p53-Dependent p21 mRNA Elongation Is Impaired when DNA Replication Is Stalled

We have previously reported that when DNA replication is blockedin some human cell lines, p53 is impaired in its ability toinduce a subset of its key target genes, including p21^WAF1/CIP1.Here, we investigated the reason for this impairment by comparingthe effects of two agents, hydroxyurea (HU), which arrests cellsin early S phase and impairs induction of p21, and daunorubicin,which causes a G₂ block and leads to robust activation of p21by p53. HU treatment was shown to inhibit p21 mRNA transcriptionrather than alter its mRNA stability. Nevertheless, chromatinimmunoprecipitation assays revealed that HU impacts neitherp53 binding nor acetylation of histones H3 and H4 within thep21 promoter. Furthermore, recruitment of the TFIID/TATA-bindingprotein complex and the large subunit of RNA polymerase II (RNAPol II) are equivalent after HU and daunorubicin treatments.Relative to daunorubicin treatment, however, transcription elongationof the p21 gene is significantly impaired in cells treated withHU, as evidenced by reduced occupancy of RNA Pol II at regionsdownstream of the start site. Likewise, in the p21 downstreamregion after administration of HU, there is less of a specificallyphosphorylated form of RNA Pol II (Pol II-C-terminal domainserine 2P) which occurs only when the polymerase is elongatingRNA. We propose that while the DNA replication checkpoint isunlikely to regulate the assembly of a p21 promoter initiationcomplex, it signals to one or more factors involved in the processof transcriptional elongation.

INTRODUCTION

Top

Introduction

The p53 tumor suppressor protein is a sequence-specific transcriptionfactor that is found mutated in over 50% of human cancers (47,70). p53 serves as a molecular guardian that responds to variousforms of stress by regulating the expression of numerous genesinvolved in cell cycle progression, senescence, apoptosis, andmitosis (57, 58, 71). The ability of p53 to serve as a transcriptionalregulator is essential to its roles in tumor suppression. Thevast majority of tumor-derived mutations of p53 occur withinthe core DNA-binding domain of the protein, and such mutationsusually prevent p53 from binding to promoters and activatingtarget genes (33). Loss of p53 activity can result in uncontrolledcell proliferation in the presence of DNA damage, accumulationof mutations, and neoplasia (63). To date, hundreds of targetgenes have been identified for p53 (48, 71, 73), with the mostwell characterized of these promoting either cell cycle arrestor apoptosis. These genes include a G₁ cyclin-dependent kinaseinhibitor, p21^WAF1/CIP1 (hereafter referred to as p21) (21,36); an inhibitor of the G₂ cyclin B/CDK1 complex, GADD45 (39,77); a p53 E3-ubiquitin ligase, MDM2 (4, 75), and proapoptoticgenes, such as PUMA, BAX, PIG3, p53AIP1, NOXA, and APAF1, toname a few (37, 56, 72). There is also a significant body ofevidence implicating cytoplasmic p53 in a transcription-independentmitochondrial apoptosis pathway (12, 16, 52), although recentwork indicates interplay between the nuclear transcriptionalactivity of p53 and its functions in the cytoplasm (11).

It is well established that in response to DNA damage causedby ionizing radiation (IR), UV light, and a plethora of otheragents (including topoisomerase inhibitors, such as daunorubicin[dauno]), the half-life (t_1/2) of p53 increases and cells undergoarrest in G₁ and/or G₂ (25). Following DNA damage, p53 becomesphosphorylated, acetylated, and subjected to other modificationswhich affect its stability, activity, and subcellular localization(2, 9, 59). Although most reports to date have studied the rolegenotoxic stress plays in activating p53, it has become evidentthat other forms of stress may also lead to the activation ofthe p53 response (57). In this study, we examined DNA replicationblock and the effect it has on p53 function.

Hydroxyurea (HU) is a chemotherapeutic drug that has been usedin the treatment of leukemia (26). HU inhibits the functionof ribonucleotide reductase, an enzyme responsible for the conversionof ribonucleotides (rNTPs) to deoxyribonucleotides (dNTPs) (65,69). Upon treatment with HU, the cell's limited supply of dNTPsrapidly decreases, leading to stalled replication forks anda reversible, p53-independent arrest in early S phase. We previouslyshowed that in response to the DNA replication block inducedby HU, p53 protein is stabilized but is impaired in its abilityto activate several but not all of its transcriptional targetgenes (32). Most relevantly, p21 protein and RNA fail to accumulate,while another p53 target, PIG3, exhibits no such accumulationdefect. Since the expression of only a subset of p53 targetgenes is affected by DNA replication block, this impairmentdoes not result from a global defect in transcription or translationafter drug treatment, but instead points to a bona fide gene-selectivemechanism. In a subsequent publication, we showed that reductionof p21 protein levels during S-phase arrest is critical forthe cell to efficiently resume the cell cycle once the replicationblock is resolved (30). We further found that, after replicationblock, p21 protein is subjected to enhanced proteasome-dependentdegradation. Thus, at least in some cell lines, the failureto accumulate p21 results from two different mechanisms, transcriptionalimpairment and increased proteolysis. Here, we investigatedthe reasons for the impairment in p21 mRNA accumulation followingDNA replication block.

MATERIALS AND METHODS

Top

Materials and Methods

Cell culture, FACS analysis, and drugs. Parental RKO and HCT116 cells (colorectal carcinoma cells expressingwild-type p53) and H1299 cells (p53-null lung epithelial carcinomacells) were grown in Dulbecco modified Eagle medium containing10% fetal bovine serum at 37°C. RKO-E6 (clone 10.2) cells,which stably express the human papillomavirus 16 (HPV-16) E6protein, or RKO-neo cells, which have been stably transfectedwith an empty vector, were grown in McCoy's medium containing10% fetal bovine serum supplemented with 500 µg/ml G418and maintained at 37°C. These cells lines were gifts fromK. Cho (29, 40). Drugs used in this study are as follows: hydroxyurea,1.7 mM (Sigma); daunorubicin, 0.22 µM (Sigma); actinomycinD, 0.4 µM (Calbiochem). Florescence-activated cell sorter(FACS) analysis was performed as previously described (32).

