Cell Cycle-dependent Usage of Transcriptional Start Sites

Cyclin B1 mRNA is expressed temporally throughout the cell cycle with peak expression in G2and M phase. Both transcriptional and posttranscriptional controls are important for this cell cycle-dependent regulation of cyclin B1 mRNA. In this study, we observed that cyclin B1 has two major transcripts: (a) a constitutively expressed transcript, and (b) a cell cycle-regulated transcript expressed predominantly during G2-M phase. These different transcripts are due to alternative start sites. The constitutively expressed transcript starts 65 bases upstream from the cell cycle-regulated message. Changes in mRNA stability did not appear to control the expression of the cell cycle-specific transcript, but we were able to identify a 24-base pair region of the cyclin B1 promoter spanning the start site of the cell cycle-regulated transcript that was critical for its cell cycle-regulated promoter activity. This suggests that transcriptional regulation is responsible for controlling the presence of each message. The 24-base pair sequence required for cell cycle regulation was notable for containing the nucleotides GGCT repeated three times. The possibility that these two transcripts might be physiologically distinct was raised when the cell cycle-specific transcript was found to be translated more efficientlyin vitro than the constitutively expressed transcript. These results characterize a novel mechanism for the regulation of cyclin B1 throughout the cell cycle that is dependent upon the use of different transcriptional start sites.

The cell cycle-dependent expression of cyclin B1, the regulatory subunit of mitosis promoting factor (MPF), is critical for the proper timing of a cell's entry into mitosis. The cyclin B1 protein accumulates during interphase and peaks at the G 2 -Mphase transition. This pattern of expression is seen not only at the protein level, but also at the mRNA level in somatic cells (1)(2)(3). The levels of cyclin B1 mRNA are regulated at both the level of transcription and the level of message stability. Previously reported data from our laboratory have shown that cyclin B1 mRNA is highly stable in G 2 and mitosis but is markedly less stable in G 1 (4). Although these variations in message stability contribute to the cell cycle regulation of the cyclin B1 gene, transcriptional regulation also influences the expression of cyclin B1 mRNA.
Several studies have defined the transcriptional start sites for cyclin B1 (5)(6)(7). However, the locations of these start sites differ from each other. Using primer extension and RNase protection assays, Piaggio et al. (6) identified a start site that is located 63 bases upstream of the start site identified by Cogswell et al. (7) using RNase protection assays. In addition, Pines and Hunter (5) defined the cyclin B1 transcriptional start site at an intermediate position.
In this report, we describe the experiments that we performed to reconcile these seemingly disparate results. Studies of the human cyclin B1 promoter show that transcriptional variation also plays a role in the regulation of the cyclin B1 gene (6 -8). The upstream region of the human cyclin B1 gene has been isolated and shown to have cell cycle-regulated promoter activity. Piaggio et al. (6) demonstrated that the activity of the cyclin B1 promoter was minimal in quiescent cells and gradually increased after serum stimulation. Our laboratory has shown that the cyclin B1 promoter has cell cycle regulation; its activity was low in G 1 and maximal in G 2 and mitosis (8). Deletions from the 5Ј end of the cyclin B1 promoter allowed us to identify a region of 90 bp 1 that retained cell cycle-regulated promoter activity. Cogswell et al. (7) have found that the activity of the cyclin B1 promoter was greater in G 2 -M phase than in S phase using in vitro transcription assays and transfected cells. They suggested that an E-box element within the cyclin B1 promoter contributed to its cell cycle regulation.
In their experiments, upstream stimulatory factor (USF) was found to bind to the E-box in a cell cycle-dependent manner. The binding of upstream stimulatory factor to the E-box was shown to be increased in cells blocked in G 2 -M phase compared with cells blocked in S phase or cycling cells. This suggested that upstream stimulatory factor may play a role in the transcriptional regulation of cyclin B1 during the G 2 -M-phase transition. In a more recent report, Katula et al. (9) found that cell cycle-regulated activity could be conferred by a smaller fragment excluding the E-box. The promoter activity of this fragment was dependent upon the presence of two CCAAT box elements (9).
