BRK/Sik Expression in the Gastrointestinal Tract and in Colon Tumors¹

Sik is an intracellular tyrosine kinase that was identified in a screen for tyrosine kinases in intestinal epithelial cells (1). Although it is related to the SRC family and contains SH2 and SH3 domains, it has a very short unique NH₂ terminus and is not myristoylated (2). Expression of Sik is restricted to epithelial cells and has been detected in the skin and in all linings of the alimentary canal. Transcription of Sik is initiated as cells migrate away from the proliferative zone and begin the process of terminal differentiation. Sik expression is first detected at mouse embryonic day 15.5 in the differentiating granular layer of the skin (2).

The role of Sik in differentiation was examined in mouse keratinocytes (3). The addition of calcium to cultured mouse keratinocytes induces a terminal differentiation program and a cascade of tyrosine phosphorylation. Sik was activated within 2 min after calcium addition to keratinocytes. It was found to bind a rapidly phosphorylated M_r 65,000 GAP⁵-associated protein (GAP-A.p65) through its SH2 domain. Overexpression of Sik in the embryonic mouse keratinocyte cell line resulted in increased expression of the differentiation marker filaggrin during calcium-induced differentiation. This suggested that Sik, the only known tyrosine kinase activated in keratinocytes within minutes after calcium addition, is involved in a signal transduction pathway that may promote differentiation.

Here we demonstrate that the breast tumor kinase BRK is the human orthologue of Sik. A portion of the BRK catalytic domain was initially cloned using PCR and degenerate primers corresponding to the conserved regions of tyrosine kinase catalytic domains and RNA isolated from involved axillary nodes from a patient with metastatic breast cancer. The full-length BRK cDNA was isolated from the MCF-7 and T-47D breast cancer cell lines (4). BRK has been detected in breast tumors and in a number of breast tumor cell lines, but not in normal breast, liver, placenta, pancreas, or other tissues (4, 5). Of 41 primary breast tumor samples quantified by Western blotting relative to cytokeratin 18, BRK was overexpressed 5-fold or more in 27% of the samples and overexpressed 2-fold or more in 61% of the samples as compared to normal breast tissue. One tumor expressed 43-fold higher levels of BRK protein (5).

BRK has also been cloned from melanoma cells and named PTK6 (6). BRK mRNA levels were undetectable in seven primary melanoma lines, two normal samples of melanocytes, and in biopsies from metastatic melanomas. However, BRK was present in 2/22 metastatic melanoma cell lines, and may be expressed in 10% of primary melanoma and melanocyte cultures (7).

Using human/hamster somatic cell hybrids, BRK was mapped to human chromosome 20 (6). Fluorescence in situ hybridization was used to further localize the BRK gene to 20q13.3 (8). 20q13.3 was one of five regions found to be amplified in homogeneously staining regions of chromatin of three primary breast carcinomas (9). In addition to being amplified in breast tumors and breast tumor cell lines (10, 11), amplification at 20q13 has been detected in other epithelial tumors including gastric and gastro-esophogeal tumors (12) and colon tumors (13). We discuss the relationship between BRK and Sik and the possible role that this kinase may play in epithelial cell cancers.

A human small intestine (duodenum) 5′ Stretch cDNA library prepared in the λgt11 vector (Clontech) was screened using a ³²P-labeled 562-bp SstI fragment of the Sik cDNA (2). cDNA clones were sequenced by the dideoxynucleotide chain termination method (14), using Sequenase (USB).

Tissues for in situ hybridization experiments were surgically resected, fixed in neutral buffered formaldehyde, and embedded in paraffin. Tissues for protein expression studies were obtained through pinch biopsies obtained during endoscopic procedures performed in the GI suite at the University of Illinois Hospital (Chicago, IL). All samples were collected after informed consent was obtained in accordance to the previously approved institutional review board protocol.

Human genomic DNA was prepared from a blood sample obtained from a healthy volunteer. Mouse genomic DNA was extracted from the liver of a CD1 mouse. Ten μg of human and mouse DNA were digested with BamHI, HindIII, SstI, and BamHI-HindIII; subjected to electrophoresis through a 0.8% agarose gel; and transferred to a nitrocellulose membrane. Probes for hybridization included a 562-bp purified SstI fragment of Sik cDNA and an 884-bp SacI-EcoRI fragment of human BRK cDNA. 100 ng of each fragment was used for random primer labeling with 50 μCi of [³²P]dCTP. Membranes were washed in 0.1% SDS and 0.1× SSC at 65°C for the corresponding species probe and at 55°C for the interspecies hybridizations and exposed to film.

