The role of gamma interferon in innate immunity in the zebrafish embryo

Over the last decades, a number of vertebrate and invertebrate species have been used as model organisms for the study of immunity. Among the classic examples, studies in amphibians, fish and birds have illuminated the function and evolution of the adaptive immune system, whereas, more recently, experiments in insects, nematodes and plants have contributed to the understanding of innate immunity. As a more recent addition to the repertoire of genetic model organisms, especially for the use of large-scale ‘forward’ genetic screens, the zebrafish has been employed for the study of disease mechanisms and, in particular, pathogen resistance and innate immunity (for reviews, see Meeker and Trede, 2008; Sullivan and Kim, 2008; Burns and Zon, 2006; Pozos and Ramakrishnan, 2004; Trede et al., 2004; van der Sar et al., 2004; Traver et al., 2003; Yoder et al., 2002).

With its comparatively cheap maintenance, accessibility for experimental manipulation and options for large-scale screens, the zebrafish represents an attractive alternative to mammalian model organisms, both for the study of the evolution of immune defence mechanisms, and for the understanding and analysis of the basic processes underlying immunity in vertebrates.

The adaptive immune system of the zebrafish develops at about 3 weeks of age (reviewed by Trede et al., 2004). This means that within the first three weeks of development the embryo and larva depend solely on innate immune defence mechanisms. Although the embryo is enclosed in a chorion for the first two days of its life, which protects it physically, it is not known whether some pathogens may not be able to penetrate the chorion. At this stage, the embryo is able already to clear experimentally created bacterial infections and can activate typical immune response target genes, suggesting that innate immune signalling and defence pathways must be active. Little is known about the innate defence mechanisms that the zebrafish uses at this, or other, stages.

Clearance of bacteria is achieved mainly by macrophages and neutrophils, which engulf bacterial pathogens shortly after infection, as shown for Mycobacterium marinum (Davis et al., 2002), Escherichia coli, Bacillus subtilis (Herbomel et al., 1999), Vibrio anguillarum (O’Toole et al., 2004), Staphylococcus aureus (Prajsnar et al., 2008) and Salmonella typhimurium (van der Sar et al., 2003; van der Sar et al., 2006). The zebrafish embryo is also able to counteract viral infections, such as spring viremia of carp virus (Sanders et al., 2003; Levraud et al., 2007), snakehead rhabdovirus (Phelan et al., 2005) and nervous necrosis virus (Lu et al., 2008). As in mammals, infection induces acute phase proteins in the zebrafish embryo. During the early phase of infection bacterial pathogens lead to the induction of il1b and tnfa (Pressley et al., 2005), whereas viral infections induce tlr3, tnfa, mxa, mxb, mxc, mxe, viperin (also known as rsad2), iκbα and ifn-φ1 (ifnphi1) (Phelan et al., 2005; Novoa et al., 2006; Levraud et al., 2007; Lu et al., 2008). The ability of the fish embryo to react to infection with an effective immune defence is both of biological interest and of importance for practical reasons. It means that embryos rather than adults can be used in assays involving pathogens, which is an advantage because embryos can be kept in small volumes apart from the main fish facilities, thereby avoiding the risk of contamination of stocks.

The major components of the immune recognition systems and defence signalling pathways are highly conserved between fish and mammals, but significant divergence has been observed as well, both in the way the adaptive immune system creates diversity, and among the innate immune recognition and effector components that are involved in direct interactions with pathogens (Traver et al., 2003; Meijer et al., 2004; Jault et al., 2004; Litman et al., 2001; Yoder et al., 2004; van der Aa et al., 2009; Stein et al., 2007). One set of fast-evolving components that have a central role in immune defence are the interferons and their receptors, of which only a small subset has been studied in the zebrafish.

Interferons, which are classified into three groups (type I, II and III) in mammals, belong to the family of class II cytokines, which mediate immune and inflammatory responses. Apart from the interferons [namely, the single IFN-γ (type II), three IFN-λs (type III), the large IFN-α (type I) expansions, and several others], this family includes the interleukins IL-10, IL-19, IL-20, IL-22, IL-24 and IL-26. Systematic bioinformatic searches and the isolation of cDNAs encoding interferons or interferon-related peptides from various fish species have shown that there is a large number of class II cytokine genes in teleosts. The only clear relationships between fish and mammalian genes are seen in the interleukin group, whereas the interferons are much more divergent (Fig. 1A). None of the predicted or experimentally characterised fish interferons closely resemble the mammalian type I or type III interferons. However, genes have been found in several fish species that are slightly more similar to IFN-γ than to any of the other mammalian interferons, and that have therefore also been named IFN-γ, even though the relationship to mammalian interferons is not supported by high bootstrap values (Stein et al., 2007). Some fish species, including the zebrafish, have two such genes (Igawa et al., 2006; Milev-Milovanovic et al., 2006; Stolte et al., 2008), whereas in other species of fish, only one gene has been found (Zou et al., 2005). The apparent ifn-γ homologs are expressed in immune organs such as the spleen and kidney, as well as in specific immune cells like natural killer (NK) and T cells (Milev-Milovanovic et al., 2006; Stolte et al., 2008; Zou et al., 2005). Zou et al. showed that, in cell culture experiments, the rainbow trout (Oncorhynchus mykiss) IFN-γ is functional and leads to the activation of the target gene Ip-10 and enhanced respiratory burst activity in macrophages (Zou et al., 2005).

