Topoisomerase IIα Binding Domains of Adenomatous Polyposis Coli Influence Cell Cycle Progression and Aneuploidy

Background Truncating mutations in the tumor suppressor gene APC (Adenomatous Polyposis Coli) are thought to initiate the majority of colorectal cancers. The 15- and 20-amino acid repeat regions of APC bind β-catenin and have been widely studied for their role in the negative regulation of canonical Wnt signaling. However, functions of APC in other important cellular processes, such as cell cycle control or aneuploidy, are only beginning to be studied. Our previous investigation implicated the 15-amino acid repeat region of APC (M2-APC) in the regulation of the G2/M cell cycle transition through interaction with topoisomerase IIα (topo IIα). Methodology/Principal Findings We now demonstrate that the 20-amino acid repeat region of APC (M3-APC) also interacts with topo IIα in colonic epithelial cells. Expression of M3-APC in cells with full-length endogenous APC causes cell accumulation in G2. However, cells with a mutated topo IIα isoform and lacking topo IIβ did not arrest, suggesting that the cellular consequence of M2- or M3-APC expression depends on functional topoisomerase II. Both purified recombinant M2- and M3-APC significantly enhanced the activity of topo IIα. Of note, although M3-APC can bind β-catenin, the G2 arrest did not correlate with β-catenin expression or activity, similar to what was seen with M2-APC. More importantly, expression of either M2- or M3-APC also led to increased aneuploidy in cells with full-length endogenous APC but not in cells with truncated endogenous APC that includes the M2-APC region. Conclusions/Significance Together, our data establish that the 20-amino acid repeat region of APC interacts with topo IIα to enhance its activity in vitro, and leads to G2 cell cycle accumulation and aneuploidy when expressed in cells containing full-length APC. These findings provide an additional explanation for the aneuploidy associated with many colon cancers that possess truncated APC.


Introduction
Mutation of the tumor suppressor Adenomatous Polyposis coli (APC) gene is considered an initiating event in over 80% of all colorectal cancers [1]. Mutations in APC have also been associated with chromosomal instability and aneuploidy in early polyps from FAP (Familial Adenomatous Polyposis) patients [2,3,4]. Although the ability of APC to suppress canonical Wnt signaling by targeting bcatenin is critical for APC to suppress tumorigenesis [5,6,7,8,9], accumulating evidence suggests that APC likely suppresses tumor development through pathways besides inhibiting canonical Wnt signaling [see review [10]].
Among the multiple functions of APC identified is cell cycle regulation [11,12,13,14,15,16,17]. APC involvement in G1/S is attributed to its recognized role in canonical Wnt signaling. APC participation in the G2/M transition involves interaction with topoisomerase IIa (topo IIa) [18]. However, the mechanism by which APC regulates the G2-M cell cycle transition is poorly understood.
Besides being an enzyme that catalyzes DNA topology changes [19,20,21,22], topo IIa is also a critical regulator of one G2/M checkpoint during cell division, the decatenation checkpoint [23]. Inhibition of topo IIa activity leads to initiation of the G2 decatenation checkpoint, resulting in G2 cell cycle arrest [23]. Topo IIb, a closely related protein, has a similar amino acid sequence and activity as topo IIa, but is dispensable for cell cycle control [24,25,26].
Previously, we found full-length endogenous APC interacts with endogenous topo IIa but not with topo IIb [18]. Expression of a central fragment of APC that binds topo IIa led to cell cycle accumulation in G2, independent of b-catenin [18]. Thus, we concluded that nuclear APC interacts with topo IIa and thus, might be involved in the regulation of cell cycle progression. In the current study, we identify a second domain in the central portion of APC that specifically binds to topo IIa but not topo IIb. Both APC central domains dramatically impact the activity of topo IIa in vitro. Cell lines with full-length endogenous APC that express either of the APC domains capable of modifying topo IIa activity accumulate in G2 and display increased aneuploidy. Using a cell line lacking topo IIb and with a mutant isoform of topo IIa, we demonstrate that cell cycle arrest triggered by expression of middle APC fragments requires normal expression of endogenous topo IIa. Our data implies that full-length APC can participate in the topo IIa-mediated regulation of the G2-M transition.