Protein extraction and immunoblotting. Cells were lysed using HEPES lysis buffer (10 mM HEPES, pH 7.9;1 mM EDTA; 10% glycerol; 400 mM NaCl; 1% NP-40; 0.5 mM phenylmethylsulfonylfluoride, and protease inhibitors [1 µM benzamidine, 3µg/ml leupeptin, 0.1 µg/ml bacitracin, 1 µg/mlmacroglobulin]). Lysates were centrifuged to remove cellulardebris, and whole-cell extract (40 µg) was resolved by12% sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresisand transferred to nitrocellulose. Membranes were probed usingthe following antibodies: p53, 1801 (supernatant solution fromhybridoma culture); p21, Waf-1 (Ab-1) (OP-64; Oncogene Research);actin, A-2066 (Sigma). Anti-mouse or anti-rabbit secondary antibodiesconjugated to horseradish peroxidase (Sigma) were incubatedwith membranes and detected by enhanced chemiluminescence (Amersham).

RNA extraction and RT-PCR analysis. Cells were seeded and treated according to the conditions describedin the figure legends for each experiment. RNA was harvestedusing a QIAGEN RNeasy Mini kit according to the manufacturer'sspecifications. First-strand cDNA synthesis and reverse transcriptase(RT)-PCRs were performed as previously described (51). InitialRT-PCRs were performed in parallel with a standard curve oftemplate DNA, and different cycling conditions were tested toconfirm that amplifications were within the linear range. A"No RT" reaction, in which RNA was subjected to the conditionsof cDNA synthesis without reverse transcriptase, was includedas a negative control in RT-PCR experiments to confirm the purityof RNA samples. In most cases, RT-PCRs were run using 2.5% agarosegels, imaged with a Gel Logic 100 imaging system (Kodak), andquantitated with Kodak 1-D image analysis software. For PCRsconducted in the presence of radioactive dCTP, samples wereresolved on 8% polyacrylamide gels, dried, and autoradiographed.In these instances, band quantitation and analysis was performedusing a phosphorimager (Molecular Dynamics). Primer constructionis as follows: RT-p21 forward (Fwd), 5'-GTTCCTTGTGGAGCCGGAGC-3';RT-p21 reverse (Rev), 5'-GGTACAAGACAGTGACAGGTC-3'; RT-PIG3 Fwd,5'-GCAGCTGCTGGATTCAATTAC-3'; RT-PIG3 Rev, 5'-GCCTATGTT CTTGTTGGCCTC-3';RT-MDM2 Fwd, 5'-GTGAATCTACAGGGACGCCAT-3'; RT-MDM2 Rev, 5'-CTGATCCAACCAATCACCTGAA-3'; RT-PUMA Fwd, 5'-TGTGAATCCTGTGCTCTGCC-3';RT-PUMA Rev, 5'-TTCCGGTATCTACAGCAGCG-3'; RT-GAPDH Fwd, 5'-CAACTACATGGTTTACATGTTC-3';RT-GAPDH Rev, 5'-GCCAGTGGACTCCACGAC-3'.

RNA stability experiments. First, RKO cells were treated with either hydroxyurea (1.7 mM,24 h) or daunorubicin (0.22 µM, 12 h) to stabilize p53protein to an equivalent degree in the two experimental sets.Next, 0.4 µM actinomycin D was added to the culture mediaand incubated for 0, 1, 2, 4, or 6 h. Cells were washed twicewith phosphate-buffered saline, harvested by scraping, and dividedinto two samples. Cells in each aliquot were pelleted by centrifugationand subjected to either RNA extraction followed by RT-PCR orprotein extraction followed by immunoblotting as described above.

Slot blots and nuclear run-on assays. Slot blots were generated as follows. Plasmid DNA containingcDNAs of interest was linearized and purified by phenol-chloroform/isoamylalcohol extraction followed by ethanol precipitation. Linearizedplasmids were boiled in the presence of 6x SSC (1x SSC is 0.15M NaCl plus 0.015 M sodium citrate), and single-stranded DNAwas slot blotted on Nytran-N membrane (Schleicher and SchuellBioscience, Inc.) that was previously wetted in 2x SSC. Blotswere UV cross-linked (autolink setting, UV Crosslinker; FisherBiotech) and incubated overnight in 1 ml of prehybridizationbuffer (2x TESS [20 mM Tris {pH 7.4}, 10 mM EDTA, 0.3 M NaCl,0.2% SDS], 4x Denhardt's reagent, 100 µg/ml tRNA) at 65°C(with rocking).

To isolate nuclei, approximately 5 x 10⁶ cells were washed incold phosphate-buffered saline, scraped, and pelleted by centrifugation(300 x g for 5 min). The cells were resuspended in 2 ml of NP-40lysis buffer (10 mM Tris, pH 7.4; 10 mM NaCl; 3 mM MgCl₂; 0.5%NP-40) and centrifuged (1,175 x g for 5 min) to pellet nuclei,and the pellet was washed with 4 ml NP-40 lysis buffer. Nucleiwere centrifuged once again, and the pellet was resuspendedin 100 µl nuclear storage buffer (50 mM Tris, pH 8.3;40% [vol/vol] glycerol; 5 mM MgCl₂; 0.1 mM EDTA) and storedat –70°C.

Run-on reactions were performed as follows. Nuclei were thawedon ice, mixed with 100 µl reaction buffer (10 mM Tris,pH 7.5; 10 mM MgCl₂; 300 mM KCl; 0.5 mM each of ATP, CTP, andGTP; 15 µl [ ${alpha}$ -³²P]UTP [3,000 mCi/ml]), and incubated at30°C for 30 min. The reactions were stopped by the additionof 0.5 ml TRIzol reagent (Invitrogen), and labeled RNA was extractedaccording to the manufacturer's specifications. RNA was resuspendedin RNase-free water and added to the hybridization solution(2x TESS; 1x Denhardt's reagent; 100 µg/ml tRNA). Approximately6 x 10⁶ cpm of probe was used for each sample. Blots were hybridizedfor 48 h at 65°C (with rocking) and then washed twice in2x SSC/0.2% SDS, washed twice in 0.2x SSC/0.2% SDS (20 min perwash at 65°C), dried briefly, and exposed. Bands were quantitatedand analyzed by phosphorimaging.

