INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

In the past decade, new findings in the fields of cell biology and molecular genetics of cancer have revealed a deregulation of the cell cycle as a critical event for the onset of tumorigenesis. Progression through the cell cycle, the sequence of events between two cell divisions, is governed by the actions of positive and negative regulators in the eukaryotic cell. The mammalian cell cycle is positively regulated by heterodimeric complexes of stable cyclin-dependent kinases (CDKs) and unstable regulatory cyclin subunits (1, 51). Mitogenic stimuli result in the phosphorylation and thereby activation of cyclin-CDK complexes by CDK-activating kinase (17, 22, 35). The activated cyclin-CDK complexes in turn phosphorylate substrates such as the retinoblastoma protein (pRb) throughout the cell cycle (16, 36, 51).

The connection between cyclins and cancer has been substantiated with G₁-type cyclins (25-27, 55). Cyclin E, a G₁ cyclin which forms complexes with CDK2, is essential for S-phase entry (41, 47) and has a profound role in oncogenesis (29, 31). In dividing cells, the expression of cyclin E increases to a maximum at the G₁/S transition, with a peak expression level near the restriction point (13, 34). When coupled to CDK2, the active kinase follows cyclin D-CDK4 in progressively phosphorylating pRb, releasing it from members of the E2F family (15). As E2F is released it activates a number of S-phase genes, including cyclin E and E2F-1 (18, 40). This state of readiness to enter S phase requires cyclin E or cells will arrest at this point in the cell cycle (42, 57). For example, the functional knock out of cyclin E by injection of anti-cyclin E antibodies into fibroblast cells causes cell arrest in the G₁ phase (42). Conversely, the overexpression of cyclin E protein causes acceleration through G₁ along with a decreased cell size (42, 48). In addition to its requirement for DNA synthesis, cyclin E also plays a key role in senescence (14), development (7, 48), and modulation of downstream signals involving pRb (58) and E2F (18, 40). Normal expression and activity of cyclin E play a crucial role in cell proliferation, and any defects in its expression could have a critical effect in oncogenesis.

We and others have reinforced the linkage between oncogenesis and cyclin E by correlating the altered expression of cyclin E to the loss of growth control in breast cancer (8, 30, 32). Furthermore, several tumor cohort studies (reviewed in reference 55) have documented a strong correlation between cyclin E overexpression and poor patient outcome (31, 44) and lack of estrogen receptor expression (21, 39, 49). In addition, patients with high cyclin E levels in their tumors have a significantly increased risk of death and/or relapse from breast cancer even if they are node negative (39, 44). Additionally, examination of the oncogenic potential of cyclin E in transgenic mice under the control of the bovine -lactoglobulin promoter revealed that lactating mammary glands of the transgenic mice overexpressing cyclin E contain hyperplasia and over 10% also develop mammary carcinomas (6). Lastly, constitutive overexpression of cyclin E (but not cyclin D1 or A) in both immortalized rat embryo fibroblasts and human breast epithelial cells results in chromosome instability (53). Collectively, these data provide strong support for the role of cyclin E in breast cancer tumorigenesis.

In normal cells, cyclin E is expressed precisely when needed and then is rapidly degraded. In tumor cells, cyclin E expression is different in several ways. First, amplification of the cyclin E gene and overexpression of cyclin E mRNA by 64-fold have been reported in a subset of breast cancer cell lines (8, 32). Second, the protein expression and associated kinase activity are no longer cell-cycle regulated (21, 50). Third, in addition to the typical 50-kDa protein, a pattern of low-molecular-weight (LMW) forms ranging in size from 49 to 33 kDa is also detected by immunoblotting (23, 31, 32). However, despite the tumor-specific expression of LMW forms of cyclin E and the compelling prognostic evidence, very little is known about what gives rise to these isoforms.

Previously we reported that although cyclin E is subject to extensive alternative splicing, these cyclin E mRNA variants do not account for the LMW forms of the protein observed in tumor cells (43). Here we show that the generation of these tumor-specific LMW forms of cyclin E is predominantly derived from proteolytic processing of the full-length cyclin E. Mass spectrometry analysis revealed that the full-length cyclin E, which is 50 kDa and is found predominantly in both normal and tumor cells, is the EL form (i.e., the 15-amino-acid [aa] elongated variant of cyclin E [42]). By using site-specific mutations and transient transfection, we have also identified two protease-sensitive domains in cyclin E. Four of the five LMW forms of cyclin E in tumor cells are accounted for by proteolysis at these two sites, with posttranslational modification of the two proteolytic forms creating closely migrating doublets. Mutation analysis of amino-acyl residues within the proteolytic cleavage domain suggests that cyclin E is a substrate for the elastase class of proteases. In addition to these proteolytically derived bands, a fifth band appears from an alternate translation start at M46 of cyclin EL. More importantly, the cyclin E LMW forms are hyperactive, as they display stronger activity toward their native substrates, such as histone H1 and GST-Rb, than the full-length form does. Additionally, these LMW isoforms of cyclin E are localized to the nucleus and can stimulate the cell to progress through the cell cycle more effectively than does the full-length form. The proteolysis of cyclin E described here is a newly discovered level of tumor-specific cyclin regulation distinct from the ubiquitin-proteasome pathway.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Materials, cell lines, culture conditions, and transfections. Serum was purchased from HyClone Laboratories (Logan, Utah), and cell culture medium was obtained from Life Technologies, Inc. (Grand Island, N.Y.). All other chemicals used were reagent grade. The culture conditions for the normal 70N and 81N cell strains, immortalized MCF-10A cell line, and breast cancer MCF-7, MDA-MB-157, MDA-MB-231, and MDA-MB-436 cell lines were described previously (21). The 76N-E6 cell line (gift from V. Band, Tufts Medical Institute, Boston, Mass.) was immortalized and cultured as described previously (1, 2). All cells were cultured and treated at 37°C in a humidified incubator containing 6.5% CO₂ and maintained free of mycoplasma as determined by Hoechst staining (24). Transfection by electroporation was carried out on the tumor cell line MDA-MB-157 at 0.27 kV and 960 µF capacitance. For each condition, 1 × 10⁷ cells were suspended in 0.5 ml of media (serum free), with 40 µg of the indicated plasmid in a 0.4-cm gap cuvette. Following transfection, cells were plated with complete medium and harvested 24 h posttransfection for analysis. Transfection with the FuGene Transfection Reagent was carried out in 70N, 81N, MCF-10A, 76NE6, and MDA-MB-436 cells according to the manufacturer's instructions (Roche Molecular Biochemicals, Indianapolis, Ind.). Cells were harvested for analysis 24 h posttransfection.

