Translational Control by an Upstream Open Reading Frame in the HER-2/neu Transcript*

Overexpression of the HER-2 (neu,erbB-2) receptor results in cellular transformation and is associated with a variety of human cancers. Multiple mechanisms, including gene amplification and transcriptional, post-transcriptional, and translational controls contribute to the regulation of HER-2 expression. One of the components of these regulatory mechanisms is a short upstream open reading frame (uORF) in the HER-2 mRNA that represses downstream translation in a variety of cell types. Here we explore the mechanism by which this uORF exerts its inhibitory effect. As judged by comparisons of protein and mRNA abundance and by polysomal distribution analyses, the uORF represses translation of the HER-2 cistron or of a heterologous reporter gene. Despite its conservation among mammalian species, the peptide sequence of the uORF is not required for this inhibitory effect. Rather, the majority of ribosomes that load on the HER-2 mRNA most likely translate the uORF and are then unable to reinitiate at the downstream AUG codon, in part due to the short intercistronic spacing. A minority of ribosomes gain access to the HER-2 initiation codon either by leaky scanning past the upstream AUG codon or by reinitiating after having translated the uORF despite the short intercistronic region. These results suggest that the HER-2 uORF controls synthesis of this oncoprotein by limiting ribosomal access to downstream initiation sites.

The HER-2 (neu, erbB-2) oncogene encodes a 185-kDa transmembrane receptor tyrosine kinase (1)(2)(3)(4). Although HER-2 is involved in normal development as evidenced by neural and myocardial defects in knock-out mice (5), most studies of HER-2 have focused on its role in cancer. Overexpression of HER2 occurs in numerous types of human cancers and has been linked to neoplastic transformation and aggressive tumor growth (6 -12). Cells from tumors in which HER-2 is overexpressed often contain amplified copies of the HER-2 gene. However, in some cases HER-2 overexpression is due to transcriptional and post-transcriptional mechanisms in the absence of gene amplification (12)(13)(14)(15). Moreover, under certain condi-tions, HER-2 receptor levels vary without changes in mRNA levels, suggesting that translational controls also participate in the control of HER-2 protein synthesis (16,17).
In eukaryotes, translational regulation of specific genes typically occurs at translational initiation and is mediated by cis-acting sequences present in the 5Ј transcript leader, such as upstream AUG codons and associated upstream open reading frames (uORFs 1 (18,19)). Although uORFs are found in only 5 to 10 percent of eukaryotic mRNAs overall, approximately twothirds of oncogenes including HER-2 and many genes involved in cellular growth and differentiation contain uORFs (20). In the well studied case of the Saccharomyces cerevisiae GCN4 gene, uORFs regulate protein synthesis by affecting which downstream AUG codons are utilized by reinitiating ribosomes (21). Ribosomes translate the first uORF in the GCN4 mRNA under all conditions. They then reinitiate at another uORF when amino acids are plentiful or, under starvation conditions, they bypass the other uORFs and reinitiate at the GCN4 start codon. Several other uORFs have been shown to act by a mechanism that depends on the uORF-encoded peptide sequence and, in some cases, involves ribosomal stalling on the mRNA (19,(22)(23)(24)(25)(26)(27)(28). For the vast majority of uORFs, insufficient data are available to enable predictions about whether they affect downstream translation and, if so, about the mechanism involved.
Previously, we demonstrated that two distinct translational mechanisms control HER-2 protein expression (29). One is a cell type-dependent mechanism that causes increased HER-2 translation in transformed cells compared with primary cells. The other is a cell type-independent repression of downstream translation mediated by an upstream AUG codon. The upstream AUG codon is in an optimal Kozak context (30) and initiates a six-codon uORF that terminates five nt from the HER-2 start codon (see Fig. 1A). Mutation of the upstream AUG codon eliminates the uORF and results in an approximately 5-fold increase in downstream translation in each of five cell types examined (29). The position and coding content of the uORF are highly conserved among mammalian species (29,(31)(32)(33)(34)(35), suggesting that these features may be important for the regulatory effects of the uORF.
