ranTLR9 GCCCGTTTAACCGGTTGGATTCCAGGAC 68 °CTLR9-5ra GGATCTCCATCTTGCACGAGCGGTC 68 °CTLR9-5ran GAGCCCTGATAAGTCATCCGACCC 68 °CTLR93race GGTCAACCTGGAGAACTCAGATCACAGGA 68 °C3raTLR9 CCCTATCTTGTGTCACAATGAC 68 °CTLR9-1R TACAACCAATACAGCGCCTC 57 °CTLR9-2R GTCCAGGTACATGAGGCTTG 59 °CTLR9-3R ACAGAACGTCCAGGCTCTC 58 °CTLR9-4F CTGACAACCTCCTGGCATATGA 59 °CTLR9-5F GTAGACTGCTCTGACAGACCA 58 °CTLR9-6F ATAATGTTTCCTCTCCCGACCT 57 °CTLR93raceN GTCTCTGTTTGGAGGAGAGGGACTGG 57 °CTLR9-7F GGAGGAGAGGGACTGGATTC 59 °CTLR9-6R TCAGCTGAAGGTACTTTAGTTTGG 59 °CRNATLR9A286F CTGGATGAGAAGGTGGATGCA 60 °CRNATLR9A397R GCCAGGACAACACAGACTTCCT 60 °CRNATLR9b334F TTTCCCCCTGCTGAGTTGTT 60 °CRNATLR9b435R AACATCTCATCGAGCAGAACCA 60 °CEF1Sa95F AAATGCGGAGGAATCGACAA 60 °CEF1Sa165R GAGCCCTTGCCCATCTCAG 60 °CAct 1F CTGGGATGACATGGAGAAGA 60 °CAct 1R CTTGATGTCACGCACGATTT 58 °C

44 R. Franch et al. / Gene 378 (2006) 42–51

html), andD. rerio (WGS ver. 3, 27 November 2003, http://www.sanger.ac.uk/Projects/D_rerio/wgs.shtml). Similarity searcheswere implemented against the downloaded databases using astand-alone version of BLAST (ver. 2.2.2, ftp://ftp.ncbi.nih.gov/blast/executable/). Initially, all knownmurine (12) and human (10)TLR protein sequences were extracted fromGenBank and used asa multiple sequence query to search the two fish genomes with thetBlastn option. The results obtained were then used to extract thecorresponding protein sequences from the predicted transcriptomeof F. rubripes (http://genome.jgi-psf.org/fugu6/fugu6.home.html),and D. rerio (http://www.ensembl.org/Danio_rerio/). Only anno-tated peptides were considered, whereas proteins originating justfrom GenScan predictions were excluded from the analysis. Basedon the predicted protein domain structure, only sequences showingboth a Toll/IL-1R (TIR) domain and one or more Leucine-Rich-Repeat domain(s) were further analyzed. The fish proteins obtainedas described abovewere then used as amultiple sequence query in aBlast search as previously mentioned. The same procedure wasthen repeated until no additional sequences could be found.

All pufferfish and zebrafish putative TLR sequences werealigned with murine and human TLRs using a Web-based versionof ClustaW (http://bioinformatics.utmem.edu/clustal/clustalw.html) under default settings. In the same alignment were includedother known piscine TLRs: Japanese flounder TLR2 and TLR22,rainbow trout TLR, goldfish TLR, and the putative sea breamTLR9 sequence obtained in this study. All other known TLR9proteins were analyzed as well: pig TLR9, rat TLR9, bovineTLR9, canine TLR9, and feline TLR9. The corresponding access-ion numbers to all TLR sequences used in the phylogenetic an-

alysis are listed in the legend to Fig. 2. Two additional putative fishorthologues of TLR9 were obtained as follows. Murine andhuman TLR9 sequences were used as queries in a Tblastn searchagainst the genome of Tetraodon nigroviridis (ver.7, 4 April 2003,http://www.genoscope.cns.fr/externe/tetranew/data/assembly/)and Orizyas latypes (6 July 2004, http://dolphin.lab.nig.ac.jp/medaka/). The sequence yielding the highest Blast score was thenextracted from each genome database. In the case ofT. nigroviridisthe corresponding annotated protein was already available fromthe Tetraodon genome browser (http://www.genoscope.cns.fr/externe/tetranew/entry_ggb.html). For the medaka sequence, thecorresponding contig was obtained. A “Blast two-sequence”alignment (http://www.ncbi.nlm.nih.gov/blast/bl2seq/bl2.html)was used to define the putative coding region in the medakagenome. The predicted protein sequence was then refinedmanually after translation of the DNA sequence. All alignedsequences were used to reconstruct a phylogenetic tree of TLRproteins. The tree building method used maximum likelihood asoptimality criterion, as implemented in the program PHYML(Guindon and Gascuel, 2003). The best fitting evolutionarymodel(WAG+gamma distribution+invariant sites) was establishedusing Prottest (Abascal et al., 2005) and implemented in PHYML.

