The ARF tumor suppressor prevents chromosomal instability and ensures mitotic checkpoint fidelity through regulation of Aurora B.

The ARF tumor suppressor is part of the CDKN2A locus and is mutated or undetectable in numerous cancers. The best-characterized role for ARF is in stabilizing p53 in response to cellular stress. However, ARF has tumor suppressive functions outside this pathway that have not been fully defined. Primary mouse embryonic fibroblasts (MEFs) lacking the ARF tumor suppressor contain abnormal numbers of chromosomes. However, no role for ARF in cell division has previously been proposed. Here we demonstrate a novel, p53-independent role for ARF in the mitotic checkpoint. Consistent with this, loss of ARF results in aneuploidy in vitro and in vivo. ARF(-/-) MEFs exhibit mitotic defects including misaligned and lagging chromosomes, multipolar spindles, and increased tetraploidy. ARF(-/-) cells exhibit overexpression of Mad2, BubR1, and Aurora B, but only overexpression of Aurora B phenocopies mitotic defects observed in ARF(-/-) MEFs. Restoring Aurora B to near-normal levels rescues mitotic phenotypes in cells lacking ARF. Our results define an unexpected role for ARF in chromosome segregation and mitotic checkpoint function. They further establish maintenance of chromosomal stability as one of the additional tumor-suppressive functions of ARF and offer a molecular explanation for the common up-regulation of Aurora B in human cancers.

INTRODUCTION

The ARF tumor suppressor is encoded by an alternate reading frame of the INK4a tumor suppressor in the CDKN2a (also known as the INK4a/ARF) locus. ARF is best known for its role in the p53 pathway. In response to cellular stress, ARF is released from the nucleolus and enters the cytoplasm, where it binds MDM2 (Pomerantz ; Zhang ), an E3 ubiquitin ligase that inhibits the transcriptional activity of p53 (Momand ) and targets it for degradation (Honda ). Binding of ARF inhibits the ability of MDM2 to ubiquitinate p53 and also promotes MDM2 degradation, thus stabilizing p53 and allowing for cell cycle arrest (Zhang ; Honda and Yasuda, 1999). However, it has been shown that ARF has tumor-suppressive roles outside of this pathway (Korgaonkar ). Mice lacking ARF and p53 develop a wider tumor spectrum and are more likely to have multiple tumors per animal than mice lacking p53 alone (Weber ). In addition, high levels of ARF expression impede entry into and progress through S phase in cells lacking p53 and MDM2 (Weber ; Yarbrough ), suggesting that ARF interacts with distinct partners to mediate cell cycle delay. ARF can also bind and inhibit numerous transcription factors independently of p53, including FoxM1B, E2F, and Myc (Kalinichenko ; Ozenne ). p53-independent functions of ARF have also been identified in ribosome biogenesis and sumoylation (Ozenne ). Thus, ARF has multiple functions outside of the p53 pathway, some of which contribute to its tumor-suppressive activity.

Aneuploidy, an abnormal chromosome complement that deviates from a multiple of the haploid, occurs in ∼85% of human tumors (Weaver and Cleveland, 2006; Zasadil ). Aneuploidy is often accompanied by chromosomal instability (CIN), the recurrent gain or loss of chromosomes during multiple divisions. Aneuploidy and CIN serve as markers of poor prognosis in multiple cancer types, including lymphomas, soft tissue sarcomas, and cancers of the lung, breast, ovaries, and colon. Several mouse models that develop aneuploidy and CIN have an elevated rate of spontaneous and/or carcinogen-induced tumorigenesis (Ricke ; Holland and Cleveland, 2009; Schvartzman ; Zasadil ). Together, the data suggest that aneuploidy and a low rate of CIN can be tumor promoting.

Chromosome missegregation during mitosis is a common cause of aneuploidy and CIN in tumor cells (Cimini ; Ganem ; Silkworth ; Ryan ). The mitotic checkpoint (also termed the spindle assembly checkpoint) is the major cell cycle regulator acting during mitosis to prevent chromosome missegregation. To produce genetically identical progeny, replicated sister chromatids are sorted and segregated on a bipolar spindle composed of microtubules. Sister chromatids attach to spindle microtubules through their kinetochores—protein structures that assemble at centromeric DNA. The mitotic checkpoint delays the separation of sister chromatids, which occurs at the transition from metaphase to anaphase, until all kinetochores have formed stable attachments to spindle microtubules (Rieder ; 1995). Sister separation occurring before formation of stable attachments to opposite poles results in random segregation of chromosomes, increasing the likelihood of producing aneuploid progeny. To prevent this, unattached kinetochores recruit mitotic checkpoint components, including Bub1, BubR1, Mad1, Mad2, and CENP-E, to catalytically generate a diffusible inhibitor of the anaphase-promoting complex/cyclosome (APC/C; Nilsson ; Kulukian ; Han ). Once all kinetochores have stable microtubule attachments, the checkpoint is satisfied and APC/C becomes active, leading to sister chromatid separation and the generation of euploid progeny (Vleugel ).

The Aurora B kinase is a member of the chromosomal passenger complex (CPC), which localizes to inner centromeres in early mitosis and relocates to the spindle midzone after anaphase onset (Cooke ). During prometaphase, Aurora B functions to release improper kinetochore–microtubule interactions that would impair accurate chromosome segregation (Lan ; Cimini ). Aurora B overexpression occurs in a variety of human cancers (Chieffi ; Vischioni ; Chen ; Lin ), and inhibitors of Aurora B are in clinical trials (Komlodi-Pasztor ). In this study we identify novel, p53-independent, tumor-suppressive functions of ARF in chromosome segregation and mitotic checkpoint signaling via regulation of Aurora B protein levels.

RESULTS

Loss of ARF causes aneuploidy in vitro

We and others reported that ARF loss results in abnormal chromosome numbers in MEFs (di Tommaso ; Silk ). However, the mechanism remained unclear. To determine how loss of ARF resulted in chromosome missegregation, MEFs were prepared from E14.5 ARF−/− embryos and wild-type littermates. Loss of ARF caused a significant increase in aneuploidy, as assessed by chromosome spreads (also known as metaphase spreads; Figure 1, A–C). Although most aneuploid ARF−/− cells had a near-diploid chromosome number, ARF−/− MEFs also exhibited a 2.4-fold increase in near-tetraploidy (Figure 1, D and E), suggestive of cytokinesis failure. Consistent with this, MEFs lacking ARF also had a 1.9-fold increase in the percentage of binucleate cells relative to MEFs with a wild-type complement of ARF (Figure 1, F and G).

