In Vivo Expression of an Alternatively Spliced Human Tumor Message That Encodes a Truncated Form of Cathepsin B

Cathepsin B is a lysosomal cysteine protease whose increased expression is believed to be linked to the malignant progression of tumors. Alternative splicing and the use of alternative transcription initiation sites in humans produce cathepsin B mRNAs that differ in their 5′- and 3′-untranslated ends. Some human tumors also contain cathepsin B-related transcripts that lack exon 3 which encodes the N-terminal signal peptide and 34 of the 62-amino acid inhibitory propeptide. In this study we show that one such transcript, CB(−2,3), which is missing exons 2 and 3, is likely to be a functional message in tumors. Thus, CB(−2,3) was found to be otherwise complete, containing the remainder of the cathepsin B coding sequence and the part of the 3′-untranslated region that is common to all previously characterized cathepsin B mRNAs in humans. Its in vitrotranslation product can be folded to produce enzymatic activity against the cathepsin B-specific substrate,N α-benzyloxycarbonyl-l-Arg-l-Arg-4-methylcoumaryl-7-amide. Endogenous CB(−2,3) from the metastatic human melanoma cell line, A375M, co-sediments with polysomes, indicating that it engages the eukaryotic translation machinery in these cells. Epitope-tagged forms of the truncated cathepsin B from CB(−2,3) are produced in amounts comparable to the normal protein after transient transfection into COS cells. Immunofluorescence microscopy and subcellular fractionation show this novel tumor form of cathepsin B to be associated with nuclei and other membranous organelles, where it is likely to be bound to the cytoplasmic face of the membranes. This subcellular distribution was different from the lysosomal pattern shown by the epitope-tagged, full-length cathepsin B in COS cells. These results indicate that the message missing exons 2 and 3 is likely to be translated into a catalytically active enzyme, and that alternative splicing (exon skipping) could contribute to the aberrant intracellular trafficking of cathepsin B that is observed in some human cancers.

Cathepsin B is a lysosomal cysteine protease whose expression and trafficking are frequently altered in transformed and malignant cells (1). Both its association with the plasma membrane and its secretion have been linked to the increased capacity of human and rodent tumors to invade and metastasize (2)(3)(4)(5). The association of cathepsin B with cell fractions containing plasma membrane-derived vesicles and endosomes was found to increase after transfection of a human breast epithelial cell line with the c-Ha-ras oncogene (6). Cathepsin B has been detected in nuclei from human ectocervical tumors and in basal and columnar cells of the normal and hyperplastic prostate (7,8). In human lung tumor cell lines cathepsin B activity was also localized to the endoplasmic reticulum, nuclear membrane, and plasma membrane (9). Forms of cathepsin B greater in size and stability than the mature, lysosomal form have been observed in tumors from humans and animals or in media conditioned by them (10 -13). At present, many of these findings are unexplained.
Normal human tissues and human tumors contain different cathepsin B mRNAs due to alternative post-transcriptional processing of their 5Ј-and 3Ј-untranslated ends and to the probable existence of more than one promoter in the cathepsin B gene (14 -17). Some human tumors also contain cathepsin B-related transcripts that differ in their 5Ј-coding region (14,17). These appear to arise from the skipping of exons 2 and 3 during splicing of the pre-mRNA (14), and from transcription initiation in exon 4 (17). Both events result in the loss of the sequences for the signal peptide and a part of the inhibitory propeptide. Reticulocyte lysates were able to translate a synthetic message missing exons 2 and 3 with high efficiency, producing a truncated form of cathepsin B that lacked the first 51 amino acid residues (⌬ 51 CB) (14). In this paper we show that the transcript missing exons 2 and 3 is likely to be a functional message and that its translation product retains the capacity to be folded into an active enzyme. When tagged with the C (connecting)-peptide of proinsulin and expressed in COS cells, the truncated protein, ⌬ 51 CB-CP, associates with the cytosolic surface of nuclei and other membranous organelles. Therefore, alternative splicing could explain some of the diversity of size and location reported for cathepsin B in some human tumors, and suggests that cathepsin B could enhance their malignant behaviors by novel and previously unsuspected mechanisms. Institute, the University of Chicago. The cathepsin B template, phCB79-2(Ϫ2,3), was described previously (14). Standard cloning procedures were used to replace the 5Ј end of the PCR product with those from phCB79-2 and phCB79-2(Ϫ2) (14). For expression in COS cells, the three cathepsin B-C peptide cDNA fusions were introduced into the XbaI and SacI cloning sites of the vector, pEUKC1 (CLONTECH Laboratories, Inc., Palo Alto, CA), adjacent to the SV40 late promoter.
cDNAs encoding cathepsin B-Flag peptide fusions were constructed using the Promega Altered Sites in vitro mutagenesis system according to the guidelines provided by the supplier. The mutagenic oligonucleotide used to insert the Flag marker peptide (22) (Kodak Scientific Imaging Systems, New Haven, CT) at the carboxyl terminus of cathepsin B in phCB79-2(Ϫ2) and phCB79-2(Ϫ2,3) was 5Ј-ACTGGGAA-AAGATCGACTACAAGGACGATGACGACAAGTAATCTGCCGTGGGC. The sequence of the primer was verified by DNA sequencing (14). PCR was then used to generate XbaI and SacI sites for directional cloning into the pEUK-C1 vector.