Sorting and analysis of transfected cells. Following cotransfection of 81N cells with cyclin EL-FLAG and green fluorescent protein (GFP) constructs, cells (1 × 10⁷) were harvested and resuspended in sterile ice-cold phosphate-buffered saline (PBS) and subjected to sorting with a Vantage fluorescence-activated cell sorter (FACS) from Becton Dickinson with a 488-nm argon ion laser. The GFP fluorescence was measured using a band-pass filter at 530/30 nm. Cell sorting was performed under sterile conditions. The sorting gates were set to sort the GFP-negative cells in the left tube and the GFP-positive cells in the right tube. Sorted cells were collected in a tube containing growth medium. The GFP transfection efficiency was 30%. GFP-positive cells were gated by separating nonfluorescent cells, as generated by mock transfection, from the fluorescent cells. On average, 3 × 10⁶ cells were collected in the GFP-negative tube and 9 × 10⁵ cells were collected in the GFP-positive tube. After sorting, an aliquot of the sorted cells was run on the FACS Vantage to check the purity of the two populations. A purity of >99% was noted for each group. The remaining cell suspensions for each condition were spun down and resuspended in ice-cold PBS and used for three different analyses. DNA content analysis was performed with 1 × 10⁵ of the GFP-positive cells collected from each transfectant group. FLAG immunohistochemical analysis was performed on 5 × 10⁴ GFP-positive and -negative cells collected from each transfectant group, following their plating on glass coverslips for 16 h. Lastly, Western blot analysis was performed on the remaining sorted and unsorted cells as described below.

Western blot, immunoprecipitation, and H1 kinase analysis. Cell lysates were prepared and subjected to Western blot analysis as previously described (45). Briefly, 50 µg of protein from each condition was electrophoresed in each lane of a 10% (cyclin E, FLAG, cyclin D1, and actin) or 13% (CDK2, CDK4, p21, p27, and actin) sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) gels and transferred to Immobilon P overnight at 4°C at 35 mV constant voltage. The blots were blocked overnight at 4°C in BLOTTO (5% nonfat dry milk in 20 mM Tris, 137 mM NaCl, 0.25% Tween, pH 7.6). After six 10-min washes in TBST (20 mM Tris, 137 mM NaCl, 0.05% Tween, pH 7.6), the blots were incubated in primary antibodies for 3 h. Primary antibodies used were cyclin E HE-12 and cyclin D1 monoclonal antibodies (Santa Cruz Biochemicals, Santa Cruz, Calif.) at 1 µg/ml, anti-FLAG polyclonal antibody (Santa Cruz Biotechnology, Santa Cruz, Calif.) at 0.25 µg/ml, monoclonal antibodies to p27, CDK2, and CDK4 (Transduction Laboratories, Lexington, Ky.) and monoclonal antibody to p21 (Oncogene Research Products/Calbiochem, San Diego, Calif.), all at 0.1 µg/ml in BLOTTO, and actin monoclonal antibody (Roche Molecular Biochemicals) at 0.63 µg/ml in BLOTTO. Following primary antibody incubation, the blots were washed and incubated with goat anti-mouse immunoglobulin-horseradish peroxidase conjugate at a dilution of 1:5,000 in BLOTTO for 1 h and finally washed and developed with the Renaissance chemiluminesence system as directed by the manufacturer (NEN Life Sciences Products, Boston, Mass.).

In vitro transcription and translation assays. To transcribe and translate the cyclin EL-FLAG constructs cloned in the pcDNA3.1 vector, we used the TNT Coupled Reticulocyte Lysate system (Promega, Madison, Wis.). Briefly, 1 µg of pCDNA3.1 plasmid containing either the cyclin EL-FLAG inserts or no insert was added to rabbit reticulocyte lysate in the presence of T7 RNA polymerase and 1 mM complete amino acid mixture in a total volume of 50 µl and incubated at 30°C for 90 min. One-microliter volumes of each of the translation products, containing both the rabbit reticulocyte lysate and the synthesized cyclin E protein, were then separated on SDS-PAGE gels and subjected to Western blot analysis using either a monoclonal antibody to cyclin E (clone HE-12) or a polyclonal antibody to FLAG (both from Santa Cruz Biochemicals).

Plasmid construction. Block deletions (e.g., from aa 67 through 69, creating DEL6769) and amino acid substitutions (e.g., methionine substituted by alanine at position 16, creating M16A) were prepared using quick-change site-directed mutagenesis (5) or PCR cloning. The plasmid pCDNA 3.1, containing the EL form of cyclin E and a FLAG epitope (boldface) 5'-GCAAGCTTTTCACTTGTCATCGTCGTCCTTGTAGTCCGCCATTTCCGGCCCGCT-3' at the carboxy terminus (pCDNA 3.1 EL-FLAG, [23]) was used as the starting template for all of the mutations. The following oligonucleotides were synthesized along with their antiparallel version to form a double-stranded DNA pair containing the desired mutation centered in the sequence: M16A, AAGGAGCGGGACACCGCGAAGGAGGACGGCGG; M46A, CCAGATGAAGAAGCGGCCAAAATCGAC; V67D, GACAATAATGCAGACTGTGCAGACCCC; C68D, AATAATGCAGTCGATGCAGACCCCTGC; A69D, AATGCAGTCTGTGATGACCCCTGCTCC; D70P, GCAGTCTGTGCACCCCCCTGCTCCCTG; D70A, GCAGTCTGTGCAGCCCCCTGCTCCCTG; DEL6769, TGGGACAATAATGCAGACCCCTGCTCCCTG; DEL6469, AGCCAGCCTTGGGACGACCCCTGCTCCCTG; DEL7075, CAATAATGCAGTCTGTGCACCCACACCTGACAAAG; DEL1621, TGCGAAGGAGCGGGACACCGCGGAGTTCTCGGCT; DEL2227, AAGGAGGACGGCGGCTCCAGGAAGAGGAAG; DEL2833, GAGTTCTCGGCTCGCAACGTGACCGTTTTT; DEL3439, AGGAAGAGGAAGGCACAGGATCCAGATGAA; DEL4045, GTGACCGTTTTTTTGATGGCCAAAATCGAC. PCRs were performed with 1 µg of template (pCDNA 3.1-EL-FLAG) using Pfu polymerase (Stratagene) in the buffer provided and 125 ng of each oligonucleotide pair for 16 cycles of 95°C for 30 s, 55°C for 60 s, and 68°C for 12 min. The methylated template plasmid (from growth in Escherichia coli host DH10B) was digested with DpnI restriction enzyme at 37°C for 1 h. The DpnI-resistant, unmethylated PCR product containing the mutation was desalted on a G-50 Sephadex spin column and electroporated into the E. coli host DH10B. The E. coli cells were plated on ampicillin plates, and colonies were picked and checked for successful mutation by restriction analysis and sequencing. The following mutations were made with a different strategy using oligonucleotides to PCR amplify the plasmid pLXSN containing the cyclin EL (42) cDNA as template. The DEL7681 forward primer (GTCTGTGCAGACCCCTGCTCCCTGATCGATGATGACCGG) was used with the reverse primer CE1247 (AGCAAACGCACGCCTCCGCTGCAACA). The DEL8287 (ACACCTGACAAAGAACCAAACTCAACGTGCAAG) and DEL8893 (ACACCTGACAAAGAAGATGATGACCGGGTTTACCCTCGGATTATTGCA) forward primers were combined with the bridge forward primer CE196 (GCAGTCTGTGCAGACCCCTGCTCCCTGATCCCCACACCTGACAAAGAA) and used with the reverse primer CE1247. The PCRs were performed with 10 ng of template (pLXSN-cyclin E EL) using Taq polymerase in the buffer provided for 18 cycles of 95°C for 30 s, 65°C for 20 s, and 72°C for 30 s. The PCR products were ligated into the T/A cloning vector pCRII and transformed into DH10B, and colonies were selected as above. Plasmids containing the correct mutation were digested with the restriction enzymes BsgI and AgeI, and the fragment was ligated into the same site of pCDNA 3.1 EL-FLAG. Transformation and colony selection were as described above.