Functional divergence between fish and mammalian TLR9swas analyzed using the program Diverge (Gu, 1999). Twosequence clusters were considered, the first one including allmammalian sequences, the other all teleost ones. The coefficientof functional divergence (θ) and the threshold of a posterioriprobability for significantly divergent sites were calculated asdescribed in Wang and Gu (2001).

45R. Franch et al. / Gene 378 (2006) 42–51

2.4. Characterization of sea bream TLR 9 intronic regions

The same four primer pairs reported above were also used toPCR amplify part of the TLR9 gene from sea bream genomicDNA. The PCR conditions used were as described previously.PCR products were directly sequenced, and sequences fromgenomic DNA and cDNA were compared to determine theexon–intron structure of the TLR9 gene.

2.5. Sequence polymorphism

The presence of a single nucleotide polymorphism at position3671 (GenBank AY751798) was assessed as follows. GenomicDNAs were extracted from muscle tissue of 38 fishes. The indi-viduals analyzed originated from different geographic popula-tions (Aegean Sea, Adriatic Sea, Thyrrenian Sea, AtlanticOcean).Primers TLR9-7F and TLR9-6R were used to PCR amplify a339 bp fragment that was subsequently digested with HpyCHIV(New England Biolabs, Beverly, MA). Restriction profiles wereevaluated on 1.8% agarose gel.

2.6. RT-PCR expression analysis

A RT-PCR analysis was carried out on total RNA extractedfrom different tissues/organs of adults and of fries of 110 days aswell as from whole larval stages. For each sample the sameamount of RNA (1 μg) was reverse transcribed using randomhexamers. Specific primers (TLR9-7F-TLR9-6R, see Table 1)were used to amplify a cDNA fragment of 239 bp in size for TLR9.Primers were located on different exons (exon 2: TLR9-7F, exon3: TLR9-6R). A fragment of the sea bream β-actin gene wasamplified to test the quality of the extracted RNA, the efficiency ofthe RT reaction, and the absence of genomic DNA contamination.β-actin primers are listed in Table 1. The PCR conditions usedwere as described in Section 2.2. TheRTenzymewas also omittedto test for the presence of genomic DNA contamination. The sizeof the expected product from β-actin cDNA is 446 bp. One ormore larger fragments are expected from genomic DNA, sinceact1F is located on exon 2 and act1R is positioned on exon 5 of theβ-actin gene.

2.7. RealTime RT-PCR quantification of TLR9 splicingisoforms

Two alternative splicing isoforms of sea bream TLR9 werefound to be expressed in most samples. To further evaluate thisevidence and to quantify the relative expression of both isoforms,a Real-TimeRT-PCR assaywas developed. Two primer pairs, onespecific for isoform A, one for isoform B, were designed(RNATLR9A286F–RNATLR9A397R, RNATLR9b334F–RNATLR9b435R, Table 1). The two TLR9 isoforms representedthe “target” genes. The housekeepingElongation Factor 1 (EF-1)gene was used as “reference” gene. To this end a specific primerpair was designed (EF1Sa95F–EF1Sa165R, Table 1). Primerdesign was carried out with the Primer Express software (ver. 2.0Applied Biosystems). PCR reactions were carried out in 96 wellmicrotiter plate wells in a 20 μl reaction volume with SYBR

Green master Mix (Applied Biosystems) with optimized con-centrations of specific primers. An ABI Prism 7000 SequenceDetection System (Applied Biosystems) was programmed for aninitial step of 2 min at 50 °C and 10 min at 95 °C, followed by 40thermal cycles of 15 s at 95 °C and 1 min at 60 °C. Every assaywas run in triplicate and negative controls (no template, templateproduced with no RT enzyme) were always included. For eachsample 1 μg of total RNA was reverse transcribed as describedabove in this section. A 1:20 dilution of each cDNAwas used astemplate in Real-Time PCR. Specificity of PCR amplification ofeach primer pair was confirmed by melting curve analysis (Ririeet al., 1997). Standard curves were generated by measuringthreshold cycle (CT) values of two-fold serial dilutions of thecalibrator cDNA. The efficiency (Ex) of each RT-PCR assay wasmeasured as Ex=10

(1− slope) where slope is the slope of thestandard curve. The efficiencies of each assay (TLR9A, TLR9B,EF-1) were found to be comparable when applying a validationtest. Briefly, the differences between CT values of reference andtarget gene at each dilution of calibrator cDNA were plottedagainst the log of cDNA concentration. The absolute value ofslope of the interpolating curve is expected to be b0.1. Since forboth target templates (TLR9A and TLR9B) this value was b0.1,fold differences for gene expression were calculated bycomparative threshold cycle (CT) method (2−ΔΔCT) as imple-mented in the software SDS (ver. 1.1 Applied Biosystems).