In Vitro Transcription and Translation-Plasmid phCB79-2(Ϫ2,3) was used as a template for the preparation of m 7 G(5Ј)ppp(5Ј)G capped RNA as described previously (14). Purified, synthetic capped mRNA (0.6 g) was translated with a rabbit reticulocyte lysate (Promega, Madison, WI), according to the instructions of the supplier. To promote protein folding, the translation mixture was dialyzed in the cold against buffer (0.05 M Tris-HCl, 20% glycerol, 0.1 mg/ml bovine serum albumin, 0.15 M NaCl, 1 mM Na 2 EDTA) to which had been added 6.0 M guanidinium chloride. At intervals of 10 or 30 min, the dialysis buffer was diluted with an equal volume of buffer to reduce the guanidinium chloride concentration by half, and the dialysis continued for two additional cycles of dilution. The final cycle was followed by dialysis against buffer only (23,24). The samples were clarified by centrifugation at 14,000 ϫ g for 15 min and quick frozen for later analysis. In an alternative procedure, some samples were treated with 0.1 volume of a 10 ϫ "oxido-shuffling" redox buffer (1 ϫ contains 0.15 M Tris-HCl, pH 8.0, 1 mM Na 2 EDTA, 0.3 mM glutathione disulfide, and 3.0 mM reduced glutathione) and incubated at room temperature for 24 h (25). Cathepsin B activity was measured fluorometrically with Z-Arg-Arg-NMec as described previously (26). Some samples were pretreated for 30 min at 37°C with pepsin (0.83 mg/ml) to remove the remainder of the inhibitory propeptide (26,27).
The TNT coupled transcription-translation kit from Promega and canine pancreatic microsomes (from R. Gilmore, University of Massachusetts) were used to test for co-translational translocation of various cathepsin B isoforms into the endoplasmic reticulum. Briefly, the reactions contained 12.5 l of the TNT rabbit reticulocyte lysate, 1 l of the reaction buffer, 0.5 l of SP6 RNA polymerase, 0.5 l of a 1 mM amino acid mixture minus methionine, 0.5 l of RNasin (40 units/l), 1 g of purified plasmid DNA, 2 l of L-[ 35 S]methionine (10 mCi/ml), and nuclease-free water to a final volume of 25 l. Some translation reactions also contained 1.5 l of nuclease-treated microsomes, 1 eq/l (28). The reactions were carried out at 30°C for 1.5 h according to the manufacturer's instructions. In order to verify the uptake of the translated proteins into the lumen of microsomes, an aliquot of the translation reaction was digested with 0.1 mg/ml proteinase K for 45 min on ice. Digestion was stopped by a further 10 min incubation in the presence of 12 mM phenylmethylsulfonyl fluoride (29). Products were analyzed by SDS-PAGE and autoradiography (30).
Polysome Isolation-Polysomes were analyzed by a modification of earlier procedures (31,32). Exponentially growing A375M cells (five 175-cm 2 flasks, ϳ20 ϫ 10 6 cells/flask) were fed 10 -12 h before harvesting. PBS-washed cells were collected by scraping into 5 ml of buffer containing 30 mM HEPES, pH 7.4, 150 mM NaCl, and 11 mM glucose. Cells were pelleted at 1,000 rpm for 5 min at 4°C and resuspended in 10 ml of lysis buffer (10 mM HEPES, pH 7.4, 100 mM KCl, 5 mM MgCl 2 , 0.5% Nonidet P-40, 1 mM dithiothreitol, 50 units/ml RNasin, 100 g/ml cycloheximide). After a 10-min incubation on ice, cells were treated with cold 0.5% (w/v) sodium deoxycholate for an additional 10 min with occasional vortexing. Cell debris was removed by centrifugation at 3,000 rpm for 30 min at 4°C and 1.0-ml aliquots were applied to continuous sucrose gradients (15-45% in 13 ml of buffer which also contained 50 mM HEPES, pH 7.0, 100 mM KCl, 10 mM MgCl 2 , 0.5 mg/ml heparin, and 150 g/ml cycloheximide). The tubes were centrifuged at 87,000 ϫ g in a Beckman SW40 rotor for 3.5 h, and polysome profiles were determined from absorbance measurements at 254 nm on isolated fractions. Total RNA was isolated from each fraction and the presence of variant cathepsin B mRNAs was determined by reverse transcription linked PCR using exon junction-specific primers.
In an independent experiment, A375M cells were suspended in 20 ml of lysis buffer, or in lysis buffer containing 30 mM EDTA in place of the cycloheximide and MgCl 2 . The cells were incubated on ice for 10 min and lysed with sodium deoxycholate as above. After centrifugation, the supernatant was loaded onto a 1.0 and 2.0 M sucrose step gradient in 18 ml of lysis buffer containing either MgCl 2 and cycloheximide, or EDTA (33). The samples were centrifuged at 20,000 rpm for 22 h at 4°C in a Beckman SW28 rotor. Total RNA was extracted from the pellets and the supernatants, and variant cathepsin B mRNAs were quantified by RNase protection (14).
