A New Family of 10 Murine Ovalbumin Serpins Includes Two Homologs of Proteinase Inhibitor 8 and Two Homologs of the Granzyme B Inhibitor (Proteinase Inhibitor 9)*

Serine proteinase inhibitors (serpins) are classically regulators of extracellular proteolysis, however, recent evidence suggests that some function intracellularly. Such “ovalbumin” serpins include the human proteinase inhibitors 6 (PI-6), 8 (PI-8), and 9 (PI-9), plasminogen activator inhibitor 2, and the monocyte/neutrophil elastase inhibitor. PI-9 is a potent granzyme B (graB) inhibitor that has an unusual P1 Glu and is present primarily in lymphocytes. In a search for the murine equivalent of PI-9 we screened cDNA libraries, and performed reverse transcriptase-polymerase chain reaction on RNA isolated from leukocyte cell lines and from lymph nodes and spleens of allo-immunized mice. We identified 10 new ovalbumin serpin sequences: two resemble PI-8, two resemble PI-9, and the remaining six have no obvious human counterparts. By RNA analysis only one of the two sequences resembling PI-9 (designated SPI6) is present in mouse lymphocytes while the other (a partial clone designated mBM2A) is predominantly in testis. SPI6 comprises a 1.8-kilobase cDNA encoding a 374-amino acid polypeptide that is 68% identical to PI-9. mBM2A is 65% identical to PI-9 and over 80% identical to SPI6. Although the reactive loops of SPI6 and mBM2A differ from PI-9, both contain a Glu in a region likely to contain the P1-P1′ bond. SPI6 produced in vitro using a coupled transcription/translation system formed an SDS-stable complex with human graB and did not interact with trypsin, chymotrypsin, leukocyte elastase, pancreatic elastase, thrombin, or cathepsin G. Recombinant SPI6 produced in a yeast expression system was used to examine the interaction with human graB in more detail. The second-order rate constant for the interaction was estimated as 8 × 104 m −1 s−1, and inhibition depended on the Glu in the SPI6 reactive center. The SPI6 gene was mapped to the same region on mouse chromosome 13 asSpi3, which encodes the murine homolog of PI-6. We conclude that even though their reactive centers are not highly conserved, SPI6 is a functional homolog of PI-9, and that the regulation of graB in the mouse may involve a second serpin encoded by mBM2A. Our identification of multiple sequence homologs of PI-8 and PI-9, and six new ovalbumin serpins, is consonant with the idea that the larger set of granule and other proteinases known to exist in the mouse (compared with human) is balanced by a larger array of serpins.

Serine proteinase inhibitors (serpins) form a superfamily of proteins that resemble ␣ 1 -proteinase inhibitor in structure (1). The majority regulate proteinase activity in extracellular processes such as fibrinolysis, blood coagulation, complement activation, and tissue remodeling (2). Serpins generally contain an exposed C-terminal loop in which the reactive center P 1 -P 1 Ј residues define the inhibitory specificity of the molecule. Inhibition of proteolysis is achieved by the formation of essentially irreversible 1:1 stoichiometric complexes between serpins and cognate proteinases (2,3).
The physiological roles of most ovalbumin serpins are unknown but MNEI, maspin, PAI-2, and SCCA may function in tumorigenesis and inflammation. MNEI has been postulated to control serine proteinases found at inflammatory sites to prevent damage to surrounding tissue (22), and maspin is a can-didate tumor suppressor (9). Increases in SCCA levels and release are associated with squamous cell carcinoma (7), while PAI-2 expression is increased in monocytes during inflammation (23).
We have recently demonstrated that PI-9 is a potent inhibitor of the cytotoxic lymphocyte granule serine proteinase, granzyme B (graB), and have suggested that PI-9 is produced by cytotoxic lymphocytes and other immune cells to counter misdirected graB released during target cell destruction (14). To investigate the physiological role of PI-9 we have decided to study it in the mouse, which has obvious advantages for studies involving the immune system.
Here we report the identification of 10 new serpin sequences from the mouse arising from a search for a PI-9 homolog. Among these are two that are probably PI-8 homologs and two that closely resemble human PI-9. The remainder represent ovalbumin serpins that have the hallmarks of functional proteinase inhibitors but have no known human counterparts. Eight of these new serpin sequences (including one of the two PI-9 homologs, designated SPI6) were found in activated lymphocytes. SPI6 was characterized further and shown to be a graB inhibitor. These results indicate that rodents have a much larger array of ovalbumin serpins than humans, which probably serves to balance a larger set of proteinases.

