Podoplanin is a membrane mucin that, among a series of tissues, is expressed on late osteoblasts and osteocytes. Since recent findings have focussed on podoplanin's potential role as a tumour progression factor, we aimed at identifying regulatory elements conferring PDPN promoter activity. Here, we characterized the molecular mechanism controlling basal PDPN transcription in human osteoblast-like MG63 versus Saos-2 cells.
We cloned and sequenced 2056 nucleotides from the 5'-flanking region of the PDPN gene and a computational search revealed that the TATA and CAAT box-lacking promoter possesses features of a growth-related gene, such as a GC-rich 5' region and the presence of multiple putative Sp1, AP-4 and NF-1 sites. Reporter gene assays demonstrated a functional promoter in MG63 cells exhibiting 30-fold more activity than in Saos-2 cells. In vitro DNase I footprinting revealed eight protected regions flanked by DNaseI hypersensitive sites within the region bp -728 to -39 present in MG63, but not in Saos-2 cells. Among these regions, mutation and supershift electrophoretic mobility shift assays (EMSA) identified four Sp1/Sp3 binding sites and two binding sites for yet unknown transcription factors. Deletion studies demonstrated the functional importance of two Sp1/Sp3 sites for PDPN promoter activity. Overexpression of Sp1 and Sp3 independently increased the stimulatory effect of the promoter and podoplanin mRNA levels in MG63 and Saos-2 cells. In SL2 cells, Sp3 functioned as a repressor, while Sp1 and Sp3 acted positively synergistic. Weak PDPN promoter activity of Saos-2 cells correlated with low Sp1/Sp3 nuclear levels, which was confirmed by Sp1/Sp3 chromatin immunoprecipitations in vivo. Moreover, methylation-sensitive Southern blot analyses and bisulfite sequencing detected strong methylation of CpG sites upstream of bp -464 in MG63 cells, but hypomethylation of these sites in Saos-2 cells. Concomitantly, treatment with the DNA methyltransferase inhibitor 5-azaCdR in combination with trichostatin A (TSA) downregulated podoplanin mRNA levels in MG63 cells, and region-specific in vitro methylation of the distal promoter suggested that DNA methylation rather enhanced than hindered PDPN transcription in both cell types.
These data establish that in human osteoblast-like MG63 cells, Sp1 and Sp3 stimulate basal PDPN transcription in a concerted, yet independent manner, whereas Saos-2 cells lack sufficient nuclear Sp protein amounts for transcriptional activation. Moreover, a highly methylated chromatin conformation of the distal promoter region confers cell-type specific podoplanin upregulation versus Saos-2 cells.
Podoplanin represents a type I integral membrane protein of the mucin-type family. Its 162 aa protein core is extensively O-glycosylated, which results in a doubling of the molecular weight up to 45 kDa. The podoplanin mRNA was identified for the first time in the murine osteoblastic cell line MC3T3-E1 , and later it was found to be expressed in murine colon carcinoma and melanoma cells , in the stromal cells of peripheral lymphoid tissue , in skeletal muscle, heart, placenta, lung, skin, bone and brain [4,5]. Podoplanin has also been detected in the ciliary epithelium of the eye , in transformed epidermal keratinocytes and fibroblasts  and in the choroid plexus and neuronal cells of the CNS . In our lab, podoplanin has been discovered as a membrane protein of glomerular epithelial cells, which was found to be downregulated in puromycin nephrosis and caused a flattening of podocytes, hence its designation "podoplanin" . Moreover, podoplanin has been found in endothelia of lymphatic vessels  and has been established as a discrimination marker between lymphatic (LEC's) and blood endothelial cells (BEC's) . The widespread localization of podoplanin has lead to a multiplicity of synonyms for its expression forms of human, murine and rat origin: murine OTS-8 , gp38 , T1α , PA2.26 antigen , RANDAM-2  or Aggrus , rat E11 antigen  or RTI40 , and human gp36 .
The exact biological function of podoplanin remains controversial and seems to be versatile. Though its highly abundant expression in a multiplicity of tissues, distinct cell-specificities and functions have been described. In the lung, podoplanin has been shown to be restricted to type I lung alveolar cells , where it also functions as the main receptor for influenza C virus . Also quite early, podoplanin has been described as a platelet aggregating factor in murine and human colon carcinoma  and murine melanoma . Podoplanin/T1α deficiency in mice lead to disruption of normal lymphatic vasculature formation and development of lymphedema . Moreover, podoplanin has been demonstrated to bind chemokines in the lymphatic vessels of renal transplants  and to promote cell motility by downregulating E-cadherin expression in oral squamous cell carcinomas . Recent studies, therefore, have focussed on its potential role as a key player in tumour vessel formation  and as a tumour progression factor in skin  and germ line carcinomas . Podoplanin becomes upregulated in cells of the invasive cancer front, where it induces filopodia formation and promotes tumour cell migration .
Rat podoplanin/E11 has been reported to be a marker for late steps in the differentiation pathway of the osteoblastic lineage, being expressed by mature osteoblasts and newly formed osteocytes . The human osteosarcoma cell lines MG63 and Saos-2 used in this study represent differentiated late stage osteoblasts, nonetheless showing different degrees of maturation. Whereas MG63 cells exhibit a premature fibroblast-like state, Saos-2 cells display an epithelial phenotype and appear more rounded . Podoplanin/E11 was detected in a restricted subset of osteoblastic cells which is in direct contact with bone matrix, suggesting a functional role as an anti-adherent molecule according to its high sialic acid content. Nevertheless, the function of podoplanin in the bone is unknown.
