The transcription factor activating enhancer-binding protein epsilon (AP–2ε) regulates the core promoter of type II collagen (COL2A1)



A. K. Bosserhoff, Institute of Pathology, University of Regensburg, Franz-Josef-Strauss-Allee 11, 93053 Regensburg, Germany

Fax: +49 941 944 6602

Tel: +49 941 944 6705



The transcription factor activating enhancer-binding protein epsilon (AP–2ε) was recently shown to be expressed during late chondrocyte differentiation, especially in hypertrophic chondrocytes. In this study, we were able to reveal that the promoter of the type II collagen (COL2A1) gene, encoding the extracellular matrix protein type II collagen, is specifically regulated by AP–2ε. Expression of COL2A1 is downregulated at the transition of chondroblasts into hypertrophic chondrocytes and our data provide evidence that AP–2ε is involved in this process. In reporter gene assays, overexpression of AP–2ε in cartilaginous cell lines resulted in a significant reduction in COL2A1 core promoter activity of ~ 45%. Furthermore, we found that this process is dose-dependent and highly specific for the epsilon isoform. Computational analysis offered only a single putative AP–2-binding motif within the chosen promoter fragment but site-directed mutagenesis revealed this motif to be regulatory inactive. After expanding our screening to motifs containing minor differences from the classical AP–2 consensus sequence (5′–GCCN3GGC–3′), we determined the sequence 5′–GCCCAGGC–3′ ranging from position −128 to −135 bp as an important regulatory site and responsible for COL2A1 downregulation through AP–2ε. Interaction of AP–2ε with the COL2A1 promoter at this site was confirmed by chromatin immunoprecipitation and electromobility shift assay. Further, our experiments suggest that at least one additional factor is involved in this process. This is the first study to prove regulation of COL2A1 by AP–2ε highlighting the role of the transcription factor as a modulator of cartilage development.


activating enhancer-binding protein epsilon


chromatin immunoprecipitation assay


type X collagen


type II collagen


electromobility shift assay


short interfering RNA


Cartilage is an avascular, noninnervated tissue that is comprised of only a single cell type – chondrocytes. Their primary task is to maintain the balance between anabolism and catabolism of the various components of their surrounding extracellular matrix [1]. Type II collagen (Col2), encoded by the COL2A1 gene on human chromosome 12, is a homotrimer consisting of three α1helices [2] and is the most abundant extracellular protein produced by chondrocytes. Together with other cartilage-specific collagens, it forms a fibrillar meshwork which provides tensile strength to cartilaginous tissues by entangling proteoglycans and glycoproteins [3]. Different forms of chondrodysplasia and cartilage degeneration shown by individuals carrying mutations in COL2A1 underline the important role of this protein for skeletal development [4].

Synthesis of type II collagen is controlled by a number of extracellular cytokines and growth factors, as well as intracellular transcription factors [5]. Analyses of the COL2A1 DNA sequence revealed multiple regulatory sites within both the core promoter and first intron regions targeted by activating or inhibiting factors like Sox9, Egr1 and Sp1 [3, 6, 7]. Interestingly, the transcription factor lymphoid enhancer-binding factor 1 was recently shown to induce a juxtaposition of the 3′ untranslated region of COL2A1 with the core promoter through gene looping resulting in upregulation of COL2A1 expression [8]. With all these different mechanisms controlled by cis- and trans-acting elements, developmental stage- and tissue-specific expression of COL2A1 during chondrogenesis is ensured. Chondrogenesis represents the first part of the complex multistep process of endochondral bone formation that forms the majority of the skeleton in vertebrates [9]. This mechanism starts during embryogenesis (mid-fourth week in the human embryo) when undifferentiated mesenchymal cells aggregate to form condensations at the locations that prefigure future skeletal elements. They start to synthesize an early extracellular matrix containing type I collagen, hyaluronan, tenascin and fibronectin [10]. Subsequent maturation of the mesenchymal cells results in chondroblasts, characterized by the synthesis of cartilage extracellular matrix proteins, including type II collagen, proteoglycans, aggrecan and cartilage-derived retinoic acid-sensitive protein, whereas expression of type I collagen stops [10]. Eventually, the cells exit the cell cycle and increase in volume – they become hypertrophic. The hypertrophic chondrocytes terminally differentiate, attract blood vessels and ultimately undergo apoptosis. Endochondral ossification is completed when osteoblasts infiltrate into the cartilaginous matrix and displace it with bone tissue [11].

The transition of chondroblasts into hypertrophic chondrocytes represents not only a phenotypic shift, but also a major change in gene expression patterns. The cells cease to express early cartilage matrix genes like COL2A1 and induce the expression of type X collagen (COL10) and vascular endothelial growth factor [1]. This switch needs to be tightly controlled to assure correct organization of the growth plate in the long bones and bone formation in general. Although several positive and negative regulators are known to influence chondrocyte hypertrophy including the Indian hedgehog/parathyroid hormone-related protein pathway, proteins of the transforming growth factor-β superfamily as well as transcription factors like runt-related transcription factor and SRY-related HMG box 9/5/6 [11, 12], the whole process is still not fully understood. Interestingly, a hypertrophy-like phenotype is often resembled by osteoarthritic chondrocytes in pathological cartilage [13, 14]. Therefore, discovery of additional factors playing a role in chondrocyte hypertrophy regulation will enhance our understanding of this complex mechanism, which in turn will help to improve prevention of osteoarthritis and will help to develop therapeutic strategies against it.

