In this study, we further characterized two forms of CTCF, CTCF-130 and CTCF-180, with regard to their PARylation status. CTCF-130 was previously suggested to be the non-PARylated form of CTCF (92
) whose predominance in all cellular contexts has been well established (26
). Further detailed analysis of CTCF-130 and CTCF-180 revealed the complex nature of CTCF PARylation. Both forms appear to be PARylated, although to different extents (Fig. ): CTCF-180 is likely to be highly PARylated, whereas CTCF-130 may contain only a few ADP-ribose residues. Two lines of evidence support this proposition: (i) CTCF-130 can be detected only by the rabbit polyclonal antibody (Fig. ), and (ii) PARylated CTCF-130 comigrates with the non-PARylated form of CTCF (Fig. ). These data demonstrate that the CTCF-130 band seen upon SDS-PAGE represents different populations of CTCF, which may contain mono(ADP-ribose), oligo(ADP-ribose), and unmodified CTCF; similar observations have been made in a recent report (91
). Our findings suggest that the two PARylated forms of CTCF, CTCF-130 and CTCF-180, may be involved in functional regulation in different cell contexts (see below).
Intriguingly, PARylation of CTCF-130 with PARP-1 (92
) and enzymatic degradation of CTCF-180 with PARG (Fig. ) result in the appearance of discrete bands of CTCF-180 and CTCF-130, respectively. The absence of a discernible CTCF protein smear, indicating varying modification levels, suggested that a distinct mechanism may contribute to the dramatic increase/decrease in the molecular mass of CTCF following PARylation and de-PARylation. This phenomenon requires further investigation.
The coexistence of differentially ADP-ribosylated forms of CTCF supports the current view of the biochemistry of the PARylation reaction, which requires preformed mono(ADP-ribosyl)ated substrates. Several enzymes, such as mono(ADP-ribosyl)transferase, PARP-1 itself, other PARP family members, and heteromers of these proteins, may mono(ADP-ribosyl)ate CTCF, which can then be used for the PARylation reaction (43
Although CTCF-180 usually represents only a minor fraction of the total CTCF population, our recent study shows that CTCF-180 is the predominant isoform in many normal human tissues and the only isoform in normal breast tissues, with CTCF-130 associated with cell proliferation, whereas CTCF-180 is associated with nondividing cells (27
). In this context, the low levels of endogenous CTCF-180 in the cell lines, as well as the absence/very low levels of CTCF-180 produced from the exogenous CTCF-expressing constructs (reference 40
and this study), may be explained by rapid degradation of PAR, typical for proliferating cells in culture, where the half-life of PAR is less than a minute (5
To specifically examine the role of PARylation in CTCF functions, we generated a CTCF mutant deficient in PARylation by mutating the cluster of glutamic acid residues, the preferred acceptors of PARylation, in the N-terminal domain of CTCF (Fig. ). To our knowledge, this is the first example of the generation of a mutant protein completely deficient in PARylation in vivo
; a recently described PARylation-deficient p53 mutant still demonstrated residual in vivo
). The non-PARylated CTCF Mut4 was identical to the CTCF WT by the following criteria: (i) protein size of 130 kDa upon SDS-PAGE, protein stability, and similar protein levels produced from the plasmid vectors (Fig. , , and and data not shown), (ii) localization and distribution patterns in the nucleus (Fig. ), and (iii) ability to bind to DNA (Fig. and ). These findings are in contrast with the reports that PARylation alters the stability and DNA binding activity of proteins (81
In the present study, in all experimental systems tested, the differences in CTCF PARylation resulted in the loss of CTCF function. In the presence of PJ34, a potent PARP inhibitor, or transfection with the plasmid expressing CTCF Mut4, the activity of the p19ARF promoter-driven reporter was considerably reduced, suggesting that CTCF function as a transcriptional regulator is PARylation dependent. Notably, the effects of PARylation in cell lines of different origins seem to vary, thus indicating that cell context may play a “tuning” role in CTCF regulation.