Farina et al. (10) also proposed a role for the E-box in cyclin B1 expression during the G 0 -G 1 -phase transition. They found that the Max protein binds to the E-box consensus element in quiescent cells, suggesting that the E-box has an inhibitory role during G 0 . Because the 90-bp region of the cyclin B1 promoter that we have identified that confers cell cycle regulation does not include the E-box element, it appears to be unnecessary for cell cycle control. These contrasting results imply that the E-box is not the only element that plays a role in the regulation of cyclin B1 expression throughout the cell cycle. It is most likely that multiple elements are important for the cell cycledependent regulation of cyclin B1 mRNA expression.
In this report, we describe the identification of two cyclin B1 * This work was supported by United States Public Health Service Grant GM 47439 from the National Institutes of Health. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
¶ To whom correspondence should be addressed: transcripts: (a) a constitutive transcript that appears in several phases of the cell cycle, and (b) a cell cycle-regulated transcript that appears only during G 2 -M phase. In addition, these two cyclin B1 transcripts may also be regulated at the translational level.

EXPERIMENTAL PROCEDURES
Cell Culture, Cell Synchronization, and RNA Extraction-HeLa cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum and 1% penicillin and streptomycin at 37°C in a humidified atmosphere containing 5% CO 2 . Cells were blocked at different phases of the cell cycle by treatment with one of the following: (a) aphidicolin (1 g/ml) for 16 h, (b) mimosine (400 M) for 24 h, or (c) nocodazole (0.04 g/ml) for 16 h. Total RNA was extracted from HeLa cells using TRIzol reagent (Life Technologies, Inc.), which is an adaptation of the guanidine isothiocyanate method described by Chomczynski and Sacchi (11).
Synthetic Oligonucleotides-Oligonucleotides were synthesized by the University of Pennsylvania Cancer Center (Philadelphia, PA). The sequences of these oligonucleotides are shown in Table I.
Primer Extension Analysis-An oligonucleotide consisting of 20 bp of the 5Ј sequence from the cDNA of the human cyclin B1 gene (from positions 11 to 30) was labeled at the 5Ј end with [␥-32 P]ATP and T4 polynucleotide kinase. 5 ϫ 10 4 cpm of the labeled primer were hybridized with 50 g of RNA in 1ϫ aqueous hybridization buffer (1 M NaCl, 167 mM HEPES, pH 7.5, and 333 M EDTA, pH 8.0) overnight at room temperature. The primer extension reaction was carried out as described previously (12). The extended products were then separated and analyzed by electrophoresis on a 6% denaturing polyacrylamide gel.
Reverse Transcription-PCR-An oligonucleotide corresponding to the human cyclin B1 cDNA sequence from positions 250 to 270 (referred to as cycB-250 RT) was used as the primer for cDNA synthesis in the reverse transcription reaction. 5 pmol of primer were annealed to 4 g of RNA. Reverse transcription was carried out at 42°C for 90 min. The cDNA synthesized from each reverse transcription reaction was used for two separate PCR amplifications. The cycB-250 RT oligonucleotide was used for the downstream primer in both PCR amplifications. For one set of reactions, the oligonucleotide ccreg.start.887 was used for the upstream primer. For the other set of reactions, the oligonucleotide con.start.848 was used for the upstream primer. The sequences of both of these oligonucleotides lie within the sequence of the human cyclin B1 promoter at positions 887-913 and 848 -873, respectively. The products of each PCR amplification were separated by electrophoresis on a 1.5% agarose gel stained with ethidium bromide.
In Vitro Transcription and Translation-The DNA templates used for in vitro transcription reactions were PCR-amplified from a plasmid containing the human cyclin B1 promoter upstream of a luciferase reporter gene (pGL2-upcB (upstream region of cyclin B1)) using the following oligonucleotides for primers: (a) T7.ccreg.start for the upstream primer for the cell cycle-regulated template, (b) T7.const.start for the upstream primer for the constitutive template, and (c) lucif.down.stop for the downstream primer for both templates. The T7.ccreg.start primer contains the T7 promoter for RNA synthesis and spans the region from positions 885 to 907, which includes the cell cycle-regulated start site. The T7.const.start primer contains the T7 promoter and spans the region from positions 823 to 845, which includes the constitutive start site. The lucif.down.stop primer is located within the luciferase coding region (positions 802-826) and includes a stop codon. Capped RNA transcripts were synthesized from these PCR-amplified templates using the RiboMAX System (Promega). 200 ng of each RNA transcript (along with 100 ng of a luciferase control transcript) were used in an in vitro translation reaction containing rabbit reticulocyte lysate (Promega), [ 35 S]methionine, and an amino acid mixture. The labeled proteins were then denatured and separated by 10% SDS-polyacrylamide gel electrophoresis. The levels of labeled proteins were quantitated using optical densitometry.