Transfection of COS Cells with Tagged Forms of Cathepsin B-COS 7 cells, grown to 70 -80% confluency in 35-mm tissue culture dishes, were transfected by either of two methods, with the LipofectAMINE reagent (Life Technologies, Inc., Grand Island, NY) according to the instructions of the supplier, or by the DEAE-dextran method (34). Twenty-four h after transfection the complete medium was replaced with serum-free medium and the incubation continued for an additional 24 h. Media samples were concentrated by centrifugation through Centricon 10 filters (Amicon, Inc., Beverly, MA), and diluted with an equal volume of 2 ϫ sample buffer for SDS-PAGE (30). Cells were washed with PBS and scraped into 2 ϫ sample buffer. The samples were subjected to electrophoresis on a 12% polyacrylamide gel in the presence of SDS and transferred to nitrocellulose with a miniblotter (Bio-Rad). To detect the C-peptide tagged proteins, the membrane was incubated overnight at 4°C in 1 ϫ TTBS buffer (35) containing 5% nonfat powdered milk (Carnation Co., Los Angeles, CA) and a 1:500 dilution of rabbit anti-proinsulin C-peptide antiserum (S. J. Chan, the Howard Hughes Medical Institute, the University of Chicago). The membrane was washed with 1 ϫ TTBS buffer for 30 min and incubated for 2 h in 1 ϫ TTBS buffer containing 5% nonfat milk and a 1:5000 dilution of goat anti-rabbit IgG conjugated to peroxidase. The blot was developed using the LumiGLO substrate kit (KPL, Gaithersburg, MD) for chemiluminescence detection according to the supplier's instruc-tions. Monoclonal mouse anti-Flag antibody, M2 (Eastman Scientific Imaging Systems), was used to detect the Flag epitope according to the supplier's instructions. In some experiments 2 g/ml tunicamycin was added to the cell culture medium 12 h prior to harvesting cells and media for analysis.
Immunofluorescence Microscopy-COS cells were seeded on 12-mm coverslips at a density of 10 4 cells/coverslip in 100-mm tissue culture dishes. Forty-eight h after transfection, expressed proteins were visualized by a prior procedure, which was modified for secondary immunofluorescence microscopy (36). Briefly, cells were fixed for 30 min in 4% paraformaldehyde, and rinsed three times, 10 min each, with PBS, pH 7.2. The cells were blocked in normal goat serum (NGS) blocking solution (PBS, pH 7.2, 10% NGS, 1% bovine serum albumin, 0.3% Triton X-100) for 30 min. Primary antibodies were added to the NGS carrier solution (PBS, pH 7.2, 1% NGS, 1% bovine serum albumin, 0.3% Triton X-100) at a dilution of 1:100 for both the rabbit anti-human C-peptide (Dr. S. J. Chan, the University of Chicago, Chicago, IL) and mouse anti-PDI monoclonal antibody (StressGen Biotechnologies Corp., British Columbia, Canada). These were incubated with fixed cells for 1 h, after which the cells were rinsed twice, 10 min each, in PBS, pH 7.2. The cells were then incubated for 1 h at room temperature with either fluorescein isothiocyanate-labeled goat anti-rabbit or anti-mouse antibody at a dilution of 1:250 in NGS carrier solution. The cells were rinsed three times, 10 min each, with PBS, rinsed once with distilled water, and mounted in glycerol on a glass slide. The slides were viewed in the dark with a Zeiss microscope equipped for fluorescence, and suitable fields were photographed (36).
Subcellular Fractionation-COS cells were washed with PBS and collected by scraping into homogenization buffer (0.25 M sucrose, 1 mM EDTA, pH 7.4). The cells were pelleted at 1,000 rpm for 10 min at 4°C, resuspended in 1.5 ml of homogenization buffer, and lysed by repeated passage (40 times) through a 27 1 ⁄2-gauge needle (37). The homogenates were fractionated by differential centrifugation (38). Nuclei were collected at 800 ϫ g for 10 min in a Sorvall SS34 rotor and the pellet resuspended in 100 l of homogenization buffer. The post-nuclear supernatant was centrifuged at 4°C in a Beckman TLA 103.3 rotor at 25,000 ϫ g for 20 min to collect mitochondria and lysosomes, and then at 100,000 ϫ g for 30 min to obtain the microsomal fraction. The pellets were resuspended in 100 l of homogenization buffer. The post-microsomal supernatant were taken to represent the soluble cytosolic proteins. The concentration of protein in each fraction was determined using the Bio-Rad colorimetric protein assay according to the suppliers instructions. After SDS-PAGE, individual proteins were identified by Western blotting. To determine whether the full-length or truncated cathepsin B was delivered to the lumen of a membrane bound organelle, some samples (10 g of total protein) were pretreated with 0.1 mg/ml of proteinase K as described above, with and without the addition of Triton X-100 to a final concentration of 0.15%.