The family of activating enhancer-binding protein–2 (AP–2) transcription factors comprises five members: AP–2α to -ε. The proteins are active as homo- or heterodimers and bind to the palindromic recognition sequence 5′–GCCN3GGC–3′ within multiple gene promoters [15]. In vitro and in vivo analyses of AP–2 knockout mice have demonstrated the importance of AP–2 genes in numerous physiological processes [16]. Members of the AP–2 family are also known to be expressed during chondrogenesis [17, 18]. For instance, it was show that AP–2α plays an inhibitory role during early chondroblast maturation. In previous studies, we were able to reveal that AP–2ε – the last identified AP–2 protein [19-21] – is expressed during late chondrocyte differentiation, especially in hypertrophic chondrocytes. There it regulates the expression of integrin alpha10 [22, 23]. AP–2ε was also shown to be upregulated in osteoarthritic chondrocytes resulting in enhanced expression of the chemokine C-X-C motif ligand 1, which in turn promotes calcification and extracellular matrix degradation [24]. Other than cartilage, AP–2ε expression was also detected in the olfactory bulb. Feng et al. discovered disorganized olfactory bulb lamination in AP–2ε deficient mice, yet the mice were able to discriminate a variety of odors [25]. Apparently, the mice do not show an obvious cartilage phenotype, but minor alterations in extracellular matrix composition or abnormalities during embryogenesis that are compensated later were not examined in detail.

In order to obtain additional information about the role of AP–2ε in cartilage we focused on typical chondrocyte differentiation markers and found the COL2A1 gene to be regulated by this transcription factor. Our data provide evidence that AP–2ε indirectly interacts with the core promoter of COL2A1 and subsequently inhibits its transcriptional activity.


A highly conserved COL2A1 core promoter fragment is negatively regulated by the transcription factor AP–2ε

In order to analyze the function of AP–2ε in cartilage differentiation we performed experiments with the prominent cartilage marker gene COL2A1. We used the ecr browser program to search for evolutionary conserved regions within mammalian COL2A1 core promoter sequences by comparing the human DNA sequence with that of different mammalian species (mouse, rat, dog and cow). General high sequence similarity was detected, with the first 550 bp upstream of the transcription starting point showing the highest level of accordance in all species (Fig. 1). Consequently, we focused our study on this region using a corresponding COL2A1 promoter construct cloned into a luciferase expression vector for our experiments: pCOL2A1–Luc [7].

Figure 1.

COL2A1 genomic analysis. Schematic diagram showing comparative genomic analysis of COL2A1 core promoter regions using the ecr browser program ( The human DNA sequence is compared with that of mouse, rat, dog and cow. Areas with > 50% sequence similarity are depicted above the black line. (Red) intergenic regions; (blue) coding sequence; (ochre) untranslated regions; (salmon) intron regions; (pink bars) evolutionary conserved regions. The black box represents the location of the 702 bp COL2A1 promoter construct used in this study.

SW1353 chondrosarcoma cells were transiently transfected with the promoter construct and luciferase activity was measured. It was ~ 40–fold more active than the control plasmid (pGL2basic), confirming its activity in the cell line. In order to analyze a modulation of COL2A1 promoter activity mediated through AP–2ε, an expression plasmid for the transcription factor was cotransfected into the cells. Interestingly, this resulted in a significant reduction in COL2A1 promoter activity of ~ 45% (Fig. 2A), which indicated that the chosen promoter fragment is negatively regulated by the transcription factor AP–2ε. In a second approach we used small interfering (si)RNA against AP–2ε in SW1353 cells to knockdown the mRNA level of the transcription factor (Fig. 2B). Compared with cells transfected with control siRNA we measured a significant upregulation of COL2A1 promoter activity of ~ 40% (Fig. 2A) and an enhanced COL2A1 mRNA level of ~ 170% (Fig. 2B).

Figure 2.

Regulation of the human COL2A1 promoter by AP–2ε. (A) Activity of pCOL2A1–Luc was analyzed in SW1353 cells using reporter gene assays. In addition, an expression construct for AP–2ε or siRNA against AP–2ε was cotransfected into the cells and promoter activity was measured. The respective control (empty vector or negative control siRNA) is set to 1. Cotransfection of an AP–2ε expression plasmid resulted in a significant reduction of promoter activity down to approximately 55%. siRNA against AP–2ε resulted in a significant upregulation of promoter activity to ~ 140%. (B) AP–2ɛ and COL2A1 mRNA expression was analyzed in SW1353 cells transfected with siRNA against AP–2ɛ via qRT-PCR (relative to β–actin). The efficiency of the AP–2ɛ knockdown was ~ 50%. This resulted in an enhanced COL2A1 mRNA level of ~ 170% compared with the control. Data are given as means ± SEM. *< 0.05; **< 0.01; ***< 0.001.