Finally, fish have one or more interferons that are not related closely to any of the mammalian interferons: these were named IFN-φs (Stein et al., 2007). This group includes an interferon that has been referred to variously as ‘zebrafish interferon’, ‘IFN A/B’, ‘type I IFN’ and ‘IFN-λ’ (Altmann et al., 2003; Lutfalla et al., 2003; Phelan et al., 2005; Robertsen, 2006; Wang et al., 2006; Levraud et al., 2007), but the phylogenetic distance of this protein from the mammalian type I and type III interferons does not support this unambiguously, and it was therefore renamed IFNφ1 (ZFIN ID: ZDB-GENE-030721-3).

In summary, from the sequence comparison alone, it is not possible to judge which of the fish genes correspond to which mammalian interferon genes, or which of the predicted fish cytokines may fulfil functions similar to those known for mammalian interferons. However, if the zebrafish is to be used as an experimental system for the study of immunity, inflammation and disease, then an understanding of the interferon signalling system is essential, and functional studies are necessary to determine the roles of the interferons.

Functional studies have been carried out in various fish and in cultured fish cell lines. The zebrafish class II cytokine genes whose functions have been analysed in vivo are ifn-φ1, ifn-φ2 (ifnphi2) and ifn-φ3 (ifnphi3). ifn-φ1 is involved in the activation of virus-induced target genes and is necessary for resistance to viral infections in the zebrafish embryo (Levraud et al., 2007); it is also involved in resistance against viral and bacterial infections in adult zebrafish (Lopez-Munoz et al., 2009). ifn-φ2 and ifn-φ3 have been shown recently to mediate resistance against viral infections in adult zebrafish (Lopez-Munoz et al., 2009). However, ifn-φ1 is the only zebrafish interferon for which the corresponding receptor could be identified (Levraud et al., 2007). It was shown to act through a receptor composed of the cytokine receptor family member B1 (CRFB1) and CRFB5 receptor chains. In view of its antiviral activity, and because of its genomic organization, Levraud and colleagues concluded that IFN-φ1 is most similar to the mammalian type III interferon, IFN-λ (Levraud et al., 2007). This further illustrates that the phylogenetic relationships suggested by sequence comparisons are not a reliable basis for deducing functional analogies.

The two ifn-γ genes in the zebrafish are inducible by both lipopolysaccharide (LPS) and polyriboinosinic-polyribocytidylic acid (polyIC) in tissue culture cells (Igawa et al., 2006). One of the genes, ifn-γ2 (also known as ifng1-2), has been shown to induce mxb and mxc in ZF4 cells and mediates resistance against viral infections in these cells (Lopez-Munoz et al., 2009). However, treatment of adult zebrafish with IFN-γ2 failed to mediate resistance against viral and bacterial infections (Lopez-Munoz et al., 2009). To date, the function of ifn-γ1 (also known as ifng1-1) is completely unknown and nothing is known about the expression, induction and function of ifn-γ1 or ifn-γ2 in the zebrafish embryo.

To determine whether the two predicted zebrafish ifn-γ genes correspond to mammalian gamma interferons in their innate immune functions, we used the zebrafish embryo to test their roles in activating target genes and defending the fish against pathogens.

The zebrafish is able to respond to bacterial infections at an early stage of development. To test whether interferon might be involved in the response, we first analysed the stages during development at which the gamma interferon genes ifn-γ1 and ifn-γ2 are expressed. Real-time PCR (RT-PCR) shows high levels of ifn-γ1 mRNA in the eggs immediately after they are laid: the mRNA must be supplied maternally, since zygotic transcription has not started at this point. ifn-γ1 mRNA continues to be expressed, although at lower levels, throughout embryonic development (Fig. 1B). ifn-γ2 mRNA was not found during the early stages of development, but was sometimes seen at very low levels at later stages (Fig. 1B).