EXPERIMENTAL PROCEDURES
General Methods-General recombinant DNA methods were performed according to Ref. 24. cDNA probes were labeled with [␣-32 P]dATP (DuPont NEN) using the random hexamer method (Prime-A-Gene System, Promega).
Cells and Tissues-The mouse cytotoxic T cell line R8 has been described (25). Mouse allo-immunizations, isolation of spleen and lymph nodes, and culture of isolated spleen and lymph node cells in interleukin-2 containing media were as described (26).
Identification of New Serpins-Degenerate primers PB117 and PB118 designed to conserved sequences flanking the PI-6 and ovalbumin serpin reactive centers were used as described (14). Total RNA was isolated from cells and tissues as described (27). cDNA was synthesized from the RNA using Moloney murine leukemia virus reverse transcriptase and oligo(dT) (24), and amplification and cloning of products as described (14). Candidate serpin sequences were identified by the presence of the hinge region motif (Gly-Thr-Glu-Ala-Ala-Ala-Ala(Thr/ Ser)) (2) and a conserved Phe-Cys-Ala-Asp sequence not encoded by the primers.
cDNA Isolation and Sequencing-The SPI6 cDNA clone was isolated from a mouse day 15 embryo library in gt10 (CLONTECH). The mBM2A and mBM17 clones were isolated from a mouse bone marrow library in gt11 (CLONTECH). Approximately 4 ϫ 10 5 plaques were screened in each case at low stringency using a partial cDNA clone of human PI-9 (14). Membranes were hybridized at 32°C in 50% formamide, 5 ϫ SSPE, 0.5% SDS, and 5 ϫ Denhardt's, then washed in 2 ϫ SSC and 0.1% SDS at 45°C. Positive plaques were purified and the EcoRI inserts were subcloned into the EcoRI site of pUC118 or pBluescript II KS Ϫ (Stratagene). Both strands of each cDNA were sequenced using the Sequenase II system (Amersham) either on overlapping clones generated by directed exonuclease III deletion (Erase-a-Base system, Promega), or by an oligonucleotide walk strategy. Sequences were assembled and analyzed using the DNASIS programs (Hitachi).
Preparation of SPI6 Protein for Determination of Complex Forming Activity-The SPI6 cDNA was subcloned into the EcoRI site of pSVTf (28) so that the 5Ј end of the clone was closest to the T7 promoter. Purified plasmid DNA was linearized with SalI prior to transcription/ translation. For control production of human PI-6, pSVPTI/P (11) was linearized with SalI. Labeled protein was produced in a coupled TNT TM Wheat Germ Extract system (Promega) using T7 polymerase and incorporating [ 35 S]methionine (DuPont NEN). Translation products were analyzed by SDS-PAGE (29) or native PAGE (30) followed by fluorography (Amplify TM , Amersham). For analysis of complex formation, 2 l of the translation products were incubated at 37°C for 30 min with 2 ng of the appropriate proteinase. Purification of active graB was as described (31). Thrombin was purified as described (32), chymotrypsin, trypsin, leukocyte elastase, pancreatic elastase, and cathepsin G were purchased from Sigma.
Production of Recombinant SPI6 -The SPI6 cDNA was modified by PCR to incorporate a hexa-histidine tag immediately after the initiating methionine. The 1.8-kilobase SPI6 cDNA in gt10 was used as a template in a PCR primed by 5Ј-TCTGCCATCATGCATCATCATCATCAT-CATAATACTCTGTCTGAAGGAAATGG-3Ј (sense) and 5Ј-TTATG-GAGATGAGAACCTGCCACA-3Ј (antisense). Amplification, cloning, verification, and expression of the modified SPI6 cDNA in Pichia pastoris was as described previously for PI-6 and PI-9 (14,33). SPI6 and GraB Kinetics-For stoichiometric determinations, 10 pmol of graB was incubated with different concentrations of SPI6 at 37°C in 20 mM Hepes pH 7.4, 100 mM NaCl, 0.05% (w/v) Nonidet-P40 (34). Residual enzyme activity was determined after 15 min by a twostage assay using Boc-Ala-Ala-Asp-S-benzyl and 5,5Ј-dithiobis(nitrobenzoic acid) (35). The rate of inhibition of graB by SPI6 was determined by incubating equimolar enzyme and inhibitor at 37°C, and determining residual activity periodically (33,36). The second-order rate constant was calculated as described (33).
