Transcriptional and Translational Regulation of the Léri-Weill and Turner Syndrome Homeobox Gene SHOX*

Regulation of gene expression is particularly important for gene dosage-dependent diseases and the phenomenon of clinical heterogeneity frequently associated with these phenotypes. We here report on the combined transcriptional and translational regulatory mechanisms controlling the expression of the Léri-Weill and Turner syndrome gene SHOX. We define an alternative promotor within exon 2 of the SHOX gene by transient transfections of mono- and bicistronic reporter constructs and demonstrate substantial differences in the translation efficiency of the mRNAs transcribed from these alternative promotors by in vitro translation assays and direct mRNA transfections into different cell lines. Although transcripts generated from the intragenic promotor (P2) are translated with high efficiencies, mRNA originating from the upstream promotor (P1) exhibit significant translation inhibitory effects due to seven AUG codons upstream of the main open reading frame (uAUGs). Site-directed mutagenesis of these uAUGs confers full translation efficiency to reporter mRNAs in different cell lines and after injection of Xenopus embryos. In conclusion, our data support a model where functional SHOX protein levels are regulated by a combination of transcriptional and translational control mechanisms.

The human pseudoautosomal homeobox gene SHOX has recently been shown to encode a cell type-specific transcription factor involved in cell cycle and growth regulation (1). 1 Haploinsufficient loss of the SHOX gene causes short stature and has been correlated with variable skeletal phenotypes frequently observed in Léri-Weill and Turner syndrome patients (2)(3)(4)(5)(6). SHOX-deficient individuals exhibit a considerable phenotypic heterogeneity ranging from mild, barely detectable skeletal malformations to severe dysplasia adversely affecting the life of these patients (7,8). This phenotypic heterogeneity is clinically well documented but not understood on a molecular level. Although dominant negative effects of mutated gene products have been discussed (1), phenotypes caused by complete gene deletions argue for a quantitative threshold of functional pro-tein to be necessary for normal development. Along these lines, the understanding of mechanisms regulating the level of functional protein will provide important clues to explain the SHOX-related phenotypes and their inherent phenotypic heterogeneity.
Although initially transcriptional regulation has been correlated with a variety of human diseases, it has become increasingly clear that post-transcriptional processing and differential translation represent equally important checkpoints in the tissue-specific and disease-related control of gene expression (9 -11). Among the various elements of mature mRNAs that are known to regulate translational efficiency, several recent investigations have emphasized the role of the 5Ј-untranslated region (UTR) 2 as major determinant controlling the initial steps of protein expression (12,13). Most eukaryotic mRNAs that exhibit high translational competence present themselves with short, 5Ј-m 7 GpppN (7-methylguanosine) cap containing UTRs that lack highly organized secondary structures. In contrast, several tightly regulated proteins including proto-oncogene products, growth factors, and their receptors as well as homeodomain proteins are encoded by mRNAs with long, complex 5Ј-UTRs with considerable secondary structures and multiple AUG codons upstream of the main open reading frame. Because these features interfere with the rate-limiting initial steps of the canonical cap-dependent translation mechanism and subsequent ribosomal scanning of the 5Ј-mRNA leader, they represent key indicators of mRNAs regulated on a translational level (14). In addition, these structural hallmarks have evoked and fueled a lively discussion about the capability of several cellular mRNAs to promote cap-independent translation by direct recruitment of ribosomes to internal ribosomal entry sites (IRES) (15)(16)(17)(18).
We investigated the structural and functional properties of the 5Ј-UTR of the SHOX mRNA by computational, in vitro, cell culture based and in vivo analyses. In vitro and cell culture studies demonstrated an alternative intragenic promotor within exon 2 of the SHOX gene. Because this dual promotor structure leads to alternative mRNAs with different 5Ј leaders, we analyzed and compared the translational efficiency of the two UTRs by direct RNA transfection assays. Here we provide evidence that the different 5Ј-UTRs exhibit differential translation efficiencies due to cell type-dependent inhibitory elements within the long SHOX mRNA variant. We finally integrate this data to propose a combinatorial model that takes into account transcriptional and translational mechanisms regulating the amount of functional SHOX protein.