FIGURE 1:

ARF loss causes aneuploidy in vitro. (A) Diploid MEF chromosome spread with 40 chromosomes. Scale bar, 10 μm. (B) Average aneuploidy in MEFs. n = 3 experiments of 50 spreads each. (C) Histogram of chromosome spreads from B. Inset, enlarged histogram showing the percentages of near-tetraploid MEFs. (D) Tetraploid MEF chromosome spread with 80 chromosomes. Scale bar, 10 μm. (E) Average tetraploidy in MEFs. n = 3 experiments of 100 spreads each. (F) Binucleate MEF. Red, F-actin stained with phalloidin. Blue, DNA. Scale bar, 20 μm. (G) Average percentage of binucleation in MEFs of the indicated genotypes. n = 250 cells from each of three independent experiments. **p < 0.001.

ARF−/− animals develop aneuploidy in vivo

To test whether ARF loss results in chromosome missegregation in an intact organism, we collected splenocytes from 5-mo-old mice for analysis of aneuploidy using chromosome spreads (Figure 2A). Indeed, ARF−/− splenocytes showed a 3.7-fold increased level of aneuploidy relative to splenocytes from wild-type mice (Figure 2B), demonstrating that loss of ARF is sufficient to induce aneuploidy in vivo. All aneuploid splenocytes had near-diploid numbers of chromosomes (Figure 2C). To further examine this, we analyzed tissue from the small intestine by fluorescent in situ hybridization (FISH) using a probe for chromosome 11 (Figure 2D). Loss of ARF resulted in a significant increase in abnormal numbers of chromosome 11 in intestinal tissue (Figure 2, E and F). Overall, these results indicate that ARF is required to maintain chromosomal stability in vitro and in vivo.

FIGURE 2:

ARF is required to maintain chromosomal stability in vivo. (A) Chromosome spread from a mouse splenocyte with 40 chromosomes. Scale bar, 10 μm. (B) Average aneuploidy in 5-mo-old mouse splenocytes. n = 3 independent experiments of 50 spreads each. (C) Histogram of chromosome spreads from B. No tetraploid spreads were observed in splenocytes. Mitotic indices of splenocytes were comparable in wild-type (0.52 ± 0.18%) and ARF-null (0.62 ± 0.03%) animals. n ≥ 930 cells from three animals. (D) Murine intestine labeled with FISH probe to chromosome 11. Top right is labeled to indicate nuclear boundaries and number of chromosome 11 signals in each cell. (E) Percentage of cells of the indicated genotypes with greater or less than two copies of chromosome 11. (F) Percentage of cells with the indicated number of copies of chromosome 11. *p < 0.05; **p < 0.001.

Loss of ARF causes mitotic defects

The increased rate of aneuploidy in ARF−/− cells suggested that passage through mitosis in the absence of ARF would result in mitotic defects. Indeed, examination of asynchronously cycling MEFs in mitosis revealed that loss of ARF resulted in a 5.2-fold increased frequency of misaligned chromosomes when the majority of chromosomes were at the metaphase plate (Figure 3, A and B), suggesting a deficit in chromosome congression. To test this directly, we treated cells with the proteasome inhibitor MG132 to prevent anaphase onset and permit additional time for chromosome alignment to occur. ARF−/− cells showed a substantial decrease in the percentage of cells that successfully aligned their chromosomes as compared with wild type. This was true whether MG132 was added to asynchronous cells or to those that had formerly established monopolar spindles due to treatment with the Eg5/KSP inhibitor monastrol (Figure 3, C and D). ARF−/− MEFs also showed an increase in chromosomes that lag behind the main bodies of segregating DNA during anaphase or telophase (lagging chromosomes; Figure 3, E and F), which are commonly used as a marker of chromosome missegregation (Thompson and Compton, 2011).

FIGURE 3:

ARF−/− cells exhibit mitotic defects. (A, B) ARF−/− MEFs have increased levels of misaligned chromosomes. (A) Left, normal metaphase. Right, cell with metaphase plate and misaligned chromosome (arrow). (B) Percentage of MEFs of the indicated genotypes containing visible metaphase plates and misaligned chromosomes. n > 50 metaphases from three independent experiments. (C, D) ARF loss causes congression defects. (C) Left, image of MG132-treated cell with fully congressed chromosomes. Right, image of MG132-treated cell with incompletely aligned chromosomes. (D) Quantitation of the percentage of wild-type and ARF−/− cells that successfully aligned their chromosomes after 3 h treatment with 10 μM MG132 (left) or after 18 h of 100 μM monastrol to induce monopolar spindles, followed by washout into 3 h of MG132 (mon→MG; right). (E, F) MEFs lacking ARF have increased levels of lagging chromosomes in anaphase and telophase. (E) Left, normal anaphase. Right, abnormal anaphase cell with lagging chromosome indicated by arrow. (F) Percentage of cells of the indicated genotype with lagging chromosomes. n >100 anaphases and telophases from thee independent experiments. (G, H) ARF−/− MEFs have an elevated frequency of supernumery centrosomes. (G) Interphase cells with two (left) or four centrosomes (right), denoted by arrows. (H) Percentage of interphase cells with abnormal numbers of centrosomes. n > 250 cells from each of three independent experiments. (I–K) ARF loss results in multipolar spindles. (I) Example of a multipolar spindle in an ARF−/− MEF. (J) Percentage of MEFs with abnormal spindles. (K) Histogram showing number of poles per spindle. n > 100 mitotic cells from each of three independent experiments. Scale bars, 5 μm. *p < 0.05.

Cytokinesis failure produces tetraploid G1 cells with two centrosomes. After centriole replication in S phase, tetraploid cells can develop supernumery centrosomes. Consistent with the 2.4-fold increase in tetraploidy in cells lacking ARF expression (Figure 1E), ARF loss in MEFs resulted in a twofold increase in cells with abnormal numbers of centrosomes in interphase (Figure 3, G and H). Because centrosomes are the primary nucleating sites of microtubules, these extra centrosomes can result in formation of multipolar spindles in the subsequent mitosis. Further analysis of mitotic cells showed that MEFs lacking ARF have a threefold increased incidence of multipolar spindles (Figure 3, I– K). Together, these mitotic defects would be expected to result in increased rates of chromosome missegregation and aneuploidy in ARF−/− cells.

