Depletion of the Poly(C)-binding Proteins αCP1 and αCP2 from K562 Cells Leads to p53-independent Induction of Cyclin-dependent Kinase Inhibitor (CDKN1A) and G1 Arrest*

The α-globin poly(C)-binding proteins (αCPs) comprise an abundant and widely expressed set of K-homolog domain RNA-binding proteins. αCPs regulate the expression of a number of cellular and viral mRNAs at the levels of splicing, stability, and translation. Previous surveys have identified 160 mRNAs that are bound by αCP in the human hematopoietic cell line, K562. To explore the functions of these αCP/mRNA interactions, we identified mRNAs whose levels are altered in K562 cells acutely depleted of the two major αCP proteins, αCP1 and αCP2. Microarray analysis identified 27 mRNAs that are down-regulated and 14 mRNAs that are up-regulated in the αCP1/2-co-depleted cells. This αCP1/2 co-depletion was also noted to inhibit cell proliferation and trigger a G1 cell cycle arrest. Targeted analysis of genes involved in cell cycle control revealed a marked increase in p21WAF mRNA and protein. Analysis of mRNP complexes in K562 cells demonstrates in vivo association of p21WAF mRNA with αCP1 and αCP2. In vitro binding assays indicate that a 127-nucleotide region of the 3′-untranslated region of p21WAF interacts with both αCP1 and αCP2, and co-depletion of αCP1/2 results in a marked increase in p21WAF mRNA half-life. p21WAF induction and G1 arrest in the αCP1/2-co-depleted cells occur in the absence of p53 and are not observed in cells depleted of the individual αCP isoforms. The apparent redundancy in the actions of αCP1 and αCP2 upon p21WAF expression correlates with a parallel redundancy in their effects on cell cycle control. These data reveal a pivotal role for αCP1 and αCP2 in a p53-independent pathway of p21WAF control and cell cycle progression.

Each ␣CP isoform contains three copies of the hnRNP K homology RNA binding domain (9). ␣CPs, along with hnRNP K, are uniquely characterized by in their strong binding preference for C-rich motifs. This subset of hnRNP K homology domain proteins has been linked to post-transcriptional controls via binding to elements in 5Ј-and 3Ј-untranslated regions (UTRs) of cellular and viral mRNAs (10 -19). For example, ␣CP1 and/or ␣CP2 regulate the stability of the mRNAs encoding ␣2-globin, tyrosine hydroxylase, and ␣1(I) collagen via binding to 3Ј-UTR motifs and mediate control over the translation of specific mRNAs, including 15-lipoxygenase, CCAAT/ enhancer-binding protein ␣, folate receptor, and phosphatase 2A, by binding to either 5Ј-or 3Ј-UTR elements. In addition to regulating the expression of several cellular mRNAs, ␣CP can also regulate a number of distinct steps in viral gene expression (11, 20 -29). Taken together, these studies indicate that ␣CPs constitute key regulators in a wide spectrum of post-transcriptional controls.
To develop an understanding of how ␣CPs impact on cell function, we have screened for in vivo binding targets. Microarray analysis of immunoenriched ␣CP2-mRNP complexes isolated from K562 cells (30) revealed 160 ␣CP2-associated mRNAs. These mRNAs could be clustered according to the function(s) of their encoded proteins, suggesting roles for ␣CP2 in coordination of post-transcriptional controls. One of the larger functional clusters consisted of mRNAs that affect cell growth and proliferation. A role for ␣CP2 in cell cycle control was consistent with prior observations that a member of the ␣CP family, ␣CP4, can induce cell cycle arrest at G 2 -M and stimulate apoptosis (31,32).
The current study was initiated to assign functions to ␣CP interactions with cellular mRNAs (30). To accomplish this goal, we acutely depleted K562 cells of ␣CP1 and ␣CP2, either separately or together, and identified mRNAs that were either induced or repressed in their steady state levels. During the course of these studies, we observed that the ␣CP1/2 co-depletion decreased cell proliferation and triggered a G 1 arrest. The basis of the mitotic arrest was explored by determining the effect of the ␣CP1/2 co-depletion on the expression of genes that play pivotal roles in cell cycle control. These studies revealed an induction of the cyclin-dependent kinase inhibitor 1A (CDKN1A) mRNA and protein. CDKN1A is also known as wild-type p53 activated fragment (p21 WAF ), and we will use this designation throughout. The induction of p21 WAF mRNA and protein correlated with the G 1 arrest. p21 WAF mRNA was found to be associated with both ␣CP1 and ␣CP2 mRNP complexes in untreated cells, and the induction of p21 WAF mRNA subsequent to ␣CP1/2 co-depletion was mechanistically linked to prolongation of the p21 WAF mRNA half-life. These data lead us to conclude that ␣CP1 and ␣CP2 play a role in cell cycle control via a p53-independent, post-transcriptional modulation of p21 WAF gene expression.
