Scientist Stories: Jonathan Watts, Spotting Indirect Effects in Oligonucleotide Therapeutics

So good morning everyone, good afternoon, good evening, depending on the place in the world you are. So first of all, apologies for the timing of the webinar. So I think there was a little bit confusing. So I think daylight savings time started already in the US. It did not start in Europe yet. So apologies for the late warnings about the starting time of the webinar. But Hopefully everyone is here, and if not, you can also watch it back. The webinar will be placed on the OTS website in a few days from, from today. So for today, we would like to welcome John Watts. So John Watts grew up in Canada, and he received his bachelor's in 2001 from the Dalhousie University, and he was winning the university medal in chemistry and conducting work in carbohydrate synthesis with Professor Bruce Grindley. After a year in the nonprofit sector in Côte d'Ivoire, John returned to science and to Canada, receiving his PhD in 2008, working with Professor Massad Dahma at McGill University. I already have seen Massad somewhere in the chat here, so welcome, Massad. His doctoral work included the development of two new oligonucleotide analogues and exploration of why to fluorooligonucleotides shows such high binding affinity, and studies on the chemical modifications of ASOs and siRNAs. He then moved to Professor David Corey's lab at UT Southwestern for postdoctoral studies where he, his work included the chemical modifications of promoter-targeted duplex RNAs and the use of oligonucleotide oligospermine conjugates as antisense and anti-gene oligonucleotides. In 2012, John started his independent research group at the University of Southampton in the UK, where he received the 2013 Young Investigator Award from the OTS and the 2015 Vice Chancellor's Award for Teaching. In the summer of 2015, he was recruited to the RNA Therapeutics Institute at UMass Medical School, where he is an associate professor. The Watts Group is developing new approaches in the medicinal chemistry of several classes of oligonucleotides along with novel tools for sequencing and synthetic biology. It's not in the biography, but I think he recently became a full professor at the same institute. And today, John will walk through a recent study that the lab carried out trying to activate frataxin expression to treat Friedreich ataxia. A lot of traditional good and right controls were done. But they still found themselves vulnerable to off-target effects. Next-gen sequencing combined with gene editing proved that observed effect was indirect. So, John, floor is yours, and looking forward to your presentation.

