|Home | About | Journals | Submit | Contact Us | Français|
The putative function of highly conserved regions (HCRs) within 3′ untranslated regions (3′UTRs) as regulatory RNA sequences was efficiently and quantitatively assessed by using modular retroviral vectors. This strategy led to the identification of HCRs that alter gene expression in response to oxidative or mitogenic stress. Databases were screened for UTR sequences of >100 nucleotides that had retained 70% identity over more than 300 million years of evolution. The effects of 10 such HCRs on a standard reporter mRNA or protein were studied. To this end, we developed a modular retroviral vector that can allow for a direct comparison of the effects of different HCRs on gene expression independent of their gene-intrinsic 5′UTR, promoter, protein coding region, or poly(A) sequence. Five of the HCRs tested decreased mRNA steady-state levels 2- to 10-fold relative to controls, presumably by altering mRNA stability. One HCR increased translation, and one decreased translation. Elevated mitogen levels caused four HCRs to increase protein levels twofold. One HCR increased protein levels fourfold in response to hypoxia. Although nonconserved UTR sequences may also have a role, these results provide evidence that sequences that are highly conserved during evolution are good candidates for RNA motifs with posttranscriptional regulatory functions in gene expression.
The proliferative or differentiative state of a cell is dynamic and requires continuous regulation to be maintained (4–6). In response to extracellular cues, cells divide or cease division and differentiate. When responses to such cues go awry, neoplasia can result. Thus, knowledge of the molecular mechanisms that control gene expression in response to changes in the environment is of fundamental importance to understanding how cells develop normally or give rise to cancer and may lead to novel targets for therapeutic interventions.
Transcriptional controls influence gene expression by determining rates of mRNA production, but posttranscriptional controls are equally important in that they determine the amount of protein produced from that mRNA. The significance of such posttranscriptional controls is particularly clear in Xenopus laevis, Drosophila melanogaster, and Caenorhabditis elegans, where early patterning in development is largely determined by controlling the distribution, stability, and translation of inherited maternal transcripts (69). The magnitude of change in gene expression due to posttranscriptional mechanisms is often relatively small, yet two- to threefold increases or decreases in mRNA or protein abundance can have a major developmental impact. In mammals, posttranscriptional control appears to be especially important in order for cells to respond to changes in the environment, such as heat shock (70), the availability of iron (32), oxygen (48, 54), or growth factors (1). Posttranscriptional mechanisms may also serve to check and balance transcriptional regulation of gene expression. Although splicing (58), integration of selenocysteine (37), editing (10), frameshifting (2), and localization (75) are documented posttranscriptional control mechanisms, two particularly well established mechanisms for changing gene expression posttranscriptionally are alterations in mRNA stability (68) and mRNA translation efficiency (39, 87).
The sequences responsible for the posttranscriptional regulation of mRNAs often reside within the 3′ untranslated region (3′UTR) of the transcript. A remarkable finding is the existence of highly conserved regions (HCRs) within 3′UTRs (19, 20). When orthologous genes were compared, stretches of more than 100 to 2,000 nucleotides were found to exhibit more than 70% conservation over 300 to 500 million years of evolution, from mammals to birds, amphibians, or fish. In the absence of selective pressure, less than 30% conservation would be expected. Moreover, in greater than 10% of the cases analyzed, the sequence conservation within these HCRs in the 3′UTR was higher than the conservation in the protein coding region of the gene. This striking sequence conservation within 3′UTRs clearly reflects a strong selective pressure and suggests an essential function. However, we failed to detect shared sequence motifs among the 326 different HCRs that have been identified, although several are likely to act by similar mechanisms. Thus, although computer analysis can be used to identify HCRs within RNAs, it currently cannot predict their function.
In order to assess the potential functional significance of HCRs, which had thus far not been tested, an experimental approach was needed that allowed sufficient information to be rapidly gathered regarding the mechanisms by which these sequences act to control gene expression in mammalian cells. Most rigorous would be an analysis in the intact cell of the effects of HCRs, independent of other gene-specific regulatory elements, on the posttranscriptional regulation of reporter genes. Such studies require the abilities to introduce and express the wild-type or mutant RNA sequences efficiently and systematically in mammalian cells. To date, most studies have relied on transient or stable transfections, both of which suffer from variability in numbers of copies of the transgene taken up by each cell. Moreover, stable transfections that yield only a few clones may not be representative after weeks to months of selection, especially if the 3′UTR has an adverse effect on growth. In an attempt to overcome some of these problems, we used efficient methods for delivering 3′UTR sequences to cells and for rapidly selecting polyclonal populations with a low copy number.
We show here that an analysis of each of 10 distinct HCR sequences within 3′UTRs was capable of altering gene expression at the posttranscriptional level, albeit by different mechanisms. Two features were salient to these findings. First, we selected evolutionarily conserved sequences (HCRs) within 3′UTRs as a means of focusing the search for untranslated regions with a role in posttranscriptional regulation of gene expression. Although we postulate that sequences that have not diverged may have important functions, this does not rule out the possibility that certain nonconserved sequences also have such functions. Second, we developed a retroviral vector system that allows delivery of 3′UTR reporter constructs to populations of thousands of cells within 1 to 2 weeks, thus avoiding problems associated with clonal analysis and long-term selection. Moreover, the vector we developed is modular and can allow for a direct comparison of the effects of different HCRs on gene expression independent of their gene-intrinsic 5′UTR, promoter, protein coding region, or poly(A) sequence. Using this system we demonstrate that specific HCRs can induce changes in mRNA or protein accumulation under steady-state conditions. Moreover, certain HCRs serve as potent sensors that respond to local stresses and changes in the cell milieu, such as growth factors and hypoxia, typical of sites of injury, ischemia, or tumor development.
