For a detailed, graphical explanation of the MaGellin work flow, please download the MaGellin Workflow Specifications Sheet, which includes all of the steps in the MaGellin workflow.
We developed the MaGellin assay to optimize the development process for site-specific methylases. Having validated the assay, we determined to design and test three site-specific methylases, two of which had never been constructed before.
The process further validated our MaGellin assay:
1. We recapitulated published results with a zinc finger-methylase and shed light on the significant magnitude of its off target effects. MaGellin is an excellent assay for this purpose, because of the noiseless chassis and because it's simpler to detect off target effects on a plasmid than a genome.
2. We further characterized our promising novel TALE-methylase and were able to de-noise this noisy, complex system. MaGellin was in agreement with bisulfite sequencing that the TALE exhibited targeted inhibition of the methylase. This has serious implications for the multitude of TALE-effector systems that have recently been developed: the TALE can inhibit the effector if the linker length and distance between the TALE binding site and target site are not optimized. The MaGellin workflow is well suited to solve this optimization problem.
3. We demonstrated the enzymatic activity of our novel dCas9-M.SssI and are now characterizing it further.
Zinc Finger-M.SssI Fusion
The zinc finger is a small DNA binding domain, with limited sequence specificity. Previous studies showed it was prone to off-target methylation, which we verified. This was also validation that the MaGellin assay accurately reports the site-specificity of methylation, effectively demonstrating our assay does everything we need it to do.
SHOW ZINC FINGER DATA
Figure 1: The ZF-M.SssI was cloned into MaGellin with and without its binding site present. We ran the standard MaGellin assay on both plasmids, using methylation sensitive restriction enzymes to report the methylase activity.
To be sure of the targeting specificity, we cloned the MaGellin plasmid with and without the zinc finger’s binding site present at the target cut site. This demonstrated how the presence of a zinc finger binding site shifts the methylation pattern (Figure 1).
TALE-M.SssI Fusion
TALEs have a greater sequence specificity than zinc fingers, and are easier to customize and less expensive to construct. They have already been validated for use in genome engineering and are replacing zinc fingers for some applications. We followed the MaGellin protocol to clone a TALE-M.SssI fusion and induced its expression. We repeated this experiment numerous times and found the TALE-M.SssI was methylating at both sites, as reported by the MaGellin software.
TALE-M.SssI actively methylates DNA as reported by our MaGellin Assay
Figure 2: A TALE-M.SssI was cloned into and expressed from MaGellin, then induced with IPTG in T7 Express cells. The NEB10 control has no T7 polymerase and no possibility of leaky expression. The linearized control is the same band length as blanket methylation.
MaGellin reported our TALE-M.SssI was detectably methylating DNA at both the target site and off-target site. We expected a certain degree of off target methylation simply because the TALEs could occupy all the binding sites on our low copy plasmid; the molar ratio is one of the problems in developing site-specific methylases that the inducible MaGellin system is designed to address. MaGellin is designed to screen multiple fusion protein constructs in a high-throughput manner, and a user would normally select only constructs that methylate in a highly site-specific manner. However, we were interested in using MaGellin to study the TALE-M.SssI further before going back to the drawing board to redesign the linker length, binding site, and other variables.
Validated COBRA is in agreement with our new MaGellin Assay
Given the novel nature of our MaGellin assay, we wanted to see if a traditional, published methylation assay would be in agreement about the TALE-M.SssI result.
Figure 3: COBRA on induced TALE-M.SssI. The plasmid was bisulfite treated and the target and off target sites were amplified with our standard bisulfite sequencing primers. The amplicons were digested with TaqαI, which only cuts methylated sites, and BamHI, which only cuts untreated DNA. In COBRA, as opposed to MaGellin, digestion means methylation, and no digestion means no methylation.
We bisulfite converted the plasmid and used our validated bisulfite sequencing primers on both the target and off target site, then used the COBRA assay. The controls recapitulated that our primers are biased for only bisulfite converted DNA, as desired. Unconverted DNA would have been digested by the control enzyme, otherwise. Methylation-sensitive TaqαI digested both the on and off target sites, confirming that the TALE was partially methylating both sites, as MaGellin reported (Figure 3). This validated our assay further, as MaGellin reports the same biological outcome as the published COBRA method, but at a fraction of the cost, time, and technical difficulty.
Varied Induction Conditions Suggests TALE-M.SssI is Prone to Off Target Activity
Given the quick turnaround and cost-effectiveness of the MaGellin assay, it was feasible to test our TALE-M.SssI construct at 20 different conditions to get a better idea of its in vivo. We hoped to find the optimal induction point for reducing off target methylation. This study would have cost us approximately $7,000 to do by bisulfite sequencing, based on the prices at our university core facility. MaGellin only required restriction enzymes and gel electrophoresis.
