Protein Products of the Rat Kallikrein Gene Family SUBSTRATE SPECIFICITIES OF KALLIKREIN rK2 AND KALLIKREIN rK9*

Two closely related kallikrein-like proteinases having little activity toward the standard synthetic amide substrates of tissue kallikreins were isolated from the rat submandibular gland. They were found to be the protein products of the rKlk2 (tonin) and the rKlk9 genes by amino acid sequence analysis (nomenclature of the genes and proteins of the kallikrein family is according to the proposal of the discussion panel from the participants of the KININ ’91 meeting held Sept. 8-14,1991, in Munich, Germany). These two proteinases of similar structure also had very similar physi- cochemical properties. They differed from other kal-likrein-related proteinases in having high pHi values of 6.20 (rK2) and 6.86 (rK9). Kallikrein rK2 was purified as a single peptide chain, whereas rK9 ap- peared as a two-chain protein after reduction. Their enzymatic properties were also very similar and differed significantly from those of other rat kallikrein- related proteinases. Unlike the

Two closely related kallikrein-like proteinases having little activity toward the standard synthetic amide substrates of tissue kallikreins were isolated from the rat submandibular gland. They were found to be the protein products of the rKlk2 (tonin) and the rKlk9 genes by amino acid sequence analysis (nomenclature of the genes and proteins of the kallikrein family is according to the proposal of the discussion panel from the participants of the KININ '91 meeting held Sept. 8-14,1991, in Munich, Germany). These two proteinases of similar structure also had very similar physicochemical properties. They differed from other kallikrein-related proteinases in having high pHi values of 6.20 (rK2) and 6.86 (rK9). Kallikrein rK2 was purified as a single peptide chain, whereas rK9 appeared as a two-chain protein after reduction. Their enzymatic properties were also very similar and differed significantly from those of other rat kallikreinrelated proteinases. Unlike the five other kallikreinrelated proteinases we have purified so far, kallikrein r K 9 was not inhibited by aprotinin. r K 9 also differed from rK2 by its tissue localization. The prostate gland contained only rK9 where it was the major kallikreinlike component.
The amino acids preferentially accommodated by the proteinase 53 to 52' subsites were identified using synthetic amide and protein substrates. Unlike other kallikrein-related proteinases, rK2 had a prevalent chymotrypsin-like specificity, whereas rK9 had both chymotrypsin-like and trypsin-like properties. Both rK2 and rK9 preferred a prolyl residue in position P2 of the substrate and did not accommodate bulky and hydrophobic residues at that position, as did most of the other kallikrein-related proteinases. This P2-proline-directed specificity is necessary for processing the precursors of several biologically active peptides. Subsites accommodating residues COOH-terminal to the scissile bond were also important in determining the overall substrate specificity of these proteinases. rK2 and r K 9 both showed a preference for hydrophobic residues in P2'. Other subsites upstream of the 53 subsite were found to intervene in substrate binding and hydrolysis. The restricted specificity of rK2 and rK9 is consistent with the presence of an extended substrate binding site, and hence with a processing * This work was supported by Grant CRE 895006 from the Institut National de la Santk et de la Recherche MBdicale (INSERM). 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. I To whom correspondence should be addressed URA CNRS 1334, Facultk de MBdecine, 2bis, Bd Tonnelli., F-37032 Tours Cedex, France. Tel.: 33-47-36-60-45;Fax: 33-47-36-60-46. enzyme function. Their P1 specificities enabled both proteinases to release angiotensin I1 from angiotensinogen and from angiotensinogen I, but rK9 was at least 100 times less active than rK2 on both substrates. The substrate specificities of rK2 and rK9 were correlated with key amino acids defining their substrate binding site. The predicted preferential sequence(s) around the cleavage site deduced from these data may be used to identify the biological substrate(s) of these proteinases.
The protein products of the tissue kallikrein gene family have not all been characterized. This is mainly because of the large number of genes in this family, particularly in the mouse and the rat, and the structural similarity of these genes (1-5). Although many of the kallikrein-related proteinases in the rat have been described, only a few have been correlated unambiguously with their corresponding gene or mRNA (6-9). Tissue kallikrein (rK1)' and tonin (rK2), encoded by the genes rKlkl (formerly rGK1) and rKlk2 (formerly rGK2), were the first to be characterized (6, 7). Four additional kallikreins have recently been characterized by amino-terminal sequence analysis and correlated with genes in the family. These include kallikreins rK7 and rK8, encoded by the the rKlk7 (formerly RSKG7) and rKlk8 (formerly rGK8) genes (8), kallikrein rK9, formerly reported as prostatic protease (9), SEV (10) or KLP-S3 (11) which corresponds to the so far unidentified rKlk9 gene (S3 mRNA (12)), and kallikrein rKlO (13), which is identical to antigen y (14) and probably to Tkininogenase (15), for which no corresponding gene or mRNA has so far been described.