Expression of the LMW forms of cyclin E in insect cells. Plasmids pVLCDK2, pVLcycE, pVLcycET1, and pVLcycET2 were constructed by subcloning the full-length cDNA fragments for CDK2, the EL form of cyclin E, and its truncated forms (i.e., T1 and T2), respectively, into the multiple cloning sites downstream of the polyhedrin promoter of baculovirus transfer vector pVL1392 (PharMingen, San Diego, Calif.) to generate the high-expression vector systems for these proteins. Recombinant viruses were produced by cotransfection of the above purified plasmids, separately, with linearized BaculoGold virus DNA (PharMingen) into Sf9 insect cells. For the resultant recombinant viruses, B.G.CDK2, B.G.cycE, B.G.cycET1, and B.G.cycET2, titers were determined by end-point dilution assay and amplifications were performed by infecting the Sf9 insect cells at a low multiplicity of infection (MOI) (0.7), maintaining the high percentage (99%) of the recombinant viruses in the large stock and therefore assuring the high expression of the desired proteins. The supernatants of viruses amplified were harvested around 60 h postinfection (p.i.). Protein extracts were subjected to Western blotting, immunoprecipitation, and H1 kinase analysis, as described above.

Porcine pancreatic elastase digestion. Plasmids containing the full-length cyclin EL, block deletions A to E, A', and point mutations of cyclin E FLAG-tagged constructs as well as the vector-alone construct were translated in vitro using a TNT kit (Promega) as described above. A total of 10⁴ U of the porcine pancreatic elastase (Sigma Biochemicals) was mixed with 0.25 µl of the TNT reaction mixture (i.e., 5 µl of a 1:20 dilution) in the elastase reaction buffer (50 mM Tris [pH 8.5], 250 mM NaCl) and incubated at 30°C for 5 min. At the end of incubation the entire reaction mixture (total volume, 20 µl) was subjected to SDS-PAGE and Western blot analysis with anti-FLAG antibody.

Cyclin E purification and MS. Cyclin E was purified from 500 mg of MDA-MB-157 cell extract prepared by sonication and centrifugation as described previously (45). The extract was precleared with three successive incubations with 1 ml of Sepharose CL6B rotated at 4°C for 2 h. After preclearing, the extract was incubated overnight with 1 ml of the monoclonal antibody HE-111 cross-linked to Sepharose (Santa Cruz). The antibody resin was washed with Buffer A (50 mM Tris [pH 8.0], 1 mM EDTA, 0.5 M NaCl, 0.5% Nonidet P-40), Buffer B (50 mM Tris [pH 8.0], 1 mM EDTA, 0.15 M NaCl, 0.1% SDS, 0.5% Nonidet P-40), and Buffer C (50 mM Tris [pH 8.0], 1 mM EDTA, 0.15 M NaCl, 0.5% Nonidet P-40) and then eluted by boiling in SDS-polyacrylamide, reducing, sample buffer. The eluted sample was loaded onto a 10% polyacrylamide gel, electrophoresed, and silver stained (omitting the glutaraldehyde step [38]). A gel slice was excised and a 1-mm fragment was electrophoresed again and evaluated by Western blotting to confirm the presence of cyclin E. The gel slice was digested using modified trypsin from Boehringer Mannheim. Reactive cysteines were reduced with 10 mM dithiothreitol (DTT) and blocked with iodoacetamide (aminocarboxy methylation). The gel slice was dehydrated in acetonitrile for 30 min and then dried in a vacuum. The slice was rehydrated in 10 mM DTT, 100 mM ammonium bicarbonate at 55°C for 1 h. This solution was replaced with 55 mM iodoacetamide, 100 mM ammonium bicarbonate in the dark for 45 min at room temperature with vortexing. The slice was then washed with 100 mM ammonium bicarbonate for 10 min, dehydrated in acetonitrile for 30 min, and dried in a vacuum. One more round of hydration-dehydration was performed, and the gel slice was swollen in digestion buffer (50 mM ammonium bicarbonate, 5 mM CaCl_2, 12.5 ng of trypsin/µl) and digested at 37°C for 18 h. The peptides were extracted and analyzed by matrix-assisted laser desorption-ionization time-of-flight mass spectrometry (MALDI-TOF MS) as follows. The gel slice was extracted twice with 0.1% trifluoroacetic acid (TFA) and 70% acetonitrile in a small volume for 45 min. The sample was concentrated using a Speed-Vac. The combined extracts were loaded onto a C₁₈ Zip Tip (Millipore), washed with 0.1% TFA, and eluted with 10 to 70% acetonitrile. One microliter of the eluted sample was spotted on target along with 1 µl of matrix (10 mg of alpha-cyano-4-hydroxycinnamic acid/ml) solubilized in 70% acetonitrile and analyzed on a Bruker Reflex delayed-extraction MALDI-TOF apparatus.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