Site-directed Mutagenesis-The SPI6 cDNA cloned in the P. pastoris expression vector pHILD2 was mutated using the Transformer TM kit (CLONTECH). The selection primer 5Ј-CGGTGAGCATGCAGACCT-TCAAC-3Ј removes a unique XbaI in pHILD2. The mutagenic primer 5Ј-GCCATCATAGCATTTTGCTGTGCC-3Ј substitutes 337 Glu in the SPI6 sequence with Ala. Following verification by DNA sequencing, the mutated SPI6 cDNA was subcloned from pHILD2 to pSVTf for use in the coupled transcription/translation system.
RNA Analysis-Mouse multiple tissue Northern blots containing 2-g samples of mRNA from various mouse tissues (CLONTECH) were probed with a cDNA comprising the 3Ј-untranslated region of the SPI6 cDNA (nucleotides 1035-1819). Hybridization was carried out at 42°C in 5 ϫ SSPE, 10 ϫ Denhardt's, 0.1% SDS, and 50% formamide. Membranes were then washed in 0.1 ϫ SSC and 0.1% SDS at 65°C and exposed to x-ray film. The membrane was subsequently stripped and re-probed under the same conditions with the mBM2A cDNA.
Chromosomal Localization of Spi6 -The chromosomal localization of Spi6 was mapped by analysis of the Jackson Laboratory (C57BL/ B6JEi ϫ SPRET/Ei)F1 ϫ SPRET/Ei backcross DNA panel (BSS) (37). A partial SPI6 cDNA clone (nucleotides 798 -1819) generated by the Erase-a-Base system was used in Southern hybridizations to identify a PvuII polymorphism in genomic DNA between Mus musculus (C57BL/ B6JEi) and Mus spretus (SPRET/Ei). This was then used to probe a membrane containing approximately 2 g of PvuII-digested genomic DNA from each of the 94 animals in the BSS panel (21). Hybridization was in 6 ϫ SSC at 68°C followed by washing in 1 ϫ SSC at 68°C. Each DNA sample was scored for the presence or absence of the M. musculus fragment and the results were analyzed by L. Rowe at the Jackson Laboratory using the Map Manager program (38).

Identification of GraB Binding Activity and Multiple Serpin
Sequences in Mouse Immune Cells-We have previously reported that PI-9 is an inhibitor of graB, and that it is predominantly expressed in immune cells (14). To assess whether a comparable inhibitor exists in mouse cells, we incubated a cytosolic extract from R8 cells with purified human graB. R8 cells are an interleukin-2-dependent mouse cytotoxic lymphocyte line that produce granules and are reminiscent of natural killer cells (25). In many respects they resemble the human YT line which we have shown produces large quantities of graB and PI-9 (14).
As shown in Fig. 1, incubation of R8 extracts with graB led to the formation of a 67-kDa complex detected by SDS-PAGE and immunblotting with an anti-graB monoclonal antibody. This resembled the SDS-stable complex formed between graB and PI-9, which is characteristic of serine proteinase-serpin interactions (2). In addition, immunblotting with anti-PI-9 antibodies weakly detected a 42-kDa protein in R8 extracts (data not shown). From these experiments we concluded that a mouse graB inhibitor and homolog of PI-9 is present in R8 cells.
