Identification of candidate p53-regulated miRNAs
In order to identify miRNAs potentially regulated by p53, we utilized the p53
wild-type HCT116 colon cancer cell line (p53WT
) and an isogenic cell line in which both alleles of p53
were inactivated by homologous recombination (p53-/-
) (Bunz et al., 1998
). Cells were treated with the DNA damaging agent adriamycin, which has previously been demonstrated to lead to induction of p53 and its downstream targets (Waldman et al., 1995
), and RNA was analyzed using a custom microarray developed in our laboratory capable of monitoring the expression of 474 human miRNAs. miRNAs exhibiting a 3-fold or greater change in expression upon drug treatment in the p53WT
cells and less than a 2-fold change in expression in the p53-/-
cells were chosen for further study. Seven upregulated miRNAs satisfied these criteria and showed the expected expression pattern by northern blotting with adriamycin-induced accumulation specifically in p53WT
cells (miR-23a, miR-26a, miR-34a, miR-30c, miR-103, miR-107, and miR-182; ). No miRNAs that were downregulated in a p53-dependent manner following adriamycin treatment were observed. A previous study in which polysome-associated miRNAs in HCT116 p53WT
cells were analyzed by microarray also identified miR-26a as a potential target of p53 (Xi et al., 2006
miR-34a is induced by p53 following DNA damage
Characterization of the miR-34a primary transcript
Among the DNA-damage induced miRNAs, miR-34a exhibited the largest magnitude of upregulation. Although there are two additional human miR-34 homologues (miR-34b and miR-34c), these miRNAs were not detected by microarray analyses suggesting that they are not expressed in this cell line. In order to determine whether miR-34a is a direct transcriptional target of p53, we first elucidated the complete structure of the miR-34a primary transcript (pri-miR-34a). This miRNA is contained within the second exon of a spliced expressed sequence tag (EST) (DB286351; Supplemental Figure 1
). Rapid amplification of cDNA ends (RACE) was used to define the complete 5′ and 3′ ends of this transcript. To facilitate these efforts, pri-miRNAs were stabilized by small interfering RNA (siRNA)-mediated inhibition of Drosha, the endonuclease which performs the first step in miRNA processing (Lee et al., 2003b
). These experiments revealed that the 5′ end of the miR-34a pri-miRNA is heterogeneous, with all transcripts initiating in an approximately 100 base-pair (bp) region in the vicinity of the 5′ end of the DB286351 EST (Supplemental Figure 1
). Although this region is within a large (>1.5 kb) CpG island, no sequences resembling a consensus TATA box (TATAAA) or initiator element (YYANTY) were present near the putative transcription start sites. 3′ RACE demonstrated that all transcripts terminate at a single position 360 bp downstream of the 3′ end of DB286351. A polyadenylation consensus site is present 22 bp upstream of this nucleotide. We also confirmed efficient splicing of the approximately 30 kb intron separating exons 1 and 2 and a lack of any intervening exons by performing reverse-transcriptase PCR (RT-PCR) with primers near the 5′ and 3′ ends of the pri-miRNA (data not shown).
p53-dependent transcriptional activity of the miR-34a promoter
We next investigated whether the miR-34a pri-miRNA is directly regulated by p53. While the region upstream of the pri-miRNA is not highly conserved between human and mouse or human and rat, a highly conserved perfect consensus p53 binding site is located just downstream of the transcription start site (). To test for transcriptional activity, a series of genomic fragments were cloned into a promoterless luciferase reporter plasmid (). Because several potential out-of-frame translation initiation codons are present between the transcription start site and the luciferase open reading frame in these reporter vectors, we also cloned an internal ribosome entry site (IRES) downstream of the putative promoter fragments. Reporter activity was then tested in HCT116 p53WT
cells. The longest construct extended from 1.4 kb upstream to 578 bp downstream of the most 5′ transcription start site and included the putative p53 binding site (, P1). This fragment yielded robust p53-dependent transcriptional activity. This activity was comparable to that observed when a similar reporter vector containing the SV40 viral promoter was transfected into this cell line (data not shown). The activity of this promoter is not limited to HCT116 cells as shown by transfection experiments in HEK293T cells (Supplemental Figure 2
). Removal of the p53 binding site by truncation or mutation abolished transcriptional activity (, P3, P6, and P1mut
). Remarkably, a 307 bp fragment containing the p53 binding site located downstream of the transcription start site (, P5) was sufficient for full promoter activity.