The expression of interferon in the absence of infection was unexpected. We wanted to determine whether the expression of the gamma interferons was truly constitutive or whether it might, perhaps, be induced by pathogens present in the fish tanks. Thus, we compared ifn expression in regularly obtained embryos with expression in germfree embryos (see Methods). We did not detect any differences in expression levels, suggesting that basal ifn expression is independent of the presence of microorganisms (data not shown).

The finding that interferons are expressed in the early embryo raises the question of whether the signalling pathway might be active. We used an overexpression approach to test whether ifn-γ1 or ifn-γ2 are able to elicit a response. First, we searched for target genes as a readout for IFN-γ signalling. We selected a set of zebrafish homologs of known mammalian IFN-inducible genes and compared their expression in untreated fish embryos with their expression in embryos injected with in vitro transcribed ifn-γ1 or ifn-γ2 mRNA. The mRNA was injected into 1–2-cell stage embryos and mRNA from pooled embryos was isolated for RT-PCR at 28 hpf.

The genes we tested were ifn-φ1, nramp, adar, mx, ifi30, cathepsin D (ctsd), p27 (cdkn1b), cr3, gp91, lmp2 (also known as psmb9a) and members of two interferon-dependent immune GTPases, the gbp and irg families of genes. These potential target genes responded in different ways. The majority showed a high basal expression, which was not influenced by overexpression of ifn-γ1 or ifn-γ2 (Fig. 2). Except for some irg family members, which are discussed below, the only genes that responded to IFN-γ were ifn-φ1 and lmp2. Both showed low levels of basal expression and were upregulated strongly following injection of ifn-γ1 or ifn-γ2 mRNA (Fig. 2).

The immunity-related GTPase (IRG, p47) family in the zebrafish consists of 11 IRGs and three ‘quasi GTPases’ (Bekpen et al., 2005). Five of the irg genes (irge3, irge4, irgf1, ir gf3 and irgf4) were induced strongly by both IFN-γs, and two (irge2 and irgg1) were induced weakly (Fig. 2). One of the quasi GTPases (irgq1) also showed a weak induction, whereas the other two (irgq2 and irgq3) showed high levels of basal expression that was not altered by ifn-γ overexpression (Fig. 2). All of the genes that reacted to interferon mRNA were induced equally well by ifn-γ1 and ifn-γ2, showing that the two interferons appear to be able to trigger the same responses. In some cases we saw a stronger response to the combination of both interferons, which hints at some non-overlap in their functions, although a particular sensitivity of these genes to the total IFN levels cannot be excluded.

Since both gamma interferons strongly upregulated ifn-φ1 expression, it was possible that their effect on some of the supposed target genes might, in fact, be indirect, i.e. mediated through the induction of ifn-φ1. To test this, we overexpressed ifn-φ1 and measured the effect on the expression of the postulated gamma interferon targets (lmp2 and the irg genes). As a control for the function of ifn-φ1 signalling, we tested the expression of the previously identified ifn-φ1 target gene viperin (Levraud et al., 2007). As expected, we found clear viperin induction upon ifn-φ1 overexpression (Fig. 3A). Expression of ifn-γ1 and ifn-γ2 was not inducible by ifn-φ1, showing that there is no feedback loop from ifn-φ1 on ifn-γ expression (Fig. 3A). The irg genes that were not inducible by ifn-γ were also not influenced by ifn-φ1. Of those genes that responded to ifn-γ, only irge3 and irge4 were also inducible by ifn-φ1 (Fig. 3A), whereas lmp2, irgf1, irgf3 and irgf4 were not. Thus, the induction of the irg genes in the zebrafish is comparable to the situation in the mouse, where most irg genes are specific γ targets and only a few are inducible by other interferons (Sorace et al., 1995; Bafica et al., 2007; Carlow et al., 1998).

The finding that ifn-φ1 is able to trigger the expression of irge3 and irge4 raises the possibility that the reaction of irge3 and irge4 to ifn-γ overexpression is mediated through ifn-φ1. To test this, we overexpressed ifn-γ1 and ifn-γ2 and, in parallel, blocked the ifn-φ1 signalling pathway using a morpholino (crfb1-MO) that was specific for the IFN-φ1 receptor chain CRFB1, which has previously been used successfully (Levraud et al., 2007). In this situation we still observed the activation of irge3 and irge4, but not of the ifn-φ1 response marker viperin (Fig. 3A,B). This shows that irge3 and irge4 can be activated by ifn-γ independently of ifn-φ1. It also shows that the CRFB1 receptor chain is not involved in IFN-γ signalling (Fig. 3B,C).