Chromatin immunoprecipitation. Chromatin immunoprecipitation (ChIP) assays were performed aspreviously described (51). The antibodies used for ChIP werep53, PAb1801/DO-1 cocktail (supernatant solution from hybridomacultures); acetyl-histone H3, 06-599 (Upstate Biotechnology);acetyl-histone H4, 06-598 (Upstate Biotechnology); Pol II (N-20),sc-899 (Santa Cruz); Pol II-CTD serine 2P (H5), MMS-129R (Covance);and TFIID/TBP (N-12), sc-204 (Santa Cruz). In cases where immunoprecipitationswere performed using antibodies directed against acetylatedhistones, 5 mM sodium butyrate and 5 µM trichostatin Awere also added to the lysis buffer to protect against histonedeacetylase activity. In cases where immunoprecipitations wereperformed using antibodies against the phosphorylated form ofRNA Pol II (anti-Pol II-C-terminal domain [CTD] serine 2P),a mixture of phosphatase inhibitors (Calbiochem) was added tothe lysis buffer to protect against phosphatase activity, andthe beads/antibody used for immunoprecipitation were preparedas previously described (43). PCRs were carried out for 30 cyclesunless otherwise indicated. Linear amplification of PCRs wasconfirmed, and amplicons were resolved by the same methods asthose described for RT-PCR. ChIP primer sequences are as follows:p21 promoter (5' p53 response element) (28 cycles) Fwd, 5'-CTGGACTGGGCACTCTTGTC-3',and p21 promoter Rev, 5'-CTCCTACCATCCCCTTCCTC-3'; p21 TATA promoterFwd, 5'-TATTGTGGGGCTTTTCTG-3', and p21 TATA promoter Rev, 5'-CTGTTAGAATGAGCCCCCTTT-3';p21 NEG promoter Fwd, 5'-GGAGTCCTGTTTGCTTCTGG-3', and p21 NEGpromoter Rev, 5'-CTTTGGCCACACTGAGGAAT-3', and p21 Intron 1 Fwd,5'-GTGATGGGCCTCTCTGGTTA-3'; p21 Intron 1 Rev, 5'-TGTTTGGGGGCTGTTCTAAG-3';p21 Exon 2 (28 cycles) Fwd, 5'-GAGCGATGGAACTTCGACTT-3', andp21 Exon 2 Rev, 5'-CAGGTCCACATGGTCTTCCT-3'; p21 Exon 3 (28 cycles)Fwd, 5'-CCCTTTCCTGGACACTCAGA-3', and p21 Exon 3 Rev, 5'-CCCTAGGCTGTGCTCACTTC-3';MDM2 promoter Fwd, 5'-TCGGGTCACTAGTGTGAACG-3', and MDM2 promoterRev, 5'-CACTGAACACAGCTGGGAAA-3'; GADD45 promoter Fwd, 5'-GGTTGCCTGATTGTGGATCT-3',and GADD45 promoter Rev, 5'-GAGTAGCTGGGCTGACTGCT-3'; PIG3 promoterFwd, 5'-CACTCCCAACGGCTCCTTT-3', and PIG3 promoter Rev, 5'-GCCCATCTTGAGCATGGGT-3';PIG3 Exon 2 Fwd, 5'-CCAGCAACATTTTGGGACTT-3', and PIG3 Exon 2Rev, 5'-TCTGGGATAGGCATGAGGAG; and PIG3 Exon 3 Fwd, 5'-ATGCAGGACTGAGTGGTGTG-3',and PIG3 Exon 3 Rev, 5'-GCGTTGCTTCAGAGAAATCC-3'; GAPDH promoterFwd, 5'-AAAAGCGGGGAGAAAGTAGG-3', and GAPDH promoter Rev, 5'-CTAGCCTCCCGGGTTTCTCT-3'.

RESULTS

Top

Results

p53 is transcriptionally impaired in response to DNA replication block. For the present study, we compared the effects of HU, whichcauses cells to accumulate at the G₁/S-phase border with thosecaused by dauno, a topoisomerase II inhibitor (45) that arrestsRKO colorectal cancer cells in G₂ (see Fig. 2D and E and reference32). The first objective was to find circumstances where HUor dauno treatment could induce equivalent amounts of p53 proteinand determine the cell's response to these conditions. RKO cellswere treated with HU or dauno for various times, and then p53and p21 protein levels were assessed by immunoblotting (Fig.1A). We also examined mRNA levels of both p21 and another p53target, PIG3, under these conditions (Fig. 1B). Time pointswere identified where equivalent levels of p53 protein weredetected (Fig. 1A, top panel, compare lanes 4 and 6 or lanes5 and 7). At these times, far less p21 protein accumulated inthe samples treated with HU than in the related samples treatedwith dauno (Fig. 1A, middle panel). The inefficient accumulationof p21 after HU treatment was linked to impaired transcriptionof the p21 gene as measured by RT-PCR, which was observed veryearly and was sustained throughout the time course (Fig. 1B,left panel and graph, compare lanes 2 to 5 with lanes 6 to 9).In contrast, even very early (6 h) after dauno treatment, therewas strong activation of the p21 gene, indicating that submaximallevels of p53 were capable of fully transactivating the p21promoter. On the other hand, under conditions where p53 levelswere equivalent, the PIG3 promoter was transactivated to similarextents with the two treatments (Fig. 1A; Fig. 1B, middle panel).In fact, in the HU 12-h sample, PIG3 mRNA accumulated to anextent similar to that seen 6 h after dauno, despite the factthat the p53 protein level in the HU sample was approximatelyhalf of that seen for dauno (Fig. 1A, compare lanes 3 and 6;Fig. 1B, compare lanes 12 and 15). These results demonstratethe selective impairment of p53 transcriptional readout followingHU treatment in RKO cells.

View larger version (31K):
[in this window]
[in a new window]

FIG. 2. Stalled DNA replication represses the p53 response to daunorubicin. (A) RKO cells were untreated (Ctrl) or were treated with HU and dauno (as described for Fig. 1A) for 24 and 12 h, respectively, as indicated. For the "HU/D" sample, cells were pretreated with HU for 12 h and then daunorubicin was added to the culture medium, which was incubated in combination with HU for an additional 12 h. H1299 cells were untreated and included as a negative control. Cells were collected, lysed, and subjected to immunoblot analysis using the indicated antibodies. (B) H1299 and RKO cells were treated (as described for panel A), and RNA was extracted and subjected to RT-PCR analysis. [ ${alpha}$ -³²P]dCTP was included in the PCR mixture for p21, PIG3, and GAPDH (top) to obtain radioactive amplicons. For PUMA RT-PCR experiments (and the accompanying GAPDH [bottom] experiments), PCRs were conducted in the absence of [ ${alpha}$ -³²P]dCTP, and amplicons were resolved by agarose gel electrophoresis. Graphs show the averages and standard deviations (error bars) for two independent RT-PCR experiments. (C to E) FACS analysis was performed using RKO cells that were subjected to different combinations of drug treatments as shown, and data were analyzed using the ModFit program. (C) Cells were treated with HU (24 h), with gamma IR (10 Gy) for 12 h ( ${gamma}$ ), or a combination of both treatments (HU for 12 h and then gamma IR for an additional 12 h [HU/ ${gamma}$ ]). (D) RKO cells were treated as described for panel A. (E) RKO cells were treated with aphidicolin (APH, 5 µg/ml) for 24 h, dauno for 12 h, or a combination of both treatments (APH for 12 h then dauno for an additional 12 h [APH/D]). Ctrl, control; D, dauno.