Statistical analyses on relative expression levels (Spearmanrank correlation test and Mann–Whitney test for independentsamples) were implemented in SPSS (ver. 12.0.1, SPSS Inc.Chigago, IL). Exact probability values were obtained.

3. Results

3.1. Molecular cloning of sea bream TLR9

The adopted RACE approach yielded a cDNA sequence of3913 nucleotides for the putativeTLR9 gene of S. aurata (SaTLR9,GenBank AY751797). Translation of the nucleotide sequenceusing the first possible start codon produced a peptide sequence of1063 amino acids. Untranslated (UTR) regions were respectively67 bp (5′UTR) and 657 bp (3′UTR). Results of similarity searchesconfirmed that the obtained sequence is most similar to vertebrateTLR9 proteins (data not shown). Prediction of protein domainsrevealed a potential signal peptide at positions 1–25 of SaTLR9(Fig. 1). A putative transmembrane region was predicted betweenresidues 846 and 866. This defined an N-terminal extracellularregion, containing 21 leucine-rich repeats (LRRs), and anintracellular C-terminal region showing a typical TIR domain(Fig. 1). These boundary positions are conserved compared tomammalian TLR9s. The exon–intron structure of the gene wasalso defined (GenBank AY751798). Two intervening sequenceswere found, respectively of 603 bp and 100 bp. The first intronwasobserved in correspondence of position 23 of the coding sequence,while the second one was located at position 2986 (Fig. 1). Thecomparison of genomic and cDNA sequences, which wereobtained from two different individuals, revealed a potential singlenucleotide polymorphism (SNP) at position 3671 (GenBankAY751798). A survey for the presence of this SNP in 38 sea

46 R. Franch et al. / Gene 378 (2006) 42–51

breams from different geographic locations, however, suggestedthat this is a rare variant, which is present in a single heterozygoteindividual.

3.2. Evolutionary analyses on vertebrate TLR9s

To further evaluate the evolutionary relationships of SaTLR9in the context of vertebrate TLRs, a phylogenetic tree wasreconstructed (Fig. 2). All vertebrate TLR9 proteins formed amonophyletic group, and the observed relationships within thiscluster reflected the taxonomic positions of the species. The onlyexception is the anomalous position of rodent TLR9s (mouse andrat) as the sister group of all the other mammalian sequences(cattle, pig, cat, dog, human), whereas rodents are expected to bemore closely related to primates than to carnivores or ungulates(Murphy et al., 2001). This evidence might be explained as a“long branch attraction effect” due to the higher rate of sequenceevolution that has been reported for the murine and rat genomes.

The phylogenetic tree reconstructed here also confirms that atleast nine TLR genes were already present in the commonancestor of mammals and teleost fishes as proposed in previousstudies. The main novelty in the tree in Fig. 2 is that murineTLR13 is included in a group of TLR sequences that waspreviously considered “fish-specific” (Oshiumi et al., 2003;Meijer et al., 2004). TLR13 is likely the mouse orthologue ofTLR21 in the teleost fish.

The analysis of functional divergence (Wang and Gu, 2001;Gu, 2003) suggested that evolutionary constraints on teleostTLR9s are significantly different (theta 0.42±0.04, pN0.001)compared to mammalian TLR9s. A posteriori analysis of func-tional constraints at each amino acid position revealed that 19 sitesshowed a significantly different rate of evolution between the twogroups (teleosts–mammals).Most functionally divergent residuesare found in the extracellular region, while only two are located in

Fig. 1. Domain architecture of the sea bream TLR9. TLR9A represents the deducpredicted signal peptide, in light grey are the Leucine-Rich Repeats, in white the tranthe TIR domain (O'Neill et al. 2003) are shadowed and referred to with Roman numbArrows correspond to functionally divergent positions (see Sections 3.2 and 4). ThAbbreviations of vertebrate names are as follows: Bt = Bos taurus, Cf = Canis familiaMm =Mus musculus, Ol = Oryzias latipes, Rn = Rattus norvegicus, Sa = Sparus aurencoded by TLR9B mRNA is represented below.

the intracellular region, outside the TIR domain (Fig. 1). Nearlyall divergent sites are completely conserved among teleost TLR9swhereas they are fast evolving in mammals (data not shown).

3.3. RT-PCR expression analysis

RT-PCR analysis of SaTLR9 expression using primersencompassing intron 2 revealed the presence of two bands(Fig. 3A). The lower band represented the expected product ofcompletely spliced mRNA (TLR9A). The upper band originatedfrom the amplification of a splicing variant that retained thepresence of intron 2 (TLR9B, GenBank AY751796). It can beexcluded that this variant originated from contaminant genomicDNA, since all RNAs were treated with DNAse and no productwas obtained when omitting the enzyme in the RT reaction (seeMethods). A similar result was obtained when using primers forthe β-actin gene (Fig. 3B,C).

