Dr. Marianne Rots, Professor of Molecular Epigenetics and Head of Epigenetic Editing Laboratory at University of Groningen, Netherlands provides an informative overview of the current state of epigenetic editing approaches.
Intro to Epigenetic Editing
Epigenetic editing exploits the two main characteristics of epigenetics. On the one hand, epigenetic marks are heritable. This means the epigenetic signature of DNA methylation and certain histone modifications are copied with every cell decision by endogenous cellular mechanisms. On the other hand, epigenetic marks are reversible. The cell can respond to external signals by adapting these marks. The effectors of both maintaining, as well adjusting epigenetic signatures, are the epigenetic enzymes, so-called writers or erasers.
These enzymes do not, by themselves, recognize DNA sequences but are recruited by, for example, transcription factors. Epigenetic marks control gene expression levels. For actively expressed genes, these marks include histone acetylation or certain other histone modifications, including H3K4 trimethylation.
Silenced genes, on the other hand, are often characterized by DNA methylation or H3K9 or K27 trimethyhlation. To rewrite the epigenetic signature for this particular gene, a DNA binding domain is engineered to recognize and bind a stretch in the expression regulation control region of the gene. Such DNA binding domains can subsequently be used to target epigenetic effector domains to the gene of interest.
This fused effector domain is a writer of repressive marks. The introduction of more repressive marks will prevent a gene from being expressed. An eraser of repressive marks lowers the amount of repressive marks and should increase gene expression.
Epigenetic editors are enzymatic domains of epigenetic writers or erasers fused to gene-targeting designer DNA binding domains. For effector domains, one can choose from an increasing list of candidates. Also for designer DNA binding domains, several options are available– both chemical as well as biotechnological.
Epigenetic Readers & Writers
Here are some examples of effector domains, with a focus on DNA methylation. The first attempt towards epigenetic editing was published by Xu and Bestor in Nature Genetics back in 1997. In this paper, induced DNA methylation was demonstrated for a predetermined genetic sequence in an oligonucleotide context. As the effector domain, a prokaryotic DNA methyltransferase, M.SssI, was fused to the DNA binding domain.
As advantages of the approach, Xu and Bestor mentioned that the silencing effect was likely to be maintained.
This would imply that the treatment agent does not need to be present continuously, providing an efficient hit and run approach. This certainly is an advantage over, for example, siRNA, which generally does not silence the source of production of the RNA. For permanent effects, siRNA does require potentially harmful integration of its expression cassette in the host genome.
Moreover, induced DNA methylation is expected to recruit endogenous gene silencing machinery. As such, the whole cell might reinforce the approach. Therefore, complete inhibition is more feasible compared to other inhibition approaches. Also, because the target pool, which in general is only two DNA copies per cell, is very small.
However, Xu and Bestor also alluded to the importance of specificity. And as also was shown by follow up studies, the activity of M.SssI cannot be restricted through fusion to its DNA binding domain. This high catalytic activity of M.SssI one disqualifies the enzyme as an effector in locus-specific epigenetic editing. Despite the finding that the activity was lowered when infusion, other approaches are required to increase specificity for epigenetic editing purposes.
To improve specificity, the so-called split enzyme approach might prove beneficial as introduced for DNA methyltransferases by Barbas in 2007. In this approach, an enzyme is cut in two parts, which on their own do not demonstrate catalytic activity anymore. However, if the two parts meet, for example, by co-targeting to two neighboring sites in the genome, the reassembly of the two parts results in an active complex.
Another case, from Hungary, has shown this to be feasible for M.SssI, which we are currently following up for epigenetic editing purposes. Another approach to increase specificity of epigenetic editing is to severely cripple the enzyme so that only upon prolonged presence of the enzyme at a predetermined site the enzyme has sufficient time to write its mark.
Initial Epigenetic Editing Experiments
Within the setting of a European Union funded project, we now set out to exploit M.SssI also for this purpose. First we aimed to demonstrate that the prokaryotic M.SssI enzyme is effective in the human setting. In collaboration with Synvolux Therapeutics we developed cellular delivery devices, which are able to actively deliver M.SssI into the nucleus of living human cells. We published this is back in 2008.