Loss of ARF results in a weakened mitotic checkpoint

To determine whether mitotic checkpoint function is compromised in ARF−/− cells, we challenged MEFs with the microtubule poison colcemid. Colcemid treatment results in loss of spindle microtubules, producing unattached kinetochores and an active mitotic checkpoint. Whereas wild-type MEFs accumulated in mitosis in response to colcemid, ARF−/− cells had a significantly reduced mitotic index (Figure 4A). Consistent with this, time-lapse analysis revealed that the duration of mitosis was significantly shorter in cells lacking ARF (Figure 4B). To confirm that this was an ARF-specific effect, we tested whether mitotic arrest in response to colcemid could be rescued by transfection with a yellow fluorescent protein (YFP)–tagged version of the ARF protein. Like endogenous ARF, ARF-YFP localizes to nucleoli during interphase (Weber ; Supplemental Figure S1, A and B). High levels of ARF overexpression were reported to cause an interphase arrest because they result in an increase in cells with a 2n, G0/G1 DNA content on flow cytometry profiles. In certain cell types, a substantial increase (20–30%) in the percentage of cells with a 2n content of DNA is observed (Quelle ; Weber ; Yarbrough ). However, in other cases, the increase in the G0/G1 population is ≤10% (Quelle ; Yarbrough ). In most cases, the percentage of ARF-overexpressing cells in G2/M is similar, or somewhat increased, compared to the percentage in control cells (Quelle ; Weber ; Yarbrough ). Indeed, when treated with the microtubule poison nocodazole, cells overexpressing ARF retain their ability to enter mitosis (Zhang ). Consistent with these data, the mitotic index in cells expressing ARF-YFP was indistinguishable from that in cells transfected with empty vector in the absence of microtubule poisons (Figure 4C, left). However, expression of ARF-YFP was sufficient to restore mitotic checkpoint activity in ARF−/− cells. The mitotic index of ARF−/- cells expressing ARF-YFP was indistinguishable from the mitotic index of wild-type cells transfected with empty vector after 16 h of colcemid treatment (Figure 4C, right). This rescue experiment confirms that the mitotic checkpoint defect in ARF−/− cells is due to an acute requirement for ARF in mitotic checkpoint signaling as opposed to one or more compensatory changes that occurred during development in response to loss of ARF.

FIGURE 4:

ARF−/− cells have a weakened mitotic checkpoint. (A) Mitotic index of MEFs treated with colcemid for the specified number of hours. n = 250 cells at each time point from each of three independent experiments. (B) Duration of mitosis in MEFs treated with colcemid, as assessed by time-lapse microscopy. n = 100 cells from three independent experiments. (C) ARF transfection rescues mitotic checkpoint activity in ARF−/− MEFs. After 32 h of transfection with ARF-YFP or empty vector, MEFs were treated with colcemid or vehicle for 16 h before analysis of mitotic index. n = 250 cells from each of three independent experiments. (D) Mitotic index of MEFs infected with retroviruses expressing empty vector or wild-type or V24E p14ARF. Green fluorescent protein (GFP) was also expressed from an internal ribosomal entry site. After 32 h of infection, cells were treated with colcemid for 16 h before analysis of mitotic index. n > 250 cells from each of three independent experiments. (E) Percentage of metaphase cells with misaligned chromosomes after 72 h of infection with retroviruses expressing empty vector or wild-type or V24E p14ARF. Both wild-type p14ARF and the MDM2-binding-domain mutant V24E rescue the occurrence of misaligned chromosomes in ARF−/− MEFs. n > 50 metaphases from three independent experiments. (F) Percentage of anaphase or telophase MEFs with lagging chromosomes after infection with empty vector or wild-type or V24E p14ARF. Both wild-type and V24E p14ARF reduce the incidence of lagging chromosomes in ARF−/− cells. n > 100 anaphases and telophases from three independent experiments. (G) Levels of the mitotic checkpoint components Mad2 and BubR1 are elevated in ARF−/− MEFs, whereas levels of Bub1 and CENP-E remain unchanged. Tubulin, loading control. *p < 0.05.

To test whether a version of ARF containing a mutation in the MDM2-binding site encoded by exon 1β is capable of rescuing the mitotic checkpoint defect caused by ARF loss, we infected ARF−/− MEFs with a retrovirus expressing either wild-type or V24E p14ARF. Valine 24 occurs in a 15–amino acid peptide (amino acids 16–30) that binds MDM2 (Bothner ), and cells expressing the V24E ARF mutant are impaired in their ability to stabilize p53 and activate its transcriptional activity (Korgaonkar ; di Tommaso ). Interestingly, the V24E mutant was as effective as the wild type in rescuing the mitotic checkpoint defect in ARF−/− cells (Figure 4D), suggesting that this function is independent of p53. Furthermore, both wild-type and V24E ARF rescued the occurrence of misaligned and lagging chromosomes in ARF−/− MEFs (Figure 4, E and F).

ARF−/− cells exhibit p53-independent mitotic defects

Aneuploidy has also been reported in p53−/− MEFs (Harvey ). To formally test whether the mitotic effects of ARF loss were p53 dependent, we bred mice deficient in ARF and p53 to produce animals heterozygous for both genes. Doubly heterozygous mice were then interbred to produce wild-type, ARF−/−, p53−/−, and ARF−/−;p53−/− MEFs, which were examined for mitotic defects. Both ARF−/− and p53−/− cells displayed misaligned chromosomes in the presence of a metaphase plate, but removal of ARF in cells lacking p53 further increased the percentage of cells with misaligned chromosomes (Figure 5A), suggesting that ARF promotes chromosome congression independently of p53. Lagging chromosomes were also observed in both ARF−/− and p53−/− cells. However, cells lacking both tumor suppressors had a higher level of lagging chromosomes than cells lacking either tumor suppressor alone (Figure 5B), indicating that ARF also uses p53-independent mechanisms to prevent lagging chromosomes.

