Complexity of Translationally Controlled Transcription Factor Sp3 Isoform Expression

cocktail) and incubated with 50 µ l equilibrated agarose wheat germ lectin (Amersham Biosciences) for 1 hour at 4°C. Matrix-bound glycosylated proteins where washed extensively in binding buffer before adding 2x SDS Laemmli buffer (Sigma). For competition experiments, 100 mM N-acetyl D-glucosamine (Roth) was added to the binding buffer. Binding of Sp proteins to agarose wheat germ lectin was analyzed by Western blotting.


INTRODUCTION
The transcription factor Sp3 is a ubiquitously expressed member of the Sp family of transcription factor that is involved in the expression and regulation of many genes including house keeping genes, tissuespecifically expressed genes, viral genes as well as cell cycle-regulated genes (1,2). Sp3 contains a highly conserved DNA-binding domain close to the C-terminus, and two glutamine-rich activation domains in the Nterminal moiety. The expression pattern, the structure as well as the DNAbinding properties of Sp3 are very similar to Sp1, which suggested originally that these two proteins exert similar functions. The physiological roles of Sp1 and Sp3, however, appear to be significantly different. Sp1 knockout mouse embryos are severely retarded in growth, and die after day different isoforms show that their transcriptional inactivity is regulated by SUMO modification. Our results demonstrate that Sp3 has many unique molecular features and is not simply a functional equivalent of Sp1.

EXPERIMENTAL PROCEDURES
Plasmids-The missing 5´-part of the Sp3 cDNA was cloned by RT-P C R o f H e L a R N A u s i n g t h e p r i m e r s 5´-actcggaattcCCTTTTGTGTTTCCCGCACAGTCA-3´ and 5´-CTGTGCAGAAGCCAAATCACCTGT-5´. The resulting 470 bp product was cut with EcoRI (artificial site at the 5´-end of the forward primer) and NotI (site within the Sp3 cDNA sequence) and cloned into EcoRI/NotIrestricted pSPT18-Sp3 plasmid that contained the original published Sp3 sequences (13,14) For in vitro transcription/translation assays of wild-type Sp3 and mutants, the T7/Sp6 promoter containing pSPT18 plasmid was used. Transient expression of Sp3 variants in Drosophila Schneider SL2 cells (15)   In the past, analyses that could clarify the nature of the various Sp3 species were hampered due to the lack of a full-length Sp3 cDNA clone.
The original Sp3 cDNA sequences (13,14) did not contain an AUG translational start codon and it was postulated that translation of the Sp3 mRNA starts at an non-AUG codon (14). In the course of a comparative sequence analysis among the evolutionary related Sp transcription factors Sp1, Sp2, Sp3 and Sp4 (19) it was suggested that the 5´-end of Sp3 is absent in the published cDNA clones. Moreover, human genomic DNA sequences were identified that encompass three exons coding for additional 85 N-terminal amino acids of Sp3 (12). Based on these data, we cloned the missing 5´-part of Sp3 by RT PCR with primer pairs specific for sequences in exon 1 and exon 3 (GenBank accession number AF494280) (12) and fused it to the Sp3 coding region. The resulting Sp3 cDNA codes for 781 amino acids when translated from the first AUG codon.
Expression of the four Sp3 isoforms observed in immunoblots can be reconstituted in a coupled in vitro transcription/translation assay using the full-length Sp3 cDNA clone (Fig. 1B, lane 2). Moreover, the additional Sp3 species observed exclusively in SDS-lysed cell extracts can also be reconstituted when in vitro translated Sp3 proteins were subjected to an in vitro SUMO modification reaction (10) (Fig. 1B,    Unlike Sp1, Sp3 is not glycosylated-The transcription factor Sp1 is posttranslationally modified by glycosylation (22). Since Sp3 is evolutionary and structurally very closely related to Sp1, we asked whether any of the isoforms of Sp3 is also modified by glycosylation. Glycosylated proteins bind strongly to wheat germ agglutinin (WGA) and WGA affinity chromatography has been used in the past as a major step to purify Sp1 from cell extracts (23).
Whole cell extracts of untransfected 293 cells and of SL2 cells transfected with Sp1, wild-type Sp3 or the SUMOylation-deficient Sp3K551R mutant were incubated with wheat germ agglutinin matrix.
Subsequently, bound proteins were subjected to immunoblot analyses using antibodies to Sp1 and Sp3. As expected, endogenous Sp1 present in 293 extracts (Fig. 5, lane 2) as well as transiently (Fig. 5, lane 16) or stably (not shown) expressed Sp1 in SL2 cells bound strongly to wheat germ agglutinin (Fig. 5). Binding was specific as it was competed by an excess of N-acetyl D-glucosamine. (Fig. 5, lane 4). Surprisingly, neither the small, the long nor the SUMO-modified isoforms of Sp3 bound to wheat germ agglutinin (Fig. 5, lanes 6 and 14). The same holds true for the SUMO modification-deficient Sp3K551R mutant (Fig. 5, lane 12). These results clearly show that the transcription factor Sp3, unlike Sp1, is not a target for posttranslational modification by glycosylation.  6A and data not shown).

