A Novel and Atypical Type One Interferon Gene Expressed by Trophoblast during Early Pregnancy*

Low stringency screening of day 14-15 pig conceptus cDNA library using an interferon (1FN)-w probe has led to the isolation of a novel type one (type I) IFN gene physiologically expressed by trophoblast during the period of implantation in the uterus. It encodes a 170-amino acid preprotein with two potential N-glycosylation sites and a predicted N-terminal signal peptide of 2 1 residues. This protein was shorter and richer in Cys residues than other type I IFNs and distantly related to all of them including IFN-T, the trophoblast IFNs of ruminants, with amino acid identities ranging between 27 and 42%. Moreover, its antiviral activity was resistant to neutralization by antisera directed against several IFN-a, -8, -w, and -7 family members. The pig genome contained only two intronless, non-allelic loci homologous to the cDNA, one of them corresponding to the cDNA. This gene had a single transcription initiation site. Its upstream putative regulatory regions had no similarity with those of other type I IFN genes and did not display the typical clustering of short sequence motifs thought to be involved in virus inducibility but, instead, a modular structure of 30-40-base pair direct repeats. These data highlight the very atypical characteristics of this gene and

Low stringency screening of day 14-15 pig conceptus cDNA library using an interferon (1FN)-w probe has led to the isolation of a novel type one (type I) IFN gene physiologically expressed by trophoblast during the period of implantation in the uterus.It encodes a 170-amino acid preprotein with two potential N-glycosylation sites and a predicted N-terminal signal pep- tide of 2 1 residues.This protein was shorter and richer in Cys residues than other type I IFNs and distantly related to all of them including IFN-T, the trophoblast IFNs of ruminants, with amino acid identities ranging between 27 and 42%.Moreover, its antiviral activity was resistant to neutralization by antisera directed against several IFN-a, -8, -w, and -7 family members.
The pig genome contained only two intronless, nonallelic loci homologous to the cDNA, one of them corresponding to the cDNA.This gene had a single transcription initiation site.Its upstream putative regulatory regions had no similarity with those of other type I IFN genes and did not display the typical clustering of short sequence motifs thought to be involved in virus inducibility but, instead, a modular structure of 30- 40-base pair direct repeats.These data highlight the very atypical characteristics of this gene and its product and suggest that it represents the first member of a yet unknown family of type I IFNs present in other mammalian species according to sequence divergence analysis.
Interferons (IFNs)' are multifunctional cytokines initially defined by their ability to induce in target cells a state of resistance against subsequent viral infection.In mammals, they fall in two groups.Type I IFNs represent a heterogeneous group of molecules sharing sequence homology, encoded by a set of clustered, intronless genes that can be expressed from a broad range of cell types.Type I1 IFN or IFN--/ is an unrelated molecule encoded by a single gene with introns and an expression pattern almost exclusively limited to activated T and natural killer cells (1, 2).
* The costs of publication of this article were defrayed in part by the payment of page charges.This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C.Section 1734 solely to indicate this fact.

222706-222708.
Viral infection has been previously determined to be the main biological event triggering the in vivo and in vitro expression of type I IFNs (1).Now, these virus-inducible IFNs fall into three distinct families: a, p, and w on the basis of sequence homology and antigenic properties (3-5).However, there are many reports about the induction of type I IFNs in a virus-independent manner during physiological processes such as regulation of cell growth and differentiation (recently reviewed in Ref. 6).Nevertheless, until recently, only virus-inducible type I IFN cDNAs and genes corresponding to the families mentioned above had been cloned (5, 7), and the possibility that the "physiological" IFNs could be assigned to special type I IFN families had not received great attention.
Recently, proteins physiologically secreted by the trophoblast of ovine and bovine conceptuses during a short period around implantation and known to mediate maternal recognition of pregnancy by maintaining the function of the corpus luteum during early pregnancy (8) were identified as type I IFNs (9-12).These molecules, originally called trophoblastins or ovine and bovine trophoblast protein-1 (oTP-1 and bTPl ) , although closely related to IFN-w, appeared to represent a new family of type I IFNs.They were designated IFN-T and recognized as the first type I IFN family whose biological role was not primarily antiviral and whose temporal and tissuespecific induction was physiologically controlled (reviewed in Ref. 13).However, IFN-T seems to exist only in some ruminant species (14).Several studies anterior and posterior to IFN-T discovery have described the physiological expression of antiviral activity in the feto-placental unit of non-ruminant species (15-23), suggesting that pregnancy could be a preferential opportunity for physiological expression of IFNs possibly to achieve a common function unrelated to maintenance of pregnancy.Such a function, together with the possibility of discovering yet unknown IFN families, emphasized the need for molecular characterization of these IFNs.
Our group and others have recently shown that a high level of IFN expression by implanting conceptuses occurs in the pig, a non-ruminant species (18-21).Interestingly, characterization of this antiviral activity revealed striking differences with ruminant trophoblast IFNs; we demonstrated that pig conceptuses express two antigenically distinct antiviral proteins (20,21).Surprisingly, one of them was clearly identified as IFN--/ (type I1 IFN), which represents a major secretory product of the trophoblast.The other antiviral protein was tentatively classified as type I IFN because of its antigenic cross-reactivity with human leukocyte-derived IFN (HuIFN-Le).These serological data were strengthened by the detection of transcripts cross-hybridizing under low stringency conditions to a porcine IFN-w probe in implanting conceptus mRNAs (24).However, until now, direct molecular evidences concerning this pig embryonic type I IFN were lacking.

