Evolution of the Inflammatory Response in Vertebrates: Fish TNF-α Is a Powerful Activator of Endothelial Cells but Hardly Activates Phagocytes¹

The cytokine TNF-α is a powerful proinflammatory cytokine released by several immune cells during infection or tissue damage and is involved in a diverse range of inflammatory, infectious, and malignant conditions. TNF-α exists in two forms, a membrane-bound and a soluble form, each possibly with its own distinct physiological role (1). Hence, TNF-α is expressed as a 26 kDa membrane-bound precursor, which is proteolytically cleaved by a disintegrin metalloproteinase (TNF-α-converting enzyme) to give a 17 kDa C-terminal active form (2, 3). TNF-α exerts its effects by binding, as a trimer, to a 55 kDa cell membrane receptor termed TNFR1, or to a 75 kDa cell membrane receptor termed TNFR2 (4, 5), both receptors being abundant in most cells (6, 7). Recent studies with knockout mice have shown that TNFR1 predominantly mediates both the proinflammatory and apoptosis pathways, whereas TNFR2 mediates signals that promote tissue repair and angiogenesis. For this reason, mammalian TNF-α has been dubbed a double-edged sword, because it induces either cell migration, proliferation, survival, and differentiation or cell death (8). Probably one of the major functions of TNF-α is the regulation of the interactions between the endothelium and leukocytes through the up-regulation of the endothelial adhesion molecules P- and E-selectins and ICAM-1, which results in reduced velocity of leukocyte rolling in postcapillary venules and their subsequent extravasation (9, 10, 11).

TNF-α has been identified, cloned, and characterized in several bony fish, including Japanese flounder (12), rainbow trout (13, 14), gilthead seabream (15), carp (16), catfish (17), tilapia (18), turbot (19), mandarin fish (20), and goldfish (21). These studies have revealed the existence of some obvious differences from their mammalian counterpart, such as the presence of multiple isoforms of TNF-α in some teleost species (14, 16); the high constitutive expression of this gene in different tissues of healthy fish and its relatively poor up-regulation by immune challenge in vitro and in vivo (13, 15, 17, 18, 22); and the dimeric nature of the recombinant protein (23). However, the most unexpected and interesting difference between fish and mammal TNF-α concerns the weak in vitro effects on phagocytes. Thus, although high concentrations (10–2500 ng/ml) of recombinant TNF-α weakly induce chemotaxis, respiratory burst, phagocytosis, and a NO response of macrophages in goldfish (21) and rainbow trout (17), TNF-α (alone or combined with LPS) fails to trigger the respiratory burst of phagocytes in turbot (19) and gilthead seabream (23). This weak in vitro activity of fish TNF-α sharply contrasts with the powerful actions exerted by the i.p. injection of recombinant TNF-α in gilthead seabream, which includes the recruitment of phagocytes to the injection site with a concomitant strong increase in their respiratory burst (23). These data raise questions concerning the evolution of TNF-α functions and, particularly, the roles played by this cytokine in the regulation of the inflammatory response in different vertebrate groups. Therefore, we have studied the in vivo biological activities displayed by fish TNF-α using an immunologically (gilthead seabream) and a genetically (zebrafish) tractable model, paying special attention to the effect of TNF-α on endothelial cells. The results show that endothelial cells are the main target cells of TNF-α, suggesting that fish TNF-α is mainly involved in the recruitment of leukocytes to the inflammatory foci rather than in their activation. In addition, we also show that TNF-α increases the susceptibility of zebrafish to viral (spring viremia of carp virus; SVCV³) and bacterial (Streptococcus iniae) infections.

Gilthead seabream (Sparus aurata L., Perciformes, Sparidae) adults (150 g mean weight) were kept in 260 liters of running seawater aquaria (flow rate, 1500 liters/h) at 23°C under a 12-h light-dark cycle and fed with a commercial pellet diet (Skretting) at a feeding rate of 15 g of dry diet per kg biomass of fish per day.

Wild-type zebrafish (Danio rerio H., Cypriniformes, Cyprinidae) were obtained from the Zebrafish International Resource Center and maintained as described in the zebrafish handbook (24). The transgenic zebrafish line that expresses enhanced GFP under the neutrophil-specific promoter for the myeloid-specific peroxidase (mpx) gene (Tg(mpx::gfp)ⁱ¹¹⁴) has been described previously (25). All animal studies were conducted in accordance with the European Union regulations for animal experimentation.