Sik is closely linked to Eef1a2 and was found on BAC clones containing this gene. Mapping of Eef1a2 was carried out using DNA from The Jackson Laboratory Interspecific Backcross BSS panel (15). This panel is made up of 94 N2 offspring derived from the cross (C57BL/6J × SPRET/Ei)F₁ × SPRET/Ei. Over 3310 loci have been mapped in this cross. Eef1a2 mapped to within the most distal group of markers on mouse chromosome 2. A BAC library constructed from 129/Sv ES cell DNA (Research Genetics) was screened with primers corresponding to Eef1a2. Two independent clones were isolated, each containing an insert of approximately 65–70 kb. Sequencing indicated that each clone contained the coding and 3′-UTR sequence of the Sik gene in addition to Eef1a2. One clone also contained the 5′-UTR sequence of Sik.

Sequences compared in Fig 3 B include mouse sequences Pltp: basepairs 67-1548 of GenBank locus U37226 (NCBl accession 1051265); Eya2: 166-1587 of U81603 (1816530); Gnas: 20-1204 of Y00703 (51127); Pck1: 1-546 of EST sequence AI037119, 470-539 of AA562908, 1-59 of AI021099, 18-478 of AA080172, 458-569 of AA286042, 9-537 of AA106463, 118-515 of AA110781; Lama5: 5906-10820 of U37501 (2599231); Eef1a2: 134-1525 of L26479 (1220409); Col9a3: 1-333 of X91012 (975686), 1-104 of AA027742 (1493761); Sik: 286-1641 of U16805 (847794); human sequences PLTP: 88-1569 of L26232 (468325); EYA2: 196-1617 of Y10261 (1834488); GNAS: 69-1253 of X04408 (31914); PCK1: 122-1990 of L05144 (189944) excluding 589-609 (for which mouse EST sequence was unavailable); LAMA5: 1-4930 of AB011105 (3043589); EEF1A2: 84-1475 of X70940 (38455); COL9A3: 1564-2012 of L41162 (1196420); BRK: 814-2169 of X78549 (515025).

Proteins extracted from mouse tissues were incubated for 1 h at 4°C with anti-Sik (C-17) antibody (Santa Cruz Biotechnology). After the addition of 40 μl of protein A-Sepharose (Pharmacia, Piscataway, NJ), samples were incubated for 1 h at 4°C. Complexes were extracted several times in 50 mm HEPES (pH 7.0), 0.15 m NaCl, and 0.1% NP40 and resuspended in kinase buffer [50 mm HEPES (pH 7.0), 1% NP40, and 10 mm MnCl]. Autophosphorylation reactions were performed by incubation of immunoprecipitated protein with 20 μCi of [γ-³²P]ATP (Amersham) for 15 min at 30°C. Reactions were stopped by adding EDTA to 20 mm, and samples were resuspended in 2% SDS, 62.5 mm Tris-HCl (pH 6.8), and 10% glycerol. Denatured proteins were separated on a 10% SDS-polyacrylamide gel, which was then treated for 2 h with 1 m KOH at 55°C, dried, and exposed to X-ray film.

Expression of Sik and BRK was analyzed by RNase protection assay, as described previously (1, 16), using [³²P]α-CTP-labeled antisense RNA probes. A pBlueScript SK II+ plasmid containing a 205-bp fragment encoding a portion of the Sik catalytic domain (1) was linearized at an XbaI site in the polylinker, and in vitro transcription was performed using T7 polymerase (Promega, Madison, WI). A 198-bp StuI-ApaI fragment of the BRK cDNA was subcloned into pBlueScript SK II−. This plasmid was linearized with HindIII, and in vitro transcription was performed with T7 polymerase. Linearized templates used for in vitro transcription were purified on 5% acrylamide gels. As controls for RNA levels and integrity, RNase protections were also performed with antisense probes for mouse or human cyclophilin (pTRI-cyclophilin-mouse and pTRI-cyclophilin-human; Ambion).

