Small Interfering RNA-Induced Transcriptional Gene Silencing in Human Cells

See allHide authors and affiliations

Science  27 Aug 2004:
Vol. 305, Issue 5688, pp. 1289-1292
DOI: 10.1126/science.1101372


Small interfering RNA (siRNA) and microRNA silence genes at the transcriptional, posttranscriptional, and/or translational level. Using human tissue culture cells, we show that promoter-directed siRNA inhibits transcription of an integrated, proviral, elongation factor 1alpha (EF1A) promoter–green fluorescent protein reporter gene and of endogenous EF1A. Silencing was associated with DNA methylation of the targeted sequence, and it required either active transport of siRNA into the nucleus or permeabilization of the nuclear envelope by lentiviral transduction. These results demonstrate that siRNA-directed transcriptional silencing is conserved in mammals, providing a means to inhibit mammalian gene function.

Small 21- to 25-nucleotide RNAs have diverse biological roles in eukaryotes, including transposon silencing and antiviral defense by small interfering RNAs (siRNAs) and developmental gene regulation by microRNAs (miRNAs) (13). siRNAs and miRNAs are processed from double-stranded precursors by the ribonuclease (RNase) III–RNA helicase Dicer (1). Argonaute proteins can bind small RNAs and are components of effector complexes that down-regulate gene expression by several mechanisms (4). Small RNAs with perfect homology to their target can cause specific mRNA cleavage (called RNA interference), whereas those with mismatches to their target mediate translational inhibition (3). Small RNA–mediated transcriptional gene silencing was first observed in plants through the use of inverted-repeat transgenes or transgenic viruses to generate siRNAs homologous to a target promoter (57). Promoter-directed siRNAs also silence transcription in the yeast Schizosaccharomyces pombe, and transcriptional silencing in Drosophila has been linked to an Argonaute protein (810). Transcriptional silencing by siRNAs probably reflects genome defense mechanisms that target chromatin modifications to endogenous silent loci such as transposons and repeated sequences (5, 1114).

Although siRNA-induced transcriptional gene silencing has not been reported in mammals, transcription of an antisense RNA has been implicated in gene silencing and DNA methylation (15, 16), and the structure of mouse pericentromeric heterochromatin may require an RNA component for its maintenance (17). Here we investigate whether siRNA-induced transcriptional gene silencing occurs in human cells.

We chose to target an elongation factor 1 alpha (EF1A) promoter–green fluorescent protein (GFP) reporter gene integrated into the genome of human 293FT cells by transduction with a feline immunodeficiency virus (FIV) vector (18). siRNA EF52 is homologous to an EF1A promoter sequence essential for transcription (AAG GTG GCG CGG GGT AAA CTG, –106 to –86 base pairs relative to the transcriptional start site) (19). A second siRNA homologous to exon 2 of the GFP coding region was designed to target posttranscriptional mRNA destruction. We transduced 293FT cells with the EF1A-GFP vector, allowed 24 hours for integration, then transfected with either EF52, GFP, or a control siRNA matching the human chemokine receptor CCR5. mRNA and DNA were analyzed 48 hours after siRNA transfection. The siRNA targeting the GFP mRNA transcript reduced expression relative to the control as measured by quantitative, real-time reverse transcription polymerase chain reaction (RT-PCR) (Fig. 1A). Potent inhibition of GFP expression was also seen with siRNA EF52 targeting the EF1A promoter (Fig. 1A, open bars).

Fig. 1.

(A) Promoter-targeted siRNA inhibits gene expression. 293FT cells were transduced in duplicate with lentivirus, then transfected with CCR5 (control), GFP (coding region), or EF52 (promoter) siRNAs. GFP mRNA was quantified by real-time RT-PCR. Gene expression was measured in triplicate; standard deviations are shown. (B) EF52 siRNA silences transcription. Nuclear run-on assays used nuclei from 293FT cells transduced with lentivirus and mock- or EF52 siRNA–transfected.