Zebrafish membrane-bound TNF-α (zfproTNFα; accession number AY427649) was obtained by PCR amplification using a proof-reading DNA Polymerase (Pfu, Fermentas). The PCR primers were proTNF-F and TNF-R, whereas cDNA from SVCV-infected zebrafish was used as template. The PCR-amplified fragments were cloned into the pGEM-T Easy vector (Promega) and then subcloned into the BamHI site of the pcDNA6/V5-His vector (Invitrogen) to express the V5/His₆-tagged protein. The powerful signal peptide of the seabream type II IL-1R (26) was inserted in frame into the NheI-KpnI site of the zfTNFα-V5/His₆ construct using the annealed IL1R2-F and IL1R2-R primers (Table I) to direct the recombinant protein to the plasma membrane. Precursor zfTNFα was found to be expressed, processed, and released in HEK293 by means of Western blot with the anti-V5 Ab (Invitrogen).

The putative mature zfTNFα (residues 74–234) was also cloned into the pET15b bacterial expression vector (Novagen). Briefly, the zfTNFα was amplified with Pfu DNA polymerase with the primers zfTNF-F2 and zfTNF-R2 and the zfproTNFα-V5/His₆ construct as template. The PCR-amplified fragment was cloned into the pCRII-TOPO cloning vector (Invitrogen) and then subcloned into the BamHI site of the pET15b vector for bacterial expression of the His₆-tagged protein. All constructs were sequenced using an ABI PRISM 377 (Applied Biosystems). Recombinant His₆-tagged zfTNFα and seabream TNF-α (sbTNFα) were produced in Escherichia coli, purified by metal affinity chromatography as described earlier (23), and stored at −80°C in small aliquots in the presence of 40% glycerol.

Seabream specimens were injected i.p. with 1 ml of PBS containing 10 μg of sbTNFα or vehicle alone (PBS plus elution buffer plus 40% glycerol) (23). Four and sixteen hours postinjection, peritoneal exudate cells and fragments of the peritoneum were collected and processed for gene expression analysis (see below), whereas the culture medium injected to obtain the cells (conditioned media) was clarified by filtration (0.45 μm) and stored at −80°C for further analysis.

Five mpx::gfp adult zebrafish were anesthetized by immersion in benzocaine (100 μg/ml) (Sigma-Aldrich) before injection. Each fish was injected in the left epaxial muscle with 1 μl of PBS containing 100 ng of pcDNA6/V5-His (mock) or zfproTNFα-V5/His₆ plasmids together with 10 ng of pDsRed-Monomer-N1 vector (Clontech). Fish were monitored daily over a 10-day period for monomeric red-fluorescent protein expression and neutrophil (enhanced GFP⁺ cells) recruitment using a Lumar stereomicroscope (Zeiss). Tissue samples from the injection site were also collected and processed for histology (see Histology).

Seabream head kidney (fish bone marrow equivalent) leukocytes obtained as described elsewhere (27) were maintained in RPMI 1640 culture medium (Invitrogen) adjusted to gilthead seabream serum osmolarity (353.33 mOs) with 0.35% NaCl; sRPMI) supplemented with 5% FBS (Invitrogen), 100 IU/ml penicillin, and 100 μg/ml streptomycin (P/S; Biochrom). Most experiments were conducted using purified cell fractions of macrophages and acidophilic granulocytes, the two professional phagocytic cell types of this species (27, 28). Briefly, acidophilic granulocytes were isolated by MACS using a mAb specific to gilthead seabream acidophilic granulocytes (G7) (27), whereas macrophages were isolated from G7⁻ cell fractions by adherence to the culture flasks after overnight incubation (28). The remaining nonadherent cell fraction was mainly constituted by lymphocytes according to the expression of the H chain of IgM (28) and the α and β chains of the TCR (unpublished results).

Seabream endocardium endothelial cells (EEC) were isolated as previously described (29) with slight modifications. The atria from 10–15 hearts were collected and incubated in the following solutions for the indicated times: 1) PBS for 30 min; 2) trypsin (0.5 mg/ml) and EDTA (0.1 mg/ml) in PBS for 5–10 min; and 3) collagenase (0.5 mg/ml collagenase in sRPMI supplemented with 0.7 mg/ml CaCl₂ and 50 μg/ml gentamicin) for 30 min. The cell pellet was resuspended in sRPMI supplemented with 10% FBS and 50 μg/ml gentamicin and seeded on 75-cm² tissue culture plastic flasks precoated with 0.1% gelatin in PBS for 30 min at room temperature. After 36 h of incubation at 23°C to allow EECs to adhere, debris and nonadherent cells were removed by vigorous shaking of flasks and by washing them twice with medium. Finally, the medium was replaced by fresh growth medium, and the cultures were left undisturbed.