Vectors were constructed by using standard cloning procedures. To construct the Hermes HRSpuro-GUS reporter retrovirus, we used the following basic components: the β-geo clone (22) (SacI/XhoI, 300 bp) for the bovine growth hormone poly(A); the pBabe puro (57) for the SV40puro cassette (SalI/ClaI, 1,000 bp) and for the 5′ long terminal repeat (LTR) (SspI/BamHI, 1,000 bp); pGUSN358→S from Clontech (Sse8387I/EcoRI, 1,800 bp) for the Escherichia coli GUS gene; the Retrotet vector (36) for the 3′LTR (SIN) and most of the retroviral backbone (BamHI/SspI, 3,000 bp); and the O7CMVm (XhoI/EcoRI, 500 bp) for the inducible promoter consisting of seven copies of the tetracycline (tet) operator followed by a cytomegalovirus (CMV) minimal promoter (positions −53 to +75). All internal elements between the two LTRs were cloned one after the other in a Bluescript plasmid (Stratagene) containing a long polylinker. The plasmid backbone was then replaced by the retroviral backbone by a simple XhoI/BamHI digest.
The HCRs were amplified out of the genomic DNA from the specified species (Fig. (Fig.1)1) or from an EST clone for the Ran and EF-1α HCRs with the primers from the regions listed in Fig. Fig.1.1. The “vim a” HCR was amplified with the primers from regions 3 to 22 and 138 to 157. The “vim b” HCR was amplified with the primers from regions 138 to 157 and 273 to 292. The amplifications were performed by using the Expand High-Fidelity PCR system (Boehringer Mannheim). The amplified fragments were purified by gel electrophoresis and cloned directly in the SrfI site of a modified pCR-Script SK(+) phagemid (Stratagene). This vector was modified by replacing the BamHI-KpnI polylinker with an AscI site and destroying the BamHI site at the same time. HCRs were sequenced and found to be identical with the published sequences of those HCRs. However, the HuD HCR was found to have a deletion at its 3′ end (the last 23 bases were missing). All of the HCRs were cloned in the Hermes HRSpuro-GUS reporter retrovirus with the restriction enzymes AscI (5′ end) and BstXI (3′ end).
The transrepressor retrovirus, RetroTetRTRb(−) was constructed by cloning the tetR-KRAB (18) NcoI-BamHI fragment into the NcoI/BamHI sites of the MFG retroviral backbone (67). The transactivator retrovirus, RetroTetRTAb(+), was constructed by cloning rtetR-VP16 (27) into the NcoI/BamHI sites of the MFG retroviral backbone.
The mouse embryonic fibroblast cell line C3H10T1/2 (hereafter referred to as 10T1/2 cells) was purchased from the American Type Culture Collection (CCL-226; batch F-11839). 10T1/2 cells were maintained in growth medium (GM) consisting of Dulbecco’s modified Eagle’s medium (DME) (Irvine Scientific) with 20% serum (15% calf serum plus 5% fetal bovine serum [FBS]; both from HyClone). The retroviral-packaging cell line Phoenix-E was a gift from Garry Nolan (Stanford University). Phoenix-E cells were maintained in DME with 10% FBS. All media were supplemented with glutamine and penicillin-streptomycin according to the manufacturer’s recommendations. Cells were grown at 5% CO2. Transcriptional activity from the tetracycline-sensitive O7CMVm promoter was controlled by the tetracycline analog doxycycline hydrochloride (dox; Sigma). Relevant concentrations are indicated in the Results section.
When necessary, cultures were selected in the presence of the drug puromycin dihydrochloride (Sigma) at a concentration of 2.0 to 2.5 μg/ml. Cells were cultured in the presence of drug for at least three generations.
Retroviral particles were produced by transiently transfecting Phoenix-E cells and subsequently transducing 10T1/2 cells as previously described (74). Retroviral titers were similar for each HCR-expressing cell population as determined by the number of integrated transgenes, which differed by only twofold (results not shown).
10T1/2 cells were loaded with the fluorescent substrate fluorescein di-β-d-glucuronide (FDGlcU; Molecular Probes) and analyzed by fluorescence-activated cell sorting (FACS) as previously described (52, 59).
The presence of the GUS reporter gene product, β-glucuronidase, was assayed by first fixing cells in 4% paraformaldehyde–0.25% glutaraldehyde (Sigma) in 100 mM sodium phosphate buffer (pH 6.6) for 3 min at room temperature, followed by reaction with 5-bromo-4-chloro-3-indolyl-β-d-glucuronic acid (1 to 2 mM) (Gold Biotechnology) in a solution of 10 mM EDTA, 0.5 mM K3Fe(CN)6, K4Fe(CN)6, and 0.5 ml of Triton X-100 (New England Nuclear) in 100 mM sodium phosphate buffer at 37°C for several hours to overnight. The glucuronic acid stock was prepared in dimethylformamide at 40 mM and stored at −20°C in the dark.