View larger version (18K):
[in this window]
[in a new window]

FIG. 1. HU treatment selectively impairs activation of p53 target genes. (A) RKO cells were untreated (Ct) or treated with HU (1.7 mM), or daunorubicin (0.22 µM) over a time course (6 to 24 h, as indicated), lysed, and subjected to immunoblot analysis with anti-p53, -p21, and -actin antibodies. Red- and blue-highlighted time points indicate respective conditions in which equivalent levels of p53 protein have been induced. (B) RKO cells were treated as for panel A, and RNA was extracted and subjected to RT-PCR using primer sets for the indicated genes. Amplicons were resolved by agarose gel electrophoresis and stained using ethidium bromide. Graphs show the averages from three independent experiments, and error bars included in the graphs represent one standard deviation. Ctrl, control.

RKO cells were also treated with a combination of HU and daunoto assess the impact both drug treatments together would haveon p53 accumulation and its downstream targets (Fig. 2). Underconditions where p53 levels were similar and dauno treatmentalone provoked a robust accumulation of p21 protein and mRNA,the cotreatment of cells with HU and dauno led to extremelylow p21 protein levels that were similar to those with HU alone(Fig. 2A). Similarly, with a combination of HU and dauno treatment,p21 mRNA was induced less well than with dauno alone, althoughit accumulated slightly better than with HU alone (Fig. 2B,compare lanes 3 and 4). As before, the p53 target gene PIG3yielded similar levels of accumulated mRNA in response to bothHU and dauno treatments, with somewhat more PIG3 mRNA beingdetected upon cotreatment with these drugs. Interestingly, thisresponse might be common among proapoptotic targets of p53,as a RT-PCR analysis of the PUMA gene shows results similarto those for PIG3 (Fig. 2B, compare lanes 12 to 14 to lane 11).

In line with these results, the cell cycle distribution causedby HU treatment predominated when cells were cotreated witheither gamma irradiation or daunorubicin (Fig. 2C and D). Whencells were blocked in S phase by aphidicolin (an inhibitor ofDNA polymerases), the results were essentially similar, indicatingthat the effects of HU are not due to a specific effect of thisdrug on cells but rather reflect its function in blocking Sphase (Fig. 2E; see Fig. S1A and B in the supplemental material).Note as well that the HU-related impairment of p53 transcriptionalactivity was also seen in another cell line, HCT116 (see Fig.S1C and D in the supplemental material). Thus, when DNA replicationis stalled, cells cannot respond to a typical p53 stress signal.

Although the transcriptional activity of p53 was drasticallyimpaired following DNA replication block, it was not completelyeliminated (Fig. 1B, compare levels of p21 mRNA in control andHU-treated samples). To determine if this small amount of p21accumulation was dependent on p53, we compared the effects ofHU in a cell line (RKO-E6) engineered to express a stable formof the HPV-16 E6 protein with a control cell line (RKO-neo)(29, 40). Since the HPV-16 E6 protein targets p53 for degradationby the proteasome (60), RKO-E6 cells have extremely low levelsof p53 protein and are incapable of stabilizing p53 followingstress. In fact, the results showed that after HU treatmentof RKO cells, the small amount of p21 mRNA that accumulatesis independent of p53 (see Fig. S2 in the supplemental material).Therefore, p53-mediated induction of p21 is specifically targetedby stalled DNA replication.

The rate of transcription from the p21 promoter is decreased in HU-treated cells. Although HU treatment impaired the ability of p53 to induceaccumulation of p21 mRNA, it was unclear whether this resultedfrom inhibition of p21 mRNA transcription or decreased mRNAstability. To address this issue, we used actinomycin D to haltnew transcription and monitor the degradation rates of p21 mRNAover time (50, 66). First, RKO cells were treated with HU for24 h or dauno for 12 h to allow accumulation of equivalent levelsof p53 protein and were then treated with actinomycin D. Cellswere harvested at increasing time points, and samples were preparedfor immunoblotting (Fig. 3A) or RT-PCR (Fig. 3B). While thehalf-life of p21 mRNA was slightly higher in drug-treated cellsthan in untreated cells (control t_1/2, 3 h; HU and dauno t_1/2,4 h), there were no significant differences between HU and dauno-treatedcells in the half-lives of p21 mRNA (Fig. 3C). Moreover, althoughPIG3 mRNA was extraordinarily stable, making it difficult toreliably estimate its half-life under these conditions (t_1/2was >6 h in all cases), there was no change in PIG3 mRNAstability under each of these conditions. Likewise, anotherp53 target gene, MDM2, whose induction is impaired after HUtreatment (32), exhibited no change in mRNA stability with thesetreatments (data not shown).

View larger version (20K):
[in this window]
[in a new window]

FIG. 3. mRNA stability of p53 target genes is not altered after HU treatment. (A and B) RKO cells were either untreated (Ctrl) or treated with HU for 24 h or daunorubicin for 12 h, followed by treatment with 0.4 µM actinomycin D (Act. D) for the time periods indicated. Cells were harvested, divided into two aliquots, lysed, and subjected to analysis by immunoblotting using the indicated antibodies (A) or RNA extraction and RT-PCR using primer sets for the indicated genes (B). (C) RT-PCR data from experiments shown in panel B was quantitated using Kodak 1-D image analysis software. Each point was plotted as a ratio of the amplified cDNA that exists at time zero. Half-life measurements were extrapolated from the graphs for each treatment used. Graphs show the averages for three independent experiments, with error bars representing one standard deviation.

