A new inhibitor of metalloproteinases from chicken: ChIMP-3. A third member of the TIMP family.

We report cDNA cloning and primary structure of a new metalloproteinase inhibitor (ChIMP-3) produced by chicken embryo fibroblasts. ChIMP-3, formerly called the 21-kDa protein, is one of five ChIMPs (Chicken Inhibitor of MetalloProteinases). In this paper, we report that of the three most abundant ChIMPs, ChIMP-3 and ChIMP-a are extracellular matrix components, whereas ChIMP-2 is found in the media conditioned by the cells. Treatment of ChIMP-3 and ChIMP-a with N-glycosidase-F indicates that ChIMP-a is N-glycosylated whereas ChIMP-3 is not. The deduced amino acid sequence of ChIMP-3 predicts a protein whose properties are consistent with experimental measurements. Analysis of sequence alignments with the two previously described members of the TIMP (tissue inhibitor of metalloproteinases) family, TIMP-1 and TIMP-2, from various species indicates that ChIMP-3 is a related but distinct protein. This conclusion is supported by lack of significant binding with anti-TIMP-1 and anti-TIMP-2 antibodies. Based on these data, its unusual solubility properties, and its exclusive location in the matrix, we propose that ChIMP-3 is a new member of this family of metalloproteinase inhibitors, a TIMP-3.

Previously, we reported a 21-kDa ECM protein from chicken embryo fibroblasts whose synthesis was stimulated during the early stages of transformation initiated by Rous sarcoma virus (Blenis and Hawkes, 1983). Synthesis of this protein was also stimulated by treatment of normal, uninfected cells with the tumor promoter phorbol myristate acetate (Blenis and Hawkes, 1984). These observations implicated the protein in the development of transformation. Recently, electrophoretic purification and partial sequence analysis strongly suggested that the 21-kDa protein is a member of the family of metalloproteinase inhibitors which includes TIMP. Furthermore, the purified 21-kDa protein displayed inhibitor activity characteristic of these molecules (Staskus et al., 1991). Based on a number of criteria, including NH2-terminal sequence, statistical analysis of amino acid composition, size, and apparent lack of glycosylation, we proposed that the 21-kDa protein was either a variant of TIMP-1 or a third member of the TIMP family (Staskus et al., 1991).
In this paper, we report that the 21-kDa protein is one of five ChIMPs (Chicken inhibitor of MetalloProteinases) produced by chickenembryo fibroblasts. Of thethree most abundant and well characterized inhibitors, ChIMP-3 (the 21-kDa protein) and ChIMP-a are ECM components and ChIMP-2 is found in the media conditioned by cells. ChIMP-b and ChIMP-c are minor inhibitors of the ECM and media, respectively. Here we describe cDNA clones which include the entire coding region of ChIMP-3 (212 amino acids). A total of 886 nucleotides have been determined including 5' and 3' noncoding sequences. Biochemical data substantiate the lack of N-glycosylation of ChIMP-3. From the comparison of this sequence with the sequences of TIMP-1 and TIMP-2, and other supporting data, we propose that ChIMP-3 is a third member of this family, a TIMP-3.

Cell Culture and Preparation of ECM and Conditioned Media-
Chicken embryo fibroblasts were prepared and cultured as described 1732 1 (Blenis and Hawkes, 1983), with the exceptions of storage in liquid nitrogen at the primary stage and seeding of tertiary cells at 2 X lo6/ 100-mm culture dish (Falcon Labware). Cells were infected as secondary cultures with the temperature-sensitive mutant of Rous sarcoma virus, LA24, clone G2. For preparation of ECM, cells were maintained at 41 "C for 12-15 h before transfer to the permissive temperature for transformation (35 "C). Ten hours after temperature shift the ECM was harvested, basically as described previously (Blenis and Hawkes, 1983). The transforming cells were removed from the culture dishes following a 15-min incubation in Ca2'-, M e -f r e e phosphate-buffered saline, containing 5 mM EGTA, 1 mM phenylmethylsulfonyl fluoride, and 1 mg/ml t-amino-n-caproic acid (pH 7.4). After several rinses in phosphate-buffered saline and water, the ECM was solubilized in a small volume of electrophoresis sample buffer without reducing agent (Laemmli, 1970). Conditioned media samples were prepared from LA24-infected cultures which had been maintained at the permissive temperature from the time of seeding. After 38 h the medium was removed, the cell monolayer was rinsed twice with phosphate-buffered saline and the cells cultured for an additional 22 h in serum-free medium. This conditioned medium was removed from the cells, centrifuged at 10,000 x g for 20 min and analyzed by protease/substrate gel electrophoresis.