In the current report, we present an analysis of the translational mechanism by which the HER-2 uORF affects downstream translation. Our results demonstrate that the HER-2 uORF inhibitory function does not depend on the peptide sequence of the uORF, the identity of the downstream cistron, or the precise 5Ј end of the mRNA. Instead, the very short intercistronic spacing between the uORF and the downstream cistron appears to be required for its inhibitory effect. Despite the optimal context of the upstream AUG codon and the short intercistronic spacing, both leaky scanning and ribosomal reinitiation after translation of the uORF contribute to HER-2 protein synthesis. These observations suggest that mammalian uORFs, like those in yeast, may control access of ribosomes to alternative downstream initiation sites.
To construct the HER-2 expression plasmids, the HER-2 ORF was isolated from SV40/erbB2 (6) (provided by S. Aaronson, National Institutes of Health) by digesting with Bsu36I, blunting with DNA polymerase (Klenow), then digesting with HindIII and inserting the HER-2coding region into the HindIII/HindII sites of pBSϩ (Stratagene). The HER-2 ORF, isolated from this plasmid by digestion with XbaI, blunting with DNA polymerase (Klenow), and cutting with HindIII, was ligated into the HindIII/PvuII sites of pEQ176. The resulting plasmid, pEQ580, contains 22 nt of the HER-2 leader immediately upstream from the HER-2 AUG codon but does not contain the upstream AUG codon. The HER-2 transcript leader from pEQ516, isolated by HindIII/ partial NcoI digestion, was inserted into the HindIII/NcoI sites of pEQ580 to generate pEQ582. Plasmid pEQ581 is identical to pEQ582 except that the uORF AUG codon has been mutated to AAG.
The transcript leaders from pEQ516 and pEQ471 were PCR-amplified with oligos #48 and #61 (GGAAGGTACCATGGTCTTAAGCT-CACTGCGG) to introduce an AflII site just downstream from the uORF. The resulting fragments were cloned as HindIII/Asp718 fragments into pEQ176, yielding pEQ526, the wt construct, and pEQ485, the corresponding AAG mutant. A 50-nt intercistronic spacer consisting of cytomegalovirus UL4 transcript leader sequences, derived by cutting pEQ239 (36) with SpeI, blunting with DNA polymerase (Klenow), and digesting with Asp718, was inserted into pEQ526 and pEQ485 that had been digested with AflII, blunted with DNA polymerase (Klenow), and cut with Asp718, yielding pEQ717 and pEQ718, respectively. pEQ719 and pEQ720 were generated by inserting a UL4 RsaI-Asp718 fragment from pEQ239 into pEQ526 and pEQ485 that had been digested with SpeI, blunted, and digested with Asp718. A fragment generated by PCR amplification of the UL4 transcript leader with primers gp48.3 (36) and #105 (CGGCCTTAAGTGAAGAGTCTATAAAG) and digestion with Afl2/Asp718 was inserted into pEQ526 and pEQ485 to generate pEQ608 and pEQ607, respectively.
A blunted BglII/SalI fragment derived from pM128 (provided by A. Hinnebusch, National Institutes of Health) containing the 3Ј-most 152 nt of the S. cerevisiae GCN4 transcript leader (37) was cloned into pEQ526 that had been cut with AflII and blunted. Plasmid pEQ741 contains this sequence in the same orientation as found in the GCN4 mRNA. The same fragment was cloned into pEQ485 to yield the corresponding AAG mutant pEQ743. The transcript leader and the 5Ј end of the HER-2 coding region isolated as HindIII, blunted-ApaLI fragments from pEQ578 and pEQ577 were inserted into pEQ176 that had been digested with XhoI, blunted, then digested with HindIII to generate pEQ673 containing the wt HER-2 leader and the corresponding AAG mutant pEQ674. pEQ573 was constructed by PCR amplification of pEQ516 with oligos #48 (see above) and #83 (GGAAGGTACCATGGTGCTCCCTGCGGC) to eliminate the uORF stop codon. The resulting fragment was digested with HindIII and Asp718 and cloned into the corresponding sites in pEQ176. The product of PCR amplification with oligos #48 and #99 (GGAAGGTACCATGGTGCTCACTGCGGCTCCGGCCCCATGGTGG-CGGCTGGACCC), having a super optimal upstream AUG codon, was cloned into pEQ176 as a HindIII/Asp718 fragment to produce pEQ559. Plasmid pEQ592, the corresponding AAG construct, was produced in the same manner using oligos #48 and #103 (GGAAGGTACCATGGT-GCTCACTGCGGCTCCGGCCCCTTGGTGGCGGCTGGACCC). pEQ-739, in which the uORF contains a super optimal AUG codon and a mutated stop codon, was produced in the same way by PCR amplification with oligos #48 and #83 using pEQ559 as a template. pEQ751, containing the full-length transcript leader with the HER-2 uORF fused to lacZ, was constructed by PCR amplification with oligos #106 (GGCCAAGCTTATTCCCCTCCATT GGGACCGGAG) and #163 (GGAAGGTACCAAGGATGCTCCCTGCGGC) using pEQ637 as a template. The insert was digested with HindIII and Asp718 and cloned into the same sites in pEQ176. pEQ752, the corresponding uORF AAG mutant was made by the same strategy using pEQ655 as the template.