Total RNA from mouse tissues was prepared by homogenization in guanidine thiocyanate solution with 2-mercaptoethanol followed by CsCl gradient centrifugation (17). Total RNA from human colon tumors and adjacent normal colon tissue was a generous gift from Robert M. Lee and N. O. Davidson (University of Chicago). Twenty μg (mouse) or 10 μg (human) of each total RNA sample or an equal amount of baker’s yeast tRNA was precipitated with ethanol and resuspended in 30 μl of hybridization buffer containing 2 × 10⁵ cpm of probe. The concentration and quality of the RNA were confirmed on stained 4-morpholinepropanesulfonic acid/formaldehyde gels.

The SuperScript Preamplification System (Life Technologies, Inc.) and 2 μg of total RNA were used for synthesis of cDNA. PCR was performed in 20 μl containing 80 ng of cDNA, 50 pmol of each primer, 0.2 mm deoxynucleotide triphosphate mix, 1.2 mm MgCl₂, 20 mm Tris-HCl (pH 8.4), and 50 mm KCl. PCR was done using the following parameters: (a) denaturation, 45 s at 94°C; (b) annealing, 45 s at 67°C; and (c) extension, 1 min at 72°C. For each combination of primers, the kinetics of PCR amplification was studied, and the number of cycles corresponding to the plateau were determined. PCR was performed at an exponential range (29 cycles for BRK; 25 cycles for keratin 8). A 224-bp BRK PCR product was generated using primers Brk-1 (5′- ATCCAGGCCATGAGAAGC-3′) and Brk-2 (5′- TGGATGTAATTCTGCGACTCC-3′), corresponding to nt 706–724 and nt 929 to 909 of the Brk sequence. A 110-bp K8 PCR product was generated using primers K8-305 (5′-TTGCCTCCTTCATAGACAAGG-3′), corresponding to nt 305–326, and K8-415 (5′-TGTTGTCCATGTTGCTTCG-3′), corresponding to nt 396–415 of the human K8 gene sequence. PCR products were separated on a 2% agarose gel and stained with ethidium bromide. Bands corresponding to each specific PCR product were quantitated using NIH Image.⁶

Total protein was isolated from pinch biopsies or cultured cells. Samples were lysed in radioimmunoprecipitation assay buffer (1 × PBS, 1% NP40, 0.5% sodium deoxycholate, and 0.1% SDS). Twenty μg (tissue) or 30 μg (cell lines) of protein per lane were subjected to electrophoresis through a 10% SDS-polyacrylamide gel and transferred to Immobolin P membranes. Filters were blocked for 1 h in 5% nonfat dry milk in buffer containing 10 nm Tris-HCl (pH 7.5), 500 mm NaCl, and 0.1% Tween 20 and then incubated for 1 h with BRK (C-17) antibody (Santa Cruz Biotechnology), or they were blocked with buffer containing 10 mM Tris-HCl (pH 7.5), 100 mm NaCl, 0.1% Tween 20 and then incubated with β-antibody for 1 h. (Sigma Chemical Co., St. Louis, MO). The higher NaCl substantially reduced background obtained with the commercially available BRK polyclonal antibody. Subsequently, the membranes were stained with appropriate horseradish peroxidase-conjugated secondary antibodies, and antibody binding was detected using the SuperSignal ULTRA chemiluminescence substrate (Pierce). In control experiments, the immunogenic peptide specifically competed out BRK antibody binding, confirming the specificity of the signal.

BRK mRNA expression in the human intestine was examined using in situ hybridization techniques as described previously (2). The cRNA probe was prepared from a template consisting of an 884-bp SacI-EcoRI fragment of human BRK cDNA cloned into pBluescript II SK. The in vitro transcription product was labeled with ³⁵S-labeled UTP and hydrolyzed to fragments of an average size of 150 nt in length for better tissue penetration. After hybridization, slides were treated with RNase A for excess probe removal, washed with increasing stringency to 0.1% SSC at 55°C, and dehydrated in graded ethanols diluted with 0.3 m sodium acetate. Sections were coated with autoradiographic emulsion and exposed for 8–11 days at room temperature.