Macrophages, acidophilic granulocytes, and EECs were stimulated for 3 h or at 23°C with 50 μg/ml phenol-extracted genomic DNA from Vibrio anguillarum ATCC19264 cells (VaDNA), 1 μg/ml flagellin (Invivogen), or 100 ng/ml recombinant sbTNFα (rsbTNFα) in sRPMI supplemented with 10% FBS and P/S. The concentrations of flagellin and VaDNA have been found to be optimal for in vitro activation of seabream phagocytes (22). Cell were processed for gene expression analysis as described below, and the conditioned media from EECs were collected, clarified by filtration (0.45 μm), and stored at −80°C. To check whether the effects of rsbTNFα were solely due to the recombinant cytokine rather than to possible contamination with LPS from E. coli, we evaluated the activity of sbTNFα samples that had been preheated at 80°C for 30 min and of samples depleted of sbTNFα by metal affinity chromatography.

The zebrafish embryonic cell line ZF4 was purchased from the American Type Culture Collection, and the epithelioma papulosum cyprinid cell line (EPC) was kindly provided by Dr. A. Estepa (Miguel Hernández University, Elche, Spain). ZF4 cells were maintained at 28°C in DMEM/F12 culture medium (Invitrogen) supplemented with 10% FBS, 15 mM HEPES, 0.5 mM sodium pyruvate (Sigma-Aldrich), and P/S, whereas EPC cells were grown at 25°C in DMEM (Invitrogen) supplemented with 10% FBS, 1 mM sodium pyruvate, 2 μg/ml amphotericin B (Biochrom), and 50 μg/ml gentamicin (Biochrom). ZF4 cells were stimulated for 1, 3, 5, and 16 h with 100 ng/ml rzfTNFα for gene expression analysis. ZF4 cells were also pretreated for 2 h with 100 ng/ml rzfTNFα before being infected with 4 × 10³ or 4 × 10⁴ 50% tissue culture-infective dose (TCID₅₀)/ml of SVCV, incubated for 48 h at 25°C in the presence of 100 ng/ml rzfTNFα, and then fixed and stained for 2 h in a solution containing 4% paraformaldehyde, 1% crystal violet, and 0.9% NaCl. The number and area of the plaques were visualized using a Nikon inverted microscope (Eclipse TE2000) and analyzed with the MIP-4.5 image analysis software (Digital Image System).

Respiratory burst activity was measured as the luminol-dependent chemiluminescence produced by seabream leukocytes stimulated for 4 and 16 h (head kidney) or 5 and 16 h (peritoneal exudate) at 23°C with 50 μg/ml VaDNA and/or 100 ng/ml rsbTNFα in sRPMI supplemented with 10% FBS and P/S (30). This was brought about by adding 100 μM luminol (Sigma-Aldrich) and 1 μg/ml PMA (Sigma-Aldrich), whereas the chemiluminescence was recorded every 117 s for 1 h in a FLUOstart luminometer (BGM; LabTechnologies). The values reported are the average of quadruple readings, expressed as the slope of the reaction curve from 117 to 1170 s, from which the apparatus background was subtracted.

Total head kidney leukocyte and lymphocyte fractions were incubated for 16 h and acidophilic granulocytes were incubated for 4 h with 10 μg/ml LPS and 50 μg/ml VaDNA in sRPMI supplemented with 5% FBS and P/S or left untreated (control). They were then labeled with the fluorescent dye 2′,7′-bis(2-carboxyethyl)-5(6)-carboxyfluorescein acetoxymethyl ester (BCECF-AM, Sigma-Aldrich) and added in sRPMI medium to a monolayer of resting or rsbTNFα-activated EECs (100 ng/ml, 24 h). After incubation for 20 min at 23°C, unbound cells were removed by three washes with sRPMI medium, and adhered cells were quantified using a fluorescence analyzer (BMG).