Several lines of evidence indicate that inhibition of GFP expression by the promoter-directed EF52 siRNA occurs at the transcriptional level. Transcriptional silencing in mammalian cells is associated with chromatin modifications that include histone deacetylation and cytosine DNA methylation (20). Silencing by EF52 siRNA was reversed by treating cells with trichostatin (TSA) and 5-azacytidine (5-azaC), inhibitors of histone deacetylases and DNA methyltransferases, respectively (20) (Fig. 1A). These agents did not affect RNA interference of the GFP transcript. We confirmed transcriptional silencing using nuclear run-on analysis, which indicated a 93% reduction in transcriptional initiation from the EF1A-GFP reporter gene in EF52-treated cells (Fig. 1B). A glyceraldehyde-phosphate dehydrogenase (GAPDH) control was unaffected by EF52 siRNA in nuclear run-on and RT-PCR experiments. The GAPDH and CCR5 siRNA controls show that promoter-directed transcriptional silencing is specific (Fig. 1B and fig. S1B).

We performed several control experiments to exclude alternative explanations and to confirm that EF52 siRNA down-regulates EF1A-GFP expression by silencing transcription. To ensure that transcription from the integrated transgenic EF1A promoter initiated in a similar position to endogenous EF1A, we performed RT-PCR (fig. S2) (19, 21). Specifically, transcripts containing the FIV rev-responsive element (RRE) upstream of the EF1A promoter were not detected, nor were other possible spliced messages initiating from the FIV long-terminal repeat (table S1 and fig. S2). These results show that siRNA EF52 targeted the transgenic EF1A promoter, not a transcribed region. Although the lentiviral vector used integrates into the chromosome and produces a transcriptionally active transgene within 24 hours (22), PCR analysis ensured that neither integration frequency nor total lentiviral DNA was affected by siRNA treatment (fig. S3). Collectively, these results demonstrate that EF52 siRNA targets a promoter region rather than transcribed RNA, does not reduce the number of transgenes, and induces transcriptional gene silencing in human cells.

Transcriptional gene silencing in mammalian cells is often accompanied by cytosine DNA methylation, and de novo DNA methylation in plants is guided by small RNAs (12, 20, 23). The EF52 siRNA target within the EF1A promoter contains a restriction site for the methylation-sensitive enzyme HinP1I. When methylated, this site is protected from digestion, and a PCR product spanning it can be amplified. The HinP1I site was unmethylated in genomic DNA from untreated cells and from cells treated with control CCR5 or GFP siRNAs. However, DNA methylation was detected in cells treated with EF52 promoter–directed siRNA (Fig. 2A) (the HinP1I assay measured DNA methylation at both the endogenous locus and the EF1A-GFP reporter). Methylation induced by EF52 siRNA was abolished by treatment with TSA and 5-azaC (Fig. 2A). Gene expression from a transfected reporter plasmid was similarly reduced by exogenous methylation (with DNA methyltransferase Sss-I) and restored by TSA and 5-azaC (Fig. 2B). These findings show that siRNA-induced transcriptional silencing in mammalian cells is associated with DNA methylation, a mark of silent chromatin at other loci (20).

Fig. 2.

siRNA-induced transcriptional silencing is associated with DNA methylation. (A) HinP1I-based DNA methylation assay of the EF1A promoter. DNA was prepared either (top) from lentiviral transduced cells transfected with CCR5 control, GFP, or EF52 siRNAs (with or without TSA and 5-azaC treatment) or (bottom) from untreated 293FT cells. HinP1I cut within the EF52 siRNA target site, preventing PCR amplification in unmethylated samples. Sss-I–methylated, EF1A-GFP plasmid DNA was a positive control. (B) Fluorescence-activated cell sorting analysis of GFP expression in cells transfected with Sss-I–methylated, EF1A-GFP plasmid, with or without TSA and 5-azaC treatment. TSA and 5-azaC counteracted transcriptional inhibition caused by DNA methylation. Standard errors of the mean are shown.