Thanks very much for the kind introduction, Ronald, and thank you all for being here. It's great to see so many of you here. So let me share my screen. So I'm calling this Spotting Stubborn Indirect Effects in Oligonucleotide Therapeutics. And as Ronald mentioned, this talk will have two parts. The first part is really a story of a project that we worked on trying to develop antisense oligos to activate frataxin expression to treat Friedreich's ataxia. And eventually, This took us quite deep into work required to differentiate direct and indirect effects and to understand the mechanism of how these compounds were activating protaxin. And then in the second part, when we finish that story, we'll look at some lessons learned for different projects. So what's Friedreich's ataxia? Friedreich's ataxia is a disease driven by a repeat expansion of a triplet repeat, a GAA repeat. And patients with Friedreich's ataxia can have hundreds of copies of this repeat. It's in the first intron of the frataxin gene. And some sequences, including purine-rich sequences, are particularly likely to form R-loops. And there's evidence that there's an R-loop formed here at this locus. For most sequences, DNA homoduplexes are more stable than heteroduplexes, DNA-RNA heteroduplexes. But purine-rich RNA sequences are one of those contexts where often actually the heteroduplex can be competitive or even more stable. And so they're vulnerable to R-loop formation. At Fritaxin, the R-loop leads to repressed expression, leads to epigenetic changes, which leads to further repressed expression of Fritaxin. That leads to insufficient expression of the frataxin protein, which leads to mitochondrial dysfunction. And so the disease pathology is most obvious, most severe in tissues with a high metabolic rate. So that includes brain, heart, skeletal muscle, and pancreas, among others. So our work on using oligos to reactivate frataxin was sort of building on and inspired by some work done in David Corey's lab where they used repeat-targeted nucleic acids binding that triplet repeat, and they observed that they could get derepression of the, presumably by displacing the R-loop, derepression of Frtaxin expression, so reactivation of Frtaxin expression. And that was published in Nature Communications in 2016. And so building on this work, my student Damon, who's just in his final year of his PhD here in my lab, set out to activate frataxin expression using non-repetitive intron-targeted antisense oligos. So of course there are some, at least theoretical, safety risks with using targeting repetitive sequence. There are GAA repeats in other genes besides frataxin. And so we wanted to know if other regions of the intron might be appropriate for, for protaxin reactivation. So this is a little schematic of the gene. Let me just make this be a laser if I can. There it is. So this is the end of exon 1 and the beginning of exon 2. And Here's the triplet repeat, the GAA repeat near the 5' end of exon 1. And so as an initial focus, we looked at regions just upstream or just downstream of that triplet repeat expansion in regions that are predicted to still contain this R-loop to show R-loop formation. And so we tiled a number of steric blocker oligonucleotides, these ones that have S in their title, steric blockers, all the way across from upstream through targeting the repeat and then downstream of that repeat. In patient-derived fibroblasts where both alleles contain the expanded repeat, they had 330 copies and 380 copies of the repeat in each of the two alleles. And so we identified this one hit, S10, that caused about a twofold activation of frataxin expression. And you'll notice in these graphs that frataxin is inherently pretty noisy. There's variability in the expression of frataxin from experiment to experiment, and so there are a lot of repeats done. But very significantly, this compound activates frataxin expression across all these repeats. We then took a second approach where we looked at a broader range of intron 1 sequences and used some gapmer ASOs, wondering whether there could be some leftovers of intron that might stick around and be longer lived than— longer lived in the nucleus potentially than the mature RNA being spliced out and might continue to play a repressive role. We were aware that Of course, gapmers could also in some cases potentially repress expression if we targeted the intron before the intron had been spliced out. But designing a series of gapmers across this site, a few of them were repressive, most of them were inactive, but we actually found one gapmer right in the middle of the intron that we called it G30, and it also provided activation of frataxin expression, statistically significant, about twofold. This red bar on the left in both of these graphs is wild-type cells with an untreated control oligonucleotide. And all the rest of the bars in both of these graphs represent the patient-derived cells with negative control oligonucleotides or repeat-targeted— or sorry, upstream or downstream of the repeat-targeted test oligonucleotides. And so by a summary of this initial screen, we found 2 hits in different regions. Each of them gave about a 2-fold activation of frataxin expression. And as you can see in both of these cases, if you compare to the wild type cells grown under the same conditions, they restored frataxin expression to a similar level relative to wild type cells. So we were encouraged. This is a sort of encouraging starting point. This was active relative to untreated. That's the UTC, is actually untreated cells, a non-targeting control, scrambled control, oligonucleotide, the other negative control, and then multiple inactive compounds. But these two active compounds gave us this very encouraging result. So one of the next steps that we did was to carry out a micro-walk. And so we wanted to know if ASOs, other antisense oligos that were targeting overlapping regions of those first two hits, were also active. And so we moved two bases at a time, just upstream all the way to just downstream of these initial hits. And we did them as both steric blockers and as gap-mers for both cases. And we were able to find