Protease/Substrate Gel Electrophoresis-Samples of ECM and conditioned media were assayed for metalloproteinase inhibitor activity by electrophoresis on protease/substrate gels as described earlier (Staskus et al., 1991). Basically, the substrate gel technique (Heussen and Dowdle, 1980;Herron et al., 1986) was modified to include a source of metalloproteinase in addition to gelatin, during polymeri- Treatment of Proteins with N-Glycosidase-F-The ECM harvested from 20 culture dishes was used to isolate ChIMP-3 and ChIMP-a. Unreduced material was first isolated from preparative polyacrylamide gels by cutting bands from the gel lanes after staining with cetyltrimethylammonium bromide to visualize ChIMP-3 (Staskus et al., 1991). As ChIMP-a could not be clearly located using this technique, several narrow bands were cut behind ChIMP-3 in the gel lanes, one of which contained ChIMP-a. After electroelution of samples (Staskus et al., 1991), without reducing agent, small aliquots were electrophoresed a second time on protease/substrate gels to identify those band eluates which contained separated ChIMP-3 or ChIMP-a activities. These two samples were dialyzed extensively against 20 mM sodium phosphate, pH 6.8, containing 0.01% (w/v) SDS at 4 "C using Spectropor tubing ( M , 3500 cut-off, Spectrum Diagnostics).
When necessary, sample volumes were reduced by coating the dialysis tubing with carboxymethylcellulose (Aquacide, Calbiochem) to absorb water. The samples were then dialyzed for 2 days against 0.2 M sodium phosphate, pH 8.0, containing 10 mM EDTA, 0.05% (w/v) SDS. To promote thorough exchange of SDS, some samples of electroeluted ChIMP-3 and ChIMP-a were also dialyzed against 6 M urea containing 0.1% (w/v) Brij 35, before dialysis against the phosphate buffers, as described above. To 38 p1 of each sample was added 2 pl of 10% (w/v) octyl glucoside in the phosphate buffer, for a final glucoside concentration of 0.5% (w/v). Each sample was split into 20-p1 aliquots. One aliquot was incubated with 0.5 units of recombinant N-glycosidase-F (N-glycanase, Genzyme Corp.) added in a volume of 2 pl. The samples were incubated overnight at 35 "C and then electrophoresed on protease/substrate gels.
Oligonucleotide Synthesis-Oligonucleotide primers for PCR were synthesized on a Milligen Biosearch automated DNA synthesizer (model 8600; Biosearch) and purified by 7 M urea, polyacrylamide gel electrophoresis. The oligonucleotide bands were visualized by UV shadow casting, excised and electroeluted by use of an Elutrap apparatus (Schleicher & Schuell), and desalted on Sep-Pak (Millipore Corp.) using standard protocols. Primer I was designed to bind to the noncoding strand of ChIMP-3 cDNA corresponding to the NH2terminal amino acids 6-11 of the mature protein (IHPQDA, Staskus et al., 1991) (nucleotides 200-216, underlined in Fig. 3). This sequence was chosen for the relatively small codon degeneracy (96-fold) and for the least possible sequence similarity to other TIMPs. The sequence of the 24-base primer consists of 17 bases specific for ChIMP-3, a 6-base Sal1 restriction site, and an extra base at the 5' end (underlined), as follows: 5'-GGTCGACATA(or C or T)CAC(or T)CCA(or C or G or T)CAA(or G)GAC(or T)GC-3'. Primer 11, which incorporates an XbaI restriction site and an extra base at the 5' end (underlined), was designed to bind to polyadenylated sequences. Primer I1 has the following sequence: 3'-(T),,AGATCTC-5'. These primers contained restriction sites for use as an alternative to blunt end ligation for subsequent cloning. Two additional oligonucleotides (22-mers) were designed to screen the Xgtll library. These primers flank the Xgtll EcoRI cloning site. Primer 111, the 5' Xgtll oligonucleotide, is upstream of the cloning site: 5"GGTGGCGAC-GACTCCTGGAGCC-3' and primer IV, the 3' X g t l l oligonucleotide, is on the complementary strand downstream of the cloning site: 3'-GTAATGGTCAACCAGACCACAG-5'. Primer V: 5"TGCTC-TCCAACTTCGGCCACT-3' was designed on the basis of partial sequence information on ChIMP-3 and corresponds to nucleotides 618-638 of ChIMP-3 cDNA. It was used to clone the 3' end of ChIMP-3. Primer VI was designed to anneal to the coding sequence of ChIMP-3 (nucleotides 640-663) in order to characterize the 5' end of ChIMP-3 cDNA. Its sequence: 3"TCCTGTGGTTCGCTTCG-TGATACGGACGTCAC-5', includes 24 nucleotides specific for ChIMP-3, one PstI restriction site, and two extra bases at the 5' end (underlined).