In a parallel approach we used RT-PCR to identify ovalbumin serpin sequences present in mouse immune tissue. Given its presence in activated lymphocytes (14), it is conceivable that PI-9 or its homologs are primarily produced during an immune response. To allow for this, we took cells from the popliteal lymph nodes and spleens of allo-immunized mice. Some of these lymph node and spleen cells were then cultured for 1-2 weeks in the presence of interleukin-2, which selectively activates and promotes the proliferation of cytotoxic lymphocytes (in vitro stimulated). R8 cells and a long-term (Ͼ6 months) culture of mouse spleen cells passaged in high levels of interleukin-2 were also examined. The latter cells (designated NK) exhibit natural killer-like, non-MHC restricted cytotoxic function. 2 RNA was prepared from each of these sources and used as a template for cDNA production and PCR. The PCR primers were degenerate oligonucleotides designed to ovalbumin serpin sequences that have been previously used to identify PI-9 (14). These amplify the serpin reactive center loop and candidate sequences are identified by conserved residues inside the primer-binding sites. In our hands this primer pair has amplified PAI-2, MNEI, PI-6, PI-8, PI-9, and heparin cofactor II from various human tissues. 3 As shown in Fig. 2, RT-PCR of the mouse cells and tissue identified nine serpin sequences (SPI3, SPI6, mNK10, mAT2, mR86, mNK9, mNK13, mNK21, and mNK26). One of these (SPI3) is the previously described mouse homolog of PI-6 (21). The relative frequency that these new sequences appeared is indicated in Table I. All but one (mNK26) appeared more than once, in independent clones arising from different samples. All of the new sequences possessed the serpin proximal hinge motif Gly-Thr-Glu-Ala-Ala-Ala-Ala(Thr/Ser) (39) inside the upstream-primer site, indicating that they are functional proteinase inhibitors. In addition, most possessed a Phe-Cys-Ala-Asp motif adjacent to the downstream primer site. The latter motif is found in the ovalbumin serpin subgroup that comprises PI-6, PI-8, PI-9, and MNEI (see Fig. 2).
Comparison and alignment of the mouse sequences with the human serpin sequences indicated that only one of the novel sequences (mNK10) has a clear human counterpart. This is probably a homolog of PI-8 since it shows 73% identity in the variable region between the proximal hinge and the Phe-Cys-Ala-Asp motif, and it has an identical P 1 Arg residue.
Surprisingly, sequence comparisons alone revealed no obvious homolog of PI-9. However, the second most frequent sequence arising from R8 cells after SPI3 was SPI6, and this sequence was identified in all the other samples. Although there is only 39% identity between the SPI6 and PI-9 variable regions, inspection and alignment of the SPI6 reactive center with the other serpins suggested that a Glu would be at, or close to, the P 1 -P 1 Ј bond. Since (i) a Glu at this position is a unique feature of PI-9 (14) and is crucial for graB inhibition, 4 (ii) SPI6 sequences arose most frequently in R8 cells which also contain a graB-binding function, and (iii) the other serpin (SPI3) observed in R8 cells does not interact with graB, we hypothesized that SPI6 is a functional homolog of PI-9.
Isolation and Sequencing of SPI6 cDNAs-To isolate a complete SPI6 cDNA, mouse bone marrow and day 15 embryo cDNA libraries were hybridized at low stringency to a partial human PI-9 cDNA. Isolates from the bone marrow library included several incomplete SPI3 and SPI6 clones, as well as two new serpins not already identified by RT-PCR (mBM2A, mBM17; Fig. 2). Analysis of these partial clones suggested that mBM17 is another homolog of PI-8, being 64% identical in the variable region with an identical P 1 Arg residue (and over 80% identical to mNK10). The clone mBM2A is highly related to SPI6 and probably represents a second PI-9 homolog in the mouse (see Fig. 3B and "Discussion" below).
Three clones were isolated from the mouse embryo library. Two corresponded to the characterized serpins SPI3 and mouse ␣ 1 -antitrypsin, while the third clone contained a 1.8-kilobase insert encoding SPI6. (The designation serine proteinase inhibitor 6 (SPI6; gene Spi6) has been allocated by the International Committee on Standardized Nomenclature for Mice.) Sequencing of the SPI6 cDNA predicted a protein of 374 amino acids (Fig. 3A). The coding region is flanked by a short 5Ј-untranslated region containing two termination codons (one in frame with the first methionine), and is followed by a 653-  (6), and the prototype serpin sequence, ␣ 1 -antiproteinase (1). Lower panel shows the mouse serpin sequences identified in this study. SPI3 is the mouse homolog of PI-6 (21). Reactive center bonds (P 1 -P 1 Ј) are indicated by the vertical arrow. Dashes indicate gaps introduced for optimum sequence alignment. The sequences used for the design of the degenerate PCR primers are indicated by the horizontal arrows above the PAI-2 sequence.
base pair 3Ј-untranslated region which includes a polyadenylation signal at nucleotides 1684. However, the cDNA extends for a further 135 base pairs and has no poly(A) tail, suggesting that Spi6 transcription can continue past this first polyadenylation signal. Transcriptional termination and polyadenylation might then be directed by another signal not represented in this clone. Two polyadenylation signals would allow transcriptional termination to occur at either of two points, and may explain the presence of two SPI6 transcripts detected by Northern blotting (see Fig. 5A).