multiple additional hits by moving 2 to 4 bases in each case, in this case up to 6 bases upstream or so, in this case, 2 bases upstream or 2 bases downstream that maintained this kind of twofold-ish activation, twofold or so activation, again, in these patient-derived cells. Then we carried out some additional work to optimize the length.. And the chemical modification of these hits, you can see that starting from our 18-nucleotide hit, we made longer ones, 19, 20, 21, or 22 nucleotides. Most of those maintained activity. That's upstream of the repeat again. And then these ones at the S30 site downstream of the repeat, again, multiple slightly longer versions. We also included LNA nucleotides in some of these. They were initially made as 2'-methoxyethylphosphorothioate oligos. LNA nucleotides also worked, maintained activation. And some of these gave slightly higher levels of activation than our initial one in these patient-derived cells. So between 2- and 3-fold in some cases. We then wondered how important the repeat length was. And so we looked at, from those patient-derived fibroblasts, some with a longer repeat. So instead of 300-something repeats in each allele. These ones had 400 and 500-something repeats in each allele of the frataxin gene. And the activation was maintained in these longer repeat cells by both the S10 and the S30 targeted oligos. And then we set out to try wild-type fibroblasts, and we saw activation in wild-type fibroblasts. And this was a little bit of a puzzle to us. So we also checked other cell lines, not for taxin development, not fibroblasts and not for production of Friedrich's ataxia patient-derived. And those wild-type cells also showed activation. And in particular, these UAD7 glioblastoma-derived cells showed really very potent activation. So we thought this is not dependent then on the formation of an R-loop. We're not releasing an R-loop. Doing something to derepress frataxin activation. We're actually activating it in some way that's not dependent on the prior presence of disease pathology. We then moved to mouse models of Friedrich's ataxia. And there are several mouse models available. The one that we used the most was called the YG8R model from the Pook lab. And after ICV injection of these compounds, either the S10 or the S30 lead compound relative to two negative controls. So this is intracerebroventricular injection, injecting into the spinal fluid in the mouse brain. We didn't see any activation even at doses where ASOs are typically very active. This was 25 or 30 nanomoles, pretty reasonably high dose. We wondered if this was an issue with something in vivo, with biodistribution in vivo, or an issue with the mouse model. And so to test whether it was a problem in the mouse model or in neurons, we took mouse embryonic neurons, grew them in culture, and then studied our different sets of compounds in these embryonic neurons using gymnotic delivery at 1.5 micromolar. Again, you can see that there was no significant activation, no activation at all in these neurons of the mouse model. This is the YG8R model, the same model that was used above, which contains two tandem copies of the expanded repeat. The YG8SR with this over 800 contains a very long repeat, a longer GAA repeat than the other mouse models. And only a single copy. But both of these mouse models did not show activation. To rule out whether it was a delivery problem, we also delivered a Malet1-targeted gapmer into the neurons from both of those mouse models. And you can see that we got really very solid silencing of Malet1 using this gapmer oligo. And so we knew it wasn't a delivery problem. There was something inherently not active about this oligo in this neuron model. So we were a little bit concerned that the activation didn't happen in vivo. And we were also aware that for some of the cell work, the cells grew more slowly when we were— when they were treated with active compounds. They weren't ridiculously toxic, but there was a slowing of cell growth. And that was something that we paid attention to. And we were concerned to confirm whether this was an on-target effect. Two really important papers by Peter Hagedorn et al. Peter Hagedorn is at Roche and was at Santeras before that. These are really important papers if you haven't read them in terms of approaches to Study sequence specificity, identify and avoid off-target effects of RNAse H-dependent antisense oligos. And one of the principles that Peter presents in these papers is that when you're using two different antisense oligos, and then study transcriptome-wide the changes in gene expression, the off-target changes from ASO1 will often be different from the off-target changes exerted by ASO2.. But the on-target changes, things like obviously the gene of interest, but also potentially downstream changes— when you silence a gene, it will of course change other things in the cell, and you want to know if those are downstream of the desired effect or if those are off-target changes. If they're on target, they should be common between those two kind of non-overlapping sequences if those sequences are both working. To silence the target of interest. And so working with the group of Athma Pye here at the RNA Therapeutics Institute, we did a transcriptome-wide sequencing study of the two hits. We were fortunate to have two active hits, and we felt that that gave us pretty good evidence that this could be an on-target effect since two different regions, non-overlapping, produced the same kind of effect. So the first piece of news struck us as good. The ASOs, the two ASOs did induce a similar transcriptomic profile. And so the first bar here is wild type cells. So healthy cells, not Friedreich's ataxin, not ataxin cells, but then the next three are negative controls. So a non-targeted siRNA, a non-targeted antisense oligo, and untreated cells, so cells that didn't have any oligonucleotide at all. They were similar to each other, those three. And then the S30 lead and two leads from that S10 locus, one that was longer and one that was still 18 nucleotides and shifted by 2 from the original S10 sequence, so our two most potent hits at that site from