A Novel Type One Interferon Gene 19761
In this report, we describe the cloning of a type I IFN gene expressed by the pig trophoblast during implantation.Surprisingly, the product of this gene appears distantly related to all previously cloned type I IFNs, including IFN-7.Upstream putative regulatory regions of the relevant gene display no similarity with corresponding regions of other type I genes.
Moreover, analysis of sequence divergence suggests that such IFN could also exist in other mammalian species, thus defining a new type I IFN family.

EXPERIMENTAL PROCEDURES
Animals, Collection of Embryos, and Culture of Trophoblastic Vesicles-Large White and Meishan gilts were observed for estrus and naturally inseminated.Conceptuses were removed by flushing the uterus with sterile medium as previously described (20).Trophoblastic vesicles were prepared from day 14 conceptuses flushed with 199 medium supplemented with penicillin (100 units/ml) and streptomycin (0.1 mg/ml) and prewarmed at 37 "C.Zones of the elongated conceptuses corresponding to tubular trophoblast were cut in 2-4mm fragments.The fragments were grown in 199 medium supplemented with antibiotics as above, human insulin (0.2 IU/ml) and fetal calf serum (lo%, v/v) in 25-cm2 culture flasks (each containing about 40 fragments) at 37 "C under constant agitation.Trophoblastic vesicles generally appeared during the following 48 h.Those larger than 2 mm in diameter were then separated from other cellular debris, placed in fresh medium, and cultured for 1 or 2 more days in the same conditions.They were then collected, rinsed in medium without serum, frozen in liquid nitrogen, and stored at -70 "C until RNA extraction.
Materials-All enzymes were used according to the supplier's recommendations.Restriction endonucleases, DNA-modifying enzymes, DNA polymerases, ribonucleases, proteinase K, and dNTPs were purchased from Boehringer Mannheim.Reverse transcriptase enzyme and buffer used in primer extension experiments were from Life Technologies, Inc. Acrylamide was from Bio-Rad and agarose from Appligene.Other reagents were "pro analysis" from Merck and Normapur from Prolabo.Oligonucleotides were synthesized on a Biosearch 8600 synthesizer using MilliGen/Biosearch reagents.Labeled dNTP and rNTP were from Amersham Corp.
Standard Techniques-DNA and RNA purification, electrophoresis, subcloning, and sequencing of DNA fragments, labeling of oligonucleotides, solution hybridization-ribonuclease protection, primer extension, and transfections were performed as described in Ref. 25.However, primer extension was modified as follows; the oligonucleotide annealing step was performed overnight at 30 "C in 0.17 M HEPES pH 7.5, 0.34 mM EDTA, 0.3 M NaCl in the presence of placental ribonuclease inhibitor (Pharmacia LKB Biotechnology Inc.).Southern blot analysis of genomic DNA was performed on nitrocellulose (Schleicher & Schuell, BA 85) as previously described (24).Northern blot analysis of RNA was done using Hybond N+ nylon membranes (Amersham Corp.) according to the manufacturer's recommendations and subsequently hybridized as described (26) except that bovine serum albumin was replaced by nonfat dry milk in the hybridization buffer.Double-stranded DNA probes were labeled by random priming ( 27) with [a-32P]dCTP or dATP (3000 Ci/mmol).Antisense cRNA probes were generated using the T7 RNA polymerase kit from Stratagene with [a-"PIUTP (800Ci/mmol) as labeled rNTP.
Construction and Screening of cDNA and Genomic Libraries-An oligo(dT)-primed pig conceptus cDNA library was constructed in X Uni-ZAP XR phage vector using the Stratagene cDNA synthesiscloning kit as recommended by the supplier.Day 14 and day 15 pig conceptus polyadenylated (poly(A+)) RNAs, twice purified by oligo(dT)-cellulose chromatography, were prepared separately from pools of several conceptuses and mixed in equal amounts.Starting from 2 pg of this mixture, 9 X 10' independent clones were obtained.After one round of amplification, the clones were plated and screened with the 795-base pair (bp) Eco0109I DNA fragment containing the complete coding region of the porcine IFN-w3 gene (24).Hybridization and subsequent washings were done under the low stringency conditions described in Ref. 24 for Northern analysis of conceptus mRNAs.A porcine genomic library was constructed by cloning partial Sau3A digest of total pig liver genomic DNA between the BamHI sites of X EMBL4 or X DASH11 phage vectors according to standard methods and screened using conditions described previously (24).
Antiuiral Assays and Seroneutralization of IFN Actiuity-Vectors designed to express coding sequences of the two cDNA allelic variants and locus 2 open reading frame (ORF) were constructed by cloning the HinPI-RsaI DNA fragment (conserved in the three sequences) containing the coding sequence in the SmaI site of the eukaryotic expression vector pSVL (Pharmacia).Constructs were transfected in COS-7 cells and antiviral activity was assayed in cell culture supernatants by inhibition of the cytopathic effect of vesicular stomatitis virus on bovine MDBK cells (28).Seroneutralization of this activity was assayed by adding the specified antiserum in the titration medium as described (28)  Saitou and Nei (35), respectively.Data base searches were performed program (38).CLUSTAL V and PHYLIP were both available on the mology between pairs of sequences were done using the BestFit French CITIZ/Bisance network (39).

