RNA modifications and epitranscriptomics - digital talks

Video Transcript

• 00:00 - 00:15:  Hello everyone, hopefully you can all hear me. So on behalf of Abcam, I'd like to welcome
• 00:15 - 00:19:  you to the very first session in our epigenetics from bench to clinic event. This is going
• 00:19 - 00:23:  on over the course of November, and there are several other sessions in the following weeks
• 00:23 - 00:27:  after this, but this is the first and therefore obviously the most exciting. Today's session
• 00:27 - 00:31:  will focus on RNA modifications and epitranscriptomics. And we're really pleased to be able to give
• 00:31 - 00:35:  you a list of exciting talks. And while there's a panel discussion at the end, which I'm hoping
• 00:35 - 00:39:  is going to be awesome. Today's session is going to be chaired by none other than Tony
• 00:39 - 00:44: Kouzarides, who I'm sure needs no introduction. But Tony is a professor of cancer biology
• 00:44 - 00:48:  at the University of Cambridge, a senior group lead at the Gurdon Institute and a director
• 00:48 - 00:52:  of Melnor Therapeutics Institute. His lab is interested in epigenetic modifications
• 00:52 - 00:57:  of histones, RNA and DNA, and their involvement in different types of cancer. He's a co-founder
• 00:57 - 01:01:  and ex-director of both Abcam and Chroma Therapeutics, co-founder and director of
• 01:01 - 01:07:  Storm Therapeutics, Cambridge Gravity, and Conquer Cancer, a cancer charity in Spain.
• 01:07 - 01:11:  With that, I'm absolutely thrilled to hand over to Tony.
• 01:11 - 01:20:  Hello, everybody. Welcome to this first Abcam RNA Modifications meeting, which is the first
• 01:20 - 01:29:  of a series of meetings on epigenetics. First, we will have two short talks from abstracts,
• 01:29 - 01:35:  and then we're going to have three plenary talks. This will be followed by a panel discussion
• 01:35 - 01:42:  where the plenary speakers will discuss the state of the field in general terms. So before
• 01:42 - 01:47:  we start, can I remind you that you can ask questions in the chat box at the bottom of
• 01:47 - 01:54:  your screen, but please ask them during the talk so that I can see them and we can start
• 01:54 - 02:01:  asking questions as soon as the talks finish. So let's start with the first talk, and this
• 02:01 - 02:07:  is from Frankie Young from the Beckman Research Institute in California. And the title of
• 02:07 - 02:16:  his talk is Arginine Methylation of Methyl-14 by PRMT1 Regulates M6A Deposition. Over to
• 02:16 - 02:17:  you, Frankie.
• 02:17 - 02:23:  First of all, I want to thank the organizers for putting together this wonderful series
• 02:23 - 02:32:  of seminars. We really add a great taste to our lockdown all over the place. And I also
• 02:32 - 02:42:  want to thank the committee for choosing my abstract for a short talk presentation. So I'm not
• 02:42 - 02:48:  going to spend my precious 10 minutes to give you the background, but I'd rather
• 02:48 - 02:58:  just grab slides from Dr. Jeffrey's recent review just to give you an idea of what summarizes
• 02:58 - 03:08:  the recent progress in the M6A field in the past 10 years. So this is about genesis of M6A,
• 03:08 - 03:16:  which involves the Reiter complex and proton M6A. So this modification can be removed by
• 03:16 - 03:22:  two different demethylases, and then it's mediated by different reader proteins that
• 03:22 - 03:31:  can catalyze different biological processes, including RNA splicing, degradation, and enhanced
• 03:31 - 03:39:  mRNA translation. Well, the work in the past 10 years, about 8 to 10 years, has really
• 03:40 - 03:46:  revealed lots of functions of M6A, but those works are all done by genetic knockout or
• 03:46 - 03:56:  knockdowns. So how M6A is regulated is largely unknown. So my lab's really mainly focused on
• 03:56 - 04:01:  studying protein arginine methylation. So arginine methylation is a quite abundant
• 04:01 - 04:08:  post-translational modification on proteins, especially on arginines, and it's catalyzed by a group of enzymes called
• 04:08 - 04:16:  protein arginine methyltransferases. So there are three types of methylation in the cells.