TLR9Awas found to be expressed in several tissues of juve-nile and adult sea bream as well as in whole larvae at differentdevelopmental stages. Liver, skeletal muscle, and brain were theonly tissues showing weak or absent expression of splicingvariant A. TLR9B was clearly observed in adult and juvenilespleen and gills, in the adult brain, and in all larval stages. Sinceamplification of TLR9A and TLR9B was carried out in the samereaction tube with the same primer pair, each template (TLR9Aand TLR9B) will compete with the other one. The PCR productof TLR9A should then be slightly favored because of its smallersize. As expected, a more brilliant amplification band is obs-erved for TLR9A in most tissues. However, in whole larvae, aswell as in adult and juvenile brain, TLR9B yielded a strongeramplification signal. This seems to indicate that the ratio bet-ween TLR9A and TLR9B mRNA varies across tissues. In seabream larvae, the isoform ratio appears to increase from earlierto later stages (Fig. 3).

ed protein from the completely spliced mRNA. A dark grey box indicates thesmembrane domain, and in black the TIR domain. The three conserved boxes iners. The position and length of the two introns is indicated with a black triangle.e alignment of specific residues across all vertebrate TLR9s is shown above.ris, Dr = Danio rerio, Fc = Felis catus, Fr = Fugu rubripes, Hs = Homo sapiens,ata, Ss = Sus scrofa, Tn = Tetraodon nigroviridis. The putative truncated protein

Fig. 2. Phylogenetic tree of vertebrate TLRproteins. Numbers at each node refer to bootstrap values (only values higher than 50% are shown). The scale bar indicates 0.2 proteinsequence divergence. Abbreviations of sequence names are as follows: Bt = Bos taurus, Ca =Carassius auratus, Cf =Canis familiaris, Dr =Danio rerio, Fc = Felis catus, Fr =Fugu rubripes, Hs =Homo sapiens, Mm=Musmusculus, Ol =Oryzias latipes, Om=Oncorhynchus mykiss, Po =Paralichthys olivaceus, Rn =Rattus norvegicus, Sa = Sparusaurata, Ss = Sus scrofa, Tn = Tetraodon nigroviridis. Numbering of TLR proteins follows the nomenclature used in the most recent literature. Accession numbers of sequencesretrieved from public data bases are as follows: BtTLR9 (NP_898904), CaTLR22 (AAO19474), CfTLR9 (NP_001002998), DrTLR1 (ENSDARP00000020678), DrTLR2(AAQ90474), DrTLR3 (ENSDARP00000014779), DrTLR4a (ENSDARP00000003522), DrTLR4b (AAQ90475), DrTLR9 (ENSDARP00000022876), DrTLR18(ENSDARP00000015475), DrTRL21 (ENSDARP00000027343), DrTRL22 (ENSDARP00000027304), FcTLR9 (AAN15751), FrTLR2 (FRUP000001157320), FrTLR3(FRUP00000138215), FrTLR5 (FRUP00000163648), FrTLR9 (FRUP00000144019), FrTLR18 (FRUP00000134585), FrTLR21 (FRUP00000138634), FrTLR22(FRUP00000143580), HsTLR1 (Q15399), HsTLR10 (Q9BXR5), HsTLR2 (O60603), HsTLR3 (O15455), HsTLR4 (O00206), HsTLR5 (O60602), HsTLR6 (Q9Y2C9),HsTLR7 (Q9NYK1), HsTLR8 (Q9NR97), HsTLR9 (NP_059138), MmTLR1 (Q9EPQ1), MmTLR11 (AY510704), MmTLR12 (AY510705), MmTLR13 (AY510706),MmTLR2 (Q8K3D9), MmTLR3 (Q99MB1), MmTLR4 (Q8K2T5), MmTLR5 (BAC29603), MmTLR6 (Q9EPW9), MmTLR7 (P58681), MmTLR8 (P58682), MmTLR9(NP_112455), OlTLR9 (this study), OmTLR22 (AJ628348), PoTLR2 (BAD01046), PoTLR22 (BAD01047), RnTLR9 (NP_937764), SaTLR9A (this study), SsTLR9(NP_999123), TnTLR9 (GSTEN00027960) (see Section 2.3 for details on how we consulted genome sequence assemblies of D. rerio, F. rubripes and T nigroviridis).

47R. Franch et al. / Gene 378 (2006) 42–51

Fig. 3. A) RT-PCR analysis of TLR9 expression in different tissues and developmental stages. Abbreviations are as follows: 4 d, 15 d, 30 d, 48 d (respectively four, 15,30, and 48 days old whole larvae), Gu (gut), Sp (spleen), He (heart), Mu (skeletal muscle), Gi (gills), Li (liver), Br (brain), Hk (head-kidney). Control templates wereRT reaction without RT enzyme, sterile water, sea bream genomic DNA. Size standard was φ×174/HaeIII (Promega). The expected size of the two fragments (TLR9Aand TLR9B) is shown. B) RT-PCR analysis of β-actin expression in larvae and juvenile tissues. Abbreviations and size standards are as in Fig. 3 A. C) RT-PCRanalysis of β-actin expression in adult tissues and controls. Abbreviations and size standards are as in Fig. 3 A.