Chemotaxis assay was performed using a 10-well transmigration chamber (Neuroprobe) as described previously (31). Conditioned media from EECs and peritoneal exudates, together with the positive (10% autologous seabream serum) and negative (medium alone) control samples, were placed in the lower wells, and a head kidney cell suspension (2 × 10⁶ cells/well) was placed in the upper wells of the transmigration chamber. Upper and lower compartments of the transmigration chamber were separated by polyvinyl pyrrolidone-pretreated polycarbonate filter with 3-μm pores (Neuroprobe). Filled transmigration chambers were incubated for 3 h at 23°C with 5% CO₂. Transmigrated cells were harvested from the lower wells, analyzed with a flow cytometer (BD Biosciences), and counted with a Neubauer chamber.

Total RNA was extracted from tissues or cell pellets with TRIzol reagent (Invitrogen) following the manufacturer’s instructions and treated with DNase I, amplification grade (1 U/μg RNA; Invitrogen). SuperScript III RNase H⁻ ReverseTranscriptase (Invitrogen) was used to synthesize first-strand cDNA with oligo(dT)₁₈ primer from 1 μg of total RNA at 50°C for 50 min. Real-time PCR was performed with an ABI PRISM 7500 instrument (Applied Biosystems) using SYBR Green PCR Core Reagents (Applied Biosystems). Reaction mixtures were incubated for 10 min at 95°C, followed by 40 cycles of 15 s at 95°C, 1 min at 60°C, and finally 15 s at 95°C, 1 min at 60°C, and 15 s at 95°C. For each mRNA, gene expression was normalized to the ribosomal proteins S18 (seabream) or S11 (zebrafish) content in each sample using the comparative C_t method (2^−ΔΔCt). The primers used are shown in Table I. In all cases, each PCR was performed with triplicate samples and repeated at least with two independent samples.

The SVCV isolate 56/70 was kindly provided by Dr. P. Fernández-Somalo (Laboratorio Central de Veterinaria, Ministerio de Medio Ambiente y Medio Rural y Marino, Algete, Spain). The virus stock was propagated in EPC cells and titrated in 96-well plates according to the report of Reed and Muench (32). The strain KFP 404 (serotype II) of S. iniae was kindly provided by Dr. A. Eldar and cultured in brain-heart infusion broth as described previously (33). Ten adult zebrafish were challenged at 26°C in 10-liter tanks by i.p. injection of 10⁵ TCID₅₀/fish SVCV (34) or 10³ CFU/fish S. iniae (35) and 10 or 100 ng/ml rzfTNFα. After challenge, the fish were monitored every 12 h during a 15-day period for clinical signs of disease and mortality.

Tissue fragments were fixed for 24 h in 1% acetic acid, 3% zinc chloride, and 5.6% formaldehyde; embedded in Paraplast Plus (Sherwood Medical); and sectioned at a thickness of 5 μm. After being dewaxed and rehydrated, some sections were stained with H&E, whereas others were subjected to an immunohistochemical method using a commercial affinity-purified rabbit anti-human/mouse caspase 3 active (R&D Systems), which recognizes a fully conserved epitope of zebrafish caspase 3.

Data were analyzed by ANOVA and a Tukey multiple range test to determine differences between groups.

We have previously demonstrated that although sbTNFα fails to activate the respiratory burst activity of seabream head kidney phagocytes in vitro, it is able to promote leukocyte infiltration and activation when injected in vivo (23). To test whether seabream phagocytes need to be activated to respond to sbTNFα, we stimulated in vitro these cells with rsbTNFα alone or in combination with different pathogen-associated molecular patterns (PAMPs) before triggering the respiratory burst with PMA. It was observed that TNF-α also failed to activate the respiratory burst of head kidney and peritoneal exudate phagocytes when used in combination with VaDNA, LPS, and flagellin (Fig. 1 and data not shown). In addition, rsbTNFα-elicited peritoneal exudate leukocytes also failed to respond in vitro to rsbTNFα (data not shown).

To understand the mechanism involved in the recruitment and activation of leukocytes upon i.p. injection of TNF-α, we next analyzed the expression of several pro- and anti-inflammatory genes. rsbTNFα-elicited leukocytes showed increased mRNA levels of several proinflammatory genes, mainly IL-6, IL-8, and COX-2, as well as the seabream type II IL-1R (Fig. 2,A). Surprisingly, cells of the peritoneal membrane (Fig. 2 B) showed not only increased mRNA levels of IL-1β and IL-6 but also those of the E-selectin, the expression of which in mammals is restricted to endothelial cells where it is commonly up-regulated by TNF-α. This result suggests that TNF-α might not exert its effects on leukocytes directly and that other molecules, probably produced by endothelial cells, mediate in the biological effects of TNF-α in vivo.