Two central regions of APC bind topo IIa
Previously, we identified an interaction between endogenous APC and topo IIa [18]. Exogenous expression of the 15-amino acid repeat region of APC (M2-APC) that interacts with topo IIa led to G2 cell cycle arrest. In the present study, we investigated whether the 20amino acid repeat region of APC (M3-APC) also interacts with topo IIa. Topo IIa specifically co-precipitates with both EGFP-fused M2and M3-APC ( Figure 1B, top blot). However, topo IIb does not coprecipitate with either APC fragment under the same experimental conditions ( Figure 1B, second blot). EGFP-M2-APC encompasses APC amino acids 959-1338 while EGFP-M3-APC contains amino acids 1211-2075 ( Figure 1A). The M3 region contains two monopartite nuclear localization signals [27]. Together, these two regions contain all 15-and 20-amino acid repeats of APC that mediate bcatenin binding and downregulation. Therefore, b-catenin coprecipitation with both M2-and M3-APC, served as a positive control ( Figure 1B, third blot).
A functional interaction between purified M2-or M3-APC and topo IIa Given that exogenous M2-and M3-APC each interact with endogenous topo IIa, we tested whether both APC fragments would also influence two different reactions catalyzed by topo IIa. Purified non-overlapping recombinant M2-and M3-APC fragments each enhanced the ability of purified topo IIa to resolve highly catenated kinetoplast DNA into decatenated mini DNA circles in vitro (Figure 2A and 2B). Neither purified M2-nor M3-APC showed decatenation activity in the absence of topo IIa (Figure 2A and 2B). Topo IIa enzyme can also convert supercoiled DNA into relaxed circular DNA and this relaxation activity was enhanced by addition of purified M2-or M3-APC ( Figure 2C). Neither purified M2-nor M3-APC relaxed the supercoiled DNA in the absence of topo IIa ( Figure 2C). While it was clear that purified M2-or M3-APC enhanced both decatenation and relaxation activities of purified topo IIa, we wanted to eliminate and M3-APC fused to the C-terminus of EGFP and used in all cell studies. (B) A GFP antibody was used to immunoprecipitate (IP) EGFP from HCT116bw cells expressing either EGFP-M2-APC, EGFP-M3-APC or EGFP. Immunoblots (IB) were probed using antibodies indicated to the left of the gel. Ten percent input equals 10 mg total protein. Topo IIa co-precipitates with both M2-APC and M3-APC, but not with EGFP. Topo IIb does not coprecipitate with M2-or M3-APC, but b-catenin does co-precipitate with both. Blots probed for a-tubulin served as a loading control for the input samples and a negative control for the immunoprecipitations. ( * ) marks migration of antibody heavy chain. Representative blots from three independent experiments are shown. doi:10.1371/journal.pone.0009994.g001 the possibility that these effects were due to increased total protein concentration in the reactions and were instead specific properties of M2 and M3-APC. Purified BSA did not enhance topo IIa activities ( Figures 2B and 2D). In contrast, when reactions performed such that the purified topo IIa alone displayed moderate activity, the addition of purified BSA protein slightly inhibited both decatenation and relaxation activities of topo IIa ( Figure 2B and 2D). These in vitro assays provide additional support for a functional interaction between APC and topo IIa. Furthermore, purified M2-and M3-APC had no area of overlap, and yet each interacted with purified topo IIa. We conclude that although the EGFP-M3-APC used for cell studies overlaps slightly with the EGFP-M2-APC, this area of overlap is not solely responsible for the topo IIa interaction and M2-and M3-APC can each interact with and affect topo IIa independently.
Expression of M2-or M3-APC results in G2 cell cycle arrest M2-and M3-APC each bind endogenous topo IIa in cells ( Figure 1B) and purified topo IIa in vitro ( Figure 2). Thus, we expressed M3-APC in HCT116bw cells (HCT116 cells that contain only a wild-type allele of b-catenin) to determine whether this expression altered cell cycle progression. Cell cycle distribution was determined by FACS analysis of living cells labeled with propidium iodide. Similar to what was previously seen using M2-APC [18], cells expressing M3-APC progressively accumulated in the G2/M phases of the cell cycle, while cells expressing EGFP did not ( Figure 3A and B, Table S1). When compared to cells expressing EGFP alone, cells expressing M3-APC for 72 hours showed a 2-fold increase in G2/M distribution and a 31% decrease in S phase distribution; cells expressing M2-APC showed a 2.4-fold increase in G2/M distribution and a 77% decrease in S phase distribution. We conclude that expression of M2-or M3-APC leads to cell cycle accumulation in G2/M. Of note, the reduction in S phase cells seen upon expression of M2-or M3-APC, suggested a second cell cycle delay prior to S phase, likely in G1. This apparent delay in G1 is consistent with a previous observation that APC regulates the G1-S transition [13].
M2-APC expression elicits cell accumulation in the G2 phase rather than in mitosis [18]. FACS analysis does not distinguish between the G2 and M cell cycle populations. Thus, to determine whether expression of M3-APC also resulted in G2 accumulation, we determined the mitotic indices in living or fixed M3-APCexpressing cells at 24 and 48 hours post-transfection. We found no significant expansion of the mitotic population in M3-APC expressing cells by either counting phospho-histone H3 positive cells (data not shown) or by estimating the percentage of living cells displaying mitotic figures as visualized with Hoechst stain ( Figure 3C and Table S2). Therefore, as previously reported for M2-APC, the expanded G2/M population of M3-APC-expressing cells determined by FACS analysis represents an accumulation in G2, not in M. Together, these results demonstrate that M3-APC and M2-APC each induce G2 cell cycle arrest when expressed in a colon epithelial cell line with full-length APC. G2 arrest triggered by M2-and M3-APC is not dependent on p53 Tumor suppressor p53 participates in various pathways that regulate the G2/M transition (reviewed in [28]). To determine if p53 is required for the M2-and M3-APC-mediated G2 arrest, we determined the consequences of M2-or M3-APC expression in the promyelocytic leukemia cell line, HL-60. HL-60 cells express full-length APC protein (data not shown) but are null for p53 [29]. When either M2-or M3-APC was expressed in HL-60 cells, the G2/M population increased significantly ( Figure 4A and Table  S3). M2-APC expression resulted in a near doubling of the G2/M population, while M3-APC expression resulted in a tripling of the G2/M population. The G2 arrest in HL-60 cells indicates that the impact of M2-and M3-APC on the cell cycle is not restricted to colon epithelial cell lines. Moreover, the G2 cell cycle arrest triggered by expression of either M2-or M3-APC is not dependent on p53.
Cells deficient in topo II do not arrest in G2 following expression of either M2-or M3-APC Based on our data, we hypothesize that exogenous M2-and M3-APC each interact with endogenous topo IIa, resulting in a p53-independent cell cycle arrest in G2. Ideally, to demonstrate that this APC-mediated cell cycle regulation is dependent upon topo IIa, we would eliminate topo IIa from the analyzed cells. Unfortunately, topo IIa is an essential protein and its perturbation typically results in cell cycle delay followed by cell death [30]. Although no cultured mammalian cell line completely lacks topo IIa, there are cell line variants such as HL-60/MX2 cells with compromised topo IIa activity. HL-60/MX2 cells were originally generated by selecting for resistance to the topo II inhibitor mitoxantrone [31]. Thus, compared to parental HL-60 cells, they are 195-fold less sensitive to drugs that target topo II [32,33,34]. This resistance has been attributed to the observation that HL-60/ MX2 cells express no topo IIb, only a low level of topo IIa, and a truncated topo IIa with reduced activity and aberrant subcellular localization [32,33,34]. In contrast to the parental HL-60 cells, we observed that HL-60/MX2 cells expressing M2-or M3-APC showed no increase in the G2/M population ( Figure 4 and Table  S3). Rather, the slight decrease in G2/M and increase in the S population of M2-and M3-APC-expressing cells were not significantly different from the EGFP-expressing HL-60/MX2 cells. Taken together with our observation that M2-and M3-APC interact preferentially with topo IIa rather than topo IIb ( Figure 1B), we conclude that topo IIa is required for M2-or M3-APC-triggered cell cycle arrest in G2.
Cells with truncated APC do not arrest in G2 following expression of M2-or M3-APC The majority of somatic APC mutations in colon cancers result in overexpression of a truncated protein that includes all of the M2-APC region and part of M3-APC (Table 1). We examined the cell cycle profile of SW480 cells which express an endogenous APC protein truncated at amino acid 1368. In SW480 cells, expression of either M2-or M3-APC did not lead to G2 arrest, or any other alterations in the cell cycle phases ( Figure 5 and Table  S4).