A Gene Coding for a Type I IFN Is Expressed by Pig
Trophoblast around Implantation-Our previous work (24) described the detection of transcripts related to the coding region of the porcine IFN-w3 gene in implanting pig conceptus.In order to clone relevant cDNAs, a 14-15-day old conceptus cDNA library primed with oligo(dT) was constructed in X ZAP vector and screened using this probe.From a total of 6 X lo5 phage plaques analyzed, 43 positive clones were isolated and rescued as plasmids.Sequence of their inserts indicated that some of them were completely identical (because of library amplification) so that there were only 32 distinct clones.They were arranged in four distinct groups of overlapping clones corresponding to four different cDNAs.Three of them (representing 17 clones) appeared devoid of long ORF and represented probably the 3"noncoding region of incomplete cDNAs with fortuitous homology to the probe.

A Novel Type One Interferon Gene
The fourth group contained 15 overlapping clones defining a cDNA of 1096 bp bearing a 170-codon-long ORF initiated by the first ATG triplet of the sequence and flanked by a 5'noncoding region of 207 bp with stop codons in the three phases (Fig. lA).The 3"noncoding region was 379 bp long and contained several potential polyadenylation signals, the most distal being probably the only functional one.As expected, the ORF displayed regions of significant homology to the porcine IFN-w3 coding sequence, but the overall homology was rather low even after optimal alignment (56%).The predicted polypeptide sequence (Fig. 1A) appeared to be clearly related to other type I IFNs preproteins since it could be aligned with members of the IFN-a, -6, -w, and -T families (Fig. 2 A ) , but unrelated to IFN-y.Moreover, the hydrophobic N-terminal end of the molecule had the characteristics of a signal peptide.Using the method of von Heijne (40), signal peptidase cleavage was predicted to occur after amino acid 21, leading to a mature protein of 149 residues.This protein had two potential N-glycosylation sites at positions 3 and 79.Interestingly, we found two types of cDNA clones, differing only at two positions in the coding sequence, resulting in amino acid changes at position 67 and 80 of the mature protein.This indicated the existence of two allelic variants of the protein: a Gly67-Tyrs0 and a Ser67-Asnso variant.Both ORFs were cloned in the proper orientation downstream from the SV40 late promoter in the pSVL eukaryotic expression vector.Antiviral activity (about 1-3 x lo3 IU/ml) was detected in the supernatant of COS-7 cells transiently transfected with these constructs, indicating that both ORFs encoded a secreted and biologically active IFN.Simple size comparison with other type I IFNs clearly showed that this IFN is the smallest natural type I IFN cloned until now.Thus, for more clarity, it will be provisionally abbreviated spI IFN (for qhort porcine type 1 IFN) hereafter in this paper.Interestingly, its antiviral activity was neutralized by antisera against virusinduced human IFN-Le as well as was part of the activity secreted by conceptuses (20).
A single 1.3-kilobase (kb) transcript homologous to the cDNA was detected in conceptus mRNAs during implantation (Fig. 1B).Its kinetics of expression were very similar to those of trophoblastic IFN-y mRNA (20); it appeared high at day 14-15 of gestation and decreased markedly until day 20, when it was hardly detectable.Around day 15 of gestation, extraembryonic trophoblast represents quantitatively the main part of the conceptus (41); thus, it seemed very likely that this type I IFN was expressed by this tissue.In order to verify this, we looked for the presence of the relevant transcript in trophoblastic vesicles, which are conceptus explants grown in uitro and composed only of trophoblast (42).Fig. 1C shows that after 4 days of culture, the transcript is detectable by RNase protection assay in total RNA from trophoblastic vesicles.

