A Family of Genes Coding for Two Serologically Distinct Chicken Interferons*

Southern blot analysis and screening of a genomic (cid:108) phage library with the previously cloned chicken inter- feron (IFN) cDNA indicated that the chicken genome contains at least 10 IFN genes. A particularly strongly hybridizing phage clone that we analyzed in more detail carried a head to tail arrangement of three intron-less IFN genes that differed from each other and from the cloned chicken IFN cDNA by only a few base changes. The primary translation products of these three IFN genes consist of 193 amino acids, and the mature pro- teins are composed of 162 amino acids. All three genes of this IFN family, designated IFN1 , yielded active chicken IFN when expressed individually in transfected COS7 cells. A weakly hybridizing phage clone contained an additional intron-less chicken IFN gene, designated IFN2 , whose product was 57% identical to chicken IFN1. Southern blot analysis suggested that the chicken genome contains a single IFN2 gene. The primary transla- tion product of IFN2 consists of 203 amino acids, and the mature protein is composed of 176 amino acids. Purified recombinant chicken IFN2 from Escherichia coli had a specific antiviral activity of about 10 6 units/mg, which was about 20-fold lower than that of chicken IFN1 puri- fied in parallel. The antiviral activity of chicken IFN2 from E. coli or from transfected COS7 cells could not be neutralized by antiserum to recombinant chicken IFN1. regions of IFN1 or IFN2 were used as hybridization probes. Hybridizations were carried out at 42 °C in 0.12 M sodium phosphate, pH 7.3, 0.25 M NaCl, 7% SDS, 1 m M EDTA, 20% formamide, and 200 (cid:109) g/ml of denatured herring sperm DNA. Linearized plasmids carrying the coding regions of IFN1 or IFN2 were subjected to 3-fold dilutions in buffer containing carrier DNA, and

Southern blot analysis and screening of a genomic phage library with the previously cloned chicken interferon (IFN) cDNA indicated that the chicken genome contains at least 10 IFN genes. A particularly strongly hybridizing phage clone that we analyzed in more detail carried a head to tail arrangement of three intron-less IFN genes that differed from each other and from the cloned chicken IFN cDNA by only a few base changes. The primary translation products of these three IFN genes consist of 193 amino acids, and the mature proteins are composed of 162 amino acids. All three genes of this IFN family, designated IFN1, yielded active chicken IFN when expressed individually in transfected COS7 cells. A weakly hybridizing phage clone contained an additional intron-less chicken IFN gene, designated IFN2, whose product was 57% identical to chicken IFN1. Southern blot analysis suggested that the chicken genome contains a single IFN2 gene. The primary translation product of IFN2 consists of 203 amino acids, and the mature protein is composed of 176 amino acids. Purified recombinant chicken IFN2 from Escherichia coli had a specific antiviral activity of about 10 6 units/mg, which was about 20-fold lower than that of chicken IFN1 purified in parallel. The antiviral activity of chicken IFN2 from E. coli or from transfected COS7 cells could not be neutralized by antiserum to recombinant chicken IFN1. Thus, like mammals, the chicken has a large number of type I IFN genes that code for at least two serologically distinct antiviral activities.
In mammals, cytokines with antiviral activity are classified as type I and type II interferons (IFNs). 1 Type I IFN includes ␣-, ␤-, -, and -IFNs, which have related structures and use a common receptor. The former three are synthesized in response to virus infection (1,2), whereas -IFN is synthesized in response to developmental stimuli in the trophoblast of ruminants (3) and humans (4). Matureand -IFNs of humans consist of 172 amino acids, whereas ␣and ␤-IFNs are composed of 165 or 166 amino acids (2). Antisera prepared against ␣-IFN or ␤-IFN showed a high degree of subtype specificity and did not neutralize the antiviral activity of -IFN (5). All mammalian type I IFNs are coded for by intron-less genes (2). The ␣and -IFNs are encoded by gene families with as many as 20 closely related members (2). In the various species, the ␤-IFNs are coded for by single genes (e.g. mouse) or by gene families (e.g. cattle). Type II IFN or IFN-␥ is synthesized by antigen-or mitogen-stimulated T cells (1). It is encoded by a single gene with introns, has pleiotropic regulatory effects on cells of the immune system (6,7), and is the principal macrophage-activating factor of mammals (8,9).