Paraffin-embedded biopsy samples sectioned at 5–8 μm thick were deparaffinized, hydrated, and preincubated in block buffer [0.1 m Tris-HCl (pH 7.5), 0.15 m NaCl, and 0.05% Tween 20; 1:50 normal goat serum] for 40 min. Sections were then incubated with 0.2 μg/ml BRK antibody (Santa Cruz Biotechnology) in block buffer overnight at 4°C, washed, and incubated with 1:250 biotinylated goat antirabbit antibody (Vector Laboratories) in block buffer for 30–60 min. After washing, the TSA-Indirect Kit (DuPont New England Nuclear) was used according to the manufacturer’s instructions. Briefly, sections were treated with strepavidin-horseradish peroxidase, reacted for 5 min with biotinyl tyramide reagent, visualized with 1:500 FITC-avidin DCS (Vector Laboratories), and mounted with Vectashield mounting medium (Vector Laboratories). For control sections, 4 ng/ng BRK antibody of the peptide from which the BRK antibody was raised (Santa Cruz Biotechnology) were added to block buffer, 10–15 min before use.

To isolate the human orthologue of Sik, we screened a normal human small intestine cDNA library (Clontech) with mouse Sik cDNA. From 10⁶ recombinant phage, three positive clones were isolated. These partial cDNA clones shared sequence identity with the nonreceptor tyrosine kinase BRK that was isolated from a human metastatic breast tumor (4).

Like Sik, BRK is also a 451-amino acid protein with SH2 and SH3 domains and putative regulatory tyrosines at the activation loop and COOH terminus. Unlike members of the SRC family, Sik and BRK each lack consensus myristoylation motifs at the NH₂ terminus. Sik and BRK also have the sequence HRDLAARN in their catalytic domains, in contrast to the HRDLRAAN sequence shared by members of the SRC family. The amino acid sequences of Sik and BRK are aligned in Fig. 1. Whereas the functional domains of BRK and Sik are conserved, Sik and BRK share only 80% amino acid identity and 83% nt identity.

The Sik and BRK mRNAs each contain repetitive elements in their 3′ noncoding regions. In Sik mRNA, a B1 repeat precedes the observed polyadenylation site by approximately 350 bp, whereas the related Alu repeat (18, 19) comprises the last 300 bp of the known BRK cDNA. The Sik B1 element is in an antisense orientation, occurs in a truncated form, and contains sequence motifs closely matching the family of mouse B1.m-B repeats (19), which is estimated by mutation rates of CpG and non-CpG positions to have been amplified in the mouse genome 4.4 ± 3.1 Myr ago (20). The BRK Alu element occurs in a sense orientation, terminating the known BRK sequence with its polyadenylation sequence, and displays hallmarks of the modern human Alu dimer Alu-Sb, the most recently amplified family of Alu repeats (21). Coincidental integrations are not improbable because Alu elements preferentially insert into regions of high GC content (22). Such repetitive elements may have regulatory effects on the mRNAs in which they are incorporated (23).

Because of the relatively low level of sequence identity (80%) between Sik and BRK, we performed Southern hybridization experiments to confirm that Sik and BRK are true orthologues. ³²P-labeled BRK and Sik probes were hybridized with restricted mouse and human genomic DNA. Stringent hybridization with probe from the same species indicated the position of bands for each gene. If the probe from the other species recognized a gene of equal or greater homology, an additional set of bands would be visible; however, both the mouse Sik probe and the human BRK probe detect the same fragments in mouse and human genomic DNA as shown in Fig. 2, confirming that Sik and BRK are the closest homologues of one another.

Comparison of the nt sequences encoding Sik and BRK indicated no apparent insertions or deletions; sequence differences largely reflect changes in the GC content of the DNA code. Of 231 differences within the coding sequence, 159 are positions where G or C is present in the human sequence, whereas the corresponding mouse nt is A or T. Many alterations in the protein sequence seem to be due to mutations of CpG dinucleotides. Of 86 CpG dinucleotides in the human, 18 are conserved with mouse sequence, which has 24 in total. Throughout the known promoter and mRNA sequence, the o/e CpG ratio of Sik averages 0.23 and does not exceed 0.3 within a 500-bp window, whereas BRK averages an o/e ratio of 0.63, peaking at 0.8 in a 5′ CpG island, but not falling below 0.4. The average level for mammalian genomic DNA is 0.26 (24). Loss of Sik CpG content is evenly distributed throughout promoter and coding sequence, with BRK always having at least double the Sik o/e CpG ratio.