If Synvolux delivery agent, is used, no DNA methylation is induced. And all tested CpGs are unmethylated, similar to the untreated control situation. Upon active delivery of the wild type enzyme, using the Synvolux delivery agent, genome wide DNA methylation was induced, and also for our gene of interest. This methylation resulted in a decrease in protein expression.
So with this we validated that M.SssI one is a very active DNA methytransferase. That it can methylate CpGs, also in a mammalian context, and that we could actively deliver it into mammalian cells. Then, together our top case, we set out to construct a catalytically crippled version of the enzyme, so-called C141S. Delivery of the enzymatically crippled mutant was indeed far less effective in inducing DNA methylation. Importantly, C141S was less toxic for the treated cells. So this crippled mutant might provide an improved writer for targeted DNA methylation.
Mammalian DNA Methyltransferase Enzymes
What about mammalian DNA methyltransferase enzymes? They might provide adequate tools for epigenetic editing, because they have a lower intrinsic activity. In addition, they would face less immunogenecity issues. We set out to use mammalian DNMT3 enzymes as effector domains in epigenetic editing.
Indeed, we showed that DNMT3 can be used to induce targeted DNA methylation on the genes and coding for SOX2, for Maspin, and for VEGF-A. These findings resulted in the first two papers on endogenous, gene-targeted DNA methylation. This finding was recently validated for yet another endogenous gene, the previously introduced EpCAM gene, by using DNA binding domains, which had been engineered in my lab.
Epigenetic Editing –Erasers
What about erasers? For histone marks, various enzymes have been known to remove the modifications. In contrast, DNA methylation was considered a very stable epigenetic mark until recently. The mark was thought to be erased only by inhibiting the DNA methyltransferase enzymes during cell division, so-called passive DNA demethylation. Also, some indications were put forward that DNA repair enzymes could be used to replace the methylated C by an unmethylated C.
Federov’s group from Harvard indeed recently validated this concept for an enzyme called TDG. Obviously, for permanent effects, epigenetic editing would indeed benefit from enzymes, which can actively remove, or modify, the methyl group of methylated cytosines without cell divisions. Excitingly, in 2009, mammalian enzymes were described which could actively start a DNA demethylation process, the Ten-eleven Translocation (TET) enzymes.
TET enzymes work by converting methyl cytosine to 5 hydroxymethylcytosine, and other modifications which have completely different biological functions. Excitingly, we and another group convincingly showed that TET enzymes indeed remove DNA methylation on the hypermethylated low side. This opens up exciting avenues for hypermethylated genes to become re-expressed without re-expressing unintended genes.
Designer DNA Binding Domains
DNA binding domains can be engineered to preferentially bind to DNA sequences at will by using chemical, as well as biotechnological approaches. The chemical approaches include triplex-forming oligonucleotides, peptide nucleic acids, and polyamides. TFOs are single-stranded DNA molecules, which can be engineered to bind particular DNA sequences in the major groove. PNAs are derivatives thereof. Polyamides are small chemicals, which combine predetermined sequences in the minor groove.
Although these approaches have shown promise for gene expression modulation, they face some limitations including the difficulties of fusing effector domains to it. This limitation is not true for protein-based targeting approaches. And engineer DNA binding protein-based approaches do receive a lot of attention these days. Despite the difficulty of fusing protein domains to chemical DNA-binding domains, my collaborators Antal Kiss and Elmar Weinhold succeeded in coupling the catalytically crippled M.SssI version called C141S to a triplex forming oligo. This oligo was designed to bind our gene of interest, EpCAM gene.
Plasmid experiments show that we could induce CpG specific DNA methylation by treating plasmids with the TFO constructs. After treatment, 18 out of 24 clones analyzed– so representing 75%– showed methylation at the targeted CpG, number 7. Of these clones, four also showed methylation at CpG number 5, on the other side of the targeted side. This project is EU-financed for further optimization of CpG-targeted methylation.