In contrast to mammals, the IFNs of birds are poorly characterized. The first cDNA for a chicken IFN was isolated only recently (10). Its sequence similarity to mammalian IFNs is marginal, but conservation of cysteine residues and inducibility by virus indicate that it represents a type I IFN. This notion was supported by the finding that recombinant chicken IFN is a potent antiviral agent that lacks other biological activities associated with IFN-␥ of mammals (11). Antibodies to the cloned chicken IFN neutralized the bulk of antiviral activity in preparations of partially purified chicken IFN from various natural sources (11), suggesting that a single serotype of IFN is predominately induced under experimental conditions. These results were compatible with the assumption that the chicken has a single gene for type I IFN (10). However, Southern blot analysis now suggested the presence of several IFN genes in chicken. A more detailed analysis showed that the chicken genome contains a family of at least 10 IFN genes, now designated IFN1, which all appear to code for one serotype of chicken IFN. A second serotype of ChIFN is encoded by a single gene, designated IFN2, that shows limited sequence conservation.

MATERIALS AND METHODS
Phage Library Screening-Approximately 1.2 ϫ 10 5 plaques (10 4 / 150-mm dish) of a chicken genomic library constructed in FIX II (Stratagene) were screened with the previously cloned ChIFN cDNA (10) that was radiolabeled by nick translation. The membranes were hybridized at 42°C in a buffer containing 10 mM PIPES, pH 6.8, 300 mM NaCl, 10 mM EDTA, 0.5% SDS, 1 ϫ Denhardt's solution, 100 g/ml denatured herring sperm DNA, and 20% formamide for 15 h before they were washed in 2 ϫ SSC containing 0.5% SDS for 1 h at 42°C and then for 30 min at 56°C. Selected positive phages were plaque-purified three times.
Genomic Southern Blot Analysis-Samples (20 g) of chicken liver DNA (a kind gift of Dr. B. Kaspers) were digested with 80 units of restriction enzyme, fragments were size-fractionated by electrophoresis through a 1% agarose gel, soaked in 0.2 M HCl for 10 min, and transferred to nylon membranes in 0.4 M NaOH. Radiolabeled DNA fragments comprising the complete coding regions of IFN1 or IFN2 were used as hybridization probes. Hybridizations were carried out at 42°C in 0.12 M sodium phosphate, pH 7.3, 0.25 M NaCl, 7% SDS, 1 mM EDTA, 20% formamide, and 200 g/ml of denatured herring sperm DNA. Linearized plasmids carrying the coding regions of IFN1 or IFN2 were subjected to 3-fold dilutions in buffer containing carrier DNA, and samples corresponding to approximately 45 pg, 150 pg, 450 pg, and 1.35 ng of plasmid DNA were loaded into individual gel slots. Assuming a complexity of the haploid chicken genome of 1.2 ϫ 10 9 (12), these amounts of plasmid DNA should yield hybridization signals that equal approximately 0.6, 2, 6, and 18 gene equivalents in 20 g of total genomic chicken DNA. The blots were washed in 2 ϫ SSC ϩ 0.5% SDS at 68°C for 1 h and then in 0.2 ϫ SSC ϩ 0.5% SDS at 68°C for 15 min.
Analysis and Subcloning of Phage DNA-DNA from phage clone 8/18, later shown to carry three IFN1 genes (Fig. 2), was digested with XbaI, the resulting 8.4-and 6.1-kb fragments were subcloned individually into pSP72, and their restriction maps were established. To create plasmid subclones carrying single IFN1 genes, the 8.4-kb fragment was further digested with KpnI, and the products were subcloned into vector pSP72. The various IFN1 expression constructs were created by cloning the IFN1 coding sequences, which are located immediately downstream of the NcoI restriction sites indicated in Fig. 2, into the vector pcDNAI. The NcoI sites were filled in with Klenow polymerase, and appropriate fragments were gel purified and ligated into suitable polylinker sites of the vector.