Transcripts Missing Exons 2 and 3 from A375 Human Melanoma Cells
Appear to be Functional Messages-In humans, the occasional loss of exon 2 during the splicing of cathepsin B pre-mRNA results in messages that differ by 88 nucleotides in the length of their 5Ј-untranslated regions. The ratio of these messages vary among different tissues. The shorter message, which was more active in an in vitro translation assay, is the predominant form in many human tumors (14). Remarkably, some human tumors also contain transcripts that are additionally missing exon 3 which encodes the signal peptide and part of the inhibitory propeptide (14,17). In Fig. 1, reverse transcription and PCR amplification were used to confirm that messages missing exons 2 and 3 are otherwise complete. For the reverse transcriptase reaction oligo(dT) was used to prime the synthesis of cDNA from transcripts containing poly(A) tracts. For the PCR, the 5Ј primer contained 9 nucleotides on each side of the exon 1-exon 4 junction. The 3Ј primer was complementary to a region of exon 11 that is found in the 3Ј-untranslated regions of both the 2.3-and 4.0-kilobase cathepsin B mRNAs from various human tissues and tumors (14 -16). The plasmid phCB79-2, and the cDNA reverse transcribed from PC3M human prostate carcinoma cellular RNA, were included to verify the specificity of the 5Ј primer for the exon 1-exon 4 junction. RNase protection assays had previously shown that PC3M contains messages with and without exon 2, CB and CB(Ϫ2), but lacks the message CB(Ϫ2,3) in which both exons 2 and 3 are skipped (14). The plasmid phCB79-2(Ϫ2,3) was a positive control. Its 934-base pair product is indistinguishable from those obtained with the two human melanoma variants, A375M and A375P, indicating that the mRNAs missing exons 2 and 3 in these cells are otherwise complete. Also note the higher levels of this message in the more metastatic melanoma variant. The identity of this product was confirmed by hybridization with a human cathepsin B cDNA (Fig. 1B).
The distribution of cathepsin B mRNA isoforms in polysomes from A375M cells was evaluated by sucrose density gradient centrifugation and reverse transcriptase-PCR, and compared with glyceraldehyde-3-phosphate dehydrogenase (39) (Fig. 2 and Table I). The results show the three splice variants, CB, CB(Ϫ2), and CB(Ϫ2,3), to be similarly distributed in the sucrose gradient. Thus, 60 -75% of each message was present in fractions 5-12, which contain polysomes (32). Similar proportions of the abundant constitutive mRNA for glyceraldehyde-3-phosphate dehydrogenase were found in these same fractions. An RNase protection assay was also used to independently verify the presence of messages missing exons 2 and 3 in polysomes isolated by centrifugation through a sucrose step gradient (33). The results in Fig. 3 confirm the recovery of a large proportion of messages missing only exon 2 (CB(Ϫ2)), and both exons 2 and 3 (CB(Ϫ2,3)), in the polysome pellet. The full-length cathepsin B mRNA, which is a minor species in these tumor cells (14), is observed only after longer exposure times. Destabilization of polysomes by chelating Mg 2ϩ with EDTA caused the variant messages to be shifted to the supernatant fraction, which is indicative of a specific association of the transcripts with polysomes.
The Truncated Form of Cathepsin B Can be Expressed in Vivo-To further assess the translational potentials of variant messages for cathepsin B from human tumors, COS cells were FIG. 1. Detection of cathepsin B mRNAs that are missing exons 2 and 3. Total RNA was reverse transcribed using an oligo(dT) primer and amplified by PCR as described under "Experimental Procedures." Plasmid DNA was amplified directly. The 5Ј primer was directed against the exon 1-exon 4 junction and the 3Ј primer was directed against an untranslated sequence present in all full-length cathepsin B mRNAs (14). The lanes are identified as follows. M, an EcoRI/HindIII digest of -phage. From bottom to top the sizes of the first four observable bands are 831, 983, 1336, and 1584 base pairs, respectively. The other lanes are: PC3M, human prostate carcinoma (metastatic); A375M, human melanoma (metastatic); A375P, human melanoma (parental, poorly metastatic); phCB79-2, harbors a cDNA clone containing 11 exons of cathepsin B including the 798-nucleotide 3Ј-untranslated region; phCB79-2(Ϫ2, 3), identical to phCB79-2, except for missing exons 2 and 3. Panel A, after electrophoresis of PCR products on a 0.7% agarose gel and staining with ethidium bromide. Panel B, after transfer of the PCR products to a Nytran membrane, hybridization to a human 32 P-labeled cathepsin B cDNA probe and autoradiography.
transiently transfected with cDNAs encoding carboxyl-terminal-tagged forms of the enzyme. The truncated procathepsin B, ⌬ 51 CB, labeled with the insulin C-peptide (Fig. 4, CB-CP(Ϫ2.3)) was readily detected in lysates of COS cells, producing a band with the predicted molecular mass of 35 kDa. ⌬ 51 CB-CP was not present in the medium, as expected for a protein that lacks a signal peptide. In contrast, the tagged products of the vectors encoding the full-length pre-proenzyme (CB-CP and CB-CP(Ϫ2) in Fig. 4) were detected in both cell lysates and media from COS cells. Their appearance in the medium agreed with prior observations that overexpression of lysosomal pre-proenzymes can result in their secretion (40,41). The presence of additional bands in lysates from cells transfected with CB-CP and CB-CP(Ϫ2) also suggests that the tagged proenzymes can undergo post-translational transport and processing in COS cells (42). Very similar results were also obtained for the flag-tagged full-length and truncated forms of cathepsin B (data not shown).