There is increasing interest in these proteinases because several of them have different substrate specificities, despite their close structural relationship, suggesting that they could serve different biological functions. Kallikrein rK1, for example, is involved in processing kinin from kininogens (16), whereas rK2 can release angiotensin I1 from angiotensinogen in vitro (17). We have recently shown that kallikreins rK7, rK8, and rKlO have substrate specificities which differ from those of rK1, particularly for residues in the P2 position (nomenclature of Schechter and Berger (see Ref. 18)), and could be involved in the processing of precursors other than kininogens and angiotensinogen (8, 13).
Kallikrein rK9, which has 84% amino acid identity with rK2 (12), has been reported to be a tonin-like enzyme having vasoconstrictive activity (10). These two proteinases therefore provide an excellent model for investigating structure-func-Nomenclature of the genes and proteins of the kallikrein family is according to the proposal of the discussion panel from the participants of the KININ '91 meeting held Sept. 8-14, 1991, in Munich, Germany. 10045 VVGGYNC*ETNSQPWQVAVIGTTF Asterisks indicate cysteine residues which were identified as phenylthiohydantoine derivatives of pyridylethyl cysteine. tion relationships in this proteinase family. We have determined their overall substrate specificity using synthetic amide and ester substrates as well as natural peptide and protein substrates. This specificity has been tentatively correlated with differences in the key amino acids which define the substrate binding site in the members of the kallikrein family.

RESULTS
Purification of Kallikreins rK2 and rK9-A rat submandibular gland extract was fractionated on a DEAE-A50 column equilibrated in 10 mM phosphate buffer, pH 6.35,O.l M NaC1. Fractions of the two main peaks of unbound material were assayed for their esterolytic activity usingp-tosyl-L-argininemethyl ester in the presence or absence of excess aprotinin. All fractions contained esterolytic activity, but part of it, under the first peak, was slightly inhibited by aprotinin (Fig. 1s in the Miniprint Section). These fractions were pooled and fractionated on a Mono S column equilibrated in 25 mM sodium phosphate buffer, pH 6.0. The unbound material eluted as a single peak of aprotinin-inhibited esterolytic activity. The bound material, which also contained esterolytic activity, was eluted at 0.16 M NaCl and was not inhibited by aprotinin (Fig. 2S, Miniprint Section). The two peaks were concentrated and stored frozen until use.
NH2-terminal Sequence Analysis-Both proteinases were analyzed for their NH2-terminal sequence after reduction and pyridylethylation and fractionation on a C4 reverse phase chromatography column.The two peaks obtained after reduction of the Mono S-bound material corresponded to the NH2terminal sequence deduced from the nucleotide sequence of rKlk9 mRNA and to a segment starting at residue 90 (numbering based on the kallikrein rK1 sequence (5)) ( Table I). These two segments correspond to the light and heavy chains of the molecule, a structural feature found in the related kallikreins rK7, rK8, rKlO (8, 13). According to the rules proposed for kallikrein nomenclature,' this protein, which is identical to the prostatic protease (9), SEV (lo), and KLP-S3 (ll), will be called kallikrein rK9.
The unbound material eluted from the reverse phase chromatography column as a single peak after reduction-pyridylethylation, and its sequence corresponded to that of tonin or rK2.
Molecular Sizes and Isoelectric Points-The presence of one chain for rK2 and two chains for rK9 was confirmed by SDS-PAGE3 ( Fig. 1). Kallikreins rK2 and rK9 appeared to be Portions of this paper (including "Experimental Procedures," Figs. 1S-3S, and Table IS) are presented in miniprint a t the end of this paper. Miniprint is easily read with the aid of a standard magnifying glass. Full size photocopies are included in the microfilm edition of the Journal that is available from Waverly Press.
Kallikrein rK2 and rK9 are probably the two most basic kallikreins present in the rat submandibular gland with pHi values of 6.20 and 6.85, respectively, as determined from isoelectrofocusing on thin layer polyacrylamide gels using a pH 3.5-9.5 Ampholine gradient.