MS identifies the predominant cyclin E as the EL form. The characteristic pattern of cyclin E processing found in tumor cells and tumor tissues includes a number of LMW forms ranging in size from 33 to 45 kDa, in addition to the 50-kDa full-length form which is found in both normal and tumor cells (31). We used MS to identify the full-length 50-kDa cyclin E in tumor cells by determining the N terminus directly from tryptic peptides of the purified cyclin E protein. To this end, we purified cyclin E from 2 × 10¹⁰ MDA-MB-157 cells (500 mg of total cellular extract) with a cyclin E monoclonal antibody (clone HE-111) affinity column. The purified cyclin E was eluted from the column and resolved by SDS-PAGE. After silver staining the gel, the band corresponding to the full-length (i.e., 50 kDa) cyclin E was excised, reduced, alkylated, and digested with trypsin according to established procedures (52). The purified cyclin E tryptic peptides were extracted from the gel slice and analyzed by MALDI-TOF MS, which yielded 10 peptides with internally calibrated masses matching predicted cyclin E peptides from both the N- and C-terminal domains of cyclin E (Fig. 1). The mass spectrum of the purified cyclin E tryptic digests as analyzed by MALDI-TOF is depicted in Fig. 1A. The alignments of the 10 peptides to cyclin E are shown underneath the schematic representation of cyclin EL (Fig. 1B). The shaded region defines the unique cyclin EL sequences not found in the 45-kDa cyclin E. Cyclin EL is the 15-aa elongated variant of cyclin E identified in 1995 (42), while the 45-kDa cyclin E corresponds to the gene identified in the yeast complementation studies first identifying cyclin E (33, 34). Four of these tryptic peptides (numbers 5, 7, 9, and 10) contain sequences from the N terminus of the EL form of cyclin E which are not present in the 45-kDa form of cyclin E (Fig. 1C). All four peptides contain an internal methionyl residue from position 16 (M16) of the intact EL form of cyclin E, which could only derive from the EL cyclin E and not the 45-kDa cyclin E, with a predicted translation start at M16. This MS analysis of cyclin E purified from breast cancer cells confirms that the 50-kDa full-length cyclin E found in these cells is the cyclin EL. Therefore, we selected the EL form of cyclin E as the backbone for all subsequent expression vectors.

View larger version (29K): [in this window] [in a new window]   FIG. 1.   MS identified cyclin EL as the 50-kDa cyclin found in tumor cells. (A) MALDI-TOF analysis of the tryptic digests of the purified cyclin E band excised from a silver-stained PAGE gel. The major identified peaks all correspond to cyclin E when compared to the available databases through Protein Prospector. Cyclin E peaks are designated with a mass and a number. Unlabeled peaks correspond to trypsin self-digestion products or were not identified. (B) Schematic representation of cyclin EL and cyclin E. Solid bars show alignments of peptides to different regions of cyclin EL. The shaded region is unique to cyclin EL. (C) Peptide sequences obtained from MALDI-TOF analysis of cyclin E tryptic digests. The numbering corresponds to the peaks identified in panel A. The internally calibrated masses submitted to Protein Prospector are within 0.06 Da of the peptides matched.

Block deletions and amino-acyl substitutions define cyclin E proteolysis in tumor cells. Previously, we showed that the processing of cyclin E required to generate the LMW forms detected only in tumor cells occurs posttranslationally (23). To identify the exact position where cyclin E is proteolytically processed, we transiently transfected the full-length and selectively deleted and/or mutated forms of C-terminus epitope-tagged, cyclin EL-FLAG into tumor cells and analyzed their expression by Western blotting using a FLAG antibody (Fig. 2B). The results revealed that in the MDA-MB-157 tumor cell line the exogenously expressed FLAG-tagged cyclin EL (Fig. 2B, lane EL) is processed very similarly to the endogenous cyclin E (Fig. 2B, lane E-Tumor), generating the same Western blotting pattern of bands. We used a numbering system to facilitate the analysis of the cyclin E LMW forms. The full-length 50-kDa form of cyclin E is EL1. Below EL1 is a doublet migrating at 45 and 44 kDa and designated as EL2 and EL3. A single band migrating at 40 kDa is EL4, and another doublet below this at 35 and 33 kDa is designated EL5 and EL6. The Western blotting patterns of cyclin EL-FLAG isoforms EL1 to -6 were comparable in other transfected breast cancer cell lines (see below).

View larger version (43K): [in this window] [in a new window]   FIG. 2.   Deletional analysis of cyclin E identifies sequences responsible for generating the LMW forms. (A) Schematic representation of cyclin EL, highlighting the sequences used for deletional analysis. Five block deletions of 6 aa each were made in a 30-aa region (aa 64 to 93) designated A-E. The block deletion A' contains a 3-aa deletion spanning aa 67 to 69. T2 represents the start of cyclin EL-Trunk2. All constructs were made with an 8-aa FLAG peptide fused at the carboxy terminus. (B) The block deletions A to E and A' as well as the cyclin EL, Trunk 2, and vector alone were transiently transfected into MDA-MB-157 cells. Protein was extracted 24 h posttransfection and subjected to Western blot analysis with an anti-FLAG antibody. Note that the block deletions A to E are each 6 aa shorter than EL and migrate faster on the gel. The band above cyclin EL1 is protein that is cross-reactive with the FLAG antibody and can be seen in all lanes including the vector-alone lane. On the left of the Western blot, the relative mobilities of the LMW bands are labeled EL1 to -6. Shown on the right is the cyclin E Western blot of the total cell extract from the tumor cell line MDA-MB-157 (untransfected), with the LMW bands (EL1 to -6) labeled. (C) Transient transfection of cyclin EL-FLAG constructs harboring the indicated point mutations in aa 67 to 70, followed by Western blotting with FLAG antibody. (D) The EL, M16A, M46A, and Trunk 2 plasmid constructs with the FLAG tag were transiently transfected into MDA-MB-157 cells. Protein was extracted 24 h posttransfection and subjected to Western blot analysis with an anti-FLAG antibody. The same plasmids were also transcribed or translated in vitro (TNT assay) and subjected to SDS-PAGE and Western blot analysis in the indicated lanes. The LMW band labeled EL4 is not translated in vitro or in transient transfection of MDA-MB-157 cells when the methionyl at position 46 is mutated to an alanyl residue (lanes 5 and 6). Vector TNT and vector MDA-MB-157 represent in vitro-translated and -transfected empty vectors, respectively. (E) Transient transfection of cyclin EL-FLAG constructs harboring the indicated block deletions in aa 16 to 45, followed by Western blotting with FLAG antibody.