the micro-WOC. So all three of those active ASOs induced a similar transcriptome profile.. You can see on this heat map. But if you look, compare the first column to these last 3 columns, the transcriptomic profile that's induced by these 3 sequences is very different from the profile of healthy or wild-type cells. So instead of taking these negative controls and taking the population of transcript expression from these negative controls and bringing it back to look more like the wild type cells, it actually moved it in a different direction altogether. There are different ways we can visualize this besides that heat map. So one of the ways that we can visualize this is to look at a principal component analysis. And you can see these 3 negative controls in the middle. And just conceptually, when grouping these compounds by transcriptome-wide similarity, you can see that wild type cells are found on the left. And all three ASO treatments move the transcriptome in the opposite direction, move it to the right quite far away. So this is not making the cells more like wild type cells. And then secondly, this is called an upset plot. And this shows the number of differentially expressed genes that are shared between different groups. And so you can see that these three antisense oligo groups, S30 and the two S10 groups, they have a lot of differentially expressed genes in common, but they don't have many relative to the wild type cells. They're not moving it to be like the wild type cells. And so All of these things told us that we were at risk of this not being an on-target effect. We were concerned that, you know, we expected that if we were reactivating Frtaxin and the changes in gene expression were downstream of Frtaxin reactivation, that this should look more like wild-type cells rather than less like wild-type cells. So we wondered if the Frtaxin activation that we were seeing might actually be downstream of some other change, that maybe it was an off-target effect. We were causing some kind of slowing of cell growth, and that that might be activating Frtaxin expression. And so Damon had the scientific courage to do the ultimate control experiment and to remove the target sites using genome editing. So he created pairs of guides, electroporated those into HEK293 cells, seeded those HEK cells at a density that allowed a lot of wells with only a single cell. And then isolated colonies, sequenced those, and expanded the clones that contained a deletion of the S10 region or a deletion of the S30 region. And so we now had clones that were only— that were not containing regions complementary to our oligonucleotides. He confirmed that that didn't affect the overall expression of frataxin or the splicing. Or any other processes upstream of being treated without being treated. When we treated these two cell lines now with our active compounds, you can see what happens. These are again untreated non-targeting controls or untreated cells. This is a cell line that's missing the target region of S10. Sorry, on the left of each group, it's the knockout clone, and on the right of each group, it's wild type HEK293 cells. And you can see that the activation is essentially the same in both of those cells. So in a cell line that's missing the S10 target site, S10 still activates for taxin expression. And similarly, at the S30 site, you can see again at the left, This is a wild type. No, sorry, here it's at the right. This is a wild type cell, and the left two are the two different knockout clones. And again, they show very similar levels of activation, even better activation in this case, when the target site is not there. And so this suggests that this is, again, or demonstrates really conclusively that this is working through an off-target effect and not a direct effect. So what's going on? Well, we first wondered if this is operating through some off-target transcript that was targeted by both sites. They're not overlapping at all. They're hundreds of bases apart, 1,000 bases apart or more in where they target within the intron of Fritaxin. But of course, they could still coincidentally be inhibiting the same off-target transcript and causing a change that way. So we worked again with Athma's lab, generated a little algorithm to calculate a pairing score of genes that matched our different ASOs. And this is a Venn diagram showing the differentially expressed genes that are complementary to the S30 set of oligos or the S10 oligo. And you can see that Fritaxin is the only transcript with high complementarity to both ASOs. So it's not like there's an easy-to-identify hybridization-mediated target that's shared between these two that could be causing the off-target effect. So the next question that we asked is, is this effect more dependent on base pairing, or is it something to do with backbone chemistry? And so we made PMO analogs, morpholino analogs. Most of you, I think, will know that PMO, phosphorodiamidate morpholino oligomer, or PMO or morpholino analogs are neutral backbone modified analogs that don't contain phosphorothioids. They don't show much protein binding. And so we wondered whether if this was a backbone chemistry phenomenon, we could identify that by changing chemistry. So these PMO analogs of non-targeted controls or of S10 and S30 did not activate relative to untreated cells. So these did not activate relative to the non-target controls. We saw no activation by any of these morpholino analogs. So maybe backbone chemistry is involved. And then finally, we wanted to know if this effect was dependent on specific motifs within the ASOs. And so we carried out a single mismatch study mutating one base at a time. In particular, there was this G-rich region here And when we mutated any of the Gs in this CGG motif, this C or G or G, these lost all activity. So even a single mismatch in this region was enough to make the effect totally go away. The other sequence also contained a G-rich motif. It contained 4 Gs at the 3' end. And again, not all sequences that contained these 4 Gs provided this activation, but G-rich motifs certainly have been seen to go through, to work through off-target effects or induce off-target