The reactive center proximal hinge regions of SPI6 (Gly 326 -Thr-Glu-Ala-Ala-Ala-Ala-Ser-Ala) and mBM2A (Gly 139 -Thr-Glu-Ala-Ala-Ala-Ala-Ser-Ala) are conserved according to the consensus sequence for inhibitory serpins (39). This predicts that both are functional proteinase inhibitors. Alignment (without introducing deletions) of the SPI6 and mBM2A polypeptide sequences with the various human and mouse ovalbumin serpins predicts that the P 1 -P 1 Ј residues in the reactive center loop are Cys 340 -Ala 341 and Cys 153 -Ala 154 , respectively. However, the optimal alignment suggests Phe 338 -Cys 339 or Tyr 151 -Cys 152 , respectively (see Figs. 2 and 3B). The presence of Cys residues at or near the reactive center suggests that SPI6 and mBM2A may be oxidation sensitive, and is consistent with an intracellular localization and role. PI-6, PI-8, PI-9, and MNEI also have cysteines in their reactive center loops, and alkylation of MNEI and PI-6 destroys activity (6,11).
Expression of SPI6 and Interaction with Serine Proteinases-[ 35 S]Methionine-labeled SPI6 was produced in vitro using a coupled transcription/translation system. The translation products were analyzed by SDS-PAGE and fluorography. As predicted from the sequence, a protein of approximately 42 kDa was produced from the SPI6 cDNA (Fig. 5).
Labeled SPI6 was then tested for complex formation with graB and a number of other serine proteinases. A feature of serpin-serine proteinase interactions is the formation of a stable complex that is not dissociated by SDS (2). Addition of labeled SPI6 to purified graB resulted in the formation of a 67-kDa complex that was apparent following reduction, boiling, and electrophoresis in SDS (Fig. 5). Human PI-9, which forms an SDS-stable complex with graB (14), was produced in the same system and incubated with graB. As expected, a 67-kDa graB-PI-9 complex was observed (Fig. 5).
SPI6 did not complex with any of the following proteinases: trypsin, chymotrypsin, leukocyte (neutrophil) elastase, pancreatic elastase, thrombin, or cathepsin G (data not shown). Similar results were obtained by native PAGE (data not shown). These experiments suggest that SPI6 is a likely mouse homolog of PI-9.
Inhibition of Human GraB by SPI6 -Hexa-histidine tagged recombinant SPI6 was produced in a yeast expression system and purified by nickel affinity chromatography. As predicted from the cDNA sequence, the purified recombinant protein had a molecular mass of 42 kDa (not shown).
Complex formation between a serpin and proteinase follows second-order kinetics, and association rate constants of 10 5 to 10 7 M Ϫ1 s Ϫ1 represent physiologically significant interactions (3). The reaction stoichiometry is usually equimolar and initial formation of a Michaelis complex is followed by the formation of a kinetically stable (locked) tetrahedral complex (2,3). To examine the interaction of SPI6 and graB in more detail, we first established the stoichiometry of the reaction by titrating a fixed amount of graB against varying amounts of inhibitor and measuring residual proteolytic activity (data not shown). The 1:1 ratio observed is typical of a serine proteinase-serpin interaction (2). The association rate constant (k a ) for complex formation was calculated as 8 Ϯ 0.8 ϫ 10 4 M Ϫ1 s Ϫ1 . This is just outside the range for physiologically important serpin-proteinase interactions (2, 3), and is somewhat lower than the constant for the graB and PI-9 interaction (1.7 ϫ 10 6 M Ϫ1 s Ϫ1 (14)). This poorer than expected inhibition may simply reflect the different species origin of the serpin and proteinase (see "Discussion").