Cell Culture, Transfection, and RNA Analyses-COS-7 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% NuSerum (Collaborative Biomedical). 2 g each of a test ␤-gal expression plasmid and the pEQ430 control were transfected into COS-7 cells in triplicate 60-mm dishes using calcium phosphate (29). 48 h post-transfection ␤-gal activity was measured by a fluorimetric substrate cleavage assay (38), then whole cell RNA was harvested by the acid guanidinium isothiocyanate method (39) and analyzed by Northern blot hybridization with a ␤-gal probe. Polysomes from COS-7 cells transfected with HER-2 expression plasmids were separated on 15-50% sucrose gradients, and the RNA content of each fraction was analyzed by Northern hybridization as described (36) Immunoblot Analysis-At 48 h post-transfection, cells were washed with phosphate-buffered saline then lysed with 2% SDS at 65°C. The resulting cell lysates were denatured at 95-100°C for 5 min and then electrophoresed through 7.5% SDS-polyacrylamide gels, and the proteins were transferred to polyvinylidene difluoride transfer membrane (TROPIX, Inc.) by electroblotting. Immunoblot analysis was carried out according to the manufacturer's recommendations using the Western-Light Plus chemiluminescent detection system (TROPIX, Inc.) with rabbit polyclonal serum directed against the 14 carboxyl-terminal amino acids of HER-2. Whole cell RNA was harvested from parallel dishes as described above and analyzed by Northern blot analyses using a HER-2 extracellular domain fragment probe (29).

Inhibition of HER-2 Translation by the uORF-Previous
studies revealed that the HER-2 uORF inhibits translation of a downstream reporter gene (29). To determine whether it also inhibits expression of the authentic HER-2 protein, we constructed plasmids having the HER-2 ORF downstream from transcript leader sequences having or lacking the uORF (Fig.  1A). After transfection of these plasmids into COS-7 cells, HER-2 protein levels and mRNA accumulation were measured by immunoblot and Northern blot analyses as described under "Experimental Procedures." The wild-type transcript leader containing the uORF repressed HER-2 protein expression compared with the leader containing a mutation in the upstream AUG codon or to the control containing only a very short leader (Fig. 1B, compare WT with AAG and SL lanes). mRNA levels were similar among all constructs, indicating that differences in protein expression did not result from variation in transcript accumulation. The low level of expression of HER-2 protein and RNA in mock-transfected cells suggested that most of the protein and RNA detected in the other samples represented products of the transgenes rather than the endogenous gene. Nonetheless, we confirmed these results using FLAG-tagged HER-2 expression plasmids with which we could unambiguously detect transgene expression (data not shown).
The transcript leader sequences in pEQ582 contained 93 nt, including the uORF, from the 3Ј end of natural HER-2 transcript leader. Although all reported HER-2 sequences share an identical 96 nt at the 3Ј end of the transcript leader, some reports have suggested that the 5Ј end of the leader may contain alternative sequences (2,3,31,32). In experiments using either ␤-gal or HER-2 expression plasmids, we found that the uORF had quantitatively similar effects on downstream translation whether it was contained in the full-length 178-nt leader (29) or only within the conserved 93-nt 3Ј region (data not shown).