effects in the past. And so we were maybe not all that surprised when mutating single or a pair of bases in that G-rich motif made the activation of this compound go away. So these are different motifs, but Both of them contain G-rich motifs, and disrupting these two eliminates the activation seen by these two sets of transcripts. So I want to highlight that sometimes off-target effects are stubborn. We were using sort of the minimum standard controls. Most of you will know this paper that came out by Keith Gagnon and David Corey in Nucleic Acid Therapeutics, our society journal as the OTS. A few years ago. And they have a little summary here of minimum guidelines for studies using synthetic oligonucleotides to control gene expression. And if you look at these sort of minimum experiments here, you know, two independent silencing siRNAs or antisense oligos, again, we had two non-overlapping sequences that produced the effect, a mismatch-containing control, a scrambled control, we were using multiple types of negative controls, including also untreated cells that we think are important to compare. And even in spite of this, we had to dig actually several layers deeper to find that this was an off-target effect. So where do we go from here? We could choose to just despair that we'll ever know whether our effects are real, but I think actually there are some important lessons to take away that we can put into practice. So lesson number 1, I think we need to be especially careful, those of you who are listening who are working with new mechanisms. New mechanisms are exciting, but when mechanisms aren't well understood, we are more vulnerable to slipping into off-target effects. And in particular, whenever we're working on upregulating things, I think we can be quite vulnerable to off-target effects. Upregulation is more vulnerable to silencing in our hands, Because among other things, they can be confounded by off-target effects based on normalization. So if a non-targeting control that we're using induces some silencing of the on-target gene, then the test ASO will appear to activate by comparing it to that control ASO, that non-targeted control. Or if the test antisense oligo induces some silencing, even modest silencing of a housekeeping gene, even let's say 50% silencing of whatever you're normalizing to as a housekeeping gene, that will also look like twofold activation. That will result in apparent activation. So when working with new mechanisms or poorly understood mechanisms, we need to check mechanism on multiple levels, and we need to include multiple controls. Lesson 2, I think, here is to be alert to unexpected phenotypes, especially to any kind of toxicity. So in our cell studies, the active compounds caused some slowing of cell division. It wasn't overt toxicity, but it was, was significant enough slowing of growth that the treated wells had fewer cells by the time of harvest. That was noticeable, and this was unexpected for our mechanism. And so it's just important to highlight that toxicity and immune responses are often accompanied by many gene expression changes, and that can include gene expression changes in the genes that you are looking at. And so if you see any evidence of toxic or immune responses, it's good to really pay attention to that. I think the lesson 3 that came out of our work is that sometimes changing chemistry can be a very good way to reveal off-target effects. So morpholino analogs, PMO analogs, didn't recapitulate the activation seen by our methoxyethyl phosphoroethioate oligos. And in contrast, if you take an example like Duchenne muscular dystrophy, skipping exon 51, the leading first set of leading splice-skipping oligos for exon 51 were trisupersen and edupersen. That was 2'-O-methylphosphorothioate and a PMO analog. And they were mostly overlapping. They showed similar effects in cells. They both made it into clinical trials. And so that provides additional evidence that that's not a backbone-dependent effect. Those were sequence-mediated on-target effects. A fourth lesson, a fourth lesson that I think is useful for many of these kinds of projects are that RNA sequencing, looking transcriptome-wide, was really key for us to understand whether our cells were moving closer to the wild type direction, or in this case, in a different direction. So looking at gene expression changes transcriptome-wide, it made it really very clear that our patient-derived cells were not becoming closer to the phenotype of wild type cells. They were actually going the opposite direction. And RNA-seq also helps avoid some of the normalization effects mentioned earlier. Because we can compare to the whole transcriptome rather than only one or two normalization genes, housekeeping genes. And finally, genome editing is a powerful method to study mechanism, especially when working with new mechanisms. So our genome editing work showed conclusively that the effect was indirect. There are other happier ending examples. Previous work by Longqing Li's group, for example, showed that an oligo working through a non-traditional mechanism was operating through a direct effect. That was, again, a promoter-targeted gene activation experiment. So this work was published late last year in Nucleic Acids Research. I've kept this short to hopefully be able to discuss more with you and leave lots of time for discussion. But if you want more details, reach out or dig deeper into this paper. I also want to highlight this book that some of you may have heard of already. It was just out last summer. This book was a book that was helpful to me as a student in its third edition. The third edition came out while I was doing my own PhD studies, and for the fourth edition, Mike Gate and I— sorry, for the fourth edition, Martin Egli and I joined the two Mikes, Blackburn and Gate, and and brought it up to date. And we really wrote it with students in mind. So I hope those of you who are listening who may be students or maybe newer to the field will find it helpful. So we negotiated hard with the Royal Society of Chemistry to keep it affordable, and hopefully your lab or your library could get a copy if they don't.