Despite such progress using chemical DNA binding approaches, biotechnological re-engineering of DNA binding domains of endogenous transcription factors has seen an explosion in attention within the past two decades. The most advanced is the field of engineered zinc finger proteins. Zinc finger proteins are the most abundant class of transcription factors in the human body. Their DNA binding domain consists of individual modules, the fingers.
Each finger is a small, 30 amino acid peptide. In general, only three amino acids decide to which DNA sequence one finger binds. Fusing three or six fingers together yields more sequence-specific proteins. A six finger zinc finger protein can recognize and bind a stretch of 18 base pairs which, by mathematics, could provide a unique address in the human genome.
Stretches of seven amino acids to be grafted into a finger has been described for the majority of various triplets of base pairs, as sets of designer zing finger proteins can be easily obtained for any promoter of choice through websites like zincfingertools.org. Zinc finger proteins are derived from mammalian systems, and thus have an advantage with respect to adverse immunological reactions upon delivery in organisms. Indeed, clinical trials using engineered zinc finger proteins did not detect any harmful events.
However, the field has taken up recently more straightforward gene targeting approaches. The transactivation like elements, derived from plant pathogens, and CRISP-Rs, which play a role in the normal bacterial defense system. Especially the CRISP-R system, has revolutionized the DNA targeting field, as it provides a cheap, flexible, and easy approach to target any effector domain to intended loci in the genome.
History of Epigenetic Editing
Epigenetic editing combines two active research fields, addressing specific loci within the human genome, so-called gene targeting, and the exploding field of epigenetics, which yields candidates for writers and erasers. When did these two fields meet?
The realization that any gene could be repressed from its endogenous locus has started over 25 years ago. First indications that genes could be repressed by interfering with the endogenous transcription machinery by DNA binding domains was obtained for TFOs in 1988 and for zinc finger proteins in 1994. The first example of an engineered zinc finger protein was reported by Sir Allan Klug, who had received the Nobel Prize for protein nucleic acids interactions earlier, in 1982.
Quickly it was realized that adding effector domains could improve efficacy and fusing inhibitors of epigenetic enzymes is one way to go. However, a more straightforward way to modulate gene expression was to fuse transcriptional repressors and even activators to DNA binding domains, to allow bidirectional down and up regulation. The first in vivo demonstration of such gene activation was reported for engineered zinc finger proteins in 2002. This was taken to the clinic shortly after that by Sangamo Biosciences in 2004.
Actually, this steadily increasing list of endogenous genes being modulated by so-called artificial transcription factors, clearly demonstrate the feasibility of modulating expression of any gene at will. The drug-able genome concept claims that all genes can be exploited as therapeutic targets, instead of the less than 2% of protein coding genes for which currently drugs are available. Surely, transcription factors engineered to target any gene at will provide a way to realize this concept.
For down regulation of gene expression, small interfering RNA approaches are used extensively in research and also tested in clinical trials. However, the messenger RNA molecules to be degraded are continuously expressed. Targeting the source of RNA production, the DNA, as obtained by artificial transcription factors, might give an advantage with respect to effectiveness.
For increasing gene expression, cDNA gene therapy can be used. However, this approach is hampered by the choice of cDNA isoforms to transfer, and by the size of some cDNAs. In general, the promoters used to drive cDNA expression cannot be easily controlled. As ATFs induce expression from the endogenous locus, all isoforms are likely to be expressed in their natural ratios. Despite the advantages of Artificial Transcription Factors over the conventional approaches, ATFs do not possess enzymatic activities, so their effect on expression modulation is considered transient.
For siRNA, cDNA, and artificial transcription factors to result in stable effects, potentially harmful integration of the expression cassettes in the host genome is required. In addition, for up regulation of gene expression, these approaches cannot control the degree of up regulation. So why not fuse an epigenetic enzyme to engineered DNA binding domains?
If one could induce a solid set of epigenetic marks, these are likely to be copied with every cell division. One series of treatments will edit the intended marks, and the cell will take over and possibly even spread the mark to further increase efficiency. This way, epigenetic editing provides a hit and run approach for sustained effects, without the need for integration in the host genome.