• 04:17 - 04:24:  There's monomethylation, dimethylation, and dimethylation can be divided into asymmetric
• 04:24 - 04:32:  dimethylation and symmetric dimethylation. So in humans, there are nine arginine methyltransferases,
• 04:32 - 04:38:  they're named from PRMT1 all the way to nine. So in recent years, there has been great interest in
• 04:38 - 04:48:  targeting PRMTs in human cancers. Well, in the cells, most of the time, if you look for
• 04:49 - 04:56:  dimethylated proteins, they tend to have this motif surrounding the methylated arginine for the ADMA.
• 04:56 - 05:04:  So you will find enrichment of glycine and prolines around arginine, or for SDMA is enriched for
• 05:04 - 05:09:  glycine and arginine motifs. And these arginine methylated proteins, you can find them mainly,
• 05:09 - 05:15:  you see, many of the proteomics studies point out that they're mainly involved in transcription and
• 05:15 - 05:26:  RNA processing, which is very relevant to this event. So when we look at Methyl-14, which is
• 05:26 - 05:32:  one of the key components of the M6A methyltransferase complex, well, it contains this
• 05:33 - 05:41:  arginine glycine-rich C-terminus tail. So you can appreciate just how concentrated this RGG is
• 05:41 - 05:49:  at the C-terminus. Well, this tail is highly disordered, where many of the crystal
• 05:49 - 05:56:  structure studies cannot really define the structure of this tail, because of its flexibility.
• 05:56 - 06:02:  And this tail is very important. If you just truncate off this tail, the Methyl-3,
• 06:02 - 06:11:  Methyl-14 complex will not be active. So we look into, because of this RGG enrichment,
• 06:11 - 06:19:  we look into potential arginine methylation of Methyl-14. And we use a set of PRMTs to methylate
• 06:19 - 06:26:  in vitro using tritium-labeled SAM as a cofactor to look for Methyl-14 methylation.
• 06:26 - 06:35:  So here we show that PRMT1, 3, and 6 can methylate Methyl-14. And this is indeed at the tails,
• 06:35 - 06:42:  because if you just truncate off this Methyl-14 RGG tail, you will not see the methylation signal.
• 06:42 - 06:47:  And this methylation also happens in the cells. We can detect it with a methylation-specific
• 06:47 - 06:55:  antibody. Again, this shows that the methylation happens on the tail. If you, even in vivo,
• 06:55 - 07:00:  if you just truncate off the tail, you lose the methylation signal. And this is detected by two
• 07:00 - 07:08:  different methylation-specific antibodies. And then we next define which enzyme methylates Methyl-14,
• 07:08 - 07:14:  because in vitro, there are a few of them that can methylate. In vivo, we show that PRMT1 actually
• 07:14 - 07:21:  can co-IP with Methyl-14. And this interaction is really mediated by the C-terminal tail,
• 07:21 - 07:26:  because if you have truncated Methyl-14, it will no longer bind to PRMT1.
• 07:27 - 07:33:  And indeed, if you overexpress PRMT1, you see an increase in the methylation on Methyl-14.
• 07:33 - 07:40:  And if you knock down PRMT1 expression by siRNA, you'll lose the, decrease the methylation of Methyl-14.
• 07:40 - 07:46:  And then we also detect this methylation in multiple cancer cell lines, including lung cancer
• 07:46 - 07:53:  cell lines and breast cancer cell lines. And we use this very potent type 1 PRMT inhibitor to block
• 07:54 - 08:00:  methylation. We see that methylation of Methyl-14 is almost completely abolished
• 08:02 - 08:09:  under the inhibitor treatment. So now this is a very, I would say, very abundant modification in
• 08:10 - 08:17:  human cells, at least. And then we also attempted to identify the methylation site and
• 08:17 - 08:23:  at least use mass spectrometry. We identified five arginines that are methylated, but because of the
• 08:23 - 08:29:  enrichment of RGG, we believe there are many sites we were not able to identify, using mass spectrometry.