FIGURE 5:

Mitotic functions of ARF are p53-independent. (A) Percentage of MEFs with misaligned chromosomes in metaphase. Loss of ARF increases the incidence of misaligned chromosomes in p53−/− cells. n > 80 metaphases from three independent experiments. (B) Percentage of MEFs with lagging chromosomes in anaphase or telophase. Although loss of either ARF or p53 increases the percentage of cells containing lagging chromosomes, reduction of ARF in p53−/− cells causes a further increase, consistent with a p53-independent role for ARF in preventing chromosome lagging. n > 100 anaphases and telophases from each of three independent experiments. (C) Mitotic index of MEFs of the indicated genotypes ± 16 h of colcemid treatment. ARF, but not p53, is required for the mitotic checkpoint–mediated increase in mitotic index in response to colcemid. n > 250 cells from each of three independent experiments. *p < 0.05.

To assess mitotic checkpoint function, we examined the mitotic index of MEFs with normal and reduced levels of p53 and/or ARF. No differences in mitotic index were observed in the absence of microtubule poisons (Figure 5C, left). After 16 h of colcemid treatment, wild-type cells accumulated in mitosis, as expected (Figure 5C, right). ARF−/− MEFs exhibited a decreased mitotic index relative to wild-type cells, consistent with the results in Figure 4, A– D. However, p53−/− MEFs accumulated in mitosis to the same extent as wild-type MEFs, indicating that p53 is not required for the mitotic checkpoint. Thus, the function of ARF in mitotic checkpoint signaling is independent of the p53 pathway.

Expression of BubR1 and Mad2 is elevated in ARF−/− cells

To determine the mechanism by which ARF functions in the mitotic checkpoint, we first examined ARF localization during mitosis. Although it is well established that ARF localizes to nucleoli during interphase (Weber ), localization of ARF in mitosis has not previously been reported. We found that endogenous ARF had a punctate localization during mitosis. A small portion of the puncta frequently colocalized with spindle poles (Supplemental Figure S1, C and D). However, ARF did not localize to kinetochores in asynchronous or colcemid-treated mitotic cells (Supplemental Figure S1, C and E).

To further understand the effects of ARF on mitotic checkpoint signaling, we compared the expression levels of the mitotic checkpoint proteins Bub1, CENP-E, BubR1, and Mad2 in cells expressing and lacking ARF. Immunoblot analysis of protein extracts from these cells showed that whereas Bub1 and CENP-E levels remained unchanged between wild-type and ARF−/− cells, both BubR1 and Mad2 levels were increased (Figure 4G). Despite the increase in overall protein levels of BubR1 and Mad2, only Mad2 showed enhanced kinetochore recruitment as assessed by quantitative immunofluorescence (Supplemental Figure S2). This could be due to more significant up-regulation of Mad2 or because of saturation of available BubR1-binding sites.

Overexpression of BubR1 does not cause mitotic defects

BubR1 is overexpressed in a variety of human cancers, including those of the breast, lung, and stomach (Weaver and Cleveland, 2006). BubR1 levels are also elevated in ARF−/− splenocytes (Supplemental Figure S3A). To determine whether overexpression of BubR1 caused the mitotic defects seen in ARF−/− MEFs, we incorporated a tetracycline-inducible, YFP-tagged human BubR1 expression vector into chromosomally stable colorectal cancer cells (DLD1) expressing the Tet repressor. Upon treatment with tetracycline, YFP-BubR1 was expressed (Supplemental Figure S3B) and localized to kinetochores (Supplemental Figure S3C). However, expression of YFP-BubR1 did not cause a significant increase in either misaligned or lagging chromosomes (Supplemental Figure S3, D and E). Nor did YFP-BubR1 expression induce supernumery centrosomes (Supplemental Figure S3F) or multipolar spindles (Supplemental Figure S3G) or alter mitotic checkpoint fidelity (Supplemental Figure S3H). Together, these data indicate that increased BubR1 expression is not responsible for the mitotic defects observed in ARF−/− MEFs, consistent with recent findings in BubR1-overexpressing mice (Baker ).

Overexpression of Mad2 does not recapitulate mitotic phenotypes of ARF loss

Because overexpression of BubR1 was not sufficient to mimic the mitotic defects in ARF−/− cells, we next examined whether over­expression of Mad2 could do so. We overexpressed human Mad2 using the same tetracycline-inducible system used for BubR1. Overexpression of Mad2 (Supplemental Figure S4A) resulted in an increase in lagging chromosomes during anaphase and telophase (Supplemental Figure S4B), whereas misaligned chromosomes in metaphase were unaffected (Supplemental Figure S4C). Overexpression of Mad2 also failed to induce abnormal numbers of centrosomes in interphase or multipolar spindles in mitosis (Supplemental Figure S4, D and E). We next examined whether overexpression of Mad2 would weaken the mitotic checkpoint, as did ARF loss. However, overexpression of Mad2 did not have a significant effect on mitotic index (Supplemental Figure S4F), consistent with previous results (Sironi ). It has been reported that reduction of p53 causes overexpression of Mad2 (Schvartzman ), a finding that we were able to recapitulate (Supplemental Figure S4G). Thus, it is unsurprising that up-regulation of Mad2, which is p53-dependent, in not responsible for the p53-independent effects of ARF loss.

Aurora B kinase is overexpressed in ARF−/− cells in vitro and in vivo

Having found that up-regulation of neither Mad2 nor BubR1 could replicate the set of mitotic defects that occur due to ARF loss, we examined other ARF binding partners for evidence of effects on known mitotic proteins. Two transcription factors inhibited by ARF, FoxM1B and Myc, have been found to regulate mitotic genes, including the kinases Aurora A and B, as well as CENP-F (Laoukili ; Courapied ; den Hollander ). We found that levels of Aurora A and CENP-F were unchanged in ARF−/− versus wild-type MEFs (Supplemental Figure S5). However, levels of Aurora B were elevated in ARF−/− MEFs in comparison to wild type (Figure 6A). Of importance, Aurora B expression was also increased in splenocytes of mice lacking ARF (Figure 6B), indicating that levels of Aurora B are increased in vivo as well as in vitro.

FIGURE 6:

Aurora B expression is elevated in ARF−/− cells in vitro and in vivo. (A) Immunoblot showing increased levels of Aurora B in MEFs with reduced expression of ARF. (B) Splenocytes from ARF−/− mice express heightened amounts of Aurora B relative to splenocytes from wild-type animals. Coomassie is used as a loading control.