We also analyzed the effects of the uORF on HER-2 translation by examining the polysomal association of mRNAs having and lacking the uORF (Fig. 2). Polysomes in cells transfected with pEQ582 (AUG) and pEQ581 (AAG) were fractionated on sucrose gradients, and transgene mRNAs in each fraction were detected by Northern blot hybridization using a probe specific for the transgene 3Ј-untranslated region that is not contained in endogenous HER-2 transcripts. Elimination of the uORF resulted in a shift of the HER2 mRNA to larger polysomes, corresponding to more efficient translation. The mean position of wild-type mRNAs was fraction 3, corresponding to disomes, whereas that for the AAG construct was fraction 5, corresponding to approximately 6-and 7-mers. Similarities of the UV absorption profiles (not shown) and of the distribution of actin mRNA (Fig. 2) between the two samples indicated that the shift to larger polysomes, resulting from elimination of the HER-2 uORF, was not an artifact of variation between the two gradients. These results demonstrate that uORF represses HER-2 expression by reducing the ribosomal loading on the mRNA and thus verify that the uORF inhibits HER-2 expression at the translational level.
Inhibition by the HER-2 uORF Is Peptide Sequence-independent-We next investigated the mechanism by which the HER-2 uORF exerts its repressive effect. Inhibition by uORFs in several other eukaryotic genes is dependent upon the peptide-coding sequence of the uORF (19,26,40). Because the HER-2 uORF sequence is conserved among mammalian species (29,(31)(32)(33)(34)(35), we tested whether it is required for inhibition of downstream translation. ␤-Gal expression constructs in which the uORF was modified by shifting the reading frame to generate a different amino acid sequence while preserving most of the nucleotide sequence (pEQ591) or by random mutagenesis with a degenerate oligonucleotide (pEQ721, -722, -723) were transfected into COS-7 cells (Fig. 3). ␤-Gal activity and mRNA accumulation were analyzed as described under "Experimental Procedures." Like the wt uORF, each of these mutant uORFs inhibited translation of the downstream ␤-gal gene, demonstrating that the HER-2 uORF functions in a peptide sequenceindependent manner.
Effect of Intercistronic Spacing on Repression of Downstream Translation-Another feature of the HER-2 uORF that might account for its inhibitory effect is the proximity of its termination codon to the initiation codon of the downstream cistron. In all mammalian species in which the sequence has been reported, this intercistronic spacing is only five nt. To test the role of this spacing on the inhibitory activity of the HER-2 uORF, we lengthened the intercistronic distance in our ␤-gal reporter construct by inserting various fragments derived from the cytomegalovirus UL4 transcript leader. The fragments used to construct these plasmids contain two adjacent AUG codons, either of which can serve as initiation sites for ␤-gal synthesis. The size of the intercistronic spacer shown in Fig. 4 for pEQ717, pEQ719, and pEQ608 is based on the assumption that initiation occurs at the second of these two AUG codons. If the first one is used, then the actual intercistronic spacing would be three nt shorter. These spacers do not contain other AUG codons and, in other experiments, did not affect down- stream reporter gene translation (36). Nonetheless, we constructed control plasmids containing the same spacer sequences but having a mutation of the AUG codon of the uORF to detect unexpected uORF-independent effects of the spacer sequences. These plasmids were transfected into COS-7 cells and analyzed as described under "Experimental Procedures" (Fig. 4). Expansion of the intercistronic spacing to 10 nt (pEQ526) had little effect on downstream translation compared with the wt spacing. However, expansion to 50, 116, or 148 nt increased ␤-gal expression approximately 2-fold. The sequences used to expand the intercistronic spacing also inhibited expression from the controls lacking the uORF. This reduction may be due to the modest reduction in ␤-gal RNA accumulation after transfection of constructs having the longer insertions (Fig. 4, right panel). The inhibitory effect of the uORF, when measured as the ratio of expression from the AUGto the corresponding AUG ϩ plasmid, decreased from 7-fold with the wt spacing to 1.4-fold when the spacing was 148 nt. These results suggest that repression by the HER-2 uORF is alleviated, at least in part, by increasing the intercistronic spacing.