Functional differences between the various isoforms-All
The BCAT-2 reporter contains two GC-boxes fused to the E1b TATA box.
Upon transfection of the wild-type Sp3 construct or cDNA mutants from which predominantly the first isoform (2.AUGm), the second isoform (1.AUGm) or exclusively the third and fourth isoforms (Δ1.+2.AUG) were expressed, the BCAT-2 reporter became not activated (Fig. 6B). This result shows that all four isoforms of Sp3 are unable to activate the BCAT-2 reporter.
The SV40 promoter bearing five GC boxes becomes activated upon co-transfection of the wild-type Sp3 construct (WT) as well as by mutants from which predominantly the first isoform (2.AUGm) or the second isoform (1.AUGm) are expressed. The small isoforms of Sp3 expressed upon transfection of the Δ1.+2.AUG construct did not activate this reporter construct (Fig. 6B).
The promoter-specific activation capacity of the long isoforms of The capacity of Sp3 to activate transcription is silenced by SUMO modification at lysine K551 (9,10). These studies also have been performed with N-terminally truncated, epitope-tagged versions of Sp3 that, according to the results described above, mimicked the activity of the long isoforms.
Since all four isoforms of Sp3 are SUMO-modified, we asked whether the inactivity of the full-length Sp3 construct on the BCAT-2 reporter as well as the inactivity of the small isoforms on both, the BCAT-2 and the SV40 promoter is dependent on the presence of the SUMO target lysine K551.
To this end, we introduced mutations that prevent SUMOylation (K551 mutations) into the full-length Sp3 construct from which only the two long isoforms are expressed, and into the Sp3Δ 1.+2.AUG construct from which only the two small isoforms are expressed after transfection (Fig. 7A). Western blot analyses demonstrate that the small isoforms are expressed at higher level as compared to the long isoforms ( Fig 7B).
However, the SV40 promoter that is not activated by the wild-type small isoforms becomes strongly activated by the Sp3si-K551D mutant (Fig. 7D).
This result shows that also the inactivity of the small Sp3 isoforms that contain a single glutamine-rich activation domain is to some extent due to posttranslational modification by SUMO.

Subcellular localization of Sp3 isoforms-We wanted to know
whether the apparent differences in regulatory properties of the small and long isoforms of Sp3 may reflect differences in their subcellular localization. Sp3-deficient KO MEFs and SL2 cells were transfected with appropriate expression constructs for the long and the small wild-type Sp3 isoforms, as well as with the corresponding SUMOylation-deficient mutants. Immunostainings with anti-Sp3 antibodies revealed that in both cell types the wild-type isoforms and the SUMOylation-deficient mutants were located in the nucleus exhibiting a sponge-like, diffuse appearance (Fig. 8). These results show that the differences in the activation capacity of the various isoforms and mutants are not due to differences in their subcellular or subnuclear localisation.

An upstream open reading frame is involved in regulating Sp3
isoform expression-Initiation site selection at the first and second AUG is dependent on a short 30-nucleotide uORF. Mutation of the uAUG codon leads to exclusive initiation at the first AUG. Thus, the uORF is essential for the synthesis of the second long Sp3 isoform. Another mutation that brings the uAUG in an optimal context leads to predominant initiation at the second AUG start codon. This result could be explained also by a leaky scanning and reinitiation model. The uAUG is in a suboptimal context. Thus, leaky scanning over the uAUG would lead to initiation at the first AUG. Other ribosomes translate the uORF and reach the terminator codon.
According to current models, the 40S subunit may hold on to the mRNA, resume scanning and reinitiate at a downstream AUG (26). Reinitiation, however, is most efficient when the uORF terminates some distance before the next AUG to re-acquire Met-tRNAi-elF-2 for downstream AUG recognition (26). Since the distance between the uORF stop codon and the next AUG is only 17 nucleotides, these ribosomes would preferentially use the second AUG. Such a scenario would explain why predominantly the second AUG was used, when the uAUG codon was placed in an optimal Kozak sequence (more ribosomes initiate at the uAUG). In addition, this model explains also the shift towards initiation at the first AUG when the distance between the uORF stop codon and the first AUG was increased (insertion mutants described in Fig. 3).

Is Sp3 isoform expression altered under certain conditions?-Sp3
isoform expression is to some extent reminiscent of the isoform expression described for the transcription factors C/EBPα and C/EBPß (28)  Previously, we have shown that mutation of the SUMO lysine K551 impaired acetylation of Sp3 in vivo (32). This finding suggested originally that the same lysine residue that is a target for SUMOylation is also a target for acetylation. However, antibodies highly specific for the acetylated IKEE motif of Sp3 did not recognise endogenous Sp3 or Sp3 overexpressed in 293 cells or SL2 cells (unpublished data). These results suggest that it may not be K551 that becomes acetylated in vivo. We rather speculate that SUMO-modified Sp3 might recruit acetyl transferases that in turn acetylate other lysine residues within the Sp3 protein.