The mouse Sik gene is linked to the gene Eef1a2 that encodes translation elongation factor α-2. A BAC library constructed from 129/Sv ES cell DNA was screened with primers corresponding to Eef1a2 (25). Two independent clones were isolated, each containing a 65–70-kb insert. Each was found by sequencing to contain the coding and 3′-UTR sequence of the Sik gene in addition to Eef1a2. One clone also contained all of the 5′-UTR sequence of Sik. The Sik gene must therefore map to the distal end of mouse chromosome 2, within 60 kb of Eef1a2, in a region of conserved synteny with human chromosome 20q13 (Fig. 3). Because BRK maps to human chromosome 20q13.3 (8), this provides further evidence that these two genes are orthologous. In wasted mice, the Eef1a2 gene is deleted (25), but the Sik gene is intact and appropriately expressed.⁷

The coding sequences of BRK and the linked genes EEF1a2 and LAMA5 contain 82–89% G + C in the third codon position, which places them well above the threshold of the H3 isochore (75%), the G + C-richest fraction of DNA that forms 3% of the human genome and contains 28% of human genes (26). H3 isochores, which are usually over 300 kb in length, are most concentrated in T (telomeric, thermally resistant, or H3+) bands and occur more sparsely in T′ (H3*) bands; one of the human genome’s 28 H3+ bands maps to the telomere of chromosome 20 (27, 28). Although overall values for G + C% and 5-methylcytosine in human and mouse are nearly identical (24), the mouse genome lacks the very GC-rich H3 isochore (29, 30). Accordingly, the mouse Sik, Eef1a2, and Lama5 genes contain 14–19% less G + C in the third codon position. Interestingly, the tightly linked Col9a3 gene also has 18% less G + C, although its third codon G + C is much lower. It has been estimated that 20% of the CpG islands present in the human have been lost in mouse orthologues, primarily in tissue-specific genes (31). The case of BRK and Sik suggests that this can occur by effects on the isochore level that increase CpG loss uniformly on a megabase scale. Thus, the conserved synteny of these genes in human and mouse chromosomes near the telomere allows the physical extent and nature of T band changes between the species to be examined directly.

Sik mRNA expression has only been detected in regenerating epithelia such as that lining the gastrointestinal tract, as well as in the liver and skin. Little or no mRNA expression was detected in the kidney, lung, spleen, testis, uterus, diaphragm, or brain (1, 2).⁸ We examined Sik mRNA expression in the mouse small and large intestine (Fig. 4 B) by RNase protection. Levels of the Sik protected fragment increase from the duodenum to the ileum, with the ileum having peak levels of Sik mRNA expression.

We could not detect Sik protein in total tissue protein extracts using standard Western blot analysis with a rabbit polyclonal anti-Sik antibody, perhaps due to its low differentiation-specific expression. Immunoprecipitations were performed, followed by in vitro kinase assays, with total proteins from brain, spleen, tongue, ileum, liver, and muscle (Fig. 4 C). Active mouse Sik, capable of autophosphorylating itself with ³²P, was detected in total protein from the ileum of the mouse small intestine. No autophosphorylated Sik was detected when Sik peptide was added to the immunoprecipitation reaction, indicating that the signal is specific. At longer exposures, some autophosphorylated Sik was also detected after immunoprecipitation with total tongue protein. These protein data are consistent with the earlier mRNA expression studies, which indicated that Sik mRNA was present in the ileum and the tongue (2).

Sik expression was also examined in the mammary gland. Mammary gland differentiation is regulated hormonally and requires pregnancy for the establishment of terminal differentiation (for a review, see Ref. 32). During involution after weaning, a number of morphological and biochemical changes occur. We examined Sik expression in the mammary glands of virgin, pregnant, and lactating mice and at different times after weaning. At no time did we detect expression in the normal mammary gland. In contrast, significant levels of Sik expression were detected in the skin and small intestine (Fig. 4 A). Sik expression was also not detected in the mammary gland at these stages using in situ hybridization and immunohistochemistry (data not shown), ruling out the possibility that Sik expression was induced in a small subset of the cells. We have also not detected Sik expression in the NMuMG normal mouse mammary gland cell line (Ref. 33; data not shown).

BRK expression was examined in the human colon carcinoma tumor cell lines SW480, HT29, T84, and Caco-2. Caco-2 cells, which differentiate spontaneously after becoming confluent, provide a model system for studying enterocyte differentiation. As these cells are maintained in culture, they polarize, form microvilli, and express increasing levels of brush border enzymes, such as sucrase isomaltase (34). Caco-2 cells are poorly tumorigenic in nude mice when compared with SW480 and HT29 cells (35). Using RNase protection assays, high levels of BRK mRNA were detected in SW480 and HT29 cells (Fig. 5 A), but low levels were present in Caco-2 and T84 cells. BRK mRNA levels increased 4-fold as Caco-2 cells differentiated, with peak levels appearing at 14 days after plating. This increase in BRK expression levels was detected in different stocks of Caco-2 cells as well as in total and polyadenylated RNAs. Expression of BRK was not detected in the Hep3B liver hepatoma cell line.