Steady-state RNA levels are the sum of RNA synthesis and RNAdegradation. Since differences in mRNA degradation had beenruled out as the explanation for reduced p21 mRNA accumulationafter HU treatment, we employed a nuclear run-on assay whichcan directly measure the rate of transcription of a particulargene in its native cellular context (20, 53). Nuclear run-onexperiments performed in the presence of HU showed a clear decreasein the rate of transcription from the p21 gene compared to daunorubicin(Fig. 4). Conversely, PIG3 mRNA synthesis showed no such decreasefollowing HU treatment. Therefore, impaired p21 mRNA accumulationafter HU treatment is the result of a defect in its synthesisrather than its posttranscriptional stability.

View larger version (24K):
[in this window]
[in a new window]

FIG. 4. HU treatment results in reduced transcription from the p21 promoter. (A) Nuclear run-on assays were performed as described in Materials and Methods. Linearized plasmids containing cDNA fragments from the p21, PIG3, and GAPDH genes were immobilized on nylon membranes and hybridized with ³²P-labeled RNA probes generated from run-on transcription in RKO cells (control [Ctrl] or HU or dauno treated, as for Fig. 2A). Approximately 6 x 10⁶ cpm of labeled total RNA was added to each hybridization reaction. Horizontal lines within the nuclear run-on images indicate that intervening slots have been spliced out. (B) Graphical representation of the averages and standard deviations (error bars) for three independent nuclear run-on experiments. Blots were analyzed by phosphorimaging, and background corrections were made for each blot by first subtracting the amount of hybridization to a negative control (linearized pBSK empty vector) and then normalizing data to the amount of GAPDH synthesis.

p53 binding to the p21 promoter is not affected by blocked DNA replication. An explanation for the decreased amount of p21 transcriptionobserved after DNA replication block could be an inability ofthe p53 protein to bind to the p21 promoter. We used the ChIPassay to examine this possibility. Various regions within thep21 locus, as well as a number of other target genes (PIG3,GADD45, and MDM2), were examined to detect bound p53 after treatmentswith HU, dauno, or a combination of both drugs (Fig. 5). Consistentwith increased p53 levels caused by these treatments (Fig. 2A),there was in each case more p53 bound to these promoters thanin untreated cells. With some genes, however (e.g., MDM2 andPIG3), the increase in binding was not proportional to the increasein detectable p53 protein in treated versus untreated cells.The analyses revealed no apparent differences in the abilitiesof p53 to bind the 5' response element within the p21 promoterfollowing HU versus dauno treatment. Furthermore, other targetgenes tested, including the PIG3 gene, showed no differencesin the abilities of p53 to bind response elements within theirpromoters after these treatments (Fig. 5B). Similar resultswere obtained for HCT116 cells treated with HU and dauno, aswell as for RKO cells treated with aphidicolin and dauno (datanot shown). Additionally, p53 association with the proximal(3') p53 response element within the p21 promoter was examinedand found to give results similar to those seen with the 5'site in RKO cells after these treatments (data not shown). Severalcontrols were performed to support the validity of these results.First, in different p53 ChIP experiments, both polyclonal andmonoclonal antibodies were used and, in all cases, greater than90% immunodepletion of p53 was observed, as determined by immunoblottinganalysis of the postimmunoprecipitate supernatants (data notshown). Second, as a control for immunoprecipitation specificity,ChIP assays were performed using the p53-null H1299 cell line(Fig. 5B, lane 6) as well as the RKO-E6 cell line (data notshown), and virtually no ChIP signal was detected. Finally,p53 binding to a far-upstream region of the p21 promoter (–5kb) was undetectable (Fig. 5B, p21 NEG panel). Taken together,these data indicate that the impairment in p53 transcriptionalactivity following DNA replication block does not result froma defect in its ability to occupy response elements within targetpromoters.

View larger version (23K):
[in this window]
[in a new window]

FIG. 5. p53 associates with p21 and other target promoters in HU-treated cells. (A) Schematic representation of the p21 locus. The locations of all PCR primers used in ChIP assays are labeled and indicated by dark blue boxes. Red areas indicate the approximate locations of exons within p21. The transcription start site (+1 bp) is indicated by a black, right-angle arrow. (B) p53 ChIP was performed in H1299 or RKO cells following HU and/or daunorubicin treatment. Cells were treated as described for Fig. 2A, and lysates were immunoprecipitated with anti-p53 monoclonal antibodies (1801 and DO-1). Approximately 0.2% of input chromatin and 0.5% of ChIP DNA was used as a template in the PCRs. Graphs show the averages and standard deviations (error bars) for two independent ChIP experiments and are representative of a total of five experiments. Data are expressed graphically as induction (n-fold) compared to untreated RKO cells. p21 NEG (2.8 kb upstream of the 5' response element) and GAPDH primer sets were included as negative controls to confirm the specificity of the ChIP assay. Ctrl, control; D, dauno; HU/D, HU and dauno.

Impaired transcription of the p21 gene is not the result of defective acetylation of core promoter histones H3 and H4. Acetylation of critical lysines within histones (along withmethylation of other lysines) causes remodeling of chromatinin the vicinity of the modification and, in turn, facilitatestranscription of the relevant gene (19, 24). p53 is known tobe functionally and physically associated with several acetyltransferases,including p300/CBP, PCAF, Tip60, and TRRAP (7, 15, 19). Theseinteractions can lead to promoter histone acetylation and increasedtranscriptional activity (1, 3, 5, 22). We examined the statusof histone acetylation within the p21 promoter TATA region followingHU and dauno treatment by performing ChIP assays with antibodiesagainst the acetylated forms of histone H3 and H4 (Fig. 6).As a control for basal p53-independent histone acetylation withinthe p21 promoter, we used H1299 cells, which are null for p53.Compared to untreated RKO cells, both HU and dauno treatmentsled to robust induction of acetylation of both histones, asdid the combination of HU and dauno (Fig. 6A). Thus, althoughthe overall cellular levels of acetyl-histone H3 and H4 proteinwere not increased following drug treatment (data not shown),the presence of these modified forms within the p21 promoterwas damage dependent and likely resulted from the binding ofp53 protein to the promoter. There were no differences seenbetween HU or dauno treatments alone or together, indicatingthat nucleosomes at the p21 promoter were modified to a similarextent whether they were being transcribed at low (HU) or high(dauno) levels.