G2 accumulation in M2-and M3-APC-expressing cells does not depend on b-catenin regulation
A major function of the central region of tumor suppressor APC is to target, b-catenin for proteasome-mediated destruction. We have previously reported that expression of M2-APC does not alter b-catenin localization, level or activity and concluded that the G2 arrest triggered by M2-APC is not likely mediated by b-catenin [18]. Although both M2 and M3-APC bind b-catenin, only the M3-APC region is capable of targeting b-catenin for cytoplasmic destruction. To test if the G2 arrest triggered by M3-APC involves b-catenin regulation, we expressed M3-APC in HCT116bm cells which produce only stabilized b-catenin that can not be targeted for degradation [35]. HCT116bm cells expressing either M2 or M3-APC displayed a near doubling of the G2/M population with an accompanying decrease in the G0/G1 population when compared to HCT116bm cells expressing EGFP alone ( Figure 6A and B and Table S5). Moreover, using a reporter construct to determine b-catenin activity as a transcription coactivator of LEF-1, it was demonstrated that expression of M3-APC led to increased b-catenin activity in HCT116bm cells, but slightly decreased b-catenin activity in HCT116bw cells ( Figure 6C). This opposing effect of M3-APC on b-catenin activities in the two cell lines supports our conclusion that a decrease in b-catenin activity is not required for M3-APCtriggered G2/M cell cycle delay. Furthermore, expression of APC amino acids 1379-2080 in SW480 resulted in decreased bcatenin activity [36]. Thus, our finding that SW480 cells expressing M3-APC do not arrest in G2 ( Figure 5) indicates that decreasing b-catenin activity is not sufficient to trigger a G2/M cell cycle delay.