48 R. Franch et al. / Gene 378 (2006) 42–51

3.4. Quantitative analysis of SaTLR9 expression

A preliminary analysis was carried out using a Real-TimeRT-PCR assay specifically designed to calculate the relativeexpression of each splicing variant. Two experiments wereperformed. In the first one, the expression of TLR9A and TLR9Bwas evaluated in normal larval and adult samples (Table 2).TLR9A was less abundant in early larval stages (4 days and30 days post hatching) than in whole larvae near to themetamorphosis (48 days). Immune-related organs had a higherlevel of TLR9A, without any difference between juvenile and

Table 2Relative expression of TLR9A and TLR9B in normal samples

Sample TLR9A TLR9B

Whole-larvae (4 days) 1.1 11.1Whole-larvae (30 days) 1.0 2.3Whole-larvae (48 days) 2.5 1.0Spleen (juvenile) 2.9 33.2Spleen (adult) 2.7 3.0Head-kidney (adult) 19.3 19.2

adult spleen. As already observed in the previous RT-PCR assay(Fig. 3), TLR9B expression decreased during larval develop-ment. Likewise, adult spleen showed a ten-fold lowerexpression of the intron-retaining isoform compared to juvenilespleen. In the second experiment, spleen tissue samples ofjuvenile sea breams (0.5 g) were analyzed. A non-parametrictest for independent samples (Mann–Whitney test) rejected thehypothesis that pathogen exposure might significantly alter theexpression of TLR9 under the present experimental conditions.The five individuals that were immersed in clean seawater(mock-infection) showed an average higher expression ofTLR9A and TLR9B (mean values were respectively 4.4 and3.8) compared to individuals exposed to low virulence P.damselae subpsecies piscicida (mean values 3.7 and 2.6), butthe observed differences were not statistically significant(TLR9A pN0.6, TLR9B pN0.3). Both isoforms showed abroad range of expression across individual fishes with aeight-fold difference between the most extreme values [TLR9A3.7±2.4 (infected animals), 4.4±2.6 (controls); TLR9B 2.6±2.0(infected animals), 3.8±2.3 (controls)]. A highly significantcorrelation (Spearman rank correlation test, rho=0.915

49R. Franch et al. / Gene 378 (2006) 42–51

pb0.00001) was found when comparing expression values ofTLR9A and TLR9B in each sample.

4. Discussion

In silico analyses of two fish genomes (Danio rerio, thezebrafish, and Fugu rubripes, the Japanese pufferfish) showedthat a putative TLR9 orthologue exists in lower vertebrates(Oshiumi et al., 2003; Meijer et al., 2004; Jault et al., 2004).However, despite its important role in the host response to bac-terial and viral infections as well in the reaction to immuneadjuvants, neither full-length sequencing nor in-depth expressionanalysis of TLR9 genes have been reported to date in vertebratespecies other than mammals. To our knowledge, the gilthead seabream TLR9 is the first non-mammalian complete sequencereported to date.

A recent study suggested that sea bream leucocytes produceIL1-β after stimulation with DNA of Vibrio anguillarum, but notwith salmon DNA (Pelegrin et al., 2004). Based on sequencehomology, SaTLR9 appears as a good candidate for the role ofbacterial DNA receptor in the sea bream. SaTLR9 shows thetypical domain architecture of TLR9 proteins, with a large extra-cellular region characterized by 21 leucine-rich repeats and acanonical TIR domain in the intracellular region. Comparativesequence analysis suggests that functional constraints at certainsites in vertebrate TLR9s might be different between mammalsand teleosts. Functional divergence, however, is limited to theextracellular domain, and it seems to be due to the presence of fast-evolving positions in mammalian TLR9s, especially in rodents.This evidence is in agreement with the observation of a species-specificity in the recognition of unmethylated CpG oligodeox-ynucleotides (ODN) for mammalian TLR9s. Murine TLR9recognizes different ODN sequence motifs than human or porcineTLR9 (Bauer et al., 2001; Shimosato et al., 2003). Therefore, theobserved functional divergence might be related to fine tuning ofspecificity in ligand-binding between mammalian species ratherthan reflecting a major shift of ligand between vertebrate lineages.

Little is known about the amino acid residues that are involvedin ligand-binding of TLR9. Recently, Rutz et al. (2004) found thatmammalian TLR9s share with methylated CpG-DNA bindingproteins (MBD1-4) a specificDNA-bindingmotif termedmethyl-CpG-DNAbinding domain (MBD). Site-directedmutation of tworesidues within the TLR9MBD-like domain strongly diminishedthe CpG-sequence-specific binding of murine TLR9. Comparingpiscine andmammalian TLR9s, these two residues are found to beperfectly conserved and the MBD-like domain displays highsequence similarity across all the analyzed species (Fig. 1). In adifferent study, Tabeta et al. (2004) reported that a chemically-induced non-conservative mutation (Leu→Pro) at position 499of TLR9 in mice abolished the sensitivity to stimulation withCpG-ODN. At this protein site, teleost TLR9s show only con-servative amino acid substitutions (Leu→Val, Val→Ala)(Fig. 1). None of these critical residues were included in the setof functionally divergent sites.