Despite considerable progress that has been made in describing the expression pattern of podoplanin under normal and various pathological conditions, little is known about the mechanisms leading to the widespread gene activity of podoplanin. Due to its certain specificities and potential role in tumour progression, the regulation of PDPN gene expression is of high interest. In an early study, podoplanin/OTS-8 could be induced by phorbolester in the mouse osteoblastic cells MC3T3-E1 , and therefore it was defined as a target gene for early response proteins. However, the molecular mechanism underlying podoplanin expression in osteoblasts remains largely unclear. In rat alveolar type I cells, Sp1 elements and a thyroid transcription (TGT3/HNF-3) factor site have been identified as essential for transcriptional activation of podoplanin/T1α [24,25]. In lymphatic endothelial cells, podoplanin/T1α is an early responder to the lymphatic-specific master regulator Prox-1 , and recently, podoplanin expression was shown to be under control of endogenously produced IL-3 in human lymphatic endothelial precursor cells . However, no direct cis- and trans-acting regulatory elements involved in cell-specific upregulation of human podoplanin have been elucidated so far.
In this study, we demonstrate a cell-specific difference between MG63 and Saos-2 cells in the binding of nuclear proteins to sequences immediately adjacent and more distal to the transcription start site of the PDPN promoter. We identify four Sp1/Sp3 binding-sites inside a 654 bp element upstream of the PDPN gene and establish that two of them mediate basal transcription in MG63 cells, whereas Saos-2 cells seem to lack sufficient amounts of nuclear Sp1/Sp3 for PDPN promoter activation. We show that these two functional Sp-binding sites may involve further activator elements. Moreover, we assess the functional hierarchy of another regulatory system, the chromatin modification, which is strongly altered in MG63 versus Saos-2 cells distal of the core promoter and leads to enhanced expression of podoplanin.
Podoplanin mRNA expression in the human osteoblastic cell lines MG63 and Saos-2
To identify cell lines which are positive for podoplanin expression, reverse transcription PCRs were performed with a series of human cell lines of different origin (Fig. 1A). The osteoblastic osteosarcoma cell line MG63 exhibited a strong RT-PCR signal, whereas Saos-2 osteosarcoma cells lacked it. This result was confirmed by real-time PCR analyses showing that MG63 cells contained three logs more podoplanin mRNA transcripts than Saos-2 cells. While in MG63 cells the podoplanin mRNA population represented approximately 3% of GAPDH mRNA population, it made up just 0.003% in Saos-2 cells (Fig. 1B). A Northern blot analysis revealed the presence of two podoplanin transcript variants in MG63 cells (Fig. 1C). A 2.8 kb transcript corresponding to the full length mRNA was seen, and a shorter species was detected at 1.1 kb. As determined by 3' RACE analysis using MG63-derived cDNA, this short mRNA species represented a transcriptional variant ending after a cryptic polyA-signal at bp 345 of the 2068 bp 3'-UTR (data not shown).
Figure 1. Podoplanin mRNA transcription and transcriptional initiation in the human osteoblast-like cell line MG63. A) PDPN transcription was analyzed using primers specific for podoplanin cDNA (see "Material and Methods"). The human cell lines are: MG63 and Saos-2, osteosarcoma cells; HUVEC, umbilical vein endothelial cells; EAhy926, endothelial-like cells; Met5A, pleural mesothelial cells; A549, type II alveolar lung carcinoma cells; Raji, Burkitt's B-lymphoma cells; JY, B-lymphoblastoid cells; K562, erythroleukemia cells; CCL, "centrocyte-like" B-lymphocytes; HL60, promyeloid cells. B) Real-time PCR analysis of PDPN transcription in MG63 versus Saos-2 cells. Results ± S.D. of three independent PCRs are shown. Transcript levels are relative to GAPDH mRNA content. C) Northern blot analysis of podoplanin mRNA fragments in MG63 versus Saos-2 cells. D) RT-PCR determination of transcription initiation sites in MG63 cells. The lane numbers correspond to the RT-PCR 5' primers in Table 1. They were tested together with 3' primer hPPrev (+114/+96) for their ability to amplify the corresponding sequences in cDNA preparations of MG63 cells. PCR products are shown on an agarose gel stained with ethidium bromide. Lanes L: DNA ladder, Lane C: Negative control containing no template DNA for PCR reaction. E) Sequencing results of 48 clones retrieved by 5' RACE from MG63 cells. The 5' cDNA endings and their frequency are indicated above the sequence stretch.
Table 1. Primer sequences used in this study.
Reverse transcription PCR applying a series of nine ascending 5' primers suggested that in MG63 cells, podoplanin mRNA was initiated at multiple sites mostly located between bp -97 and +1 of the 5'-flanking region (Fig. 1D). In 5' RACE experiments using MG63-derived cDNA, the 5'-length of mRNA transcripts and their frequency of occurrence was determined. The positions and frequencies of cDNA variants derived from 5' RACE analysis are summarized in Fig. 1E. The major transcription start site of podoplanin mRNA was allocated to the position designated +1 (Fig. 1E and 2A).