The Truncated Form of Cathepsin B Can be Folded in Vitro-The propeptide in lysosomal enzymes is thought to guide their folding after synthesis (41,43,44). The expected protein prod-uct of the cathepsin B mRNA missing exons 2 and 3 retains 28 of the 62-amino acid propeptide. Therefore, it was of interest to determine whether ⌬ 51 CB had lost information essential to its correct folding. To initiate folding, the product of in vitro translation was treated in one of two ways: with an "oxido shuffling" buffer to facilitate disulfide bond formation, and with decreasing concentrations of guanidinium chloride, as described under "Experimental Procedures." The ability to acquire enzymatic activity is a highly stringent test for the folding of a protein into a native conformation. The results in Fig. 5 demonstrate that the product of the message missing exons 2 and 3 has the potential to acquire biological activity. Also noteworthy is the finding that the guanidinium chloride-treated product was active (33%) prior to treatment with pepsin to remove the remainder of the inhibitory propeptide. This indicates that removal of the entire propeptide is not needed for activation. In contrast, in vitro translation reactions which lacked the synthetic mRNA templates gave no activity after any of these treatments (data not shown).
Localization of the Truncated Form of Cathepsin B in COS Cells-Indirect immunofluorescence microscopy was used to  Table I.

TABLE I Distribution of variant cathepsin B mRNAs in polysomes from A375M
melanoma cells A375 melanoma cells were lysed in the presence of 100 g/ml cycloheximide and 5 mM MgCl 2 as described under "Experimental Procedures." One-ml aliquots were fractionated by centrifugation at 87,000 ϫ g on a 15-45% sucrose gradient, in the presence of cycloheximide and Mg 2ϩ , in a Beckman SW40 rotor. Fractions were collected and total RNA isolated from each. The amount of specific mRNAs were then estimated by reverse transcriptase-PCR as described in Fig. 2 and under "Experimental Procedures." Fraction 5 corresponds to the position of the 80 S ribosome peak (32). Shown are the amounts RNA or PCR product recovered in the indicated fractions as a percent of the total. The primers used to amplify the glyceraldehyde-3-phosphate dehydrogenase (GAPDH) cDNA were 5Ј-GGTCGGAGTCAACGGATTTG-GTGG (sense), and 5Ј-CCTCCGTACGCCTGCTTCACCAC (antisense) (34). RNA   locate C-peptide-tagged forms of cathepsin B when expressed in COS cells from the SV40 late promoter in pEUKC1. The full-length pre-proenzyme produced a punctate pattern of fluorescence, showing localization of the protein in peripheral lysosomes (Fig. 6, panels A and B). Some of the tagged protein was also concentrated in the juxtanuclear region of the cell. Similar patterns have been seen previously for lysosomal enzymes in COS cells (45). In contrast, the tagged, truncated cathepsin B from CB-CP(Ϫ2,3) acquired a predominantly juxtanuclear, vesicular distribution, unlike that expected for a cytosolic protein that lacks a signal peptide (Fig. 6, panels C and D). A fluorescent halo that surrounds the nucleus was also evident at higher magnification, suggesting that some of the tag may have accumulated in the perinuclear space or been bound to the surface of the nuclear envelope. Non-transfected cells and cells transfected with vector alone produced little discernible fluorescence under these same conditions (not shown). As the pattern obtained with ⌬ 51 CB-CP appeared ERlike, its localization was compared with that of the ER-resident protein, PDI (Fig. 7). The tubular and vesicular pattern of labeling observed for the truncated enzyme (panel A) clearly overlapped that of PDI (panel B), suggesting that the two were associated with the same membranous structures.
In order to substantiate the results of indirect immunofluorescence microscopy, transfected COS cells were lysed under isotonic conditions and fractionated by differential centrifugation. The fractions, which nominally correspond to nuclei, mitochondria plus lysosomes, microsomes, and cytosol (38), were analyzed for the presence of the C-peptide tagged proteins, cathepsin L, and PDI (Fig. 8). Panel A shows the results for cathepsin L. This antibody is reported to detect three major bands (46): the 43-kDa precursor which was seen in all the fractions; a 34-kDa intermediate form which was present in the microsomal fraction; and the 25-27-kDa mature form in the fraction containing mitochondria plus lysosomes. The results for PDI are shown in panel B. As expected, the majority of this marker (53 kDa) was found in the microsomal fraction (47). In panel C, the full-length, C-peptide tagged procathepsin B (43 kDa) was found in highest amounts in the 25,000 ϫ g fraction which is enriched in mitochondria and lysosomes. The proenzyme was also present in other fractions, which are likely to contain intermediary compartments in the lysosomal biosynthetic pathway. In panel D, the majority of the product from CB-CP(Ϫ2,3) (35 kDa) was found to sediment with nuclei and other membranous organelles. This is in general agreement with the results of indirect immunofluorescence microscopy in which the truncated protein was not uniformly distributed throughout the cytoplasm, but was instead associated with the nuclear envelope and a juxtanuclear network of vesicles and tubules.