PCR (Polymerase Chain Reaction) and Cloning-A Xgtll cDNA library derived from 10-day old chicken embryos was purchased from Clontech. The library was screened using the PCR as described by Friedman et al. (1990). Initially, primer I, coding for IHPQDA, was used along with primer 11, modified oligo(dT) sequence, on 1 pl of the library (1-9 X lo9 phages/ml). The PCR was run in a DNA thermal cycler (Perkin-Elmer Cetus) for 30 cycles. Each cycle consisted of heating at 98 "C for 1 s, annealing at 50 "C for 15 s, and polymerization at 60 "C for 4 min. This reaction consistently yielded a single 483base pair product (P483) detected on a 1% agarose gel representing a partial ChIMP-3 cDNA. After treatment with the Klenow enzyme, P483 was cloned into the HincII site of pUC19 (GIBCO/BRL) as a blunt-end fragment. Six independent subclones were selected. After sequence analysis of P483, a specific primer (V) was designed to determine the 3'-end sequence of the cDNA. This primer along with primer I11 or IV (the 5' or 3' Xgtll oligonucleotides described above) were used to amplify cDNA from the Xgtll library. The PCR was run for 30 cycles. Each cycle consisted of heating at 94 "C for 30 s, annealing at 61 "C for 2 min, and polymerization at 72 "C for 5 min. The resulting single 269-base pair fragment (P269) was cloned as described before in pUC19. Two independent subclones were analyzed.
A 32P-labeled probe, P483, was generated by PCR and end labeling for subsequent screening of the Xgtll library. The PCR was run using primer I only (see "Results") for 30 cycles with each cycle consisting of heating at 94 "C for 30 s, annealing at 61 "C for 2 min, and polymerization at 72 "C for 5 min. The resulting product was gelpurified and labeled with [(u-~*P]~CTP using the multiprimer labeling system from Amersham Corp. Incorporated nucleotides were separated from unincorporated nucleotides on a Sephadex (2-50 column (Boehringer Mannheim).
Screening the cDNA Library-Approximately lo6 phages were grown on six 150-mm plates, lifted in duplicate onto supported nitrocellulose transfer membrane (BAS-NC from Schleicher & Schuell), and hybridized to the partial cDNA probe, 32P-labeled P483, described above. Hybridizations were performed overnight at 42 "C in 5 X Denhardt's solution (Denhardt, 1966), 5 X SSC (SSC is 15 mM sodium citrate, 150 mM NaCl), 50 mM sodium phosphate (pH 6.5), 0.1% SDS (w/v), 250 pg/ml fish sperm DNA, 50% deionized formamide, 1% dextran sulfate (w/v). The filters were washed in 0.1 X SSC containing 0.1% SDS (w/v) at 60 "C. Five positively hybridizing plaques were purified. Two independent clones, C and D, were chosen. After extraction and purification, the DNA was analyzed by PCR using both 5' and 3' Xgtll oligonucleotides (primers I11 and IV). The phage DNA from clone C was amplified by PCR after the first round to characterize the 5' end of ChIMP-3 cDNA. The amplified DNA of purification using primer VI and either primer I11 or IV in order from clones C and D was subsequently cloned into pUC19, as described before, and sequenced. Standard protocols for cDNA library screening, X phage purification, agarose gel electrophoresis, and plasmid cloning were employed (Maniatis et al., 1982).