The above results prompted us to examine the effects of rsbTNFα on endothelial cells and compared these effects with those of macrophages. For this, we isolated seabream EECs and found that they have a morphology different from that of macrophages (Fig. 3,A). In addition, they proliferated in culture for up to four weeks and could be successfully subcultured and frozen (data not shown). In addition, the mRNA levels of E-selectin were high in EECs and could be increased by stimulation with VaDNA, flagellin or rsbTNFα (Fig. 3,B). In sharp contrast, E-selectin mRNA was hardly detected in macrophages and could not be up-regulated by PAMPs or rsbTNFα. Surprisingly, the gene coding for the macrophage CSF receptor was constitutively expressed in both EECs and macrophages, although at higher levels in the latter (Fig. 3,B). rsbTNFα was responsible for all these effects, given that preheated and depleted rsbTNFα samples were unable to increase the mRNA levels of E-selectin by EECs (Fig. 3 C). This result confirms our previous study which showed the specificity of the rsbTNFα (23).

We next analyzed the expression of different pro- and anti-inflammatory genes in these two cell populations and found that VaDNA and flagellin drastically increased the mRNA levels of the proinflammatory molecules IL-1β, IL-6, TNF-α, and COX2 in both EECs and macrophages, whereas rsbTNFα slightly increased the mRNA levels of IL-1β in both cells (Fig. 4). In addition, rsbTNFα strongly increased the mRNA levels of CCL4 and IL-8 in EECs but had a very weak effect on the expression of these two chemokines in macrophages (Fig. 4). In contrast, VaDNA and flagellin similarly increased the mRNA levels of CCL4 and IL-8 in EECs and macrophages. As regards the TGFβ1, both PAMPs and rsbTNFα had a negligible effect in both cell types (Fig. 4). However, rsbTNFα reduced the mRNA levels of TGFβ1 in EECs but increased them in macrophages (Fig. 4). Finally, although both PAMPs increased the mRNA levels of the decoy receptor for IL-1β (IL-1RII) (26), stimulation with rsbTNFα had no effect on the expression of this gene in EECs and macrophages.

Because EECs were able to respond to VaDNA and flagellin, they would be expected to express TLR9 and TLR5. Although both macrophages and EECs expressed TLR5 at similar levels, the mRNA levels of the TLR9 were much higher in macrophages (Fig. 5). Strikingly, PAMP or rsbTNFα stimulation failed to increase the mRNA levels of TLR9 in macrophages but drastically increased them in EECs (Fig. 5). On the other hand, EECs and macrophages also expressed the TLR22 gene, which codes for an orphan receptor with no apparent orthologs in mammals (36), and stimulation with PAMPs and rsbTNFα increased its mRNA levels to a similar degree (Fig. 5). In addition, EECs constitutively expressed the homolog of the mammalian LPS-binding protein and bactericidal permeability increasing protein at higher levels than macrophages. However, we found a strong increase in the mRNA levels of this gene in macrophages after stimulation with VaDNA, flagellin, or rsbTNFα (Fig. 5), whereas rsbTNFα lowered these levels in EECs (Fig. 5).

We finally checked the expression of the genes coding for two molecules involved in Ag presentation, CD83 and MHCIIα (spau-daa gene). Both genes were constitutively expressed at higher levels in macrophages than in EECs (Fig. 5). In addition, the mRNA levels of CD83 increased after PAMP or rsbTNFα stimulation in macrophages, whereas PAMPs, but not rsbTNFα, had an effect on ECCs. Strikingly, the mRNA levels of MHCIIα in macrophages were unaffected by PAMP or rsbTNFα stimulation but increased in EECs in response to either stimulus.

The induction of E-selectin in EECs by TNF-α stimulation in vitro and in vivo led us to study the functional relevance of this observation by means of a classical leukocyte-endothelial cell adhesion assay. Stimulation of EECs with rsbTNFα resulted in the enhanced adhesion of LPS-/VaDNA-activated head kidney leukocytes (data not shown). Both lymphocytes and granulocytes bound more efficiently to EECs when they had previously been stimulated with rsbTNFα (Fig. 6). In addition, although rsbTNFα is unable to attract seabream leukocytes, the conditioned media from rsbTNFα-stimulated EECs and from the peritoneal exudates of rsbTNFα-injected fish showed strong chemoattractant activity toward these cells (Fig. 7). Collectively, these results suggest that TNF-α promotes leukocyte recruitment in vivo through the up-regulation of E-selectin and the production of chemokines by EECs.