In Vitro RNA Synthesis and in Vitro Translation-Synthetic mRNAs were generated from various SHOX-UTR-Luciferase fusion constructs using the mMESSAGE mMACHINE TM or MEGAscript TM reaction systems (Ambion) according to the supplier's recommendations. Briefly, 1 g of plasmid DNA linearized with XhoI were in vitro-transcribed with T3 polymerase in the presence of 6 mM CAP analog (m 7 G(5Ј)ppp(5Ј)G or A(5Ј)ppp(5Ј)G). The resulting mRNAs were digested with DNase I for 15 min at 37°C, purified by column chromatography (Ambion), and analyzed on a formaldehyd containing 1% agarose gel. Purified mRNAs were photometrically analyzed and quantitated according to standard procedures. These synthetic mRNAs were used to prime in vitro translation reactions carried out with the Rabbit Reticulo Lysate System (Promega) according to the protocol provided. All reactions contained 1 g of mRNA and 35 l of nuclease-treated reticulo lysate in a final volume of 50 l. 5 l were withdrawn from this reaction in 10 min intervals, combined with an equal volume of 2ϫ Passive Lysis Buffer (Promega), and firefly luciferase activity was determined as described above. Alternatively, pHPL-and pRF-derived plasmids were linearized with XhoI, column-purified (Qiagen), and directly used to prime coupled in vitro transcription/translation reactions (T N T) according to the manufacturer's instructions (Promega).
RNA Transfection and Quantitation of Translation Efficiency-Synthetic mRNAs generated with the mMESSAGE mMACHINE TM or ME-GAscript TM reaction systems (Ambion) were directly transfected into different cell lines using 2 g RNA and 8 l of TransMessenger TM Transfection Reagent (Qiagen) per well of a 6-well plate. Transfection procedure was carried out following the recommendations of the supplier. Transfected cells were trypsinized, washed once with phosphatebuffered saline, and the cell pellet was resuspended in 200 l phosphate-buffered saline. 40 l of this cell suspension were mixed with 10 l of 5ϫ Passive Lysis Buffer (Promega) and directly used for Luciferase assays. The remaining 160 l of the cell suspension were used for isolation of total RNA by standard procedures (Qiagen). 1 g aliquots of total RNA were reverse transcribed and analyzed by quantitative real time RT-PCR amplification with the FastStart reaction system (Roche Applied Science) in a LightCycler® instrument (Roche Applied Science) using the following primers: Luc1for, 5Ј-GGAGAGCAACT GCATAAG-GC-3Ј, and Luc1rev, 5Ј-CATCGACTGAAATCCCTGGT-3Ј. The reactions contained 60 nmol MgSO 4 in a final volume of 20 l and were carried out under the following conditions: 95°C/10 min; (95°C/15 s, 60°C/5 s, 72°C/15 s) 35X. Luciferase activities were corrected for RNA transfection efficiencies with the results from real time PCRs.