It is clearly important whether endogenous EF1A expression is silenced by promoter-directed siRNAs. However, in mammalian cells that have not been transduced with lentivirus, transfected small RNAs lack an efficient nuclear transport mechanism, and mammalian cells have specialized export pathways for hairpin-containing miRNA precursors (2426). This obstacle is circumvented by lentiviral transduction, which permeabilizes the nuclear membrane before siRNA transfection (22). In order to assess the effect of siRNA on the endogenous EF1A promoter, we transfected 293FT cells with EF52 and control [human immunodeficiency virus (HIV)–1 polymerase-specific] siRNAs using MPG, a bipartite amphipathic peptide incorporating a fusion peptide from HIV-1 gp41 transmembrane protein and the SV40 virus nuclear localization sequence (26). MPG facilitates the nuclear import of nucleic acids, including siRNAs (26). Cells transfected with EF52 and MPG showed significantly reduced endogenous EF1A expression by real-time RT-PCR, relative to the control (Fig. 3A). Transfection with EF52 and the conventional Transfast liposome reagent did not cause silencing, despite a transfection efficiency comparable to or greater than that of MPG (Fig. 3A). Silencing in cells transfected with EF52 and MPG was abolished by treatment with TSA and 5-azaC, indicating that it occurs at the transcriptional level (Fig. 3A). Furthermore, HinP1I digestion of the EF1A promoter was blocked in cells treated with EF52 and MPG, but not in cells transfected with EF52 and Transfast (Fig. 3B), indicating that siRNA-induced transcriptional silencing of the endogenous EF1A promoter is associated with DNA methylation.

Fig. 3.

(A) Promoter-targeted siRNA inhibits endogenous EF1A. EF1A expression was quantified by real-time RT-PCR in cells transfected with control (HIV-1 polymerase) or EF52 siRNAs with either MPG (a nuclear import–mediating peptide) or conventional Transfast reagent. Black columns represent MPG-transfected cells treated with TSA and 5-azaC. Standard errors of the mean were derived from four independent experiments. (B) siRNA-induced silencing of the endogenous EF1A promoter is associated with DNA methylation. DNA methylation of the endogenous EF1A promoter was assayed by the HinP1I method in cells that were transfected with control HIV-1 polymerase or EF52 siRNAs with either MPG or Transfast. (C) Nuclear-imported siRNAs inhibit transgenic EF1A-GFP long after lentiviral transduction. Lentiviral-transduced cells were sorted for GFP expression, grown for 8 weeks, then transfected as in (A). GFP mRNA expression was measured by real-time RT-PCR. The results represent two experiments with three independent samples per experiment; standard errors of the mean are shown.

As the effect of lentiviral transduction on the nuclear membrane is probably transient, we examined the need for nuclear transport of siRNAs in silencing an integrated EF1A-GFP reporter gene well after transduction (Fig. 3C). For this experiment, we transduced 293FT cells, isolated a GFP-positive population after 72 hours, and grew the cells for 8 weeks. As observed for endogenous EF1A, silencing of the integrated EF1A-GFP in this population depended on transfection with MPG and was reversed by TSA and 5-azaC (Fig. 3C). This suggests that siRNA-induced transcriptional silencing of an integrated reporter is not strictly dependent on lentiviral transduction, but rather on the ability of siRNAs to gain access to the nucleus. Silencing of endogenous EF1A and of the integrated reporter gene 8 weeks after transduction was less efficient than silencing of newly integrated EF1A-GFP (compare Fig. 3, A and C, to Fig. 1A). It is possible that MPG is less efficient than lentiviral transduction at transporting siRNAs into the nucleus. Alternatively, newly integrated EF1A-GFP transgenes may be more accessible to siRNAs because of their intrinsic chromatin structure; newly transformed transgenes are more susceptible to de novo DNA methylation and silencing in Arabidopsis (26).

Transfected siRNAs are generally retained in the cytoplasm of mammalian cells, where they mediate efficient mRNA cleavage but cannot target chromatin (24). siRNAs transcribed from hairpin transgenes are exported from the nucleus because they may resemble pre-miRNAs, which are produced by the nuclear RNase III Drosha and cleaved into mature miRNAs by cytoplasmic Dicer (1, 3, 28). These intrinsic features of mammalian cell biology may indicate why siRNA transport into the nucleus is necessary for transcriptional silencing.

Our findings confirm that siRNA-directed transcriptional gene silencing is conserved in mammalian cells. Small RNAs may guide mammalian transcriptional silencing in many different biological contexts, including the establishment of genomic imprints and targeting of DNA methylation to retroviruses and repeated transgenes (12, 29, 30).

Supporting Online Material

Materials and Methods

Figs. S1 to S3

Table S1

References and Notes

References and Notes

View Abstract

Navigate This Article