Specific downregulation of COL2A1 promoter activity by the AP–2ε isoform

To determine whether this effect is dependent on the concentration of AP–2ε we used a series of different amounts of AP–2ε expression plasmid. Low doses of plasmid (5 and 25 ng) led to no, or just minor, downregulation of the COL2A1 promoter (Fig. 3A). However, 50 ng, significantly reduced the activity of pCOL2A1–Luc. This correlates with a 6- to 10–fold overexpression of AP–2ε on mRNA level (data not shown). Higher concentrations did not result in a stronger downregulation of the promoter construct, implying that transfection of 50 ng AP–2ε expression plasmid was sufficient for further experiments. Next, we analyzed whether this effect is specific for the epsilon isoform. Neither AP–2α nor -β – the other two AP–2 molecules expressed during chondrogenesis [17, 18] – were able to suppress the activity of the COL2A1 promoter (Fig. 3B). Nevertheless, the inhibitory function of AP–2ε was dependent on the intracellular amount of other AP–2 isoforms. Cotransfection of equal amounts or an excess of AP–2α nullified the negative effect of AP–2ε (Fig. 3C), while minor concentrations of AP–2α (10 ng) did not.

Figure 3.

Inhibiting effect of AP–2ε on the COL2A1 promoter is highly specific. (A) Different amounts of AP–2ε expression plasmid were transfected into SW1353 cells together with pCOL2A1–Luc and luciferase activity was measured; 500 ng pCMX is set to 1. Low doses of plasmid (5 and 25 ng) led to no, or just minor, downregulation of the COL2A1 promoter, whereas 50 ng or more significantly reduced the activity of pCOL2A1–Luc. (B) AP–2α and -β expression plasmids (50 ng) were transfected into SW1353 cells together with pCOL2A1–Luc and luciferase activity was measured. pCMX is set to 1. By contrast to AP–2ε, AP–2α and -β did not suppress promoter activity of the COL2A1 promoter construct. (C) Different amounts of AP–2α expression plasmid were transfected into SW1353 cells together with 50 ng AP–2ε expression plasmid and pCOL2A1–Luc; 250 ng pCMX is set to 1. Cotransfection of equal amounts or an excess of AP–2α nullified the negative effect of AP–2ε, whereas 10 ng AP–2α were not able to do so. (D) An expression plasmid for the splice variant AP–2α Isoform B was transfected into SW1353 cells together with pCOL2A1–Luc and promoter activity was measured; 50 ng pCMX is set to 1. Overexpression of AP–2α IsoB resulted only in minor, nonsignificant downregulation of COL2A1 promoter activity. Data are given as means ± SEM. ns, not significant; *< 0.05; **< 0.01.

A naturally occurring AP–2α splice variant that lacks parts of the domain necessary for DNA binding (AP–2α isoform B) was shown to be an efficient inhibitor of AP–2 transactivator function by heterodimerization with ‘active’ AP–2 proteins [26, 27]. We used this variant as additional control to confirm the specificity of the effect observed with AP–2ε. Because none of the AP–2 isoforms activated the promoter construct we did not expect a strong negative effect of AP–2α isoform B (Fig. 3D). In fact, we were able to observe only insignificant, minor reduction of luciferase activity upon transfection of cells with AP–2α IsoB. This finding indicates that AP–2ε mediated suppression of COL2A1 is not due to entrapping other AP–2 isoforms but rather a different mechanism.

Mapping the site responsible for the AP–2ε effect within the COL2A1 promoter sequence

To analyze whether AP–2ε directly interacts with the COL2A1 promoter, we studied the nucleotide sequence in closer detail. Using matinspector software of Genomatix a single putative AP–2-binding site within the 702 bp promoter fragment was detected, occupying −309 to −317 bp (AP–2.1) (Fig. 4). The screening is based on known consensus binding sites for transcription factors.

Figure 4.

Sequence of the utilized COL2A1 promoter fragment. Computational screening of the sequence with matinspector software offered one putative AP–2-binding site (AP–2.1; −309 to −317 bp). Manual screening for further potential binding motifs offered two more sites: AP–2.2 (−212 to −221 bp), and AP–2.3 (−128 to −135 bp). All sites are underlined and highlighted in gray. The TATA-box is underlined. The transcription starting point (+1) is marked with an arrow. Location is given in bp relative to the transcription starting point.

To analyze whether this site is involved in the inhibitory process induced by AP–2ε we disrupted the palindromic binding motif via site-directed mutagenesis (AP–2.1*) and again measured luciferase activity. For this and all subsequent mutations we exchanged two to three nucleotides within the first three base pairs of the putative AP–2 motifs (GCC) for A/T, as this was shown to be effective in prior studies [24]. Contradicting our expectations, forced AP–2ε expression led to the same observation made with the wild-type construct: AP–2ε significantly reduced the activity of pCOL2A1–Luc/AP–2.1* (Fig. 5A). This implied that AP–2.1 was no binding site of AP–2ε.

