On a technical level, insertion of SPF motifs into the desired target regions using CRISPR/Cas9 has been successful. In most cases, this insertion has led to local demethylation and in some cases, it has also led to an increase in target gene expression. Since overall, both demethylation and activation of gene expression were met with heterogeneous outcomes and neither has been achieved in a robust manner, it is clear that our efforts cannot be regarded as a robust proof-of-concept for this approach.
It is important to note that the motifs of CTCF and SOX2 were previously tested experimentally for their ability to specifically and efficiently bind to their corresponding PFs, either by ChIP experiments, or DNA/protein microarray and EMSA. Therefore, we made the assumption that the TFs bind to the WT motif in our setting. However, one could confirm that this event is indeed taking place by performing chromatin immunoprecipitation (ChIP) experiments, assessing CTCF or SOX2 enrichment at the target regions.
In case of TSG promoter targeting, insertion of the wild-type CTCF motif has led to, albeit sometimes modest, local demethylation, which could not be observed for the scrambled motif (Fig. 3.5 and Fig. 3.6). Only for Rassf1a, which also experienced the largest extent of demethylation, we could observe an increase in mRNA expression, that could not be confirmed on the protein level. It is unknown whether this is due to regulatory events at the post-transcriptional level, or simply because the amount of protein is below the detection limit. For p16 we could not detect any increase in gene expression. Some possible explanations could be insufficient extent or bad positioning of demethylation at the target region, steric hindrance due to CTCF occupancy, TF motif disruption due to motif insertion, or absence of other key TFs
131 that are needed for activation. However, the fact that we obtained a moderate p16 reactivation with dCas9-TET1 in the same cell line indicates that the conditions to obtain a detectable level of activation, including the presence of TFs necessary for activation, are present (Fig. 3.1). In comparison to CTCF motif insertion, targeting with dCas9-TET1 resulted in more widespread demethylation, which also included the transcription start site (TSS). This is probably due to the fact that multiple copies of the effector protein were spread out over the promoter region, while the insertion just occurred in one spot. This activation, while significant, was relatively low and not detectable on the protein level, so one should be careful when using it as a reference point. CTCF occupancy has yet to be confirmed by ChIP, however, steric hindrance is probably lower in comparison to the recruitment of up to four larger proteins in form of dCas9-TET1.
On the other hand, CTCF recruitment could also lead to the disruption or the establishment of new DNA loops, which may influence gene expression. The importance of a CTCF-mediated chromatin boundary site about -1 kb from the TSS has previously been shown in regards to regulating p16 expression, with its loss resulting in gene silencing in various cancer cells lines661. Similar results have been observed for a boundary site around 1.8 kb upstream of the Rassf1a gene661. Changes in chromatin architecture, such as new loop formations upon CTCF motif insertion, could be assessed via circular chromosome conformation capture (4C) or related techniques.
Using CTCF, only insertion of the reverse wild-type motif into the Dazl part of the Dazl-Snrpn reporter lead to an increase in GFP expression (Fig. 3.9 and Fig. 3.11). While this was accompanied by some degree of demethylation, moderate rates of demethylation have also been observed in clones with scrambled or other wild-type motif insertion that did not show an increase in GFP expression. One extreme example was the insertion of the wild-type CTCF motif in forward direction, which resulted in two clones with identical sequences, but two vastly different bisulfite profiles (Fig. 3.9F) and no apparent change in expression. For SOX2 motif
132 insertion into the Snrpn part, the clone with the wild-type motif in forward direction showed the strongest increase in GFP expression, while retaining complete methylation. At the same time, the clone with the scrambled motif experienced a slight decrease in methylation but no changes in expression. These findings open up several questions: (1) Is the observed demethylation caused by the SPFs? (2) Is demethylation necessary for the activation of the Snrpn-Dazl reporter? (3) Is the observed increase in gene expression facilitated through other functions of the SPFs? – It has been shown that the insertion of unmethylated fragments can lead to demethylation of neighboring sequences, which may be an alternative explanation for the first question. Moreover, a part of the answer for the second question is provided by the fact that we have shown that 5-aza treatment results in demethylation and activation of Snrpn-Dazl reporter, which indicates that demethylation is sufficient, but not whether it is necessary, for the activation of the reporter. Indeed, other 5-aza-dependent demethylation events occurring at different locations might be necessary for this activation. Both CTCF and SOX2 are known to interact with a variety of other proteins influencing gene expression and might act through these mechanisms. As mentioned above, CTCF is an architect of chromatin topology, and may act through loop formation. SOX2, has been shown to recruit Krüppel-like factor 4 (KLF4), another identified SPF, as well as to interact with some other 143 proteins, several of which are transcriptional activators and coactivators662,663. Taken together, the results at their current state are not sufficient to support a proof-of-concept for SPF motif insertion-mediated reactivation of gene expression via demethylation. Further ChIP and 4C experiments, on the other hand, should elucidate the mechanisms at play.
Interestingly, WGBS in VAT either from RT or CAx treated mice, revealed many DMRs, which were hypomethylated after CAx treatment. This result is in concordance with previous studies using reduced representation bisulfite sequencing (RRBS) and RNAseq in BAT and WAT, and that identified mostly hypomethylated DMRs in BAT versus WAT at the identified DEG
133 promoters420. In addition, injection of mice with 5-aza has resulted in an increase in UCP1 and Ppargc1α expression in SAT, as well as in a decrease in inflammation factors IL1β and TNFα664. None of our identified DMRs, however, overlapped with a DEG promoter region, nor, after looking at H3K4me1 and H3K27ac marks in WAT or BAT, with any apparent enhancer. As of yet, no candidate genes for epigenome editing in VAT have been revealed. This may change by changing the parameters for DMR recognition or by reducing the heterogeneity of the sample with above-mentioned FACS methods.