View larger version (44K):
[in this window]
[in a new window]

FIG. 6. HU treatment does not impair acetylation of histone H3 and H4 at the p21 promoter. (A) H1299 and RKO cells were treated as described for Fig. 2A, and ChIP assays were performed to compare p21 promoter histone acetylation following HU or daunorubicin treatment. Lysates were immunoprecipitated with polyclonal antibodies specific for acetylated histones H3 and histone H4. A sample in which lysates were incubated with beads lacking antibody (no Ab) was included to determine background levels of chromatin pulled down by beads alone. Approximately 0.2% of input chromatin and 0.5% of ChIP-DNA was used as a template in the PCRs with [ ${alpha}$ -³²P]dCTP in the reaction mix. Radioactive amplicons were resolved on 8% acrylamide gels, dried, and autoradiographed. ChIP data were analyzed by phosphorimaging and normalized to their respective inputs. Graphs show the averages and standard deviations (error bars) for two independent ChIP experiments, expressed as induction (n-fold) compared to untreated RKO cells. (B) Lysates from RKO cells were immunoprecipitated with acetylated histone antibodies (as described for panel A), and PCR was performed using primers to amplify different regions of the p21 promoter and gene as indicated (see Fig. 5A for primer locations). Amplicons were resolved by agarose gel electrophoresis. Ctrl or C, control; D, dauno; HU/D, HU and dauno.

When the extent of acetylated histones H3 and H4 was assessedwithin other regions of the p21 gene (Fig. 5A, diagram of p21locus and primer positions), the highest level of acetylatedhistones occurred around the p53 5' response element withinthe p21 promoter, and strong signals were obtained as far downstreamas intron 1 (+3 kb), correlating with both drug treatment andhigh levels of p53 binding (Fig. 6B). Conversely, very low levelsof acetylated histones were detected both in a distal regionof the p21 promoter (p21 NEG, approximately 2.8 kb upstreamof the 5' p53 binding site, where no p53 binding is observedby ChIP) or within a region further downstream (p21 Exon 2,+5.5 kb). Even at these distal regions, somewhat more acetylationof these histones was still detected after drug treatment. Whetherthe extent of acetylation was relatively high or low, no differencesbetween HU and dauno-treated cells were seen. Thus, althoughwe cannot rule out subtle variations in acetylated histone H3and H4 residues, our data indicate that HU-mediated transcriptionalimpairment occurs downstream of p21 promoter histone acetylation.

Association of TATA-binding protein and RNA polymerase II (RNA Pol II) with the p21 promoter is not affected by stalled DNA replication. Recruitment of the TATA-binding protein (TBP) as part of itsholocomplex, TFIID, is one of the first events in the assemblyof the transcription initiation machinery, and this proteinhas been shown to functionally and physically interact withp53 (10, 49, 61). We examined the ability of TBP to assembleat the p21 promoter TATA box region following DNA replicationblock by using ChIP with anti-TBP antibodies. Consistent withprevious reports, basal binding by TBP was quite high (23, 27);however, an approximately 1.5- to 2.5-fold increase in occupancywas reproducibly seen upon damage. Here, too, there was no differencein TBP binding in cells treated with HU or daunorubicin (Fig.7A and B).

View larger version (29K):
[in this window]
[in a new window]

FIG. 7. TATA-binding protein and RNA polymerase II are recruited to the p21 promoter following HU treatment. (A) RKO cells were treated as described for Fig. 2A, and ChIP assays were performed using lysates with a polyclonal TBP antibody to immunoprecipitate TBP/chromatin complexes. Approximately 0.2% of input chromatin and 4% of ChIP-DNA was used as a template in the PCRs including [ ${alpha}$ -³²P]dCTP in the reaction mix. (B) Graphical representation of the data shown in panel A. The graph shows the averages and standard deviations (error bars) for three independent experiments, expressed as induction (n-fold) compared to untreated RKO cells. (C to E) H1299 and RKO cells were treated as for Fig. 2A, and ChIP assays were performed using a polyclonal antibody against the large subunit of RNA Pol II. DNA isolated from anti-RNA Pol II ChIPs was subjected to PCR analysis (as in panel A) using primers for the p21 TATA region (C, p21 TATA) or a p21 5' distal region as shown in the diagram in Fig. 5A (D, p21 NEG). (E) Graphical representation of the data shown in panels C and D. Data are the averages from five (C, p21 TATA) or three (D, p21 NEG) independent experiments, with error bars representing one standard deviation. As in Fig. 6, ChIP assays were performed in parallel with a no-antibody control (no Ab) to determine background levels of chromatin pulled down by beads alone. Ctrl, control; D, dauno; HU/D, HU and dauno.

Having found that bound p53, acetylated histones, and TBP recruitmentwere essentially unaffected by HU treatment, we went on to examinethe recruitment of RNA Pol II to the p21 promoter. ChIP assayswere performed using antibodies against the largest subunitof RNA Pol II, and PCRs were performed using primers flankingthe TATA box region (Fig. 7C). Recruitment of polymerase tothe p21 promoter TATA box region occurred in a stress- and p53-dependentmanner (Fig. 7C, compare lane 1 and 2 to lanes 3 to 5; Fig.7E), and the significant presence of RNA Pol II was not detectedwithin a distal region 5 kb upstream of the start site withinthe p21 promoter (Fig. 7D, p21 NEG PCR; Fig. 7E). Nevertheless,we found RNA Pol II poised to initiate transcription to similarextents in cells that have been treated with HU, dauno, or acombination of both drugs. Taken together, these results indicatethat impaired p21 transcription in response to HU treatmentwas most likely regulated through a postinitiation mechanism.

HU suppresses p21 transcription by inhibiting RNA polymerase II elongation. To determine if transcriptional elongation was affected in cellstreated with HU, ChIP assays were performed to examine RNA PolII occupancy downstream within the p21 gene (Fig. 5A shows keyregions within the p21 gene that were examined). Indeed, thepresence of RNA Pol II was decreased further downstream withinExon 2 (+5.5 kb) in cells treated with HU than in samples fromdaunorubicin-treated cells (Fig. 8A and C). By contrast, whenthe PIG3 gene was examined, there were no significant differencesin the amount of RNA polymerase detected within the transcribedregion of the PIG3 second Exon (+2.2 kb) (Fig. 8B and C). Hence,RNA Pol II is less efficient at traversing the p21 gene afterHU treatment than it is in samples from dauno-treated cells.Since the level of Pol II found at exon 2 of the PIG3 gene wasquite high prior to drug treatment (Fig. 8B, lane 1 comparedto lanes 2 to 4), we wanted to extend this observation by examiningPol II occupancy at regions further downstream within the p21and PIG3 genes. When Pol II levels were examined at exon 3 ofthe p21 (Fig. 8D, +7.5 kb) or PIG3 (Fig. 8E, +5.5 kb) gene,the results were essentially the same as they were for exon2 of these genes, except that a twofold increase in Pol II occupancywas observed at PIG3 exon 3 following drug treatment and, again,there was no significant change seen after HU or daunorubicintreatment (Fig. 8E, compare lane 1 to lane 2 to 4; Fig. 8F).It was noted that occupancy of RNA Pol II in exon 2 and exon3 of the p21 gene was somewhat higher in cells treated withthe HU/dauno drug combination than that seen with HU alone,indicating that the presence of dauno leads to a partial rescueof transcriptional impairment. These data are consistent withour RT-PCR data (Fig. 2B, compare lanes 3 and 5 to lane 4) andsuggests that the small amount of increased p21 mRNA seen withthe combination of both treatments can be attributed to RNAPol II elongation within the p21 gene.