Western Blot Analysis-Radioimmune precipitation assay buffer (1% IGEPAL CA-630, 0.5% sodium deoxycholate, 0.1% SDS in phosphate-buffered saline) lysates were isolated (Santa Cruz Biotechnology, Inc., Santa Cruz, CA), and protein was quantified using the Bio-Rad D/C kit. Extracts prepared from HCT116 cells that had been treated with 50 M irinotecan for 24 h were provided (kind gift of N. Finnberg; W. El-Deiry laboratory, University of Pennsylvania). ␣CP1, ␣CP2, and rpL7 antibodies were generated by our laboratory; antibodies to p53 and p21 WAF were purchased from Santa Cruz Biotechnology; and lamin A/C, CCNH, and RB antibodies were purchased from Cell Signaling Technology. Antibody-bound proteins were visualized by Western analysis using ECL Plus (Amersham Biosciences).
RNA Isolation and Microarray Analysis-Analysis of a G4112A hybridization microarray (Agilent) representing 25,584 human genes was performed by Mogene using a 2-g aliquot of total RNA isolated from cells 2 days post-siRNA transfection (RNAeasy; Qiagen).
qRT-PCR Analysis-cDNA was synthesized from total RNA (1 g) (Reaction Ready First Strand cDNA Synthesis; Super-Array), and the cDNA product, diluted 13-fold with H 2 O, was used as a template for quantitative PCR (RT 2 Sybr Green/Rox; SuperArray or Taqman reagents; ABI). The following sets of primers were purchased from ABI: FADS1 (fatty acid desaturase 1) (Hs00203685_m1), RASSF5 (Ras association domain family protein 5) (Hs00739100_m1), and RIG-I (Hs00204833_m1). The following sets of primers were purchased from SuperArray: CCNH (PPH00969A), HIF1␣ (PPH01361A), PHGDH (PPH07199A), and UBCH2 (PPH18206A). Reactions were run in triplicate on an ABI Prism 7700, and the data were analyzed using Sequence Detection Software version 1.9.1. The Ct values obtained are an average of the triplicates.
Analysis of the Cell Cycle and Viability-Fluorescence-activated cell sorting analysis was performed on cells 4 and 5 days post-siRNA transfection (Easycyte Mini; Guava). Cell cycle reagent (containing propidium iodide) was used to analyze the cell cycle, and a minimum of 2000 cells were analyzed according to the manufacturer's protocol (Guava). The Viacount Dye Exclusion Assay was used to measure viability (Guava).
mRNA Half-life Determination-Actinomycin D (Sigma) was added to the media (5 g/ml) at 2 days post-siRNA transfection and total RNA was collected at subsequent 0, 2, and 4 h time points. 4-g samples were analyzed by Northern blotting (34) using a p21 WAF cDNA probe (Origene) labeled with 32 P (RadPrime DNA Labeling Kit; Invitrogen). Band intensities were quantified on a Storm Phosphor-Imager (Amersham Biosciences).
Cross-linking Immunoprecipitation Analysis-Plasmids containing a T7 promoter upstream of various regions corresponding to the p21 WAF 3Ј-UTR were a kind gift of P. Leedman (University of Western Australia) (33) and are depicted in Fig. 7A. These plasmids were linearized with HindIII prior to transcription. The p21 WAF 3Ј-UTR regions designated WAF1-A to WAF1-E (Fig. 7A) were amplified by PCR. The primers are indicated in Table 4. Note that each forward primer contained . Protein lysates were analyzed by Western blotting using antibodies directed against the corresponding proteins. The two bands detected with the ␣CP2-specific antibodies represent ␣CP2 full-length protein (upper band) and ␣CP2KL, its major splice variant. The two bands in the lamin A/C Western blot panel represent the two lamins, A and C. B, the simultaneous depletion of ␣CP1 and ␣CP2 results in decreased cyclin H protein expression. Cells were mock-transfected or transfected with a mixture of siRNAs directed against ␣CP1 and ␣CP2 (␣CP1 ϩ 2) or directed against lamin A/C (Lam). Protein lysates were analyzed by Western blotting using antibodies directed against cyclin H or an antibody that recognizes RB (as a loading control). a T7 promoter (TAATACGACTCACTATAGG) at its 5Ј-end, which is not included in the primer sequences listed in Table 4. The p21 WAF cDNA (Origene) was used as a template for the PCR. PCR was performed using the Platinum Pfx DNA polymerase (Invitrogen) according to the manufacturer's instructions except that we used 2ϫ Pfx amplification buffer, 0.2 g of template, 0.4 l of Pfx DNA polymerase, and 1ϫ PCRx enhancer per reaction. The conditions were 94°C for 5 min; 30 cycles of 94°C for 15 s, 55°C for 30 s, 68°C for 84 s; and 68°C for 7 min. Fragments were gel-isolated using the QIAquick Gel Extraction Kit (Qiagen). Linearized plasmids or PCR products were used as templates for transcription of radiolabeled thiolated RNAs as described (30). These RNAs contained thiolated uridines, which allow for cross-linking of the thiol group to a binding protein located within a few Å of the thiol moiety. The RNAs were incubated with cytoplasmic extract from K562 cells and irradiated at 312 nm to activate the protein/RNA crosslink. Following irradiation, the samples were treated with RNase A to remove the unprotected RNA. The samples were then immunoprecipitated with antibodies specific to ␣CP1 or ␣CP2 (both generated by our laboratory) or c-Myc antibodies (Santa Cruz Biotechnology) and analyzed by SDS-PAGE as described (30).