Less conserved across vertebrates is the intron–exon structureof the TLR9 gene. In mammals, a single intron is present, close tothe start codon (Du et al., 2000). In teleosts, no introns have been

reported for zebrafish TLR9 (Meijer et al., 2004), whereas a singleintron has been found in fugu, but in a different position, dividinginto two exons the region encoding the TIR domain (Oshiumiet al., 2003). In the sea bream, two introns were observed (Fig. 1).The first one is likely homologous to the only intron found inmammals, the second one is in the same position of the intron infugu. In medaka and tetraodon a single intron corresponding tothe second intron of SaTLR9was observed (data not shown). Thepresence of a first intron located in the 5′ region of the gene inmammals and sea bream and its absence in fugu, tetraodon,medaka, and zebrafish is likely due to incomplete gene predictionof genomic sequences in these species. The position of this intronis close to the 5′ UTR, which is often difficult to predict withoutexperimental evidence of full-length mRNA sequence. Thesecond intron in SaTLR9 is found in all the other teleost species,except for zebrafish, while it is absent in all mammals examinedthus far. A possible explanation is that this intron is an evolu-tionary novelty that appeared after the separation between Osta-riophysi (zebrafish) and Neoteleostei (fugu, tetraodon, medaka,sea bream) (Ishiguro et al., 2003). The observed difference in thegenomic organization of the TLR9 gene between these two line-ages of teleost fishes is extremely relevant because the secondintron was found to be involved in the alternative splicing ofSaTLR9. Retention of this intron generates an incompletelyspliced transcript (TLR9B). Intron-retention represents one ofthe four major types of alternative splicing. In a recent surveyof the human transcriptome, intron-retaining transcripts havebeen found in 14.8% of over 21,000 known genes (Galanteet al., 2004). Recent studies support the potential of both extra-and intracellular signalling to the splicing machinery as ameans of regulating gene expression, and indicates that thisform of gene control is widespread and mechanistically com-plex (Black, 2003; Shin andManley, 2004). Analysis of TLR9Aand TLR9B expression provides experimental evidence thatthe ratio between these two alternative transcripts is variableamong different tissues and developmental stages. On theopposite, the expression of the two isoforms appears to behighly correlated when the same tissue and developmentalstage are investigated.

The presence of two alternative isoforms might affect thefunction of SaTLR9 at themRNA stage and at the protein level. Inthe first case, the generation of incompletely spliced transcriptsmight reduce the efficiency of producing mature TLR9 mRNAsand compete for nuclear export. In the second case, the TLR9Btranscript encodes a premature termination codon producing aputative protein lacking 58 amino acids of the C-terminal domain.The putative TLR9B protein is likely to be defective, since one ofthe three conserved boxes identified in the TIR domain ofvertebrate TLRs is missing (Fig. 1). Experimental deletion of 32amino acids at the C-terminal of murine TLR9 showed that thedeleted protein is inactive, suggesting that this region plays acritical role in intracellular signalling. The C-terminus deletedprotein also suppressed the activity of wild-type mTLR9 in adose-dependent manner (Takeshita et al., 2004). However, itremains to be clarified whether TLR9BmRNA undergoes nuclearexport and translation as for completely spliced transcripts. It willalso be interesting to investigate the presence of intron-containing

50 R. Franch et al. / Gene 378 (2006) 42–51

TLR9 isoforms in other teleost species that, contrary to mammalsand zebrafish, appear to have acquired an intervening sequence inthe 3′ end of the coding region.

While the potential for alternative splicing of the 3′ intron islimited to higher teleost fishes (Neoteleostei), the expression ofTLR9A can be compared across all the vertebrate species inves-tigated so far. In mammals, TLR9 is mainly expressed in dendriticcells and B lymphocytes, with species-specific differences be-tween humans and mice. More in general, TLR9 is found in abroad array of tissues and organs. This observation likely reflectsthe presence of antigen-presenting cells in TLR9-positive tissue,although also non-immune cells show expression of TLR9,especially epithelial cells. In agreement with this evidence, spleenand head-kidneywere found to bemarkedly positive for TLR9A inthe sea bream, together with mucosal–epithelial barriers (skin,gills, gut). Liver, brain, and skeletal muscle showed weak orabsent expression. Where a comparison is possible, these resultsare in good agreement with evidence for human (Zarember andGodowski, 2002), swine (Shimosato et al., 2003), cattle (Gold-ammer et al., 2004), fugu (Oshiumi et al., 2003), and zebrafishTLR9 (Jault et al., 2004). A relevant exception is the absence ofTLR9 in the gills of F. rubripes (Oshiumi et al., 2003).

