Purification, Characterization, and cDNA Cloning of a Kunitz-type Proteinase Inhibitor Secreted by the Porcine Uterus*

The porcine uterus synthesizes a proteinase inhibitor (M, 14,000) under the influence of progesterone that is relatively specific for plasmin and trypsin, but that also has weak affinity for chymotrypsin. Several isoforms of this uterine plasmidtrypsin inhibitor were purified by a procedure whose final two steps involved affinity chro- matography on immobilized chymotrypsin and cation exchange chromatography. Amino-terminal sequencing showed that at least three of the isoforms were closely related. An oligonucleotide probe based on the protein sequence was used to identify a cDNA that contained an open reading frame coding for a mature protein (M, 10,295) of 93 amino acids. The inhibitor had a well de-fined, but unique, Kunitz domain of 64 residues at its amino terminus that shared 67% sequence identity to bovine pancreatic trypsin inhibitor. Its P, residue was arginine rather than lysine. Northern analysis showed the presence of a single mRNAspecies (700 bases) that in adult female pigs appeared to be confined to the uterus. UPTI was and significantly By reached maximal concentra- in data are consistent with an earlier that the inhibitor serves to neu- tralize the activities of one or more serine proteinases generated by the proliferating trophoblast during the

In most animals, implantation involves intrusion of the trophoblast into the uterine wall, thus bringing the developing placenta into close proximity of maternal blood vessels, a process that facilitates the transfer of nutrients from the dam to the fetus. Implantation has been associated with proteinase secretion by the developing blastocysts (1)(2)(3). In mice (4, 5 ) a n d humans (6,7), plasminogen activator (PA),' which can generate the broadly specific proteinase plasmin from its precursor plasminogen, has been implicated in this process. However, other proteinases are likely involved since murine blastocysts lacking both urokinase and tissue plasminogen activator have recently *This work was supported by Grant HD21980 from the National Institutes of Health and Grants 89-37240-4586 and 92-37203-7995 from the United States Department of Agriculture. This is Contribution 12026 from the Missouri Agricultural Experiment Station Journal Series. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
to the GenBankTM/EMBL Data Bank with accession number& L14282.
The nucleotide sequencers) reported in this paper has been submitted Wahpeton, ND 58075. ine plasmidtrypsin inhibitor; bp, base paids). been shown to implant normally (8). In contrast to most species, placentation in the pig does not involve erosion of the uterine wall but is a noninvasive process in which the microvilli on the surface of the trophoblast interdigitate with the epithelial cells lining the uterine endometrium (9,10). However, cultured pig trophoblast has been demonstrated to secrete PA(l1) and when transplanted to an ectopic site, such as the kidney capsule or oviduct, shows features of invasiveness (12,13). Uterine flushings from pregnant pigs contain plasminogen, thus providing a potential for production of plasmin within the uterine lumen of the pig (14).
The failure of the pig blastocyst to invade the uterine wall, despite production of proteinases, raised the possibility that the maternal uterus might secrete proteinase inhibitors. Such inhibitory activity was subsequently identified in the uterine secretions of pigs during pregnancy (11,14). The activity could be attributed to a group of basic, low molecular weight proteins and was directed toward plasmin, trypsin, and, to a lesser extent, chymotrypsin (15). Further characterization of the uterine plasmidtrypsin inhibitor (UPTI) indicated its presence in uterine secretions of pigs during the luteal phase (11) and pseudopregnancy (151, i.e. when levels of progesterone were high. Plasmidtrypsin inhibitory activity could also be detected in the uterine flushes of ovariectomized gilts treated with either progesterone or progesterone and estrogen in combination but not with estrogen alone (11,15,16). UPTI was identified immunocytochemically in the surface and glandular epithelium of the endometrium (16), and its initial secretion appeared to be triggered by the release of estrogen from conceptuses as they elongated from the spherical to the filamentous form between Days 11 and 13 of pregnancy (14). The aims of the present study have been to purify UPTI, to clone its cDNA, and to examine its expression in the uterus during pregnancy.