The cyclin E 45-kDa form is not expressed in tumor cells. We suspected that bands EL2 and -3 represented protein translated from the M16 start site originally thought to be the primary form of cyclin E (33, 34). However, we were surprised when an M16A substitution did not affect the expression of bands EL2 and -3 (Fig. 2D). For these analyses, we compared the in vitro-translated products of cyclin EL1 and of cyclin EL1 mutated at M16 or M46 to an alanyl residue, with the proteolytically processed protein products from transfected tumor cells (Fig. 2D). When the protein from EL1-FLAG-transfected MDA-MB-157 was run side by side with the in vitro-translated cyclin E products from the same vector, the bands EL2 and EL3 clearly migrated faster than the in vitro translation products from the M16 translation start site (Fig. 2D, compare lanes 1 and 2). This difference in migration between the M16 in vitro-translated band and EL2 and -3 in the tumor cells was a hint that M16 may not be required to generate EL2. EL2 is the LMW band we suspected was the 45-kDa cyclin E wild type (33, 34). We then replaced the methionyl residues at positions 16 and 46 with an alanyl residue, knocking out any alternative translation initiation from these sites (Fig. 2D). The M16A and M46A substitutions clearly eliminated translation starts from these sites when translated in vitro (lanes 3 and 5). However, only the M46A substitution eliminated EL4 from transfected MDA-MB-157 cells (lane 6), while the M16A substitution had no effect on the generation of EL2 (lane 4). Even though EL2 and -3 migrate closely to the cyclin E translated from M16, these bands are most likely created by proteolysis rather than by alternate translation at M16. The in vitro mutagenesis of cyclin E shown here and the MS data (Fig. 1) collectively provide strong evidence that the high-molecular-weight, full-length form of cyclin E found in tumor cells is translated from the cyclin EL mRNA (42) and not the cyclin E wild-type form initially cloned as this G₁ cyclin (33, 34). Furthermore, while the EL5 and -6 and EL2 and -3 LMW forms are due to proteolytic processing, EL4 can be accounted for by an alternate translation start site, as shown by the elimination of this band after the mutation of the methionyl at residue 46 into alanyl, both in transfected cells and when the protein is translated (i.e., TNT assay) in vitro (Fig. 2D, lanes 5 and 6).

The LMW forms of cyclin E are hyperactive. To examine the biochemical activity of the LMW forms of cyclin E compared to that of the full-length form, we used three different constructs of cyclin E that were FLAG tagged at the C terminus in transfection assays using normal and tumor cells. The three constructs are cyclin EL-FLAG, Trunk 1, and Trunk 2, as described previously (23). Trunks 1 and 2 bracket only the LMW isoforms of cyclin E and not the full-length form, while cyclin EL represents the full-length form. Trunk 1 initiates at aa 40 and brackets EL2-6 (Fig. 2E, lane 8), and Trunk 2 initiates at aa 65 and brackets EL5 and EL6 (Fig. 2D, lane 7). We have examined the biochemical activities associated with the protein products of cyclin EL, Trunk 1, and Trunk 2 constructs in several normal and tumor cell lines (Fig. 3).

View larger version (43K): [in this window] [in a new window]   FIG. 3.   The LMW forms of cyclin E are hyperactive. Cyclin EL-FLAG and cyclin E-FLAG constructs Trunk 1 and Trunk 2 were transfected into two mortal cell strains (81N and 76N), two immortalized cell lines (MCF-10A and 76NE6), and two tumor cell lines (MDA-MB-157 and MDA-MB-436), harvested 16 h posttransfection, and subjected toWestern blot analysis with anti-FLAG antibody, histone H1 and GST-Rb kinase analysis, or immune complex formation with CDK2. For Western blot analysis, 50 µg of protein extract from each condition was analyzed with polyclonal antibody to FLAG. For kinase activity and immune complex formation, equal amounts of protein (250 µg) from cell lysates were prepared from each condition and immunoprecipitated with anti-FLAG antibody (polyclonal) coupled to protein A beads. Histone H1 and GST-Rb were used as substrates in the kinase reaction. For each condition, the resulting autoradiogram of the histone H1 and GST-Rb SDS-PAGE are shown. Immune complex formation with CDK2 was assessed by subjecting the anti-FLAG immunoprecipitates to Western blot analysis using a monoclonal antibody to CDK2.

View larger version (49K): [in this window] [in a new window]   FIG. 4.   The LMW forms of cyclin E are biologically active. 81N mortal human mammary epithelial cells were cotransfected with either vector alone backbone (i.e., 3.1), cyclin EL-FLAG, cyclin E-Trunk 1-FLAG, or cyclin E-Trunk 2-FLAG and a GFP-containing vector at a 3:1 ratio. (A) Sixteen hours posttransfection, cells were harvested, GFP-containing cells were sorted, and the sorted and unsorted cells were then subjected to Western blot analysis with the indicated antibodies. (B and C) The GFP-positive cells were then stained with PI and subjected to flow cytometry analysis. Panel B shows the FACS scan raw data of the GFP-sorted cells, and panel C graphically shows the distribution of cells in each phase of the cell cycle. (D) Sorted cells were also plated on microscope slides, allowed to attach for 16 h at 37°C, and stained with FLAG antibody. Following antibody staining, samples were mounted and examined with a Nikon Optiphot microscope equipped with an oil immersion 60×, 1.4 numerical aperture, differential interference contrast objective. Digital images were collected with a SPOT camera (Diagnostics Instruments Inc.) using filters for fluorescein and Texas red, and for each channel, the exposure times were kept constant. Bar, 10 µm.

The LMW forms of cyclin E facilitate S-phase entry. Next, we examined the biological activity of the LMW forms of cyclin E by addressing if their overexpression in normal cells would have a mitogenic effect on the cell cycle and increase progression of cells through S phase. For these studies, we transfected the three cyclin E constructs (i.e., cyclin EL-FLAG, Trunk 1-Flag, and Trunk 2-Flag) into a normal mammary epithelial 81N cell strain. These constructs were cotransfected with a plasmid containing enhanced GFP (EGFP). Sixteen hours following cotransfection, cells were harvested and subjected to FACS analysis to isolate the GFP-positive cells from GFP-negative cells. Following sorting of cells, the EGFP-positive cells were divided into three samples. The first group of EGFP-positive cells was extracted and prepared for Western blot analysis along with the EGFP-negative sorted and unsorted cells (Fig. 4A). The second group of EGFP-positive cells was fixed and prepared for cell cycle analysis (DNA content with PI) using flow cytometry (Fig. 4B and C). The third sample was plated onto coverslips for analysis by anti-Flag immunofluorescence (Fig. 4D).