DNA Sequencing-Double-stranded cDNA cloned into pUC19 was sequenced by the dideoxy terminator method (Sanger et al., 1977) using sequencing kits purchased from Pharmacia LKB Biotechnology Inc. or U. S. Biochemical Corp. (Sequenase version 2.0). Each cDNA subclone was sequenced using an M13 universal primer, a reverse sequencing primer (Pharmacia LKB), or internal primers. The sequencing strategy used is presented in Fig. 4. In all cases, both strands were analyzed in order to confirm the sequence results. For this purpose, in some cases, smaller fragments were subcloned into pUC19 using the restriction sites indicated in Fig. 4. At least two independent subclones were sequenced to identify possible errors caused by PCR.
Sequence Analysis-DNA and deduced amino acid sequence analyses were performed using the EuGene Sequence Analysis Package from the MBIR Molecular Biology Information Resource, Department of Cell Biology, Baylor College of Medicine. Comparison of the deduced amino acid sequence of ChIMP-3 with other sequences in the data bank was performed using the Pattern-Induced Multisequence Alignment (PIMA) algorithm of Smith (1990, 1992) which employs secondary structure-dependent gap penalties for comparative protein modeling.

RESULTS
Analysis of cells, ECM, and conditioned media from cultures of normal and transforming chicken embryo fibroblasts by protease/substrate gel electrophoresis indicates five distinct IMP activities which range in size from M, -20,000 to M, -28,000.' Until we have firmly established the relationship of these proteins to one another and to the two existing members of the TIMP family we propose to call them ChIMP-2 and -3 and ChIMP-a through -c. In this paper we report that the three most abundant are ChIMP-2, ChIMP-3, and ChIMP-a. ChIMP-2 is a major inhibitor found in the conditioned media and ChIMP-3 and ChIMP-a are found in the ECM (Fig. 1). ChIMP-b and ChIMP-c are minor inhibitors detected in the ECM and media, respectively.' ChIMP-3 was formerly called the 21-kDa protein and was studied because its synthesis is stimulated during oncogenic transformation. Early characterization of this protein indicated that it was probably not N-glycosylated (Blenis and Hawkes, 1983). To examine this question further, isolated ChIMP-3 was incubated with N-glycosidase-F and then analyzed by protease/substrate gel electrophoresis. As shown in Fig. 2 (left panel), the electrophoretic mobility of ChIMP-3 was not altered by treatment with N-glycosidase-F. Under identical conditions, the apparent relative mass of ChIMP-a was decreased by approximately 5 kDa after enzyme treatment. These data support our earlier proposal that ChIMP-3 is not N-glycosylated whereas ChIMP-a is an N-glycoprotein and, like TIMP-1 (Stricklin, 1986), does not require carbohydrate for its activity. ChIMP-3 and deglycosylated ChIMPa have slightly different relative mobilities in SDS-polyacrylamide gels, with ChIMP-3 migrating slower than ChIMP-a ( Fig. 2, right panel). Thus ChIMP-3 appears to differ from ChIMP-a by more than just the absence of N-linked carbohydrate, suggesting that ChIMP-3 is a distinct protein.
We have previously determined the primary structure of the NH2 terminus of ChIMP-3 by direct amino acid sequencing (Staskus et al., 1991). This information was used to prepare a mixture of synthetic oligonucleotides (primer I) coding for amino acids 6-11, IHPQDA (numbering of residues from the NH, terminus of the mature protein, Staskus et al. (1991)). This sequence was chosen for the relatively low degeneracy of its codons and because it exhibits the most differences with published sequences of TIMP-1 and TIMP-2. Primer I and an oligo(dT) primer (11) were used in a PCR on the Xgtll chicken cDNA library. A single 483-base pair product (P483) resulted. This was cloned into pUC19 and sequenced. Sequence analysis of P483 revealed that only primer I was used in the amplification of partial ChIMP-3 cDNA (nucleotides 200-682, Fig. 3). This primer, designed to anneal to the noncoding strand of the sequence (nucleotides 200-216, Fig. 3), was also able to anneal to the coding strand (nucleotides 666-689, Fig. 3) with a total of six mismatches, of which two were in the sequence specific for ChIMP-3 ( denotes treatment with N-glycosidase-F (-) indicates no treatment. Standards are the same as those indicated in the legend to Fig. 1. 3). There was no evidence of a PCR product containing a poly(A) tail resulting from the annealing of the oligo(dT) primer to the cDNA. From this we conclude that primer I1 did not take part in the PCR.