In addition to effects due to intercistronic length, the sequence of the intercistronic region may affect reinitiation fre-quency (41,42). To further examine the role of intercistronic sequence on translational inhibition, we tested the effects of a second spacer sequence derived from a portion of the S. cerevisiae GCN4 transcript leader that has no upstream AUG codons or other known translational regulatory elements. Constructs containing the uORF or the upstream AUGmutation with an intercistronic spacing of 171 nt were transfected into COS-7 cells, and ␤-gal expression was analyzed (Fig. 5). For unknown reasons, the GCN4 spacer greatly reduced the abundance of reporter gene transcript accumulation from pEQ741 and pEQ743 compared with plasmids lacking the GCN4 sequences. Nonetheless, this intercistronic spacer also reduced the inhibitory effect of the uORF to 1.6-fold, similar to the results using the CMV UL4 spacer sequences (Fig. 4).
To evaluate the reinitiation potential of ribosomes that have translated the authentic uORF, we constructed plasmids more closely resembling the structure of the natural HER-2 mRNA. In pEQ673, the ␤-gal ORF initiates 92 nt downstream from the uORF termination codon. The "intercistronic" spacer in this case includes the five nt between the uORF and the HER-2 AUG codon followed by the first 79 nt of the HER-2 ORF and 7 nt of polylinker-derived sequences. In the natural HER-2

FIG. 4. Translational repression by the HER-2 uORF is spacing-dependent.
Plasmids containing either the wt HER-2 uORF or a mutation of the upstream AUG codon to AAG and having varying lengths of a translationally neutral CMV UL4 spacer sequence between the HER-2 uORF and the downstream ␤-gal gene (pEQ516, 471 ϭ 5 nt; pEQ526, 485 ϭ 10 nt; pEQ717, 718 ϭ 50 nt; pEQ719, 720 ϭ 116 nt; and pEQ608, 607 ϭ 148 nt) were transfected into COS-7 cells and analyzed as described in the legend to Fig. 3. Fold inhibition by the uORF was calculated as the ratio of ␤-gal expression from each AUGconstruct to that from the corresponding AUG ϩ construct.

FIG. 3. Translational inhibition by the HER-2 uORF is sequence-independent.
Expression constructs containing the truncated wt (pEQ516) and AAG mutant (pEQ471) transcript leaders or vectors in which the sequence of the HER-2 uORF were modified by frameshift mutations (pEQ591) or random mutagenesis with degenerate oligonucleotides (pEQ721-723) were transfected into COS-7 cells. Controls (light bars) include pEQ176, which expresses ␤-gal with no HER-2 leader sequences (pEQ176), and a truncated enzymatically inactive ␤-gal (pEQ430), which was included in each sample as a control for transfection efficiency and RNA recovery. At 48 h post-transfection, ␤-gal activity and whole cell RNA was harvested for analysis of accumulated ␤-gal mRNA levels. The means and S.D. of ␤-gal activities from triplicate dishes are shown. mRNA there are two AUG codons, positioned 96 and 138 nt downstream from the end of the uORF that encode methionines within the HER-2 extracellular domain. Thus, the ␤-gal AUG codon in pEQ673 is the third AUG codon from the 5Ј end and is in a position closely approximating that of the first of the two in-frame internal AUG codons present in the authentic HER-2 mRNA. Transfection assays of pEQ673 revealed only a low level of ␤-gal (ϳ4%) compared with the control having no HER-2 leader. However, the effect of the uORF was to increase ␤-gal expression approximately 2-fold compared with the corresponding upstream AUGmutant pEQ674 (Fig. 5). Together with the results using heterologous intercistronic spacers ( Fig. 4 and 5), these data suggest that some ribosomes that translate the uORF are able to reinitiate only after having traversed ϳ50 or more nucleotides downstream from the uORF termination codon.