Figure 5.

Mutagenesis of putative AP–2-binding sites in the COL2A1 promoter. (A, B) COL2A1 promoter constructs with mutated AP–2-binding sites were subcloned into pGL2basic, and reporter gene assays were performed in SW1353 cells. In addition, an expression construct for AP–2ε was cotransfected into the cells and promoter activity was measured. pGL2basic or pCMX, respectively, is set to 1. Disruption of AP–2.1 via site-directed mutagenesis did not alter promoter activity or the negative effect induced by AP–2ε (A), neither did disruption of AP.2.2 (B). Disruption of AP–2.3 strongly reduced basic promoter activity (B), indicating that this site has an important regulatory function. In addition, downregulation of promoter activity after cotransfection of AP–2ε was completely prohibited. (C) Sequence alignment of the three putative AP–2 binding motifs in human, mouse, rat, dog and cow. The 5′–GCCGGC–3′ palindrome of the site AP–2.3 in the COL2A1 promoter is fully conserved in all species, whereas the palindromic structure of AP–2.1 and AP–2.2 is disrupted in at least one species. (D) Repeating the experiments from (B) in another model cell line (C–28/I2) revealed results comparable with SW1353. pGL2basic (left) or pCMX (right), respectively, is set to 1. Data are given as means ± SEM. ns, not significant; *< 0.05; **< 0.01.

Because the computational analysis offered no further potential AP–2-binding sites within the respective promoter sequence, we expanded our search to motifs that slightly differed from the classical AP–2-binding site 5′–GCCN3GGC–3′. More precisely, we screened the sequence for motifs that still contained the typical GCC–GGC palindrome, yet separated by 3 ± 1 intermediate base pairs rather than exactly 3. Two more sites matched our specifications: AP–2.2 (−212 to −221 bp; 5′–GCCACTCGGC–3′) and AP–2.3 (−128 to −135 bp; 5′–GCCCAGGC–3′) (Fig. 4). Mutation of AP–2.2 did not alter the activity of the promoter construct or the negative effect induced by AP–2ε (Fig. 5B). However, disruption of AP–2.3 strongly reduced basic promoter activity in SW1353 by about threefold implying that this region has a prominent regulatory function. Most important, cotransfection of an expression plasmid for AP–2ε did not show an effect on luciferase activity after mutation of the site AP–2.3 (Fig. 5B). A sequence comparison revealed that the 5′–GCCGGC–3′ palindrome – a characteristic feature of AP–2 binding motifs – of AP–2.3 is fully conserved in all species analyzed with the ECR browser, whereas the palindromic structure of AP–2.1 and AP–2.2 is disrupted in at least one species (Fig. 5C). This finding underlines that sites 2.1 and 2.2 have no generally conserved regulatory function in COL2A1 promoter regulation.

To further validate our data we repeated this experiment in the immortalized human chondrocytic cell line C–28/I2. We obtained similar results for the individual mutations although promoter inhibition was only ~ 30% in these cells (Fig. 5D). This minor difference might be attributed to the fact that basic promoter activity was only about one-third in C–28/I2 compared with SW1353.

Indirect interaction of AP–2ε with the identified site

Interaction of AP–2ε with the promoter of COL2A1 in vivo was demonstrated by chromatin immunoprecipitation (ChIP) assays with chromatin from osteoarthritic chondrocytes, in which AP–2ε is known to be upregulated [23]. Localization of AP–2ε at the AP–2.3 site within the COL2A1 promoter could be revealed using chromatin precipitated with a specific AP–2ε antiserum and a primer pair spanning this site (Fig. 6A). As expected, no PCR product was generated in the corresponding control reactions on chromatin precipitated with an IgG and a polymerase II antibody, respectively. Consistent with the data from the reporter gene assays, primer pairs spanning sites AP–2.1 and AP–2.2 did not produce a band on chromatin precipitated with the AP–2ε antiserum (Fig. 6A), implying that AP–2ε does not interact with the COL2A1 promoter at these regions. Specific immunoprecipitation was further confirmed in PCR with GAPDH primers and negative control primers provided by the manufacturer (Fig. 6A). In addition, an electromobility shift assay (EMSA) with a labeled oligonucleotide spanning this site was performed. Figure 6B depicts that incubation of the oligonucleotide with nuclear extract of SW1353 cells transfected with AP–2ε led to the generation of a number of protein–DNA complexes, probably derived from multiple proteins interacting with the oligonucleotide (lane 2). Incubation with an excess of unlabeled oligonucleotides (comp) could completely displace those complexes (lane 3). AP–2ε binding was confirmed by incubation with a specific AP–2ε antibody (lane 4) leading to a supershifted complex. In order to show direct binding of AP–2ε to the AP–2.3 site we repeated the EMSA with in vitro translated AP–2ε, which was confirmed by western blotting, but did not observe a different shift pattern compared with the control (pCMX) (Fig. 6C). Therefore, we assume that AP–2ε interaction with the COL2A1 promoter is indirect and dependent on at least one additional factor.