On the basis of the sequence of P483 another primer (V) was synthesized in order to determine the 3' end sequence of the cDNA. This primer was used with an oligonucleotide representing sequences of the flanking EcoRI cloning site (5' or 3') of Xgtll in a PCR on the Xgtll chicken cDNA library. This reaction resulted in the amplification of a single 269base pair product (P269), which was cloned into pUC19 and sequenced. P269 represents the nucleotides 618-886 (Fig. 3). The 5' end sequence along with the complete sequence of ChIMP-3 was determined by screening the chicken cDNA library using 3'P-labeled P483 as a probe. A total of lo5 independent plaques were screened. Five positive clones were isolated and two analyzed (clones C and D). A composite of c GCG ilcA CAS iicc ccc TC7 6 / / C CI/C cc/J ccc A X clic :x LGA 4 3 ACA GGC GAG CCT CCA CTT iicc cclj iicli C I u LXG ccc CTC IIG CTC Be C b A ccc c*c CCC ccc C ' I GC' ACC au AL' CC' I " CTC SSL TTC 131 m%t C n r ala "P ,e> IlY phe 7 :Tc ccc CTC TIC CTC TCC ACC ICC iicc CIC ccc ilC CTC iTG scc :,e le" a l a "dl phe le" cy, ser t r p 9Fr le" arg a s p lei "a, a l a 22 5a.C ccc TSC *CT TCC CTC ccc ULLAcCcT. CAC G*C CEG rl; TCC 223 g," ala 0,. t h r c y r "a1 pro Ile h l , pro q l n a l p a:a Ph* CY'

3-
Iuc TCC G*C ATC CTC a x CCT CCT -CTT CTC YCC Aac Iu$ I T C * L e asp lis "a, t,e arq ala lys "a1 "a, gly lys lY' 1 P 1 52 Nucleotide sequence and deduced amino acid sequence of ChIMP-3. The first nucleotide presented in the sequence followed the EcoRI cloning site of Xgtll (not shown) and is referred to as nucleotide 1. The initiation codon (ATG) and the terminal codon (TGA) are underlined. The nucleotide sequences corresponding to primers I (200-216), V (618-638) and VI (640-663) and the amino acid sequence corresponding to a potential glycosylation site (208-210) are also underlined. The first cysteine (amino acid 25) of the mature protein is outlined. The nucleotides which represented mismatches with primer I in the carboxyl-terminal part of the sequence, as discussed under "Results," have been marked above the nucleotides by an asterisk (*) for the sequences specific for ChIMP-3 and by a plus sign (+) for the Sal1 site of the primer. the physical map of the cDNA clones is shown in Fig. 4. One clone (D) was missing the first 13 nucleotides and nucleotides 18 and 19. We speculate that these deletions, which are very close to the EcoRI cloning site may result from cDNA cloning in Xgtll.
The composite nucleotide sequence of ChIMP-3 is shown in Fig. 3. The full-length cDNA contains 886 nucleotides. The first ATG appears at nucleotide 113 and is followed by a long open reading frame of 636 nucleotides. This open reading frame was the largest one observed and by comparison to the other TIMP sequences is the most likely candidate to initiate translation. The longest sequence derived from 5' noncoding sequences is composed of 112 base pairs. The 0.636-kilobase pair coding sequence encodes pro-ChIMP-3 of 212 amino acids. This protein sequence includes a 24-residue signal peptide and a mature ChIMP-3 of 188 amino acids. One potential N-glycosylation motif, Asn-X-Thr (Beeley, 1985), is present in the carboxyl-terminal sequence at position 208-210 (X = Ala).