Translational Initiation at the HER-2 AUG Codon-Since the AUG codon of the HER-2 uORF is in a very good context for initiation (gccAUGg), we expected that most ribosomes that load onto the HER-2 mRNA would initiate at this AUG codon. Because of the short intercistronic spacing, these ribosomes would not be expected to reinitiate efficiently at the HER-2 AUG codon. These considerations raise the question of how HER-2 is ever translated. One possibility is that some ribosomes leak past the upstream AUG codon despite its context. Alternatively, a few ribosomes that translate the uORF may reinitiate at the HER-2 AUG codon despite the short intercistronic spacing. To evaluate these possibilities we first mutated the stop codon of the uORF (TGA 3 GGA; pEQ573), creating an extended uORF that terminates at the next in-frame stop codon, 41 nt downstream from the ␤-gal initiation codon. This mutation greatly reduces the possibility that ␤-gal can be made by reinitiation after translation of the uORF. As shown in Fig.  6, deletion of the stop codon reduced ␤-gal expression by approximately half (compare pEQ573 to pEQ516), suggesting that a portion of expression downstream from the HER-2 uORF occurred by ribosomal reinitiation. The residual ␤-gal expression from this construct could be due to leaky scanning (43) or, conceivably, backward scanning after translation of the extended uORF. Although backward scanning has been reported (44), our previous studies of ribosomes terminating in this region suggested that backward scanning, if it occurs at all, is very inefficient (45).
To further evaluate the contribution of leaky scanning, we mutated the context of the uORF AUG to gccgccaccAUGg, a sequence shown by Kozak to yield maximum initiation frequency in higher eukaryotes (46,47). This "super" optimal AUG context mutation (pEQ559) reduced ␤-gal expression by approximately 50%, supporting the conclusion that some leaky scanning occurs at the wt AUG codon despite its excellent context. Mutation of the uORF AUG codon in the super optimal context to AAG (pEQ592) eliminated the repression, confirming that the observed results were due to a translational effect of the uORF. Combining the stop codon and super optimal AUG codon context mutations (pEQ739) reduced ␤-gal expression to about half that for either single mutation. Although we do not know how the small amount of ␤-gal that is synthesized from this mutant is made, these data support the conclusion that both leaky scanning and ribosomal reinitiation contribute to HER-2 expression.
Translation Initiation at the uORF AUG Codon-Previous results indirectly suggest that the HER-2 uORF is translated.
To determine more directly whether uORF translation occurs, we constructed a plasmid in which the HER-2 uORF was fused in-frame to the ␤-gal ORF (pEQ751). In this plasmid the uORF stop codon and the ␤-gal AUG codon were mutated, and a single nucleotide was inserted to fuse the uORF to the ␤-gal ORF such that ␤-gal synthesis should only occur if ribosomes initiate translation at the uORF AUG codon. pEQ751 and its AUGderivative (pEQ752) were transfected into COS-7 cells, and ␤-gal activity and mRNA accumulation were analyzed (Fig.  7). The high level of ␤-gal activity from pEQ751 (AUGϩ) compared with the background level expressed by pEQ752 (AUG -) confirms that the HER-2 uORF AUG codon is utilized as a translational initiation site in this mRNA.

FIG. 5. Effect of alternative intercistronic spacers on translational repression by the HER-2 uORF.
To determine the effects of other spacer sequences, two sets of constructs with either a 171-nt GCN4 spacer (pEQ741, 743 ϭ AUGϩ/Ϫ) or a portion of the HER-2 gene with authentic uORF-HER-2 spacing, an intact HER-2 start codon, and the 5Ј-most 79 nt of the HER-2 ORF (pEQ673, 674 ϭ AUGϩ/Ϫ) were produced, transfected into COS-7 cells, and analyzed as described in the legend to Fig. 3.   FIG. 6. Analysis of the mechanism by which expression downstream of the HER-2 uORF occurs. Expression constructs with various mutations affecting the AUG context and stop codon of the HER-2 uORF were utilized to study their effects on translation of the downstream ␤-gal gene. Plasmids contained the wt (pEQ516; start codon consensus ϭ gccggagccATGg) and AAG mutant (pEQ471) transcript leaders, a mutation of the uORF stop codon from TGA to GGA (pEQ573), an improved consensus (gccgccaccATGg) start site (pEQ559, AUG ϩ ;pEQ592 AUG -), and a combined improved start codon consensus-stop codon mutant (pEQ739). The constructs were transfected into COS-7 cells and analyzed as described in the legend to Fig. 3.