G2 arrest triggered by M2-and M3-APC is accompanied by increased aneuploidy
Although derived from human colon cancer tissue, the HCT116 cell line retains a stable diploid karyotype. Not only did HCT116 cells expressing either M2-or M3-APC show progressive G2 arrest, but they displayed a significant increase in aneuploidy (Figure 7 and Table S1). In contrast, SW480 cells are originally aneuploid [ Table 1]. Expression of either M2-or M3-APC in SW480 cells did not lead to a significant increase in aneuploidy (Figure 7 and Table S4).

Discussion
In this study, we identified a novel topo IIa binding domain (M3) in the central region of APC that enhances both decatenation and relaxation activities of purified topo IIa. Cells expressing M2-or M3-APC accumulated in the G2 phase of the cell cycle and showed increased aneuploidy; however, this result was not observed in cells with endogenous truncated APC missing part of the M3 domain. The G2 arrest was also independent of p53 but was dependent on topo IIa. Our data indicate the central region of APC interacts with topo IIa and thereby regulates G2-M cell cycle progression.

M2 and M3-APC each trigger G2 cell cycle arrest independent of topo IIb
Topo IIa and topo IIb are 75% identical in protein sequence and share some functional similarity in promoting DNA topology changes. However, successful generation of cell lines and mouse models completely lacking topo IIb [24,25,26] indicates that topo IIb is dispensable for cell cycle progression. In contrast, topo IIa is essential for cell cycle control. Our finding that M2-and M3-APC interact with endogenous topo IIa but not topo IIb (Fig. 1B) is consistent with our previous observation that full-length endogenous APC interacts with endogenous topo IIa but not topo IIb [18]. Based on this information, we conclude that APC-mediated G2 cell cycle arrest is not dependent on topo IIb.

Abnormal nuclear morphology is not associated with G2 arrest
We previously reported that expression of M2-APC in HCT116bw cells leads to abnormal nuclear morphology and G2 cell cycle arrest [18]. In contrast, although expression of M3-APC in HCT116bw cells also resulted in G2 arrest, the abnormal nuclear morphology was not observed until 72 hours posttransfection (data not shown). Since both M2 and M3-APC interact with and affect topo IIa, it seems most likely that the altered nuclear morphology observed shortly after M2-APC expression is topo IIa-independent. Furthermore, the abnormal nuclear morphology was also seen in SW480 cells [18] which do not undergo a G2 arrest following expression of either M2-or M3-APC ( Figure 5). Thus, the abnormal nuclear morphology is not associated with G2 arrest.