Figure 2. Structure and activity of the human PDPN 5'-flanking region in MG-63 and Saos-2 cells. A) Nucleotide sequence and putative regulatory elements within bp -857/+205 of the 5'-flanking region of the human PDPN gene. The major transcription start site is indicated with an asterisk at +1. Putative regulatory elements highlighted by underlining were identified using the Transfac database. Shown are only sites which are conserved between the human, murine and rat promoter. Numbers on the left and right refer to the nucleotide position relative to the major transcription start site. The scheme below shows the relative positions of these putative transcription factor binding sites within bp -1885/+205. AP, activating enhancer binding protein; C/EBPβ, CCAAT/enhancer binding protein β; ETF, EGFR-specific transcription factor; ICSBP, interferon consensus sequence binding protein; NF-1, nuclear factor 1. B) Schematic representation of stepwise 5' deletions of the PDPN 5' flanking region cloned upstream of the luciferase reporter gene into the pGL3 vector. Negative numbers indicate the 5' end of the promoter fragment relative to the major transcription start site at +1. C) Luciferase activity of 5' deletion constructs were assessed in MG63 and Saos-2 cells 24 h after transfection with equal stoichiometric plasmid amounts and a constant amount of internal control Renilla luciferase plasmid as described under "Methods". Luciferase activity was normalized to Renilla activity and is relative to the longest bp -1885/+171 construct, which was set at 100%. The average ± S.D. for three to six independent transfection experiments, each performed in triplicates, is shown. Statistical analysis was performed using student's t-test, and is as follows: *, p < 0.05, **, p < 0.01, ***, p < 0.001. The statistical analysis indicates a significant promoter activity difference between the consecutive deletion-constructs.
Isolation and structural analysis of the human PDPN promoter
The human PDPN gene 5' flanking region was isolated by PCR amplifications from a genomic BAC clone and subsequent staggered cloning. When comparing the isolated promoter sequence with the human chromosome 1 sequence, sequence identity was confirmed despite a single base difference at bp -415 (C in BAC clone versus A in the database). DNA sequence analysis using the Transfac database showed that the 5' flanking region lacked a consensus TATA box, whereas a GATA box (GATAAA) was localized 31 nucleotides upstream of the main mRNA transcription initiation site (Fig. 2A). The PDPN gene 5'-flanking region was characterized by a high GC content and revealed the existence of several potential Sp1 transcription factor sites. Moreover, putative binding sites for the basic transcription factors AP-2, AP-4, C/EBP, and NF-1, but not for osteoblast-specific factors were detected along the isolated PDPN promoter sequence (Fig. 2A).
The PDPN promoter is transcriptionally regulated in MG63 and Saos-2 cells
Genomic DNA sequencing of the promoter sequence in Saos-2 cells confirmed that lack of PDPN transcription was not due to deletion mutations (data not shown). Thus, MG63 and Saos-2 cells proved to be a suitable cell system for investigation of transcriptional podoplanin regulation. To assess which portions of the promoter were involved in regulation of podoplanin expression, reporter plasmids harboring progressive deletions from the 5' end of the bp -1885/+171 PDPN flanking region were set according to putative binding sites of transcription factors (Fig. 2A). The sixteen promoter fragments were inserted upstream of a luciferase reporter gene (Fig. 2B) and were evaluated in transient expression analyses. Luciferase transcription was consistently 30 times more active in MG63 than in Saos-2 cells (Fig. 2C). In MG63 cells, bp -1104/+171 construct provided highest promoter activity (761 Luc/Ren × 100), whereas Saos-2 cells exhibited highest normalized luciferase activity with bp -498/+171 construct (27 Luc/Ren × 100). Transcriptional activity diminished progressively with further deletion of the 5' region, showing three activity minima with the constructs bp -666/+171, -261/+171 and -76/+171 in MG63 cells, and with constructs bp -186/+171 and -76/+171 in Saos-2 cells, which indicated the loss of positive regulatory elements (Fig. 2C). The low promoter activity of the smallest bp -26/+171 construct in both cell lines suggested that the 171 bp spanning 5'-UTR of podoplanin mRNA was not involved in upregulation of promoter activity. These data showed that the 1028 bp 5' region contained motifs providing basic PDPN promoter activity, whereas rather silencing motifs bound to the more upstream bp -1885/-1104 region.
Nuclear proteins of MG63 and Saos-2 cells bind to the PDPN promoter
To more precisely define cis-acting DNA sequences responsible for cell-type specific PDPN transcription, seven overlapping DNA probes encompassing bp -1169 to +171 of the PDPN promoter (Fig. 3A) were used for in vitro DNaseI footprinting assays. With MG63 nuclear extracts, eight protected regions, fp8 – fp1, were identified with five probes (Fig. 3B), while probes EB and SB exhibited no differential DNaseI digestion pattern (data not shown). Complexes fp6, fp5 and fp4 were detected on both the coding and the noncoding DNA strand (Fig. 3B), and footprinting region fp1 became also visible with Saos-2 nuclear extracts (Fig. 3B). Moreover, digestional modified regions flanking the protected regions emerged along the probes (Fig. 3, arrowheads), additionally indicating the presence of DNA-binding proteins.