Levels of BRK protein were examined by Western blotting. SW480 and HT29 cells express high levels of BRK protein, equal to or exceeding that of the breast tumor cell line MCF-7 (Fig. 5 B). Although peak levels of BRK mRNA were detected at day 14 after plating in Caco-2 cells, peak levels of BRK protein were found at 7 days after plating, suggesting BRK posttranscriptional regulation. A BRK doublet in the Caco-2 and HT29 cells indicates that a modified form of the protein is also expressed in these cell lines.

Biopsy samples from esophagus, stomach, duodenum, and colon epithelia were obtained, and total proteins were extracted. The pinch biopsy samples were composed primarily of surface epithelial tissue. Using a rabbit anti-BRK polyclonal antibody, BRK protein expression was detected in all human gastrointestinal tissues that we examined (Fig. 6 A). Significant levels of Sik expression have also been detected in the epithelium of the mouse fetal stomach (2) and adult stomach and esophagus.⁸

Levels of BRK mRNA expression in five human intestinal tumors were compared with expression in adjacent normal mucosa by RNase protection (Fig. 6,B). BRK expression in three moderately differentiated colon tumors (T1, T2, and T5) was 2–3.5 times higher than in adjacent normal tissue (N1, N2, and N5). No difference in BRK levels was detected in a moderately differentiated rectal tumor (T3 and N3) or in one tubular adenoma from the cecum (T4 and N4). Expression of cyclophilin, our standard control, also increased approximately 2–3-fold in the tumor samples in a pattern very similar to that observed with BRK (data not shown). A 2–3-fold increase in cyclophilin expression in tumor samples has been reported previously (36). Other controls such as β-actin and glyceraldehyde-3-phosphate dehydrogenase also exhibit similar levels of increased expression in tumor samples (37). We performed semiquantitative RT-PCR to examine BRK expression in the paired samples and the expression of keratin 8, an epithelial specific marker expressed in the colon. When levels of BRK were normalized to the amount of keratin 8 expressed, 2-fold increases in BRK expression were observed for tumors T1 and T5 (Fig. 6 C).

We examined BRK expression in the normal human colon and colon tumors using in situ hybridization and immunocytochemistry. BRK mRNA can be detected at the highest levels in the upper crypts in cells that are exiting the cell cycle and undergoing terminal differentiation in the normal colon (Fig. 7,A). These data coincide with the findings obtained for the murine orthologue Sik (2). Higher levels of hybridization grains were apparent over the disorganized cells of moderately differentiated human colon adenoma (Fig. 7, C and D, open arrow) than in adjacent normal epithelium (closed arrow). A higher magnification of a portion of the tumor is shown in Fig. 7, E and F, which shows variability in the levels of BRK RNA expressed within the tumor.

We examined BRK expression in additional 12 archival colon tumor samples by immunohistochemistry. Several samples stained strongly positive for BRK expression specifically in the epithelial cells of the tumor. A statistical analysis was not possible because some of the samples obtained were of poor quality and did not stain with BRK or various control antibodies, and no conclusion about BRK expression could be made from those samples. BRK expression in epithelial cells of three archival tumors is shown in Fig. 8. BRK protein in the normal colon and in the tumor tissue appears to be primarily cytoplasmic.

From its initial cloning from breast tumor tissue, BRK has been suspected to be a proto-oncogene. BRK is a highly diverged intracellular kinase of the form SH3-SH2-YK, in which SH3 is a polyproline binding motif, SH2 recognizes phosphorylated tyrosine in a sequence-specific context, and YK is the tyrosine kinase catalytic domain. BRK is most closely homologous (45%) to SRK1 (P42686), a SRC-like tyrosine kinase from Spongilla lacustris, and it shares a 45% homology and six of seven introns with Dsrc41, a Src-like gene from Drosophila. Nonetheless, BRK is highly diverged, with nearly equivalent homology to the proto-oncogenes p60-YRK (Q02977), p59-FYN (P27446), p90 v-YES (61504) and its cellular homologue c-YES, FRK/RAK (P42685), and c-Src itself, all with a 44–45% protein identity. The BRK gene has only two intron boundaries conserved with the SRC family members (38, 39), further suggesting that BRK and Sik represent a distinct family of nonreceptor tyrosine kinases.