Since extrinsic RNA can act as a strong immune stimulus, it was possible that activation of the target genes was not the result of IFN-γ production from the injected mRNA, but rather the result of an immune reaction to the injected RNA. To test this, we injected enhanced green fluorescent protein (eGFP) mRNA and monitored the effect on gene expression. eGFP mRNA induced the expression of ifn-φ1, irge3 and irge4, whereas the other genes that were upregulated by IFN-γ1 and IFN-γ2 (irgf1, irgf3, irgf4 and lmp2) did not respond to eGFP mRNA injection (Fig. 3D). Thus, the activation of ifn-φ1, irge3 and irge4 may be part of an antiviral response, fits well with the finding that IFN-φ1 is necessary for an antiviral defence (Levraud et al., 2007). By contrast, irgf1, irgf3, irgf4 and lmp2 appear to be true IFN-γ targets.

In conclusion, these data show that the IFN-γ signalling pathway can be activated in the early zebrafish embryo and that all signalling components needed to induce specific IFN-γ target genes must be present.

Having found that ifn-γ mRNA is present at low levels in the embryo, and that the system necessary to respond to interferon is functional, we wanted to determine whether bacterial infections were able to stimulate interferon signalling. We used two different bacterial which strains, E. coli (DH5α) and Yersinia ruckeri, a natural fish pathogen (for a review, see Tobback et al., 2007), for this purpose. Both bacterial strains carried plasmids that constitutively expressed the red fluorescent protein dsRED so that their presence in the infected fish could be monitored easily (Fig. 4C,D). We initially infected the animals by incubation with bacteria. The bacteria were taken up and were detectable in the intestine, but did not spread from there and did not kill the animals. We therefore turned to injecting bacteria into the posterior blood island at 28 hpf. Embryos survived the injection of doses of up to ~3000 colony-forming units (c.f.u.) of E. coli (DH5α), and were able to clear the infection within 2–3 days (Fig. 4A). Thus, we used high doses of E. coli (2000–3000 c.f.u.) for subsequent experiments. By contrast, the injection of as few as ~15–20 c.f.u. of Y. ruckeri killed the embryos within 2 days (Fig. 4B).

The expression of target genes was measured by RT-PCR at 24 hours after infection. To monitor the inflammatory response, we assayed the expression of tnfa, which has been shown previously to be upregulated strongly in zebrafish infected with Edwardsiella tarda (Pressley et al., 2005).

Both E. coli and Y. ruckeri induced a strong upregulation of tnfa expression showing that the infection was recognised and that innate immune signalling cascades were activated (Fig. 4E). ifn-φ1 was also upregulated, but much more efficiently by Y. ruckeri than by E. coli, whereas neither pathogen induced upregulation of ifn-γ1 or ifn-γ2 (Fig. 4E), even when tested early after infection [3 and 6 hours post-infection (hpi), not shown].

We also tested whether Y. ruckeri infection leads to the activation of the irg genes or other ifn-γ target genes. Consistent with the finding that ifn-φ1 is upregulated upon infection, expression of the ifn-φ1-inducible irg genes irge3 and irge4 was also upregulated (Fig. 4F). Unexpectedly, irge5 transcription was also activated upon Y. ruckeri infection, although this gene did not respond to any of the interferons in the overexpression studies, indicating that it must be activated by an alternative pathway or by an as-yet-unidentified member of the interferon family.

Despite the fact that ifn-γ1 and ifn-γ2 mRNA levels remained constant after infection, their target genes irgf1 and lmp2 were activated upon Y. ruckeri infection (Fig. 4F). This cannot be an indirect result of IFN-φ1 being upregulated by Y. ruckeri, since irgf1 and lmp2 do not respond to IFN-φ1. Thus, bacterial infection either induces these genes through an interferon-independent pathway, or the interferon produced from the existing low levels of mRNA, which is not sufficient to induce target genes on its own, may be sufficient to do so in the presence of signals from a separate pathway triggered by bacterial infection [e.g. Toll-like receptor (TLR) signalling]. Alternatively, the pathway triggered by infection may lead to the post-transcriptional upregulation of interferon production from the low level of mRNA.

The fact that interferon signalling is functional in the embryo and responsive to infection suggests that it might have a role in protecting the embryo against pathogens. To test this, we analysed whether raising or lowering the levels of interferons affected the reaction of the embryos to bacterial infection. We either overexpressed ifn-γ1 and ifn-γ2 or carried out knockdowns using antisense morpholino oligonucleotides (MO) against the 5′ untranslated region (UTR) of ifn-γ1 and ifn-γ2 and compared the effect on infections with that of untreated sibling embryos. To block IFN-φ1 signalling, we used the morpholino directed against the IFN-φ1 receptor chain CRFB1.