Plasmids and Transient Transfection Assays-The pGL2-upcB plasmid contains the full-length promoter of human cyclin B1 upstream of a luciferase reporter gene (8). Plasmids containing cyclin B1 promoter sequences downstream of either the constitutive start site or the cell cycle-regulated start site regulated by the SV40 promoter were constructed by using PCR. The appropriate fragments were amplified and subcloned into pGL2-PROMOTER (Promega), which contains the SV40 promoter upstream of the luciferase reporter gene. The plasmid containing the SV40 promoter-constitutive start site sequences is referred to as SV.const.start. The plasmid containing the SV40 promoter-cell cycle-regulated start site sequences is referred to as SV.ccreg.start.
A luciferase plasmid containing the cyclin B1 promoter up to position 842, which includes the constitutive start site, was generated by PCR amplification of the appropriate fragment followed by subcloning into pGL2-BASIC (Promega). This plasmid is referred to as upcB.const.842. A plasmid containing the cyclin B1 promoter up to position 896, which includes the cell cycle-regulated start site, was generated by digesting a plasmid containing the cyclin B1 promoter that contains a BglII linker from positions 897 to 908 with BglII to excise the sequences downstream of the cell cycle-regulated start site. This remaining plasmid was re-ligated and is referred to as upcB.ccreg.896.
The pGL2-upcB⌬873 plasmid contains the full-length human cyclin B1 promoter with a 76-bp deletion at its 3Ј end from positions 873 to 949. This was accomplished by removing the cyclin B1 promoter from pGL2-upcB by digestion with KpnI and XhoI. The cyclin B1 promoter was then digested with DraI to remove the 76-bp region from the 3Ј end before re-ligating it back into the pGL2-BASIC plasmid.
HeLa cells were transiently transfected with various plasmids using the calcium-phosphate method as described previously (8,13). The pSV␤-gal plasmid was co-transfected as an internal control for transfection efficiency. The transfected cells were then treated with aphidicolin to block them in G 1 -S phase or with nocodazole to block them in G 2 -M phase. Cells were harvested using reporter lysis buffer (Promega) and assayed for luciferase and ␤-galactosidase activity as described previously (8).

Variation of the 5Ј Ends of the Cyclin B1 mRNA with the Cell
Cycle-To determine whether the transcriptional start sites of the cyclin B1 gene vary throughout the cell cycle, we performed primer extension analysis on total RNA extracted from HeLa cells blocked in different phases of the cell cycle. HeLa cells were treated with one of the following drugs before RNA extraction: (a) aphidicolin to block cells at G 1 -S phase, (b) mimosine to block cells in late G 1 , or (c) nocodazole to block cells in mitosis. Using flow cytometry, we have previously found that 70 -75% of a population of cells were arrested at the G 1 -S-phase boundary by aphidicolin, 65-75% of cells were arrested in G 1 by mimosine, and 85-90% of cells were blocked in G 2 -M phase by FIG. 1. Primer extension analysis reveals two transcriptional start sites for human cyclin B1. Primer extension was carried out using total RNA extracted from HeLa cells that had been arrested in late G 1 by mimosine, arrested in G 1 -S phase by aphidicolin, and arrested in G 2 -M phase by nocodazole. The radiolabeled primer consisted of sequences from the 5Ј end of human cyclin B1 cDNA. Arrows indicate the presence of two transcripts: (a) a 117-bp extended product seen in all phases, and (b) a 52-bp product seen only in G 2 -M phase. The sequencing ladder on the left was generated with CircumVent Thermal Cycle Sequencing (New England BioLabs) using the same radiolabeled primer that was used in the primer extension reactions and a plasmid containing the cyclin B1 promoter. nocodazole (8). An oligonucleotide corresponding to sequences from the 5Ј end of the human cyclin B1 cDNA was used as the radiolabeled primer. As shown in Fig. 1, two major transcripts were found: (a) a cell cycle-regulated transcript, and (b) a constitutively expressed transcript. A longer transcript that mapped to 117 bases from the 3Ј end of the primer was observed in equivalent amounts in HeLa cells blocked in late G 1 , G 1 -S phase, and G 2 -M phase. A shorter transcript that mapped to 52 bases from the primer was observed only in HeLa cells blocked in mitosis. As expected, we observed a greater amount of total cyclin B1 mRNA in G 2 -M phase than in late G 1 or G 1 -S