EXPERIMENTAL PROCEDURES
Purification of UPTZ-Uterine flushings obtained as described previously (17) were centrifuged at 5,800 x g for 45 min at 4 "C, and the supernatant fraction was dialyzed (4 "C) against 0.01 M Tris-HC1 (pH 8.2). Basic proteins were enriched by allowing them to bind to CMcellulose at pH 8.2 in 0.01 M Tris-HC1 buffer. They were then eluted in the same buffer containing 0.5 M NaCI. These basic proteins were then size-fractionated by gel filtration over a Sephadex G-100 column (3.5 X 100 cm, Pharmacia Biotech Inc.), and the low molecular weight fraction (range of M, 10,000 to 20,000) was pooled, dialyzed against Tris-HC1 buffer (0.01 M, pH 8.2), and loaded onto either an anhydro-chymotrypsin-Sepharose column (see below) or a similar column carrying bound chymotrypsin. After washing with 0. Preparation of Anhydro-chymotrypsin and Chymotrypsin Affinity Matrices-Anhydro-chymotrypsin was prepared according to the procedure of Matta et al. (20) and affinity-purified over a lima bean inhibitor-Sepharose column, which was prepared as described previously (21).
The resulting material was dialyzed against 0.01 M NaHCO, (pH 8.3) and coupled to CNBr-activated Sepharose (1 mg of proteid1 g of gel, Sigma).
In subsequent experiments, it was noted that chymotrypsin (type VII, l-chloro-3-tosylamido-7-amino-2-heptanone-treated; Sigma) could substitute as an effective affinity substrate for anhydro-chymotrypsin and did not cleave UPTI. Therefore, chymotrypsin was coupled directly to CNBr-activated Sepharose in the same manner as used for anhydrochymotrypsin.
lZypsin Inhibition Assay-Samples (0-100 p1) were incubated with 100 pl trypsin (20 pg/ml, type XIII, L-1-tosylamido-2-phenylethyl chloromethyl ketone-treated, Sigma) at room temperature for 15 min. The volume was adjusted to 450 pl by addition of 0.05 M Hepes (pH 7.0) and 50 p1 of 50 m M N-benzoyl-DL-arginine-p-nitroanilide (stored as a 50 m M stock in dimethyl sulfoxide, Sigma) were added. After incubating 15 min at 37 "C, 500 pl of soybean trypsin inhibitor (10 pg/ml, type I-S, Sigma) were added. Absorbance at 410 nm was measured. Since UPTI had previously been shown to bind trypsin very tightly in a 1:l stoichiometry (15), the amount of functional UPTI in any sample could be measured accurately.
Amino Acid Sequencing-UPTI was subjected to NH,-terminal amino acid sequence analysis by the Edman degradation method on an Applied Biosystems model 470 protein sequencer with on-line analysis for phenylthiohydantoin derivatives (Protein Core, University of Missouri, Columbia, MO).
Design of Oligonucleotide for Library Screen-NH,-terminal amino acid sequencing of UPTI indicated considerable similarity to porcine leukocyte inhibitor (22), bovine pancreatic trypsin inhibitor, and bovine serum trypsin inhibitor (23). Aregion of UPTI highly conserved in these other proteins (amino acids 16-27, see Fig. 2) was chosen, and an oligonucleotide, UPTI-5, was made corresponding to this region. The putative nucleotide sequence, 5'-GGC CCC TGC AGG GCC CAC TTC ATC AGG TAC TTC TAC-3', was constructed by assuming that point mutations accounted for the majority of the amino acid differences between U P T I and the porcine leukocyte inhibitor and by utilizing the most common codon usages observed for the pig (24).