The half-life of Aurora B is increased in ARF−/− cells

Expression of Myc-GFP increases Aurora B promoter activity ∼30-fold (den Hollander ). ARF binds and inhibits the transcriptional activity of Myc (Qi ), suggesting that transcription of Aurora B is elevated in ARF−/− cells. However, quantitative PCR analysis using two distinct primer sets revealed that Aurora B mRNA levels are only modestly increased in ARF−/− cells compared with wild type (Figure 7A). To determine whether the elevated levels of Aurora B in ARF−/− cells were due to increased protein stability, we treated cells with the protein synthesis inhibitor cycloheximide and followed them over 12 h to determine the half-life of Aurora B protein. The Aurora B half-life in wild-type MEFs was similar to the half-life reported for human Aurora B in HeLa cells (Nguyen ). However, the Aurora B half-life was increased approximately threefold in ARF null cells compared with wild type (Figure 7, B and C), revealing that the overexpression of Aurora B that occurs in the absence of ARF is likely due to effects on protein stability rather than transcription.

FIGURE 7:

Elevated levels of Aurora B in ARF−/− cells are due to increased protein stability. (A) Quantitative real-time PCR examining relative Aurora B mRNA levels in ARF+/+ and ARF−/− cells. Two different primer sets show a modest, nonsignificant increase in Aurora B transcript levels in ARF−/− primary MEFs. p = 0.285 and 0.1088 for primer sets 1 and 2, respectively, as assessed by Wilcoxon signed rank test. n = 3. (B) Immunoblots showing Aurora B levels after the indicated number of hours of cyclohexamide treatment to prevent new protein synthesis. (C) Quantitation of Aurora B protein levels in ARF+/+ and ARF−/− cells after cycloheximide treatment. Aurora B is substantially more stable in ARF−/− than in ARF+/+ cells. n = 3 independent experiments. Mean values for Aurora B protein levels are as follows (90% mean confidence intervals are shown in parentheses): ARF+/+, 1 h, 0.7615 (0.5816, 0.9414); 2 h, 0.5503 (0.4635, 0.6371); 3 h, 0.6008 (0.4052, 0.7964); 6 h, 0.3308 (0.2618, 0.3997); 8 h, 0.4255 (0.3474, 0.5037); 12 h, 0.3417 (0.2518, 0.4315); ARF−/−,1 h, 0.7777 (0.6665, 0.8889); 2 h, 0.9211 (0.8913, 0.951); 3 h, 1.061 (0.8448, 1.277); 6 h, 0.9289 (0.8282, 1.03); 8 h, 0.7647 (0.5965, 0.933); and 12 h, 0.5516 (0.2819, 0.8213).

Overexpression of Aurora B phenocopies ARF loss

To determine whether overexpression of Aurora B was sufficient to induce the mitotic defects seen in ARF−/− cells, YFP-tagged Aurora B was transfected into wild-type MEFs. Aurora B-YFP was expressed (Figure 8A) and localized appropriately to inner centromeres in prometaphase (Figure 8B) and the midbody in telophase (Figure 8C). When challenged with colcemid to activate the mitotic checkpoint and arrest cells in mitosis, wild-type cells transfected with Aurora B exhibited a significantly lower mitotic index than cells transfected with empty vector (Figure 8D), indicating that overexpression of Aurora B alone is sufficient to weaken the mitotic checkpoint.

FIGURE 8:

Overexpression of Aurora B phenocopies ARF loss. (A) Immunoblot showing expression of Aurora B-YFP in primary MEFs. Tubulin is shown as a loading control. (B, C) Aurora B-YFP localizes appropriately. Aurora B-YFP transiently transfected into wild-type MEFs localizes to (B) inner centromeres in prometaphase and (C) the midbody in telophase. Scale bars, 5 μm. (D) Reduced mitotic index in wild-type cells expressing Aurora B-YFP after 16 h of treatment with 100 ng/ml colcemid, showing that overexpression of Aurora B is sufficient to weaken mitotic checkpoint signaling. n > 250 cells from each of three independent experiments. (E–I) Wild-type MEFs transfected with Aurora B-YFP have elevated levels of (E) misaligned chromosomes in metaphase (n ≥ 50 cells with metaphase plates from three independent experiments), (F) lagging chromosomes in anaphase and telophase (n > 100 anaphases and telophases from three independent experiments), (G) binucleate interphase cells (n > 250 cells from each of three independent experiments), (H) supernumery (>2) centrosomes in interphase (n > 250 cells from each of three independent experiments), and (I) multipolar spindles in mitosis (n > 100 mitotic cells from each of three independent experiments), as compared with wild-type MEFs transfected with empty vector (EV). *p < 0.05.

Elevated expression of Aurora B also caused other mitotic defects. Wild-type MEFs transfected with Aurora B had increased frequencies of misaligned chromosomes in metaphase (Figure 8E) and lagging chromosomes during anaphase and telophase (Figure 8F), similar to what was observed in ARF−/− MEFs (Figure 3, A– F). Increased Aurora B expression was also sufficient to cause elevated levels of binucleate cells (Figure 8G), consistent with cytokinesis failure and the formation of tetraploid cells. As would be expected after tetraploidization, expression of Aurora B-YFP resulted in supernumery centrosomes during interphase (Figure 8H) and multipolar spindles during mitosis (Figure 8I). Together, these results demonstrate that overexpression of Aurora B is sufficient to phenocopy ARF loss.

Partial depletion of Aurora B rescues mitotic defects in ARF−/− cells

To further confirm that elevated levels of Aurora B were responsible for the mitotic phenotypes observed in ARF−/− cells, Aurora B was depleted to near wild-type levels by transfection of two distinct small interfering RNA (siRNA) sequences. Transfection of either siRNA or a combination of both depleted Aurora B in ARF−/− cells to a level similar to that in wild-type MEFs (Figure 9A). Partial depletion of Aurora B rescued mitotic checkpoint function in ARF−/− MEFs and restored their ability to accumulate in mitosis in response to microtubule depolymerization (Figure 9B). Reducing levels of Aurora B in ARF−/− MEFs also rescued their chromosome congression defect (Figure 9C) and significantly reduced their incidence of lagging chromosomes (Figure 9D). These experiments show that normalization of Aurora B levels is sufficient to ameliorate the mitotic defects characteristic of ARF−/− cells.