Enzymatic Properties-pNPGB-titrated rK2 and rK9 were assayed for their amidolytic and proteolytic activities using a variety of synthetic fluorogenic substrates and native or denatured protein substrates. All the amide substrates tested, corresponding to different combinations of PI, P2, and P3 residues (listed under "Experimental Procedures," Miniprint Section), were poorly hydrolyzed by both proteinases since most of the kcat/Km values were below 1 mM" This suggests that either rK2 and rK9 have little amidolytic activity or that subsites other than S1 and S2, and possibly S3, are critical for substrate binding. This possibility was investigated using protein substrates and two peptide substrates with intramolecularly quenched fluorescence, Abz-Phe-Arg-Ser-Arg-EDDnp and Abz-Phe-Arg-Leu-Val-Arg-EDDnp (19).
The former was significantly hydrolyzed by both proteinases (Table 11) though at a far lower rate than that reported for porcine or horse kallikrein (19). It was sensitive enough, however, to measure the activity of nanomolar concentrations of both proteinases.
Peptides and proteins, including reduced and pyridylethylated lysozyme, oxidized insulin B chain, and also substance P, renin tetradecapeptide substrate (angiotensinogen 1-14), and human big endothelin-1 which may be related to the biological activity of these enzymes (both have been reported as proteinases with vasoconstrictive activity), were then used to define the specificity of rK2 and rK9, and discriminate between their activities. Both proteinases were incubated under the same experimental conditions with the same amount of each substrate. The enzyme/substrate molar ratios varied between 1/15 and 1/23,000, depending on the substrate used, but were kept constant for both enzymes. Cleavage products were separated by reverse phase chromatography on a C18 column and analyzed for their NH2-terminal sequence to locate the preferential cleavage sites. As shown in Table  111, the two proteinases cleaved most of the substrates used at different sites, indicating that they have different specificities. Their overall specificity for P3 to P2' residues was analyzed, taking into account all rK2 and rK9 cleavage sites identified in the six peptide substrates as well as the frequency of distribution of individual residues. Amino acids were classified into six groups as described under "Experimental Procedures" (Miniprint Section). Results are reported in Fig. 2. Unlike other members of the kallikrein family, rK2 and rK9 were able to accommodate aromatic residues in position P1. rK9 differed from rK2 in that it also accommodated basic residues at that position. Both proteinases accommodated negatively charged residues in P2 but showed a preference for a prolyl residue at that position. Bulky and hydrophobic residues, which are preferentially accommodated in S2 by kallikreins K1 and several other members of the family, did not fit the rK2 and rK9 requirements for hydrolysis. Both proteinases also showed a preference for hydrophobic residues in P2', which confirms the presence of an extended substrate binding site on these proteinases. Angiotensin 1 1 Release from Rat Angiotensinogen and from Angiotensin I-The results obtained using the renin substrate tetradecapeptide showed that both enzymes cleaved at the angiotensin I1 site. The direct generation of angiotensin I1 from its protein precursor was further investigated by radioimmunoassay using rat angiotensinogen as substrate. Though both proteinases released angiotensin I1 but not angiotensin I from the rat precursor, rK9 was about 300 times less efficient than rK2 under the same experimental conditions. When angiotensin I was used as a substrate, rK9 was about 100 times less efficient than rK2 to release angiotensin I1 as revealed by reverse phase HPLC analysis on a C18 column.
Inhibition by Proteinase Inhibitors-The different susceptibilities of rK2 and rK9 to serine proteinase inhibitors represent another way of discriminating between these two proteinases. This feature was exploited to complete their purification. Table IV shows the results obtained using several representative serine proteinase inhibitors.
Only soybean trypsin inhibitor was as good an inhibitor of both proteinases, with apparent Kt values lower than 1 pM. Unlike all other kallikrein-related proteinases reported so far (8, 20), rK2 was poorly inhibited by aprotinin and rK9 was insensitive to this inhibitor. Rat al-PI, which does not inhibit kallikrein rK1, binds both rK2 and rK9, although rather slowly. Neither rK2 nor rK9 were significantly inhibited by ovomucoid trypsin inhibitor, lima bean trypsin inhibitor, or chymostatin.

DISCUSSION
In spite of their close structural similarities, with homologies of 72-89% (5), the proteinases of the rat tissue kallikrein  2. Analysis of the appearance of amino acids around the cleaved bonds in the hydrolysis of pyridylethylated lysozyme, oxidized insulin B chain, renin substrate tetradecapeptide, big-endothelin-1, and substance P by kallikreins rK2 and rK9. Amino acids were classified into six groups, and the preference for each group for P3 to P2' positions in the substrates was calculated as described under "Experimental Procedures." The cleaved bond in rat angiotensinogen, which is the same as that in renin substrate, was not included in the calculation. Upper panel, preference of rK2; lower panel, preference of rK9.