Figure 3. In vitro DNaseI footprint analysis of protein-binding to the PDPN promoter. A) DNA probes used in DNaseI footprinting depicted above a restriction map of the -1885/+205 bp PDPN 5'-genomic region. These were: probeEB (bp -1169 to -646), probe(-571) (bp -857 to -571), probe(-360) (bp -666 to -360), probeBM (bp -649 to -385), probeSB (bp -498 to -85), probeSA (bp -261 to +31), and probeBX (bp -89 to +171). Asterisks indicate the positions of radioactive labeling as described in the experimental procedures section. Six of the probes analyze the coding DNA strand, whereas probe BM tracks the noncoding strand. B) DNaseI footprinting of probes using MG63 and Saos-2 nuclear cell extracts. End-labeled restriction fragments were incubated alone (Control) or with MG63 or Saos-2 cell nuclear extracts prior to digestion with DNase I. Footprinting was carried out in the sense and antisense direction. A sequencing reaction of the same region was run in parallel for exact size assignment of the protected regions. Protected sequences are indicated with single lines. Arrowheads denote sites hypersensitive to DNaseI.
To determine whether the detected elements indeed bound nuclear proteins, gel shift assays were performed using eight double stranded DNA oligonucleotides spanning the DNase I footprint regions (Table 1). Probes alone produced no shifted bands (data not shown). Probes fp7, fp5, fp2 and fp1 produced a similar pattern of low-mobility DNA-protein binding (complexes I – III) when adding nuclear extract of MG63 cells (Fig. 4, lanes -). The same bands became faintly visible with Saos-2 extracts except for probe fp7, which did not show complexes I and II. Consistent with the footprints, oligos fp6 and fp4 bound nuclear proteins strongly from MG63, but weak from Saos-2 cells (Fig. 4, lanes -). Although covering a broad footprint region, probe fp4 only yielded a binding pattern of several rapidly migrating proteins, and fp8 and fp3 did not reveal significant DNA-protein complexes (data not shown).
Figure 4. EMSA analysis of footprint regions. Sequence-specific proteins bound to oligos fp1, fp2, fp5 and fp7 of the human PDPN promoter. Nuclear extracts from MG63 and Saos-2 cells were incubated with 32P end-labeled double-stranded oligonucleotides, in the absence (-) or presence of a 100-fold molar excess of unlabeled specific (s) and unspecific (ns) oligonucleotide, as described in the "Methods" section. As unspecific probes, Oct-1 (fp1, fp2, fp5, fp7) or AP-4 (fp4 and fp6) consensus oligos were applied. Arrows denote the bands of altered mobility resulting from the interactions of nuclear proteins with the oligonucleotides.
Specificity of the gel shifted bands was demonstrated by competition assays in which nuclear extracts were preincubated with a 100-molar excess of unlabeled specific (Fig. 4, lanes s) or unrelated probe (Fig. 4, lanes ns) as competitor. As all bands of the gel shifts were eliminated or attenuated by excess of unlabeled probe, these moieties may bind the respective oligos in a manner requiring concomitant association. A computer analysis revealed consensus motifs for transcription factors NF-1 (fp7, fp6), C/EBP-α (fp5, fp4), HNF-3 (fp4), Oct-1 (fp4) and AP-4 (fp1) in the applied probes. However, a panel of antibodies against these factors did not interfere with the observed low molecular weight DNA-protein complexes in supershift assays (data not shown). Nonetheless, these data suggested the presence of several sequence-specific proteins bound to the proximal PDPN promoter in MG63 cells.
Sp proteins bind to fp7, fp5, fp2 and fp1in the PDPN promoter
Probes fp7, fp5, fp2 and fp1 contained GC-rich regions exhibiting high homology to the consensus Sp1-binding motif. To test the binding of Sp-proteins to these elements, supershift analyses with anti-Sp1 and anti-Sp3 antibodies were performed. When anti-Sp1 antibody was added, complex I disappeared in favor of a supershifted band, which showed the predominant contribution of Sp1 (Fig. 5A, lanes Sp1). Addition of anti-Sp3 antibody blocked protein-DNA binding, which resulted in loss of complex II, whereas no supershifted complex appeared (Fig. 5A, lanes Sp3). When used together, Sp1 and Sp3 antibodies supershifted or inhibited both complexes (Fig. 5A, lanes Sp1+Sp3). To further test whether the core Sp1 sequence was required for binding of nuclear proteins, gel shift assays with the same oligonucleotide probes, but mutated at residues known to be critical for Sp-binding (Table 1), were performed with nuclear extracts of MG63 cells. The mutant oligos failed to form the observed complexes, indicating that the chosen mutations were sufficient to eliminate the binding capacities of these sites (Fig. 5B). These studies demonstrated that four elements within the PDPN promoter, herewith designated Sp.4 – Sp.1, bound Sp1/Sp3 proteins from MG63 and also weakly from Saos-2 cells. The results furthermore indicated that Sp1 and Sp3 were not bound coordinately to the DNA stretches, since the anti-Sp1 antibody did not shift Sp3-DNA complexes and vice versa. This suggested that Sp1 and Sp3 occupied the same or overlapping sites of the PDPN promoter.