To test the efficacy of the morpholinos, GFP constructs were generated that contained the morpholino target sequences upstream of eGFP. In vitro transcribed fusion RNAs (crfb1–eGFP, ifn-γ1–eGFP and ifn-γ2–eGFP) were injected into 1-cell stage embryos with or without the corresponding morpholinos. Injection of the fusion eGFP-RNAs led to strong fluorescence in the injected embryos, which was effectively shut down by the corresponding morpholinos (Fig. 5A). Thus, all three morpholinos blocked translation efficiently.

We assayed the survival rates of embryos that were injected with crfb1, ifn-γ1 and ifn-γ2 morpholinos and infected with ~3000 c.f.u. of E. coli. Embryos in which any one of the signalling pathways had been blocked were able to clear the bacteria and showed survival rates that were comparable to untreated embryos (Fig. 5B), indicating that knockdown of a single interferon signalling pathway does not abolish resistance to E. coli. It also shows that the morpholinos that we used had no unspecific negative effects on resistance to bacteria. However, only 24% of embryos injected simultaneously with ifn-γ1 and ifn-γ2 morpholinos (ifn-γ1+ifn-γ2 MO-injected embryos) survived the E. coli infection (Fig. 5B). This was also reflected in the extent of bacterial clearance in the surviving embryos at 24 hpi. Whereas the infection was moderate in untreated embryos at 24 hpi (Fig. 5C), the ifn-γ1+ifn-γ2 MO-injected embryos showed a strong infection that was more widespread (Fig. 5C). The ifn-γ1+ifn-γ2 MO-injected embryos developed completely normally and did not show any signs of retardation before infection (supplementary material Fig. S1). This is also reflected by the high survival rate of their non-infected siblings (Fig. 5B). In order to further strengthen the conclusion that the failure in the defence of the embryos against bacteria is the result of interferon downregulation, it would be desirable to rescue the defects in the morpholino-treated embryos by co-injection of a morpholino-insensitive interferon mRNA. This is impossible because the expression of interferon from injected mRNA also increased sensitivity to infection (described in more detail below). However, the observed redundancy, that is, the fact that resistance was lost only when both morpholinos were used together, makes it unlikely that the reduced resistance was the result of unspecific effects of one or both morpholinos.

Knockdown of both ifn-γ1 and ifn-γ2 also increased the mortality caused by infection with Y. ruckeri (15 c.f.u.). Only 44% of these embryos survived the first 24 hours after infection compared with 80% of untreated embryos (Fig. 5D). In contrast to E. coli infections, Y. ruckeri infection also led to slightly increased death rates in embryos in which ifn-γ1, ifn-γ2 or crfb1 had been knocked down individually (~65% survival at 24 hpi) (Fig. 5D). At 48 hpi, the effect of the ifn-γ single knockdowns was similar to that seen in embryos with the double knockdown.

These experiments show that IFN-γ1 and IFN-γ2 can protect the early zebrafish embryo against bacterial infections. The finding that either of the gamma interferons is sufficient to protect the embryo against E. coli suggests that they have largely or completely overlapping functions, which is consistent with their effects on target gene activation. The fact that the single knockdowns nevertheless had an effect on the defence against Y. ruckeri might be because of a quantitative effect – perhaps the amount of protein produced from only one of the interferon genes is simply not sufficient to protect against highly virulent pathogens. To test whether higher levels of IFN-γ can protect the embryo against the lethal effect of Y. ruckeri infection, we overexpressed ifn-γ1 and ifn-γ2 and assayed the survival of the embryos. However, rather than protection, we observed a higher death rate in the ifn-γ-overexpressing embryos. Whereas the infected control embryos showed survival rates of ~95% at 1 dpi, only 15% of the ifn-γ RNA-injected embryos survived to this time point when infected with 15 c.f.u. of Y. ruckeri (Fig. 6A). A comparable effect on survival after infection has been observed for tumour necrosis factor a (Tnfa). Roca et al. found that zebrafish show a higher susceptibility to viral (spring viremia of carp virus) and bacterial (Streptococcus iniae) infection after Tnfa pre-treatment (Roca et al., 2008). Hence, we were interested to see whether the ifn-γ overexpression influences tnfa expression in the early embryo. When we compared tnfa expression in control embryos with ifn-γ-overexpressing embryos, we found an upregulation of tnfa upon ifn-γ overexpression (Fig. 6B). Furthermore, the level of tnfa in ifn-γ-overexpressing embryos was enhanced further when the embryos were infected with bacteria (Fig. 6B). A similar effect was seen for il1b expression, although the induction of il1b by ifn-γ is weaker than the expression of tnfa (Fig. 6B). Thus, the strong activation of ifn-γ might lead to a burst of expression of downstream cytokines, which appear to be deleterious for the infected embryo.