DNA of the phage clone 1/4 was digested with combinations of various restriction enzymes and analyzed for IFN-related sequences by Southern blotting. A hybridizing 2.3-kb EcoRI-BglII fragment, later shown to contain the complete coding region of the IFN2 gene, was subcloned into pSP65. PCR was used to generate the IFN2 expression constructs. Primer 1 (5Ј-CGACGGAATTCCCAGCAGAACACAAGT-CCC-3Ј) corresponded to nucleotide positions 103-121 of the cloned DNA fragment and introduced an additional EcoRI restriction site. Primer 2 (5Ј-GTGCACTCGAGACAGTCACTGGGTGTTGAG-3Ј) was reverse complementary to nucleotide positions 749 -728 and provided an new XhoI restriction site. The PCR product of these primers was digested with EcoRI and XhoI and cloned in the corresponding sites of the eukaryotic expression vector pcDNAI. Primer 3 (5Ј-GCGTACATATGT-GCAACCATCTTCGTCACCAGG-3Ј) corresponded to nucleotide positions 212-233 of the cloned DNA fragment and introduced an additional NdeI restriction site. Primer 4 (5Ј-TCACGTAGGATCCAGTCACT-GGGTGTTGAG-3Ј) corresponded to nucleotide positions 745-728 of the cloned DNA fragment and introduced an additional BamHI restriction site. The PCR product of this primer combination was digested with NdeI and BamHI and cloned into the corresponding sites of the bacterial expression vector pET-3a.
DNA Sequence Analysis-Nucleotide sequences were determined by the dideoxy chain determination method using the T7 polymerase kit from Pharmacia Biotech Inc.
Production of Recombinant ChIFNs in COS Cells and Escherichia coli-Transfections of COS7 cells with the various pcDNAI constructs were performed as described (13). At 72 h post-transfection, the culture supernatants were harvested.
Expression of the pET constructs in E. coli was performed as described (13). Purification of recombinant ChIFN1 and ChIFN2 was done by a method that includes solubilization of recombinant proteins in guanidine hydrochloride and binding to nickel-chelate agarose, a protocol originally established for the purification of recombinant duck IFN (14).
IFN Titrations-The antiviral activity of the various ChIFNs was measured by a cytopathic effect inhibition assay as described (13).
Virus Yield Reduction Assays-These assays were performed with chicken CEC-32 cells infected with vesicular stomatitis virus (VSV) as described (13). Rabbit antiserum to purified recombinant ChIFN1 from E. coli was used at a dilution of 1:200 (13). Preimmune serum (1:200) from the same animal served as negative control.

RESULTS
A Family of Genes Coding for Chicken IFN-To estimate the size of the IFN gene family, we performed genomic Southern blot analyses of liver DNA from blood group antigen-syngenic White Leghorn chicken. When the blots were hybridized to a radiolabeled probe derived from the previously cloned chicken IFN (ChIFN) cDNA (10), a simple pattern of strong and a complex pattern of weak hybridization signals was observed (Fig. 1A). By comparing the signal intensities to those of defined amounts of linearized plasmid DNA containing ChIFN cDNA (10), we calculated that 10 or more IFN genes were present in the chicken genome. Support in favor of this view came from the screening of a phage library of genomic chicken DNA: the ChIFN cDNA probe identified a total of 165 positive phages among the approximately 120,000 phage plaques analyzed, which was roughly 100 times as many as expected for a single copy gene. On Southern blots, the ChIFN cDNA probe detected two prominent signals in BamHI-digested DNA, but only one prominent signal in PstI-or HindIII-digested DNA. Since the cloned chicken IFN cDNA has a single BamHI restriction site, but no PstI or HindIII sites, this result suggested that the prominent signals originated from multiple intron-less genes with similar structures. Cloning and sequencing experiments supported this interpretation (see below).
Three Intron-less IFN1 Genes on a Single 14.5-kb Fragment of Chicken DNA-To characterize the IFN gene family in more detail, we performed PCR analyses on several positive phages with primers that should amplify the coding region of chicken IFN (13). Reactions with phage DNAs yielded products that could not be distinguished from those of PCRs performed with IFN cDNA, again suggesting that the chicken IFN genes were intron-less. Partial sequencing of some PCR products showed that members of this gene family, which we now designate IFN1, had almost identical sequences.