Figure 4: The TALE-M.SssI with and without the TALE binding site present was induced with 0, .1, 1, and 2 mM of IPTG for 0, 2, 6, and 24 hours to find optimal expression conditions. Representative bands’ intensities were quantified and normalized by background intensity. The plot is missing samples marked * because we ran out of miniprepped DNA. The dotted white circle marks the conditions of our initial experiments.
We varied induction conditions, expecting one might be more optimal than our previous inductions. As expected, the "without binding site" negative control showed significant off target effects, presumably because the TALE did not bind and the methylase is still active. However, as we induced for longer times and with more IPTG than we had originally used, MaGellin reported off target methylation increasing even for the TALE with the binding site.
This makes sense, as the TALEs could saturate the binding sites on a low copy plasmid or genome. We expect this has implications for recently published TALE-effector systems, which may be noisier than expected as it is much more difficult to assay off-target effects in mammalian systems. We propose the use of split-reconstitution systems that are activated by co-localization of effector subunits at a target site. Those are our future directions, optimization of the system will be quick and easy with MaGellin. We would not have noticed the off target methylation so quickly on a mammalian genome, but it was simple to de-noise this complex system with MaGellin.
Bisulfite Sequencing Confirms TALE Binding
Figure 5: MaGellin standard bisulfite sequencing primers were used to bisulfite sequence 500 bp including the target site. Rows are individual clones, circles are CpG sites, and distances reflect their distance in bp. Filled in circles are methylated CpGs.
The next step for characterizing functional constructs with MaGellin is to use the standard bisulfite sequencing primers for a higher resolution look at individual CpG methylation. We bisulfite sequenced the target site of the induced TALE-M.SssI with the TALE binding site present 4 nucleotides upstream of the target AvaI cut site. The data further validated the enzymatic activity of our TALE-M.SssI in the region of the TALE binding site. It was also the first demonstration that the MaGellin assay and standard primers can easily be used for bisulfite sequencing with a TOPO cloning kit with ~7 day turnaround. Interestingly, no clones were methylated at the CpG site that is within the AvaI site itself (Figure 5).
This data, in combination with the varied induction conditions and COBRA results, led us to hypothesize a new model for the TALE’s mechanism that successfully explains each result. Although 4 nucleotides between the zinc finger binding site and the target cut site was optimal for a published zinc finger fusion with a short linker. That distance is too short for use with a TALE fusion with our linker length. The TALE is at least three times the size of the zinc finger and so TALE binding occludes that CpG from interacting with the methylase.
We have now shown that the novel TALE-M.SssI binds to its binding site strongly, as it almost fully protected that site from the methylase for 24 hours (Figure 4). We have shown that it exhibits methylating activity (Figure 2). Perhaps most interestingly, we have demonstrated that the TALE is large enough to physically occlude neighboring nucleotides from access to its linked effector, which has significant consequences for the recently published slew of TALE fusions – including the TALE-histone methylases, TALE-histone demethylases, and TALE-DNA demethylases for epigenetic engineering. We expect the same result will hold for Cas9-effector fusions, and are in the process of validating that hypothesis. We have already cloned the first dCas9-methylase fusion and are going to characterize its activity.
Summary
MaGellin was developed to optimize the development of robust tools for site-specific methylation. To those ends, we successfully cloned and expressed three fusion methylases, two of which are novel constructs with advantages over the previously published zinc finger. Our constructs have shown methylase activity and DNA binding activity, which we could measure with our new assay. They are ready to be further optimized, using our workflow.
To gain our new insight into a fundamental shortcoming of recently developed genome engineering tools, we used MaGellin to its full extent: swapping out DNA binding domains and binding sites, varying induction conditions, applying COBRA, bisulfite sequencing, and depending on our original algorithm to properly predict methylation-sensitive digestion patterns. Importantly, we could not have reached this result without MaGellin, because the one-plasmid system in a noiseless chassis makes it simple, even unavoidable, to detect off target methylation. Conversely, for the previously published work in mammalian systems, it was not feasible to detect off target effects across a long genome with background signal. Based on our data, future improvements on genome engineering tools should include the construction of two targeted fusions with subunits of effectors that only dimerize and show activity at the binding sites, along the lines of how TALE-Nucleases cleave DNA. That could be the best way to construct epigenetic engineering tools with the specificity necessary for clinical applications.
Moreover, we have demonstrated the importance of studying the distance between the binding site and the target site, and shown the ideal distance will be very different between different DNA binding domains. This boils down to an optimization problem between choosing binding sites and linker lengths; this is exactly the sort of problem that the MaGellin system is designed to solve in a fast and affordable manner.
"