RESULTS
siRNA-mediated Depletion of ␣CP1 and ␣CP2 in K562 Cells-␣CP1 and ␣CP2 siRNAs were transfected either individually or in combination. Western blot analyses revealed that the ␣CP1 and ␣CP2 siRNAs selectively depleted their targeted proteins and that both ␣CPs isoforms were depleted when the two siRNAs were used in combination (Fig. 1A). Two controls were included to document specificity of siRNA actions: a "mock" transfection lacking only siRNAs and a transfection with an unrelated siRNA directed against lamin A/C. Expression of  lamin A/C protein was unaffected by either of the ␣CP siRNAs but was effectively cleared by the lamin A/C siRNA.
Alterations of mRNA Steady State Levels in ␣CP1/2-codepleted Cells-To identify cellular mRNAs whose expression is modulated by ␣CP1 and ␣CP2, RNA isolated from control and ␣CP1/2-co-depleted cells were compared by microarray analysis. The study included six sets of microarray hybridizations, beginning each time with an independent siRNA transfection; three of these studies compared combined treatment with ␣CP1 and ␣CP2 siRNAs ("␣CP1/2 co-depletion") with mock-transfected cells, and the other three compared the codepletion with lamin A/C siRNA treatment. 41 mRNAs were altered 1.7-fold or more in both of these comparisons. Table 1 identifies each mRNA by GenBank TM accession number, average -fold changes, and putative function(s) of the encoded proteins, as identified by GO (gene ontology) terms, OMIM (online Mendelian inheritance in man), and/or manual literature searches. The mRNAs are ranked according to the average -fold change of all six experiments. There were 27 down-regulated mRNAs and 14 up-regulated mRNAs. The mRNAs encoding ␣CP1 and ␣CP2 ranked highest on this list of down-regulated mRNAs, with ␣CP2 (or its variants) ranked at number 1, 2, and 6 with -fold changes of Ϫ7, Ϫ7, and Ϫ3 compared with mock and Ϫ5, Ϫ4, and Ϫ3 compared with lamin A/C. Likewise, ␣CP1 ranked at number 3 with a -fold change of Ϫ4 compared with mock or lamin A/C. These results validate the effective siRNA targeting of ␣CP1 and ␣CP2 mRNAs. The fourth highest ranking candidate is the mRNA encoding the cell cycle regulator, CCNH (cyclin H). This mRNA was decreased by 6-fold when ␣CP1/2 co-depletion was compared with mock depletion and by 3-fold when compared with lamin A/C. The mRNA encoding HIF1␣ (hypoxia-inducible transcription factor 1␣) ranked at number 10. In the list of mRNAs that were up-regulated by the ␣CP1/2 codepletion, an anonymous mRNA ranked the highest, with a 6-fold increase in its steady state level when compared with mock and an 11-fold increase when compared with lamin A/C.
Verification of the microarray data was carried out on selected mRNAs by targeted qRT-PCR ( Table 2). The level of each mRNA was determined as the -fold change versus the mock transfection control and the -fold change versus the Lamin A/C siRNA control. The mRNA encoding CCNH was 3and 2-fold lower in the ␣CP1/2 co-depleted cells compared with mock or lamin A/C knockdowns, respectively. Likewise, the mRNAs encoding HIF1␣, FADS1 (fatty acid desaturase 1), and RASSF5 (Ras association domain family protein 5) were decreased in the ␣CP1/2-co-depleted cells, all in agreement with the microarray analysis. The increase in the levels of the mRNAs encoding PHGDH (phosphoglycerate dehydrogenase), RIG-1 (retinoic acid-inducible gene 1), and UbcH2 (ubiquitinconjugating enzyme E2H), as determined by the microarray analysis of ␣CP1/2 co-depleted cells, were confirmed by the qRT-PCR analysis, as was the lowest ranked up-regulated mRNA (rank number 14) in the microarray analysis (RIG-I; Table 1). These qRT-PCR studies support the reliability of the microarray data set.