• 08:29 - 08:37:  So, and it indeed shows here, even if you mutate all the five arginines, you kind of only lose
• 08:37 - 08:43:  about 50% of the methylation signal. So we have to really, if you really want to create a
• 08:43 - 08:47:  methylation-deficient construct, we have to go all the way to mutate all the arginines.
• 08:47 - 08:55:  Well, anyway, the results show you here is that PRMT1 is the major arginine methyltransferase
• 08:55 - 09:01:  that can methylate Methyl-14 tails. So the tail is kind of heavily arginine methylated.
• 09:01 - 09:06:  Then the question is what's the impact of arginine methylation on Methyl-14 function?
• 09:07 - 09:12:  So we look into, because this tail has been shown to be involved in RNA binding.
• 09:12 - 09:20:  So we purified Methyl-14 under inhibitor treatment, which gave us hypomethylated Methyl-14.
• 09:20 - 09:28:  And we also included RK mutant, which is a methylation-deficient Methyl-14. And we performed
• 09:28 - 09:34:  three independent experiments to show that without arginine methylation, Methyl-14 binding with
• 09:35 - 09:41:  RNA is greatly reduced. This is an RNA cooldown experiment, this is a gel shift,
• 09:41 - 09:48:  and this is a fluorescent polarization assay. And as a consequence of loss of or reduced RNA binding,
• 09:48 - 09:56:  so the hypomethylated Methyl-14 together with Methyl-3 shows greatly reduced
• 09:57 - 10:04:  methyltransferase activity. Well, this study in vitro biochemical assay shows that activity
• 10:04 - 10:12:  decreased. Well, in vivo, people have shown that because M6A is deposited co-transcriptionally,
• 10:12 - 10:19:  RNA Polymerase II can interact with Methyl-14 and Methyl-3 complex. And we show here that arginine
• 10:19 - 10:25:  methylation-deficient Methyl-14 greatly reduces its interaction with RNA Polymerase II. Its interaction with
• 10:25 - 10:33:  RNA Polymerase III is not affected at all. And this is also true if you just inhibit Methyl-14 methylation
• 10:33 - 10:40:  using this inhibitor. So RNA Polymerase II interactions also
• 10:40 - 10:51:  reduce. And because of its interaction with RNA, we even treated with RNase A to abolish the
• 10:51 - 10:58:  association of RNA, we see a great reduction of RNA Polymerase II with Methyl-14, suggesting that it is really
• 10:58 - 11:04:  the binding to RNA that contributes to the association of Methyl-14 with RNA Polymerase II.
• 11:05 - 11:12:  So, so far, I basically show you this model that PRMT1 can methylate the tail of Methyl-14, which
• 11:13 - 11:20:  greatly enhances association with RNA as well as RNA Polymerase II. If we, if you inhibit
• 11:21 - 11:29:  PRMT1, this interaction is greatly diminished. So then the next question is, so what's the
• 11:29 - 11:38:  biological function of Methyl-14 arginine methylation in vivo? So to address that question, we obtained
• 11:38 - 11:46:  mouse embryonic stem cells from Jacob Hanna where he published the pioneer paper in science
• 11:46 - 11:51:  describing Methyl-14 and Methyl-3. So in the knockout Methyl-14 cells, we rescued expression
• 11:52 - 11:59:  using lentivirus wild-type and RK mutant Methyl-14 to have these isogenic cell lines.
• 12:00 - 12:09:  And with the help of a postdoc in Chang-Yi's lab, who helped us to define quantitatively
• 12:09 - 12:16:  the M6A versus A ratio in our cell lines. So we see a striking decrease in RK mutant
• 12:17 - 12:24:  cells because of the rescued expression of Methyl-3 and Methyl-14 is not comparable to the wild-type
• 12:24 - 12:31:  cell. That's why we see even the rescued wild-type is not reaching the real wild-type level, but clearly
• 12:31 - 12:41:  the RK mutant shows much less M6A in vivo. And we did M6A sequencing and the peak calling
• 12:41 - 12:49:  also shows the same trend. It was different degrees of loss of Methyl-14 arginine methylation. We see
• 12:49 - 12:55:  a gradual decrease of M6A peaks caught by the clam using different criteria.