View larger version (52K):
[in this window]
[in a new window]

FIG. 8. HU treatment suppresses RNA polymerase II elongation of the p21 gene. (A to F) DNA samples from the RNA Pol II ChIPs in Fig. 7C to E (RKO cells) were also subjected to PCR using primers for p21 Exon 2 (A, p21 Exon 2) as well as the second exon within the PIG3 gene (B, PIG3 Exon2). (C) Graphical representation of three independent experiments as shown in panels A and B. The same DNA was then subjected to PCR using primers generated within the third exon of p21 (D, p21 Exon 3) or PIG3 (E, PIG3 Exon 3). (F) Graphical representation of the averages for three or two experiments, shown in panels D and E, respectively. (G) ChIP assays were performed as described in Materials and Methods, using a phosphospecific antibody that recognizes phosphorylated serine 2 within the RNA Pol II CTD (anti-Pol II-CTD serine 2P), and PCRs were carried out using primers to amplify exon 3 of the p21 gene. (H) DNA samples generated (panel G) were then used in PCRs containing primers against the TATA region of the p21 gene as a negative control. (I) Graphical representation of the data shown in panels G and H. Data are the averages for three independent experiments. ChIP data are normalized to their respective inputs and expressed as occupancy (n-fold) compared to untreated RKO cells. PCRs were carried out in the presence of [ ${alpha}$ -³²P]dCTP in all cases except for those shown in panels D and E, in which radioactive nucleotides were excluded from the reaction mixture. For these experiments, amplicons were resolved by agarose gel electrophoresis. In all graphs, error bars represent one standard deviation. C or Ctrl, control, D, dauno; HU/D, HU and dauno.

A substantial amount of regulation of RNA Pol II activity occursthrough posttranslational modification of the large, catalyticsubunit of the complex. The CTD of this subunit contains multiple(52 in human RNA Pol II) repeats of a motif (YSPTSPS) that canbe phosphorylated at different sites to either maintain poisedRNA polymerase at the initiation site (serine 5) or activateRNA polymerase for transcription elongation (serine 2; hereafterreferred to as Pol II-CTD serine 2P) (43, 64). Gomes et al.(27) reported that doxorubicin treatment leads to a marked increasein the amount of Pol II-CTD serine 2P within the p21 gene locusat regions far downstream of the start site. We confirmed andextended this observation using antibodies against Pol II-CTDserine 2P to show that dauno treatment increased levels of thisform of RNA Pol II within exon 3 of the p21 gene (Fig. 8G).Relative to cells treated with daunorubicin, there was a significantdecrease in the amount of Pol II-CTD serine 2P present in thisregion after HU treatment (Fig. 8G, lanes 2 and 4; Fig. 8I).By contrast, there was less Pol II-CTD serine 2P at the TATAbox, and no differences were seen for HU- and daunorubicin-treatedcells at this site (Fig. 8H and I). These data indicate thatp21 transcriptional impairment in our system is the result ofreduced elongation of the p21 gene by RNA Pol II and pointsto a postinitiation mechanism that is regulated by signals occurringas a result of blocked DNA replication.

DISCUSSION

Top

Discussion

Upon stalled DNA replication caused by HU, p53 protein can accumulate,bind to the promoters of its target genes, facilitate promoterhistone acetylation, and recruit key components of the basaltranscriptional machinery. Under the same conditions, however,transcription of the p21 gene is substantially reduced comparedto other forms of stress, such as daunorubicin, and this reductioncorrelates with decreased elongation of RNA Pol II along thep21 gene. These observations have some interesting implications,which are discussed below.

It was unexpected that HU treatment leads to high levels ofstabilized p53 that can bind efficiently to its some of itstarget genes without a subsequent increase in their transactivation.A few previous reports, however, document the accumulation ofa wild-type protein that is incapable of fully activating transcription.Hypoxic conditions cause p53 levels to rise but selectivelyaffect its posttranslational modifications and ability to regulatetranscription (35, 44). Relevant to our findings is the observationthat hypoxia also elicits an S-phase arrest and activation ofATR (34). Another drug shown to produce high levels of inactivep53 is 5,10-dideazatetrahydrofolate, an inhibitor of de novopurine synthesis. Here, too, p53 is unable to transcribe thep21 gene, although in this case decreased transcription is associatedwith a lack of complete p53 phosphorylation and acetylationas well as inadequate histone acetylation (8). Although we cannotaddress whether all modifications of p53 were similarly affectedafter HU and dauno treatment, p53 is phosphorylated at S15,S46, and S392 and acetylated at K382 to similar extents aftergamma irradiation and HU treatment (32). Therefore, we do notthink posttranslational modification of p53 is likely to explainthe defect seen after HU treatment, even though we found thatlevels of p53 generally reflect the extent of its promoter binding,as others have also reported (38, 67).

It was initially unanticipated that acetylation of histonesH3 and H4 was markedly and equivalently increased by HU anddaunorubicin. In addition to the link between p53 and histoneacetylation, p53 has been shown to activate transcription throughits association with histone methyltransferases, such as CARM1,PRMT1 (1), and SET9 (13), and with ubiquitin-conjugating enzymes,such as the hBRE1 histone E3-ubiquitin ligase (41), and p53has been shown to repress at least one gene by mediating dimethylationof lysine 9 of histone H3 (55). Although we were unable to detectany deficiencies in histone acetylation within the p21 genethat would explain HU-induced transcriptional impairment andwe cannot rule out the possibility that other changes in histonemodifications may be at play, preliminary experiments did notindicate that altered methylation of histone H3 lysines explainsreduced p21 transcription (data not shown). In light of ourfindings that impairment of p21 expression is occurring downstreamof transcription initiation, it is not surprising that thesemodifications are unaffected by HU treatment.