The Truncated Cathepsin B Is Not Resident in the Lumen of the ER-The overlap in the staining pattern for the truncated cathepsin B with that of PDI suggested that it might be associated with the endoplasmic reticulum. To determine if the truncated enzyme contained a cryptic signal peptide-like sequence, we examined the translation of tagged and untagged messages in the presence of canine microsomes. This reconstitutes an in vitro system to study signal recognition particledependent transport into the lumen of the ER (28). Extraluminal and luminal proteins were distinguished by their sensitivity to digestion with proteinase K. As shown in Fig. 9, the full-length, C-peptide-tagged message produced two bands in the presence of microsomes. The lower 40-kDa band, which was also seen in the absence of microsomes, was sensitive to proteinase K digestion and corresponded to the extraluminal, tagged, pre-procathepsin B. The upper, 44-kDa band, which appeared only in the presence of microsomes, was largely insensitive to relatively high concentrations of proteinase K, and is likely due to a glycosylated proenzyme in the lumen of the microsome (42). In contrast, the message for CB-CP(Ϫ2,3) produced only a single band of about 35 kDa that was proteinase K sensitive, irrespective of the presence of microsomes. Identical results were obtained for the untagged forms of cathepsin B (not shown), indicating that the C-peptide did not affect the translocation of the variant forms of cathepsin B into microsomes.
Human cathepsin B contains two conserved asparagine residues, Asn-P21 in the propeptide and Asn-113 in the mature enzyme, which are sites for N-glycosylation of cathepsin B in FIG. 5. Refolding of a truncated cathepsin B produced by the  in vitro translation of mRNA missing exons 2 and 3. Capped synthetic mRNA, prepared with phCB79-2(Ϫ2,3) (14) as a template, was translated by a reticulocyte lysate. Refolding was initiated by sequential dialysis against decreasing concentrations of guanidinium chloride (GdmCl) with buffer changes at 10-or 20-min intervals, or by treatment with a glutathione disulfide/reduced glutathione "oxido-shuffling" buffer ("Experimental Procedures"). Enzymatic activity was measured by the release of 7-amino-4-methylcoumarin (AMC) from Z-Arg-Arg-NMec before and after pretreatment of the folded enzyme with pepsin to remove the remainder of the propeptide. No cathepsin B activity could be detected in the absence of capped, synthetic mRNA in the translation reaction. the lumen of the rough ER. The truncated enzyme retains the glycosylation site at Asn-113, and would likely be glycosylated were it to reach the lumen of the ER by a signal peptideindependent pathway (49 -51). This was tested by comparing the expression of C-peptide-tagged forms of cathepsin B in COS cells in the absence and presence of the oligosaccharide transferase inhibitor, tunicamycin (52). In Fig. 10, the tagged, fulllength protein (CB-CP) that was synthesized in the presence of tunicamycin traveled faster on an SDS-PAGE gel than the protein synthesized in the absence of drug. Inhibition of glycosylation by tunicamycin did not block the secretion of the proenzyme. In contrast, tunicamycin had no effect on the electrophoretic mobility of the product of CB-CP(Ϫ2,3), indicating that it was not N-glycosylated in COS cells. It is very likely, therefore, that ⌬ 21 CB-CP did not reach the lumen of the endoplasmic reticulum. An alternative explanation in which the truncated enzyme adopts a conformation that results in an altered pattern of protein glycosylation (53) is unlikely, as shown below.
Cathepsin B Is Associated with the Extraluminal Surfaces of Isolated Membrane Organelles-Immunofluorescence microscopy and subcellular fractionation suggested that the truncated cathepsin B was associated with membrane-bound organelles in COS cells. However, as documented above, it is not likely to be present in the lumen of the endoplasmic reticulum or any other compartment connected to the ER by vesicular transport. Therefore, the overlap in the distribution of the C-peptide-tagged, truncated cathepsin B and PDI in COS cells suggests that the former may be associated with the extraluminal surface of the ER. This possibility is supported by the results in Fig. 11, which shows the truncated enzyme in isolated membrane fractions was sensitive to proteinase K digestion. In contrast, the full-length enzyme was resistant to digestion by proteinase K unless the membranes were first solubilized with Triton X-100. DISCUSSION Cathepsin B is a lysosomal cysteine proteinase which may not be exclusively localized to lysosomes. Plasma membrane and secreted forms, possibly resulting from missorting in the trans-Golgi network (26), are thought to promote tumor invasion and metastasis. Cathepsin B activity and immunoreactivity have also been found in the nucleus (7) and endoplasmic reticulum (9) in some cancer cells. This finding suggests that cathepsin B could have other important and as yet unsuspected functions in cancer. Molecular mechanisms that might explain the abnormal intracellular locations of cathepsin B are largely unknown. We and others have examined transcripts of the cathepsin B gene in human tumors for evidence of an altered protein which lacks the capacity to be targeted to lysosomes. Most of the messages detected differed in their 5Ј-and 3Јuntranslated ends due to differences in mRNA splicing (14 -17) or the selection of alternative transcription start sites (17). These might contribute to the regulated expression of cathepsin B. In addition, two transcripts were found which lacked sequences encoding the signal peptide and part of the inhibitory propeptide (14,17). Despite their different presumptive origins, the coding regions in both begin with the same initiation codon in exon 4 and predict a product that lacks the signal peptide and 34 of the 62 amino acids which comprise the inhibitory propeptide. However, the methods employed to detect these putative messages, RNase protection, reverse transcriptase-PCR, and 5Ј-rapid amplification of cDNA ends, cannot distinguish between incomplete or nonfunctional transcripts and functional, full-length mRNAs. To better characterize the transcript missing exons 2 and 3, we have used oligo(dT) primers to reverse transcribe messages containing poly(A) tracts from two human melanoma cell lines, and PCR to amplify the specific messages that contain the exon 1-exon 4 junction. A single product was obtained which encompassed all the remainder of the coding region and the 3Ј-untranslated tail characteristic of normal cathepsin B mRNAs. In addition, the majority of this message was found to associate with polysomes in a manner sensitive to EDTA treatment. Collectively, these results support a claim that transcripts missing exons 2 and 3 are functional messages that can recruit ribosomes to initiate translation in the A375 melanoma cells.