The Type I IFN Gene Expressed by Pig Trophoblast Codes for a Protein Distantly Related to Other Type I IFNs Including
IFN-T-To compare spI IFN to other type I IFNs, we created a multiple sequence alignment including all IFN-a, -P, -w, and -7 non-allelic precursor sequences published to date (not shown).This multiple sequence alignment, shown in part in Fig. 2.4, revealed several notable features of spI IFN primary sequence.The shortness of the molecule is due to the presence of a 7-amino acid deletion that occurs between residues 103 and 104 of the presumptive mature protein and to a trimmed C-terminal end, which lacks 11-14 residues compared to IFNa, 9 compared to IFN-P, and 17 compared to IFN-w and T. This is not due to premature termination caused by nonsense mutation in a pre-existing longer reading frame because the   In most cDNA clones, the poly(A) tail was found after the most distal of these sites as figured here.In some cDNA clones, it was found after one of the five overlapping poly(A) signals of the (AAAT),AAA sequence, probably because oligo(dT) primer annealed to the downstream A-rich sequence of the mRNA.However, one cannot exclude the possibility that these poly(A) sites may function in some cases.B , Northern blot analysis showing the kinetics of expression of the type I IFN mRNA in peri-implantatory pig conceptus tissues.On each lane, 5 pg of poly(A+)-enriched RNA from conceptus sampled at the appropriate day of gestation were analyzed.After hybridization with a cRNA probe corresponding to positions 161-818 of the cDNA, the blot was re-hybridized with a double-stranded DNA human @-actin probe as control.Size markers (in kilobases) are indicated at right.C, ribonuclease protection assay showing the expression of the type I IFN mRNA in porcine trophoblastic vesicles after 4 days of in uitro culture.The antisense cRNA probe corresponded to positions 161-465 of the cDNA flanked by 50 nucleotides of unrelated plasmid-derived sequences.Samples of RNA (5-10 pg) from 3 independent batches of trophoblastic vesicles (numbered 1-3) were analyzed.Samples were run on a 6% polyacrylamideurea gel.Signal obtained from day 14 total conceptus RNA (CSP) is shown as a control.C-T, control experiment without target RNA sequences; C-R, control experiment without RNase treatment; M, end-labeled DNA markers (405, 308, 243, and 239 nucleotides).Arrows indicate the position of the full-length probe (f.l.p., 355 nucleotides) and of the protected fragment (pi., 305 nucleotides).- This alignment is part of the multiple alignment constructed as described under "Experimental Procedures."Dashes indicate gaps inserted to optimize the alignment.Potential N-glycosylation sites are underlined.Presumptive signal peptide and mature protein sequence of spI IFN are numbered SI to S21 and 1 to 149, respectively.The amino acid changes due to allelic variation of the protein are indicated above the sequence as well as Cys residues (asterisks).Residues of spI IFN identical in other type I IFN sequences are shaded.B, phylogenetic tree of 24 type I IFN mature protein sequences including spl IFN and selected members of type I IFN families from 12 mammalian species (one subtype per species per family).This tree was established using the Fitch-Margoliash distance matrix method as described under "Experimental Procedures."Horizontal branch length is proportional to the amount of sequence divergence.The branch linking the IFN-/3 family t o all other type I IFNs was divided a t midlength in order to root the tree.The first two letters of sequence names refer to the species (Hu, human; Bo, bovine; Ou, ovine; Go, goat; mo, musk ox; Po, porcine; Mu, murine; Do, dog; Ca, cat; Eq, equine; Rb, rabbit).Accession numbers or references for the sequences analyzed are listed as follows according to the top to bottom display order of the tree (where Sw is SwissProt and Gb is nucleotide sequence downstream from the stop codon is devoid of homology with C-terminal coding regions of other type I IFNs.The cysteine content of spI IFN also appeared unprecedented among type I IFNs as the preprotein and mature protein contain 9 and 7 Cys residues, respectively, of which only 2 of the 4 canonical Cys residues found in most type I IFNs are conserved (3, 4).This pair of residues (Cys30 and C Y S ' ~~) is conserved in all type I IFNs except murine IFN-P and is known to form a disulfide linkage crucial for the antiviral activity (43).The remaining 5 Cys residues occurred at a rather atypical position; 3 of them have never been described at such positions (Cys',Cys5',Cys'07) and the 2 others (Cys5'j and were each found at a similar position in only four and one type I IFN sequences previously reported, respectively (not shown).These atypical cysteines could have important implications for the structure of spI IFN molecule.Interestingly, alignment of the primary sequences of spI IFN and MuIFN-/3 molecules, the three-dimensional structure of which was recently determined (44), and examination of the spatial arrangement of MuIFN-p residues homologous to the atypical Cys residues of spI IFN suggest that 4 of them could form two additional disulfide linkages (between Cys5'j and Cyslo7 and between CysSs and Cys14') thus linking distant regions of the molecule.