Total RNA was isolated with the RNeasy Kit (Qiagen) according to the manufacturer’s instructions. The RNA was denatured, electrophoresed in a 0.8% agarose formaldehyde (Mallinckrodt) gel (14), transferred overnight to Nytran 0.45-μm-pore-size membranes (Schleicher & Schuell), and UV cross-linked in a Stratalinker (Stratagene) at an intensity of 70 mJ. Hybridization and detection steps were conducted as described previously; a stronger chemiluminescent substrate, CDP Star (Tropix), was used (72). The blots were imaged for 5 min, followed by a 30-min exposure with the Lumi-Imager, and the signal was quantified with the LumiAnalyst software (Boehringer Mannheim).
Digoxigenin-labeled riboprobes were synthesized and tested according to the Boehringer Mannheim instructions provided with the RNA labeling kit. The GUS riboprobe (1.8 kb) was synthesized with the T7 RNA polymerase, and the rpL32 (41) riboprobe (0.5 kb) was synthesized with the T3 RNA polymerase.
Cells were harvested by centrifugation at 1,000 rpm, and the cell pellet was lysed in lysis buffer consisting of Z buffer (60 mM Na2HPO4, 40 mM NaH2PO4, 10 mM KCl, 1 mM MgSO4; pH 7.0) with 0.2% Nonidet P-40 (Sigma). For each lysate, a dilution series in Z buffer was plated in a 96-well plate format with each well containing 100 μl of sample volume. An equal volume of chemiluminescent substrate, Glucuron (Tropix), diluted 1:100 in water, was added by using a multichannel pipetter to each well according to the recommendations of the supplier. The plate was then incubated at room temperature for 3 to 4 h. Subsequently, 100 μl of Light-Emission-Accelerator solution (Tropix) was added, followed immediately by analysis of the 96-well plate with the Lumi-Imager. The total protein content in the extracts was determined by using the Bio-Rad total protein assay and a microplate reader (model 450).
To generate hypoxic conditions, 70 × 103 to 100 × 103 cells were plated in glass dishes and grown overnight in normoxic GM. The medium was then changed to fresh GM. One set of dishes was returned to normoxia (21% O2, 5% CO2), and one set was subjected to hypoxia by placing the dishes in sealed aluminum jigs and exchanging the gas for five to six cycles with 95% nitrogen and 5% CO2 (<5 ppm of O2). This oxygen level has been documented as existing in spontaneous human tumors (35). After 15 h, both sets of dishes were harvested by trypsinization, washed one time with cold phosphate-buffered saline, pelleted, and frozen at −80°C.
To test the hypothesis that HCRs constitute functional regions within 3′UTRs, we analyzed the role of 10 different HCRs in the posttranscriptional regulation of gene expression (Fig. (Fig.1A).1A). Three of the HCRs tested (c-myc, transferrin receptor [TfR], and c-fos) were from 3′UTRs that had previously been shown to have significant effects on posttranscriptional control mechanisms and served as positive controls for the assays we developed (68). They were also selected for study because the regulatory elements identified by others based on functional analysis included HCRs identified by us based on computer analysis. In the case of the other seven HCRs selected for study, the potential functions of the 3′UTR have not previously been rigorously investigated. In each functional assay, the nonresponding HCRs demonstrated that the observed effects are sequence specific.
The HCRs selected for study were chosen because they not only exhibited a high degree of evolutionary conservation but were also associated with encoded products with known functions in cell cycle control, differentiation, and cancer. For example, bcl2 prevents apoptosis and provides a selective growth advantage for many cell types (82). The gene encoding ornithine decarboxylase (ODC) is amplified in certain tumors (81), and ODC levels correlate with proliferative capacity (56). Extracellular matrix components, such as fibronectin, alter growth and differentiation of many cell types (38). Overexpression of proteins involved in translation control, such as the eukaryotic elongation factor EF-1α and its mutants, have been found to be associated with cancer (76). Ran is a GTPase that has been implicated in numerous processes, including nuclear-cytosolic trafficking and cell cycle progression (66). Vimentin, like c-myc, increases in mitogen-stimulated cells (21). HuD has been implicated in cell differentiation and development (26). We therefore explored the possibility that the HCR of an mRNA has a function in determining the level of expression of the protein with which it is associated. This could occur either by altering mRNA stability or translation at steady state or in response to stresses such as changes in mitogen levels or oxygen deprivation.
To determine whether HCRs within 3′UTRs have a role in regulating gene expression, we constructed a retroviral vector, Hermes HRSpuro-GUS, that allows a rapid and efficient assessment of UTR function in populations of thousands of clones (Fig. (Fig.1B).1B). Two transcription units were included that encode a reporter gene (the bacterial β-glucuronidase, GUS) and a selectable marker (the puromycin resistance gene product, designated puro). The modular design of the vector, which has convenient 5′- and 3′-polylinker sequences flanking the GUS open reading frame, allows UTRs to be inserted as desired and permits their effects on a standard reporter mRNA or protein to be studied in the absence of their own 5′UTR, promoter, or polyadenylation signal sequences.
Another feature of the Hermes HRSpuro-GUS vector is that transcription from the CMV minimal promoter used in all constructs is antisense to transcription from the viral LTR. This feature allows the retroviral expression of a 3′UTR sequence with a nonviral polyadenylation signal, either its own or a well-characterized bovine growth hormone (bGH) poly(A) sequence (63). In the sense orientation, inclusion of a polyadenylation signal in the test 3′UTR would not be possible, as it would arrest transcription prior to the transcription stop signal located in the 3′LTR, thereby impairing the replication and production of infectious viruses. The potential to use the endogenous polyadenylation signal is important because many HCRs are juxtaposed to their polyadenylation signals and cannot therefore be easily separated from them (Fig. (Fig.1A).1A). For those HCRs located distal to the polyadenylation signal, the bGH poly(A) sequence was used.
