Disparate tissue-specific expression of members of the tissue kallikrein multigene family of the rat.

To understand the regulatory diversity of the rat family of linked kallikrein genes, we have assayed the expression of family members in 20 major organs. Reverse transcription-polymerase chain reaction analysis using primers and hybridization probes specific for each of the 10 expressed kallikrein genes showed that no two family members share the same organ-specific pattern of expression. The only common site of expression for all 10 known active genes is the submandibular gland. The presence of the mRNA for at least one family member is detected in 19 of these 20 organs (liver excepted), from as few as three organs to as many as 18 for individual family members. For individual genes there can be more than a 105-fold variation in mRNA levels among organs, from a limit of detection of slightly less than 1 mRNA molecule/10 cells to more than 10,000 mRNA molecules/cell. Despite high sequence conservation and close linkage, the members of this family are expressed in very different and complex patterns. A gradient of diversity of expression corresponds to the order of the genes within the kallikrein family locus.

To understand the regulatory diversity of the rat family of linked kallikrein genes, we have assayed the expression of family members in 20 major organs. Reverse transcription-polymerase chain reaction analysis using primers and hybridization probes specific for each of the 10 expressed kallikrein genes showed that no two family members share the same organ-specific pattern of expression. The only common site of expression for all 10 known active genes is the submandibular gland. The presence of the mRNA for at least one family member is detected in 19 of these 20 organs (liver excepted), from as few as three organs to as many as 18 for individual family members. For individual genes there can be more than a 10 5 -fold variation in mRNA levels among organs, from a limit of detection of slightly less than 1 mRNA molecule/10 cells to more than 10,000 mRNA molecules/ cell. Despite high sequence conservation and close linkage, the members of this family are expressed in very different and complex patterns. A gradient of diversity of expression corresponds to the order of the genes within the kallikrein family locus.
Eukaryotic gene family members with similar nucleotide sequences, functions, and regulatory characteristics are often clustered. Sequence conservation among family members is mediated by nonreciprocal exchange of DNA sequence between family members through gene conversion and unequal crossing over (Dover, 1982). This sequence homogenization of linked family members is not restricted to structural gene sequences and may extend through transcriptional control regions as well (Dover and Tautz, 1986;Spoerel et al., 1989;Wines et al., 1991). As one consequence, transcriptional regulatory properties are often shared throughout the family (Dover and Tautz, 1986;Hibner et al., 1988;Kedes, 1980;Kafatos et al., 1987). In addition, family members in close proximity may share a common dominant regulatory element, such as the ␤-globin family locus control region (its LCR) 1 (Crossley and Orkin, 1993). Nonetheless, clustered family members can be differentially expressed as well (Wines et al., 1991;Kafatos et al., 1987;Gubits et al., 1984), which raises problems of transcriptional control different from those of dispersed single copy genes.
The tissue kallikrein genes are a tightly clustered family whose members share some common sites of expression and have disparate sites of expression as well. The kallikrein family encodes a group of simple serine proteases of wide tissue distribution and varied function (reviewed by MacDonald et al., 1988;Clements, 1989;Carretero et al., 1993). Although most mammals have tissue kallikrein families with five or fewer members, the family is greatly expanded in the murine lineage. Rats have an estimated 13 members, with two or three pseudogenes (Wines et al., 1991Gerald et al., 1986); mice have 24 kallikrein family members, half of which are pseudogenes (Evans et al., 1987). The 10 active kallikrein genes and two of the pseudogenes of the rat family have been mapped to two linked contigs, which span a total of 440 kb (Southard-Smith et al., 1994). The mouse family is clustered as well (Evans et al., 1987;Mason et al., 1983). Nucleotide sequence conservation within the rat and mouse families generally exceeds 85% and extends beyond coding regions, across introns, and into 5Ј and 3Ј gene flanking regions (Wines et al., 1991;Evans et al., 1987). Thus, the family mRNAs are very similar and difficult to distinguish, and the family proteins have similar physical properties and enzymatic activities. However, the presence of short divergent exonic regions (generally encoding surface polypeptide loops) provides the means to selectively detect individual genes and mRNAs of the family through the design of member-specific oligonucleotide probes Ashley and MacDonald, 1985b;van Leeuwen et al., 1986).