Hyperactive LMW forms of cyclin E have increased affinity for CDK2. The results from our transfection experiments (Fig. 3) revealed that the tumor-specific LMW forms of cyclin E promote more kinase activity than the full-length protein, explaining their selection in tumor cells. Furthermore, these LMW forms of cyclin E can also induce the expression of more total CDK2 protein in transfected human cells (Fig. 4A). These data raise the question of whether the LMW forms are hyperactive due to their preferentially inducing the expression of more CDK2 or whether these forms are hyperactive regardless of how much CDK2 is present in the cell. To directly address this question quantitatively, we overexpressed full-length cyclin E and its LMW forms and CDK2 in insect cells using a baculovirus expression system (Fig. 5). Insect cells were coinfected with the recombinant baculovirus containing CDK2 and either full-length cyclin E, cyclin E-T1, or cyclin E-T2 cDNAs (Fig. 5). As a control for expression of individual proteins, insect cells were also infected with only one baculovirus, representing each of the aforementioned cDNAs. Sixty hours p.i., cells were harvested and subjected to Western blotting (Fig. 5A) and histone H1 and GST-Rb kinase (Fig. 5B) and immunoprecipitation (Fig. 5C) analyses. Western blot analysis with FLAG or CDK2 antibodies showed that there were similar levels of expression of the three cyclin E forms and of CDK2 in the infected insect cells. Furthermore, whether CDK2 was infected alone or in combination with the three cyclin E forms, equal levels of the protein were expressed in these cells. CDK2 kinase assays using either histone H1 or GST-Rb as substrate revealed that the truncated forms of cyclin E activate a greater amount of kinase activity than the EL form (Fig. 5B) under conditions where equal amounts of CDK2 were immunoprecipitated from each sample (Fig. 5C). In the coinfected insect cells, Trunk 1 and Trunk 2 phosphorylated histone H1 and GST-Rb between two- and sixfold more effectively than the full-length cyclin E. (The bands corresponding to histone H1 and GST-Rb were excised and quantitated by scintillation counting [data not shown].) Interestingly, the fold increase in the hyperactivity of the LMW forms of cyclin E over the full length was very similar in the insect cell baculovirus expression system and the transfection experiments in normal mammary epithelial cells (Fig. 3). Immunoprecipitation analysis revealed that the hyperactivity of the LMW forms of cyclin E was not due to different levels of CDK2 in the cell, since the amount of CDK2 immunoprecipitated was identical in all the coinfected samples (Fig. 5C). What is different among the three forms of cyclin E is their ability to bind to CDK2. The LMW forms of cyclin E bind more effectively to CDK2 than the full-length form, as was apparent in CDK2 immunoprecipitates blotted with FLAG (Fig. 5C) (cyclin E immunoblots showed identical results [data not shown]). These experiments suggest that even though the same amounts of cyclin E and CDK2 are present in the cell and equal amounts of CDK2 are immunoprecipitated from each sample, the LMW forms of cyclin E can bind to CDK2 more effectively than the full-length form, resulting in their hyperactivity. The results from the insect cell baculovirus expression system (Fig. 5) clearly recapitulated the results from our transient-transfection assays (Fig. 3 and 4) and collectively suggest that the LMW forms of cyclin E are indeed hyperactive.

View larger version (37K): [in this window] [in a new window]   FIG. 5.   The LMW forms of cyclin E bind more effectively to CDK2 than does the full-length form of the protein. Cell lysates were prepared from insect cells coinfected with baculovirus containing the indicated cyclin E constructs and CDK2. (A) At 60 h p.i., equal amounts (50 µg) of protein were added to each lane; the gel was then subjected to Western blot analysis with the indicated antibodies. (B) Histone H1 and GST-Rb kinase assays were also performed on the same cell extracts by immunoprecipitating equal amounts of cell lysate with polyclonal antibody to CDK2 coupled to protein A beads, using histone H1 or GST-Rb as substrates. The autoradiograms of the histone H1 and GST-Rb SDS-PAGE are depicted. (C) Immune complex formation with CDK2 was assessed for the same samples by subjecting the anti-CDK2 immunoprecipitates to Western blot analysis using monoclonal antibodies to CDK2 and FLAG.

Identification of an elastase class of enzymes as the protease generating the LMW forms of cyclin E in vitro. Since the LMW isoforms of cyclin E have a significant biochemical and biological function in inducing G₁-to-S-phase progression, we next set out to identify the putative protease responsible for the aberrant cleavage of cyclin E in tumor cells. The results shown in Fig. 2 suggest a novel type of proteolysis cleaving cyclin E into EL5 and EL6. We refer to the cyclin E protease domain stretching from A66 to P71 as AVCADP. This is a short hydrophobic stretch of amino acids followed by aspartate-proline and embedded in a random coil (GOR4-predicted) secondary structure (termed loop 2) bearing a striking similarity to the scissile domains of the serine proteinase inhibitors plasminostrepsin and streptomyces subtilisin inhibitor, making the AVCADP sequence accessible to the protease (3). A second cleavage point nearer to the N terminus in the domain from K32 to K48 (anticipated to generate EL2/3) has the sequence VFLQDP, similar to AVCADP, appears to be in a similar random coil structure (GOR4), and is termed loop 1.

View larger version (52K): [in this window] [in a new window]   FIG. 6.   Elastase digestion of in vitro-translated cyclin E. (A and B) Plasmids containing full-length cyclin EL and its block deletions and point mutations were used to synthesize cyclin E using the TNT reticulocyte lysate system. After digestion with porcine pancreatic elastase, the proteins were separated by SDS-PAGE and analyzed by Western blotting with a FLAG antibody. (C) M16A and EL TNT protein products were incubated in the presence of no elastase, elastase alone, or elastase plus the indicated elastase inhibitors. CMK, methoxysuccinyl-Ala-Ala-Pro-Val-chloromethylketone; the arrows point to the elastase-mediated cleavage sites at EL3 (top arrow) and EL6 (bottom arrow).