TABLE IV Kinetic parameters for the interaction of kallikreins rK2 and rK9
with some representative serine proteinase inhibitors Apparent inhibition constants for the interaction of rK2 and rK9 with aprotinin and soybean trypsin inhibitor (SBTI) were determined as described under "Experimental Procedures" (Miniprint Section).
The inhibition of both proteinases by a 1-proteinase inhibitor ( a 1-PI) was characterized by the second-order rate constant for inactivation k: , (see Miniprint Section for details). Results were the mean of at least two independent experiments. family can be relatively easily separated from the same tissue homogenate on the basis of their different electric charges. This was first exploited by Brandtzaeg et aE. (21), who used anion-exchange chromatography to fractionate a rat submandibular gland homogenate into four kallikrein-containing pools. DEAE-A50-unbound material was used to purify four kallikreins, including kallikreins rK2 and rK9. The other two kallikreins, rK7 and rK8, were purified and characterized as described previously (8). The different sensitivities of kalli-kreins to proteinase inhibitors have also been used to achieve their purification. In the present study, rK2 and rK9 were separated during their purification by their different susceptibilities to aprotinin. Previously, rat kallikreins rK7 and rK8 were separated by taking advantage of the fact that rK8 is insensitive to soybean trypsin inhibitor (8). Purified rK9 occasionally showed microheterogeneity in some preparations, depending on the starting material and the purification procedure used. Similar microheterogeneity has been reported for kallikrein rKlO (13) and shown to be due in part to alternative cleavage of the initial peptide chain in the region of the kallikrein loop (22), generating the light and heavy chains of this molecule, and to a variable glycosylation of the light chain. Such a process can be expected for kallikrein rK9, which is the only other member of the family to have two potential cleavage sites after Arg residues in this region, one giving rise to an NH2-terminally blocked heavy chain starting with a Gln residue and therefore unidentifiable by sequence analysis (13). The amino acid sequences of rK2 and rK9 are 84% identical, one of the highest percentages of homology between proteinases of the kallikrein family (12). This is most significant, as this similarity includes the key amino acid residues thought to be involved in the interaction with the substrate ( Table   V). As a consequence, the two proteinases could be expected to have similar functions. In accordance with this hypothesis, Yamaguchi et al. (10) reported that kallikrein rK9, like rK2 (7, 17), has vasoconstrictive activity in uitro. However, the mechanism by which this activity occurs seems to be different (10). We demonstrated by radioimmunoassay that both proteinases specifically release angiotensin I1 from its precursor and this was confirmed using a synthetic renin substrate (angiotensinogen fragment 1-14) and angiotensin I which were cleaved at their angiotensin I1 releasing site. However, the total amount of angiotensin I1 released from both substrates by rK9 was far lower than that released by rK2 under the same experimental conditions, suggesting a different biological activity for each proteinase in spite of their similar specificity.
rK2 and rK9 are similarly regulated by androgens, but their tissue localizations differ significantly. Both are found in the submandibular gland, but only rK9 is present in the rat prostate, where it is the main kallikrein-like component. The level of rK9 mRNA in this tissue has been reported to decrease dramatically upon castration (9, 23), whereas it is affected less, and to the same extent as rK2 mRNA, in the submandibular gland (23). Kallikrein-related proteinases have also been identified in the human and dog prostate glands (24,25).
The key amino acids of the human protease, reported as prostate-specific antigen, differ from those in rat kallikreins rK9 and rK2. In particular Asp183, which is highly conserved throughout the kallikrein family, is replaced by a Ser residue in human prostate-specific antigen (26). Other critical resi- dues such as Hisg3, GlYo5, and Ala217 in rK2 and rK9 are replaced by Ser, Trp, and Gly, respectively, in human prostate-specific antigen, indicating that the proteinases do not have the same specificity and therefore serve different functions. rK9 can be readily distinguished from the other kallikreins in the rat submandibular gland by its resistance to aprotinin inhibition. There is, as yet, no structural feature which can explain this property of rK9. Both kallikreins rK2 and rK9 are strongly inhibited by soybean trypsin inhibitor, a property they share with rK7 and rKlO (8,13) but which makes them different from kallikreins rK1 and rK8, which are not susceptible to this inhibitor. The way in which the activities of the kallikrein family proteinases are regulated by physiological proteinase inhibitors is not yet well understood; only rK1 significantly modifies blood pressure when injected in minute amounts into the circulation (16). Whether or not this is related to the fact that all the kallikreins except rK1 are inhibited by a1-PI remains questionable, given the low rate at which this reaction occurs in uitro. However, an inhibitor of tissue kallikrein which irreversibly binds rat rK1 has been recently described (27). This kallikrein-binding protein corresponds to the negative acute phase reactant SPI-2 (28,29), but its ability to inactivate other kallikrein-related proteinases remains to be investigated.