To confirm the results obtained in the seabream in a different, phylogenetic distant teleost species, we analyzed the recruitment of mpx::gfp-positive neutrophils, in zebrafish injected i.m. with an expression construct for the precursor form of zfTNFα. We found a strong recruitment of neutrophils at 4 and 8 days postinfection (dpi; Fig. 8), as in the gilthead seabream. However, the mRNA levels of several pro- and anti-inflammatory molecules were unaffected in the injection sites 6 and 10 dpi, whereas injections of different PAMPs resulted in up-regulation of proinflammatory cytokines (data not shown). The number of apoptotic cells, assayed by immunohistochemistry using an Ab to active caspase 3, was also similar in mock and zfTNFα-injected fish (data not shown).

This led us to examine the role played by rzfTNFα in bacterial (S. iniae; Fig. 9,A) and viral (SVCV; Fig. 9,B) infections. Unexpectedly, i.p. injection of rzfTNFα resulted in increased mortality of fish to both pathogens. However, injection of fish with rzfTNFα did not induce mortality (Fig. 9,B). Because the increased susceptibility of fish injected with rzfTNFα to SVCV was apparent, we analyzed cytokine expression and viral load in those fish to clarify the mechanisms responsible for the increased susceptibility. The results showed that rzfTNFα-injected fish had slightly higher mRNA levels of the different antiviral (IFN1, IFN2, IFN3, and MxC) genes examined as well as of those coding for anti-inflammatory cytokines (IL-10 and TGFβ1), chemokines (CCL4 and IL-8), and cell adhesion molecules (VCAM1 and ICAM1) (Fig. 9 C and data not shown). Moreover, the viral load, assayed as the mRNA levels of the N protein of the virus, was similar in the early stages of the infection (12 h postinfection, hpi) but it was significantly higher at 36 hpi. These results suggest that TNF-α might facilitate the replication of the virus in host cells rather than promote its dissemination.

The above hypothesis was confirmed by infecting the ZF4 embryonic cells line with SVCV in the presence of rzfTNFα. We first confirmed that ZF4 cells expressed both TNF-αRs (Fig. 10,A) and that stimulation with TNF-α for 1–16 h increased the mRNA levels of different cytokines, including TNF-α, IL-12, lymphotoxin A, IFN-γ, and MxC (data not shown). However, stimulation with TNF-α had no effects on the expression of other cytokines, such as IL-1β, IFN1, IL-10, and TGFβ1. Interestingly, stimulation of these cells with rzfTNFα resulted in an enhanced replication capacity of the virus, as assayed by the number of plaques (Fig. 10,B) and their area (Fig. 10 C). Pretreatment of ZF4 cells with zfTNFα increased the number of plaques but had no effect on their area, suggesting that zfTNFα might facilitate both viral adhesion and replication.

Although the TNF-α gene has been cloned in some teleost fish, the functions displayed by this cytokine in lower vertebrates remain poorly understood. We have previously showed that gilthead sbTNFα shows conserved proinflammatory activities when injected i.p.; i.e., it promotes phagocyte recruitment and activation (23). However, sbTNFα was unable to prime in vitro the respiratory burst of phagocytes when used alone (23) or in combination with different PAMPs (this study). This led us to speculate that this cytokine mainly regulates endothelial cell functions but has little effect on professional phagocytes, in contrast to its effect in mammals. Using complementary fish models, we show that the main pro-inflammatory effects of fish TNF-α are mediated through the activation of endothelial cells. Thus, TNF-α promotes the expression of E-selectin and different CC and CXC chemokines in endothelial cells, which would explain the recruitment and activation of phagocytes observed in vivo in both seabream and zebrafish. It is unlikely that the isolation procedures or the leukocyte sources explain the low responsiveness of fish phagocytes to TNF-α, given that similar results were found with head kidney and peritoneal exudate phagocytes using several complementary in vitro and in vivo techniques. Therefore, our results suggest that the ability of TNF-α to directly activate bird and mammalian neutrophils (37, 38) and macrophages (39, 40, 41) may have appeared after the divergence of fish and tetrapods.