phase. The presence of the longer transcript was confirmed by reverse transcription-PCR. An oligonucleotide composed of a sequence from within the coding sequence of human cyclin B1 (cycB-250 RT) was used as the downstream primer. An oligonucleotide (con.start.848) that can detect only the longer transcript was used as the upstream primer. The predicted PCR product of 351 bp was detected in cells blocked in all phases of the cell cycle, as shown in Fig. 2. An upstream primer (ccreg-.start.887) that can detect both transcripts was also used as a positive control. The resulting PCR product amplified from these reactions is 312 bp. A larger amount of this PCR product is observed in cells blocked at G 2 -M phase because the upstream primer used in these reactions detects both cyclin B1 transcripts that are present at this time during the cell cycle. No products were seen when reverse transcriptase was omitted from the reactions.
These experiments indicate that there are two different cyclin B1 transcripts. One transcript is constitutively expressed and is transcribed from a start site that has been mapped by Piaggio et al. (6). The other, shorter transcript varies markedly throughout the cell cycle. The start site for this transcript corresponds to the start site that was mapped by Cogswell et al. (7). These experiments reconcile the reports of two different transcriptional start sites (shown in Fig. 3). One start site appears to be differentially regulated and cell cycle dependent, whereas the other is constitutively active throughout the cell cycle.
Cyclin B1 Upstream Sequences Do Not Alter Message Stability-We asked whether the longer cyclin B1 transcript containing an additional 65 bases had a different stability than the shorter transcript. To examine the effects of these sequences, we placed either the constitutive start site or the cell cycleregulated start site under the control of the SV40 promoter in a luciferase reporter plasmid. The SV40 promoter has fairly uniform levels of activity throughout the cell cycle (8). Thus, equal amounts of mRNA would be transcribed from this promoter. Any cell cycle variation observed in luciferase activity would be the result of changes in mRNA stability. The activities of these plasmids were tested by transient transfection into HeLa cells that were subsequently treated with various cell cycle inhibitors (Fig. 4). The plasmid containing the SV40 promoter upstream of the cell cycle-regulated start site, SV.ccreg-.start, had almost equal levels of luciferase activity in both

FIG. 2. Confirmation of the presence of the constitutively expressed cyclin B1 transcript by reverse transcription-PCR.
Reverse transcription-PCR analysis was performed on RNA extracted from HeLa cells arrested in late G 1 , G 1 -S phase, and G 2 -M phase. An oligonucleotide consisting of sequences within the coding sequence of human cyclin B1 (cycB-250 RT) was used as the primer for cDNA synthesis and the downstream primer. In one set of reactions, an oligonucleotide referred to as con.start.848 was used as the upstream primer. This primer is located 16 bases downstream of the constitutive start site and detects only the longer, constitutive transcript. In another set of reactions, an oligonucleotide referred to as ccreg.start.887 was used as the upstream primer. This primer spans the cell cycle-regulated start site and detects both the constitutive transcript and the cell cycle-regulated transcript. As a negative control, one set of reactions did not include reverse transcriptase. Markers are as follows: M1 represents a 1-kb ladder (517, 396, 344, 298, and 220 bp), and M2 represents ⌽X174 HaeIII (603, 310, 281, 234, and 194 bp). G 1 -S phase and G 2 -M phase. The plasmid containing the SV40 promoter upstream of the constitutive start site, SV.const-.start, also had approximately the same amounts of activity in both G 1 -S phase and G 2 -M phase. Both of these plasmids behave similarly to pGL2-CONTROL, a plasmid containing the SV40 promoter upstream of a luciferase reporter gene. The pGL2-upcB plasmid, which contains the full-length cyclin B1 promoter, was included in this experiment as a control for cell cycle regulation. The cyclin B1 promoter was found to be approximately 4-fold more active in G 2 -M phase than in G 1 -S phase. These results would suggest that the sequences within the 5Ј untranslated regions of the cyclin B1 transcripts do not confer changes in mRNA stability that could regulate the expression of the two different cyclin B1 transcripts.