With respect to up regulation, as only the chromatin is affected by epigenetic editing, the cellular signaling mechanisms are required to induce and control the expression level of the targeted gene. This implies that epigenetic editing is uniquely suited to closely mimic nature. Epigenetic editing does better fulfill the requirements to fully exploit a drug-able genome, potentially even introducing a cure for the incurable concept.
Now, what are the indications that epigenetic editing actually will work? First of all, a lot of research efforts have been ongoing in the last decades to target epigenetic enzymes to artificial reporter genes. These proof of principal studies clearly show that epigenetic marks can directly instruct gene expression. We have reviewed such efforts in our NAR review, 2012, where we also listed all epigenetic effector domains which have been targeted to predetermined sequences.
Of particular interest is a seminal study where H3K9 methylation was induced on the endogenous VEGF-a gene, resulting in gene repression. This study was published back in 2002 but did not receive a lot of attention. That is likely explained by the timing. The book of life, our DNA sequences, was decoded in 2000. And variations in people’s DNA were extensively investigated to predict susceptibility for diseases.
Now all attention was focused on DNA sequence, and not at all on DNA function. And a technology was needed to interfere with genetic information. For this genome editing, genome surgery approaches, the DNA binding domains just described to you were exploited. Towards this end, DNA binding domains were fused to nucleases, which are some kind of DNA scissors.
They can site specifically introduce DNA damage. The cell will try to repair the damage. And such repair mechanisms can be used to inactivate genes. This method of gene editing was received with great enthusiasm, and declared method of the year 2011. And breakthrough runner ups 2012 by Science. Indeed, Sangamo Biosciences also shifted focus from the Artificial Transcription Factor approach to targeted nucleases and initiated genome editing clinical trials based on engineered zinc finger proteins.
As also described in this recent publication, no adverse effects of the engineered zinc finger proteins were detected in a very promising study on gene editing to inactivate the HIF receptor. More recently, the focus has been diluted away from the DNA sequence. And DNA function is getting into the picture. The encyclopedia of DNA elements, ENCODE for short, gained enormous attention by claiming that the majority of DNA is being expressed somewhere, somehow.
Now we witnessed the demand for technologies to address the function of DNA. And epigenetic editing provides an exciting research tool to address questions that cannot be answered using conventional technologies.
Moreover, many diseases are associated with disturbed gene expression profiles, be it protein coding or noncoding genes. The approach of epigenetic editing allows modulating of expression of any given gene thereby increasing the drug-ability of the genome. With the advancements in DNA binding, especially the introduction of CRISP-Rs, the technology becomes feasible to many labs and has been declared a method to watch in 2014. So altogether it seems that epigenetic editing is at the verge of a breakthrough.
Pioneering Epigenetic Editing Discoveries
I will now walk you through the exciting, pioneering discoveries on epigenetic editing to date. Before doing so, it is important to realize that only a few years ago I encountered much disbelief when advocating the epigenetic editing concept. One important dogma, which was thought to hamper re-expression of sleeping genes, was that a gene that is not expressed was assumed to be packaged in tight heterochromatin configuration. This is not accessible to transcription factors.
We, and others, have described many examples of Artificial Transcription Factors being capable of re-expressing such silenced genes. Actually, in my projects I have not encountered a single epigenetically silenced gene that could not be re-expressed by Artificial Transcription Factors. These studies clearly show that heterochromatin is not preventing accessibility, per se, although larger DNA binding domains or effector domains are somewhat hindered in accessing their target gene.
Another dogma some people strongly believed in states that that epigenetic marks do not instruct gene expression but merely reflect the expression status of a gene. This dogma can only truly be addressed by epigenetic editing. And a 2002 paper clearly shows that H3K9 methylation does instruct gene repression.
To convincingly demonstrate that it is indeed the mark that instructs the gene activity, experiments always need to include a catalytically dead mutant to address the indirect contribution of recruited effectors. The current papers on epigenetic editing do indicate that the effects are through catalytic activity and not indirect via recruitment of other players.