Figure 6.

Indirect interaction of AP–2ε with the COL2A1 promoter. (A) ChIP assay demonstrates the interaction of AP–2ε with the COL2A1 promoter. DNA samples of ChIP reactions (chromatin precipitated with Pol II, unspecific IgG, and AP–2ε antibodies, respectively) and input DNA were used in PCRs with different primer pairs (GAPDH, negative control primers and primer pairs spanning the three putative AP–2 sites). All PCR fragments could be detected in the input DNA sample. A PCR product was generated using the AP–2ε ChIP DNA and the primer pair spanning the binding site AP–2.3 but not with primers spanning AP–2.1 and AP–2.2, respectively. (B) Electromobility shift assay to demonstrate the interaction of AP–2ε with the COL2A1 promoter. AP–2ε binding was shown using labeled oligonucleotides spanning the AP–2-binding site AP–2.3 (lane 2) incubated with nuclear extract of SW1353 cells transfected with 50 ng of AP–2ε expression plasmid. For competition experiments, unlabeled oligonucleotides (lane 3) were used in excess. A specific AP–2ε antibody led to generation of a supershifted complex, while bands marked by a black arrow slightly vanish (lane 4). (C) (Upper) Western blot analysis to confirm successful in vitro translation of AP–2ε. Compared with the control (pCMX), an additional band is clearly visible, resembling AP–2ε. (Lower) Electromobility shift assay with in vitro translated AP–2ε and oligonucleotides used in (B). No direct binding of AP–2ε could be determined.

Taken together, we showed that interaction of AP–2ε with the COL2A1 promoter at the identified site results in downregulation of its activity. Because AP–2ε was shown to be expressed in hypertrophic chondrocytes, this regulatory mechanism could support the shift from COL2 to COL10 expression in these highly differentiated cells.


In recent studies we revealed that AP–2ε is a positive regulator of integrin alpha10 and C-X-C motif ligand 1 in cartilage [22, 24]. This study was designed to further analyze the role of the transcription factor during chondrogenesis. As AP–2ε expression was detected in hypertrophic chondrocytes via mRNA analysis and immunohistochemistry, we focused our study on this important step of cartilage development. When chondrocytes enter hypertrophy, they exit the cell cycle, increase in volume, downregulate the expression of COL2A1 and induce the expression of COL10 [1]. We speculated that AP–2ε possibly plays a role in the COL2A1 to COL10 transition either by inducing COL10 promoter activity or inhibiting COL2A1 promoter activity.

Therefore, we analyzed the promoters of the two genes for a possible regulation through the transcription factor. Our experiments did not reveal a specific effect of AP–2ε on COL10 promoter activity (data not shown). However, we determined the promoter of COL2A1 to be consistently downregulated by AP–2ε by ~ 45% in reporter gene assays in SW1353 cells. Furthermore, siRNA against AP-2 epsilon resulted in a significant upregulation of COL2A1 promoter activity and mRNA expression. We concentrated on a highly conserved region spanning the first 500 bp upstream of the transcription starting point. It is already known that this proximal promoter region contains several regulatory elements mediating inhibitory effects, indicating its important role in COL2A1 downregulation [28-30]. In our experiments, transfection of only 50 ng expression plasmid per six-well plate resulted in a significant inhibition of promoter activity, whereas higher concentrations of 100 and 500 ng did not enhance this effect. This implies that a relatively low amount of AP–2ε was enough to saturate our regulatory system. The fact that such low doses show clear effects could already be seen in previous studies with AP–2ε and underlines its impact on regulatory processes [24].

Further assays revealed that individual overexpression of AP–2α and -β, two other AP–2 proteins influencing cartilage development [17, 18], did not significantly alter the activity of the chosen promoter fragment. Nevertheless, cotransfection of AP–2ε with equal amounts or an excess of AP–2α nullified the negative effect of the ε isoform. Because AP–2 transcription factors are only active after homo- or heterodimerization it is possible that cotransfection of a sufficient amount of AP–2α leads to predominant generation of a dimer combination that does not affect the COL2A1 promoter. Most likely, only AP–2ε homodimers can make interactions necessary for the downregulation of the COL2A1 promoter, but not α–ε heterodimers. Hereby, overexpression of AP–2α reduces the level of AP–2ε homodimers and thereby abolishes the reduction of promoter activity through AP–2ε. Taken together, these data show that downregulation of COL2A1 promoter activity by AP–2ε is highly specific.

The AP–2α splice variant AP–2α isoform B, lacking amino acids necessary for DNA-binding, interacts with normal AP–2 proteins resulting in AP–2 heterodimers unable to bind DNA [26, 27]. In this way, it acts as an inhibitor of AP–2 transactivator function. Overexpression of AP–2αB did not result in a significant downregulation of COL2A1 promoter activity. This suggests that AP–2ε-dependent suppression of COL2A1 must be based on a differential molecular mechanism.