It has been described that diverse microbial stimuli affect theexpression of their cognate TLR as well as of other TLR genes.TLR9 expression in human leucocytes is decreased upon stimu-lation with either CpG-ODN (Hornung et al., 2002) or liveEscherichia coli (Zarember and Godowski, 2002). Detailedanalysis of promoter activity in B-cell lines confirmed that humanTLR9 is down-regulated either by stimulation with CpG-ODN,LPS or over-expression of NF-κB and c-Jun (Takeshita et al.,2004). In cattle, experimental infection of the udder with variousbacterial species did not affect significantly TLR9, whereas it up-regulated TLR2 and TLR4 expression (Goldammer et al., 2004).Other studies, however, have reported CpG-ODN- or LPS-mediated induction of TLR9 expression in immature B cells(Bernasconi et al., 2003; Bourke et al., 2003). In the zebrafish,non-quantitative RT-PCR analysis revealed up-regulation ofTLR9 in animals experimentally infected with Mycobacteriummarinum (Meijer et al., 2004). In the sea bream, preliminary datafrom quantitative RT-PCR indicated that expression of TLR9Aand TLR9B is not significantly altered after exposure to a lowvirulent Gram-negative bacterial pathogen. This evidence is inagreement to what was obtained under similar experimentalconditions in cattle (Goldammer et al., 2004). Expression ofTLR9appears therefore to be differently regulated depending on thestimuli, the cell type, and the species involved. The broad range ofTLR9 expression across individual fish for the same organ anddevelopmental stage (eight-fold differences) requires furtherinvestigation to estimate inter-individual variation in the capacityof mounting innate immune responses.

In conclusion, sequence and expression data reported here forthe gilthead sea bream suggest overall conservation of the proteinprimary structure and interesting similarities in the transcriptionalregulation of TLR9 across vertebrates. At the same time, a uniquepattern of alternative splicing was observed. It remains to beassessed whether this pattern is shared among the most derivedgroup of teleost fishes (Neoteleostei). Neoteleostei include the

Perciform order, which comprises more than 7000 species world-wide, several of which are of great ecological and/or commercialimportance (e.g. tunas, billfish, tilapia, basses, sea breams). Al-though Perciforms are likely not monophyletic, they represent amore derived group compared to Ostariophysi (zebrafish, carp.goldfish) in the evolution of teleost fishes and are distantly relatedto traditional fish models such as the zebrafish (Chen et al., 2004).Therefore, the gilthead sea bream might serve as a phylogenet-ically closer model to investigate the functioning of the immunesystem in higher teleost fishes in addition to traditional fishmodels (zebrafish, goldfish).

Acknowledgements

We thank Amedeo Manfrin, Mirko Volpin, and FabioBorghesan for their help with experimental infection, PierpaoloPatarnello and Roberto Menegatti at Valle Ca' Zuliani for pro-viding living animals. This work was partially funded by a YoungInvestigator Research Grant from the University of Padova toBarbara Cardazzo.

All the experiments carried out in the present study complywith the Italian regulations for experimental activity with livinganimals and hazardous reagents.

References

Abascal, F., Zardoya, R., Posada, D., 2005. ProtTest: selection of best-fit modelsof protein evolution. Bioinformatics 21, 2104–2105.

Aderem, A., Ulevitch, R.J., 2000. Toll-like receptors in the induction of theinnate immune response. Nature 406, 782–787.

Akira, S., Hemmi, H., 2003. Recognition of pathogen-associated molecularpatterns by TLR family. Immunol. Lett. 85, 85–95.

Bauer, S., et al., 2001. Human TLR9 confers responsiveness to bacterial DNAvia species-specific CpG motif recognition. Proc. Natl. Acad. Sci. U. S. A.98, 9237–9242.

Bernasconi, N.L., Onai, N., Lanzavecchia, A., 2003. A role for Toll-likereceptors in acquired immunity: up-regulation of TLR9 by BCR triggeringin naive B cells and constitutive expression in memory B cells. Blood 101,4500–4504.

Black, D.L., 2003. Mechanisms of alternative pre-messenger RNA splicing.Annu. Rev. Biochem. 72, 291–336.

Bourke, E., Bosisio, D., Golay, J., Polentarutti, N., Mantovani, A., 2003. Thetoll-like receptor repertoire of human B lymphocytes: inducible andselective expression of TLR9 and TLR10 in normal and transformed cells.Blood 102, 956–963.

Chen, W.J., Orti, G., Meyer, A., 2004. Novel evolutionary relationship amongfour fish model systems. Trends Genet. 20, 424–431.

Du, X., Poltorak, A., Wei, Y., Beutler, B., 2000. Three novel mammalian toll-like receptors: gene structure, expression, and evolution. Eur. CytokineNetw. 11, 362–371.

Galante, P.A., Sakabe, N.J, Kirschbaum-Slager, N., de Souza, S.J., 2004.Detection and evaluation of intron retention events in the humantranscriptome. RNA 10, 757–765.