A self-inactivating (SIN) sequence within the LTR is necessary in order to express CMV-directed transcripts in antisense. After infection of the target cell, the SIN vector, which contains a deletion in the enhancer and promoter sequences of the 3′LTR, transfers this deletion to the 5′LTR, resulting in the transcriptional inactivation of the provirus (89). This inactivation is essential in order to prevent production of transcripts antisense to the reporter-UTR mRNA directed by the strong regulatory elements located in the 5′LTR of wild-type provirus. Formation of double-stranded RNA could indeed complicate the functional analysis of HCRs by altering RNA stability or activating PKR or RNase L (24, 33, 43).
GUS was selected as the reporter gene because of its many substrates that allow a range of assays. The distribution of cells expressing the GUS enzyme can be analyzed in a transduced population, and live cells can be separated and isolated according to their GUS activity by FACS; GUS activity can be histochemically monitored at the single-cell level, and GUS protein can be quantitated by using a highly sensitive chemiluminescent enzyme assay. A further advantage of GUS is that endogenous GUS activity in most mammalian cells is low, especially at a neutral pH that is optimal for the bacterial but not for the mammalian enzyme (23). Finally, the GUS coding sequence is relatively small (1.8 kb), allowing a total of 2.5 kb of exogenous sequences to be cloned into the vector. This will accommodate essentially all HCRs, since they rarely exceed 2.0 kb (20).
To characterize the retroviral vector system shown in Fig. Fig.1B,1B, three HCRs from UTRs well known to alter mRNA stability and therefore protein accumulation were tested. In each vector tested, only the HCRs differed. Mouse embryonic fibroblasts (10T1/2) were transduced with retroviruses containing either no HCR or HCRs from the c-myc, c-fos, and TfR 3′UTRs (Fig. (Fig.1A).1A). Populations of thousands of cells obtained 1 week after transduction were analyzed for GUS expression by FACS. To control for potential differences due to copy number, these initial studies were carried out with populations which harbored a single integrant. This was achieved by transducing the cell populations at different viral titers and by using only those populations in which less than 20% of the cells expressed GUS as determined by FACS. According to the Poisson distribution, more than 96% of the cells should have only one copy of the retrovirus. For each of the three HCRs, the subpopulation of transduced cells expressing GUS above background levels was collected, expanded in culture, and analyzed again by FACS. The enrichment was successful, since the FACS plots revealed that essentially all of the cells analyzed expressed significant GUS activity. Similar results were achieved by selecting the transduced population with puromycin (results not shown). The shape of the plots, or the range of GUS expression, was similar for each of the HCR-expressing cell populations (Fig. (Fig.2A).2A). This range presumably reflects the random integration of the retrovirus in regions of the genome that differ in their transcriptional activities.
For each of the three specific HCR sequences, the peak or mean value of GUS activity was determined for the cell population (Fig. (Fig.2A).2A). The x axis depicts GUS activity on a logarithmic scale, and the shift of the peaks relative to the (−)HCR controls is due to the effect of the specific HCR sequences on the posttranscriptional regulation of expression of the GUS gene in the different cell populations. The c-fos HCR had the most marked effect, followed by c-myc and TfR. The mean value of the (−)HCR control cell population was 73, compared to 54 for the TfR HCR, 45 for the c-myc HCR, and 34 for the c-fos HCR. These data show that the FACS analysis is sufficiently sensitive to provide a rapid qualitative indication of the effects of a given HCR on protein expression levels in large polyclonal populations of cells.
A single-cell analysis of GUS activity by histochemical assay paralleled the FACS results, revealing a general range of activity in the cell population that was specific for each HCR. The c-fos HCR population had barely detectable GUS activity in this assay, followed by increasing amounts of blue staining in the c-myc HCR, TfR HCR, and (−)HCR cell populations, respectively (Fig. (Fig.2B),2B), thus corroborating the results of the FACS analyses. The range in GUS expression exhibited among individual cells of a given HCR-expressing population provided further evidence that the cell populations containing a single copy of the integrated retrovirus are polyclonal. Such heterogeneity of expression levels further indicate that the effect of an HCR on the posttranscriptional regulation of GUS expression cannot be accurately assessed by using cells derived from one or a few clones of stable integrants. These results underscore the need to study a polyclonal population of cells with a broad range of integration sites.
To analyze the effects of the HCRs on mRNA, the steady-state levels of expression of GUS mRNA in the three HCR-expressing cell populations were determined by Northern blotting (Fig. (Fig.2C).2C). The blot was hybridized simultaneously with two different digoxigenin-labeled RNA probes that detect the mRNA coding for GUS and the mRNA coding for the ubiquitous ribosomal protein L32 (rpL), which served as an internal control and allowed correction for RNA loading. In this assay, as in the previous two assays, the steady-state levels of GUS mRNA were most profoundly altered by the c-fos HCR, followed by the c-myc and TfR HCRs, which also led to reduced levels of GUS transcripts relative to controls.
Thus, as determined by FACS, histochemistry, and Northern blot analyses, all three HCRs led to significantly reduced amounts of reporter protein and mRNA levels. These findings confirm the effects of the 3′UTRs of these genes on mRNA destabilization reported by others (68), providing three positive controls that validate the vector system and method of analysis.