Cell Cycle Regulation by the Cyclin B1 Promoter-To identify a region of the human cyclin B1 promoter that is critical for its cell cycle-regulated promoter activity, we constructed a series of plasmids containing various deletions that span the cell cycle-regulated transcriptional start site, as diagrammed in Fig. 5A. The cell cycle-regulated activity of these plasmids was tested by transfecting them into HeLa cells, followed by treatment with cell cycle inhibitors to block the transfected cells at particular points in the cell cycle. The results of these experi-ments are shown in Fig. 5B. The full-length cyclin B1 promoter, upcB, had approximately 4-fold greater activity in G 2 -M phase compared with that in G 1 -S phase. The cyclin B1 promoter with the sequences downstream of the cell cycle-regulated start site

FIG. 4. Comparison of the stabilities of cyclin B1 transcripts.
To determine whether the longer cyclin B1 transcript containing an additional 65 bases has a different stability than the shorter transcript, we placed the constitutive start site or the cell cycle-regulated start site under the control of the SV40 promoter in a luciferase reporter plasmid. The plasmid containing the sequences for the shorter transcript is referred to as SV.ccreg.start. The plasmid containing the sequences for the longer transcript is referred to as SV.const.start. The pGL2-CON-TROL plasmid contains the SV40 promoter upstream of the luciferase reporter gene. The pGL2-upcB plasmid contains the full-length cyclin B1 promoter upstream of the luciferase reporter gene. The activities of these plasmids were tested by transient transfection into HeLa cells that were subsequently arrested in G 1 -S phase or G 2 -M phase.
FIG. 5. Identification of sequences within the human cyclin B1 promoter that are critical for cell cycle regulation. A, deletions were made from the 3Ј end of the cyclin B1 promoter. B, the activities of the 3Ј cyclin B1 promoter deletions were tested by transfecting these plasmids into HeLa cells, followed by treatment with cell cycle inhibitors. Ⅺ, G 1 -S phase; f, G 2 -M phase. deleted, upcB.ccreg.896, also had approximately 4-fold greater activity in G 2 -M phase compared with that in G 1 -S phase. The cyclin B1 promoter constructs containing deletions that span the cell cycle-regulated start site, upcB⌬873 and upcB-.const.842, had approximately equivalent levels of activity in both G 1 -S phase and G 2 -M phase; i.e. cell cycle regulation was not observed. The sequences that were deleted in both of these plasmids contain the cell cycle-regulated transcriptional start site. Thus, the presence of this transcriptional start site is critical for the increased level of promoter activity during G 2 -M phase. These results also identify a 24-bp sequence, which is shown in Fig. 5A, that is necessary for cell cycle-regulated promoter activity.
Translational Efficiency of Each Cyclin B1 Transcript in Vitro-The presence of two cyclin B1 transcripts raises the possibility of differences in translational efficiency between the constitutively expressed transcript and the cell cycle-regulated transcript that could result in different levels of protein synthesis throughout the cell cycle. To determine if this might be possible, we compared the in vitro translational efficiency of each transcript. Template DNA for in vitro transcription was PCR-amplified from a cyclin B1 promoter-luciferase plasmid using a primer within the luciferase coding region and either a primer that spans the constitutive start site or a primer that spans the cell cycle-regulated start site. Both of these primers contain promoter sequences for T7 polymerase. After in vitro transcription, equal amounts of each mRNA transcript were translated in vitro in a rabbit reticulocyte lysate system. As shown in Fig. 6, the level of protein synthesized from the cell cycle-regulated transcript was approximately 3-fold greater than the level of protein synthesized from the constitutive transcript, as determined by optical densitometry. Thus, the increased translational efficiency of the shorter, cell cycle-regulated transcript may contribute to the higher level of cyclin B1 protein during G 2 -M phase.