• 12:56 - 13:05:  Well, the peak of the M6A in the wild-type and RK mutant, the motif remains the same,
• 13:05 - 13:12:  suggesting that the M6A peaks we caught upon M6A sites and distribution is
• 13:13 - 13:23:  as a wild-type is made in the CDS and 3 prime UTR regions. So we analyzed differential M6A sites
• 13:23 - 13:29:  and we identified about 1,700 differential M6A sites in the wild-type and RK mutant cells.
• 13:29 - 13:35:  And they located about 1,300 genes. And interestingly, these differential
• 13:35 - 13:45:  M6A sites are mainly located in long axons. And so on this 1,300 genes, we did a
• 13:45 - 13:50:  GEO analysis. We find, although it's very clear that there's stem cell maintenance, which
• 13:51 - 13:58:  is known function of M6A is enriched, but we observe interesting enrichment of error-prone
• 13:59 - 14:04:  translation DNA synthesis pathway. And here shows you that one example of
• 14:04 - 14:14:  a triplet that M6A is greatly reduced in RK mutant cells. And we tested a series of proteins
• 14:14 - 14:20:  in this pathway. And then we show that without Methyl-14 arginine methylation, their protein expression
• 14:20 - 14:28:  levels decrease. And this decrease is not due to mRNA stability or transcription. Rather,
• 14:28 - 14:34:  it's associated with reduced association with polysomes, suggesting that M6A
• 14:34 - 14:39:  on these transcripts is involved in enhancing their translation.
• 14:39 - 14:46:  So because the DNA repair pathway is dysregulated, we observed increased
• 14:46 - 14:55:  sensitivity to two different DNA damage drugs in RK mutant cells. And similarly,
• 14:55 - 15:01:  if we just use PRMT1 knockout embryonic stem cells, we observed that these cells are also
• 15:01 - 15:06:  sensitive to the DNA damage drug. As well as if you just inhibit PRMT1 activity,
• 15:07 - 15:14:  you also sensitize the cells to the DNA damage drug. So in summary, what I've showed you here
• 15:14 - 15:22:  is that we characterize a PRMT1-mediated arginine methylation of Methyl-14, which is important
• 15:22 - 15:30:  for the M6A deposition, especially a group of DNA repair genes and is helping cells to
• 15:30 - 15:38:  recover from DNA damage. Yeah, with that, I want to thank the people who did the work.
• 15:39 - 15:48:  Zhihao is the postdoc in my lab, mainly lead the project, was a co-first author with a graduate
• 15:48 - 15:55:  student in Lanling and Yixin's lab who did the M6A bioinformatics analysis. And the postdoc in
• 15:55 - 16:01:  Changyu's lab, Brian, helped us with the M6A mass spec. And we have open positions for postdoc
• 16:01 - 16:07:  fellows if you're interested. And I'll stop there and take any questions.
• 16:09 - 16:15:  Great. Thank you, Frankie. I'm going to ask the first question. It's a prerogative of the
• 16:15 - 16:25:  chairman, unfortunately. Sorry. So does arginine methylation of Methyl-14 affect the binding to
• 16:25 - 16:32:  Methyl-3? Does it have a higher affinity to Methyl-3, perhaps, with the opening up of the structure
• 16:32 - 16:41:  by arginine methylation? No. We tested, we basically excluded lots of possibilities in
• 16:41 - 16:48:  protein-protein interactions. We tested Methyl-3 interaction, tested WTAP interaction. It's not
• 16:48 - 16:55:  affected at all. Yeah. Thank you. Xuan, you raise your hand. Yes.
• 16:55 - 17:02:  Just to follow Tony's question, have you tried proteomics? Does the methylation of Methyl-14
• 17:03 - 17:09:  sort of switch it to become a transcriptional factor, as Tony showed previously and others?
• 17:10 - 17:14:  Because it could be methylation-independent. You clearly show the methylation-dependent
• 17:14 - 17:17:  function, but there could also be additional methylation-independent function.