Figure 5. Sp1 and Sp3 interact with the proximal PDPN promoter in vitro and in vivo. A) Supershifts were performed by incubating nuclear proteins either with anti-Sp1 antibody or with anti-Sp3 antibody, both antibodies together, or non immune IgG before addition of probe. Anti-Sp1 antibody supershifted complex I (ss-Sp1), while anti-Sp3 antibody specifically blocked the formation of protein-DNA complex II. B) Mutation of Sp binding sites (mut) abolished the binding complexes I and II versus the wild-type probe (wt) with nuclear extracts from MG63 cells. The mutated oligonucleotides used in this study are listed in Table 1. C) Deletion mutants of the four Sp-elements in the PDPN promoter. The sequences above the Wild-Type construct exhibit the core sequences of the Sp binding sites. Above each mutation construct, the respective mutated nucleotides are drawn in bold italics. In construct Sp.4-1mut, all four Sp elements are mutated. D) The wild type reporter construct and five mutation constructs were transiently transfected into MG63 and Saos-2 cells. Firefly luciferase values were divided by those of the internal control Renilla luciferase to represent the absolute promoter activity. Stars indicate the statistical difference of the promoter activities between the unmutated and mutated transfection experiments. *, p < 0.05; **, p < 0.01.
GC boxes Sp.4 and Sp.2 are essential for PDPN promoter activity
To assess the functional role of Sp binding elements Sp.4 – Sp.1, PCR-directed mutagenesis was performed at the same residues that had blocked Sp1/Sp3 binding in gel shift assays and were tested for their reporter activity. The four Sp binding sites were altered individually and in combination (Fig. 5C), and mutations were introduced into the promoter luciferase variant -857/+171 that supported basal PDPN transcription. The transactivation pattern was similar for MG63 and Saos-2 cells, indicating that Sp1/Sp3 promotion took place through the same elements in both cell lines (Fig. 5D). The utmost relevance was seen for Sp-binding site Sp.2 at bp -100/-95, showing an activity reduction by 73% in MG63 and by 52% in Saos-2 cells upon mutation. Deletion of Sp-binding site Sp.4 at bp -700/-696 decreased luciferase activity by 57% in MG63 and by 16% in Saos-2 cells, whereas mutation of Sp-binding site Sp.3 (bp -570/-564) did not alter luciferase activity substantially. Mutation of site Sp.1 (bp -48/-43) lead to a slight activating trend, while combined mutation of all four sites reduced activity to more than 80% in both cell lines. Taken together, these data suggested that Sp-binding sites Sp.4 and Sp.2 alone, but especially in concert, were central regulatory elements necessary for constitutive PDPN promoter activity in both MG63 and Saos-2 cells, whereas Sp.3 and Sp.1 provided no single activating importance.
Stimulatory Sp1 and repressive Sp3 coordinatively activate the PDPN promoter
To directly confirm the effect of Sp1 and Sp3 on PDPN promoter activation, bp -857/+171 wild-type promoter luciferase construct was cotransfected with increasing amounts of CMV-promoter driven Sp1 and Sp3 expression plasmids alone and in combination into MG63 and Saos-2 cells (Fig. 6A). In both cell types, Sp1 overexpression resulted in a weak but dose dependent stimulation of luciferase activity with 1.8-fold induction in MG63 and 1.6-fold induction in Saos-2 cells relative to Sp1 mock-transfected cells. Similarly, Sp3 overexpression lead to a 1.4-fold and 1.8-fold induction in MG63 and Saos-2 cells, respectively. When both factors were coexpressed, no further activation above 1.8-fold promoter activity was seen. In contrast, luciferase activity of the combined mutational variant Sp.4-1mut was not altered in these experiments. These results suggested that Sp1 and Sp3 were able to stimulate PDPN transcription through Sp1/Sp3 binding sites in the proximal promoter region independently.
Figure 6. Sp1 and Sp3 are critically involved in PDPN promoter activation. A) Cotransfection of the core promoter construct (-857/+171) without (Wild-Type) and with Sp-binding site mutations (Mutant) using increasing amounts of pCMV-Sp1 and pMCS-Sp3 or a combination of both expression plasmids along with pRL-TK into MG63 and Saos-2 cells. The total amount of expression plasmids was filled to 0.17 μg with empty vector. Luciferase values were normalized to the internal control Renilla luciferase and fold-induction versus mock-transfected cells was evaluated. B) Real-time PCR analysis of podoplanin transcription in MG63 and Saos-2 cells upon transfection with Sp1 and Sp3 expression vectors. Transcript levels are relative to GAPDH mRNA content. The average ± S.D. from two independent experiments performed in triplicates is shown. C) Drosophila SL2 cells were transiently transfected with Wild-Type promoter, Sp.4-1mut construct and pGL3-Basic. Cells were cotransfected with Drosophila expression vectors pPac-empty, pPac-Sp1, pPac-Sp3, or both. Fold promoter activation versus pPac-empty transfections was evaluated. The average ± S.D. from three independent experiments performed in triplicates is shown. D) The role of Sp.1 for PDPN promoter activity. Cotransfection of the shortened promoter constructs (-76/+171) and (-38/+171) with pCMV-Sp1 or pMCS-Sp3 along with pRL-TK into MG63 and Saos-2 cells. The statistical difference of the promoter activity in all panels is shown against mock-transfected cells. *, p < 0.05, **, p < 0.01, ***, p < 0.001.