Structural Features of the SHOX 5Ј-UTR-The SHOXa and
SHOXb mRNAs (accession numbers NM_000451 and NM_006883) share a complex 5Ј-untranslated region encoded by two exons joined via canonical donor-acceptor splice sites. This 5Ј-UTR comprises 694 nucleotides and exhibits an inconspicuous GC content of 63%. Beyond its over average length, however, it exhibits several unusual features. First, it contains a terminal oligopyrimidine tract of 12 nucleotides (TOP) and 7 AUG codons upstream of the major open reading frame (uAUGs). The environment of these uAUGs ranges between 44 and 67% identity to Kozak's consensus sequence (A/G)CC(A/ G)CCAUGG (21,22) with uAUG 1 and uAUG 3 complying with the Ϫ3(A/G)/ϩ4G rule for favorable translation initiation (23). All uAUGs are associated with open reading frames (uORFs) putatively encoding peptides with sizes between 2 and 29 amino acids, none of which exhibits any significant homologies to known peptides (Fig. 1). Second, predictions of its secondary structure using the mfold 3.1 algorithm (24) reveals an exceptionally strong folding of this 5Ј-UTR with a Gibbs free energy of ⌬G ϭ Ϫ282 kcal/mol for the most stable configuration. The individual stem loops, potentially interfering with ribosomal scanning (25), exhibit free energies in the range of ⌬G ϭ Ϫ11.2 to Ϫ24.2 kcal/mol. The capacity to form extraordinary stable secondary structures, the terminal oligopyrimidine tract, and the presence of multiple uAUGs strongly suggest regulatory properties of this 5Ј-UTR related to the control of SHOX gene expression on a translational level.
Exon 2 of the SHOX Gene Harbors Alternative Promotor Activities-The calculated high degree of folding is indicative for IRES (15) and has prompted us to first analyze the SHOX 5Ј-UTR translational activity by transfection assays using the reporter constructs pHPL/UTR and pRF/UTR ( Fig. 2A). In pHPL a stable hairpin at the 5Ј-end of the SV40-driven transcription unit represses translation of a firefly luciferase encoding mRNA generated from the parental vector (19). As shown in Fig. 2B, the SHOX 5Ј-UTR is able to rescue this reporter gene expression if inserted between the 5Ј hairpin structure and the firefly luciferase encoding sequence. Within different cell lines, the observed up-regulation varies between 25-fold (HEK293), 72-fold (COS7), and 95-fold in U2Os cells over wild type pHPL activity (data not shown). Also, the SHOX 5Ј-UTR up-regulates protein expression from the downstream reporter when inserted between the sea pansy (Renilla reniformis) and firefly encoding sequences of the bicistronic reporter vector pRF (Fig. 2, A and B).
In contrast to this remarkable up-regulation in cell culturebased assays, we did not detect any expression enhancing effects of the SHOX 5Ј-UTR in vitro, using pRF/UTR primed coupled T N T assays (data not shown). This prima facie discrepancy between cell culture-based and in vitro-derived results can be explained by the absence of cell-specific factors necessary for IRES utilization or the presence of intrinsic promotor elements that remain undetectable in T 3 polymerase-driven in vitro T N T reactions. To distinguish between these two possibilities, we analyzed total RNA preparations from cells transfected with pHPL, pRF, and the corresponding 5Ј-UTR containing constructs. As shown by Northern blot analysis, all transfections yield RNAs complying in size with the expected SV40-driven transcription (Fig. 2C). However, we observed a significantly higher steady state level of firefly luciferase RNA in pHPL/UTR-transfected cells as compared with cells transfected with the parental plasmid (Fig. 2C, pHPL). The size of this transcript furthermore suggests a transcriptional start point about 600 bp downstream of the SV40 promotor within the SHOX 5Ј-UTR. In agreement with this interpretation, we detected a second firefly luciferase encoding RNA in pRF/UTR, but not in pRF-transfected cells corresponding in size to a monocistronic mRNA that originates from a transcription initiation event within the intercistronic SHOX 5Ј-UTR fragment (Fig. 2C, pRF). Both results clearly disfavor internal ribosomal entry underlying the observed up-regulation of reporter activity, but rather point to transcriptional initiation within the SHOX 5Ј-UTR. To corroborate the hypothesis of an alternative promotor, we generated the plasmid constructs pEX2-Luc, pEX2⌬1-Luc, pEX2⌬2-Luc, and pEX2⌬3-Luc by fusing different portions of genomic DNA residing upstream of the SHOX AUG start codon to a firefly luciferase reporter unit (Fig. 2D). Luciferase assays using extracts from cells transiently transfected with these plasmids confirmed transcription-promoting functions with the highest activity observed within a fragment of 298 bp directly preceding the SHOX AUG start codon (Fig. 2D).