To analyze the effects of the seven “test” HCRs on mRNA, the steady-state levels of expression of GUS mRNA in the different HCR-expressing cell populations were first determined by Northern blotting. 10T1/2 fibroblasts were transduced with retroviral vectors containing either no HCR, an HCR from one of the three well-characterized 3′UTRs (c-fos, c-myc, and TfR) or one of the seven “test” HCRs with little or no previously documented function in posttranscriptional regulation. The 11 cell populations were sorted by FACS for GUS expression and expanded by growth in culture. To expedite and facilitate the analysis of several HCRs in parallel, cell populations were isolated irrespective of their transduction efficiencies.
In this case, to control for the number of integrants per cell for each cell population, we took advantage of the SV40puro transcription unit, precluding the need for a genomic Southern blot. Both GUS and puro transcription units are on the same vector. They differ only in that expression of the GUS transcription unit is affected by the HCR, whereas expression of the puro transcription unit is not. Thus, puro mRNA expression is proportional to the number of retroviral integrants.
The Northern blot (Fig. (Fig.3A)3A) was hybridized simultaneously with digoxigenin-labeled RNA probes specific to transcripts for GUS and puro. Accurate quantitation of the differences in chemiluminescent signals (Fig. (Fig.3)3) was made possible by using a highly sensitive luminometer that exhibits a linear dynamic range of over 1:10,000. This linear range, which is 100-fold greater than that obtained with X-ray films, allows a quantitation of a wide range of strong and weak signals at a single exposure time on the same blot. To determine the relative effects of the GUS-HCR mRNAs on mRNA steady-state levels in each cell population, the values obtained for GUS were divided by the values obtained for puro and expressed as a percentage of the values obtained for the (−)HCR population. Thus, the results obtained either by using puro to normalize multicopy integrants or by selection of single-copy integrants (Fig. (Fig.2C)2C) were similar, thereby validating the simpler, less labor-intensive multicopy approach used in all subsequent experiments.
The results shown in Fig. Fig.33 demonstrate the striking effects that HCRs can have on mRNA levels. The Ran HCR yielded mRNA steady-state levels that were almost comparable to the (−)HCR control population (90%). By contrast, the HCR of c-fos was the most potent and led to an accumulation of less than 10% of the levels of the control (−)HCR transcripts. The level of c-fos transcripts was readily quantitated with the luminometer, but it could not be imaged with the exposure parameters used to generate the print (Fig. (Fig.3).3). ODC, the second most potent HCR sequence in this assay, caused a reduction of GUS mRNA levels to 30% of the levels of the control (−)HCR transcripts. This effect of the ODC 3′UTR on mRNA steady-state levels has not previously been reported. The remaining five test HCRs—EF-1α, vimentin, fibronectin, bcl2, and HuD—all caused a 30 to 50% decrease in steady-state GUS mRNA levels.
In order to determine the effects of the HCRs on translation, we compared the steady-state levels of GUS mRNA (Fig. (Fig.4A,4A, upper panel) to the steady-state levels of GUS activity (Fig. (Fig.4A,4A, middle panel) for a number of HCR-expressing cell populations. In each assay, the HCRs that did not alter gene expression served to demonstrate that the observed effects were sequence specific. Thus, each experiment was internally controlled.
Most of the HCRs analyzed did not alter the translation efficiency of the GUS transcripts. This was clear from a comparison of the top panel, GUS mRNA steady-state levels, with the middle panel, GUS activity or protein accumulation. These measures paralleled one another in most cases, indicating that the amount of mRNA dictated the amount of protein. This is most apparent in the bottom panel of Fig. Fig.4A,4A, which shows the ratio of GUS activity to GUS mRNA. However, two HCRs are notable, as they led to marked and opposite effects, suggesting that they have an important role in regulating translation. The c-fos HCR decreased both mRNA and protein accumulation; however, the net effect on translation alone was a fivefold repression of GUS transcript translation. By contrast, the vimentin HCR significantly enhanced translation by twofold. These results demonstrate that specific HCRs can alter gene expression posttranscriptionally at the level of mRNA translation.
Since the vimentin HCR (Fig. (Fig.1A)1A) has two 80 to 90% conserved regions, we postulated that one of these two regions might suffice for the observed enhancement of translation. We therefore divided the vimentin HCR into two sequences, “vim a” and “vim b,” which overlapped by 20 bp, each of which contained one of the two conserved regions. We cloned and assayed these two regions separately and determined the GUS activity/GUS mRNA ratio for each transduced cell population. With either HCR fragment, the effect of the vimentin HCR on GUS mRNA translation was lost and the levels were characteristic of the (−)HCR control population. These results indicate that the vimentin HCR cannot be disrupted in this manner without altering its function.
We tested whether the HCRs could alter gene expression at the posttranscriptional level in response to a stress, such as a change in growth factor concentration in the culture medium. For that purpose, we grew the HCR-transduced cell populations in low-serum media for 24 h. Thereafter, half of the cultures remained in serum-poor media, whereas the other half were shifted to serum-rich media. We determined the GUS activity in the different HCR-expressing cell populations 48 h later (Fig. (Fig.5).5). Five of the HCR-expressing cell populations (c-fos, TfR, bcl2, EF-1α, and vimentin) showed no difference in response to changes in mitogen levels. By contrast, four of the HCRs (ODC, fibronectin, HuD, and Ran) responded to mitogen stimulation by inducing a 1.6- to 2.3-fold increase in GUS protein levels. Whether these changes were due to increased mRNA stability or translation is unclear, since mRNA levels were not studied in these experiments. Nonetheless, the data suggest that certain HCRs provide a posttranscriptional mechanism by which cells can significantly increase the levels of specific growth-related proteins in response to changes in mitogen levels.