Great.

So this has been work that has been mainly led by my student Damon Wong, but with help from the rest of my research group, of course. Our sequencing work was helped by Athma Pai and her student Ezekiel. And some of the work in human neuron models that didn't make it into this talk and other work on frataxin epigenetics was done in collaboration with our colleague Michael Green, who sadly passed away just last month, and his postdoc Mingyang Feng. So thanks very much for your attention and your time, and I'm looking forward to discussing together.

Wonderful. Thank you so much, John, for a really great talk with lots of information and tips for doing ASO-mediated therapeutic development, but also to understand some disease mechanisms. I was wondering if you have seen similar things with different targets? So if you just include that G-rich motif at the end of a sequence that's targeting something else, do you see any that similarly upregulate gene expression, or have you just focused on FrtXM so far?

So there are other targets. There are a few that we've worked on for activation, both when I was a postdoc in David Corey's lab, there were some targets there. We've been working on a few targets, but they're earlier stage. So certainly activation is harder, but there are multiple mechanisms that can lead to restoration of gene expression. One of the— two of the targets that we've been working on are relevant to neurological disease and that have a more defined mechanism. One is called STATHMIN2, where there's a pseudoexon that gets spliced into the STATMIN2 transcript with ALS TDP-43 pathology. We can use an ASO to skip that pseudoexon and restore expression of STATMIN2. There's another one called FMRP that's involved in Fragile X. This is work that is quite new. This is with my colleague Joel Richter. Again, Joel, through good sequencing work, has identified that FMRP is mis-spliced in Fragile X and that we can restore that. Expression by actually skipping a pseudoexon. So those are some examples of, I think, meaningful and on-target gene activation.

Yeah, absolutely. And do you think that— so you said in your talk that in the cancer cell models, you do see that slowing of cancer cell growth through this mechanism. And do you think it would be possible that in an in vivo model of a tumor, for example, you might be able to slow progression specifically in those tumor cells, and potentially this off-target effect might be an on-target therapeutic for something other than ataxia?

Yeah, I think it's a great question. It could certainly be, I think, the beginning of a project for both cancers or other things where you want to slow growth, and also for, for frataxin itself. If somebody wants to dig into the transcriptomics and understand how was this off-target effect working to activate frataxin, and how was it working to slow cell growth. Those may be different questions, or they may overlap. Probably these sequences might not be the right things to take forward as drugs, but the study, the data, all of the sequencing data is available with the paper. If somebody wants to dig in and try to study those pathways and turn them into a project that's valuable for either of those.

Awesome. And then moving to the questions in the Q&A, and please feel free to add questions to the audience members. From Christina, assuming Assuming that the transcriptional expression of frataxin might be extremely low, so around zero in individuals with disease, how would a 2- to 3-fold increase that like what we see in your studies be relevant clinically? And can you point us towards any data that shows the relative expression of frataxin between healthy and diseased fibroblasts and patients?

Yeah, so great question. Thanks for that question, Christina. So actually in these bars, we had wild-type cells on the left in most of those. Those were the red bars. Those were from healthy fibroblasts. And so it wasn't actually— it wasn't down to zero in these patient-derived fibroblasts. And the ASOs were able to reactivate the frataxin expression to be similar to that of wild-type cells. Again, that's fibroblasts. That would need to be done in probably more biologically important, therapeutically important cells like neurons, to be sure. But that's part of the answer to your question. And then looks like you're asking about Western blots as well. We did do some protein expression, but for Texin was pretty low level. And so our Westerns were pretty messy. It was hard to get really good clean Westerns for Texin from fibroblasts. I think that's good to highlight for the community that it's another way that we can gain confirmation about mechanism to check things at both the RNA and protein levels. And so I think that's a good call, Christina, that would be another level that would add value.

And a similar question from Jamie, so we'll skip over that one. From Jan, did you see any sequence or chemistry-dependent aggregation of the ASOs?

That's a really good question. We did not see them doing the things that would make aggregation obvious. We didn't see them precipitating out or losing, you know, A260 over time. So they were pretty well-behaved ASOs overall.

Nothing obvious there. Thank you. And then from Christina again, citing Wang et al., and that there's a mention of seeding only 3,000 fibroblasts into each well of the 96-well plates. And she was wondering if you lysed and if you extracted the total RNA directly from the well or you did trypsinization and a wash spin and what your total RNA yield was. And then a follow-up question for the RNAiMAX lipofectamine, what was the exact amount of stock ASO to lipofectamine that you used versus the manufacturer suggestion?