Cyclin E processing takes place intracellularly. The data presented in Fig. 6 suggest that the LMW forms of cyclin E are enzymatically produced, very likely by the elastase class of proteases. Since elastase is a secreted enzyme, it may be possible that upon breaking open the cells, the secreted elastase can come into contact with the intracellular cyclin E and induce the cleavage in lysed cells. We have addressed this issue and found by using three different approaches that the fragmentation pattern observed with cyclin E in tumor cells occurs intracellularly prior to homogenization of cells (Fig. 7A to C). Initially (Fig. 7A), we blocked the membrane-bound proteases from contacting cyclin E during the lysis procedures and found that the LMW forms of cyclin E in tumor cells preexist prior to lysis in intact cells. Our usual procedure for preparing tumor and normal cell extracts is to sonicate the cells in PBS containing a protease inhibitor cocktail (Fig. 7). Under these conditions, we detected the full complement of the LMW forms of cyclin E (Fig. 7A, lane 3). Additionally, in order to prevent active extracellular proteases from contacting cyclin E during lysis, sample buffer (containing -mercaptoethanol and 10% SDS) was added directly to cells, immediately boiled for 10 min, and then subjected to SDS-PAGE and Western blot analysis with cyclin E antibody (lane 1). We also prepared cell extracts by sonicating them in only PBS (lane 2). This experiment revealed that the fragmentation pattern was the same regardless of the buffers used. Next, to address directly whether any extracellular elastase activity could potentially cleave cyclin E during the lysis procedure, resulting in the generation of LMW forms, we added a cocktail of elastase inhibitors during lysis (lane 4). The pattern of cyclin E LMW forms remained unchanged by the elastase inhibitors (compare lanes 1 and 4). These results indicate that the LMW forms of cyclin E in tumor cells are likely to preexist in the cells prior to extraction.

View larger version (50K): [in this window] [in a new window]   FIG. 7.   Cyclin E proteolysis occurs in the cell, not during protein extraction. (A) Cells from the human breast tumor cell line MDA-MB-157 were harvested, and the cell pellets were suspended in either SDS-PAGE sample buffer and boiled immediately (lane 1) or PBS (lane 2), protease inhibitor cocktail (25 µg of leupeptin/ml, 25 µg of aprotinin/ml, 10 µg of pepstatin/ml, 1 mM benzamidine, 10 µg of soybean trypsin inhibitor/ml, 0.5 mM phenylmethylsulfonyl fluoride, 50 mM NaF, 0.5 mM vanadate) (lane 3), or protease inhibitor cocktail plus 25 µM concentrations of the elastase inhibitors elastatinal and MPCMK (lane 4). Following resuspension, cells in lanes 2 to 4 were sonicated in a circulating ice water bath, centrifuged for 45 min at 100,000 × g, and subjected to Western blot analysis with cyclin E antibody. (B) Cyclin E-FLAG recombinant protein (4 µl of TNT reticulocyte lysate reaction mixture) was added either to tumor cell extracts following sonication (lanes 2) or directly to cells prior to sonication (lanes 3). Following sonication in the indicated buffers, the cyclin E-FLAG-tumor cell extract mixture was subjected to Western blot analysis with an anti-FLAG antibody. Lane 1 of each buffer set depicts the addition of reticulocyte lysate reaction mixture with empty vector to the tumor cell extract. The "TNT alone" lane is cyclin E-FLAG TNT reticulocyte lysate reaction mixture without mixing of the tumor cell extracts. The arrow points to the cleaved cyclin E. (C) Tumor cells were immune-depleted of cyclin E by passing tumor cell extracts on a cyclin E affinity column for three rounds. The immune-depleted cyclin E tumor cell extracts were then mixed (1:1 ratio) with normal cell extracts, incubated for the indicated times at 37°C, and subjected to Western blot analysis with anti-cyclin E antibody (lanes 1 to 5). Lanes 6 and 7 depict cyclin E levels in tumor cell extracts prior to (lane 6) and following (lane 7) immune depletion. N, normal cells; T, tumor cells; ID, immune-depleted cells; S, starting material. (D) MDA-MB-157 cells were incubated with the indicated concentrations of LLnL for 36 h and subjected (50 µg/lane) to Western blot analysis with antibodies to cyclin E and p21. The bracket in the p21 panel indicates the high-molecular-weight laddering of p21, which is diagnostic of polyubiquitination.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

In this report we describe a novel proteolysis of cyclin E, occurring only in tumor cell lines and tissues, which generates the LMW forms of cyclin E in tumor cells. A total of six cyclin E proteins were detected in tumor cells ranging in molecular mass from 50 to 33 kDa. We have termed the 50-kDa cyclin E as EL1 and the five cyclin E protein products from 45 to 33 kDa as the LMW isoforms cyclin EL2 to EL6. All of the detectable LMW forms of cyclin E found in tumor cells can be explained as the result of three mechanisms at work. First, alternate translation at M46, but not at M16, accounts for the band we call EL4 (Fig. 2D). M16 alternate translation does not occur in cells, even though it can readily occur in reticulocyte lysate (Fig. 2D and 6A and B). Second, proteolysis of cyclin E occurs around the carboxyl side of A69, generating bands at EL5 and -6 (Fig. 2B and C). These bands are a doublet resulting from the third process, a posttranslation modification such as phosphorylation or deacetylation. Finally, bands EL2 and -3 arise from a similar proteolysis and posttranslational modification as EL5 and -6 at aa 40 to 45. We have also identified the full-length cyclin E as the EL form (Fig. 1 and 2D) described previously as the 15-aa alternatively spliced elongated form of cyclin E (42).

The proteolytic processing of cyclin E is unique to tumor cell lines and tissues. This tumor-specific pattern is not the result of overexpression or constitutive expression since normal cells transfected with cyclin E under a strong constitutive promoter do not further process cyclin E into its LMW forms (23; Fig. 3). The proteolysis we have described also differs from the proteasome-dependent proteolysis of cyclin E described by two other laboratories (11, 62). In these studies, the ubiquitination of cyclin E was dependent upon a specific phosphorylation at threonine 380. Although the destruction of cyclin E may be mediated through the ubiquitination-proteasome pathway in tumor cells, the generation of the LMW forms observed under steady state conditions in tumor cells and tissues is not likely to be generated by the proteasome pathway. Additionally, the LMW forms of cyclin E are present constitutively in the tumor cells and are not subject to cell cycle regulation (29). Furthermore, the LMW forms of cyclin E do not seem to represent the intermediate proteolytic products of degradative machinery (Fig. 7D). We conclude that these LMW forms of cyclin E are more likely due to the action of a different protease.