To test whether genotoxic stress induces transcriptional activity of the miR-34a promoter, reporter constructs were transfected into HCT116 p53WT
cells in the presence or absence of adriamycin. Precisely mirroring the expression of endogenous miR-34a, luciferase activity produced from the P1 and P5 reporters was induced by DNA damage in a p53-dependent manner (). As expected, mutation of the p53 binding site abolished this response. Providing further evidence that p53 directly regulates this miRNA, a previously published analysis of genome-wide binding sites for p53 using chromatin immunoprecipitation revealed that this protein directly binds to the genomic region defined here as the miR-34a promoter (Wei et al., 2006
). This earlier study did not associate this binding site with regulation of miR-34a, likely because the miRNA is located more than 30 kb away. Taken together with the data provided here, these findings provide compelling evidence that miR-34a is a direct transcriptional target of p53.
Expression of miR-34a promotes apoptosis
In order to investigate the phenotypic consequences of miR-34a expression, we transiently transfected this miRNA into HCT116 p53WT and p53-/- cells and measured apoptotic cell death by flow cytometric measurement of DNA content () and Annexin V staining (). Cells transfected with control oligonucleotide exhibited low levels of apoptosis [p53WT, 6.6 ± 3.4% apoptotic cells; p53-/-, 4.1 ± 1.6% apoptotic cells (means and standard deviations from 3 independent experiments reported)]. In contrast, transfection of p53WT cells with synthetic miR-34a potently induced apoptotic cell death (24.2 ± 3.8% apoptotic cells). Interestingly, apoptosis was substantially decreased but not completely abolished following transfection of miR-34a into p53-/- cells (9.3 ± 2.5% apoptotic cells), suggesting both p53-dependent and p53-independent mechanisms of miR-34a-induced cell death. These data suggest that miR-34a participates in the apoptotic program triggered by p53 activation.
Expression of miR-34a promotes apoptosis
Loss of miR-34a expression occurs frequently in pancreatic cancer cells
We next examined miR-34a expression in pancreatic cancer cells which frequently exhibit p53 loss-of-function. Two non-transformed pancreatic ductal epithelial cell lines (HPNE and HPDE) (Lee et al., 2003a
; Ouyang et al., 2000
) as well as 15 pancreatic cancer cell lines were analyzed by northern blotting (). miR-34a was highly expressed in HPNE and HPDE cells, demonstrating that this miRNA is normally expressed in this cell type. Remarkably, all 15 pancreatic cancer cell lines showed at least a two-fold reduction in miR-34a expression as compared to expression in HPNE and HPDE. 11/15 cell lines exhibited a ten-fold reduction or complete absence of this miRNA.
Expression of miR-34a is frequently lost in pancreatic cancer cells
Although p53 loss would be expected to reduce miR-34a expression, it is unlikely that this mechanism can account for the reduced miRNA expression in all the pancreatic cancer cell lines. There is not a direct correlation between bi-allelic loss of p53 and the magnitude of miR-34a downregulation ( and Supplemental Table 1
, compare BxPc3, PK9, CAPAN1, and Su86.86 all of which exhibit p53 loss of function). Similarly, cell lines with wild-type p53 status also exhibit low levels of miR-34a (Panc04.14 and E3LZ10.7). Thus, other mechanisms in addition to p53 inactivation likely contribute to the reduction in miR-34a abundance. In fact, deletion of the genomic interval encompassing this miRNA (1p36) is an extremely frequent event in diverse types of cancer and miR-34a is located within a recently mapped 5.4 megabase minimally-deleted region of 1p36 that commonly occurs in gliomas (Bagchi et al., 2007
; Bello et al., 1995
; Bieche et al., 1993
; Moley et al., 1992
; Mori et al., 1998
; Poetsch et al., 2003
). It has also very recently been reported that reduced miR-34a expression associated with loss of 1p36 is a frequent event in neuroblastoma (Welch et al., 2007
). We therefore examined data from a previously published high-resolution copy number analysis of the genomes of pancreatic cancer cell lines (Calhoun et al., 2006
). This study included 11/15 of the lines we studied by northern blotting experiments. Notably, hemizygous loss of the miR-34a locus was observed in 3 lines (Supplemental Table 1
; Panc 02.13, Panc 04.14, and Panc 08.13). Thus, lack of transcriptional transactivation by p53, deletion, and additional unknown mechanisms likely contribute to loss of this miRNA in pancreatic cancer. Our results, together with previous studies, demonstrate that loss of miR-34a is a frequent event in diverse cancer subtypes and raise the possibility that miR-34a loss of function contributes to cancer pathogenesis.