To learn more about the IFN1 gene family, we decided to perform a more detailed analysis of one particular phage that yielded a very strong hybridization signal, suggesting that it contained more than one IFN1 gene. Restriction analysis and partial sequencing confirmed that this phage contained a 14.5-kb fragment of chicken DNA that harbored three intronless genes, designated IFN1-1, IFN1-2, and IFN1-3 (Fig. 2). Their sequences were almost identical to that of the previously cloned chicken IFN cDNA (10), except for an A to G transition at position 248 in IFN1-1 and IFN1-2, and a C to T transition at positions 202 and 227 in IFN1-3. All three IFN1 genes coded for putative precursor proteins of 193 amino acids that may be processed to mature proteins of 162 amino acid residues. ChIFN1-1 and ChIFN1-2 have identical amino acid sequences but differ from the originally cloned ChIFN by an Asp to Ser change at position 34. ChIFN1-3 differs from the originally cloned ChIFN by a Leu to Phe change at position 19 and a Pro to Leu change at position 27. To determine whether these natural variants of ChIFN1 are biologically active, we cloned appropriate fragments of the phage DNA carrying the individual IFN1 genes into an eukaryotic expression vector, transfected the resulting constructs into COS7 cells, and tested the supernatants for chicken IFN activity. At 72 h post-transfection, the supernatants of transfected cells contained between 20,000 and 50,000 units/ml of antiviral activity, indicating that all IFN1 constructs coded for active IFN. Rabbit antiserum to purified recombinant ChIFN (11) efficiently neutralized the antiviral activities of the various COS7 cell supernatants (data not shown), suggesting that the various IFN1 genes coded for a single serotype of chicken IFN.
Identification of a Novel Chicken IFN Gene-One phage showed particularly weak hybridization signals when probed with the ChIFN probe. No product was obtained when PCR was performed with DNA of this phage and IFN1-specific oligonucleotide primers, suggesting that it may contain a novel IFN gene. Sequence analysis of an EcoRI-BglII fragment of this phage revealed an intron-less gene, designated IFN2, whose coding sequence was 73% identical to the IFN1 genes (Fig. 3). To estimate the number of IFN2 genes in the chicken genome, we performed Southern blot analyses with a radiolabeled probe that comprised the complete coding region of the IFN2 gene. This probe cross-reacted weakly with the various DNA fragments that contained IFN1 genes. More importantly, it recognized additional fragments of approximately 13, 3, and 3.5 kb in chicken DNA digested with BamHI, HindIII, or PstI, respectively (Fig. 1B). By comparing the intensities of the IFN2-specific signals of the BamHI and HindIII digests with appropriate plasmid standards (Fig. 1B), we concluded that the chicken genome most likely contains a single IFN2 gene. We noted that the signal in the PstI digest was stronger than those in the neighboring lanes. Its increased intensity was probably artifactual due to a fortuitous superpositioning of the IFN2 signal with an IFN1 signal.
The IFN2 gene codes for a polypeptide of 203 amino acids (Fig. 3), whose N terminus lacks charged amino acids, suggesting that it may function as a signal peptide. An alignment of the ChIFN2 sequence with the prototype sequence of ChIFN1 (10) and the product of a recently cloned duck IFN gene (14) is shown in Fig. 4. This comparison indicated that the cysteine residue at position 28 is the N-terminal amino acid of mature ChIFN2. Secreted ChIFN2 thus seems to be 14 residues longer than ChIFN1: it is composed of 176 amino acids and has a calculated molecular mass of 20,372 Da. Sequence conservation between ChIFN1, ChIFN2, and duck IFN is pronounced in most regions, except for the signal peptides and the C termini. When the first 150 amino acids of the mature proteins are considered, ChIFN2 is 57% identical to ChIFN1 and 61% identical to duck IFN. The C-terminal 26 amino acids of ChIFN2 show no obvious similarity to either the C-terminal 12 amino acids of ChIFN1 or the 11 C-terminal amino acids of duck IFN (Fig. 4). Remarkably, the four cysteine residues (marked by asterisks in Fig. 4), which are highly conserved in all ␣-, -, and -IFNs of mammals, are conserved in both chicken IFNs as well as in duck IFN.
The Activity of ChIFN2 Cannot Be Neutralized by Antiserum to ChIFN1-To determine whether IFN2 codes for an active chicken IFN, we amplified its complete coding region by PCR and cloned the resulting fragment into an eukaryotic expression vector. COS7 cells transfected with this construct secreted an activity into the culture supernatant that protected chicken CEC-32 cells from destruction by VSV (data not shown). Interestingly, such supernatants did not contain more than 2,000 units of antiviral activity/ml, whereas parallel cultures of COS7 cells transfected with an expression construct for ChIFN1 contained 20,000 -50,000 units of antiviral activity/ml, suggesting that ChIFN2 had a lower specific activity. To further investigate this issue, we produced recombinant ChIFN2 in E. coli. Using PCR we constructed an expression plasmid that coded for ChIFN2 devoid of signal peptide. By following a single step purification protocol that was successfully used for the production of recombinant duck IFN (14), we obtained recombinant ChIFN2 that was more than 80% pure. Purified units/mg, which was about 20 times lower than that of ChIFN1 which was purified in parallel.