Previous Northern hybridization analyses of the expression patterns of individual rat and mouse family members suggested that the expression of each might be limited to a few organs, and that these sites were often shared by different members (see e.g., Wines et al., 1989;van Leeuwen et al., 1986;Ashley and MacDonald, 1985b;Brady et al., 1989;Clements et al., 1990;Ma et al., 1992). Comparisons were constrained, however, because generally few organ sites were examined, and those examined were often different for various family genes. Moreover, the analyses were compromised by the likelihood of cross-detection of the highly related family members . When more organs were examined by the more sensitive means of RT-PCR for the rat family members KLK1 and KLK8 2 (Southard-Smith et al., 1992;Saed et al., 1990), many new expression sites were detected, raising the possibility that the expression of the tissue kallikreins, although organ-specific for individual family members, is widespread.
To gain a more useful understanding of the tissue-specific regulation of the rat kallikrein family, we examined the expres-sion of all 10 active members in the same set of 20 different organs. To help ensure selective detection of the mRNA for individual family members, we have devised (Southard-Smith et al., 1994) and now tested a rigorous RT-PCR assay which is 3-fold redundant for specificity. The use of a highly sensitive RT-PCR-based assay also ensures that potentially relevant sites of expression are not overlooked due to the relatively low sensitivity of Northern hybridization. The results of this survey demonstrated remarkably diverse expression for such a tightly clustered gene family. Each member has a unique pattern of expression, with the submandibular gland as the only universally shared expression site.

MATERIALS AND METHODS
PCR Amplification of Kallikrein mRNAs-Total RNA was isolated from rat organs by the guanidine thiocyanate procedure described previously (Chirgwin et al., 1979;MacDonald et al., 1987). PCR (Mullis and Faloona, 1987) and RT-PCR (Veres et al., 1987) were performed as described elsewhere (Southard-Smith et al., 1992, 1994. In all cases 30 PCR cycles (except 35 for ␤-actin) of 94°C for 1 min, the optimal T A for 1 min, and 72°C for 1.5 min were performed. The optimal annealing temperatures (T A ) for the primer pairs for each gene/mRNA were calculated with the OLIGO 4.0 primer analysis program (National Biosciences, Inc.) and were 53°C for rKLK2 and rKLK3; 54°C for rKLK6 and rKLK8; 55°C for rKLK7 and rKLK9; 56°C for rKLK1 and rKLK12; and 57°C for rKLK4, rKLK10, and rat ␤ actin. The RT-PCR primers for ␤-actin (GGCCAACCGTGAAAAGATGAC and ATTGCCGATAGTGAT-GACCTG) were derived from codons 113-120 and 245-251 of the reported rat cytoplasmic gene sequence (Nudel et al., 1983).
PCR and RT-PCR products were analyzed by electrophoresis in composite agarose gels of 2% NuSieve, 1% agarose in TBE buffer (100 mM Tris borate, 2 mM EDTA, pH 8.3). The PCR products were transferred to Zeta-Probe membranes by capillary blotting and cross-linked by UV irradiation. The transferred products were detected by Southern blot hybridization with 32 P-labeled oligonucleotide probes specific for each family member. Oligodeoxynucleotide primers for PCR and probes for Southern hybridization were synthesized by phosphoramidite chemistry with an Applied Biosystems 494 DNA synthesizer. Hybridization with the oligonucleotide probes was performed at 37°C for 4 h. in a solution of 50 mM sodium phosphate (pH 7), 5 ϫ SSC (0.75 M sodium chloride, 0.75 M sodium citrate), 10 ϫ Denhardt's solution, 1% SDS, 0.1% sodium pyrophosphate, and 250 g/ml herring sperm DNA. To help ensure stringent detection of cognate genes, the membranes were routinely rinsed twice at room temperature in 6 ϫ SSC, 0.1% sodium pyrophosphate, and 0.5% SDS; then washed twice at 42°C for 5 min in the same solution; then at 42°C for 5 min in 6 ϫ SSC; and rinsed once at room temperature in 2 ϫ SSC. Cross-hybridization was monitored by including DNA dot blots of cloned family genes in the hybridization mixes. In the some instances, the following more stringent wash conditions were used to help ensure specificity: 46°C for rKLK3 and rKLK7 and 50°C for rKLK10.