Expression of Cyclin H Is Decreased in ␣CP1/2-co-depleted Cells-Since CCNH mRNA was the most strongly down-regulated mRNA in the ␣CP1/2 co-depleted cells (excluding ␣CPs), we assessed the corresponding impact on CCNH protein. Western blot analysis was consistent with the mRNA analysis, revealing that CCNH protein was reduced by ϳ50% in the ␣CP1/2-co-depleted cells (Fig. 1B). The levels of CCNH protein in cells individually depleted of either ␣CP1 or ␣CP2 were reduced by ϳ20 -30% in each case (data not shown). These data suggest that ␣CP1 or ␣CP2 can each regulate CCNH protein expression, but together the effect is additive. Taken together, the data support the conclusion that CCNH mRNA and protein are markedly and coordinately reduced in cells depleted for ␣CP1 and/or ␣CP2.
Co-depletion of ␣CP1 and ␣CP2 Results in a G 1 Cell Cycle Arrest-The observed reduction of CCNH protein in ␣CP-depleted cells suggested that ␣CP1 and ␣CP2 levels might impact on cell cycle kinetics. To test this possibility, ␣CP1 and ␣CP2 were depleted from K562 cells both individually and in combination, and cell replication parameters were evaluated in comparison with mock transfection and lamin A/C siRNA transfec-tion controls. This analysis revealed a 53% reduction in cell number subsequent to co-depletion of ␣CP1 and ␣CP2 compared with mock-treated cells, and a 41% reduction in cell number when the ␣CP1/2 co-depletion was compared with lamin A/C siRNA-transfected cells at 4 days post-transfection of siR-NAs. In contrast, there was no significant decrease in the density of cells individually depleted of ␣CP1 or ␣CP2 or in the controls (data not shown). Fluorescence-activated cell sorting analysis of cells transfected with siRNAs revealed that the ␣CP1/2 co-depletion resulted in accumulation of cells in the G 1 phase that was not apparent in the mock-transfected or lamin A/C-transfected controls (Fig. 2). The G 1 arrest in the ␣CP1/2 knockdown was observed in three independent knockdown studies. When ␣CP1/2-co-depleted cells were compared with control cells, the differences in the accumulation of G 1 phase cells were highly significant (p ϭ 0.0009 when compared with mock-transfected and p ϭ 0.0023 when compared with lamin A/C siRNA-transfected cells). This G 1 arrest was accompanied by a reciprocal decrease of cells in S phase (Fig. 2) that was significant when the double ␣CP1/2 knockdown was compared with mock-transfected (p ϭ 0.0043) or with lamin A/C-transfected (p ϭ 0.0068) cells. These alterations in the cell cycle were not observed in cells individually depleted for ␣CP1 or ␣CP2 (Fig. 2). There was no substantial impact on viability among the various treatments when assessed by dye exclusion (data not shown). Taken together, the data reveal that co-depletion of ␣CP1 and ␣CP2 reduced cellular proliferation and resulted in a G 1 arrest.
Phosphorylation of Serine 795 on the Retinoblastoma (RB) Protein Is Increased in the ␣CP1/2-co-depleted K562 Cells-The RB protein is a pivotal factor in the G 1 to S transition of the cell cycle, and phosphorylation of specific residues in RB has been implicated in this activity (35). The observation that cells co-depleted of ␣CP1 and ␣CP2 accumulate in G 1 led us to monitor for changes in the phosphorylation status of RB. Western blot analysis revealed that RB phosphorylation at Ser 795 was increased in the ␣CP1/2-co-depleted cells relative to controls ( Fig. 3A, top). There was also a slight but consistent increase of Ser 795 phosphorylation in cells treated with lamin A/C siRNA relative to the mock treatment. Analysis of cells individually depleted for ␣CP1 revealed a slight increase in Ser 795 phosphorylation, whereas the modification of this residue in the cells individually depleted of ␣CP2 remained unchanged (data not shown). The Western analysis of Ser 780 phosphorylation (Fig.  3A, middle) failed to reveal any changes in any of the conditions tested. The overall levels of RB protein were also found to be unaltered in any depleted cells (Fig. 3, A and B, bottom panels). Phosphorylation of RB at Ser 807/811 was marginally increased by the ␣CP1/2 co-depletion relative to the mock control, but a marginal increase was also observed in the cells treated with the lamin A/C siRNA (Fig. 3B). Taken together, these data reveal a selective increase in phosphorylation of RB at Ser 795 in cells co-depleted of ␣CP1 and ␣CP2. However, since previous reports (36) have shown that phosphorylation of RB at Ser 795 correlates with entry into S phase, the linkage of this change to the observed G 1 arrest appeared unlikely. For this reason, we decided to search for additional targets of ␣CP that might be causative in the observed G 1 arrest in the ␣CP1/2co-depleted cells.