To confirm the effect of Sp1 and Sp3 on the activity of endogenous PDPN gene expression in MG63 and Saos-2 cells, real-time PCR was performed after transfection with Sp1- and Sp3-expression plasmids. The results corresponded to the observed potency of the respective factors to stimulate promoter activity. Podoplanin mRNA contents were clearly increased by overexpressed Sp1 and Sp3 (Fig. 6B). Overexpression of both, Sp1 and Sp3, did not further change mRNA levels in MG63 cells but strongly increased them in Saos-2 cells. These results showed that Sp1 and Sp3 indeed were able to upregulate podoplanin expression.
The roles of Sp1/Sp3 function on the PDPN promoter were more clearly established in Drosophila SL2 cells, which lack endogenous Sp-proteins. Drosophila-specific pPac-Sp1 and pPac-Sp3 expression plasmids were cotransfected with the wild-type and Sp-site mutant promoter variant Sp.4-1mut (Fig. 6C) and fold promoter activation versus pPac-empty transfections was evaluated. Cotransfection of Sp1 lead to a 16-fold luciferase activity increase of the wild-type promoter. In contrast to the results seen with MG63 and Saos-2 cells, overexpression of Sp3 reduced promoter activity 8-fold, which discovered a strong repressive action of solely present Sp3. Nevertheless, coexpression of both, Sp1 and Sp3, further increased promoter activity up to 21-fold. Since Sp1 activity was not decreased in the presence of pPac-Sp3, Sp3 seemingly did not compete with Sp1 for binding to the same site. The same effects were visible on a reduced level with mutant variant Sp.4-1mut and pGL3-Basic, indicating background activity of these constructs. Basically, these results suggested opposed action of single Sp1 or Sp3 on the PDPN promoter, while presence of both factors caused a concerted activation.
Mutation of site Sp.1 at bp -48/-43 (Sp.1mut) showed a slight increase in promoter activity (Fig. 5D), which corresponded to lowered activity of bp -76/+171 versus bp -38/+171 deletion construct (Fig. 2C). In order to explore a potential repressive role of Sp3 on this site, reporter constructs bp -76/+171 and -38/+171 were cotransfected with Sp1 and Sp3 expression plasmids and the multiple of promoter activation was determined (Fig. 6D). Consistently with luciferase reporter assays (Fig. 2C), construct -38/+171 conferred a roughly 2-fold higher promoter activity than -76/+171 in both cell lines (data not shown). Sp1 activated both constructs up to 1.4-fold in both cell lines. Consistent with the overexpression studies shown in Fig. 6A, Sp3 overexpression did not exert repressive effects on the bp -76/+171 construct. In MG63 cells, activity of the bp -76/+171 construct was not altered by additional Sp3-overexpression, while it even increased in Saos-2 cells. Therefore, we were not able to demonstrate a repressive Sp3 action on site Sp.1. We rather suggest a suppressive interaction of Sp-proteins bound to Sp.1 in combination with another yet unknown factor located more downstream from bp -38, which is abolished through shortening of the promoter or by deletion mutation of site Sp.1. A more detailed analysis of this site will be of future interest.
Conclusively, these results clearly demonstrated the pivotal importance of Sp1 and the auxiliary importance of Sp3 for basal transactivation of the PDPN promoter.
Saos-2 cells exhibit reduced nuclear Sp protein levels and low Sp protein amounts bound to the PDPN promoter in vivo
To further clarify the transcriptional difference between MG63 and Saos-2 cells, we focussed on the observation that nuclear extracts of Saos-2 cells produced weak Sp1/Sp3 EMSA signals compared to those derived with MG63 extracts (Fig. 5B). A comparative Western blot analysis showed little difference in overall Sp1/Sp3 protein contents between MG63 and Saos-2 cells (Fig. 7A, lanes 1 and 2, and densitometric analysis), whereas nuclear Sp1 and Sp3 amounts were lowered about 82 and 88%, respectively, in Saos-2 versus MG63 cells (Fig. 7A, lane 3 and 4, and densitometric analyis).
Figure 7. Sp1/3 cellular versus nuclear levels and in vivo occupation of the PDPN promoter. A) Immunoblot analysis and densitometric quantification of Sp1 and Sp3 protein contents in MG63 and Saos-2 cell lysates and nuclear extracts. Total cellular protein and nuclear extracts from MG63 and Saos-2 cells were resolved on 10% SDS-PAGE and blotted onto nitrocellulose. Levels of endogenous Sp1 and Sp3 proteins were determined by probing the blots with anti-Sp1 and anti-Sp3 antibodies, respectively. Sp1 shows a 105 kDa band, whereas Sp3 shows two bands at 120 and 75 kDa. Blots were reprobed with anti-GAPDH and anti-TFIID antibodies as loading control. After densitometric evaluation, Sp1/Sp3 levels were normalized to respective GAPDH and TFIID levels from the same blot. Data are given as the mean ± S.D. of samples of the experiment repeated three times. B) and C) Association of Sp1 and Sp3 to the PDPN promoter in vivo. Formaldehyde crosslinked chromatin was precipitated with an antibody specific for Sp1 (panel B) and Sp3 (panel C). PCRs were performed with primer pairs flanking the four Sp binding sites. L, 100 bp DNA ladder; +, immunoprecipitation with anti-Sp1 or anti-Sp3 antibody; co, precipitation control with unspecific rabbit IgG; -, negative control without antibody; Input, dilution series of control input DNA; no DNA, PCR negative control.