The mixing experiment described in Fig. 2 tested the specificity of the PCR assay for individual kallikrein genes when the cognate target gene was present at a 1000-fold lower level than the other kallikrein genes. One reaction mix contained 1 amol of the target gene template in the presence of 1 fmol of the nine other genes. To verify that none of the other members could be amplified, the results of this reaction were compared to those of a second reaction mix of identical composition, except for the absence of the target gene. The sources of the cloned DNA for the genes have been described in: Wines et al. (1989) for rKLK1, -2, -3, -4, -6, and -8; Chen et al. (1988) for rKLK7 and -12; Ashley and MacDonald (1985a) for rKLK9 cDNA; and Ma et al. (1992) for rKLK10 cDNA.

Detection of the mRNAs for Individual Rat Kallikrein Family Members
Discrimination of individual kallikrein family members is made difficult by the high nucleotide sequence conservation of family members within a species, due to concerted evolution of the linked genes (Wines et al., 1991). Despite 85-95% nucleo-tide sequence identity within exons, however, divergent regions of 20 -30 nucleotides occur in each of the five exons ( Fig.  1). The divergent oligonucleotide regions can be exploited to distinguish individual rat kallikrein family members. We have devised sets of PCR oligonucleotide primers and hybridization probes complementary to the divergent regions, which selectively detect a single known kallikrein gene or mRNA. The sequences of these primers and probes, their position in the mRNAs, and the mismatch nucleotide differences among all active family members are given in Table I.
We previously demonstrated that each primer/probe set would selectively detect its cognate gene in a mixture of all 10 expressed kallikrein genes when tested with template genes at equimolar concentration (Southard-Smith et al., 1994). However, for an RT-PCR assay in which the mRNA of the target member may be present at a much lower level than other family mRNAs, the potential for cross-detection is greater, and a much more stringent test is required to verify specificity. Therefore, we tested the specificity of the RT-PCR primer pairs and hybridization oligonucleotides for each of the 10 kallikrein genes in the presence of 1000-fold molar excess of the other nine genes by comparing the results of two PCR reaction mixes. The first contained 1 amol of the cloned target gene template and a mixture of 1 fmol of template for each of the nine other genes. The second mix contained 1 fmol of each of the nine other genes, but not the target gene. In each instance a strong Southern hybridization signal was obtained for the PCR amplification with the mixture containing the target gene, and no signal was detectable in its absence (Fig. 2). These results indicate that the PCR assay selectively detects the correct target gene in the presence of at least a 1000-fold molar excess of all nine other genes.

Sensitivity of the RT-PCR Detection of Kallikrein Family mRNAs
Although this RT-PCR strategy permits a comprehensive analysis of the presence of an mRNA, its sensitivity may detect very low mRNA levels of unclear physiological relevance. Moreover, the level of detection is set arbitrarily; the limit of detection can be altered by changing the number of amplification cycles, the amount of cellular RNA template, and the length of autoradiographic exposure. We set the conditions of 30 cycles with 1 g of total RNA to cover a wide range of potentially relevant mRNA levels. These conditions were chosen to provide a qualitative, but not quantitative, survey of the expression of the family.
To estimate the level of detection and the range of signal response under these RT-PCR conditions, we tested a series of RNA samples representing 0.1-10,000 rKLK1 mRNA molecules/cell (Fig. 3). The series was derived by consecutive dilutions of total RNA isolated from male rat submandibular gland (11,000 rKLK1 mRNAs/cell) (Brady and MacDonald, 1990).
FIG. 1. Oligonucleotide regions in kallikrein family mRNAs with highly divergent sequences among family members. The full-length mRNA for a typical kallikrein family member is diagrammed, with short 5Ј-and 3Ј-untranslated regions (horizontal lines) extending from the amino acid coding region designated by the rectangle. The positions of the diversity regions D2-D5 are indicated. Codon numbering is based on the amino acid coding sequence for rat kallikrein KLK1 (Swift et al. 1982) with the codon for the amino terminus of the mature, active enzyme ϩ1. The diversity regions are named according to their association with the exon organization of the genes indicated by the numbered head-to-head arrows (top).