Targeted Analysis of mRNAs Encoding Cell Cycle Control Proteins Reveals a Subset of mRNAs Whose Expression Is Altered by the ␣CP1/2 Co-depletion-To further explore the mechanism by which ␣CPs impact on cell cycle control(s), we screened a set of 84 human genes that play key roles in cell cycle regulation for alterations in mRNA levels subsequent to the ␣CP1/2 co-depletion (RT 2 Profiler PCR array). Eleven mRNAs that were found to be either up-or down-regulated in the ␣CP1/2-codepleted cells when compared with the mock and the Lamin A/C depleted cells are shown in Table 3. Included in this set were the mRNAs encoding p53 and p21 WAF . When compared with mock-or lamin A/C-depleted cells, the ␣CP1/2 co-depleted cells had levels of p53 mRNA and p21 WAF mRNA that were increased by 3.1-and 2.6-fold and 3-and 4.1-fold, respectively. It should be noted that the p21 WAF gene was not detected on the microarray platform used in our initial study (Table 1). This may be due to differences in the detection limits of qRT-PCR and microarray analysis. The increases in p53 and p21 WAF mRNA levels seen in the ␣CP1/2-co-depleted cells were of particular interest, since either could contribute to the G 1 arrest. p21 WAF Protein Is Up-regulated in the Cells Co-depleted of ␣CP1 and ␣CP2 via a p53-independent Mechanism-The observation that the levels of p53 and p21 WAF mRNAs were FIGURE 4. Co-depletion of ␣CP1 and ␣CP2 results in p53-independent induction of p21 WAF expression. A, K562 cells were mock-transfected (M) or transfected with siRNAs directed against ␣CP1, ␣CP2, ␣CP1 and ␣CP2 (␣CP1 ϩ 2), or lamin A/C (Lam). HCT116 cells that lack (Ϫ/Ϫ) or contain the wild type (WT) p53 gene were treated with irinotecan to induce p53 expression and were used as controls. Protein lysates were analyzed by Western blotting using antibodies directed against p53. The loading control is the ribosomal protein L7 (rpL7). B, K562 cells were transfected with siRNAs as above and analyzed by Western blotting using antibodies directed against p21 WAF . An antibody directed against RB was utilized as a loading control. both enhanced in cells co-depleted of ␣CP1 and ␣CP2 was followed by determining whether these changes were reflected at the protein level. Western blot analysis of extracts derived from cells individually transfected with ␣CP1 siRNA, ␣CP2 siRNA, or a combination of ␣CP1 and ␣CP2 siRNAs was compared with mock-transfected and lamin A/C siRNA-transfected controls (Fig. 4A). The human colon cancer cell line (HCT116) was used as a control for p53 detection. HCT116 cells containing the p53 gene (WT) and derivative HCT116 cells lacking the p53 gene (Ϫ/Ϫ) were treated with irinotecan to induce p53 expression. p53 was robustly and selectively induced by irinotecan in the wild type HCT116 cells. Parallel analysis of the K562 cells revealed a complete absence of p53 protein. This lack of p53 was consistent with previous reports (37) showing that the p53 gene is inactivated in K562 cells. The mutation consists of an insertion of a cytosine between codons 135 and 136. This insertion creates a frameshift, leading to a truncated protein of 147 amino acids. In contrast to the lack of p53 expression, p21 WAF was strongly induced in the cells co-depleted for ␣CP1 and ␣CP2 (Fig. 4B). Interestingly, the individual ␣CP1 and ␣CP2 depletions had no apparent effect on p21 WAF protein levels. In summary, the data confirm that p21 WAF protein is strongly induced by co-depletion of ␣CP1 and ␣CP2 and that this effect is p53-independent. The mRNA encoding p21 WAF Interacts with both ␣CP1 and ␣CP2 in Vivo-Since p21 WAF mRNA and protein were both found to be up-regulated by the ␣CP1/2 co-depletion and since ␣CPs are known to modulate gene expression via targeted binding to mRNA, we asked whether ␣CP1 and ␣CP2 interacted with p21 WAF mRNA in vivo. ␣CP1and ␣CP2-containing mRNPs were individually enriched from K562 cytosolic extracts by immunoprecipitation with isoform-specific antibodies. mRNAs isolated from both sets of mRNP immunoprecipitations were assessed for enrichment of p21 WAF mRNA by a semiquantitative RT-PCR analysis (Fig. 5). The analysis revealed that the p21 WAF mRNA was enriched by ϳ3and 3.4fold in the ␣CP1 and ␣CP2 mRNP isolates, respectively. In contrast, levels of ␥-globin mRNA levels were not significantly different in the ␣CP and control immunoprecipitates. These data lead us to conclude that both ␣CP1 and ␣CP2 bind to the p21 WAF mRNA in vivo.