In conclusion, these results demonstrate that the SHOX gene is transcribed from at least two alternative promotors (P 1 and P 2 ) generating distinct mRNAs that encode identical proteins, but vary in their 5Ј-UTR sequences. These transcripts are hereafter referred to as type 1 and type 2 transcripts, respectively.
The SHOX 5Ј-UTRs Differentially Regulate Translation Efficiency-As the promotor activity within exon 2 of the SHOX gene interferes with DNA transfection experiments, all subsequent analyses were directly based on in vitro generated mRNAs. To address the mechanism by which the two SHOX mRNA variants are translated, we generated in vitro transcripts from pBSK/UTR-Luc (type 1 mRNA) and pBSK/EX2⌬3-Luc (type 2 mRNA) in the presence of either a generic CAP analog (m 7 G(5Ј)ppp(5Ј)G; G-CAP) or an unmethylated variant (A(5Ј)ppp(5Ј)G; A-CAP) interfering with CAP-dependent translation initiation. The results from in vitro translation assays primed with these transcripts are shown in Fig. 3B. Although type 1 transcripts are translated at very low levels independently of the CAP structure, the unmethylated A-CAP substantially interferes with the high translation efficiency of type 2 variant. These results suggest that both type 1 and 2 transcripts do not support internal ribosomal entry, but require a CAP-dependent mechanism of translation initiation. Furthermore, they uncover considerable differences in translation competence between the 5Ј-UTRs of type 1 and type 2 transcripts. We next analyzed nested deletions of the type 1 SHOX 5Ј-UTR (Fig. 3A) by in vitro translation assays. As shown in Fig. 3C, translation efficiencies are indeed inversely related to the length of the 5Ј-UTR with an 85-fold difference between type 1 and type 2 variants. To confirm this in vitro-derived data in living cells, we transfected these mRNAs and quantitated their translational efficiency. As expected, we observed the same inverse correlation between length and translation activities in the osteogenic cell line U2Os (Fig. 3D). We next transfected the same in vitro generated mRNAs into different cell lines and determined their translational competence (Fig. 3E). Interestingly, this analysis not only confirms the translation-attenuating properties of the long 5Ј-UTR variant but also suggests some cell line-dependent responsiveness to the translation inhibitory elements presented by the different forms of the SHOX 5Ј-UTR (Fig. 3E). Taken together, these results provide compelling evidence that type 1 and type 2 transcripts generated from the alternative promotors P 1 and P 2 exhibit differential translation efficiencies due to inhibitory elements within the long SHOX 5Ј-UTR variant.
Upstream AUG Codons within the SHOX 5Ј-UTR Inhibit Translation in Vitro and in Vivo-Because the use of nested 5Ј deletions does not allow to discriminate structural effects from the influence of uAUGs within the SHOX 5Ј-UTR, we generated constructs harboring mutations of individual uAUGs or different combinations thereof. In vitro transcription of these constructs yield mRNAs identical in size and with similar overall structural features of their 5Ј-UTRs (Fig. 4, A and B). These in vitro transcripts were directly transfected into U2Os cells and their translation efficiency determined. As shown in Fig. 4B, mutation of all seven uAUGs within type 1 transcripts yields translational activities comparable with the high efficiency of type 2 mRNAs. Therefore, we can exclude length and overall structure, but rather define the uAUGs as critical determinants for the observed translational down-regulation. We next investigated if this effect can be attributed to individual uAUGs by transfection of reporter RNAs harboring individual uAUG mutations. We can show that none of the individual mutations is able to confer substantial activity increase, whereas combined mutations of uAUG 3-7 yield high translation efficiencies. These results argue for a concerted function of the SHOX uAUGs with a major contribution of uAUGs [3][4][5][6][7] . To confirm these results in vivo, we fused the wild type and mutated type 1 and the type 2 SHOX 5Ј-UTRs to the coding