DISCUSSION
Although HER-2 overexpression in tumor cells is often attributed to gene amplification, transcriptional, post-transcriptional, and translational mechanisms also contribute to the regulation of HER-2 expression (12)(13)(14)(15)(16)(17). Previously, we reported that translation of HER-2 mRNA is repressed in primary cells compared with transformed cells as measured by polysome distribution analyses (29). However, even in the transformed cells, the efficiency of HER-2 translation is suboptimal. The preservation of the uORF and its repressive effect on downstream translation in all cell types examined thus far, including diploid human fibroblasts, human mammary epithelial cells, BT-474 and MCF-7 breast cancer cell lines, and COS-7 cells, suggests that the uORF is a major determinant of HER-2 protein expression.
The current studies strengthen the hypothesis that the HER-2 uORF represses HER-2 protein synthesis. First, consistent with observed effects of the uORF on a lacZ reporter (29), we found that the uORF inhibited expression of the natural downstream cistron (Fig. 1). Second, the observed shift of mRNA to larger polysomes upon elimination of the uORF (Fig. 2) supports the conclusion that the uORF acts at the translational level.
Several considerations raised the possibility that the HER-2 uORF may function in a sequence-dependent manner, similar to regulatory uORFs found in several other eukaryotic genes (19,26,40). Reminiscent of the conservation of key codons between the sequence-dependent uORFs in the arg-2 gene of Neurospora crassa and the homologous CPA-1 gene of S. cerevisiae (27,40), the HER-2 uORF is conserved in sequence among all mammalian species examined to date (29). The peptide product of the sequence-dependent uORF2 of cytomegalovirus is synthesized and mediates repression of translation termination and ribosomal stalling on the UL4 gene mRNA (22,48). Based on the high level expression of a HER-2 uORF:␤-gal fusion gene (Fig. 7), we hypothesize that the HER-2 uORF peptide product is also synthesized, although we have not yet detected it. Despite these similarities, the HER-2 uORF does not act in a sequence-dependent manner since multiple missense mutants retain the inhibitory effect on downstream translation (Fig. 3). In computer data base searches, we have not detected the uORF sequence in any genes, including other epidermal growth factor receptor gene family members, except for HER-2. Thus we currently do not understand the significance of the conservation of the peptide sequence. It might be required for an unidentified function of the uORF other than inhibition of downstream translation.
The coding sequences of several uORFs can be mutated without altering their effects on downstream translation (21, 49 -51). In these cases, ribosomes presumably translate the uORF but are then unable to reinitiate efficiently at the downstream AUG codon. Parameters such as the length of the in-tercistronic region affect the efficiency of ribosomal reinitiation. For example, intercistronic spacing shorter than ϳ80 nucleotides has been reported to hinder downstream reinitiation (52). The 5-nt intercistronic region between the uORF and the HER-2 AUG codon is conserved among mammalian HER-2 genes. We found that increasing the intercistronic spacing by inserting additional sequences reduced the inhibitory effect of the uORF (Figs. 4 and 5). However, interpretation of our results is complicated by the contribution of both increased expression from the AUG ϩ constructs and decreased expression from the corresponding AUG Ϫ mutants. The inserted spacer sequences might exert a general inhibitory action affecting expression from both the AUGand AUG ϩ constructs. For example, the constructs having GCN4 transcript leader sequences (Fig. 5) did express lower amounts of ␤-gal RNA. Alternatively, by increasing the distance that ribosomes must traverse before encountering an AUG codon in the AUGconstructs, the spacer sequences might cause ribosomes to fall off the message before they ever reach the ␤-gal AUG codon. This effect would be expected to reduce ␤-gal translation from the AUG Ϫ mutants only. Until we identify a spacer sequence that does not alter expression from AUG Ϫ mutants, we cannot be certain that the reinitiation block in the natural HER-2 mRNA is due solely to the short distance. However, increasing the intercistronic spacer length to greater than ϳ50 nt results in a nearly 2-fold increase in expression downstream from the uORF, suggesting that at least some ribosomes that translate the uORF are unable to reinitiate at the HER-2 AUG codon because of its proximity to the uORF termination codon.