Role of Glu 337 in SPI6 Inhibitory Function-A unique feature of PI-9 is the presence of an acidic residue (Glu) at the P 1 position, which is consistent with its role as an inhibitor of a proteinase cleaving after acidic residues (Asp or Glu). Substitution of the P 1 Glu in PI-9 dramatically lowers inhibitory activity against graB. 4 Because SPI6 and mBM2A have shorter reactive center loops than PI-9 and other ovalbumin serpins, it is difficult to identify the likely P 1 residues by sequence alignment alone. However, both SPI6 and mBM2A have only one acidic residue (Glu) in their reactive center loops, and it is likely that this is important for inhibition of graB. To test this we used site-directed mutagenesis to produce a derivative of SPI6 in which Glu 337 has been substituted by Ala. This derivative was produced in the coupled transcription/translation system and in the yeast expression system. As shown in Fig. 5, labeled mutant protein did not form a detectable complex with graB on SDS-PAGE. In addition, although we were able to produce and purify equivalent amounts of the mutant from the yeast system (compared with SPI6), we were unable to demonstrate any graB inhibitory activity (data not shown). We conclude that Glu 337 is crucial for the inhibitory activity of SPI6, and that it may represent the P 1 residue. Tissue Distribution of SPI6 and mBM2A mRNA-To determine the tissue distribution of SPI6 and mBM2A, a membrane containing mRNA from mouse brain, heart, kidney, liver, lung, skeletal muscle, spleen, and testis was sequentially hybridized to 32 P-labeled SPI6 and mBM2A cDNA probes. With an SPI6 3Ј-untranslated region probe, two transcripts of approximately 2.4 and 3.6 kilobases were detected in most tissues, but predominantly in heart, lung, spleen, and kidney. The highest levels were in lung and spleen (Fig. 6). The nature of the two transcripts is unknown but they may arise by differential splicing or (as discussed above) differential transcriptional termination.
When the same membrane was stripped and hybridized to the mBM2A cDNA, the most prominent signal was obtained in testis (Fig. 6). Here a single 1.7-kilobase transcript was observed, that was not detected in other tissues. Fainter bands of identical size, distribution, and relative intensities as those seen with the SPI6 probe were also observed. These are likely to represent cross-hybridization of the mBM2A probe with SPI6 transcripts. We conclude that while mBM2A expression is essentially restricted to testis, SPI6 expression is somewhat broader.
Chromosomal Localization of Spi6 -The PI-6, human MNEI, and PI-9 genes co-localize on human chromosome 6p25 (17,18,20,40), while the PAI-2, SCCA, and maspin genes are clustered on human chromosome 18q21 (8). If SPI6 is a new member of the PI-6 family it is likely that the gene will localize to the same region on chromosome 13 as the gene encoding the mouse PI-6 homologue, SPI3.
The chromosomal localization of Spi6 was determined by hybridization of a partial SPI6 cDNA to a panel of DNAs from interspecific backcross mice (37). The SPI6 fragment (798 -1819 base pairs) detected a PvuII polymorphism between M. musculus (C57BL/BJEi) and M. spretus (SPRET/Ei) and was then hybridized to PvuII digested genomic DNA from each of the 94 animals derived from a (M. musculus ϫ M. spretus)F1 ϫ M. spretus backcross. After scoring each sample for the presence or absence of the M. musculus fragment, Spi6 was localized to mouse chromosome 13 in the same region as mouse Spi3 (Fig. 7). Spi3 and Spi6 could not be separated by this analysis, indicating a distance between them of 0 -3.8 centimorgan. Data in Table II show the recombination frequencies for the nearest linkage markers for Spi6. Spi6 is localized approximately 20 centimorgans from the centromere of chromosome 13 between the placental lactogen 1 gene (3.2 centimorgans proximal) and the motif-primed PCR marker D13Bir9 (1.1 centimorgans distal).

DISCUSSION
In this study we report the identification of 10 previously unknown serpin sequences in the mouse. Most of the new sequences were identified by RT-PCR using primers specifically designed to amplify ovalbumin serpin inhibitory regions, and sequence analysis shows that each new serpin has a Phe-Cys-Ala-Asp motif in the distal portion of this region that is present in most human ovalbumin serpins but not in other members of the serpin superfamily. Thus these new mouse FIG. 3. Sequences of SPI6 and mBM2A. A, cDNA and polypeptide sequence of SPI6. B, amino acid sequence comparison of SPI6 and the partial sequence mBM2A. The inhibitory region containing the reactive center loop (2, 39) is boxed.
serpins are probably members of the ovalbumin serpin subgroup. Based on the presence of consensus proximal hinge motifs each of these serpins is likely to be a proteinase inhibitor, but in the absence of functional data possible targets of these serpins cannot easily be predicted. It should be noted, however, that recent studies have shown that serpins are capable of inhibiting more than one class of proteinase (34,41). Hence the cognate proteinases of these mouse serpins may include serine proteinases and cysteine proteinases (caspases or papain-like cathepsins).