sequence of Xenopus noggin. Expression of noggin is well documented to cause dorsalization (loss of trunk and tail structures) during early embryogenesis (26). Injection of these constructs into early Xenopus embryos yields three distinct phenotypes that can be classified as wild type, mild, and severely dorsalized embryos (Fig. 4D). In agreement with cell culture-derived data, type 2 mRNAs induce the strongest effects, whereas type 1 mRNAs cause milder phenotypes only at significantly higher RNA levels. In contrast, reporter mRNAs harboring mutations in all uAUGs cause a phenotype comparable with type 2 mRNAs with respect to both severity and dose response (Fig. 4D). DISCUSSION Haploinsufficiency of the human pseudoautosomal gene SHOX causes short stature and skeletal malformations associated with the Léri-Weill and Turner syndromes with a remarkable phenotypic heterogeneity observed among SHOX-deficient patients. Besides other possible factors, this heterogeneity might be caused by variations in functional protein levels. Here we describe the concatenation of transcriptional and translational mechanisms regulating SHOX expression. We have identified an alternative, intragenic promotor residing within exon 2 of the SHOX gene. Using several reporter constructs, we were able to narrow down this promotor to a region of 300 bp upstream of the AUG start codon. Interestingly, this region contains a canonical TATA box and a CAAT box at positions Ϫ137 and Ϫ257, respectively. Deletion of the CAAT box containing fragment (pEX2⌬2) dramatically reduces reporter activity, indicating its relevance for the function of the identified P 2 promotor. In addition to these core promotor characteristics, our analysis reveals regulatory elements between Ϫ432 and Ϫ298 that negatively control the activity of this intragenic promotor. Although the physiological function of the dual promotor structure remains to be elucidated, it is tempting to speculate that the two SHOX promotors are utilized in a developmental-and/or differentiation-specific manner. Recent studies (27)(28)(29) in Drosophila have led to the concept of "core promotor competition" within single or among neighboring genes. This model suggests that different core promotors selectively interact with each other by competing not only for basic transcription factors but also for tissuespecific enhancer elements. This competition leads to a mutually

FIG. 4. Upstream AUGs down-regulate translation of the long SHOX 5-UTR in vitro and in vivo.
A, schematic of the in vitro-transcribed mRNAs used to examine the impact of uAUGs on translation efficiency. The 5Ј-UTRs of these mRNAs are identical in size and exhibit comparable structural features as determined by the free energy values for their predicted folding. B, translation efficiency of in vitro mutagenized RNAs after transfection into U2Os cells. Translation attenuating effects of the type 1 transcripts can be attributed to uAUGs present in their 5Ј-UTR because mutations of uAUG [3][4][5][6][7] or all seven uAUGs restore translation activities comparable with type 2 transcripts. C, schematic and gelelectrophoretic analysis of UTR-noggin fusion RNAs used for Xenopus embryo injection. Type 1 UTR (1), type 2 UTR (2), and mutagenized type 1 UTR (3) are fused to a noggin cDNA. D, these mRNAs yield three distinct dorsalized phenotypes. The weakest dorsalizing effect was observed with type 1 transcripts (1, wild type), whereas injection of both, type 2 (2), and mutagenized type 1 transcripts (3) resulted in embryos with more severe trunk and tail defects. Injection of increasing amounts (50, 100, and 500 pg) of different transcripts (1, 2, and 3) result in stronger dorsalization of the embryos. The phenotypes are characterized as wild type (blue), moderately (red), and strongly dorsalized (yellow). The effects of all three mRNAs exhibit a typical dose respose. exclusive utilization of individual core promotors. Application of such a competition model will be particularly interesting, because alternative utilization of the SHOX promotors yields distinct mRNAs with diverse 5ЈUTRs but identical coding capacities.