Overview of the Expression Patterns of the 10 Known Active Kallikrein Family Members
The presence of the mRNAs for the 10 expressed members of the family was assayed by RT-PCR conditions specific for each mRNA in 20 major organs (Fig. 4). For comparative purposes, the autoradiographic exposures were adjusted for each gene to give similar signal intensities for submandibular gland RNA samples. The results demonstrated that no two kallikrein genes have the same expression pattern. Individual members are detectably expressed in as few as three (rKLK2) to as many as 18 (rKLK3) of the 20 organs.
In many instances the intensity of the RT-PCR signal was below that of the equivalent of 0.1 rKLK1 mRNA molecules/cell (see Fig. 3). Such a low level detected in an RNA sample from an entire organ may represent a high message level in a small subpopulation of cells in which its presence is indeed physiologically relevant. Alternatively, if the assay is indeed detecting one mRNA in every 10 cells or less, its presence at this level may not be physiologically relevant. Further analysis of each  Table I), PCR amplification was performed for two different template combinations: ϩ lanes, 1 amol of the cognate gene plus 1 fmol of the nine other genes; Ϫ lanes, 1 fmol of the nine non cognate genes. Shown are the autoradiographs of the Southern blots hybridized with the appropriate oligonucleotide probe for each gene (see Table I). The molecular sizes (in base pairs) of the PCR products are indicated below each panel. Genomic clones were used for all the kallikrein family members except rKLK9 and rKLK10, for which cloned cDNAs were used.

FIG. 4. Southern hybridization results for the RT-PCR survey of the sites of expression of individual kallikrein family members.
RT-PCR analyses were performed as described under "Materials and Methods" with 1 g of total cellular RNA from each of the 20 organs listed. The kallikrein PCR products were resolved by electrophoresis in agarose gels, blotted to Zeta-Probe filters (Bio-Rad), and hybridized with gene-specific oligonucleotide probes. Some of the samples in the rKLK1 panel have two additional bands; the lower of the two has the mobility of the PCR product of the gene sequence containing the 95-bp intron 3 and amplified from contaminating genomic DNA, and the upper band has the mobility expected for a cDNA-genomic DNA heteroduplex. The bottom panel shows the ethidium bromide-stained gel for the RT-PCR amplification of ␤-actin mRNA as proof for the integrity of each RNA preparation. site of very low expression would be necessary to assess its physiologic significance. Moreover, although the results presented in Fig. 2 indicate that 1,000-fold greater levels of noncognate genes are not detected, we cannot completely exclude the possibility that some of the faint signals represent the inappropriate and inefficient amplification and hybridization of a different family member present at very high levels in a given organ. This is a particular concern when the range of mRNA concentrations may span 5 orders of magnitude (i.e. from 0.1 to 10,000 molecules/cell), so that 0.00001% crossdetection of one mRNA present at 10,000 molecules/cell could incorrectly indicate its presence in the RNA sample at 0.1 mRNA/cell. We applied additional criteria to exclude most potential cases of cross-detection, such as experimental verification of the specificity of the gene-specific hybridization probes and appropriate correlation of DNA staining with hybridization signals for the RT-PCR Southern hybridizations. These precautions preclude the cross-detection for those low to moderate signals in Fig. 4, but cannot rigorously exclude some of the signals that may represent 0.1 mRNA/cell or lower. We distinguish these very faint RT-PCR signals that we cannot rigorously confirm with a Ϯ in the summary Fig. 5.

Expression of Individual Kallikrein Family Genes
rKLK1-The RT-PCR assay detected rKLK1 mRNA in the RNA preparations from 12 of the 20 tissues examined (Fig. 4). In addition, the mRNA for this gene was variably detected with very low signal intensity in brain RNA in several (but not all) experiments (data not shown). rKLK1 encodes the true kallikrein enzyme (Swift et al., 1982). rKLK1 mRNA has previously been detected in each of these 13 organs either through Northern hybridization, direct cDNA cloning, or RT-PCR (brain: Chao et al. (1983), Southard-Smith et al. (1992); pitui-tary: Pritchett and Roberts, (1987), Clements et al. (1989); the salivary glands: Gerald et al. (1986), Ashley and MacDonald (1985b), Southard-Smith et al. (1992); kidney: Ashley and Mac-Donald (1985b), Inoue et al. (1989); pancreas: Ashley and Mac-Donald (1985b), Swift et al. (1982); stomach, duodenum, and colon: Fuller et al. (1989)). This sensitive assay did not detect rKLK1 mRNA in the prostate, consistent with previous reports (Ashley and MacDonald, 1985b;Clements et al., 1988;Wines et al., 1989); the detection of rKLK1 mRNA by Northern hybridization by Chen et al. (1988) was probably due to cross-hybridization of the oligonucleotide probe with other mRNAs of the family.