We tested whether HCRs could alter gene expression in response to the stress induced by changes in oxygen (Fig. (Fig.6).6). Regions of tumors are often hypoxic and cells that are transformed adapt to such changes and continue to grow. Populations of cells expressing the (−)HCR, the c-fos HCR, or the bcl2 HCR were exposed to hypoxic conditions (4 ppm of O2) for 15 h and compared with controls exposed to the usual 21% O2 and 5% CO2. After 15 h of hypoxia, the c-fos HCR-expressing cell population exhibited a fourfold increase in GUS protein levels compared to the bcl2 HCR- or (−)HCR-expressing cells. Indeed, hypoxia had an opposite effect on bcl2 HCR or (−)HCR cultures, reducing GUS activity by 50%. This reduction in protein accumulation is therefore independent of the HCR and is probably due to a general effect of hypoxia on cell physiology or on the CMV minimal promoter that the c-fos HCR is able to override. However, whether these effects are at the level of mRNA stability or translation remains unknown since mRNA levels were not determined in these experiments. Taken together, the results show that in response to an environmental stress such as low oxygen, previously unrecognized HCRs such as that of c-fos can have marked posttranscriptional effects on gene expression that may provide an adaptive mechanism critical for cell survival.
The ability to modulate gene dosage by reducing or increasing HCR expression would enhance the study of HCR function, especially when HCRs have growth-inhibitory effects (64, 65). The Hermes HRSpuro-GUS vector used in these studies allows for this possibility, since seven copies of the tet operator (O7) precede the CMV minimal promoter (Fig. (Fig.1).1). Cells containing the vector without an HCR were superinfected with a second retroviral vector, RetroTet RTAb(+) or RetroTet RTRb(−), expressing either a tet transactivator (rtTA) (27) or transrepressor (tTR) (18), respectively. In the presence of dox, the tTR is sterically inhibited from binding the Tet operator adjacent to the CMV minimal promoter, and transcription is no longer repressed and proceeds at a basal level. In the absence of dox, tTR binds the tet operator and represses transcription. By contrast, the rtTA binds the tet operator in the presence of dox, stimulating transcription. In the absence of dox, the rtTA is inhibited from binding, and transcription remains at a basal level. The results (Fig. (Fig.7)7) show that the basal level of transcription used in all of the studies reported thus far can be decreased or increased at will in the same HCR-expressing cell population by altering the concentration of dox. By converting the system to a binary retroviral vector system, the effects of dosage of HCR containing mRNAs on cell proliferation and differentiation can be readily assessed.
The results presented here provide evidence that evolutionarily conserved sequences within the 3′UTR are good candidates for functional regulatory elements. Indeed, most of the HCRs studied here had a significant effect on gene expression at the posttranscriptional level. The approach we employed is not intended to be exhaustive; some important regulatory elements will be missed, since they are species specific and not conserved during evolution. However, this phylogenetic approach serves as a means of focusing on those sequences that are likely to have an important regulatory function. Clearly, certain nonconserved regions will also have posttranscriptional effects on gene expression, but by focusing on sequences that have not diverged, the search for such regulatory elements becomes less random. For example, of the 10 HCRs studied, each altered gene expression at the posttranscriptional level. Three were well characterized, and we were able to show that the HCRs had all of the functions previously ascribed to the 3′UTRs (c-fos, c-myc, and transferrin receptor), localizing the functional regions to those that had not diverged significantly from chicken to mouse. Conservation did not provide insight into the type of regulatory function, since the 10 HCRs had different effects. Indeed, only a subset of possible functions were tested here and other conditions may show as-yet-unidentified HCR effects. Our findings demonstrate that HCRs are potential regulatory regions that provide an important check-and-balance on transcription, both under steady-state conditions and especially in response to environmental stresses such as changes in mitogen levels or oxygen deprivation.
Three HCRs were chosen to validate our method of analysis which we then extended to the study of seven relatively uncharacterized HCRs. These HCRs were selected for study because of their evolutionary conservation and because the protein products encoded by the mRNAs with which they are associated have known functions in growth control, differentiation, and cancer. For all three well-characterized HCRs (c-fos, c-myc, and TfR), the reported effects of the 3′UTR on mRNA destabilization were demonstrated. The system was then used to compare and rank the effects of the diverse HCRs on posttranscriptional regulation. This was possible because all of the other vector components, such as the CMV minimal promoter, 5′UTR, GUS reporter, and polyadenylation signal, remained constant.
Five of the ten HCRs tested led to a two- to tenfold reduction in mRNA steady-state levels. With the notable exception of the 3′UTR of α-globin, which increases mRNA stability (85), 3′UTR regulatory sequences have generally been proposed to reduce mRNA stability, which is in agreement with our findings (68). Thus, HCRs may provide a safeguard against overexpression of proteins with a role in cell growth control and differentiation, thus maintaining a critical balance of such proteins.