DISCUSSION
The expression of cyclin B1 mRNA is regulated at both the transcriptional and posttranscriptional levels (4,5). Several different studies have described the structure of the cyclin B1 promoter along with its cell cycle regulation (6 -8). These studies have also reported different transcriptional start sites for the cyclin B1 gene. Our experiments have clarified these discrepancies as well as identified a novel aspect of the regulation of cyclin B1. We have identified two cyclin B1 transcripts that differ by 65 bases in length. The longer transcript is constitutively expressed and appears throughout the cell cycle. The shorter transcript is cell cycle regulated and appears only at G 2 -M phase. The start sites for these transcripts map to the start sites reported by Piaggio et al. (6) and Cogswell et al. (7), respectively. The experiments described in these reports used total RNA from unsynchronized cells. This would explain why the shorter transcript was not identified as being cell cycle regulated and the longer transcript was not identified as being cell cycle independent.
Cell cycle-dependent usage of transcriptional start sites has not been described previously. The presence of an additional cyclin B1 transcript at the G 2 -M-phase boundary may contribute significantly to the increased level of cyclin B1 expression at this point in the cell cycle. This cell cycle-regulated transcript was also found to be translated more efficiently than the constitutively expressed transcript. Approximately three times more protein was synthesized in vitro from the M-phase-specific transcript than from the constitutive transcript. Thus, translational control may be an important factor in the regulation of cyclin B1 expression.
Several different mechanisms could be responsible for the presence of the additional, shorter M-phase-specific transcript. It could be regulated transcriptionally and be specifically transcribed only at G 2 and M phase. It could also be regulated posttranscriptionally; i.e. it could be more stable in G 2 and mitosis than it is in G 1 and S phase. Changes in message stability have already been shown to be responsible in part for the differences in cyclin B1 mRNA throughout the cell cycle (4). However, these differences in message stability were determined after examination of the total cyclin B1 mRNA and not that of a specific transcript. When the sequences for both cyclin B1 transcripts were placed under the regulation of the SV40 promoter, which is expressed uniformly throughout the cell cycle, we did not observe differences in mRNA stability using a transient transfection assay. Thus, changes in message stability most likely do not regulate the differential expression of the M-phasespecific transcript compared with that of the longer transcript.
To identify a region of the cyclin B1 promoter that is critical for cell cycle regulation, we made a series of deletions from its 3Ј end and tested them in HeLa cells in a transient transfection assay. These deletions span the cell cycle-regulated transcriptional start site. From these experiments, we were able to identify a 24-bp sequence that is necessary for cell cycle-regulated activity of the cyclin B1 promoter. This region includes the cell cycle-regulated start site and the sequence GGCT repeated three times. Although this motif or region has not been previously shown to be important for cell cycle-regulated genes that are active at G 2 -M phase, it may be important for cyclin B1 regulation. The CDE element has been described as an important DNA sequence for the regulation of the cdc25C, cyclin A, and cdc2 genes (14,15). These genes are active in late S phase and G 2 and are not expressed in G 0 . The CDE element was shown to be necessary for the repression of these genes in G 0 . The cyclin B1 promoter contains a CDE element, but it does not appear to be critical for its regulation in G 2 -M phase, because the deletion of this region does not alter the cell cycle-regulated promoter activity. Katula et al. (9) have recently identified two CCAAT elements that appear to be important for cyclin B1 regulation. Both of these CCAAT elements were shown to be necessary for the induction of the cyclin B1 gene during S phase. The NF-Y transcription factor was shown to bind to these CCAAT elements. Both of the CCAAT elements and the GGCT repeats may contribute to the regulation of transcription of cyclin B1 during G 2 -M phase. The GGCT repeats are located in the same region as the cell cycle-regulated transcriptional start site that we have shown to be important for the regulation of the cyclin B1 gene.