rKLK2-The rKLK2 gene has the most restricted expression of the family. The presence of rKLK2 mRNA was detected only in submandibular and sublingual glands and barely in the thymus. rKLK2 encodes the enzyme tonin, which is able to process angiotensinogen to angiotensin II in vitro (Grise et al., 1981). The demonstration ) that the original detection of rKLK2 in prostate (Ashley and MacDonald, 1985b) was due to cross-hybridization of the oligonucleotide probe with rKLK3 is consistent with the absence of an RT-PCR signal in the lane for prostate RNA in this study (Fig. 4).
rKLK3-The rKLK3 gene has the most widespread expression of the family. rKLK3 mRNA was detected readily for 18 organs; only liver and muscle did not contain mRNA. mRNA levels in brain, pituitary, and thymus were just above the level of detection for the standard RT-PCR assay. Expression in the submandibular gland (see, e.g., Ashley and MacDonald, 1985b;Fuller et al., 1988) has been reported, but low levels of rKLK3 mRNA were overlooked for parotid, kidney, spleen, pancreas, stomach, duodenum, colon, parotid, and testes. Because rKLK3 is the only family member detected in adrenal gland RNA, the generic glandular kallikrein previously detected in adrenal glands (Nolly et al., 1993) appears to be that encoded by rKLK3.
rKLK4 -rKLK4 mRNA was detected in only three sites (submandibular gland, thymus, and kidney), consistent with its limited distribution detected by Northern analysis (Wines, 1989). The Southern hybridization signal for rKLK4 is a doublet (Fig. 4) for the submandibular gland and the kidney, but not the thymus. The presence of a doublet is indicative of co-amplification of a second family member present in submandibular and kidney RNAs, but not in thymus RNA, and the formation of hybrid molecules of altered electrophoretic mobility. The lower band has the mobility of the expected 396-bp product, whereas the upper band has mobility equivalent to that of a 450-bp double-stranded DNA. The slower mobility of hybrid cDNA duplexes is due to an altered structure caused by base mispairings. A single band at 396 bp, which did not hybridize with the rKLK4 probe, was detectable by ethidium bromide staining from the RT-PCR reactions for nine of the other organs (data not shown) and is consistent with the amplification (but not hybridization) of the other family member in these tissue RNAs as well. Expression in these other tissues plus submandibular and kidney, but not thymus, is consistent with the distribution of rKLK1 mRNA. To determine whether this interfering mRNA is that of rKLK1, we sequenced the 396-bp RT-PCR products from the sublingual and pancreatic RNAs. The nucleotide sequence was identical to rKLK1. Moreover, rehybridization of the stripped Southern blot of the rKLK4 survey (from Fig. 4) with the rKLK1-specific oligonucleotide probe (rKLK1-D3b probe, Table I) detected PCR products in organs that matched the rKLK1 pattern. Therefore, despite the apparent specificity of the rKLK4 primers (Fig. 2), rKLK1 mRNA was also amplified under the RT-PCR conditions, although the rKLK1 product was not detected with the rKLK4 hybridization probe. Consequently, the specificity of the FIG. 5. Correlation of the diversity of expression and the position of family members in the kallikrein genetic locus. The expression data from Fig. 3 and other RT-PCR experiments were compiled, and the number of expressing tissues is related to the gene map of the kallikrein family locus. The locus comprises one 175-kb contig fortuitously containing the odd-numbered genes separated by an estimated 74 kb from a 225-kb contig containing the even-numbered genes (Southard-Smith et al., 1994). The RT-PCR signals equivalent to or greater than that representing 0.1 mRNA/cell for rKLK1 (see Fig. 4) are taken to indicate significant expression in an organ and scored ϩ (see text). Detectable signals below this intensity are indicated as Ϯ. The absence of a detectable RT-PCR signal is scored Ϫ. The solid pluses indicate those sites in which individual mRNAs have been detected previously by Northern hybridization. hybridization probe is an important part of the overall 3-fold redundant specificity of the RT-PCR assay.