In addition to the length of the intercistronic region, its nucleotide sequence can influence translational reinitiation (41,42). For example, a uORF terminates seven nt upstream of the CCAAT/enhancer-binding protein ␣ (C/EBP␣) AUG codon and represses C/EBP␣ expression (41). However, minor changes in the intercistronic sequence greatly reduce the inhibitory effect of the uORF, indicating that the eukaryotic ribosomes can reinitiate after a very short intercistronic distance in certain cases. We have not yet found any substitution mutations of the intercistronic spacer sequences that alleviate the HER-2 uORF inhibitory effect. 2 The increased efficiency of ribosomes reinitiating after a longer intercistronic spacer led us to evaluate whether ribosomes translating the natural HER-2 mRNA reinitiate downstream from the HER-2 AUG codon. In the natural HER-2 mRNA, the next two AUG codons after the one initiating the HER-2 ORF are located 91 nt and 133 nt downstream from the HER-2 AUG codon. Both of these codons are in the same reading frame as HER-2, and both are flanked by a moderately 2 S. J. Child, M. K. Miller, and A. P. Geballe, unpublished data. FIG. 7. Translation initiation at the uORF AUG codon. To assess whether the HER-2 uORF is translated, a HER-2 uORF fusion construct (pEQ751, AUG ϩ ) and the corresponding AUGmutant were made. In these plasmids the uORF stop codon and the ␤-gal AUG codon were mutated, and a single nucleotide was inserted to fuse the uORF to the ␤-gal ORF. Unlike the plasmids shown in Figs. 3-6, these plasmids contained the entire 178 nt HER-2 leader (29) rather than the 93-nt conserved region leader. Transfections were performed as described in the legend to Fig. 3. strong initiation sequence (gacAUGa and gacAUGc, respectively) (30). By positioning the ␤-gal ORF in the approximate position of the first of these AUG codons, we found that the uORF aids ribosomal reinitiation at this downstream start site, although the overall efficiency of reinitiation is low (Fig. 5).
How do ribosomes ever gain access to the HER-2 AUG codon despite the very good context of the uORF AUG codon and the short intercistronic spacing? The results shown in Fig. 6 suggest that a small percentage of ribosomes bypass the uORF AUG codon and that additional ribosomes are able to reinitiate at the HER-2 start site after translation of the uORF. The presence of a purine at the Ϫ3 position relative to the AUG codon is usually thought to be sufficient for efficient initiation. However, an A can be superior to G at Ϫ3, in some cases increasing translational initiation by more than 3-fold (30). Thus, it is really not too surprising that some ribosomes leak past the HER-2 uORF AUG codon even though the uORF AUG codon has the context gccAUGg.
Our results suggest the hypothesis that the HER-2 uORF may serve to control the access of ribosomes to downstream AUG codons, in some ways similar to the role of first uORF in the GCN4 mRNA. Ribosomes translate the first GCN4 uORF and then, depending on growth conditions, they either reinitiate at another upstream AUG codon or at the further downstream GCN4 AUG codon (21). In the case of HER-2, some ribosomes that have translated the HER-2 uORF reinitiate at the nearby HER-2 AUG codon, whereas others reinitiate further downstream. At present, we do not know whether the HER-2 uORF acts in a constitutive or regulated manner. Analogous to the GCN4 system, the efficiency with which ribosomes reinitiate at the alternative downstream AUG codons after translating the uORF might be affected by cell type or growth conditions. This model predicts that an amino terminus-truncated HER-2 protein might be produced at a low level from the natural mRNA. Although we have not yet detected such a protein ( Fig. 1 and data not shown), additional studies are needed to establish conditions for resolving the full length HER-2 from this putative truncated form. What function might an amino terminus-truncated HER-2 protein serve? Other studies show that deletions of the amino terminus of the protein increase the tyrosine kinase activity and transforming efficiency of the rat neu gene (53,54). Thus, even if it is produced only at a low level, this putative alternative form of HER-2 may be biologically important.
Although HER-2 is essential for development (5), its overexpression can lead to cellular transformation and tumor growth. Given the requirement for precise control of the timing, location, and abundance of HER-2 protein expression, perhaps it is not surprising that multiple regulatory mechanisms, including repression by the uORF, have evolved. In fact, an unusually large proportion of mRNAs from genes involved in cellular growth control contain uORFs (20). As illustrated by the recent report of uORF regulation of the S. cerevisiae cell cycle regulator CLN3 (50), uORFs may serve as a general strategy for linking cellular growth to proliferation. Thus, in addition to providing information about control of HER-2 expression, studies such as these are needed to further our understanding of the variety of mechanisms by which uORFs affect gene expression.