A Novel
Our multiple alignment also indicated that spI IFN displays a weak sequence homology with the 4 families of type I IFNs.The ranges of percentage of identity to the various mature IFN-a, -P, -a, and -7 subtypes (excluding positions with gaps) were 30-46%, 27-33%, 37-43%, and 37-40% respectively.Interestingly, no preferential identity to IFN-7 was noted, thus ruling out the possibility that spI IFN could belong to this family of trophoblast IFN.The lowest identity was to MuIFN-P (27%), and the highest was to human IFN-a2 (46%).Homology of coding sequences was also very low and probably explains the failure of some previous attempts to identify type I-related transcripts in pig conceptus tissues (11,18).
To analyze more rigorously the degree of similarity and the evolutionary relatedness between spI IFN and members of the four type I IFN families, we performed a phylogenetical analysis of these sequences, assuming that they evolved independently from a common ancestor by successive gene duplications (5).Phylogenetic trees were generated from the multiple sequence alignment previously mentioned using the least square method of Fitch (34) and the neighbor-joining method of Saitou and Nei (35).For more clarity, Fig. 2B shows a Fitch tree generated only from a representative selection of a phylogenetically broad range of type I IFN mature sequences (24 sequences, one per family and per species).This tree indicated that spI IFN is clearly distinct and divergent from the cluster defined by the IFN-P sequences on one hand and the cluster defined by IFN-a, -a, and -T on the other hand.Moreover, although spI IFN appeared less related to IFN-6 than to IFN-a, -u, and -7, it is clear that members of the IFN-a, -u, -7 cluster are more related to each other than to spI IFN.These facts were very consistent since the same general topology was found in all trees generated from the two methods mentioned above, even when all type I IFN sequences were included in the analysis (not shown).From an evolutive point of view, they suggest that the duplication event leading to spI IFN gene was posterior to the divergence of IFN-P but anterior to that of IFN-CY, -a, and -7 families.
We also tested the antigenic relationships of spI IFN expressed in vitro from transfected COS-7 cells with other type I IFNs.Several neutralizing antisera directed against a single member of IFN-CY, -p, -w, and -7 families were unable to neutralize the antiviral activity of spI IFN (Table I).This was notably the case for a monospecific antiserum raised against recombinant human IFN-CYB, which is the type I IFN showing the highest identity to spI IFN.These data strongly suggest that this IFN is devoid of antigenic cross-reactivity with IFNa, -p, -u, and -7, although the paradoxical neutralization of its antiviral activity by antisera raised against HuIFN-.Le was confirmed using three distinct antisera.Structure of the Type I IFN Gene Expressed by Pig Trophoblast-In order to investigate the number and the structure of porcine genomic sequences homologous to spI IFN cDNA, we performed Southern blot analysis of pig genomic DNA under normal stringent conditions.The DNA from two individuals from different breeds (European Large White and Chinese Meishan) was analyzed.Fig. 3A shows that, using enzymes with no cleavage site in the cDNA probe sequence, only two hybridizing genomic fragments were detected, whereas BglII, which has a single cleavage site in the cDNA, generates four hybridizing fragments.This strongly suggested the existence of only two non-allelic loci homologous to spI IFN cDNA.T o confirm this, a genomic library was constructed from the DNA sample of the Large White pig analyzed in Fig. 3A and screened with the cDNA.Two strongly hybridizing clones were isolated and named X 1 1 and X44.As expected, restriction mapping of their genomic insert (Fig. 3B) indicated that they were non-allelic and that each of them contained a single -1-kb-long region hybridizing to the probe.Moreover, all restriction fragments detected on genomic blots had a same sized equivalent hybridizing to the probe in one of the two clones.This clearly indicated that genomic sequences homologous to the cDNA had been exhaustively cloned and that consequently, the pig genome contained only two non-allelic loci homologous to spI IFN cDNA.They were designated locus 1 (clone X44) and locus 2 (clone All).Moreover, differences in genomic blot patterns observed between the two individuals (Fig. 3A, XbaI and BamHI lanes) indicated polymorphism of locus 2.
Nucleotide sequence of the two loci revealed that both contained an ORF highly homologous to that of spI IFN cDNA and devoid of intervening sequences (Fig. 4).We determined the sequence of =l kb upstream and -0.6 kb downstream from the coding sequences.Sequence comparisons clearly indicated that spI IFN mRNA was expressed from locus 1; the sequence of the cDNAs encoding the Gly67-Tyrm allelic form of spI IFN was identical to homologous TABLE I Neutralization of spl IFN activity by antisera Seroneutralization of IFN activity was assayed by a constant antibody method (28) in which the specified dilution of the antiserum was applied to MDBK cells before serial 3-fold dilution of a COS-7 cell-derived spI IFN preparation of known titer.Residual activity was the ratio of units measured in control and serum treated sample after VSV challenge.All sera were against purified recombinant molecules except those against HuIFN-Le and IFN-P.regions of locus 1.The 5"noncoding region of the cDNAs was only known for the one encoding the Ser6"-Asnm form and differs from locus 1 by two single base pair substitutions, which most likely reflect further differences between the 2 alleles (Fig. 4).Locus 2, although highly homologous to locus 1 (95% identity for sequences presented in Fig. 4, excluding gap positions) displayed in its 5'-and 3"noncoding regions several substitutions and insertions/deletions that clearly distinguished it from spI IFN cDNA and locus 1 sequences.As we found no cDNA derived from locus 2 among the 32 cDNA clones isolated, this gene is probably not transcribed or transcribed a t a very low level in day 14-15 pig conceptus.Moreover, alignment of locus l and locus 2 coding sequences showed that the latter underwent a single base pair insertion in codon 163, resulting in a premature stop codon and a deduced translation product lacking 7 C-terminal residues when compared to spI IFN.We detected no antiviral activity in the supernatant of COS-7 cells transfected with a pSVL -locus 2 """""""""""""-c"""""-c""".