Co-depletion of ␣CP1 and ␣CP2 Stabilizes Endogenous p21 WAF mRNA in K562 Cells-The induction of p21 WAF mRNA levels in the ␣CP1/2-co-depleted cells and the observation that both ␣CP proteins interacted with the p21 WAF mRNA in vivo suggested that control over p21 WAF mRNA levels might be mediated by an effect of ␣CP on p21 WAF mRNA stability. This model was tested. At 2 days post-siRNA transfection, the cells were treated with the transcriptional inhibitor actinomycin D, and RNA harvested at subsequent time points was quantified for p21 WAF mRNA by Northern analysis. This analysis revealed that the rates of p21 WAF mRNA decay in mock-or lamin A/C siRNA-transfected cells were similar, with a half-life of ϳ3 h (Fig. 6). In contrast, the ␣CP1/2 co-depletion resulted in prolongation of the half-life to ϳ13 h. This alteration in p21 WAF mRNA stability in the ␣CP1/2-co-depleted cells is consistent with the observed increase in levels of p21 WAF mRNA and protein. FIGURE 5. The mRNA encoding p21 WAF is associated with ␣CP1 and ␣CP2 in vivo. K562 cell extracts were immunoprecipitated (IP) with antibodies to ␣CP1 or ␣CP2 or a c-Myc control (Cont) antibody. Following immunoprecipitation, RNA was isolated from the RNP complexes and subjected to RT-PCR analysis to detect p21 WAF or ␥-globin mRNAs. The ␥-globin image for the ␣CP1 and control immunoprecipitate was from different areas of the same gel. ␣CP1 and ␣CP2 Bind to a 127-nucleotide Fragment of the p21 WAF 3Ј-UTR-Since we found that the p21 WAF mRNA interacted with ␣CP1 and ␣CP2 in vivo (Fig. 5) and since deple-tion of both proteins stabilized the p21 WAF mRNA (Fig. 6), we decided to test whether regions of the p21 WAF 3Ј-UTR interacted with ␣CP. Our rationale for this experiment was that several examples exist in the literature where ␣CP binding sites occur in the 3Ј-UTR of specific mRNAs and regulate mRNA stability (14 -16). We obtained a series of plasmids each containing a T7 promoter that drives the synthesis of different regions of the p21 WAF 3Ј-UTR (33). The map of these regions is shown in Fig. 7A. We synthesized thiolated RNA corresponding to these regions and used them in cross-linking assays. After UV cross-linking and RNase A digestion, the resulting mRNP complexes were immunoprecipitated using antibodies directed against ␣CP1, ␣CP2, or a c-Myc control. The results are shown in Fig. 7B (top). Actin antisense RNA and the ␣-globin 3Ј-UTR were used as negative and positive controls, respectively. We did not detect any cross-linked immunoprecipitated product using any of the antibodies when the actin antisense RNA was used in the assay. In contrast, when the ␣-globin 3Ј-UTR was used, we observed a cross-linked product when the immunoprecipitation was carried out using ␣CP2or ␣CP1-specific antibodies. Both of these products were of the appropriate size for ␣CP2 or ␣CP1. The only fragment of the p21 WAF 3Ј-UTR that interacted with both ␣CP1 and ␣CP2 was WAF 1-879, which corresponds to nucleotides 879 -1512 of the p21 WAF 3Ј-UTR. Next, we further mapped the ␣CP binding site on the 3Ј-UTR by generating smaller fragments of WAF 1-879 (WAF1-A to WAF1-E) to be used in the cross-linking assay ( Fig. 7A and Table 4). The cross-linking results are shown in Fig. 7B (bottom). WAF 1-879 was used as a positive control, and we observed the expected immunoprecipitation of ␣CP1 and ␣CP2. The only subfragment of WAF 1-879 that showed significant binding to ␣CP was WAF1-A. WAF1-D had very faint cross-linked products, but it was not reproducible. Therefore, we conclude that the major binding determinant of both ␣CP1 and ␣CP2 resides in the WAF1-A fragment. The sequence of WAF1-A is indicated in Fig. 7C. The triplication of C-rich regions (underlined) bears striking resemblance to the triple C-rich motifs previously identified as the ␣CP binding site in human ␣-globin mRNA (3,16).