The observed reduction in certain mRNA steady-state levels could be due in part to specific sequence motifs, such as the well-defined AU-rich elements (AUREs) (8). Of interest is the finding that HCRs located in 3′UTRs are generally AU-rich (particularly U-rich) relative to the non-HCR region of the 3′UTRs (20). These sequences are known to be present in several unstable mammalian oncogene and lymphokine mRNAs (7), as well as in some of the HCRs tested here (c-fos, c-myc, and bcl2). However, since the degree of instability conferred by those HCRs that harbor AUREs differs, context is likely to play a major role in their function. In fact, all five of the HCRs studied here that contain AUREs extend beyond those sequences and include additional sequences that are highly conserved across species. Such domains could well include other destabilizing motifs or constitute binding sites for modulators of the AURE-specific degradation machinery. Indeed, the c-fos HCR, which exhibited the greatest degree of destabilization by far, is more than threefold the size of the three consecutive AU-rich domains encompassing the AURE; in addition it contains a 20-nucleotide U-rich sequence that has been shown previously to play a role in the deadenylation of the message leading to degradation (86, 88). Regulation of polyadenylation is thought to play an essential role in mammalian mRNA decay as in yeast cells (16) and can be initiated by shortening the poly(A) tail, followed by decapping and 5′- and 3′-exonucleolytic degradation of the transcript (79). Although it is clear that the number, spacing, and conserved sequences flanking AUREs can all affect their destabilization potential, the rules that govern these effects and the roles of independent but synergistic domains remain unknown. The context within a given mRNA is important not only to AURE function but also to cell physiology. For example, when primary T cells are activated by exposure to antibodies to the CD3 and CD28 receptors, several AU-rich mRNAs, including granulocyte-macrophage colony-stimulating factor, interferon, and interleukin-2, are stabilized, whereas the AU-rich c-myc mRNA is not (50). Although the c-fos, c-myc, and transferrin 3′UTR sequences are all well-known to alter mRNA stability, the remaining 3′UTR sequences could alter mRNA steady-state levels by other mechanisms. The Hermes HRSpuro-GUS retroviral vector described here will allow the rapid analysis of mutant and truncated HCRs and HCRs located 3′ and 5′ to the reporter. It should prove useful in comparing diverse HCRs under a range of conditions and in the elucidation of the properties critical to AURE function in mammalian cells.
A second well-documented mechanism that leads to mRNA decay is endonucleolytic cleavage within the transcript. Examples include the destabilization of mRNAs encoding TfR and insulin-like growth factor II (55), both of which contain stem- loop structures in the 3′UTR. In the case of TfR, the HCR contains five stem-loop structures (A to E) designated as iron-responsive elements (IREs). These structures bind transacting proteins (iron regulatory proteins) that mask the endonucleolytic cleavage site present between IREs C and D, thereby stabilizing the mRNA (3). Whether the presumed destabilization induced by the remaining HCRs studied here is determined by sequences that decrease the binding of poly(A) binding protein, decrease poly(A) length, or provide sites for endonucleolytic cleavage remains to be determined, and the approach described here will enable such an analysis.
The amount of protein produced by translation of an mRNA is regulated at multiple levels (24, 33, 43). The mRNA contains regulatory elements that interact with transacting factors that modulate translation initiation, elongation, and termination. The rate of initiation is known to be strongly influenced by certain sequences (61) or secondary structures in the 5′UTR of mRNAs, as in the case of ODC (31, 53) or ferritin (30). In contrast to 5′UTR-mediated control, the mechanisms by which 3′UTRs influence the translation process are understood at a more rudimentary level. Recent studies in yeast cells suggest that 3′UTRs may alter translation by a looping of the mRNA via the binding of the poly(A) tail and its binding protein (PABP) to the initiation factor eIF4G, which is part of the cap-binding protein complex eIF4F (77, 78). Biochemical evidence suggests that a similar interaction may occur in mammals (46). Such an approximation of the terminal portion of the 3′UTR with the 5′ end of the mRNA could explain both the negative and positive effects of 3′UTRs on translation efficiency. Alternatively, the 3′UTR could be bound by proteins that lead to the sequestration of the mRNA into an untranslatable mRNP particle (73). The best-documented cases for translational control via 3′UTRs are the gradients of regulatory molecules that lead to pattern formation in developing Drosophila embryos (13) and the temporal control of expression of erythroid 15-lipoxygenase mRNA in mammals (60).
Only two of the 3′UTR sequences tested here had marked effects on translation efficiency. The c-fos HCR repressed translation fivefold. The effect of the c-fos 3′UTR on translation repression had been previously noted in nondividing Xenopus oocytes (45) but not in mammalian proliferative somatic cells. Most 3′UTR sequences, like the tra-2 and GLI element (TGE) within the GLI 3′UTR (40), repress translation. The vimentin HCR sequence enhanced translation twofold. With the possible exception of the amyloid protein precursor mRNA (17), reports of 3′UTRs that promote translation are rare. To test whether the findings (90) of a Y-shaped secondary structure localized within the “vim a” domain (Fig. (Fig.4)4) might mediate the effects of the vimentin HCR on translation, we divided the HCR in half and introduced each half separately into different cell populations. The stimulation of translation of each half was reduced to control (−)HCR levels, indicating that the structure described by Zehner and colleagues does not confer the observed translational induction. Of note is the finding that the HCR of vimentin has a dual effect. This HCR not only stimulates translation twofold but also decreases mRNA levels to 70% of the control levels. The relative roles of these two opposing HCR-mediated mechanisms of controlling mRNA stability and translation could change in response to environmental cues and play a critical adaptive role, in agreement with proposed models that invoke coupling of mRNA stability to translation (80).