TK T C T G C T G T C C A T
TGC GTA GAG AAG AGG GAA ATC T T C AAA 3 3 1

AGC C T T G A T C l T GGC CTG AGA AGA CTG GAG CAT ATC AAA AAA GAC AAT ATG GAT TGT CCC CAT GTA GGA T C T G C T C T C CGA AAG T A T T T C CAA 6 3 7
2 ------"-"-"-  -------------------.... ----------------------------------------T-------------------C----------  construct designed to express the locus 2 presumptive product, thus indicating that it is most likely devoid of biological activity.We thus concluded that locus 2 had the characteristics of a pseudogene.To identify the transcription start site of the spI IFN gene, we performed a primer extension analysis using an oligonucleotide complementary to the 5' end of the mRNA (Fig. 5).A single major extension product was found, which indicated that the transcription initiation was located 217 or 218 bp upstream from the ATG codon.The former position was considered as the +1 position of spI IFN gene.Fig. 4 shows that the region upstream from this site contained a TATA box starting at position -33.We performed searches for locus 1-related sequences in the GenBank data base using BLAST and FastA programs and also pairwise comparisons of this sequence with many IFN-a, -8, -w, and -7 loci using the local homology searching program BestFit.These analyses clearly showed that the significant homology between locus 1 and other type I IFN genes was strictly limited to coding regions.Noncoding regions of spI IFN gene, including those located upstream from the cap site and supposed to be involved in the transcriptional regulation of the gene, displayed neither significant identity nor similarity to upstream regions of other type I genes.A common feature of all type I loci sequenced to date including IFN-7 is the presence in the 100-150-bp region upstream from the cap site of several clustered copies of short motifs known to be the target of transcription factors regulating virus-inducibility: AA(A/G)(T/G)GA hexamers (45), which are potential binding sites for IFN regulatory factor 1 and 2 (IRF-1 and 2) and/or GAAANN (N being any nucleotide) and most importantly GAAATG hexamers (46,47).Locus 1 upstream sequences contained no potential IRF-l/IRF-2 binding sites and only four GAAANN hexamers (but none of the GAAATG type) not closely clustered and located relatively far upstream from the TATA box (positions -193, -530, -567, and -575).On the other hand, several GCrich sequences bearing elements identical or quasi-identical to consensus binding sites for transcription factors Spl and AP-2 were found in this region (Fig. 4).However, the most striking feature was the general organization of the 450-bp region located upstream from the TATA box.As shown in Fig. 4, it contained 2 adjacent 40-41-bp direct repeats (RZZ and RZZZ) and a third imperfect and 5"trimmed copy of this repeat (RZ) further upstream.Those repeats contained a GCrich core reminiscent of an AP-2 binding site.Moreover, the most distal element was flanked on its 5' end by a 267-bp region that we identified as a repetitive DNA sequence belonging to the porcine family of short interspersed element (SINE) called PRE-1 (49).Such elements are retroposons (50), which have been identified in the noncoding region of many porcine genes (51).All these elements were also present in the highly homologous locus 2 except the 40-bp repeat proximal to the TATA box, which was completely deleted in this locus.