Screening of cDNA Library-UPTI-5 was used to screen approximately 100,000 plaque-forming units of a A-gtll cDNA library constructed from polyadenylated RNA isolated from uterine endometrium of pigs at Day 60 of pregnancy (27) by standard procedures (22) under appropriate low stringency conditions (28). UPTI-31 was then subsequently used to screen approximately 100,000 plaque-forming units of a Uni-Zap (Stratagene Cloning Systems, La Jolla, CA) cDNA library constructed from polyadenylated RNA isolated from porcine endometrium that had been obtained from pigs between Days 8 and 13 of pregnancy. Positive phage plaques were identified by exposure to x-ray film (XAR-5, Eastman Kodak Co.) at -80 "C for 16-24 h and purified by secondary and tertiary screening.
Subcloning and Sequence Analysis-Inserts corresponding to two uterine clones (UPTI-31 and UPTI-33b) from the Day 60 library were amplified by polymerase chain reaction by employing primers flanking the A-gtll EcoRI insertion site and subcloned into Bluescript +/plasmids. In the Day 8-13 library, Bluescript plasmids containing UPTI cDNA inserts (UPTI-1 and UPTI-3) were isolated by means of R408 helper phage and the in vivo excision protocol for Uni-Zap, as described by the manufacturer (Stratagene Cloning Systems). Plasmid inserts from both libraries were sequenced by the dideoxy method (29) with M13 forward and reverse primers (30) and internal UPTI primers.
Source of Tissues for mRNA-Sexually mature cross-bred gilts were observed daily for estrus in the presence of intact boars. The onset of estrus was designated as Day 0. Pseudopregnancy was induced by daily injection of 2.5 mg of estradiol benzoate on . Hysterectomy was performed on Day 45 of pseudopregnancy (n = 2). Pregnant gilts ( n = 6) were slaughtered on Days 30, 60, and 90. Cervix, spleen, heart, and liver were obtained from a local abattoir. Total cellular RNA for muscle, perinephric fat, and kidney was kindly provided by the laboratory of Drs. T. G. Ramsay and M. White (Ohio State University).
Isolation of RNA from Porcine Tissues and Northern Analysis-RNA was isolated from porcine tissues as described previously (32). For Northern blot analysis, total cellular RNA was separated by formaldehyde-agarose gel electrophoresis and transferred to nylon filters (Magna NT, Micron Separations Inc., Westboro, MA) according to the method of Sambrook et al. (22). The filters were hybridized in a buffer containing the 32P-labeled UPTI-31 and 32P-labeled uteroferrin cDNA probes as described previously (33).
Statistical Analysis-The intensity of autoradiographic signals obtained from Northern blots was determined by densitometry with a Bio-Rad model 620 video densitometer (Bio-Rad). Signal intensity is expressed as the height in optical density units. The differences in intensity of signal for UPTI and uteroferrin were examined by one-way analysis of variance performed by the General Linear Model procedure of the statistical analysis system (SAS Institute, Cary, NC).

RESULTS
Purification of UPTZ-Chromatography of CM-cellulosepositive material over a Sephadex G-100 column yielded five major peaks (I-V) of protein (15,341 (data not shown). Fraction V contained a mixture of low molecular weight, basic proteins, including the majority of the trypsin inhibitory activity (15).
Pooled Fraction V was loaded onto the anhydro-chymotrypsin affinity column, and the trypsin inhibitor, which bound quantitatively, was eluted with a glycine buffer, pH 2.4. Electrophoretic analysis of the eluted protein material revealed a single broad band on one-dimensional SDS-polyacrylamide gel electrophoresis gels (data not shown). Treatment of sample with P-mercaptoethanol just prior to electrophoresis did not result in a change in migration of the band representing UPTI.  (Fig. 1) were subjected to NH,-terminal amino acid sequence analysis. Sequence information was obtained for the first 24 residues from Fraction 17 and the first 29 residues from Fraction 23 (Fig. 2). Fraction 17 yielded two signals, a minor one, starting with a valine residue followed by arginine, and a major signal starting with an alanine. These two signals were clearly variants of the same sequence, with the major component lacking the two terminal residues found on the minor component. Fraction 23 gave a single sequence identical to the minor signal present in Fraction 17.