Cells are known to regulate the expression of different genes in response to changing environmental conditions, such as nutrient or oxygen supply. Our data indicate that HCRs act as sensors and can alter gene expression in response to such stresses at a posttranscriptional level. Four HCRs (Ran, ODC, fibronectin, and HuD) responded to an increase in mitogens by increasing protein levels 1.5- to 2-fold. Here we show that the ODC HCR can alter gene expression independently of 5′ or coding components of the endogenous mRNA. Therefore, the translational stimulatory effect of the HCR is evident even in the absence of the endogenous 5′ UTR of ODC, which is known to contain an extensive secondary structure that represses the translation of ODC, an effect partially relieved by its 3′UTR (31, 51, 53). Although increases in proteins in response to serum stimulation have also been previously reported for Ran (11) and fibronectin (62), evidence that these increases could be controlled at a posttranscriptional level as shown here has not been apparent. HuD has not previously been shown to be regulated at a posttranscriptional level. Whether regulation occurs at the level of mRNA stabilization or translation remains to be determined. Nonetheless, these findings demonstrate that HCRs can be sensors of mitogen concentrations, leading to altered protein levels that may be essential to cell survival.
Hypoxia and reoxygenation often accompany injury, ischemia, and stroke. In addition, evidence is also accumulating that tumor hypoxia plays an integral role in the malignant progression of cancers (29). Solid tumors typically have regions that are necrotic, and this can be accompanied by perinecrotic hypoxia. The expression of many genes is altered at the transcriptional level in response to hypoxia, and this regulation is mediated in part by the heterodimeric transcription factor, hypoxia-inducible factor 1 (HIF-1) (28, 83). Other transcription factors have been implicated in the hypoxia response, such as c-fos, which together with c-jun presumably acts in the AP-1 transcription complex, which has been shown to be partially responsible for the expression of tumor metalloproteases stromolysin or type 1 collagenase (12). We show here that the induction of c-fos by hypoxia is regulated not only at the transcriptional level (84) but also at the posttranscriptional level by the HCR in the 3′UTR. This mode of c-fos regulation has not been previously reported. It will be of interest to determine whether the underlying posttranscriptional regulatory mechanisms are similar to the ones described for the hypoxic induction of VEGF (49) or erythropoetin (54). The importance of this class of posttranscriptional regulation of genes in response to hypoxia has become apparent because the loss of the tumor suppressor VHL results in the loss of this type of regulation (25). Given the magnitude of the effect we observe, additional hypoxia-responsive HCRs could be identified by functional genomic screening of retroviral cDNA libraries and FACS analysis (34, 42). Such HCRs may provide a molecular switch that responds to the inhibitory conditions in the microenvironment of solid tumors.
Finally, the modulation of levels of 3′UTR expression through the use of regulator retroviruses such as RetroTet RTAb(+) and RetroTet RTRb(−) will now facilitate the in-depth analysis of 3′UTR sequences with a role in growth inhibition and differentiation, such as those previously described (15, 47, 64, 65), since expression can be suppressed during cell expansion and induced specifically at the time of analysis. Moreover, our inducible retroviral system should allow the study of mRNA decay kinetics without perturbing cellular physiology, as is the case for transcription inhibitors such as actinomycin D or inducible systems based on transient expression of the c-fos promoter after serum stimulation. The inducible expression of HCRs will facilitate control of the concentration of HCR-containing mRNA molecules in the cell by varying the amount of Tet in the culture medium (44). An excess of exogeneous HCR molecules could titrate out UTR regulatory binding proteins and modify the steady-state level of expression of the endogenous gene as reported for creatine kinase B (9) and ODC (51), which could in turn lead to a pleiotropic effect on gene expression and the consequent alteration of cell physiology. The use of the reporter retroviral vector used here should diminish the risks of overexpressing the HCR-containing reporter compared to transient-transfection experiments, due to the low copy number of transgenes introduced and the use of a minimal promoter. Moreover, superinfection with the Tet-regulatable transrepressor retrovirus can be used to further decrease transcription and thus HCR-mRNA dosage.
The findings described here may have applications to the treatment of viral and malignant diseases. Tet-inducible overexpression of exogenous HCR sequences could provide a means to alter the balance of genes involved in growth control or hypoxia. Posttranscriptionally mediated therapies could be designed that mimic mechanisms used by viruses. For example, competition between the c-fos 3′UTR instability elements and the papillomavirus late mRNAs for the same poly(U) binding proteins has been postulated to lead to elevated Fos protein levels in infected cells (71). Thus, HCR expression in a time- and dose-dependent manner could be useful as an adjunct to traditional antiviral and cytostatic agents.
We thank our colleagues for critiquing the manuscript, Bruce Blakely for expert assistance in the preparation of this work for publication, and Najja Bracey and Dan Spiegel for technical assistance.
This work was supported by a postdoctoral fellowship from the Swiss National Science Foundation (823A-46704) to A.S., a grant from the CNRS (Centre National de la Recherche Scientifique) to L.D., a collaborative grant from NATO (CGR 971161) to L.D. and H.M.B., a postdoctoral training grant 5T32CA09302 to N.C.D. from the NIH grant CA73832, and grants from the NIH (AG09521, CA59717, and HD18179) to H.M.B.
A.S. and O.M.G. contributed equally to this work.