Computational analysis offered a single putative AP–2-binding site within our promoter fragment (AP–2.1), but we could not confirm this site to be regulatory active. Interestingly, a literature study revealed that AP–2-binding sites differing from the classical sequence motif 5′–GCCN3GGC–3′ could be identified. An example is the SV40 enhancer element 5′–CCCCAGGC–3′, indicating that AP–2-binding sites may represent miscellaneous GC-rich elements [16]. Because the COL2A1 promoter is generally highly GC-rich, we further focused on motifs still containing the typical GCC–GGC palindrome, yet divergent in the number of intermediate base pairs and found two such sites (AP–2.2/AP–2.3). Disruption of AP–2.2 did not influence COL2A1 promoter regulation in our experiments. Then again, mutation of AP–2.3 resulted in two different observations.

On the one hand, cotransfection of AP–2ε no longer affected promoter activity, providing a strong hint that this region must be responsible for the effect induced by AP–2ε. On the other hand, to our surprise, general COL2A1 promoter activity was strongly reduced. This finding most likely is explained by an unknown activator also binding to this very region, providing evidence that the DNA sequence surrounding AP–2.3 is an important regulatory element within the COL2A1 core promoter where multiple transcription factors can bind. Although several factors like Sox-proteins and SP1 [3] are known to positively regulate the COL2A1 core promoter we could not identify a known transactivator binding site overlapping the mutated bases within the AP–2.3 sequence.

Finally, interaction of AP–2ε with the COL2A1 promoter in vitro and in vivo at AP–2.3 could be clearly shown utilizing ChIP and detecting a super shift in EMSA. Surprisingly, direct binding of AP–2ε to the site using in vitro translated AP–2ε protein could not be found. Therefore, we assume that AP–2ε indirectly interacts with the COL2A1 promoter by binding to it only when another factor is present, which must be missing in the rabbit reticulocyte lysate used for in vitro translation, yet is expressed in the two cartilaginous cell lines SW1353 and C–28/I2. Taken together, the following model of interaction explains our observations: with a functional (i.e. wild-type) AP–2.3 site, the activator, mentioned above, can bind to the COL2A1 promoter and induce its activity. AP–2ɛ, if present, can interact with this protein (and possibly the DNA as well), resulting in downregulation of COL2A1 promoter activity. Mutation of the respective DNA-binding site AP–2.3 results in a strong reduction of promoter activity and a loss of responsiveness to AP–2ɛ regulation, both because the two factors are unable to interact with the promoter (Fig. 7). The involvement of additional proteins in this process is in line with the detection of several bands in the EMSA and explains why we detected a supershift without bands fully diminishing. Furthermore, this hypothesis would not necessarily require a new variation of the AP–2-binding site. Additional experiments will be needed to fully solve this issue.

Figure 7.

Putative correlation between AP–2ɛ and the COL2A1 core promoter. Most likely, AP–2ε indirectly interacts with the COL2A1 core promoter by binding only to the AP–2.3 site when another factor is present (i.e. a so far unknown activator) and subsequently evokes a downregulation of promoter activity when it is (over-)expressed. After mutation of the respective DNA-binding site, activity of the promoter fragment is strongly reduced and no longer responsive to AP–2ɛ, because binding of both factors is blocked. Arrows indicate promoter activity.

Nevertheless, our results demonstrate that AP–2ε is a negative regulator of the COL2A1 promoter and enlighten its function in hypertrophy regulation. We hypothesize that AP–2ε indirectly interacts with the COL2A1 promoter, when it is synthesized in hypertrophic chondrocytes, resulting in downregulation of COL2A1 promoter activity.

Materials and methods

Cell culture

The human chondrosarcoma cell line SW1353 was obtained from the American Type Culture Collection (ATCC, #HTB-94). Human immortalized C–28/I2 chondrocytes were kindly provided by Mary B. Goldring (Hospital for Special Surgery, New York, NY, USA) [31]. Both cell lines were maintained in Dulbecco's modified Eagle's medium high-glucose (PAA, Pasching, Austria) supplemented with penicillin (400 U·mL−1), streptomycin (50 μg·mL−1) (both Sigma, Deisenhofen, Germany), and 10% fetal bovine serum (PAN Biotech GmbH, Aidenbach, Germany) and were incubated in humidified atmosphere containing 8% CO2 at 37 °C. Cells were passaged using trypsin-EDTA solution (Invitrogen, Karlsruhe, Germany) at a 1 : 8 ratio (SW1353) or 1 : 6 ratio (C–28/I2) every 4–5 days. For transfection experiments with C–28/I2 cells the culture medium was switched to differentiation medium Dulbecco's modified Eagle's medium/F12 1 : 1 (PAA) supplemented with penicillin (400 U·mL−1), streptomycin (50 μg·mL−1) (both Sigma) and 1% insulin/transferrin (ITS; BD Biosciences, Franklin Lakes, NY, USA).