In order to assess whether Sp proteins bound to the identified DNA stretches in vivo, chromatin immunoprecipitations were performed using antibodies against Sp1 and Sp3 and crosslinked chromatin from MG63 and Saos-2 cells (Fig. 7B). After immunoprecipitation with anti-Sp1 antibody, extensive PCR amplifications from the Sp-binding sites relative to control input DNA were seen with MG63 derived chromatin (Fig. 7B, upper panels). In contrast, Saos-2 cells exhibited weak amplification levels versus input DNA (Fig. 7B, lower panels), which indicated low Sp1 occupation of these positions. PCRs from anti-Sp3 antibody precipitated chromatin resulted in overall weaker amplification results (Fig. 7C), which corresponded to the low DNA-bound Sp3 amounts seen in EMSAs (Fig. 5A). Site Sp.4 was stronger occupied by Sp3 in MG63 than in Saos-2 cells, while in both cell types site Sp.3 was scarcely occupied by Sp3. Sites Sp.2 and Sp.1 were acquainted with Sp3 to an equal, yet moderate extent in both cell lines. These data raised the possibility that drastically reduced nuclear Sp protein concentrations were responsible for weak PDPN promoter activity in Saos-2 compared to MG63 cells.
The PDPN promoter is strongly methylated in MG63 but not in Saos-2 cells
To further unravel the transcriptional difference between podoplanin expression in MG63 and Saos-2 cells, epigenetic effects were considered. The PDPN promoter contained a high CG content which provided putative methylation sites. Correspondingly, two CpG islands were detected , one encompassing bp -183/+16 that spanned the core promoter, and the other stretching from bp +76 to +517, including exon 1 (bp +203 to +267) plus 249 bp of intron 1 (Fig. 8A). A PstI genomic DNA-fragment that covered bp -2799 to +1191 of the PDPN promoter and gene was used to analyze a potential methylation-sensitive restriction pattern. The fragment contained 121 palindromic CpG dinucleotides, sixteen of which represented isoschizomeric HpaII/MspI restriction sites accumulating in three clusters. Five Southern blotting probes were situated beside and between these clusters (Fig. 8A). Digestion of genomic DNA with PstI alone produced the anticipated 3990 bp fragment (Fig. 8B, lanes P), and double-digestion with PstI plus methylation-insensitive enzyme MspI yielded the smallest possible fragments (1047, 1044, 399, 257 and 700 bp, respectively) (Fig. 8B, lanes M). When using PstI together with methylation-sensitive HpaII, complete digested bands were observed with Saos-2 derived genomic DNA, while in MG63 cells the distal probes P/B and F2/K mainly detected undigested 2.2 and 2.7 kb bands besides small portions of complete digested fragments (Fig. 8B, lanes H). Moreover, with probe F5/Bs only the digestion resistant 2.7 kb fragment was visible. These results indicated strong methylation of region bp -2799/-464 in MG63 cells, while it was totally unmethylated in Saos-2 cells.
Figure 8. The distal PDPN promoter is fully methylated in MG63 but non-methylated in Saos-2 cells. A) Schematic representation of HpaII/MspI binding sites and CpG islands within a PstI genomic DNA fragment spanning from bp -2799 to +1191 of the PDPN promoter with respect to the transcription initiation start site (arrow at +1). The graphs depict the CpG blot created by the EMBOSS program. Exon 1 is drawn as an open box in the sequence. HpaII/MspI recognition sites are represented by vertical lines along the sequence, and CpG islands are indicated as empty boxes above. Probes used for Southern blot analysis and stretches that were amplified after bisulfite genomic DNA sequencing are indicated as black bars and lines, respectively, below the sequence. B) Genomic DNA from MG63 and Saos-2 cells was double-digested with PstI, and with PstI in combination with the methylation-sensitive restriction enzyme HpaII or the methylation-insensitive restriction enzyme MspI. DNA then was fractionated by agarose gel electrophoresis followed by Southern blotting using the five DNA probes depicted in Fig. 8A. C) CpG methylation pattern of the PDPN promoter and of exonic and intronic CpG islands in MG63 cells as determined by sodium bisulfite genomic DNA sequencing. The CpG number indicates the order of the CpG site inside the PstI genomic fragment, and the CpG position denotes its position in relation to the transcription start site. The color-tone of the circle reflects the degree of methylation: black: 100%, dark blue: 99-67% blue: 66-34%, light blue: 33-1%, white: 0% methylation. Asterisks indicate CpG motifs that are part of a HpaII/MspI restriction digest site.
To clarify the PDPN promoter methylation status in MG63 cells in detail, genomic DNA was analyzed by bisulfite sequencing. Genomic DNA was treated with sodium bisulfite under conditions where cytosines are converted to uracils, while methylated cytosines remain unmodified . Three PCR reactions were set to inspect the methylation level of the HpaII/MspI restriction site clusters (Fig. 8C). When PCR1 (bp -198 to +131) was considered, the methylation level was found to be less than 5%. With PCR2 (bp -759 to -485), except CpG sites 32 and 33 that were methylated to 25% and 50%, respectively, all CpG sites were fully methylated. PCR3 (bp -2005 to -1572) comprised three CpG sites, which were highly (site 8 to 86%, site 9 to 57%) or fully (site 10) methylated. The observed methylation pattern corresponded to the fragmentation results seen in Southern blotting and identified a highly methylated DNA status of the PDPN promoter upstream of bp -485 in MG63 cells.