We have found that the IFN-γ signalling system can be activated in the early zebrafish embryo and is able to upregulate several target genes. Many of the homologs of known mammalian gamma interferon-inducible genes that we tested showed high levels of constitutive expression and did not respond to IFN-γ (nramp, adar, mx, ifi30, cathepsinD, p27, cr3, gp91, gbp1-4). At least at this early stage, these genes must therefore be subject to a different mode of regulation. Their constitutive expression is surprising because, in mammals, most of them show no detectable expression in the absence of interferon, and the expression of immune genes often comes with a cost to the organism.

It is worth speculating about the biological relevance of the constitutive expression of innate immune defence genes in the early zebrafish embryo. A role in immune defence should not be needed during the first two days because the embryo is enclosed in a chorion that protects it from pathogens. We consider it unlikely that these genes have an alternative function in the early embryo. One possibility, a developmental function, is unlikely since developmental mechanisms in vertebrates are highly conserved and no such roles have been described for these genes in other species. However, as soon as the embryo hatches, it is immediately exposed to the pathogens of the environment, so perhaps the pre-existence of certain components of the defence system is of advantage at this point.

In contrast to the immune GTPases of the guanylate binding protein (GBP) and myxovirus resistance (MX) type, which are expressed constitutively in the zebrafish embryo, some members of the p47 IRG family show different behaviours. The three irgq genes (irgq1-3) are also expressed constitutively, whereas the members of the irge and irgf group are not (or only minimally) expressed in untreated fish embryos. Based on the absence of a functional nucleotide-binding motif, it has previously been suggested that the function of the Irgq group members has diverged from that of the other IRGs (hence the name q or quasi GTPases) (Bekpen et al., 2005). The difference in their regulation in the zebrafish embryo supports this notion.

irge3, irge4, irgf1, irgf3 and irgf4 were induced strongly by interferon; irge2 was activated weakly. Since no cDNAs had been reported previously for irgf4 and irge2, this is the first experimental support for the prediction of these two genes. irgg1 is activated very weakly by interferon; this gene lies in tandem with the constitutively expressed irgq1, which showed some activation above its basal level upon interferon stimulation. These two genes have in fact been annotated as a fusion gene where the sequences encoding the irgg1 exon are spliced to the downstream irgq1 exon. The different expression and regulation of the sequences amplified by our two exon-specific primer sets indicates that independent transcripts must be made. Furthermore, RT-PCR using primer pairs that should amplify transcripts containing both irgg1 and irgq1 did not yield any such products, suggesting that there are two independent genes. Nevertheless, it is possible that the response we see for irgq1 after interferon treatment may be because of the promoter activity of the upstream irgg1 gene.

The remaining IRG family members, irge1, irge5, irge6 and irgf2 are not expressed in either the absence or presence of interferon. For irgf2, this finding is consistent with the fact that no expressed sequence tags (ESTs) have been found. An EST does exist for irge1 and we also detect expression in adult tissues (unpublished data) showing that it can be expressed, but apparently it does not respond to interferon.

The genes of the irge and irgf groups are the strongest interferon responders that we have found in the zebrafish embryo. This was possibly unexpected for the members of the irgf group. The zebrafish irgf genes resemble the ‘irgc-like’ genes in mammals in that they have an intron in the same place in the middle of the main exon that encodes the G-domain (Bekpen et al., 2005) (Christoph Rohde and Jonathan C. Howard, unpublished results), and because the mammalian genes in this class are not inducible by interferon.

The zebrafish genome and many other teleost genomes contain two ifn-γ genes that show only a low degree of identity to each other (18.8% amino acid identity between zebrafish ifn-γ1 and ifn-γ2) (Zou et al., 2005; Igawa et al., 2006; Milev-Milovanovic et al., 2006; Stolte et al., 2008). Our results indicate that ifn-γ1 and ifn-γ2 act redundantly despite their low degree of sequence similarity. We did not find any target genes that were inducible by one interferon and not the other, and knockdown of either interferon alone did not interfere with the ability of the embryos to clear E. coli that were injected into the posterior blood island.