• 17:18 - 17:26:  Yeah. Very good question. We actually follow that direction. We conducted proteomics studies on
• 17:26 - 17:32:  WTAP, RK mutant Methyl-14, and we're working on that. We have two biological repeats. We have
• 17:32 - 17:37:  some interesting candidates to work on, but that's a good point to look into the
• 17:37 - 17:44:  M6A independent function of Methyl-14. Yeah. We did have a list of proteins that show
• 17:44 - 17:51:  differential interaction with Methyl-14 when it's WTAP versus RK mutant. Yeah.
• 17:52 - 17:54:  So, Frankie? Yeah.
• 17:54 - 18:02:  My question is related to, again, this interaction. You showed that Methyl-14 was not coming down
• 18:02 - 18:11:  with RNA polymerase when it wasn't methylated, but Methyl-3 was. And that's where Tony's question
• 18:11 - 18:20:  was about regulating the Methyl-3, Methyl-14 heterodimer. Your data clearly shows that
• 18:20 - 18:26:  Methyl-14's binding to polymerase, which is presumably due to its ability to bind Methyl-3,
• 18:26 - 18:31:  is dependent on methylation states. So, I guess I'm trying to understand how you think
• 18:31 - 18:37:  Methyl-14 can differentially associate with the polymerase when it's known to form a very tight
• 18:37 - 18:42:  complex with Methyl-3. And I would also like to add, if you know anything about phase separation,
• 18:42 - 18:49:  you know, these RGG domains are really strongly linked to phase separation. And I'm wondering
• 18:49 - 18:56:  if phase separation could somehow contribute to the effects of this methylation. So, sorry,
• 18:56 - 18:59:  two questions there. Yeah. Yeah. So, the first question,
• 18:59 - 19:08:  I think maybe I somehow confused you. So, for the RNA Polymerase II interaction, Methyl-14 has been,
• 19:08 - 19:14:  we detect Methyl-3 all the time with Methyl-14 regardless of its methylation status. It is
• 19:14 - 19:22:  the RNA Polymerase II interaction that is sensitive to Methyl-14 arginine methylation. So, in that sense,
• 19:22 - 19:29:  we don't think that there's a Methyl-3 independent function of Methyl-14 in interaction with RNA Polymerase II.
• 19:30 - 19:34:  Yeah. Second question. So, we, yeah, we initially,
• 19:34 - 19:41:  we spent about a year and a half working on phase separation of Methyl-14. We worked with
• 19:42 - 19:49:  Paul Taylor from St. Jude Hospital. So, we, at least in our hand or in his hand,
• 19:49 - 19:57:  some of the proteins he has been working on as easily phase-separated targets, we were not able
• 19:57 - 19:59:  to see Methyl-14 phase separation.
• 20:00 - 20:06:  First, we have not tried, I can tell you this, we have not tried to add Methyl-3 together.
• 20:06 - 20:14:  We're only working on Methyl-14 wild-type versus RK mutant. And we have not added RNA in the
• 20:14 - 20:21:  system. So, those are the things, it's there. So, we just have some, need to have some good
• 20:21 - 20:27:  hands to test that out. Thank you, Frankie. I'm going to choose one
• 20:27 - 20:38:  question from the audience because we're running out of time. So, John Cui asks, does
• 20:38 - 20:47:  Methyl-14 arginine methylation mutant affect, what's the effect on RNA Polymerase II's activity,
• 20:47 - 20:52:  either in elongation speed or transcription level?
• 20:52 - 20:58:  Very good question. So, we are now having, halfway down, we're working on manuscript.
• 20:58 - 21:04:  I really look at when we have RK mutant Methyl-14, what happened to RNA Polymerase II elongation.
• 21:04 - 21:11:  We have not taken a genome-wide approach, but rather focusing on those long axons where
• 21:11 - 21:17:  we see this differential M6A happen. Yeah, thank you for the question.
• 21:18 - 21:24:  Great. Thank you, Frankie. That's fantastic. We're going to move on now to the second short
• 21:24 - 21:32:  talk by Abdul-Qadir Abakir from the Babraham Institute in Cambridge. And he's going to
• 21:32 - 21:38:  tell us about the role of M6A in R-loop biology. Over to you, Abdul-Qadir.