So epigenetic editing provides a unique tool to address many unanswered questions in chromatin biology, the cause-versus-consequence issue, which seems resolved for H3K9 and DNA methylation, might not be valid for all marks or in all chromatin environments. Many more indications on epigenetic marks instructing gene expression have been obtained, as previously reviewed by us. But what is the effect of the artificial context which have been mainly used in such studies? Also we know that epigenetic marks and enzymes interact and reinforce one another. But what is the order of events? How can we ensure our sustained defense?… just to name a few open questions.
Epigenetic editing on endogenous genes has now been described in 10 papers. In three studies, we have shown that DNA methylation can be induced on endogenous genes using mammalian DNMT3. And that this resells in gene repression for four different genes.
In the Rivenbark paper, we even provide some indications of heritability. For active DNA demethylation, three papers now demonstrate feasibility of this to induce gene re-expression. The Maeder study targeted TET 1 while Gregory and Federov used the TDT mismatched repair enzyme to remove methylated C.
With respect to histome modifications, we validated the 2002 VEGF-a study by targeting G9A and sub 39H1 to her2/neu. And also, this overexpressed oncogene, we could induce gene repression. Konermann and Mendelhall targeted various histone modifiers to core promoters or to enhancers, to successfully induce gene repression.
In these papers, we therefore used engineered zinc finger proteins as DNA binding domains, which are currently the most advanced towards clinical applications.
The other publications focus on chromatin function, and mainly exploit tails as DNA binding domains, as also highlighted in this Nature Biotechnology News & Views. These three papers and the Gregory paper highlight potency of epigenetic editing as research tools to address basic research questions, including functional validation of enzyme and obtaining insights into gene expression regulation. The introduction of the CRISP-R technology will allow widespread application of the approach and many innovative insights are expected.
For example, in my lab we now construct cell line sub panels, where genes are activated or repressed at will, either constitutively or inducibley. This will allow mechanistic insights and complex interplays, for example between oxidative stress, epigenetics, and carcinogenesis.
Epigenetic Editing in Translational Biomedical Research
In recent years, huge efforts have been devoted to identify epigenetic mutations associated with diseases. And the field has witnessed initiations of interesting subdisciplines, including the study of effects of diets, or metals on epigenetics in nutritional and environmentalist epigenetics. For all subdisciplines it would be of great importance if we could step beyond the diagnostic marker application and start validating the biological consequences of the observed epi-mutations. This would not only establish cause versus consequence, but it would also allow exploitation of epigenetic mutations as future therapeutic targets.
The field of clinical epigenetics is obviously expanding quickly. Specialized journals, meetings, and summer schools are organized to ensure synergistic interactions between researchers of various research fields. The most advanced knowledge, with respect to epigenetics and disease has been gained for cancer. Epi-mutations include genome-wide blocks of DNA hypomethylation, potentially underlying genome instability. And local DNA hypermethylation results in tumor suppressor gene inactivation. Also, these regulated histone modifications are observed very frequently. The field has been trying to identify epi-mutations to be exploited as early and sensitive diagnostic markers. But epigenetics has also moved to clinical practice, with FDA-approved inhibitors of epigenetic enzymes for treatment of hematological malignancies.
From this field, we also know that epi-mutations not only affect protein-coding genes. Many epi-mutations are observed on yet unknown loci. By mimicking such mutations, target genes can be identified, or effects on cellular physiology can be established. Projects in my team have focused on hypermethylated DNA targets, which were investigated for use as diagnostic markers. We set out to functionally validate the effect of the epigenetic mutations by enforcing up regulation of such epigenetically silenced genes. This way we identified several new tumor suppressor genes, like C13 of 18 for cervical cancer. We are currently investigating if we can sufficiently, effectively, remove the hypermethylated signature, especially from essential regions of the promoter, by targeted active DNA demethylation.