Methylation of GC boxes Sp.4 and Sp.2 does not affect Sp-protein binding
In MG63 cells, the heavy methylation pattern stretched accross functional Sp1/Sp3 binding site Sp.4 (CpG number 23, Fig. 8C), whereas binding site Sp.2 was situated in the virtually unmethylated proximal promoter region (CpG number 52, Fig. 8C). To test whether this situation gave rise to modulatory binding of Sp-proteins, we compared the capabilities of respective oligos fp7 (site Sp.4) and fp2 (site Sp.2) to inhibit the formation of Sp1-/Sp3-DNA complexes with nuclear extracts of MG63 cells in methylated as well as unmethylated form. As shown in Fig. 9A, protein-DNA complexes were competed by increasing concentrations of both, methylated and unmethylated forms of probes fp7 and fp2 to an equal extent. These data suggested that Sp binding sites Sp.4 and Sp.2 were able to interact with Sp1/Sp3 even when methylated.
Figure 9. The role of PDPN promoter methylation in MG63 cells. A) EMSA competition experiments were performed using cold methylated and unmethylated fp7 (site Sp.4) and fp2 (site Sp.2) probe as competitors. Competitors were used at 50-fold excess at the small end of the triangle and at 100-fold excess at its large end. B) Densitometric quantification of both podoplanin mRNA transcripts isolated from MG63 cells which had been treated with 5-azaCdR and TSA alone or in combination. The abundance of podoplanin transcripts was normalized to GAPDH transcripts in the same samples. The data are given as the mean ± S.D. of triplicate samples of the experiment repeated twice. The statistical analysis denotes the significant difference between non-treated MG63 cells and cells treated with 5-azaCdR or/and TSA. *, p < 0.05, ***, p < 0.001. C) Whole pGL3-constructs or selected inserts thereof were methylated or mock-methylated in vitro and transfected into MG63 and Saos-2 cells. Results are given as luciferase activity normalized to cotransfected pRL-TK activity. The experiment was repeated three times independently in triplicates. The statistical analysis denotes the significant difference between mock methylated (hatched) and respective methylated (black) constructs. *, p < 0.05, **, p < 0.01, ***, p < 0.001.
Distal promoter methylation enhances PDPN promoter activity
To further test the role of DNA methylation and histone acetylation for control of podoplanin expression in vivo, MG63 and Saos-2 cells were treated with methyltransferase inhibitor 5-aza-2'-deoxycytidine (5-azaCdR) and histone deacetylase inhibitor trichostatin A (TSA). Cell viability was tested by increasing concentrations of the agents over a cultivation period of 3 days to exclude mRNA downregulation effects upon apoptosis. Then, cells were treated either with 5-azaCdR and TSA alone or with 5-azaCdR followed by TSA, and podoplanin mRNA amounts were analyzed by quantitative Northern blot analysis. In Saos-2 cells, the drugs were not able to initiate de novo synthesis of podoplanin mRNA (data not shown), which demonstrated that lack of PDPN transcription was not under epigenetic control. In MG63 cells, variations of both podoplanin mRNA bands seen in Northern blot were evaluated (Fig. 9B). 5-azaCdR treatment did not change PDPN transcription levels when compared to untreated cells, while 0.6 μM TSA lead to a respective 28% and 48% reduction of PDPN transcription. Moreover, a combination of 1 μM 5-azaCdR and 0.6 μM TSA decreased PDPN transcription by 50% and 57%, respectively. These data indicated that in MG63 cells, in vivo modification of the promoter chromatin structure through combined demethylation and histone hyperacetylation rather repressed than enhanced PDPN transcription.
To investigate whether methylation of the distal PDPN promoter portion indeed assisted activation of gene transcription, further luciferase reporter gene assays were performed (Fig. 9C). Complete in vitro methylation of the promoter plasmids pGL3(-1885/+171) and pGL3(-857/+171) lead to more than 90% loss of promoter activity in MG63 and Saos-2 cells, indicating that the vectors were susceptible to inhibition by methylation. To further assess promoter activity upon selective methylation of defined promoter regions, fragments bp -1885/+171 and -857/+171, or fragments bp -1885/-189 and -825/-189, which comprised the region that wasmethylated in vivo in MG63 cells, were generated by restriction digest with SacI/XhoI or SacI/ApaI, respectively (Fig. 9C). The four fragments were methylated by SssI methylase or mock-methylated, religated into the unmethylated reporter plasmid backbone and transfected into MG63 and Saos-2 cells. In vitro methylation of all promoter CpGs until bp +171 revealed a weak trend towards transcriptional deactivation being more pronounced in Saos-2 than in MG63 cells (Fig. 9C). This result supported the observation that methylation of site Sp1.2 did not affect Sp1/Sp3 binding (Fig. 9A). Conversely, selective methylation of CpG sites upstream of bp -189 containing 35 or 24 CpG sites, respectively, led to considerable promoter activity increase versus mock methylated constructs, which added up to more than 40% and 30% in MG63 and to more than 25% and 60% in Saos-2 cells. These data indicated that a putative methylation-dependent enhancement mechanism, possibly efficiently altering chromatin structure, was able to act on PDPN transcription.