Although the identified intragenic promotor P 2 generates mRNAs with a short 5Ј-UTR (type 2), utilization of the upstream promotor P 1 yields mRNAs with a long and highly structured 5Ј-UTR (type 1). Interference of such complex 5Ј-UTRs with the translational scanning mechanism has led to the concept of alternative, CAP-independent translation mechanisms mediated by internal ribosomal entry sites. IRES-regulated translation, initially described for viral RNAs (30 -32), was more recently suggested as the underlying mechanism controlling the expression of several cellular transcripts (15,(33)(34)(35). We have therefore investigated the possibility of internal ribosomal entry into type 1 SHOX mRNAs by transient transfection of mono-and bicistronic vectors, in vitro T N T assays, and direct mRNA transfection into different cell lines. From these analyses, we conclude that the 5Ј-UTR of type 1 SHOX mRNAs does not contain significant IRES activities and that both SHOX transcripts are translated by the canonical CAP-dependent pathway.
We next analyzed nested deletions of the long SHOX 5Ј-UTR by mRNA transfection assays and observed an inverse correlation between length and translation activities of the SHOX 5Ј-UTRs in different cell lines. However, because nested deletions of the SHOX 5Ј-UTR increasingly disturb the overall secondary structure and sequentially remove important sequence characteristics, these results do not automatically substantiate length itself as the critical determinant for the low translation efficiency of type 1 mRNAs.
After having established its translation attenuating effects, we focused on the seven uAUGs within the type 1 SHOX 5Ј-UTR. The presence of such uAUGs has been reported to regulate protein expression from the major open reading frame by interfering with ribosomal scanning of the 5Ј leader sequence (36 -38). In agreement with these reports, we could demonstrate that mutations of these uAUGs lead to a markedly up-regulation of translation efficiency after direct mRNA transfection. In fact, type 1 mRNAs harboring mutations in all seven uAUGs resemble or even augment the activity of type 2 SHOX mRNAs in the osteogenic cell line U2Os. This observation excludes that the overall secondary structure or length itself interferes with the CAP-dependent scanning but rather attributes the low translation efficiency of type 1 SHOX mRNAs to the presence of uAUGs within its 5Ј leader. Similar effects of uAUGs have been reported for the translational control of several mRNAs encoding proto-oncogenes, growth factors, transcription factors, and other highly regulated proteins (13). For some of these cases, it could be demonstrated that translation initiates at one or more uAUGs and scanning is reinitiated after translation of the respective uORF. Such selective reinitiation after uORF translation was described for the GCN4 mRNA of Saccharomyces cerevisiae (39) and the murine mRNA for the activating transcription factor 4 (40). Upstream open reading frames within the 5Ј-UTR of both mRNAs promote an up-regulation of translation in response to the stress-dependent phosphorylation of the eukaryotic initiation factor eIF2. It will be particularly interesting to find out if type 1 SHOX transcripts generated from the upstream promotor P 1 are regulated in a similar manner, because SHOX expression has been observed in hypertrophic chondrocytes and may be related to cell cycle regulation and apoptosis in osteoblastic and chondrocytic cells. 1 In summary, the results reported here can be integrated into a combinatorial model of SHOX gene expression as depicted in Fig. 5. According to this model, SHOX expression is regulated on the transcriptional level by alternative promotors generating type 1 or type 2 transcripts with identical coding capacities but distinct translation efficiencies. This unique situation suggests that P 2 is utilized in situations with immediate needs of high SHOX amounts, whereas P 1 allows the generation of transcripts that facilitate fine tuning of protein levels by translational control mechanisms possibly related to cellular stress situations. Therefore, this model certainly merits further investigation to identify the molecular details determining alternative promotor utilization and developmental-and/or differentiation-dependent translational regulation of the SHOX gene.
FIG. 5. Combinatorial model SHOX regulation. Transcription of the SHOX gene is regulated by alternative promotors generating different type 1 or type 2 transcripts with identical coding capacity but distinct 5Ј-UTRs. These 5Ј-UTRs exhibit significant differences in their translation activity due to the presence of seven uAUGs in type 1 transcripts. Differential utilization of the alternative promotors therefore defines the amount of functional protein generated from the SHOX gene.