DISCUSSION
Our current observations lead us to conclude that ␣CP1 and ␣CP2 play a significant role in the control of p21 WAF expression. This control appears to reflect a direct in vivo association of these proteins with the p21 WAF mRNA with consequent mRNA stabilization and increase in p21 WAF protein expression. Co-depletion of ␣CP1 and ␣CP2 results in a decrease in cell proliferation and a G 1 cell cycle arrest (Fig. 2). These functions appear to be mechanistically linked to the increase in p21 WAF protein levels (Figs. 4 -6). This direct, post-transcriptional control of p21 WAF expression by the ␣CP proteins is consistent with the observation that this control is independent of p53, the major transcriptional modulator of p21 expression. Of note, the alteration in cell cycle kinetics and increase in p21 WAF expression in the ␣CP1/2 co-depleted cells were not apparent in cells individually depleted of ␣CP1 or ␣CP2 (Figs. 2  and 4). These data suggest that the ␣CP1 and ␣CP2 isoforms have overlapping and/or redundant functions that are required for control of p21 WAF expression and normal progression through the cell cycle.
CCNH Expression in the ␣CP1/2-co-depleted Cells-The mechanism of the G 1 arrest in the cells co-depleted for ␣CP1 and ␣CP2 was investigated by defining alterations in the expression of mRNAs that encode proteins involved in cell cycle control. We found that the CCNH mRNA was downregulated by the ␣CP1/2 co-depletion (Table 1) with a 50% decrease in protein expression (Fig. 1B). The results of the individual ␣CP1 or ␣CP2 knockdowns suggested that these isoforms can individually and additively regulate CCNH expression. CCNH is a regulatory subunit for a Cdk (cyclin-dependent kinase)-activating kinase involved in multiple cell cycle transitions (38). Selective inhibition of Cdk7 (a Cdk-activating kinase subunit) delays entry into S phase (39). Therefore, it is possible that a decrease in CCNH expression in cells depleted of ␣CP1 and ␣CP2 could disrupt Cdk-activating kinase function(s) and contribute to the observed G 1 arrest. However, the observation that individual ␣CP1 and ␣CP2 depletions are sufficient for repression of CCNH expression and yet fail to trigger the G 1 arrest leads us to conclude that the additive effect of the combined knockdowns on CCNH may be contributory to but are not the defining determinants of the G 1 arrest seen in the combined ␣CP1/2 depletion.
RB Phosphorylation in the ␣CP1 and ␣CP2 Co-depletion-Phosphorylation of RB plays a major role in RB-mediated cell cycle controls. We observed an increase in the phosphorylation of RB at Ser 795 in cells co-depleted for ␣CP1 and ␣CP2 (Fig. 3). However, it seems unlikely that this alteration in RB phosphorylation is the cause of the G 1 arrest subsequent to ␣CP1/2 co-depletion. Cells treated with the control lamin A/C siRNA FIGURE 7. ␣CP1 and ␣CP2 bind to a 127-nucleotide fragment of the p21 WAF 3-UTR. A, fragments of the p21 WAF 3Ј-UTR used in the initial cross-linking assay (33) and sequences corresponding to subfragments of WAF 1-879 used in higher resolution mapping. B, UV cross-linking assay of fragments of the p21 WAF 3Ј-UTR. Cytoplasmic extracts from K562 cells were incubated with thiolated, 32 ACTAGGGTGCCCTTCTTCTTGTGTGTCCC (reverse) a Each forward primer has a T7 promoter linked to its 5Ј-end (sequence not listed in Table 4). Each primer sequence is shown 5Ј-3Ј.

G 1 Arrest in ␣CP-depleted Cells
showed a reproducible, albeit moderate, increase in Ser 795 phosphorylation without a corresponding alteration in the cell cycle. In addition, previous reports in osteogenic sarcoma (SAOS-2) cells show that phosphorylation of RB at Ser 795 correlates with entry into S phase rather than G 1 arrest (36).
Although it is possible that the impact of the RB phosphorylation in K562 cells may differ from that in other cells, this linkage remains untested.