Induction of miR-34a leads to widespread alterations in gene expression
To determine the effects of miR-34a induction on gene expression, a retroviral construct was used to generate HCT116 p53WT
cells with enforced expression of this miRNA. These cells exhibited 3-4-fold higher miR-34a expression levels than cells transduced with empty virus (), a magnitude of induction very similar to the endogenous upregulation caused by adriamycin treatment (). Affymetrix gene expression profiling was then used to examine the transcriptomes of retrovirally-infected cell populations (empty virus versus miR-34a virus) in the absence of genotoxic stress. Although bona fide
miR-34a targets that are not affected at the level of mRNA abundance will be missed by this approach, other groups have employed this strategy successfully to identify miRNA targets (Krutzfeldt et al., 2005
; Lim et al., 2005
; Voorhoeve et al., 2006
). Surprisingly, cells with enforced expression of miR-34a showed a dramatically altered gene expression profile with upregulation of 532 transcripts and downregulation of 681 transcripts (Supplemental Table 2
). This is a significantly greater number of transcripts showing altered expression than observed when similar experiments were performed previously with other miRNAs (Krutzfeldt et al., 2005
; Lim et al., 2005
). Select transcripts examined by northern blotting exhibited the expected expression changes, validating the quality of the microarray dataset (). To determine what fraction of the downregulated genes are likely to be direct targets of miR-34a, we examined the frequency of the hexamer complementary to the miR-34a seed sequence (ACUGCC) in the 3′ untranslated regions (UTRs) of the downregulated, unchanged, and upregulated transcripts. This motif is statistically-significantly enriched (p value = 0.023) among the top 100 downregulated transcripts (transcripts exhibiting a 2.364-fold or greater downregulation). Among transcripts showing the same fold increase in expression, this motif is statistically-significantly underrepresented (p value = 5.6 × 10-4
). At this cut-off, 32/100 downregulated genes (32%), 2190/9534 unchanged genes (23%), and 18/149 upregulated genes (12%) contained the hexamer complementary to the miR-34a seed in their 3′ UTRs (). These findings demonstrate that the miR-34a-downregulated genes are enriched for direct targets of this miRNA and the upregulated genes are depleted of direct targets. Nevertheless, the majority of gene expression changes are likely an indirect consequence of enforced miR-34a expression. For example, only 47/185 transcripts downregulated greater than two-fold have sites complementary to the miR-34a seed sequence in their 3′ UTRs.
Expression of miR-34a leads to widespread alterations in gene expression
To better understand the global effects of expressing miR-34a on the transcriptome, we examined the Gene Ontology classifications of the up- and downregulated genes. Strikingly, the most highly enriched gene ontology category among the upregulated transcripts was ‘cell cycle’ (p value = 1.8 × 10-48
). Genes classified as ‘DNA repair’ (p value = 6.6 × 10-23
), ‘mitotic checkpoint’ (p value = 1.9 × 10-8
), and ‘DNA integrity checkpoint’ (p value = 3.6 × 10-6
) were also highly enriched among the upregulated transcripts. Moreover, genes assigned to the term ‘cell proliferation’ (p value = 1.9 × 10-4
) and ‘angiogenesis’ (p value = 4.2 × 10-3
) were significantly enriched among the downregulated genes. Similar categories of genes are known to be activated and repressed by p53 (Giono and Manfredi, 2006
). Western blotting confirmed that p53 levels were unchanged in the miR-34a-expressing cells (Supplemental Figure 3
). We note that while p53 and miR-34a regulate genes involved in overlapping pathways, they do not necessarily regulate the same genes. Thus, by inducing this miRNA, p53 is able to directly and indirectly regulate pathways that dictate the cellular response to genotoxic stress.
The role of miR-34a in p53-mediated apoptosis merits further discussion. Close inspection of our gene expression profiling results reveals that among the transcripts downregulated by miR-34a are several with well-documented anti-apoptotic functions including B-cell CLL/lymphoma 2 (BCL2
), baculoviral IAP repeat-containing 3 (BIRC3
), and decoy receptor 3 (DcR3
also known as TNFRSF6B
) (Liston et al., 2003
; Sheikh and Fornace, 2000
; Willis et al., 2003
). Indeed, the Gene Ontology term ‘programmed cell death’ is significantly enriched among the genes downregulated by miR-34a expression (p value = 7.8 × 10-5
). These findings are consistent with our results indicating that miR-34a expression is sufficient to induce apoptosis. The set of miR-34a responsive genes described here will likely continue to provide mechanistic insight as additional functions of this miRNA are discovered.
The identification of target genes of the p53 tumor suppressor protein has been the focus of intense interest. In this report, we demonstrate that miRNAs are important components of the p53 transcriptional network. Our results show that p53 directly activates expression of miR-34a and potentially other miRNAs. Expression of miR-34a is sufficient to induce apoptosis through p53-dependent and independent mechanisms. Moreover, miR-34a induction leads to dramatic reprogramming of gene expression. Much like the known set of p53-regulated genes, miR-34a-responsive genes are highly enriched for those that regulate cell-cycle progression, cellular proliferation, apoptosis, DNA repair, and angiogenesis. Interestingly, loss of 1p36, the genomic interval harboring miR-34a, is common in diverse human cancers. Consistent with this observation, we show that reduced expression of miR-34a is a very frequent feature of pancreatic cancer cells. Together, these findings suggest an important role for miR-34a in mediating p53 tumor suppressor function.