To determine whether ChIFN1 and ChIFN2 represent two distinct serotypes of chicken IFN, we evaluated the cross-neutralizing potential of a rabbit antiserum that we had prepared against E. coli-produced ChIFN1 (13). CEC-32 cells were incubated with 100 units/ml of the various IFNs and 0.5% of either rabbit antiserum or preimmune serum, before the cultures were challenged with VSV. In the presence of preimmune serum, ChIFN1 as well as ChIFN2 were very effective and reduced the VSV yields by almost 10 4 -fold (Fig. 5). In the presence of antiserum, the activity of ChIFN1 was neutralized quite effectively. By contrast, the antiserum did not significantly reduce the antiviral activity of ChIFN2 from E. coli or from transfected COS7 cells (Fig. 5), suggesting that it represents a novel serotype of chicken IFN. DISCUSSION We have shown here that the chicken contains at least two serologically distinct subtypes of IFN. The first subtype is coded for by a gene family of more than 10 members, while the second subtype seems to be encoded by a single gene. All these genes lack introns, like the mammalian genes for the various type I IFNs. Thus, although the primary structures of avian IFNs are poorly conserved, the presence of a gene family rather than a single gene and the lack of introns are highly conserved features of type I IFNs.
The search for additional subfamilies of chicken IFNs is complicated by the fact that the IFN1 gene family is quite large. Southern blot analysis suggested the presence of about 10 IFN1 genes. However, the high frequency by which positive phages were identified in a chicken genomic library suggested the presence of as many as 100 IFN genes in the chicken genome. It is possible that the latter high frequency is an artifact of the phage library. Alternatively, the library screen may indeed have revealed the true complexity of the chicken IFN superfamily. We assumed that the complex pattern of weak hybridization signals on the Southern blot (Fig. 1A) resulted from single IFN1 genes with altered restriction sites in their flanking regions. However, we cannot exclude the alternative possibility that these signals resulted from weak crossreactivity to a novel IFN gene family.
One cross-reactive phage that we characterized contained a novel chicken IFN gene that we designated IFN2. Southern blotting experiments suggested that it is a single copy gene. Functional studies showed that IFN2 codes for a chicken IFN that escapes neutralization by antibodies to ChIFN1, a finding that may be explained by the fact that the amino acid sequence of ChIFN2 is only 57% identical to that of ChIFN1. Interestingly, mature ChIFN2 is 14 amino acids longer than ChIFN1. In mammals, "long" variants of ␣-IFN with extra amino acids at the C terminus are known asand -IFNs (2). Their sequences are about 60% identical to ␣-IFNs, and their activities cannot be neutralized by antisera that neutralize ␣-IFNs (5). Although this suggests that ChIFN2 represents the avian homolog of mammalianor -IFNs, we believe that the familiar terms ␣-, ␤-, -, and -IFN, which refer to subtypes of mammalian IFNs with specific biological properties, should not be used at present for the chicken IFN system. Detailed studies on the induction of IFN2 in response to virus or developmental stimuli will be required to determine whether this assumption is correct.
An unexpected result of our studies was that the specific antiviral activity of purified recombinant ChIFN2 was about 20 times lower than that of ChIFN1. Evidence that this difference was not an artifact of the purification procedure came from experiments with supernatants of COS7 cells that were transfected with expression constructs for ChIFN1 or ChIFN2. Because the two constructs were identical except for the IFN coding regions, it seems reasonable to assume that similar amounts of recombinant protein were produced in the two cultures. Nonetheless, the supernatants of ChIFN1-producing cultures contained about 20-fold more antiviral activity than supernatants of ChIFN2-producing cultures, strongly suggesting that this difference reflects a true biological difference of the two ChIFN subtypes. This result further indicated that the low specific activity of E. coli-produced ChIFN2 cannot be explained by simply assuming that ChIFN2 needs glycosylation for full activity. This situation found here for ChIFNs is reminiscent to that described for the various ␣-IFN subtypes of humans and mice, whose individual specific activities were also found to differ significantly (15,16).