Locus-Targeted DNA Demethylation
I will discuss some actual data from my group on locus-targeted DNA demethylation, published last November in Nucleic Acids Research. In this study, by Hui Chen, in collaboration with Guo-Liang Xu from Shanghai we used the zinc finger protein generously provided by Carlos Barbas, one of the pioneers on zinc finger protein engineering. Currently we are also validating the obtained results on novel genes, for which we engineered zinc finger proteins ourselves.
Here you see the promoters region of the gene called ICAM. The engineered DNA binding domain is referred to as CD5-4, and binds an 18 base pair region which contains two CpGs. The DNA binding domain is fused to VP64, which is derived from a viral transcriptional activator and acts as a positive control. Alternatively, the DNA binding domain is fused to catalytic domains of the TET proteins, TET 1, TET 2, and TET 3.
As the TET catalytic domains are still large in size, transduction of cells with retroviruses to express the epigenetic editing agent was challenging. And we had to sort cells based on GFP reporter gene expression. The data shows that ICAM is 80% methylated for CpG number 10, which is in the binding side of the zinc finger protein.
Interestingly, expression of the DNA binding domain without any effector domain was very efficient in lowering DNA methylation This interesting finding can be explained by the very high expression of the zinc finger protein, occupying the CpGs after replication, thereby preventing DNA methylation maintenance processes from accessing this region. DNA demethylation was also observed for the DNA binding domain fused to VP64, the positive control.
Sorting of the cells and infection with the control virus did not affect DNA methylation. In the sorted cell population, the DNA binding domain fused to TET domains were not inducing demethylation to the same extent as observed for DNA binding domain only of for the VP64 fusion. This reflects the low expression level per gene of the fusion proteins, also for the sorted cell population. Similar data were observed for the second CpG within the DNA binding side, which made us to conclude that this is not active DNA demethylation, but merely reflects prevention of remethylation by binding of the constructs.
For the effector domain targeted CpGs, we observed in the unsorted cells that the effect of DNA binding domain only was not detectable for these CpGs. Interestingly, the expression of the VP64 fusion did effect methylation on the neighboring CpGs. This seems to reduce with increasing distance. This finding suggests that indirect recruited proteins might form larger complexes, which in turn could prevent remethylation over an increased size region.
Interestingly, the effect of TET 1, and especially TET 2, in sorted cells decreased DNA methylation, even though their expression seemed to be much lower on the per cell basis. These data indicate that TET can actively induce DNA demethylation upon targeting to a predetermined locus. To confirm that the observed DNA demethylation is indeed due to the catalytic activity of TET, we constructed catalytically inactive mutants. In a separate set of experiments, both for TET 1 and TET 2, the mutants did not induce demethylation, whereas the wall type domains did.
Interestingly, the induced DNA demethylation, which only lowered the methylation level to some extent, was associated with the doubling in expression levels. And this was not observed for the mutant. Currently we are investigating ways to further increase the efficiency of gene re-expression.
Future of Epigenetic Editing
Although the pioneering papers on epigenetic editing only scratched the surface of the wide spectrum of possibilities of the technology, the data show the targeted overwriting of epigenetic marks is feasible, and effective. Many dogma-challenging insights are expected for the field of chromatin biology as we now can mimic and reverse epigenetic mutations. Moreover, if specificity issues are adequately addressed, a new way to start to think about a cure for the incurable is offered, as now any gene, protein or non-protein coding, can be up or down regulated.
However, as with all novel, highly promising approaches, we have to be careful in not overplaying our hand. We need to know more about epigenetic mechanisms and gene expression regulations. But these insights are within reach with epigenetic editing tools. So I expect epigenetic editing to benefit a broad spectrum of applications, ranging from functional epigenetics, with answers to cause versus consequence, and order of events, all the way to clinical epigenetics. And maybe we can even design a cure for the incurable.
More About Dr. Marianne Rots
Dr. Rots has a lot going on. Check out some of the links below to get an idea of everything that she’s involved with lately:
- Rots Lab Website at the University of Groningen
- Clinical Epigenetics journal
- UNICAM Summer School on Nutrigenomics