Even though our search for mouse ovalbumin serpins was largely restricted to activated lymphocytes, the number of new sequences identified suggests that the ovalbumin serpin family in the mouse is much larger than its human counterpart. The existence of many more serpins in the mouse has already been suggested by the description of two gene clusters on mouse chromosome 12 that have only single human gene equivalents (42,43). These clusters have almost certainly arisen by multi- FIG. 4. Comparison of SPI6 with the PI-6 ovalbumin serpins. The numbering, conserved amino acids and structural elements indicated by lines above the sequence are based on the structure of ␣ 1 -antiproteinase (1). The signal sequence removed to generate mature ␣ 1 -antiproteinase is boxed. Reactive center bonds (P 1 -P 1 Ј) are indicated by the vertical arrow. Dashes indicate gaps introduced for optimal sequence alignment. Sequences indicated are ␣ 1 -antiproteinase (1), SPI3 (21), PI-6 (11), PI-8 (13), PI-9 (13,14), and MNEI (6).

FIG. 5.
Interaction of SPI6 and an SPI6 inhibitory region mutant with graB. 35 S-Labeled SPI6, PI-9, mutated SPI6 (Glu 337 to Ala), and mutated PI-9 (Glu 340 to Ala) were produced in a coupled transcription/translation system. Products were incubated with or without 2 ng of human graB for 15 min at 37°C, reduced, and complex formation was analyzed following 12% SDS-PAGE and fluorography.
FIG. 6. Tissue distribution of SPI6 and mBM2A mRNA. Upper panel, a membrane containing 2-g samples of mRNA from a variety of mouse tissues was hybridized to a 32 P-labeled SPI6 probe consisting mainly of the 3Ј-untranslated region (from nucleotide 1035). The final wash was in 0.1 ϫ SSC at 60°C for 30 min. Lower panel, the membrane was stripped and hybridized to a 32 P-labeled mBMB2A probe consisting of coding and 3Ј-untranslated sequences. Washing conditions and exposure times were as exactly the same as the first hybridization. ple gene duplications and encode serpins resembling ␣ 1 -antiproteinase (Spi1) or ␣ 1 -antichymotrypsin (Spi2). It is likely that, although closely related, individual members of each cluster have different inhibitory capacities. The family of mouse ovalbumin serpins identified here have at least three human counterparts (PI-6, PI-8, and PI-9) and almost certainly also target a diverse range of proteinases. The fact that Spi6 maps to the same chromosomal region as Spi3 suggests that a serpin gene cluster comprising some of the mouse ovalbumin serpins is present on chromosome 13, and that SPI6 is a member of the PI-6 ovalbumin serpin subfamily and not the PAI-2 group. This is supported by sequence comparisons which show that SPI6 more closely resembles mouse PI-6 (SPI3, 55%) than mouse PAI-2 (40%). (Fig. 4 shows a comparison of SPI6 with the other members of the PI-6 group.) It remains to be seen whether Spi6 resembles Spi3 in gene organization and lacks an intron present in PAI-2, SCCA-1, SCCA-2, and ovalbumin (21).
The larger array of inhibitory serpins in the mouse suggests that rodents have many more proteinases that must be regulated. This is illustrated by a comparison of the granule proteinase complements of cytotoxic lymphocytes in humans and rodents. In humans there are five granule serine proteinases whereas in rodents there are at least nine (44). It is conceivable that some of the new serpins we have identified that lack obvious human counterparts are involved in regulating granule serine proteinases specific to the mouse. Since most of these new serpins are likely to be lymphocyte products it is probable that other mouse leukocytes have a distinct ovalbumin serpin complement, perhaps with novel members not identified here. For example, monocytes and neutrophils should have one or more counterparts of PAI-2 and MNEI. At present only one mouse homolog of PAI-2 is known (20). Due to the primer design, we may not have amplified non-inhibitory serpin sequences so it is also conceivable that non-inhibitory mouse ovalbumin serpins await discovery.