M2-and M3-APC are not identical in their interaction with topo IIa
Although both M2-and M3-APC bind topo IIa and trigger G2 cell cycle arrest when expressed in HCT116bw, HCT116bm, or HL60 cells, the cellular response to the two fragments is not identical. Altered nuclear morphology was observed 24 hours after expression of M2-APC [18]. In contrast, nuclear morphological alterations were not observed until 72 hours after expression of M3-APC (data not shown). In HCT116bw cells, expression of M2-APC for 72 hours resulted in a more robust G2 cell cycle accumulation than did expression of M3-APC ( Figure 3B). In HL60 cells, expression of M3-APC resulted in a more robust G2 cell cycle accumulation than did expression of M2-APC (Figure 4). In transfected HCT116bw cells, M2-APC protein levels were approximately 3-fold higher than that of full-length endogenous APC and 2.7-fold higher than M3-APC levels (data not shown). However, M2-and M3-APC appeared able to co-precipitate equivalent amounts of endogenous topo IIa ( Figure 1B). In general, it appeared that transfected HCT116bw cells expressed more M2-APC than M3-APC, but M3-APC was better able to bind endogenous topo IIa than M2-APC. Experiments using purified non-overlapping recombinant M2-and M3-APC fragments revealed that M3-APC enhanced the decatenation activity of purified topo IIa at a lower molar concentration than M2-APC ( Fig. 2A). The opposite was observed in in vitro relaxation assays, where M2-APC appeared to enhance the ability of topo IIa to relax supercoiled DNA more effectively than M3-APC (Fig. 2C). Together, these observations are consistent with M2-and M3-APC binding to different regions of topo IIa and thus modifying topo IIa activity by slightly different mechanisms. Further investigation is required to identify the specific topo IIa binding sites and delineate the underlying mechanisms.

Potential molecular mechanism and physiological relevance
Over 60% of FAP polyps display aneuploidy [2,3,4]. It has been proposed that APC mutations contribute to chromosomal instability (CIN) through loss of spindle-kinetochore attachment or misregulation of the cytoskeleton [for review, see [37]]. More recently, an association of truncated APC fragments with mitotic checkpoint protein Mad2 was reported to inactivate the mitotic checkpoint, thus providing another potential mechanism for CIN [38]. We propose a novel mechanism that might contribute to increased aneuploidy following mutation of APC. Our model is based on evidence that middle fragments of APC bind to topo IIa, affect topo IIa activity, and result in G2 cell cycle accumulation and increased aneuploidy when expressed exogenously.
We propose that expression of M2-or M3-APC causes altered topo IIa activity, thus activating the G2 decatenation checkpoint, which leads to G2 arrest. Aneuploid cells would result from altered topo IIa activity in the small percentage of mitotic cells that escape the G2 decatenation checkpoint. The G2 decatenation checkpoint is vital for cell cycle control and genomic integrity. Cells lacking the G2 decatenation checkpoint become aneuploid [39,40]. A variety of topo II inhibitors have been shown to arrest cells in G2 by activating the G2 decatenation checkpoint [23]. It is possible that only cells with an intact G2 decatenation checkpoint can be arrested in G2 by expression of middle regions of APC. Consistent with this hypothesis, all cell lines we observed to undergo G2 arrest upon M2 or M3-APC expression possessed an intact decatenation checkpoint [41,42]. HL60-MX2 cells have no decatenation checkpoint [32,33,34] and did not arrest in G2 (Fig. 4). In our study, the display of aneuploidy in cells expressing M2-or M3-APC increased steadily over time (Figure 7). We suggest that the basis for this aneuploid accumulation is compromised topo IIa activity in cells that escape the G2 decatenation checkpoint. We further predict that in FAP patients, truncated APC fragments which contain the M2 region would similarly interact with topo IIa and this might result in aneuploidy. In support of this prediction, a literature review of colorectal cancer cell lines reveals a general trend that cells with full-length APC are diploid with stable karyotypes (Table 1). In contrast, cell lines that express a truncated APC that includes M2-APC are mostly aneuploid, with very few exceptions. Our results provide a potential explanation for the presence of aneuploidy in early FAP adenomas. Further experiments comparing topo IIa activity in colon cancer cell lines harboring various truncating APC mutations are needed to establish a direct link between topo IIa activity and aneuploidy. The current study expands the repertoire of molecular factors implicated in the pathogenesis of colorectal cancer, illuminating new areas for future development of treatment strategies.