The 138-nucleotide 3"untranslated region does not contain a polyadenylation signal, and no poly(A) sequence was found. Therefore, we conclude that either our cDNA clones are truncated or that the ChIMP-3 mRNA does not have a poly(A) tail. The cDNA library (Clontech) was prepared by use of an oligo(dT) primer to select for polyadenylated RNA. The vast majority of cellular mRNAs are polyadenylated, but some exceptions have been described (Brawerman, 1981). A search of nucleotide sequences in GenBank/EMBL for the sequence AATGAAA (nucleotides 840-846 in ChIMP-3), which we note is a variant of a poly(A) signal, indicated that it is also present at the 3' end of human (Docherty et al., 1985;Carmichael et al., 1986) and rabbit (Horowitz et al., 1989) TIMP-1 cDNA and precedes the putative polyadenylation consensus sequence signal, AATAAA.
A comparison of the deduced amino acid sequence of ChIMP-3 with sequences of TIMP-related proteins in the GenBank/EMBL is shown in Fig. 5. Analysis of these data indicates that 52 amino acid residues are conserved among all the proteins listed. This represents -25% sequence identity. There is -28% identity between a consensus sequence for the TIMP-1 proteins and the ChIMP-3 sequence and a closer agreement (-42% identity) between a consensus sequence for the TIMP-2 proteins and ChIMP-3. These calculations do not take into account the significance of any gaps in the alignments. In this respect, in the region of residues 58-64 in TIMP-2 (where there is a GNDIYGN insert) ChIMP-3 is more like TIMP-1. On the other hand, between residues 83-84 of TIMP-2 where there is a &residue insert (AXXXA) in the TIMP-1 sequences and at the carboxyl-terminal end, where there is an extended sequence, ChIMP-3 is more like TIMP-2. In 93 of 212 positions ChIMP-3 is unique. The region of greatest similarity is at the amino-terminal end of the mature proteins where ChIMP-3 shares 18/25 (72%) and 17/25 (68%) identical residues with the consensus sequences for TIMP-1 and TIMP-2, respectively. TIMP-1 is proposed to contain two sites of N-linked glycosylation. By comparison with TIMP-1 these would be expected at amino acid residues 54-56 and 99-101 in ChIMP-3. The corresponding sequences in ChIMP-3 do not contain N-glycosylation motifs.

DISCUSSION
The NH,-terminal amino acid sequence obtained for ChIMP-3 (Staskus et al., 1991) was used to design primers for the isolation of cDNA clones of ChIMP-3. The nucleotide sequence codes for a mature protein of 188 amino acids, of which 27 residues (residues 26-53, Fig. 3) are identical to the NH2-terminal amino acid sequence measured directly by chemical methods (Staskus et al., 1991). One exception is residue 52, which was determined to be Leu instead of Ile; however, this assignment was not made at full confidence originally (Staskus et al., 1991). The deduced amino acid FIG. 5. Comparison of the deduced amino acid sequences of ChIMP-3 and TIMPs. Sequences were compared by the Pattern-Induced Multisequence Alignment program (PIMA) of Smith (1990, 1992). Boldface letters indicate conserved amino acids; lowercase letters indicate amino acids conserved between TIMP-2 proteins and ChIMP-3; underlined letters indicate amino acids conserved between TIMP-1 proteins and ChIMP-3. Dots (. . .) represent amino acids unique to ChIMP-3; asterisks (***) represent putative TIMP-1 glycosylation sites; the bracket ([) indicates the beginning of the mature proteins. Footnotes: a, the murine IMP sequence represents the 16C8 (Edwards et al., 1986) and the TPA-S1 (Johnson et  al., 1987) sequences, which are identical. b, Gewert et al. (1987); c, Freudenstein et al. (1990); d, Docherty et al. (1985), Gasson et al. (1985), Carmichael et al. (1986); e, Horowitz et al. (1989); f, Boone et al. (1990); g, Boone et al. (1990), Stetler-Stevenson et al. (1990). Da. This is in close agreement with estimations of the mass of the pure protein as determined by SDS-polyacrylamide gel electrophoresis: 22,000 under reducing conditions (Staskus et al., 1991). The sequence indicates the presence of 12 cysteines which are conserved relative to other members of the TIMP family and substantiates electrophoretic data which had indicated the presence of disulfide bonds in ChIMP-3.3 The deduced amino acid composition predicts a very basic protein which is consistent with an estimated PI > 9.0.3 Thus, on the basis of NH2-terminal amino acid sequence, size, potential disulfide bonding, and PI, we conclude that the nucleotide sequence presented in Fig. 3 encodes ChIMP-3.