And— Yeah, yeah, great. So Christina, if you have— if you want to dig deeper into some of these specifics, feel free to reach out directly to— it sounds like some of them may be maybe, maybe quite specific, and it'd be good to get Damon involved in some of these conversations. I can answer a few. We do tend to use less lipofectamine than the manufacturer recommends. We just find that cells for across a lot of project cells tolerate it a little better when we use a lower ratio, and we still see really good transfection efficiency. Also, of course, good for the budget to spend less lipofectamine if we, if we need less to see it working. So we tend to use less lipofectamine. For both of those reasons. I should mention that we've used both qPCR and the Quantogene bDNA assay. So most of the 96-well— in fact, I think all of the 96-well assays were done with the Quantogene bDNA assay, where it wasn't a question of sort of trizol and then separating RNA. It was making cell lysates and then carrying on with this non-amplification-based RNA detection method, this branched DNA, bDNA, Quantigene assay. But qPCR and the Quantigene assay gave the same results. So we did check both of those, both of those routes.

And I think both John and Christina bring up an important point about lipofectamine. And we see very similar results between ASOs and siRNAs that the toxicity of lipofectamine can really be enhanced when you combine it with an oligonucleotide for these cells. And so we find it's really important to each time you're using a new type of lipofectamine or a new dosage of lipofectamine to redo your dosage scale, your validation, so that you know that your top dose in combination with your lipofectamine isn't going to be toxic and give you, you know, a dose-response curve that doesn't really make any sense. So doing that initial validation each time you change the lipofectamine amount or even the source of lipofectamine can be really important to get really nice non-toxic toxic-related results. Yeah, I agree. Thanks. And from Jamie about the indirect effect of genome editing, is it possible that ASO treatment is generating miRNAs that affect pathways involved in the disease?

Yeah, so this is a good question. I'm not sure exactly how the ASO treatment would generate new microRNAs. You could imagine it might change the regulation of microRNAs that are involved, that are already expressed by the cell. But ASOs are structurally quite different from microRNA themselves. But some of the same principles might be there, right? I mean, nucleic acids can work through multiple mechanisms. So I think microRNA should be considered along with all of the other on- and off-target pathways.

And from Michael, wants to tell you an excellent talk and likes book as well.

So that's a little bit cheeky from Mike because one of my favorite chapters in the book is Mike Booth's chapter on sequencing. Those of you who don't know, um, the, the chemistry of next-gen sequencing, um, please go read Mike Booth's chapter. I think it's one of the best pieces out there. That's, I think, chapter 7 or something, chapter 8. Um, and, um, I think it meets a huge need in the field. So I'm gonna take your very kind shout out, Mike, and turn it into, hopefully, I think, a pertinent shout out back to you for that work.

Awesome. So did you have any challenges with the delivery of the different ASOs to the suite of different cells that you used?

I don't think delivery was a limiting factor, a limiting step here. You saw in the neurons, that we had added, used a MALAT1 positive control, a positive control for delivery, where by trying to silence MALAT1 with a MALAT1 gapmer, we wanted to make sure that these oligos were taking up cells under the, you know, under the conditions of our experiment. And that was, you know, fibroblasts, we know that we can transfect them efficiently. We have lots of other studies that those conditions work well. So For anybody working with a new cell type and figuring out concentrations of a particular set of reagents, I highly recommend that you include, even if it's for a different target at first, include an oligo that silences something that you know works. MALAT1 is a great one because it's easy. And then hopefully as your project moves on, you can stop using that different targeted control and just include a positive control, which is then an active compound for your mechanism for your project. And that can really help. Make sure that each of your transfections is working well, that you kind of know what you're doing and don't accidentally include, you know, something that's failed experimentally and assume that means sort of that the next generation of things in your project aren't actually working.

And then from Dan Curtis in the chat, do you know why the effects of activating frataxin were lost when you went in vivo with the mouse models?

I think that's a really good question. Of course, the mouse models have frataxin on a different locus. They have it on a transgene expressing the human frataxin. And so there may be something about the way that in its endogenous genomic locus, frataxin expression is regulated by whatever else is changing in these cells. And that's not true in the case of— that may not be true in the case of these human transgene for Taxin constructs, where they are in the model mouse, the mouse model genome. That's probably only one of several hypotheses, but that's an easy one to start with.

And then from Veronica, you tested two different PMO sequences for each of the two targets. Is it sufficient to say that this chemistry doesn't work at these targets? Could it be a poor transfection efficiency of the PMOs? Or is there some reason why they might not work other than just the change in chemistry.