FIGURE 9:

Partial knockdown of Aurora B to near wild-type levels rescues mitotic defects in ARF−/− cells. (A) Immunoblot demonstrating partial knockdown of Aurora B in ARF−/− primary MEFs to levels similar to those found in wild-type MEFs. Tubulin is shown as a loading control. (B) Partial knockdown of Aurora B rescues the mitotic checkpoint defect in ARF−/− cells. n > 250 cells from each of three independent experiments. (C) Partial depletion of Aurora B by siRNA rescues the chromosome alignment defect in ARF−/− cells. n > 50 cells with visible metaphase plates from each of three independent experiments. (D) Knockdown of Aurora B by siRNA to near-wild-type levels in ARF−/− cells partially rescues the occurrence of lagging chromosomes. n > 100 anaphases and telophases from each of three independent experiments. *p < 0.05.

DISCUSSION

Evidence for p53-independent tumor-suppressive functions of ARF has existed for more than a decade (Weber ), but a full understanding of these numerous effects has remained elusive. At high levels, ARF can impede progression through interphase (Weber ; Yarbrough ; Eymin ; Datta ) and sequester—and thereby inhibit—the potentially oncogenic transcription factors Myc and E2F in a p53-independent manner (Ozenne ). There is also evidence that ARF participates in the DNA damage response in a p53-independent, as well as a p53-dependent, manner (Eymin ). In this context, we provide evidence that ARF maintains chromosomal stability in a p53-independent manner. Because CIN is a hallmark of human cancers, this is likely to be an additional mechanism by which ARF enacts tumor suppression independently of p53.

It was previously proposed that p53 is essential for the mitotic checkpoint after p53−/− MEFs showed a decreased mitotic index and increased ploidy in response to exposure to the microtubule poison nocodazole (Cross ). However, subsequent studies found that p53−/− cells delay in mitosis for an equivalent amount of time as wild-type cells after treatment with microtubule poisons in both fixed (Kienitz ) and live (Lanni and Jacks, 1998) assays. The increased ploidy of the nocodazole-treated p53−/− MEFs was the result of a G1 function of p53 in preventing rereplication of DNA after mitotic slippage (Di Leonardo ). Thus, p53 is not required for an intact mitotic checkpoint response. Consistent with these previous studies, we show that the mitotic checkpoint defect observed in ARF−/− cells is independent of p53 (Figure 5C).

The role of Aurora B in the mitotic checkpoint has also been a matter of some debate. Budding yeast contain a single Aurora family homologue, Ipl1. Ipl1 and Aurora B play well-established roles in destabilizing improper kinetochore–microtubule attachments that do not place sister chromatids under tension (Biggins ; Biggins and Murray, 2001; Ditchfield ; Hauf ; Lan ). This error-correction activity results in the production of unattached kinetochores, which activate the mitotic checkpoint. Whether Ipl1/Aurora B activates the mitotic checkpoint merely by generating unattached kinetochores or whether additional functions are required has been disputed. In budding yeast, Ipl1 is required for a mitotic checkpoint arrest caused by lack of tension but not by lack of attachment (Biggins ; Biggins and Murray, 2001; King ). However, the fission yeast homologue, ark1, is required for the attachment-sensing checkpoint as well (Petersen and Hagan, 2003). Experiments in human cells with the Aurora B inhibitors Hesperadin and ZM447439 demonstrated that Aurora B–inhibited cells were unable to mount a robust, long-term mitotic checkpoint response to the microtubule-stabilizing drug Taxol, which produces attached kinetochores that are not under tension. However, the checkpoint response was maintained in the absence of kinetochore–microtubule attachments caused by the microtubule-destabilizing drug nocodazole (Ditchfield ; Hauf ). This was consistent with the findings in budding yeast. Because Aurora B–inhibited cells were capable of maintaining arrest in nocodazole, when all kinetochores are unattached, these data were often interpreted as indicating that Aurora B kinase activity is not required for mitotic checkpoint signaling. However, more recent evidence indicates that Aurora B kinase activity is incompletely inhibited by standard drug concentrations. Higher concentrations of Aurora B inhibitors anticipated to inhibit >95% of kinase activity prevent mitotic arrest in the absence of microtubules, demonstrating a requirement for Aurora B, and its associated kinase activity, in sensing attachment as well as tension (Santaguida ). Consistent with this, a mutant of the Aurora B binding partner INCENP that is sufficient for correction of erroneous kinetochore–microtubule interactions is defective in supporting mitotic checkpoint activity (Vader ). In addition, function-blocking Aurora B antibodies abrogate both the tension- and attachment-based mitotic checkpoints (Kallio ), as does RNA interference–mediated depletion of Aurora B (Ditchfield ). Thus, the weight of the evidence favors the interpretation that Aurora B functions in the mitotic checkpoint response to unattached kinetochores in higher eukaryotes. Intriguingly, our data show that overexpression of Aurora B also impairs mitotic checkpoint signaling.

Up-regulation of Mad2 and BubR1 in ARF−/− cells is likely to be a p53-dependent phenotype. Mad2 levels are elevated in p53−/− MEFs (Schvartzman ; Supplemental Figure S4G) and in cells lacking the p53 transcriptional target and CDK inhibitor p21 (Schvartzman ). Conversely, ectopic expression of p21 in wild-type MEFs results in lower levels of Mad2 protein. This reduction in Mad2 does not occur in cells triply negative for Rb and the pocket proteins p107 and p130, indicating that p21 is acting through the Rb pathway. In wild-type cells, p21 inhibits cyclin D/CDK4, which hyperphosphorylates and inactivates Rb, leading to activation of E2F transcription factors and S-phase entry. Consistent with the known pathway of cyclin D/CDK4–mediated inactivation of Rb, p21 mutants that cannot bind CDKs were unable to repress Mad2 promoter activity (Schvartzman ). Thus, reduced stability of p53 is likely to contribute to elevated levels of Mad2 in ARF−/− cells. A similar mechanism is likely to explain up-regulation of BubR1 in p53−/− (Schvartzman ) and ARF−/− cells, although data supporting the opposite conclusion—that p53 activates BubR1 transcription—have also been reported (Oikawa ).