• 21:38 - 21:44:  Well, thank you so much, Professor Kosmideridis, and for the organizers for this wonderful
• 21:45 - 21:54:  chance to share our work, which we've done in the RISOF hub in Nottingham. The title of which is
• 21:54 - 21:59:  the role of M6-methyladenosine in R-loop biology. And currently, I'm working, basically,
• 21:59 - 22:02:  I just started my postdoc in the RIOT Lab at the Babraham Institute.
• 22:02 - 22:15:  So, as Frankie has just elegantly introduced, I don't probably need to revisit this introduction
• 22:15 - 22:22:  of methylation of adenosine. But what is really important is that the discovery of this FTO and
• 22:22 - 22:30:  ALKBH5 has brought this field out of dormancy. This was basically one of the contributions of
• 22:30 - 22:36:  the Zhuanghe Lab, and the ability to map them and as well as the identification subsequently of
• 22:36 - 22:46:  M6A-binding proteins led to the identification of several M6A-dependent RNA-binding proteins that
• 22:46 - 22:54:  subsequently implicated M6A in almost all of the metabolic steps of RNA.
• 22:54 - 23:04:  But the nuclear function of this modification is much less clear. So, to make things a little
• 23:04 - 23:11:  bit more complicated, in 2015, there were three papers that appeared in Cell to show that what
• 23:11 - 23:23:  had been restricted to bacteria, basically the presence of 6mA or N6-methyldeoxyadenosine
• 23:23 - 23:30:  in the genomic DNA of bacteria has been shown to exist also in multicellular eukaryotes,
• 23:31 - 23:38:  such as nematodes, fruit flies, and green algae. However, the findings in mammals have been a
• 23:38 - 23:47:  little bit contentious. So, some labs report that 6mA exists in mammalian cells, and some
• 23:47 - 23:53:  other labs also do not seem to corroborate those findings. So, we wanted to study this.
• 23:54 - 23:59:  So, using a highly sensitive immunochemical method that I helped to develop during my
• 24:00 - 24:06:  MSc, and there is a flap. Basically, what we do here, if you have a low abundant DNA modification,
• 24:06 - 24:12:  for example, normally, if there is an antibody to it, you could identify it with the help of
• 24:12 - 24:20:  secondary antibody. But what we've done is we use a second, horseradish peroxidase conjugated
• 24:20 - 24:26:  secondary antibody, which in the presence of hydrogen peroxide, can convert an inactive
• 24:26 - 24:31:  tyramide to an active form. This activated tyramide can be deposited all
• 24:31 - 24:39:  along the tyrosine residues in the vicinity of your DNA modification. And basically, using this
• 24:39 - 24:46:  method, what we have found is that there is a transient accumulation of active DNA demethylation
• 24:46 - 24:52:  intermediates, like 5-formylcytosine and 5-carboxylcytosine, during the lineage specification
• 24:52 - 24:58:  of neural stem cells. So, using this highly sensitive immunochemical method, what we've
• 24:58 - 25:04:  wanted to look at is whether human pluripotent stem cells, which are kind of typically normal,
• 25:04 - 25:11:  whether they also possess what have been reported in some other studies, DNA-based
• 25:12 - 25:18:  adenosine methylation. So, we've done immunostaining of M6A or 6mA here with
• 25:18 - 25:24:  5-carboxylcytosine marked in red. And as you can see here, we see a really nice signal
• 25:25 - 25:33:  of M6A in green in the cytoplasm as well as in the nucleus. Since we were interested in
• 25:33 - 25:44:  DNA modification of 6mA, we removed what would be the RNA source of this cytoplasmic signal.
• 25:44 - 25:51:  And to our surprise, what we saw was that there is an RNA-resistant nice M6A signal.
• 25:53 - 25:57:  So, we were excited a little bit by this, thinking that this
• 25:57 - 26:07:  is a strong signal in these mitotic cells. Basically, it's coming from a DNA source.