The formation of mismatched heteroduplexes is a useful diagnostic indicator for the cross-amplification of more than one family member mRNA. None of the tissue surveys for the other nine kallikrein mRNAs had more than one ethidium bromide-staining band (data not shown), a further indication of the selectivity of the RT-PCR assay.
rKLK6 -The detection of rKLK6 mRNA in submandibular RNA confirmed the sole site of expression detectable by Northern hybridization (Wines, 1989). In addition, the RT-PCR assay detected lower levels of rKLK6 mRNA in RNA isolated from brain, sublingual gland, thymus, prostate, and testis, and very small amounts in stomach RNA.
rKLK7-rKLK7 was initially cloned and characterized from a kidney cDNA library (Brady and MacDonald, 1990). It encodes esterase B, a weak kininogenase (Berg et al., 1992b) with marked preference for cleavage after arginyl residues (Elmoujahed et al. 1990). At the sensitivity of Northern blot analysis, rKLK7 mRNA was detected in RNA from submandibular gland as well as from kidney, but not in the RNA of 10 other tissues examined, including sublingual gland, duodenum, muscle, prostate, and testis (Brady and MacDonald, 1990) (Brady and MacDonald, 1990;Fuller et al., 1988;Chen et al., 1988). The RT-PCR results of Fig. 4 demonstrated that rKLK7 mRNA is also present, albeit at lower levels, in the sublingual gland, colon, and muscle, and at barely detectable levels (estimated below 1 mRNA/10 cells) in the thymus, duodenum, prostate, and testis.
rKLK8 -Of the 10 organs with detectable rKLK8 mRNA (Fig. 4), Northern hybridization only detected it in submandibular and prostate (Fuller et al., 1989). Kidney, ileum, and colon had mRNA levels near the minimum of detection by the standard RT-PCR assay. rKLK8 encodes a serine protease with a high preference for cleavage after arginyl residues (Elmoujahed et al. 1990), but of unknown physiological function.
rKLK9 -rKLK9 mRNA was found in the RNA isolated from 16 organs, four at barely detectable levels. rKLK9 encodes an enzyme called SEV, which has direct vasoconstrictor activity (Yamaguchi et al. 1991). The mRNA had previously been shown to be present by Northern hybridization in the submandibular gland and prostate, but lower levels were overlooked in parotid, kidney, duodenum, colon, and testes, (Ashley and MacDonald, 1985b;Fuller et al., 1988Fuller et al., , 1989Wines et al., 1989). These results confirm the presence of the mRNA encoding the SEV kallikrein in kidney (Saed et al., 1992) .
rKLK10 -This family mRNA, which encodes T-kininogenase (Ma et al., 1992), was found at relatively high levels in the submandibular and sublingual glands, thymus, and stomach, and at barely detectable levels in the pituitary, parotid, prostate, testis, and kidney (Fig. 4). This exceedingly low level in kidney was verified with an independent kidney RNA preparation. Using a similar RT-PCR assay, but with different primers, Ma et al. (1992) detected a relatively high level in the kidney. Examination of the RT-PCR primers and hybridization probed used by Ma et al. (1992) suggested that rKLK4 mRNA may be detected as well. Because rKLK4 mRNA is present at a relatively high level in the kidney (Fig. 4), the strong signal for the kidney detected by Ma et al. 1992 may be due to crossdetection of rKLK4 mRNA. Of all the seven sites of expression detected by the RT-PCR assay, rKLK10 mRNA is detectable by Northern hybridization only in the submandibular gland (Ma et al., 1992).
rKLK12-rKLK12 mRNA is present in submandibular and sublingual glands, kidney, and prostate and is just detectable in thymus. The levels in sublingual gland, kidney, and prostate are below detection by Northern hybridization (Chen et al., 1988). 3

DISCUSSION
This study demonstrated that the tissue kallikrein family members are disparately expressed in a wide range of organs (summarized in Fig. 5). No two family members have the same expression pattern. Members can be detected in as few as three organs (e.g. rKLK2) to as many as 18 (rKLK3). With the exception of liver, at least one family member is expressed in each of the 20 organs. Of the potential 200 expression sites examined for the total of 10 genes, nearly half (89) are positive. Whereas most of the organs examined express multiple family genes, several (brain, pituitary, lung, and adrenal) express only one, and it is not the same one. The submandibular gland is the only tissue in which all the family genes are expressed. Although the salivary glands are a common site for high expression of the family, not all the salivary glands express all the family members; rKLK4 mRNA is not detectable in the sublingual gland, and the mRNAs for five members are not detectable in the parotid.