The relationship between the structure of kallikreins and their substrate specificity was investigated using a variety of peptide and protein substrates. Unlike other kallikrein proteinases, both rK2 and rK9 hydrolyze amide fluorogenic substrates very slowly, but they are the only two that have chymotrypsin-like specificity in addition to trypsin-like specificity. These two proteinases hydrolyzed the commonly used kallikrein substrates Pro-Phe-Arg-MCA and Z-Phe-Arg-MCA 10-100 times slower than kallikreins rK1, rK7, rK8, and rKlO (8, 13). Intramolecularly quenched fluorogenic peptide substrates were used to study the P1' and P2' specificity of proteinases. Such substrates have been recently developed by Chagas et al. (19) to measure tissue kallikrein activity. One of them, Abz-Phe-Arg-Ser-Arg-EDDnp, was hydrolyzed much faster than methyl-coumarylamide peptide substrates by both rK2 and rK9. However, the main substrate cleavage site, as identified by high performance liquid chromatography analysis, was not at the expected Arg-Ser bond found for rK1 and rK10, but at the amide bond involving the second Arg residue of the peptide substrate. This means that subsites upstream of S1 and S2 in the proteinase are important for defining the unusual specificity of rK2 and rK9. There is also evidence from structural studies that rK2 has a specificity for residues far from the scissile bond (30). The low activity of rK2 and rK9 toward the amide bond of the fluorogenic methylcoumarylamide substrates may thus be explained by the reduced length of the peptide chain, which generally has not more than 3 residues. Protein substrates are therefore appropriate tools for investigating the substrate specificity of rK2 and rK9, and possibly discriminating between their activities. The P 3 to P2' overall specificity of rK2 and rK9, as defined by the cumulative data for the cleavage sites in all the protein substrates used, shows a striking resemblance between the two proteinases. The main difference is in their P1 specificity: rK9 has both trypsin-like and chymotrypsin-like specificity cleaving after aromatic as well as basic residues, whereas rK2 is more strictly chymotrypsin-like, although still able to accommodate basic residues in PI. rK9 is therefore more related to the other members of the kallikrein family. On the other hand, both proteinases preferentially accommodate a prolyl residue in position P2, whereas a bulky, hydrophobic residue is generally preferred by other kallikreins (8, 13). It has been suggested that a proline-directed arginyl cleavage is important in the processing of peptide precursors (31). rK9, which obeys this specificity, could therefore be a good candidate for the processing of such precursors.
The differing specificities of both rK2 and rK9 from those of other members of the family agree well with the unusual structural features of these proteinases, and explain why they cannot accommodate bulky, hydrophobic residues in P2, as do most of the other proteinases of the family. This specificity is thought to be due to the presence of a hydrophobic sandwich between Tyrg3 and TrpZo5 of the kallikreins, which traps hydrophobic P2 residues (22). Kallikreins rK2 and rK9 are the only members of the family to have a glycyl residue in position 205, and Tyr93 is replaced by a histidyl residue. The results reported here confirm the importance of these residues for determining the size, shape and ionic character of the substrate binding pocket of kallikreins. The Gly/Ser residue at position 217 in most members of the kallikrein family may also be important for substrate binding specificity. This residue is replaced by an Ala in rK2 and rK9, and also in rK7, which has the same histidyl residue in position 93 (5), and has been shown to have a P2 specificity not restricted to bulky, hydrophobic residues (8).
This approach, which allows investigation of the structuresubstrate specificity relationships of these two closely related proteinases, may provide a method for elucidating their biological function. However, there is, to date, no evidence that the vasoconstrictor activity reported in uitro for both rK2 and rK9 (10,11,17) is of biological relevance. With the notable exception of rK1, which is a physiological kinin-releasing enzyme, the function of all other members of the rat kallikrein family also remains to be established. Identifying the physiological substrates of these proteinases therefore represents a considerable challenge.