The inhibitory role of CTCF in cell proliferation was also compromised by CTCF Mut4 (Fig. ). It is conceivable that this regulation occurs through transcriptional regulation of CTCF target genes controlling proliferation, such as p19ARF
), and possibly p21
). Recent studies also implicated CTCF in the regulation of ribosomal biogenesis (84
) and replication timing (12
). Thus, since CTCF effects on cell proliferation are likely to involve various regulatory pathways, alterations in CTCF PARylation may lead to global changes in cell biology.
In our previous investigations, we used similar assays to assess the effects of CTCF phosphorylation by protein kinase CK2 on CTCF functions, which generally resemble those of CTCF PARylation (29
). Cross talk between the two pathways may occur and may explain the similar functional outcomes; these observations require further examination. Functional regulation by both modifications, PARylation and phosphorylation, is not unusual and has been reported previously for p53 and histones (30
Finally, the insulator function of CTCF, tested using a stable isogenic insulator reporter cell system and a mouse-human hybrid cell line, was also perturbed by PJ34 and CTCF Mut4 (Fig. ). Using this system, we explored a hypothesis that CTCF functional activities regulated by PARylation are facilitated mechanistically by the close proximity of CTCF and PARP (Fig. ). In this model, PARylation of CTCF is established by PARPs present at the same DNA site and can be reversed by PARG. We were not able to establish the presence of PARG at the insulator or association with PARP-1 and CTCF due to the absence of ChIP-grade anti-PARG antibodies; however, interaction between PARP-1 and PARG has been reported in the literature (36
In the proposed model, a feedback axis can be envisaged whereby CTCF can directly regulate PARP-1, which in turn may lead to DNA hypomethylation (40
). Cooperation between CTCF and PARP-1 may provide a link between PARylation and DNA methylation, thereby adding another layer of complexity to the model. These aspects deserve further investigation.
The notions that CTCF is stably bound to its DNA targets and that regulatory signals are responsible for the modulation of CTCF activity are supported by previously published observations (63
). However, such close association between CTCF and PARP-1 may not be required for CTCF regulation by PARylation at all CTCF targets (91
The genome-scale association between CTCF and the poly(ADP-ribosyl)ation mark was shown previously with the mouse microarray library of CTCF binding sites (71
). In the present investigation, considerable overlap between CTCF and PARP-1 in the nuclei in different cell lines was confirmed (Fig. ; see also Fig. S7B in the supplemental material). Furthermore, bioinformatics analysis of the distributions of CTCF (13
) and PARP-1 (59
) in the ENCODE regions of the genome in human cell lines revealed sizable areas of overlap between CTCF and PARP-1 on different chromosomes (Fig. ; see also Fig. S8 in the supplemental material), with three of the identified CTCF/PARP-1 colocalizaton sites validated (Fig. ). Although the ChIP-ChIP data for CTCF and PARP-1 were obtained from different cells lines, previous reports indicate that CTCF occupancy at its binding sites is conserved among different cell lines (9
In this context, it is important to acknowledge the substantial overlap between CTCF and cohesin binding sites throughout the genome, which is believed to have important functional implication in the regulation of gene expression and is the subject of much current research (references 38
, and 74
and references therein). The results of our study, indicating the frequency of colocalization of CTCF and PARP-1, lead us to hypothesize that CTCF, PARP-1, and cohesin may all appear in one complex. This hypothesis deserves further investigation.
A common theme of this study is that CTCF and PARP-1 are important elements of the same regulatory pathway(s) and likely to be linked by a feedback mechanism. Therefore, PARP-1 dysfunction may lead to loss of CTCF PARylation and, as a result, dramatic changes in gene regulation and development of disease, in particular cancer. Alternatively, deregulation of PARP-1 by aberrant CTCF protein may also contribute to tumorigenesis. In this context, it is significant that both CTCF and PARP-1 display features of tumor suppressors (2
). However, the interrelationship between CTCF and the PARylation enzymatic machinery appears to be more complex, as CTCF and PARylation enzymes are likely to be shared by different functional complexes. Such sharing may be important for functional integration of different cellular processes in response to various stimuli.