Interestingly, both reduced and elevated expression of mitotic checkpoint components, including Mad2, BubR1, and Aurora B, can result in similar phenotypes. Reduction of each of these components results in chromosome missegregation (Dobles ; Michel ; Ditchfield ; Hauf ; Honda ; Baker ; Meraldi ). Overexpression of Mad2 (Sotillo ; Supplemental Figure S4B) or Aurora B (Figure 8, E and F), but not BubR1 (Baker ; Supplemental Figure S3), also causes segregation defects. With respect to the mitotic checkpoint, a decrease of Mad2, BubR1, or Aurora B impairs mitotic checkpoint signaling (Hoyt ; Li and Murray, 1991; Kallio ; Ditchfield ; Hauf ; Santaguida ). Elevated expression of Mad2 delays fission yeast (He ; Kim ), Xenopus extracts (Chen ; Fang ), and certain tissue culture cells (Howell ; Sotillo ) in metaphase. In contrast, overexpression of BubR1 does not hyperactivate mitotic checkpoint signaling. Elevated levels of BubR1 have no notable effect on mitotic index in asynchronously cycling cells (Baker ; Supplemental Figure S3H) and can actually rescue a checkpoint defect caused by heterozygous loss of Rae1 (Baker ). Here we show that Aurora B overexpression results in a third outcome—reduction of mitotic checkpoint activity (Figure 8D). With respect to their tumor phenotypes, mice heterozygous for Mad2 or BubR1 have a relatively subtle increase in spontaneous (Mad2; Michel ) or carcinogen-induced (BubR1; Dai ; Baker ) tumors. Overexpression of Mad2 causes a substantially more severe tumor phenotype than reduction of Mad2, in which tumors occur with increased penetrance and earlier onset (Sotillo ). BubR1 overexpression, on the other hand, actually decreases tumor incidence and extends lifespan (Baker ). Animal models with ubiquitously elevated levels of Aurora B have not been reported, but Aurora B overexpression is common in human tumors, in which it correlates with poor prognosis (Bischoff ; Ehara ; Araki ; Sorrentino ; Chieffi ; Vischioni ). Thus, both increased and decreased expression of these genes sometimes results in similar phenotypes. Whereas previous work showed that reduction of Aurora B protein or kinase activity results in mitotic defects, the data reported here reveal that overexpression of Aurora B causes similar defects, indicating that levels of Aurora B must be tightly controlled to prevent aneuploidy.

It is interesting to note that the duration of the mitotic checkpoint response to microtubule poisons is substantially longer in human than in rodent cells (Rieder and Maiato, 2004). It has long been known that rodent cells are less sensitive than human cells to perturbations of the microtubule cytoskeleton (Gupta, 1985). Rodent cells are also more likely to exit mitosis and resynthesize their DNA in the presence of microtubule poisons than are human cells (Kung ). In both cases, Chinese hamster cells are the least responsive to microtubule perturbation, whereas mouse cells exhibit an intermediate phenotype between hamster and human cells. Rodent cells exhibit reduced uptake of vinblastine, colchicine, and Taxol compared with human cells, suggesting that lower internal concentration may account for the decrease in mitotic arrest and cell death (Gupta, 1985; Parekh and Simpkins, 1996). However, more recent experiments suggested that differences in phosphoregulation of the N-terminus of Mad1 confer species-specific differences to the stringency of the mitotic checkpoint (Haller ). It will now be of interest to determine the consequences of ARF loss on the longer mitotic arrest observed in human cells.

Although independent of p53, our results demonstrate that the mitotic checkpoint defect caused by loss of ARF is a result of overexpression of Aurora B. Elevated Aurora B levels have been observed in a number of human malignancies, including non–small cell lung cancer (Vischioni ), hepatocellular carcinoma (Lin ), epithelial ovarian cancer (Chen ), and prostate cancer (Chieffi ). Further, high expression of Aurora B was correlated with poor prognosis in these cancers, and inhibitors of Aurora B are in clinical trials (Komlodi-Pasztor ). Our results show that overexpression of Aurora B is sufficient to weaken the mitotic checkpoint and results in chromosome missegregation and CIN. ARF loss, by gene deletion, mutation, or promoter methylation, is likely to be one mechanism responsible for Aurora B overexpression in tumors.

MATERIALS AND METHODS

Animals, cell culture, and treatments

Animals were maintained in a C57BL/6 background and handled in accordance with the policies of the Institutional Animal Care and Use Committee of the University of Wisconsin–Madison. Primary MEFs were generated from embryonic day 14.5 (E14.5) embryos in 6-cm dishes with 3 ml of chilled 0.05% trypsin/EDTA. After the head, liver, and tail (for genotyping) were removed, embryos were minced using scissors, pipetted up and down using a 10-ml pipette, and incubated at 37°C for 15 min. Embryos were pipetted up and down with an additional 2 ml of trypsin/EDTA and a 5-ml pipette before further incubation at 37°C for 10 min. The solution was transferred to a 15-ml conical tube with 5 ml of primary MEF media. After permitting debris to settle for 1–2 min, we transferred the supernatant to a second 15-ml conical tube. MEFs in the supernatant were pelleted and resuspended in high-glucose DMEM (Invitrogen, Carlsbad, CA) containing 15% fetal bovine serum (FBS; Sigma-Aldrich, St. Louis, MO), 0.1 mM nonessential amino acids (Invitrogen), 1 mM sodium pyruvate (Invitrogen), 1 μM 2-mercaptoethanol (Acros Biological/Thermo Fisher, Waltham, MA), 2 mM l-glutamine (Invitrogen), and 50 μg/ml penicillin/streptomycin (Invitrogen) and cultured in 3% O2 and 10% CO2 at 37°C. Experiments were performed on MEFs between passages 3 and 20. Retroviruses (a kind gift from Dawn Quelle, University of Iowa) were added along with 8 μg/ml Polybrene 48 h before harvest of cells.

Flp-In TRex DLD1 cells (Invitrogen) were grown in high-glucose DMEM, 10% FBS, 2 mM l-glutamine, and 50 μg/ml penicillin/streptomycin and cultured in 5% CO2 at 37°C. Unless otherwise indicated, colcemid (Enzo, Farmingdale, NY) was used at 100 ng/ml, and mitotic indices were collected after 16 h of treatment. To induce expression of Mad2 and BubR1 transgenes, DLD1 cells were treated with 0.25 μM tetracycline for 48 h before analysis.

siRNA, 40 nM, directed against luciferase GL2 (control) or Aurora B (SASI_Mm01 00056605 and 00056608; Sigma-Aldrich) was transfected using Dharmafect 4. Analysis was performed 48 h posttransfection.

Chromosome (metaphase) spreads

Chromosome spreads were performed as in Weaver ). Briefly, cultured cells were grown to 80% confluence in 6-cm dishes before being treated with 100 ng/ml colcemid for ∼4 h. Cells were then harvested, pelleted, and resuspended with 5 ml of room temperature 75 mM KCl for 11 min. Cells were treated with 1 ml of fresh fix (3:1 methanol:acetic acid) and pelleted before being placed in 4 ml of fresh fix overnight at 4°C.