The NH, terminus of UPTI was similar to the NH, terminus of native bovine pancreatic trypsin inhibitor and porcine leukocyte inhibitor (Fig. 2). The degree of sequence identity in the region of overlap was 66.7% for bovine pancreatic trypsin inhibitor and 75.0% for porcine leukocyte inhibitor. A key difference between UPTI and the other two inhibitors was the putative active site residue, P, (35). UPTI contained an arginine residue, whereas bovine pancreatic trypsin inhibitor (36) and porcine leukocyte inhibitor (22) contained a lysine residue at P,. However, the overall sequence identity over 8 residues on either side of the active site residues was very close (82%).
Cloning and Sequencing of UPTI cDNA-A 17-base oligonucleotide probe was designed to represent a region of UPTI, amino acids 16-27 (Fig. 21, which is highly conserved in other Kunitz-domain proteins such as bovine pancreatic trypsin inhibitor and porcine leukocyte inhibitor. The putative nucleotide sequence was derived by assuming that amino acid differences between the porcine leukocyte inhibitor (22) and UPTI arose as the result of point mutations and by examining frequencies of codon usage in the pig (24). This probe was used to screen approximately 100,000 plaques of a porcine endometrial cDNA library from Day 60 of pregnancy (27). Only 7 positive plaques were identified, indicating that the UPTI cDNA was not abundant in this library. After secondary and tertiary screening, the two longest clones, UPTI-31 (575 bp) and UPTI-33b (500 bp), were chosen and sequenced completely in both directions. Both clones were identical and both contained the poly(A) tail. The UPTI-33b clone did not contain the first 70 bp of UPTI-31. UPTI-31 had an open reading frame that coded for a mature protein of 93 amino acids with an estimated molecular weight of 10,295 but lacked the 5' end of the transcript, including part of the open reading frame.
A second porcine endometrial cDNA library, prepared from pooled mRNA collected at Days 8-13 of pregnancy, was therefore screened with the UPTI-31 cDNA. Numerous positive plaques were identified, and two clones with the longest inserts, UPTI-1 and UPTI-3, were sequenced completely in both directions (Fig. 3). Their only difference was an additional 4 nucleotides, located at positions -17 to -14 on UPTI-1. This insertion did not alter the inferred site for translational initiation. Both UPTI-l and UPTI-3 were identical in the regions where they overlapped with the clones isolated from the older  Day 60 library. Therefore, there was no evidence that the protein isoforms arose from distinct transcripts. The complete open reading frame of UPTI-1 started at position 31 and coded for a 122-amino acid polypeptide. The inferred signal peptide was similar in length and sequence to that of bovine pancreatic trypsin inhibitor (36).

491
Tissue Specificity of UPTI mRNA Expression-Northern analysis of total cellular RNA (20 pg) obtained from the endometrium of pseudopregnant gilts at Day 45 (n = 2) indicated a single band of approximately 700 bases (Fig. 4). In contrast, a signal for UPTI was not detected in total cellular RNA isolated from cervix, heart, spleen, or liver (Fig. 4) or from muscle, perinephric fat, or kidney (data not shown).
Expression of UPTI mRNA during Pregnancy-Northern analysis was performed on total cellular RNA (5 pg) obtained from the endometrium of gilts during pregnancy (Fig. 5). A strong signal was detected for UPTI on Day 30 of pregnancy. However, signal intensity decreased significantly (p < 0.05) on Days 60 and 90 of pregnancy. In contrast, uteroferrin mRNA expression significantly increased ( p < 0.05) on Days 60 and 90 when compared t o Day 30, as has been observed previously (26). DISCUSSION UPTI consists of a group of basic, low molecular weight proteins secreted by the porcine uterus under the influence of progesterone (11,(14)(15)(16), which has inhibitory activity toward plasmin, trypsin, and, to some extent, chymotrypsin (15). The purification procedure described in this paper separates several of these forms, three of which were subjected to NH,terminal amino acid sequencing. The sequences obtained were identical except for an apparent proteolytic cleavage of the first two amino acids, Val and Arg, in one form. In addition, all four UPTI cDNA sequenced were identical in the regions where they overlapped, which included most of the open reading frame. It seems likely that the different forms of UPTI arise as the result of posttranslational modifications of a single gene product.