One finding that might argue against a complete functional equivalence of ifn-γ1 and ifn-γ2 is the increased mortality of Y. ruckeri-infected embryos when expression of only one of the two interferons is blocked. However, this may simply be the result of halving the total interferon levels. Clearly, the embryos are more sensitive to Y. ruckeri (it is lethal even at low doses) than to E. coli. Thus, the reduction in overall interferon levels might lead to a more severe loss in resistance against Y. ruckeri than against E. coli. The functional equivalence of the two interferons in the response to Y. ruckeri is also consistent with the finding that knockdown of each interferon results in a similar reduction in survival rates, showing that, in each case, the remaining interferon is equally effective at inducing a low-level defence.

One detail in which the two zebrafish interferon sequences differ from each other is that IFN-γ2 contains a nuclear localisation signal (NLS) in the same position as in mouse and human IFN-γ, whereas IFN-γ1 does not. Nuclear localisation of mammalian IFN-γ by its NLS has been proposed to be necessary for full biological activity (Subramaniam et al., 2000), although a mechanism by which IFN crosses the plasma membrane to reach the cytosol has not been described. Our finding that IFN-γ1, without an NLS, is able to trigger a signalling cascade as efficiently as the NLS-containing IFN-γ2 indicates that an NLS is not a general requirement for IFN-γ signalling.

Overexpression of ifn-γ1 and ifn-γ2 in the embryo leads to a dramatically increased susceptibility to bacterial infection [a recent report suggests that this reaction is less pronounced in adult fish (Lopez-Munoz et al., 2009)], in a similar way to that shown to occur in mice (Kohler et al., 1993). The increased sensitivity to E. coli in mice treated with IFN-γ has been ascribed to a septic response. Sepsis is a dysregulation of inflammation in which an initial systemic inflammatory response, which includes the upregulation of inflammatory cytokines such as TNF-α and IL-1, is not balanced properly by a compensatory anti-inflammatory response (reviewed by Buras et al., 2005). It is possible that the effects we observe also resemble sepsis. We find that ifn-γ does indeed induce the expression of the tnfa and il1b genes, which are upregulated further upon infection, indicating an overactivation of the systemic inflammatory response. This overactivation may also be the cause of the high mortality effects that have been described in adult zebrafish treated with Tnfa, which show enhanced susceptibility to viral and bacterial infection (Roca et al., 2008). However, in contrast to the situation in mammals, injection of high numbers of E. coli alone does not induce a septic shock, and fish show a high tolerance to LPS (Novoa et al., 2009). This may be related to the fact that fish recognise bacterial surface components, such as LPS, in a different manner from mammals (Sepulcre et al., 2009). The extent to which a septic response in fish is related to sepsis in mammals will need to be investigated in the future.

Morpholino oligonucleotides (MO) were synthesised by Gene Tools and injected into 1–2-cell stage zebrafish embryos. The injection solution consisted of 0.6 mM of the respective MO, 0.2% Phenol Red and 0.1 M KCl of which, on average, 2 nl were injected per embryo. Injections were carried out using an Eppendorf FemtoJet and micromanipulator.

The MO sequences used were: ifn-γ1, 5′-TTTCTGTGCTGTGAACCAAGTGATG-3′; ifn-γ2, 5′-TGAAGGCGTTCGCTAAAGTTAGAGT-3′; crfb1, 5′-GAGTCACACTTTAGCAATGATGAAG-3′.

To test the binding of the different MOs, the 5′ sequences of the respective genes were amplified with primers that contained restriction sites for EcoRI and NcoI or BamHI and EcoRI (ifn-γ1-5′utrecofor, 5′-AGGAATTCCCTTTGCCTATATAAACCCC-3′; ifn-γ1-5′utrncorev, 5′-ATCCATGGTTGTGTTTCTGTGCTGTGAA-3′; ifn-γ2-5′utrecofor, 5′-AGGAATTCTCACTCGAACACATGCATTC-3′; ifn-γ2-5′utrncorev, 5′-ATCCATGGAGTTGAAGGCGTTCGCTAAA-3′; crfb1-5′utrbamfor, 5′-GACTGGATCCGAACCGGGTGTAGTGACGTT-3′; crfb1-5′utrecorev, 5′-CTAGGAATTCCACAAACAAAAGCAGAGTCACA-3′). Digested and purified PCR products were cloned into the pCS2+/eGFP vector (kindly provided by Alexander Reugels). Capped RNA was synthesised in vitro using the Message Machine kit (Ambion), according to the supplier’s manual. RNA was injected at 200 ng/μl, with and without the respective MO (0.6 mM), into 1–2-cell stage embryos. On average, the injected volume was 2 nl. GFP fluorescence was visualised at 28–32 hpf using a Leica MZ 16FA stereomicroscope.