Another important observation of this study is that fish endothelial cells express several TLRs, which would allow them to respond to different PAMPs. Strikingly, TNF-α regulates the expression of TLRs, including TLR9, TLR5, and TLR22, in these cells but has negligible effects on macrophages. Similarly, mammalian endothelial cells are endowed with several TLRs, which allows them to cooperate with professional phagocytes in the recognition of microbes and the induction of inflammation (42, 43). Interestingly, TNF-α and TLR ligands also substantially increased the expression of MHC class II and CD83 in endothelial cells, which might suggest a role for fish endothelial cells and TNF-α in Ag presentation. Murine liver sinusoidal endothelial cells are microvascular endothelial cells with a unique phenotype reminiscent of dendritic cells and a unique function as APCs for CD4⁺ T cells (44). In addition, it has been shown that murine dendritic cells derived from peripheral monocytes express endothelial markers and, in the presence of angiogenic growth factors, differentiate into endothelial-like cells, which show Ag-presenting capacity (45, 46). Collectively, these observations together with the unexpected expression of the macrophage CSF receptor gene by seabream endothelial cells suggest a closer relationship between the monocyte/macrophage and the endothelial cell systems than previously supposed. This might explain the elusive identification of dendritic cells in fish and amphibians as well as the poor affinity maturation observed in these animals even though somatic mutants are generated during the course of their immune responses at the same rate as in other vertebrates (47).

In mammals, TNF-α plays an important role in bacterial and viral infections (48), inducing the expression of α/β IFNs (49, 50) and exerting a strong direct antiviral activity against avian, swine, and human influenza viruses with a potency greater than that of IFN (51). The observed differential bioactivities of TNF-α in fish and higher vertebrates suggest that this cytokine might play a distinct role in lower vertebrates. For example, we found that TNF-α increases the susceptibility of the zebrafish to infection with the rhabdovirus SVCV and the Gram-positive bacterium S. iniae, the effect being particularly apparent in the SVCV infection. Although the powerful actions of fish TNF-α on endothelial cells suggested that it might facilitate pathogen dissemination, we actually found that TNF-α increased antiviral genes and, more importantly, the viral load was unaffected by TNF-α during early infection stages. Although we cannot completely rule out that TNF-α promotes virus dissemination in vivo, because it also increased the expression of cell adhesion molecules, these results strongly suggest that it facilitates viral replication. This was confirmed in ZF4 cells because stimulation of these cells with TNF-α resulted in an increased number of plaques and their area even though it moderately triggers the expression of Mx in these cells. In sharp contrast, using recombinant rabies virus engineered to express either soluble or membrane-bound TNF-α, it has recently been shown that brains of mice infected intranasally with virus expressing soluble TNF-α show significantly less virus dissemination than mouse brains from a mock infection (52). In addition, none of the soluble TNF-α-infected mice succumbed to rabies virus infection, whereas 80% of mock-infected mice died. Interestingly, however, the expression of either form of TNF-α was not associated with increased cell death or induction of α/β IFNs in a cell line (52). Furthermore, reduced virus spread in soluble TNF-α recombinant virus-infected mouse brains was matched by enhanced CNS inflammation, including T cell infiltration and microglial activation, suggesting that TNF-α exerts its protective activity in the brain directly through an IFN-independent, as yet unknown, antiviral mechanism and indirectly through the induction of inflammatory processes in the CNS. Collectively, these results further support the conclusion that fish TNF-α is unable to activate professional phagocytes, and underscore the evolutionary complexity of the regulation of innate immunity by cytokines. The mechanisms orchestrating the increased viral replication in fish host cells treated by TNF-α open the way to future studies that will illuminate not only basic scientific questions regarding the phylogeny of TNF-α signaling in vertebrates but also many aspects of the applied sciences of importance to the aquaculture industry.

We thank Inma Fuentes for her excellent technical assistance, the Servicio de Apoyo a la Investigación of the University of Murcia for the maintenance of cell lines and help with the image analysis, A. Eldar (The Hebrew University-Hadassah Medical School, Jerusalem, Israel) for the KFP404 strain of S. iniae, M. P. Somalo (Laboratorio Central de Veterinaria, Ministerio de Medio Ambiente y Medio Rural y Marino, Algete, Spain) for the 56/70 strain of SVCV, and A. Estepa (Miguel Hernández University, Elche, Spain) for the SVCV titration.

The authors have no financial conflict of interest.

The authors have no financial conflict of interest.