Kunitz-type inhibitors are represented by a gene superfamily that is found throughout the animal and plant kingdoms. Members of this superfamily typically (but not invariably) are of low molecular weight and basic isoelectric point and have at least one Kunitz domain, which presents a broad spectrum of inhibitory activity toward serine proteinases (37). The inhibitory or Kunitz domain consists of a well conserved -60 amino-acid sequence, which is most commonly located at the NH, terminus on single domain inhibitors. An important feature of the Kunitz domain is the presence of 6 cysteine residues, whose spacing is crucial for proper folding (36). The three disulfide bridges that are formed appear to contribute to the high thermal stability of Kunitz domains (38). Kunitz inhibitors form unusually stable enzyme-inhibitor complexes, due to the ability of the inhibitor to act as a very effective substrate analog (39). Two binding loops define their specificities and form interfaces with the proteinase target (40). The residue located at the center of the first loop, the PI residue, mimics bound peptide substrate and defines the primary specificity toward particular classes of proteinases (41). The PI residue is most commonly occupied by a lysine. Additional specificity is conferred by the residues on each end of the binding loops, which interact with proteinase atoms around the perimeter of the catalytic site (42). UPTI has many of the characteristics of a classic Kunitz-type inhibitor, including its low molecular weight and basic isoelectric point, its single inhibitory domain located at the NH, terminus, and strict conservation of its 6 cysteine residues. By comparison with bovine pancreatic trypsin inhibitor, with which it shares 66.7% sequence identity, the two proteinase-binding loops are most likely located a t residues 15-23 and 3 8 4 5 i n UPTI (41).
The putative PI residue is occupied by arginine, a conservative replacement for lysine. Both residues are common for those Kunitz inhibitors that cleave at basic residues (36,421. Differences among the other residues in the binding loops between UPTI and other Kunitz inhibitors likely contribute to more subtle changes in inhibitor specificity among these proteins (41).
Bovine pancreatic trypsin inhibitor is synthesized as a larger precursor, containing both a signal peptide and a pro-region (42). Sequence analysis of the cDNAfor UPTI strongly suggests that it must also be proteolytically processed. The pre-region has the characteristics of a signal peptide (43), but the site of signal peptide cleavage is difficult to predict (43, 44), and the length of the pro-region (if one indeed exists) is unclear. Most likely cleavage occurs between Se?' and Th?', since this site conforms more closely than any other to the (-3,-1) rule, in which the residue a t -3 is not aromatic, charged, or bulky, and the residue a t -1 is small (44). The remaining 9 amino acids would then constitute a short pro-region. Significantly, U P T I does not have a cysteine residue in a position analogous to that found in the pro-region of bovine pancreatic trypsin inhibitor. Such a cysteine has been suggested to have a crucial role in folding of the Kunitz domain (45). Its absence in UPTI suggests that the requirement for a pro-region cysteine is not universal among Kunitz inhibitors.
The synthesis of UPTI is strongly under the influence of progesterone (15). Amounts produced, therefore, are high during the luteal phase of the estrous cycle (11, 161, as well as during pregnancy (14) and pseudopregnancy (15). Our results from the Northern analyses are consistent with this hypothesis, since expression of the UPTI mRNA was detected during all stages of pregnancy examined. However, expression was clearly stronger at Day 30 of the 114-day gestation period, a time at which rapid expansion and development of the allantois is occurring (46). This decrease in expression of the UPTI mRNA during later pregnancy is in contrast to uteroferrin ( Fig. 5; Ref. 261, the uterine serpin (471, and retinol-binding protein (481, whose mRNA expression remain high. Thus, it seems likely that UPTI functions mainly in early pregnancy, at which time it may, as previously suggested, help to control proteolytic cascades initiated by the potentially invasive pig trophoblast (11, 14-16).