DISCUSSION
We report here the molecular cloning of a type I IFN gene transiently expressed by the trophoblast of the pig conceptus during implantation.Although confirmation of the presence of spI IFN protein in peri-implantatory conceptus secretions is necessary, this protein probably corresponds to the type I IFN-like activity secreted by conceptuses, since the activity of both proteins share similar reactivity with anti-HuIFN-Le serum (20,21).It is also probably identical to the 22-kDa type I IFN-related protein detected by Cross et al. (18) by Western blot analysis in conceptus secretory products because we found that the N-glycosylated form of spI IFN immunoprecipitated from the supernatant of transfected COS-7 cells exhibits a similar apparent molecular mass (data not shown).
Nevertheless, our data provide the first clear and unambiguous characterization of a type I IFN gene physiologically expressed by the trophoblast of a non-ruminant mammalian species.Although spI IFN clearly belongs to type I IFNs, several points of our study highlight its very atypical nature and novelty.(i) The primary structure of the protein exhibits several unique features of size and residue composition.(ii) It is distantly related to all IFN-a, -p, -w, and -T subtypes and clearly segregates from these four families according to a phylogenetical analysis of sequence divergence.(iii) The biological activity of the protein is antigenically different from several members of these four families.(iv) The sequence and the organization of the presumptive regulatory regions of its gene have no equivalent among type I IFN genes.Taken together, these results indicate that this IFN is a structurally distinct subtype and strongly suggest that it must be classified in a new family of type I IFN that will require designation by a specific Greek letter.3We insist on the fact that its designation as spI IFN is quite provisional and was used in this In a previous poster communication (52), IFN-J.was proposed, but this nomenclature may be confusing and should be abandoned.
paper for more clarity only.The efficient neutralization of spI IFN activity by sera against virus-induced HuIFN-Le but not by sera against HuIFN-a2 and HuIFN-wl, which are two major components of 54), is surprising and could mean that spI IFN cross-reacts with some other IFN-a or -w subspecies.However, a single IFN-w protein seems to exist in man (54), and IFN-a:! is the known human protein sharing the highest homology to spI IFN so that cross-reactivity with a less homologous IFN-a is doubtful.A tempting and more satisfactory hypothesis is that spI IFN cross reacts with a yet unknown, closely related molecule present in IFN-Le and that consequently a virus-inducible member of the spI family exists in man.
Besides the characterization of a previously unknown type I IFN, our study reveals the striking disparity between the type I IFN genes expressed by trophoblast of pig and ruminants, which are relatively close species belonging to the same mammalian order.This raises some interesting questions about the comparative evolutive origin of these genes.IFN-TS are expressed from a multigene family composed of several active genes and pseudogenes (14, 29).Several arguments strongly suggest that they have evolved relatively recently from virus-inducible type I IFN genes.IFN-T protein and genes are closely related to IFN-w, which are virus-inducible; upstream regulatory regions of IFN-T display in their proximal region some sequence homology with IFN-a, -p and -w as well as clustering IRF-1/2 binding sites and GAAANN hexamer typical of these virus-inducible genes (30); finally, IFN-T have retained a slight virus inducibility in non-trophoblastic cells (55).Moreover, their probably recent divergence from IFN-w could explain why their existence in mammals seems limited to ruminants (14 will be interesting to characterize.One should note that the divergence time proposed above is probably underestimated.Indeed, the evolutionary relationships between spI IFN gene and other type I families proposed in Fig. 2B, which are based on divergences calculated from protein sequences, might not reflect the actual situation because of gene conversion events occurring between coding sequences and counteracting divergence resulting from nucleotide changes.Such a phenomenon is known to occur among type I IFN loci (5) and could explain why, whereas noncoding sequences of spI IFN are unrelated to those of other type I genes, spI IFN protein shares more identity with IFN-a, -u, and -7 than with IFN-(3.It is possible that the spI IFN progenitor diverged earlier than IFN-(3 and IFN-a-u-7 ancestors and underwent later gene rectification events preferentially from coding sequences of genes belonging to the IFN-a, -u, and -T lineage.It is clear that answers to these questions and clarification of the origins of spI IFN loci require characterization of homologous genes in other mammalian species and analysis of their virus inducibility and/or trophoblast-specific expression. However that may be, the absence of similarity between sp1 IFN and IFN-T regulatory sequences is striking and indicates that although both type of genes are expressed by homologous tissues at the same developmental stage, they are probably regulated by very different transcriptional mechanisms.The presence of Spl and AP-2 binding sites consensus sequences suggests that these transcription factors could be involved in the regulation of the gene.Interestingly, AP-2 binding sites have been shown to mediate transcriptional activation in response to phorbol esters and CAMP elevating agents (59), thus suggesting that protein kinase C-and/or protein kinase A-dependent signal transduction pathways could regulate the expression of spI IFN gene.However, temporally and spatially regulated expression of AP-2 mRNA was recently reported in the mouse embryo, mainly in neural crest cells (60).This suggests that AP-2 belongs to the group of developmentally regulated transactivators that play an important role during vertebrate embryogenesis.Such factors (like Oct-4) were recently shown to bind and transactivate IFN-T gene promoter in some embryonic and choriocarcinoma cell lines (61), suggesting that initiation of IFN-T expression in ruminant early trophoblast cells is a genetically programmed event (61).Such developmentally controlled expression could also be relevant for spI IFN gene if AP-2 is involved in its trophoblast specific regulation.The presence of a SINE element in the very close proximity of presumptive regulatory regions of locus 1 represents an original feature never reported for a type I IFN gene.Since such elements generally cause transcriptional repression in cis of downstream transcriptional units (see, for example, Refs.62 and 63), it would be interesting to evaluate the contribution of this element to the regulation of the spI IFN gene.Further work is required to determine the functional significance of spI IFN gene upstream sequences, which are the most interesting feature of this novel type I IFN gene.If it is confirmed that locus 2 is actually not transcribed in trophoblast during implantation, comparison of locus 1 sequence to this highly homologous silent locus could help to delineate sequences involved in trophoblast specific transcription.Thus, the 40-bp repeat proximal to the TATA box, which is deleted in locus 2, could be consequently a good candidate for such a region.
In the pig, maternal recognition of pregnancy does not seem to be mediated, as in ruminants, by conceptus secretory proteins (64), so there is no preferential hypothesis about the target cell and the specific role of sp1 IFN in the early maternal-fetal interplay.Although it is clear that type I IFN is expressed in the pig at much lower levels than in ruminants, one should keep in mind that porcine trophoblast type I IFN is co-expressed with IFN-7.This represents a unique physiological situation in which type I and type I1 IFN gene and protein expressions are temporally co-regulated.It is known that type I and type I1 IFNs can act synergistically through different receptors and signal transduction pathways to control IFN-stimulated response element-dependent transcription (65,66).Consequently, the lower amount of IFN mixture expressed by pig conceptus could be sufficient to trigger the same response as in ruminants where only type I IFN is expressed.
Our results stress that trophoblast IFN could represent a very heterogeneous category of mammalian IFNs and raise doubts about the uniqueness of IFN-T as trophoblast IFN.They suggest consequently that distinct groups of mammals even relatively close, might have evolved trophoblast IFN from different type I families.Finally, characterization of this novel type I IFN clearly demonstrates that our present view of the mammalian IFN system is not yet exhaustive and further increases the interest of characterizing IFNs expressed during physiological processes as important as early development.
A?G GCT CAG ATT TAC TTG GTA ATG GCA GGA GTG ATG CTC TGC TCC ATC H TGC TTT CTT GAC CAG IWC TTG TCT GCT GTC CAT TGC GTA GAG 304 AAG AGG GAA ATC TTC AAA CAT TTG CAA GAG ATA AAA AAG ATC CCT TGC TTA AAG GAC AGA ATC GAC TTC AAA TTT CCT TGG AAA AGA CTG ATG CTT CAG CAG ATC TCC AGT GTC TTC TCC AAA AAG GAC AGC CTG GGC CTG AGA AGA CTG GAG CAT ATG AAA AAA GAC AAT ATG GAT TGT 592 ccc CAT GTA GGA TCT GCT CTC CGA AAG TAT TTC CAA GGA ATc CM CTC P FIG.1.A, nucleotide sequence of the type I IFN cDNA isolated from a day 14-15 pig conceptus cDNA library.The nucleotides are numbered at left.The longest open reading frame and its type I IFNrelated deduced sequence (in italics) are shown.Amino acid residues are numbered at right.The presumptive N-terminal signal peptide (21 residues) and the potential N-glycosylation sites are underlined with a thin line.The two substitutions causing allelic variation (positions 469 and 509) are indicated below the nucleotide sequence, and the corresponding amino acid changes are indicated above the polypeptide sequence.Potential polyadenylation signals are underlined with a thick line.In most cDNA clones, the poly(A) tail was found after the most distal of these sites as figured here.In some cDNA clones, it was found after one of the five overlapping poly(A) signals of the (AAAT),AAA sequence, probably because oligo(dT) primer annealed to the downstream A-rich sequence of the mRNA.However, one cannot exclude the possibility that these poly(A) sites may function in some cases.B , Northern blot analysis showing the kinetics of expression of the type I IFN mRNA in peri-implantatory pig conceptus tissues.On each lane, 5 pg of poly(A+)-enriched RNA from conceptus sampled at the appropriate day of gestation were analyzed.After hybridization with a cRNA probe corresponding to positions 161-818 of the cDNA, the blot was re-hybridized with a double-stranded DNA human @-actin probe as control.Size markers (in kilobases) are indicated at right.C, ribonuclease protection assay showing the expression of the type I IFN mRNA in porcine trophoblastic vesicles after 4 days of in uitro culture.The antisense cRNA probe corresponded to positions 161-465 of the cDNA flanked by 50 nucleotides of unrelated plasmid-derived sequences.Samples of RNA (5-10 pg) from 3 independent batches of trophoblastic vesicles (numbered 1-3) were analyzed.Samples were run on a 6% polyacrylamideurea gel.Signal obtained from day 14 total conceptus RNA (CSP) is shown as a control.C-T, control experiment without target RNA sequences; C-R, control experiment without RNase treatment; M, end-labeled DNA markers (405, 308, 243, and 239 nucleotides).Arrows indicate the position of the full-length probe (f.l.p., 355 nucleo- FIG.2.A , alignment of spI IFN preprotein sequence with several members of the four families of type I IFN (IFN-a, -/3, -w, and -T).