TIMP-1 is a glycoprotein (Mr -28,000) with two proposed oligosaccharide chains (Murphy and Werb, 1985;Carmichael et al., 1986;Stricklin, 1986) whereas proteins in the TIMP-2 group are smaller with M, -18,000-23,000 under nonreducing conditions (Murray et al., 1986;De Clerck et al., 1989;Stetler-Stevenson et al., 1989) and based on their inability to bind concanavalin A are proposed to be unglycosylated (Murray et al., 1986;De Clerck et al., 1989). ChIMP-3 exhibits metalloproteinase inhibitor activity (Staskus et al., 1991) and in terms of molecular mass and apparent lack of glycosylation it resembles TIMP-2. Polyclonal antibodies to either human TIMP-1 or human TIMP-2 do not demonstrate significant binding to ChIMP-3 and antibody to ChIMP-3 does not bind to human TIMP-1 or TIMP-2 (Staskus et al., 1991). It is clear that ChIMP-3 is related to but distinct from the groups of TIMP-1 and TIMP-2 proteins whose deduced amino acid sequences are compared in Fig. 5. In common with other TIMPs, the ChIMP-3 sequence predicts a hydrophobic leader sequence consistent with it being a secreted protein and 12 cysteine residues which are conserved among all members of this family. The mature ChIMP-3 (188 amino acids) is intermediate in size between TIMP-1 and TIMP-2 (181-184 and 196 amino acids, respectively). As discussed under "Results," the region of greatest similarity for all TIMPs and ChIMP-3 is at the NH2 terminus of the mature protein. Indeed, Murphy et al. (1991b) have shown that the activity of TIMP-1 resides in the NH2-terminal half of the protein. Some features of the ChIMP-3 sequence resemble the TIMP-1 sequences, others resemble the TIMP-2 sequences, and -44% of the amino acid residues of ChIMP-3 are unique.
The mechanism of inhibition of MMPs by their inhibitors has not yet been elucidated. However, Woessner (1991) has noted two fairly conserved regions of TIMP-1 and TIMP-2 where a negatively charged residue precedes several hydrophobic residues in the primary structure. He has speculated that binding of the hydrophobic amino acids to the Sl'-S2'-S3' region of MMPs would place the negatively charged residue in a position to interact with the zinc and render the enzymes inactive. The first sequence, beginning at Asp-16 of mature TIMP-1 and TIMP-2, is also conserved in ChIMP-3; however, the second, beginning at Glu-82, is not conserved in the chicken protein. The Glu is replaced by Gln in ChIMP-3. Thus, if such a mechanism is operative in the inhibition of MMPs by TIMPs, our data support the potential role of Asp-16 rather than Glu-82 in this process.
The two sites of potential N-linked glycosylation in TIMP-1 (Docherty et al., 1985;Carmichael et al., 1986) are not found in the ChIMP-3 sequence. The lack of incorporation of D-[2-3H]mannose into ChIMP-3 (Blenis and Hawkes, 1983) and the absence of any change in the electrophoretic mobility of ChIMP-3 treated with N-glycosidase-F or synthesized by tunicamycin-treated cells: indicate that ChIMP-3 is probably not N-glycosylated. Although one potential site of N-linked glycosylation is apparent in the deduced amino acid sequence at the carboxyl terminus of ChIMP-3, not every sequence of this type is glycosylated in secreted proteins (Beeley, 1985). The carbohydrate cannot contribute significantly to the molecular mass of the protein because estimates of the molecular weight of ChIMP-3 determined by electrophoretic analysis and calculation from the deduced amino acid sequence show a close correspondence, as indicated under "Results." Previously, we proposed, on the basis of amino acid composition data, that ChIMP-3 was more closely related to members of the TIMP-1 group of proteins than the TIMP-2 group (Staskus et al., 1991). This analysis, and additional C. J. Henrich and S. P. Hawkes, unpublished data.