The way cyclin E is processed into LMW isoforms is also different than the proteolysis of other cyclins in the cell. For example, cyclin D1 is degraded by a calpain protease (10). Cyclin A is cleaved in vitro by a p27-dependent protease which removes the destruction box and allows evasion from proteasomal degradation (4). The proteolysis we describe here for cyclin E at A69 to D70 does not appear to occur to other cyclins. This site is homologous to elastase protease substrate sequences containing essential small hydrophobic residues (3). When A69 is replaced with an aspartyl residue, the proteolysis is nearly completely blocked (Fig. 2 and 6). A D70P substitution completely blocks degradation, presumably due to the unique imino-peptide bond afforded by the proline residue within the scissile bond. The effect of proline at position 70 is not surprising since this is commonly used in the design of peptide inhibitors of proteases (46). The involvement of an aspartyl residue in the cleavage site raises the possibility that a caspase protease might be involved, yet the typical DEVD motif is not present. In addition, the cyclin E protease appears to cleave on the N-terminal side of D70, contrary to a caspase cleavage on the C-terminal side of aspartyl residues. For example, a caspase activity which cleaves cyclin A during apoptosis in Xenopus embryos cleaves downstream of a critical aspartyl residue (54). In the case of cyclin E, the aspartyl residue is not essential since the D70A mutation does not block proteolysis of cyclin E and in fact may encourage it (Fig. 2).

The sequence analysis of the specific regions in cyclin E cleaved to produce EL5/6 (loop 2, A66 to P71; the AVCADP motif) and EL2/3 (loop 1, K32 to K48; the VFLQDP motif) suggested that the protease cleaving these regions is a serine protease of the elastase class (3). These regions within cyclin E form a hairpin loop type of structure making them accessible to the protease. We showed that in vitro-translated wild-type and mutant cyclin E proteins all digest in vitro with a porcine pancreatic elastase, generating a Western blot pattern similar to the in vivo pattern (compare Fig. 2 and 6). The striking similarity of in vivo and in vitro cleavage is supporting evidence that the elastase class of enzymes may actively cleave cyclin E in the cell. There are two possible ways to interpret how cyclin E is cleaved into its LMW forms by the elastase class of enzymes in tumor cells but not normal cells: (i) there is more elastase activity in tumor cells than in normal cells; or (ii) normal cells express high levels of elafin, an inhibitor of elastase. There is precedent for both these possibilities. Several studies have shown that the elastases are overexpressed in human cancer (20, 63, 64). Additionally, recent studies have reported that elafin, a potent elastase inhibitor for human leukocyte elastase and porcine pancreatic elastase (60), is differentially expressed in normal versus tumor-derived mammary epithelial cells, with very high levels in normal cells and undetectable levels in breast cancer cells (65). Collectively, these studies suggest that normal cells may be protected from the proteolytic effects of elastase by overexpressing elafin, while in tumor cells elastase expression may promote increased cyclin E activation and subsequent accelerated S-phase entry.

The specific cyclin E proteolysis at the AVCADP and VFLQDP motifs may separate a substrate-targeting domain (28) from the active cyclin E-CDK2 complex or affect the binding of an inhibitor protein. In addition, since the efficiency of proteasome proteolysis does not leave partial peptides, we presume the proteolysis of cyclin E generating bands EL2 through EL6 occurs prior to the reported proteasome degradation (11, 61). Hence, although the proteasome pathway is involved in degrading cyclin E, the proteolysis of cyclin E described here is a newly discovered level of cyclin regulation distinct from the ubiquitin-proteasome pathway. The elastase proteolysis of cyclin E may have a specialized role in the regulation of cyclin E-CDK2 activity, substrate selection, or subcellular localization rather than a role regulating the timely destruction of cyclin E.

The LMW forms of cyclin E seem to have a regulatory role distinct from that of the full-length protein. Our results clearly show that the LMW forms are biochemically hyperactive, as they can phosphorylate substrates such as histone H1 and GST-Rb much more efficiently than the full-length form. We used two different systems to examine the biochemical activity of the LMW forms of cyclin E compared to the full-length form. The two systems used, namely the transient-transfection assay in normal human mammary epithelial cells and the insect cell baculovirus expression system, both revealed that the LMW forms of cyclin E are two- to sixfold more active than the full-length form. Furthermore, the insect cell system, which is more quantitative, revealed that the hyperactivity of the LMW forms of cyclin E is in part due to an increased affinity of these forms for CDK2. Cyclin kinase inhibitors may also play a role in the cyclin E LMW form hyperactivity (data not shown). This high level of kinase activity associated with the LMW forms of cyclin E may be a significant contributor to the tumor phenotype directly through the action of the overexpressed kinase. The effect of the LMW forms (i.e., T1 and T2 transient expression) on normal cell growth also reflects this stimulated CDK2 kinase activity. When normal cells, which are devoid of any LMW forms, are forced to express the LMW forms of cyclin E, the cells progress through S phase very effectively. This fits the pattern seen in tumor cells, which already overexpress the LMW forms of cyclin E. Apparently, the tumor cells have adapted to this state since they are less responsive to the T1 and T2 transient expression (Fig. 3 and data not shown). Lastly, we showed that the LMW forms of cyclin E appear to retain the ability to localize to the nucleus even with a significant loss of the N terminus. Obviously, the nuclear targeting or retention signal is not found in the N terminus of cyclin E (19). The conclusion from these studies is that the LMW forms of cyclin E are not only functional but also hyperactive compared to the full-length form, providing the tumor cells with an added growth advantage. The LMW forms of cyclin E can phosphorylate substrates more effectively than normal cells, which only express the full-length form, and as a result tumor cells can progress through G₁ and into S phase, bypassing the restriction point. The oncogenic potential of these LMW forms in vivo is currently under study and should elucidate how extensive a role these LMW forms have in the transformation phenotype.

Lastly, the LMW forms of cyclin E could provide a novel target for rational drug design for the treatment of metastatic breast cancer without harming normal proliferating cells in the body. Our studies have implicated a protease that is induced during metastatic progression (20, 63, 64) in the proteolytic regulation of cell cycle progression. By identifying the specific protease (i.e., of the elastase class) which cleaves cyclin E into its LMW forms that is found exclusively in tumor cells and tissues, we may be able to design cyclin E-specific protease inhibitors. These inhibitors would then help control the progression through the cell cycle in invasive cells, thus limiting the ability of these cells to populate distant metastatic sites.