For overexpression of ifn-γ1, ifn-γ2 and ifn-φ1, the coding regions were amplified with primers that contained restriction sites for XhoI and XbaI (ifn-γ1XhoI-for, 5′-ATCTCGAGATGGATTCCTGCCTCAAAAT-3′; ifn-γ1XbaI-rev, 5′-ATTCTAGACTAAGACATGTGTAAATGCCG-3′; ifn-γ2XhoI-for, 5′-GCCTCGAGATGATTGCGCAACACATGAT-3′; ifn-γ2XbaI-rev, 5′-CGTCTAGATCAACCTCTATTTAGACT-3′; ifn-φ1XhoI-for, 5′-CGCTCGAGATGTGGACCTATATTTTTGTG-3′; ifn-φ1XbaI-rev, 5′-ATTCTAGATCAAGGATTGACCCTTGCGT-3′). The PCR product was gel purified, digested and cloned into the pCS2+ vector (Turner and Weintraub, 1994). For overexpression, we either injected plasmids at a concentration of 100 ng/μl or RNA (200 ng/μl), as described above, into 1–2-cell stage embryos.

RNA from zebrafish embryos was isolated from regularly kept embryos or from germfree embryos. To remove germs, the embryos were bleached immediately after fertilization. Bleaching was performed by two cycles of 4-minute incubations in 0.012% sodium hypochlorite followed by two 3-minute washes in sterile 1/3 zebrafish Ringer solution. To test the efficacy of bleaching, we homogenised embryos and plated them on bacteriological plates (LB medium and Middlebrook 7H10). The Middlebrook 7H10 and LB medium plates were incubated at 28°C and 37°C for 7 days. No colonies were detected for bleached embryos, whereas plates with untreated embryos contained ~1000 colonies at 28°C and 37°C. RNA from zebrafish embryos was isolated using the μMacs mRNA isolation kit (Miltenyi), according to protocol, by addition of a DNAseI digest for 7 minutes to remove traces of genomic DNA. First-strand synthesis was performed by utilizing the Superscript III system (Invitrogen) using 300 ng of mRNA and random hexamer primers. For RT-PCR, 7.5 ng first-strand synthesis (7.5 ng) and RedTaq DNA polymerase (Sigma-Aldrich) were used. RT-PCR was performed as follows: 2 minutes at 94°C; 28–35 cycles of 94°C for 15 seconds, 55°C for 30 seconds, 72°C for 1 minute; 7 minutes at 72°C. See supplementary material Table S1 for the primer sequences of the tested genes. All of the primer pairs that were used for RT-PCR were able to amplify the predicted products when tested on genomic DNA. The PCR products were controlled by agarose gel electrophoresis.

E. coli (DH5α) and Y. ruckeri (strain 4015726Q) carrying the dsRED-expressing pGEMDs3 plasmid (van der Sar et al., 2003) were grown in standard LB medium containing ampicillin (50 μg/ml). For infections, a 3 ml overnight culture was centrifuged for 3 minutes at 4000 g, the supernatant was removed, and the pellet washed with PBS. Washing was repeated three times to remove traces of LB and metabolites. After washing, the pellet was resuspended in 500 μl PBS. The final injection solution contained the washed E. coli or Y. ruckeri diluted in PBS containing 1/10 Phenol Red.

Zebrafish were bred at 26.5°C on a 14-hour light/10-hour dark cycle. Embryos were collected by natural spawning and staged according to Kimmel et al. (Kimmel et al., 1995).

At 28–30 hpf, zebrafish embryos were anesthetised using tricaine (Sigma-Aldrich), according to standard protocols (Westerfield, 1995), and arranged on agarose plates for infection. Infection was achieved by injecting the bacteria into the embryonic blood island. To determine the amount of injected bacteria, a control injection was performed in 100 μl of sterile PBS and plated on LB-Amp plates to count the colonies. After infection, embryos were washed in 1/3 zebrafish Ringer solution (Westerfield, 1995) and incubated further at 28.5°C in 1/3 zebrafish Ringer solution to score survival or for RNA isolation. The infection in untreated and ifn-γ knockdown embryos was monitored by visualizing dsRed fluorescence using a Leica MZ 16FA stereomicroscope.

We are grateful to Juliane Hancke for technical assistance. We thank Julia Hunn, Georges Lutfalla and Jonathan Howard for critical reading of the manuscript. We are especially grateful to Röbbe Wünschiers for the statistical evaluation of the survival data. This article is freely accessible online from the date of publication.