Chicken Metalloproteinase Inhibitor:
ChIMP-3 data, led us to speculate that ChIMP-3 was either a variant of TIMP-1 or a third, new member of the TIMP family. However, the deduced amino acid sequence, which is presumably more accurate than chemical measurements of amino acid composition, indicates that ChIMP-3 is more like TIMP-2 than TIMP-1, but clearly a distinct protein. ChIMP-3 has 42% identity with a consensus sequence of TIMP-2 and 28% identity with a consensus of TIMP-1. The only TIMP-2 sequences reported so far (human and bovine) are highly conserved (94% identity). One could reasonably argue that TIMP-2 from mammalian species would not be expected to share the same degree of identity with an avian protein.
However, we have recently purified another protein, ChIMP-2, which is undoubtedly the chicken equivalent of TIMP-2. This conclusion is based upon a number of criteria including its co-purification in a complex with pro-MMP-2, NHp-terminal amino acid sequence, amino acid composition and binding with anti-TIMP-2 antibody.2 Thus we propose that ChIMP-3 is neither TIMP-1 or TIMP-2 but a new member of the TIMP family, TIMP-3. Although in some respects ChIMP-3 is similar to TIMP-1 and TIMP-2, in others it is unique. In particular, ChIMP-3 is relatively insoluble (Blenis and Hawkes, 1984) and found exclusively in the ECM (Staskus et al., 1991), unlike most members of the TIMP-1 and TIMP-2 groups of proteins that are isolated from tissue fluids or cell culture media. In contrast to TIMP-2 (and ChIMP-2) ChIMP-3 is not isolated in a complex with a particular pro-MMP.
Since most of the chicken MMPs have not yet been purified, the inhibitory specificity of ChIMP-3 remains to be determined. We have not yet characterized a TIMP-1 from the avian system. One possible candidate for this assignment is ChIMP-a, which is clearly an N-glycosylated metalloproteinase inhibitor. Relative to ChIMP-3, ChIMP-a is a minor activity in the ECM, and its possible localization in conditioned media has not been rigorously examined.
Evidence for the existence of a family of TIMP-like proteins has been accumulating for a number of years. Only in two species (human and bovine) have both TIMP-1 and TIMP-2 been completely sequenced. Three metalloproteinase inhibitors have been reported in rabbit brain capillary endothelial cells: TIMP-1 (Mr = 30,000), IMP-1 (Mr = 22,000), and IMP-2 (Mr = 19,000) (Herron et al., 1986). The same three inhibitors and an additional IMP-3 ( M , = 16, 500) have been detected in human glioma cell lines (Apodaca et al., 1990). We originally reported four (Staskus and Hawkes, 1989) and now five inhibitor activities in chicken embryo fibroblasts. Recently, Chen et al. (1991) reported the purification of a chicken 70-kDa gelatinase associated with a 22-kDa protein.
Based on this association the protein was assumed to be TIMP-2. Craig et al. (1991) have reported the detection of three discrete metalloproteinase inhibitors in the culture media conditioned by 11-day chick tibiae. Since two of these activities were not produced by skin fibroblasts or cultures of calvariae, the relationship of these proteins to the five Ch-IMPS discussed above remains to be clarified. The third inhibitor (M, -23,000), which was not characterized further, was assumed to be chicken TIMP-2.
As discussed earlier we propose to call the chicken inhibitors ChIMPs (Chicken inhibitor of MetalloEroteinases). A letter designation (ChIMP-a, -b, and -c) will be used for those whose relationship to other ChIMPs and TIMPs remains to be clarified. The letter will be converted to a number desig-nation (as for ChIMP-2 and ChIMP-3) when the identity of the inhibitor is established. In summary, based on its primary structure and biochemical properties, we propose that ChIMP-3 is a new, matrix-specific TIMP-3. It is of particular interest because its synthesis is stimulated during oncogenic transformation and purified ChIMP-3 has been shown to promote some of the phenotypic properties of transformed cells Hawkes, 1989, 1992).