Yeah, I think that's right, Veronique, that there could easily be different reasons why the PMOs did not work. We did use the conditions for delivering PMOs to cells that were recommended by GeneTools, where we purchased those PMOs. We don't make the PMOs in our lab. We make all the other chemistries in our lab. I think I would not want to take that PMO experiment on its own to prove that this was a phosphorothioid-mediated off-target effect. But I think in concert with all the other experiments, that just provided one more piece of evidence in our thinking that maybe this was not a sequence-dependent effect.

And from Joshua Elkington, is there a specific chemical modification inducing degradation and steric blocking?

Yeah, so depending on the mechanisms, degradation of course requires specific chemical modifications or patterns. So if you want to recruit RNase H, you need to include at least 7 or 8 bases in a row of DNA or a DNA analog in your oligonucleotide. If you want to recruit RNAi, again, you need a specific double-stranded RNA pattern or mostly double-stranded RNA pattern to recruit Argonaut and induce RNAi silencing. For steric blocking, there are a lot more options. So most modifications that maintain high affinity base pairing can be used for steric blocking. And I think that's just a more flexible approach. And so things like morpholinos, PMOs, that are just very different structurally and charge-wise, those are useful for things like splice switching or other kinds of steric blocking approaches. And there are a lot of other examples, peptide nucleic acids, as well as, you know, 2' modified analogs that work well for steric blocking approaches since they don't have to recognize enzymes that do the silencing. You can still use those to silence If you use them, for example, to skip an essential exon, that can lead to silencing through a steric blocking approach.

Excellent. And then another question in the chat. Thank you for the wonderful investigation. When I do knockdown of c-Myc by siRNA, there was overexpression of mRNA, but the protein was downregulated by Western. Has your student checked the protein knockdown in these cells? And do you see consistent knockdown? I think you kind of already touched upon this, but—

Yeah, we didn't see significant protein activation, not a huge amount of protein activation, but the Westerns were hard. The Westerns were kind of noisy. So I don't put a whole lot of weight on the Westerns.

Yeah, thank you for the good question. I think it's out of some things that feel like setbacks, we can learn a lot about mechanisms and the underlying controls of using these types of therapeutics. Are you and is Damon continuing the project in terms of continuing to try to upregulate frataxin, or is it something that you're stepping away from moving forward? And what is your like new approach since this one hasn't really been effective?

Yeah, great, great question, Chantal. So we are doing some work on using genome editing for frataxin. So my colleague Wen Xue, is one of a few groups who identified that paired prime editors can delete expanded repeats in other regions of the genome precisely and efficiently. And so we're working with WEN and some of our other genome editing colleagues, Eric Sontag and Scott Wolf, to try to contract the repeat and re-express frataxin that way. Yeah, that's one of the ways in which frataxin lives on in our lab. Yeah.

Excellent.

And it looks like Brian asked too, Brian Johnson, in the Q&A, did we also investigate GAA targeting oligos to see if they also have off-target behavior? And we didn't look deeply, Brian, in answer to that one. We included a few just to see how they compared in terms of activation. We found that Some of our intron-targeted ones that turned out to be indirect, but at first we were actually encouraged that we were able to get even better activation with our non-repeat-targeted compounds than with the GAA-targeted ones. So at that point, we kind of put the GAA-targeted ones on the shelf and didn't dig further into those.

Thanks for the question. So we have a few more minutes before the hour. So if anybody has any other questions or any other topics that you'd like to talk about that are somewhat related to John's talk or the work that John is doing, please feel feel free to put them in the Q&A. Otherwise, John, do you have any other words of wisdom for us, or Ron, do you have any questions on your mind?

No, I think something that I forgot to do at the beginning of the webinar, I really would like to thank Chantal for co-hosting the webinar today. I think you were an excellent moderator of the Q&A session today. So thank you very much. And Of course, John, thanks, thanks a lot for giving the webinar today. I received your book during the last OTS, and I think it's a great book. And if you have the chance to buy it or to have a look in the book, I think you should definitely do it. And then the webinar will be posted online, I assume, later today or tomorrow, so you can have a second watch at the webinar if you like. And then I would like to announce the next webinar. So the next webinar is in 2 weeks, so it's March 30th, and then we have Irene Vazquez Dominguez and Paolo Piggini. And the webinar will be at 5 PM European time, so it will be clearly posted on the OTS website and on the social media account. So once again, John, thank you very much for today.

See you later.