Materials and Methods
Cell culture and DNA constructs HCT116bw (containing one wild-type allele of b-catenin) and HCT116bm (containing one mutant allele of b-catenin) cells (a gift from Dr. Bert Vogelstein) and SW480 (ATCC) were grown in McCoy's 5A medium (Gibco) supplemented with 10% FBS (Hyclone). HL60 cells (ATCC) were grown in Iscove's Modified Dulbecco's Medium (ATCC) supplemented with 20% FBS (Hyclone). HL60/MX2 (ATCC) were grown in RPMI 1640 medium (Cellgro) supplemented with 10% FBS. Expression constructs for APC fragments fused to EGFP were kindly provided by Dr. Naoki Watanabe and have been described previously [43]. His and S dual-tag fused M2-APC was made as described [18]. To generate recombinant N-terminal His and S dual-tag fused APC fragment M3, the corresponding cDNA for APC (amino acid 1330-2058) was amplified using PCR and subcloned into a pET-30a(+) vector.

Electroporation and FACS analysis
Cells grown on plastic were treated with trypsin to obtain a single cell suspension. A total of 2 mg of EGFP, EGFP fused M2-, or M3-APC expression plasmid were electroporated using Nucleofector I (Amaxa) according to the manufacturer's protocol. Electroporation programs used were: HCT116bw (program D-32), HCT116bm (program D-32), HL60 (program T-19), and HL60/MX2 (program X-03). SW480 cells were transfected with Metafectine (Scientifix, Australia). Forty-eight hours post-transfection, single cells in suspension were stained with 0.5 mg/ml Hoechst 33342 (Invitrogen) for 30 minutes at 37uC. FACS analysis was performed using both UV and 488 nm lasers on a 5-laser BD LSRII flow cytometry (BD Bioscience). Ten thousand EGFPpositive cells were collected for each sample. Data were analyzed using BD FACSDiva Software (BD Bioscience) and plotted using WinMDI 2.9.

Antibodies and immunofluorescence
Cells transfected with EGFP or EGFP-fused M3-APC were fixed with 4% paraformaldehyde, and immunostaining was performed using anti-phospho-histone H3 (1:500, Upstate) as described [46]. One hundred EGFP-positive cells were randomly chosen, and only cells also positive for phospho-histone-H3 were counted. As a second method to determine mitotic indices, living cells were stained with 0.5 mg/ml Hoechst 33342 at 24 and 48 hours post-transfection and mitotic figures were counted for 100 cells in each category. The mitotic indices are presented as an average 6 s.d. of three independent experiments.

Reporter gene assay
HCT116bw and HCT116bm cells grown in 24-well plates were co-transfected using Metafectine reagent (Scientifix, Australia) with 2 mg of the EGFP-M2-APC, EGFP-M3-APC or EGFP expression construct, 100 ng of the TCF-reporter construct SuperTOP-flash or FOPflash (Upstate Biotechnology, Lake Placid, NY), and 50 ng of the pRL-TK Renilla luciferase construct (Promega, WI) as a control to normalize for transfection efficiency. After 48 hours, cells were harvested and luciferase activities were determined using the Dual-LuciferaseH assay system (Promega) and a Turner Designs TD-20/20 luminometer. SuperTOP-flash and FOPflash luciferase activities were first normalized by pRL-TK Renilla luciferase, and then the normalized SuperTOP-flash luciferase activity was divided by normalized FOPflash luciferase activity to calculate relative b-catenin activity.

Supporting Information
Table S1 Cell cycle distribution in HCT116 bw cells expressing GFP, M2-APC, or M3-APC. Transfected cells were stained with Hoechst blue, and the cell cycle distribution G0/G1 (2N), S (between 2N and 4N), and G2/M (4N) was determined by FACS at three time points post-transfection. For each transfection, 10,000 GFP-positive cells were analyzed.