Relationship of ␣CPs to p21 WAF Expression and Cell Cycle Controls-Our data are most consistent with a pathway in which co-depletion of ␣CP1 and ␣CP2 lead to an induction of p21 WAF protein expression (Fig. 4B) via stabilization of p21 WAF mRNA (Fig. 6). The increase in p21 WAF protein correlates with the G 1 arrest; both occur only in the ␣CP1/2-co-depleted cells and not in cells where the mRNAs encoding these two proteins are individually targeted. The impact of the increased p21 WAF expression on cell growth is fully concordant with the known ability of p21 WAF to mediate a G 1 block of the cell cycle (40). Therefore, our data suggest that activation of p21 WAF protein expression, triggered by depletion of ␣CP1 and ␣CP2, mediates G 1 arrest in the K562 cells.
p21 WAF is a direct mediator of cell cycle arrest at the G 1 phase (41). p21 WAF can inhibit specific Cdks, resulting in inhibition of RB phosphorylation (40). Unphosphorylated RB protein binds to several proteins involved in the regulation of the G 1 to S transition, including the E2F family of transcription factors (42). The RB-E2F complex acts as a transcriptional repressor whose targets include several genes required for S phase, contributing to the mechanism of G 1 arrest (43). p21 WAF also appears to be required for maintaining the G 2 checkpoint in human cells (44). The pathways by which p21 WAF levels in the cell are controlled and modulated appear to be complex and remain to be fully defined.
Since p53 has been shown to transcriptionally up-regulate p21 WAF (45,46), an important parameter of the p21 WAF -induced G 1 arrest in the K562 cells is that this effect occurs in the absence of p53. The p53 gene in K562 cells contains a single base insertion that leads to a translational frameshift and a truncated protein (37). Sequencing of the p53 locus in K562 cells reveals only the mutant sequence, indicating that the wild type allele has been either lost or converted to the mutant allele (37). Consistent with this mutation, the p53 mRNA could be detected (Table 3), whereas the protein was not detected by Western blotting (Fig. 4A).
Post-transcriptional Regulation of p21 WAF mRNA Expression-Although p21 WAF expression is under p53-mediated transcriptional control, an extensive body of literature documents that p21 WAF expression is also subject to post-transcriptional modulation. For example, the RNA-binding proteins hnRNP K (47) or Msi-1 (48) can block translation of the p21 WAF mRNA, and p21 WAF mRNA stability can be altered by a number of mRNA-binding proteins. The half-life of the p21 WAF mRNA can be increased by the binding of HuR in response to UV light (49) or prostaglandin A2 treatment (50) or by the binding of RNPC1a (51). In the last situation, the stabilization of p21 WAF mRNA is accompanied by G 1 arrest (51). Treatment of cells with hydroxyurea has been shown to stabilize p21 WAF mRNA (52), although the mechanism remains undefined. The 3Ј-UTR of p21 WAF mRNA has also been shown to be bound by a number of RNA-binding proteins, including ␣CP1, although the functional impact of ␣CP1 binding in that study was not explored (33). In that case, recombinant ␣CP1 was used, and binding was detected using the WAF1-1/6 fragment (referred to as WAF1-571 in Fig. 7A). In contrast, our study indicates that ␣CP1 and ␣CP2 both bind to WAF 1-879, and we did not detect binding to WAF1-1/6 (Fig. 7D). It is possible that differential RNA binding can occur with recombinant ␣CP versus cellular extracts containing ␣CP. Taken together, these reports indicate that expression of p21 WAF is subject to multiple layers of post-transcriptional control.
What is the mechanism of the increased expression of p21 WAF mRNA in the current study? Since previous work has linked ␣CP to the regulation of a number of mRNA targets, we monitored the impact of the siRNA treatments on p21 WAF mRNA stability (Fig. 6). These studies revealed that the p21 WAF mRNA half-life was increased in cells co-depleted of ␣CP1 and ␣CP2. Interestingly, ␣CP1 and ␣CP2 each bind to the p21 WAF mRNA in vivo (Fig. 5). We mapped the ␣CP1 and ␣CP2 binding site on the 3Ј-UTR of p21 WAF to a 127-nucleotide sequence. This sequence contains three CU-rich patches, reminiscent of the ␣CP binding site on the ␣-globin 3Ј-UTR (3,16). The finding that co-depletion of ␣CP1 and ␣CP2 stabilizes the p21 WAF mRNA suggests that under normal conditions, the p21 WAF mRNA is destabilized by these two proteins. This finding is of particular interest, since ␣CP binding has been previously linked to mRNA stabilization rather than destabilization. Thus, the present study points to a novel activity of these hnRNP K homology domain proteins. However, our data suggest that this control via mRNA destabilization may not be unique; Table 1 lists 14 mRNAs whose steady state levels are increased in cells co-depleted of ␣CP1 and ␣CP2. The questions of whether these mRNAs are coordinately stabilized in the co-depleted cells in some manner, whether they are all direct binding targets of ␣CP1 and/or ␣CP2, and whether the alteration in any of these additional mRNAs contributes to the cell cycle arrest in the co-depleted cells can now be addressed.