An amino-terminal propeptide is often required for correct folding of the corresponding proprotein (41,43,44). However, some exceptions have been described (54,55). In yeast carboxypeptidase Y, the entire propeptide is not absolutely required for folding in vivo, provided the vacuolar signal and C-terminal region of the propeptide are retained (55). The role of the propeptide in cathepsin B folding is unknown. Our results show that the retention of 28 amino acids from the Cterminal end of the 62-amino acid propeptide is sufficient for recovery of a native-like conformation in an in vitro folding assay, as deduced by activity measurements. Furthermore, the resulting truncated proprotein has about 1/3 the Z-Arg-Arg-AMC hydrolase activity of the mature enzyme generated by removal of the entire propeptide. These results suggest that the truncated cathepsin B, when expressed in tumor cells, could have biological activity in the absence of further proteolytic processing.
The ability to detect the product of messages missing exons 2 and 3 in tumor cells required distinguishing the truncated enzyme from other forms of cathepsin B which can arise from the proteolytic processing, glycosylation, and glycolytic modification of the endogenous enzyme (42). We have used two different epitope tags to follow the synthesis and localization of the full-length and truncated cathepsin B in COS cells. The crystal structure of cathepsin B shows its C terminus to be exposed to solvent and disordered (56), suggesting that a Cterminal peptide extension might not interfere with protein folding. This was confirmed by the finding that when overex-pressed in COS cells, the secreted flag-and C-peptide-tagged procathepsin B acquired enzymatic activity after pepsin cleavage of the inhibitory propeptide. 2 In addition, the flag peptide and C-peptide are not likely to directly affect protein targeting. The flag peptide is highly polar and resembles no known protein targeting signal. The C-peptide is poorly conserved among species, it does not cause retention of proinsulin when expressed in a non-insulin producing cell, and recombinant insulins that lack the C-peptide are efficiently delivered to secretory granules in insulin-producing cells (57,58). It was recently reported that the isolated human C-peptide can produce biological effects in the diabetic rat in a manner which suggests that it could be membrane active (59). However, its polarity, and the kinetics with which it is metabilized, would seem to argue against an association with cell membranes (60). The Flag-and C-peptide-tagged full-length and truncated cathepsin B accumulated to similar levels in COS cells, confirming that the message missing exons 2 and 3 is translationally active in vivo, and that its product is sufficiently stable in COS cells to be readily detected. The presence of the tagged, fulllength proenzyme in the media results from partial missorting of overexpressed lysosomal enzymes to the secretory limb of the exocytic-lysosomal biosynthetic transport pathway (40,41). The truncated cathepsin B was not secreted, as would be predicted of a protein that is missing the signal peptide.
To begin to identify the potential function of the truncated cathepsin B in cancer, it was necessary to determine its cellular location. The truncated cathepsin B lacks the N-terminal signal peptide and was expected to be directed to the cytyosol of COS cells where it should have produced a diffuse cellular fluorescence. Unexpectedly, indirect immunofluorescence microscopy showed most of the label to be concentrated in regions of the cell proximal to the nucleus and to appear to be associated with tubules and interconnected vesicles. Staining of the nuclear envelope was also evident. This distribution of label 2 S. Mehtani and A. Frankfater, unpublished results. guage needle, and the lysates were fractionated by differential centrifugation to obtain: nuclei, 800 ϫ g pellet; mitochondria ϩ lysosomes, 25,000 ϫ g pellet; microsomes, 100,000 ϫ g pellet; and cytosol, 100,000 ϫ g supernatant. The pellets were suspended in 100 l of buffer and the individual fractions were analyzed by 12% SDS-PAGE and immunoblotting. Each lane contained 10 g of protein.
The distributions of the CB-CP (C) and ⌬ 51 CB-CP (D) were compared with those of the lysosomal marker, cathepsin L (A) and the ER marker, protein disulfide isomerase (B). For cells transfected with CB-CP, the total amount of protein in each fraction was: nuclei, 280 g; mitochondrial ϩ lysosomes, 358 g; microsomes, 164 g; and cytosol, 480 g. For cells transfected with CB-CP(Ϫ2,3), the total amount of protein in each fraction was: nuclei, 328 g; mitochondria ϩ lysosomes, 378 g; microsomes, 337 g; and cytosol, 686 g. The subcellular distribution of PDI and cathepsin L is shown for cells transfected with CB-CP(Ϫ2,3). Similar results for the two markers were also obtained with CB-CP transfected cells.