Plasmid constructs and mutagenesis

AP–2α, -β and -ε expression constructs were obtained from M. Moser (Max-Planck-Institute of Biochemistry, Martinsried, Germany). An AP–2α isoform B expression vector (AP–2α IsoB) was obtained from R. Büttner (University Hospital of Cologne, Cologne, Germany) [26]. A luciferase expression vector (pGL2-Basic; Promega Corp., Madison, WI, USA) containing a 702 bp core promoter region of COL2A1 (−577 bp up to +125 bp relative to the transcription starting point) was obtained from L. J. Sandell (Washington University School of Medicine, St. Louis, USA) and named ‘pCOL2A1–Luc’ in this study [7]. To screen the COL2A1 promoter sequence for consensus AP–2 binding motifs the matinspector program was used (Genomatix Software GmbH, Munich, Germany). Mutation of the three putative AP–2-binding sites within the COL2A1 promoter construct was carried out with the QuikChange II Site-Directed Mutagenesis Kit (Agilent Technologies, Waldbronn, Germany) according to the manufacturer's instructions. Primer pairs used for individual mutagenesis PCR are shown in Table 1. Successful mutagenesis of all constructs was confirmed by DNA sequencing performed at Entelechon GmbH, Bad Abbach, Germany. Plasmids were prepared for transfection employing the Qiagen Plasmid Midi Kit (Qiagen, Hilden, Germany).

Table 1. Primer pairs used for site-directed mutagenesis of putative Ap-2 sites. Base-exchanges are underlined. Location is given in bp relative to the transcription starting point
AP-2 motifPrimerLocationSequence (5′–3′)

Transient transfection and luciferase assay

DNA transfection of SW1353 and C–28/I2 cells was performed using Lipofectamine LTX (Invitrogen, Carlsbad, CA, USA) as suggested by the manufacturer. Cells were cultured in six-well plates for 24 h to a density of ~ 60%. Each cationic lipid⁄plasmid DNA suspension was prepared using 0.5 μg of luciferase reporter plasmid, 0.1 μg of the pRL–TK Renilla luciferase control vector and optionally AP–2 expression plasmids and the corresponding control vector pCMX-PL1, respectively, (for amounts compare individual experiments) in the transfection solutions. Cells were harvested 24 h later, lysed and analyzed for luciferase activity with a luminometer, using Promega dual-luciferase assay reagent. Transfection efficiency was normalized to Renilla luciferase activity. COL2A1 promoter activity was normalized to the control pGL2basic.

siRNA transfection of SW1353 cells was performed using Lipofectamine RNAiMAX (Invitrogen, Carlsbad, CA, USA). Cells were cultured in six-well plates for 24 h to a density of ~ 40% and transfected with 60 pmol of AP–2ε siRNA (Hs_TFAP2E_8; Qiagen) or negative control siRNA (Qiagen), respectively. Luciferase reporter plasmids and pRL–TK Renilla luciferase control vector were transfected into the cells 24 h later, as described above. A further 24 h later the cells were lysed and luciferase activity was measured. Each experiment was repeated at least three times.

RNA isolation, RT-PCR and quantitative real-time PCR

Total RNA of cells was isolated using e.Z.N.A. MicroElute Total RNA Kit (peqlab Biotechnologie GmbH, Erlangen, Germany) as described by the manufacturers. Purity and concentration was measured in a NanoDrop (peqlab Biotechnologie GmbH). cDNA was generated by RT. The RT reaction was performed in 20 μL reaction volume containing 500 ng of total RNA, 4 μL of 5× first-strand buffer, 2 μL of 0.1 m dithiothreitol (both Invitrogen), 1 μL of dN6 primer (10 mm) (Roche, Mannheim, Germany), 1 μL of dNTPs (10 mm) (Amersham Pharmacia Biotech, Pittsburgh, PA, USA). The reaction mix was incubated for 5 min at 70 °C and 1 μL of Superscript II reverse transcriptase (Invitrogen) was added subsequently. RNA was transcribed for 1 h at 37 °C. Finally, reverse transcriptase was inactivated at 70 °C for 10 min, and RNA was degraded by digestion with 1 μL RNase A (10 mg·mL−1) (Roche Applied Science, Mannheim, Germany) at 37 °C for 30 min.

Quantitative RT-PCR was carried out with the Lightcycler480 system from Roche. A volume of 1 μL cDNA template, 0.5 μL of forward and reverse primer (20 mm) for each of human collagen type II, AP–2ε or β–actin, 10 μL of SYBR-Green Premix (Roche) and 8 μL water were combined to a total volume of 20 μL. PCR primers were obtained from Sigma (Table 2). The following PCR program was used: 95 °C for 10 min (initial denaturation); 4.4 °C·s−1 temperature transition rate up to 95 °C for 10 s; 60 °C for 10 s; 72 °C for 20 s, 80 °C acquisition mode single, repeated for 45 times (amplification). The PCR product was evaluated by melting-curve analysis. Each sample was analyzed in duplicate. The expression ratios of the analyzed genes were normalized to the expression level of the housekeeping gene β-actin.