Goldammer, T., et al., 2004. Mastitis increases mammary mRNA abundance ofbeta-defensin 5, toll-like-receptor 2 (TLR2), and TLR4 but not TLR9 incattle. Clin. Diagn. Lab. Immunol. 11, 174–185.

Gu, X., 1999. Statistical methods for testing functional divergence after geneduplication. Mol. Biol. Evol. 16, 1664–1674.

Gu, X., 2003. Functional divergence in protein (family) sequence evolution.Genetica 11, 133–141.

Guindon, S., Gascuel, O., 2003. A simple, fast, and accurate algorithm toestimate large phylogenies by maximum likelihood. Syst. Biol. 52,696–704.

51R. Franch et al. / Gene 378 (2006) 42–51

Hemmi, H., et al., 2000. AToll-like receptor recognizes bacterial DNA. Nature408, 740–745.

Hornung, V., et al., 2002. Quantitative expression of toll-like receptor 1–10mRNAin cellular subsets of human peripheral bloodmononuclear cells and sensitivityto CpG oligodeoxynucleotides. J. Immunol. 168, 4531–4537.

Ishiguro, N.B., Miya, M., Nishida, M., 2003. Basal euteleostean relationships: amitogenomic perspective on the phylogenetic reality of the “Protacanthop-terygii”. Mol. Phylogenet. Evol. 27, 476–488.

Jault, C., Pichon, L., Chluba, J., 2004. Toll-like receptor gene family and TIR-domain adapters in Danio rerio. Mol. Immunol. 40, 759–771.

Latz, E., et al., 2004. TLR9 signals after translocating from the ER to CpG DNAin the lysosome. Nat. Immunol. 5, 190–198.

Meijer, A.H., et al., 2004. Expression analysis of the Toll-like receptor and TIRdomain adaptor families of zebrafish. Mol. Immunol. 40, 773–783.

Murphy, W.J., et al., 2001. Resolution of the early placental mammal radiationusing Bayesian phylogenetics. Science 294, 2348–2351.

O'Neill, L.A.J., Fitzgerald, K.A., Bowie, A.G., 2003. The Toll-IL-1 receptoradaptor family grows to five members. Trends Immunol. 24, 286–289.

Oshiumi, H., Tsujita, T., Shida, K., Matsumoto, M., Ikeo, K., Seya, T., 2003.Prediction of the prototype of the human toll-like receptor gene family fromthe pufferfish, Fugu rubripes, genome. Immunogenetics 54, 791–800.

Pelegrin, P., Chaves-Pozo, E., Mulero, V., Meseguer, J., 2004. Production andmechanism of secretion of interleukin-1beta from the marine fish giltheadseabream. Dev. Comp. Immunol. 28, 229–237.

Ririe, K.M., Rasmussen, R.P., Wittwer, C.T., 1997. Product differentiation byanalysis of DNAmelting curves during the polymerase chain reaction. Anal.Biochem. 245, 154–160.

Rutz, M., et al., 2004. Toll-like receptor 9 binds single-stranded CpG-DNA in asequence- and pH-dependent manner. Eur. J. Immunol. 34, 2541–2550.

Shimosato, T., et al., 2003. Swine Toll-like receptor 9 recognizes CpG motifs ofhuman cell stimulant. Biochim. Biophys. Acta 1627, 56–61.

Shin, C., Manley, J.L., 2004. Cell signalling and the control of pre-mRNAsplicing. Nat. Rev., Mol. Cell Biol. 5, 727–738.

Tabeta, K., et al., 2004. Toll-like receptors 9 and 3 as essential components ofinnate immune defense against mouse cytomegalovirus infection. Proc. Natl.Acad. Sci. U. S. A. 101, 3516–3521.

Takeda, K., Kaisho, T., Akira, S., 2003. Toll-like receptors. Annu. Rev.Immunol. 21, 335–376.

Takeshita, F., Suzuki, K., Sasaki, S., Ishii, N., Klinman, D.M., Ishii, K.J., 2004.Transcriptional regulation of the human TLR9 gene. J. Immunol. 173,2552–2561.

Trede, N.S., Langenau, D.M., Traver, D., Look, A.T., Zon, L.I., 2004. The use ofzebrafish to understand immunity. Immunity 20, 367–379.

Wang, Y., Gu, X., 2001. Functional divergence in the caspase gene family andaltered functional constraints: statistical analysis and prediction. Genetics158, 1311–1320.

Zappulli, V., et al., 2005. Direct identification of Photobacterium damselaesubspecies piscicida by PCR-RFLP analysis. Dis. Aquat. Org. 65, 53–61.

Zarember, K.A., Godowski, P.J., 2002. Tissue expression of human Toll-likereceptors and differential regulation of Toll-like receptor mRNAs in leukocytesin response to microbes, their products, and cytokines. J. Immunol. 168,554–561.