FIG. 3 .
FIG.3.The pig genome contains two non-allelic loci homologous to spI IFN cDNA.A, Southern blot analysis of porcine genomic DNA with the spI IFN cDNA probe.The DNA analyzed was obtained from two individuals belonging to different pig breeds (LW, Large White; MS, Meishan).Fifteen pg of DNA restricted with the specified enzyme were loaded on each lane of a 0.8% agarose gel.The blot was hybridized with a probe bearing the complete coding sequence of spI IFN (positions 161-818 of the cDNA).The positions of molecular weight standards (in kilobases) are shown at left.B, restriction map of the genomic insert of two X clones ( X 1 1 and X44) bearing a spI IFN-related locus.The thick line on each clone marks the limit of location of spI IFN cDNA-related sequences as determined by restriction mapping.These clones were isolated from a library constructed from the DNA of the Large White individual analyzed in A.

FIG. 4 .
FIG. 4. Nucleotide sequence of the 2 loci homologous to spI IFN cDNA.Sequences of locus 1 (from which spI IFN mRNA is transcribed) and locus 2 (a related pseudogene) are aligned, and positions of the ORFs are shown.Gaps (dots) were introduced to maximize homology.Dashes in locus 2 sequence indicate identity with locus 1.Nucleotide numbering refers to locus 1 sequence in relation to the cap site (vertical arrow).The TATA consensus sequence and start and stop codons are boxed.The large gray box indicates location of the PRE-1 element.Horizontal, black half-arrows indicate the position of the 40-41 bp direct repeats (RZZ and RZZZ) and of the 5"trimmed 30 bp repeat that lacks the first 11 bp of repeats I1 and 111.In locus 1 sequence, the 5' end and the 3' end (position of the poly(A) tail) of the longest cDNA clones are indicated by black triangles.Positions showing allelic variations (see "Results") are indicated by asterisks.Sequence elements located upstream from the cap site and matching perfectly with Spl and AP-2 consensus binding sites defined in release 5.2 of the TFD data base of Gosh (48) are indicated above the sequence by gray and black bars, respectively.Those matching almost perfectly with AP-2 consensus binding sites are indicated by hatched bars.The four GAAANN hexamers found in this region are underlined.

FIG. 5 .
FIG.5.Analysis of the transcription start site of spI IFN mRNA.Primer extension analysis was performed using a 32P-labeled synthetic primer (39-mer) complementary to the 5"noncoding region of the mRNA (positions 99-61 of the cDNA sequence in Fig.lA).Experiment was performed using 2.1 pg of poly(A+) RNA from day 14 (A) and day 15 ( B ) conceptus.Extension products were sized on a 6% acrylamide-urea gel using the sequence ladder numbered in nucleotides on the left.