We isolated BRK encoding cDNA clones from a normal human small intestine cDNA library using a mouse Sik probe. The genes encoding Sik and BRK do not appear to be tightly conserved. In contrast to mouse and human SRC, which share a high degree of sequence identity (99%), mouse Sik and human BRK share only 80% amino acid sequence identity. Because of this relatively low level of homology, we confirmed that Sik and BRK are orthologues of one another by performing a series of Southern blot experiments. We found that radiolabeled probes specific for BRK and Sik recognized an identical simple set of bands in both mouse and human genomic DNA, indicating that no genes with closer homology existed in either genome. In addition, we mapped the Sik gene to the distal portion of mouse chromosome 2, which shows conservation of synteny with human chromosome 20q13.3 where BRK is located.

Because BRK was initially cloned from metastatic breast tumor RNA, we examined its expression during breast development in the mouse to determine whether it plays a role in normal differentiation. We were unable to detect Sik expression at any stage of normal mammary gland development in the mouse. BRK expression in breast tumors and breast tumor cell lines, but not in normal breast tissue, has suggested a role for BRK expression in carcinomas. BRK was found to be expressed at appreciable levels in approximately two-thirds of the breast tumors examined (5). Overexpression of BRK in the HB4a human mammary cell line mitogenically sensitizes these cells to epidermal growth factor. In addition, overexpression of BRK in these cells resulted in increased growth in soft agar, indicating that BRK overexpression can contribute to a transformed phenotype (40).

We determined that BRK is expressed in tumor cell lines derived from adenocarcinomas of the colon. Using different methodologies, we also detected moderate increases in BRK RNA and protein expression in primary human colon tumor samples. Whereas the BRK colon tumor data are only of preliminary statistical significance, they consistently favor the notion of a modest increase in BRK in colon tumor tissue, which falls short of the more dramatic induction reported for breast carcinomas. The increase detected in colon tumors may not be related to factors such as gene amplification, which would be expected to yield larger increases. The activity of BRK in colon tumors has not been examined, and it is possible that increases in BRK expression do not reflect increased kinase activity. Mutations of the BRK gene in tumors have not been reported. The sequence of BRK isolated from tumor cells and normal cells appears to be identical thus far (38, 39), suggesting that BRK overexpressed in tumor cells is the normal protein.

Several studies suggest that the related SRC family tyrosine kinases participate in the development of colon cancer. SRC tyrosine kinase activity was found to be increased in human colon tumor tissue and in a variety of colon carcinoma cell lines when compared with normal adjacent tissues and normal colonic epithelial cells (41, 42, 43, 44). SRC activity was found to progressively increase as adenomas become carcinomas, and the highest levels of SRC activity were found in metastatic lesions in the liver (45). Recently, an activating mutation in SRC codon 531 was identified in 12% of advanced colon tumors examined, providing the first genetic evidence for a role for SRC in colon cancer (46). A significant increase in the activity of YES has also been observed in colon tumors and in colon carcinoma cell lines (47, 48), although no increase in the activities of some other SRC family members such as LCK, FYN, HCK, or FGR was detected (47).

We found that BRK is present throughout the normal human gastrointestinal tract, in the esophagus, stomach, duodenum, and colon. We localized BRK expression to differentiating epithelial cells in the colon, where the highest levels of protein and mRNA were found in epithelial cells in the middle and upper colonic crypts. We also found that BRK expression increased during the early stages of Caco-2 cell differentiation in vitro. These data support the hypothesis that BRK may play a role in a signal transduction pathway associated with differentiation. In previous studies, we found that mouse Sik is expressed in a differentiation-specific manner in regenerating epithelial linings. Sik was shown to associate with a GAP-binding protein, possibly linking it to the Ras pathway. It will be important to determine the role of this epithelial-specific tyrosine kinase during normal differentiation and to unveil its potential relationship to the development of breast and colon cancers.

We thank Drs. Nicholas O. Davidson and Robert M. Lee for generously sharing tumor RNA samples with us and Doreen Chambers for excellent technical assistance. We also thank Dr. Michael Mihalov for providing archival tumor samples and Drs. Thomas Layden, Allan Halline, and Richard Benya for helpful discussions and comments.