An interesting observation arising from this work is the apparent existence of multiple mouse homologs of at least two human ovalbumin serpins. Specifically, the clones mBM17 and mNK10 are probably homologs of human PI-8, while the clones mBM2A and SPI6 are homologs of human PI-9. It is possible that the different homologs play distinct roles in the mouse. For example, SPI6 and mBM2A may both regulate graB, but do so in separate contexts. This is supported by the fact that mBM2A and SPI6 have a different tissue distribution. Alternatively, these two serpins may regulate related but distinct proteinases within the mouse. Similarly, mBM17 and mNK10 probably have different roles since a different tissue distribution is implied by the presence of mNK10, and absence of mBM17, in lymphocytes.
On the basis of its properties and distribution, we have proposed that PI-9 functions to protect cytotoxic lymphocytes and other immune system components, such as antigen-presenting cells, from mis-directed graB (14). It is most likely that SPI6 is the true PI-9 homolog because (i) it has over 68% sequence identity with PI-9; (ii) it is also a graB inhibitor; (iii) inhibition depends on a glutamic acid residue in the reactive loop; and (iv) like PI-9 it is present in lymphocytes and spleen cells. However, SPI6 shows a broader tissue distribution than PI-9, being present in heart, lung, and kidney. At present we do not know whether this indicates an additional function for a graB inhibitor in the mouse, or whether it simply represents a different distribution of antigen-presenting or cytotoxic (CD8 ϩ and CD56 ϩ ) cells in mouse tissues.
Although we have been unable to do functional studies with mBM2A, its expression pattern (predominant in testis, absent in lymphocytes and spleen cells) rules it out as a direct homolog of PI-9. If mBM2A is a graB inhibitor, its presence in mouse testis may be related to the immune-privileged status of certain cells within the testis. It has been shown recently that Sertoli cells in the mouse testis express Fas ligand, and can induce apoptosis of cytotoxic lymphocytes that come in contact with them (for review, see Ref. 45). It may be that mouse Sertoli cells also produce a graB inhibitor (mBM2A) to protect against the granule-mediated arm of the CL killing machinery. Alternately, mBM2A may regulate a different proteinase to graB. Recent evidence suggests that there is more than one serine aspase in rodents.
One problem with the proposal that SPI6 and PI-9 are homologs is the relative inefficiency with which SPI6 inhibits human graB, but this may be a consequence of the lower than expected sequence conservation (39%) in the SPI6 and PI-9 inhibitory loops. Specifically, the SPI6 loop is two amino acids shorter than PI-9, there is an aromatic amino acid near the P 1 -P 1 Ј residues in SPI6 but not in PI-9, and the putative P 1 glutamic acid residue is much closer to the proximal hinge in SPI6 than in PI-9. Obviously these differences may reflect structural variations in the active sites of mouse and human graB, and in their substrate specificity. Formal resolution awaits the availability of practical quantities of purified mouse graB to test with SPI6 and PI-9. We expect that SPI6 will inhibit mouse graB more efficiently than human graB.
In conclusion, we have provided evidence for a large family of ovalbumin serpins in mouse lymphocytes and marrow, and the existence of multiple homologs of two human intracellular serpins. The distribution and properties of these new serpins suggests that rodents possess additional proteolytic systems that contribute to immune function. This might reflect differing evolutionary pressures placed on rodent and human immune systems, and suggests that caution should be exercised in extrapolating results from rodent models to humans.
Acknowledgments-We thank J. Rose (University of Melbourne) for FIG. 7. Linkage map depicting the location of Spi6 on chromosome 13. The midregion of chromosome 13 is shown as a solid line with the genes in this region indicated. Only one of the many motif-primed PCR markers (D13Bir9) in the region is shown (37), and markers in the flanking regions (dotted lines) are not included. Gpld1, glucosyl phosphatidylinositol-specific phospholipase D gene; Pl1, placental lactogen gene; Zfp65, zinc finger protein 65; Ctla2, ctla2␣ gene; Tpbp, trophoblast-specific protein gene. the BSS membranes, Lucy Rowe (The Jackson Laboratory) for analyzing the chromosomal localization data, and the anonymous reviewer who in the early stages of the work encouraged us to consider that the similarities between SPI6 and PI-9 outweighed the differences.