Functional Diversity of the Kallikrein Family-The high potential functional diversity of the kallikrein gene family is generated by two distinct mechanisms (Wines et al., 1991). The first is the variation in family members at key amino acid residues that alter the substrate specificity of the enzymes and thereby directly affect enzymatic function. This variation has created a collection of very similar proteins with distinctly different polypeptide cleavage specificities (Elmoujahed et al. 1990;Gutman et al., 1991) and potentially diverse physiological roles Mason et al., 1983;.
The second mechanism is the disparate cell-specific expression of family members described in this report. In this analysis of 20 major organs, no two have the same complement of kallikrein mRNAs. Liver was the only organ examined that did not express a tissue kallikrein gene at a significant level. The pituitary and adrenal glands each expressed a single kallikrein, KLK1 and KLK3, respectively. All other organs expressed at least two.
Together these two mechanisms provide a very complex expression pattern of hydrolytic proteases with diverse specificity. Assuming that each family member is enzymatically distinct, this expression pattern provides a unique combination of kallikrein-like proteolytic activity in each organ. The debate now is whether this potential functional diversity is utilized in vivo. Humans appear to have at most five kallikrein members, so that many of the active murine genes may encode enzymes with either no physiologic roles or redundant ones. Only two family members have been shown clearly to have physiological roles conserved among mammals. True kallikrein (EC 3.4.21.8), encoded by KLK1, participates in the regulation of blood flow through its proteolytic activation of kininogen to form the vasodilator lysyl-bradykinin (reviewed by Fiedler (1979)). In addition, enzymes orthologous to the human prostate specific antigen are expressed in the prostate of several mammalian species and appear to be commonly involved in the modification of semen (Lilja, 1985;Chapdelaine et al., 1984;Dunbar and Bradshaw, 1987). The availability of recombinant cDNA clones for the mRNAs of the murine kallikrein family genes now makes it feasible to identify and examine the biochemical and physiological role of each member of this complex gene family.
Expression of Kallikrein Family Members-All 10 known 3 E. Kroon and R. J. MacDonald, unpublished observations. active tissue kallikrein genes of the rat have been cloned and mapped to two contigs spanning 175 kb (containing the odd numbered genes) and 225 kb (containing the even numbered genes) (Southard-Smith et al., 1994). The two contigs are separated by approximately 74 kb. The data of this report show that the level of expression differs for the genes on the two contigs. The genes of the odd contig generally are expressed in a wide range of organs (from 10 to 18 of the total of 20 organs examined), whereas the genes of the even contig are generally expressed in many fewer organs (from 3 to 10; summarized in Fig. 5).
The organization of the two contigs differs in two ways. First, although all the genes on each contig are transcribed in the same direction, this direction appears divergent for the two contigs. Thus, the organization in two divergent orientations may reflect a regulatory mechanism controlling the diversity of transcription. Second, a CpG island is located at the extreme end of the odd contig, farthest from the even contig, and may reflect a source for a gradient of transcriptional diversity. Family members with the most diverse expression reside at this end of the cluster.
We are currently investigating the transcriptional control strategy of this locus to determine the nature of this apparent transcriptional activity gradient. One possibility is the presence of a single regulatory region such as an LCR at one end of the kallikrein gene cluster. In this instance the diversity of expression may decrease with the distance from the LCR. Alternatively, each gene may have associated all the regulatory information necessary for its characteristic pattern of expression. In this case the difference in the diversity of expression of the genes on the two contigs may be due to the divergent evolution of the regulatory elements of the two contigs, facilitated by their divergent orientation, which may limit sequence information exchange between contigs. Preliminary evidence from the analysis of transgenic mice bearing extended cloned regions of the locus indicates that rather than a single dominant LCR, each gene has its own controlling elements for its unique pattern of expression. 4