FIG. 9. In vitro transcription and translation of cDNAs encoding the variant forms of cathepsin B. cDNAs encoding the C-peptide tagged, full-length cathepsin B (Ϫ2CP) and truncated cathepsin B (Ϫ2,3CP), under the control of an SP6 promoter, were used in a coupled transcription-translation assay as described under "Experimental Procedures," and the products analyzed by 12% SDS-PAGE and fluorography. The cDNA for the C-peptide tagged, truncated cathepsin B gave only a single 35-kDa product, whether or not microsomes (ϩMIC) were present in the translation reaction. Its sensitivity to proteinase K (ϩPK, 0.1 mg/ml) indicated its failure to gain access to the interior of the microsomes. In contrast, the cDNA for the tagged, full-length protein produced two products. The upper, 44-kDa band, was only generated in the presence of canine microsomes, was proteinase K insensitive, and likely represents the normal glycosylated form of procathepsin B transported into the lumen of the microsomes. The 40-kDa band was proteinase K sensitive and represents the nonglycosylated, pre-procathepsin B which failed to gain entry to the microsomal space. resembled that of the ER resident protein, PDI. In agreement with the results of immunofluorescence microscopy, differential centrifugation of COS cell lysates demonstrated that the majority of the tagged, truncated cathepsin B co-sedimented with nuclei and other membranous organelles.
An analysis of CB(Ϫ2,3) with PSORT 3 failed to detect any sequence that could function as a cryptic signal peptide for its transport into the lumen of the endoplasmic reticulum. In agreement, the truncated enzyme was not taken up by canine microsomes in an in vitro translation reaction. In addition, it is also unlikely that the truncated enzyme was targeted to the lumen of the ER by a putative signal recognition particle and signal peptide independent mechanism (50,51). First, when the synthesis of CB-CP(Ϫ2,3) was examined in COS cells in the presence and absence of tunicamycin, no evidence could be obtained for glycosylation at Asn-113, as would have been expected if the truncated enzyme had been directed to the lumen of the ER. Second, treatment of membrane fractions from COS cells with proteinase K indicated that CB-CP(Ϫ2,3) was accessible to proteolytic digestion and unlikely to be in the interior of the isolated organelles. Therefore, it appears that the association of the truncated form of cathepsin B with membranous fractions from COS cells may involve interactions with the cytosolic surfaces of these organelles.
Observations of a novel form of cathepsin B, which is associated with intracellular membranes in human tumors, suggests the possibility that this enzyme may have other important functions in cancer beyond its likely participation in cell invasion and metastasis. Thus, a number of important regulatory proteins are also associated with the cytosolic surface of the ER and other organelles. These include the anti-apoptotic protein Bcl-2 (61, 62), the pro-apoptotic protein, Presenilin 2 (63,64), and members of the Ras superfamily of small GTPbinding proteins. The later are involved in the regulation of diverse biological functions which include the assembly of actin stress fibers and focal adhesions (65), the expression of genes involved in cellular proliferation and transformation (66,67), the maintenance of membranous organelles, and the transport of vesicles between them (65,68). It may be relevant that limited proteolysis of Presenilin 2 produces C-terminal fragments (69) which can act in a dominant negative manner to inhibit apoptosis (64). It is also interesting to note that v-Ha-Ras and c-Ha-Ras bind with high affinity to cathepsin B (K I values of 48 and 165 nM, respectively) (48,70).
In conclusion, we have shown that mRNAs missing exons 2 and 3, which encode a truncated form of cathepsin B, are likely to be functional messages. They contain the remainder of the propeptide and mature enzyme, and are found to associate with polysomes in the A375 human melanoma. The product of in vitro translation retains the capacity to fold into a native-like protein which is partially active in the absence of pepsin pretreatment to remove the remainder of the propeptide. Thus, alternative splicing may partly account for observations of forms of cathepsin B in human tumors that have unexplained locations, and may be greater in size and/or stability than the normal, mature enzyme. Immunofluorescence microscopy revealed a distribution for the truncated enzyme which appeared tagged with the C-peptide. Cells were either left untreated (Ϫ), or were exposed to 2 g/ml of the N-glycosylation inhibitor, tunicamycin, 12 h prior to harvesting the cells and media. Cell lysates and media samples were then analyzed by electrophoresis on a 12% SDS-PAGE gel and immunoblotting. Tunicamycin produced a significant decrease in the size of the full-length cathepsin B (from 46 to 41 kDa) but had no effect on the size of the truncated protein (35 kDa).

FIG. 11. Effect of proteinase K on the variant forms of cathepsin B in membrane fractions from COS cells. COS cells were
transfected with the cDNAs for the C-peptide tagged, truncated (upper panel) and full-length (lower panel) forms of cathepsin B. After 48 h, the cells were disrupted by repeated passage through a 27 1 ⁄2 guage needle, and fractions were collected by centrifugation at 800 ϫ g (nuclear pellet) and 100,000 ϫ g (high speed membrane pellet and supernatant). Proteinase K complete degraded the truncated enzyme present in nuclear and high speed membrane pellets. In contrast, complete digestion of the full-length enzyme in the high speed pellet required addition of the detergent, Triton X-100 (ϩTX-100).
to overlap that of the ER resident protein, PDI. The absence of glycosylation and the sensitivity of CB-CP(Ϫ2,3) in isolated membrane fractions to proteinase K digestion suggested that it was located on the cytosolic surface of intracellular membranes. In these locations the truncated protein could be expected to interact with important regulatory proteins such as Bcl-2 and Ras-related proteins. Further work is required to determine it the truncated tumor form of cathepsin B could act to alter the apoptotic program in tumor cells or to modify other important regulatory pathways known to be dysfunctional in cancer.