Table 2. Primer pairs used for quantitative real-time PCR
GenePrimerProductSequence (5′–3′)
β-Actin hβ-Act_735for 384 bp CTACGTCGCCCTGGACTTCGAGC
Col2A1 hCol2_4452for 296 bp AGGGCAATAGCAGGTTCACG

Preparation of nuclear extracts

Nuclear extracts of SW1353 cells transfected with 50 ng of AP–2ε expression plasmid per six-well plate for 24 h were prepared with the method described by Dignam et al. [32].


EMSA was performed as described previously [23]. This method is based on the binding of nuclear AP–2ε to 32P-labeled oligonucleotides containing an AP–2-binding site derived from the analyzed promoter construct of COL2A1. A double-stranded oligonucleotide was created and endlabeled using T4 polynucleotide kinase and [α32P]: COL2a1_oligo: 5′–GGCAGGGCCCAGGCGGGCTC–3′.

This fragment corresponds to a promoter region −141 to −122 upstream of the transcription starting point of COL2A1 spanning the third AP–2ε binding site (AP–2.3). For competition, unlabeled oligonucleotides were added at a 400–fold molar excess. To demonstrate specific binding of AP–2ε a specific antiserum against AP–2ε [20] was used.

ChIP assay

The ChIP assay was performed using the ChIP-IT™ Express kit following the manufacturer's instructions (Active Motif, Carlsbad, CA, USA) and as previously described [23]. Chromatin isolation was performed with chondrocytes from osteoarthritis patients, in which AP–2ε expression was shown to be upregulated compared with normal chondrocytes [23]. Samples were immunoprecipitated with a specific primary AP–2ε antiserum. An RNA polymerase II antibody and an IgG antibody were used as controls, following the protocol provided with the control kit (ChIP-IT control Kit-human; Active Motif). DNA samples from the ChIP experiments were used for analysis by PCR.

PCR was performed on four DNA templates: The input DNA (1 : 5), DNA isolated through RNA polymerase II ChIP (Pol II), DNA isolated through the negative control IgG ChIP (IgG), and DNA isolated through the AP–2ε ChIP (AP–2ε). A control reaction with no DNA template was also performed (H2O).

Five sets of specific primer pairs were used for the PCR: The positive control GAPDH and the negative control primer pairs provided in the kit, as well as primer pairs spanning the putative AP–2ε binding sites within the promoter of COL2A1. The latter are shown in Table 3. PCR fragments were analyzed on a 1.5% agarose gel.

Table 3. Primer pairs used for PCR analysis of DNA samples from ChIP experiments. Primer pairs span the putative AP-2ε binding sites within the promoter of COL2A1. Location is given in bp relative to the transcription starting point
AP-2 motifPrimerLocationProductSequence (5′–3′)
AP-2.1COL2a1_ChIP_for3−427 to −447212 bp CTCCGCTGCTCCTTTCTACC
AP-2.2 COL2a1_ChIP_for5 −255 to −272101 bp GCAGCCCCGGGTTTGGG
AP-2.3 COL2a1_ChIP_for2 −173 to −193106 bp GAAGCGTGACTCCCAGAGAG

In vitro translation and western blotting

For generation of in vitro translated AP–2ε the TnT® T7 Quick Coupled Transcription/Translation System (Promega) was used. Briefly, 1 μg of AP–2ε expression plasmid and the respective control vector pCMX was added to a rabbit reticulocyte lysate containing all components necessary for in vitro translation including a T7 RNA polymerase. After incubation for 90 min at 30 °C western blot analysis was performed to control successful translation. Here, 4 μL of the reticulocyte lysates were denatured at 70 °C for 10 min after addition of Roti-load-buffer (Roth, Karlsruhe, Germany) and subsequently separated on a 12% SDS/PAGE gel. After blotting onto a polyvinylidene difluoride membrane (Bio-Rad, Richmond, CA, USA) and blocking for 1 h with 4% BSA/TBST the membrane was incubated overnight at 4 °C with a specific primary AP–2ε antiserum. After three washing steps with TBST (0.1% Tween) the membrane was incubated for 1 h with an alkaline phosphate-coupled secondary anti-rabbit IgG (Chemicon, Hofheim, Germany) and then washed again. Finally, immunoreactions were visualized by BCIP/NBT (Sigma) staining.

Statistical analysis

Results are expressed as mean ± SEM. Comparison between groups was made using the Student's paired t-test. A P–value < 0.05 was considered statistically significant (*). A P–value < 0.01 is depicted by ** and a P–value < 0.001 is depicted by ***. All calculations were performed using the graphpad prism software (GraphPad software Inc., San Diego, CA, USA).


This work was supported by grants from the Deutsche Forschungsgemeinschaft to AKB. Special thanks go to M. Moser, R. Büttner and L. J. Sandell for providing us with AP–2 expression plasmids and the COL2A1 promoter construct, respectively. Additionally, we would like to thank M. B. Goldring for the C–28/I2 cell line.