For splenocyte spreads, spleens were shredded using two pairs of forceps. Single cells were obtained by pipetting using a P1000 and a Pasteur pipette and transferred to 15-ml tubes. After fragments settled, cells in the supernatant were pelleted and resuspended in 4 ml of RPMI 1640 (HyClone) with 100 μg/ml gentamicin (Life Technologies, Grand Island, NY), 11 μg/ml phytohemagglutinin (Sigma-Aldrich), 75 μg/ml lipopolysaccharide (Invitrogen), 10% FBS, and 5 μg/ml colchicine and cultured in 15-ml tubes at 37°C and 10% CO2 for 12 h. Splenocytes were pelleted, resuspended in 5 ml of 75 mM KCl prewarmed to 37°C, and incubated at 37°C for 45 min. One milliliter of 3:1 methanol and acetic acid fix was added before centrifugation. Cells were resuspended in 4 ml of fixative and stored overnight at 4°C.

Cultured cells and splenocytes were pelleted and resuspended in fresh fix twice before being dropped onto precleaned microscope slides and dried at 75°C. DNA was visualized with 4′,6-diamidino-2-phenylindole (DAPI).

Immunoblots

Cells from ∼90% confluent 6-cm dishes were trypsinized, washed with phosphate-buffered saline (PBS), resuspended in 100 μl of ELB lysis buffer (250 mM NaCl, 0.1% NP-40, 50 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid [HEPES], pH 7, 5 mM EDTA) and 25 μl of 5× sample buffer, boiled for 10 min, and stored at −80°C. We ran 30 μg of samples on 12% acrylamide gels and transferred them to nitrocellulose. Primary and secondary antibodies were diluted in 5% milk in Tris-buffered saline plus 0.1% Tween 20. Primary antibody dilutions were as follows: Bub1, 1:500; BubR1, 1:200 (Taylor ); CENP-E, 1:200 (Brown ); Mad2, 1:200 (Kops ); Aurora B, 1:500 (Cell Signaling); α-tubulin, 1:250 (DM1a; Sigma-Aldrich), and ARF, 1:250 (ab80; Abcam).

Immunofluorescence microscopy

Cells were grown in 12-well plates with 18-mm round coverslips until ∼80% confluent. Coverslips were washed with microtubule stabilizing buffer (MTSB; 100 mM 1,4-piperazinediethanesulfonic acid, pH 6.9, 30% glycerol, 1 mM ethylene glycol tetraacetic acid, and 1 mM MgSO4) and fixed with precooled methanol at −20°C for 5 min or 4% formaldehyde or 0.5% glutaraldehyde (Tousimis, Rockville, MD) in MTSB at room temperature for 10 min. Glutaraldehyde was quenched with 0.2% NaBH4 for 20 min. Coverslips were washed twice with PBS and blocked in triton block (0.2 M glycine, 2.5% FBS, and 0.1% Triton X-100 in PBS) at 4°C. For formaldehyde and glutaraldehyde fixations, cells were preextracted with 0.5% Triton X-100 in MTSB for 1–5 min at 37°C. Primary antibody dilutions were as for immunoblots except for Bub1 (1:200), Aurora B (1:2500), and α-tubulin (YL1/2; 1:500). Images were acquired on a Nikon Eclipse Ti-E inverted fluorescence microscope using a CoolSNAP HQ2 camera and a 100×/1.4 numerical aperture (NA) oil objective. Chromosome spread images are from a single z. Other images are maximum projections of 0.2-μm z-stacks deconvolved using the AQI module in Nikon Elements unless otherwise indicated.

Fluorescence in situ hybridization

Small intestine was dissected, washed with PBS, and fixed in 10% buffered Formalin (Fisher, PIttsburgh, PA), followed by three washes in 70% ethanol every 24 h. Paraffin sections, 8 μm, were dewaxed in xylene and hydrated in an ethanol series before cross-linking reversal in 8% sodium thiocyanate (Fisher) for 30 min at 80°C. Tissue was digested with 0.5 mg/ml proteinase K for 5 min at 37°C and washed in 2× SSC buffer (300 mM NaCl, 30 mM sodium citrate) for 2 min before dehydration in an ethanol series. A 200-kb FISH probe recognizing the thymidine kinase locus on chromosome 11 (Kreatech KBI-30501, Buffalo Grove, IL) was added to each tissue before coverslipping and sealing with rubber cement. Slides were placed on an 80°C metal plate for 5 min for denaturation before hybridizing 48 h at 37°C. After removal of coverslips and rubber cement, tissues were washed in 0.4× SSC plus 0.3% NP40 at 72°C for 2 min before a second wash in 2× SSC plus 0.1% NP40. After dehydration in an ethanol series, 0.3 μg/ml DAPI in Vectashield was used to mount the samples and counterstain. Images were acquired using a Nikon Eclipse Ti-E fluorescence microscope using a 100×/1.4 NA objective and a Hamamatsu Orca Flash 4.0 camera.

Quantitative real-time PCR

RNA was isolated from MEFs using TRIzol. One microgram of the resulting RNA was reverse transcribed (iScript; Bio-Rad, Hercules, CA). cDNA was resuspended in nuclease-free water in a total volume 100 μl. For real-time PCR analysis, 1 μl of resuspended cDNA was used in a total reaction volume of 25 μl containing 300 nM forward and reverse primers and 1× IQ SYBR Green Supermix (Bio-Rad). An annealing temperature of 55°C was used during thermocycling for all primers. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as an internal control, and melting curves of products were performed at the end of each experiment to confirm the amplification of a single product. Relative mRNA levels were calculated using the ΔΔC method. Primer sequences: Aurora B set 1, forward, 5′-CAGAAGGAGAACGCCTACCC; Aurora B set 1, reverse, 5′-GAGAGCAAGCGCAGATGTC-3′; Aurora B set 2, forward, 5′-TCAGAAGGAGAACGCCTACCC-3′; Aurora B set 2, reverse, 5′-GACTCTCTGGGACAACGTGTT-3′. GAPDH forward, 5′-CCA ATG TGT CCG TCG TGG ATC-3′; reverse, 5′-GTT GAA GTC GCA GGA GAC AA-3′.

Statistical analysis

Error bars represent mean ± SE unless otherwise specified. Statistical significance was concluded at p < 0.05.