Our data indicate that one or more components involved in thecomplex process of RNA elongation are affected by agents, suchas hydroxyurea, that produce a DNA replication checkpoint. Transcriptionalelongation involves numerous factors and complexes, includingelongins, ELL proteins, TFIIH, p-TEFb, DSIF, NELF, FACT, andmany others, that play roles such as direct stimulation of elongation,alleviation of transcriptional pausing, and chromatin remodelingto facilitate elongation through chromatin (64). Although themechanisms of transcriptional elongation are not entirely understood,recent work has revealed several levels of complexity and mechanismsthat p53 could regulate. In fact, it was shown that p53 interactswith several polypeptide components of the basal transcriptionfactor TFIIH (42, 46, 76), which has been shown to be involvednot only in initiation but also in promoter clearance by RNAPol II (17, 18, 28, 46). Functional and physical interactionsbetween p53 and the RNA Pol II elongation factor ELL have alsobeen reported (62). While both p53 and ELL work to enhance genetranscription in most situations, the proteins functionallyantagonize each other when they are physically associated andcan lead to decreased transcription of p21 (62). The ELL geneis frequently involved in a chromosomal translocation whichfuses a large portion of the gene to the trithorax-like MLLgene, generating the MLL-ELL protein, which is often found inacute myeloid leukemias (68) and which was also found to associatewith and downregulate p53 (74). Considering the historical evidencefor the success of HU in the treatment of leukemia, it is possiblethat HU affects p21 transcriptional elongation through the p53-ELLfusion axis. Finally, a recent study reported that p53 is involvedin a feedback loop with cdk9 (a subunit of the p-TEFb elongationfactor), in which p53 interacts with and is phosphorylated onS392 by cdk9 (14). However, since we previously found that p53is equivalently phosphorylated at S392 after HU treatment andgamma irradiation, which have opposing effects on p21 transcription,it is unlikely that cdk9 has a direct impact on p53 in our system(32). Although we do not yet know whether stalled DNA replicationis regulating events at the promoter clearance stage or at apoint within the elongation phase of the transcription cycle,one can imagine many scenarios in which stalled replicationforks could signal to p53 or its associated elongation factorand alter gene expression.

Looking ahead, we have identified two critical directions tofollow. The first is to identify those RNA elongation targetsthat are affected by blocking DNA replication. Particularlyrelevant to our observations is a recent elegant study fromGomes et al. (27), who showed that RNA Pol II, normally presentand paused at the p21 promoter, initiates transcription in ap53-dependent manner upon stress caused by doxorubicin (an agentsimilar to the dauno used in our experiments). As RNA Pol IItransits the p21 gene, it becomes more phosphorylated at serine2 of its CTD, consistent with previous studies linking thisphosphorylation event to actively transcribing polymerase (43,64). Gomes et al. also showed that a number of elongation factors,including the aforementioned ones, are recruited during thisprocess (27). It will be interesting to determine if the functionsof these are affected by stalled DNA replication and, if so,the mechanism by which this occurs. The second direction isto determine the signaling pathway that affects the processof elongation of p21 RNA transcripts. The upstream respondersto stalled DNA replication caused by either depleted dNTPs afterHU treatment or blocked DNA elongation by aphidicolin includeRP-A, H2AX, ATR-ATRIP, claspin, and the repair clamp and clamp-loadercomplexes, Rad 17 and 9-1-1 (6, 31). Activated Chk1, the mediatorkinase that is a key target of this complex, may regulate oneor more of the elongation factors that are involved in transcriptionof the p21 gene. It is hoped that future experiments will shedlight on these questions.

It is fascinating to ask why, when p21 is so regulated, otherp53 targets, such as PIG3, are not affected by this process.Even though RNA Pol II was shown by Espinosa and Emerson (23)to be present and poised at a subset of p53 targets, includingp21, prior to p53 activation, others, such as PIG3, do not havemuch preloaded RNA polymerase II. These authors also found thatonce p53 is induced, there is a net loss of RNA Pol II at thep21 initiation site at early time points, while in at leastone case (FAS/Apo1), there is no such loss at the FAS/Apo1 promoter,suggesting that the latter has a much higher rate of reinitiationthat could conceivably overcome a reduced rate of elongationat this gene. Whether this is the case with PIG3 remains tobe determined. Alternately, poised RNA polymerase at the p21TATA region may recruit one or more distinct elongation factorsthat are not involved in transcription of other genes, suchas PIG3, and such factors are the direct targets of the DNAreplication checkpoint.

It should be mentioned here that our results are not seen inevery human cell line; p21 mRNA accumulates in HU-treated MCF7cells that contain wild-type p53 (54; M. Mattia and C. Prives,unpublished observations). Nevertheless, they are not uniqueto RKO cells, since we also observed that markedly less p21mRNA accumulated after HU treatment than after daunorubicintreatment in another cell line (HCT116) (see Fig. S1C and Din the supplemental material). It will be interesting to eventuallydetermine the reasons for cell-type differences in responsesto stalled DNA replication. As we understand more about theupstream signaling pathways and their relationships to factorsthat regulate transcription, such differences should be elucidated.Clearly, there is much to do to further understand the processby which some transcription units are differentially affectedwhen DNA replication is stalled.

ACKNOWLEDGMENTS

We thank James Manley, Lili Yamasaki, Ron Prywes, and JoaquinEspinosa for advice and suggestions. We are grateful to EllaFreulich for expert technical assistance, Masha Poyurovsky andAndrew Zupnick for critical readings of the manuscript, andother members of the Prives lab for constant help and support.

This work was supported by NIH grant CA58316 and a fellowshipfrom the Leukemia and Lymphoma Society to V.G.

FOOTNOTES

* Corresponding author. Mailing address: Department of Biological Sciences, Columbia University, New York, NY 10027. Phone: (212) 854-2557. Fax: (212) 865-8246. E-mail: clp3{at}columbia.edu.

Published ahead of print on 11 December 2006.

Supplemental material for this article may be found at http://mcb.asm.org/.

Present address: Fundación Instituto Leloir, Buenos Aires1405, Argentina.

Present address: Dana-Farber Cancer Institute, Harvard MedicalSchool, Boston, MA 02115.

REFERENCES

Top