• 26:09 - 26:15:  So, to validate or to see whether this is the case, we've isolated genomic DNA from both
• 26:16 - 26:21:  iPS cell lines as well as human embryonic stem cell lines, whether they are mitotically enriched
• 26:21 - 26:27:  with CarrierMax, which carries basically a calcium agent that suspends cells from progressing
• 26:27 - 26:33:  during mitosis or just normal untreated cells. So, what we could detect is we could see
• 26:33 - 26:45:  5-hydroxymethylcytosine, 5-formylcytosine, which are at a very low level, but we couldn't really
• 26:45 - 26:55:  detect any 6mA. So, there is clearly some discrepancy here. On the one hand, you see
• 26:55 - 27:03:  a nice RNA-resistant signal by real staining, but by mass spectrometry, which is the gold
• 27:03 - 27:09:  standard for detection of this DNA modification, we don't see it. So, what is this source of
• 27:10 - 27:19:  mitotic chromatin-associated 6mA? So, we reasoned that it may be actually on an RNA component that
• 27:19 - 27:26:  is part of a DNA-RNA hybrid or so-called R-loops. So, these are three nucleotide structures formed
• 27:27 - 27:31:  predominantly co-transcriptionally. So, the RNA, basically, during transcription,
• 27:32 - 27:37:  as it's coming out of the C-terminal domain, it can basically re-anneal
• 27:37 - 27:44:  to its template DNA, and this leaves out this single-stranded DNA vulnerable to genotoxic
• 27:44 - 27:48:  attacks. So, these structures are actually quite abundant in our genome, ranging from 4% to 10%,
• 27:50 - 27:57:  and they can also have several implications. For example, they can protect promoters from
• 27:57 - 28:04:  methylation, they are involved in transcription regulation, and they can be quite unpleasant if they
• 28:04 - 28:09:  accumulate at excessive levels. So, there are two main ways to study these structures: Either
• 28:09 - 28:15:  use a monoclonal antibody called S9.6 antibody, which binds to the DNA-RNA hybrid,
• 28:16 - 28:26:  or, basically, even in bacteria, there is this RNAse ribonuclease that specifically cleaves RNA
• 28:27 - 28:33:  molecules when they are only in hybrid with DNA.
• 28:35 - 28:40:  So, we wanted to test if, indeed, this M6A we see in mitotic cells is coming from
• 28:41 - 28:46:  this RNA component of the hybrid. So, what we've done, first, we repeated
• 28:47 - 28:52:  this immunostaining, first, without any RNAse treatment, and as before, we see
• 28:53 - 29:00:  M6A signal here in green, and we communostained with the phospho-Histone H3 to identify the mitotic
• 29:00 - 29:07:  chromatin, and M6A is detectable both in the cytoplasm and basically in the nucleus of these
• 29:07 - 29:14:  cells, and it goes away pretty much in all the interphase cells, but there is M6A in mitotic
• 29:14 - 29:23:  chromatin, and when we introduce RNAse treatment prior to the immunostaining, remarkably, this M6A signal
• 29:23 - 29:28:  in the mitotic cells completely disappears or reduces substantially. So, this suggests that at
• 29:28 - 29:38:  least this mitotic chromatin M6A signal is present on the RNA component of the RNA-DNA hybrid.
• 29:38 - 29:43:  To validate this, basically, what we've done is to isolate genomic DNA,
• 29:43 - 29:49:  separate it into two fragments, either treating it with RNAse, which degrades
• 29:49 - 29:55:  the RNA component of the hybrid, or without any RNAse treatment, and we pass this through
• 29:55 - 30:04:  ultrafiltration, and we look by mass spectrometry in the retentate. So, what we find is that M6A
• 30:04 - 30:12:  is actually substantially released after RNAse treatment, and interestingly, other RNA modifications
• 30:12 - 30:17:  like 5-methylcytidine do not really change substantially even after RNAse treatment,
• 30:17 - 30:25:  suggesting that at least M6A is indeed present on RNA. So, this is a very important finding
• 30:25 - 30:33:  because it illustrates that, for the first time, DNA-RNA hybrids have a non-canonical base modification.
• 30:33 - 30:47:  We wanted to know where these modifications exist in our genome, so we adapted what is called DNA-RNA hybrid immunoprecipitation, or DRIP.
• 30:47 - 30:05:  In this protocol, what we did is we isolated genomic DNA fragments and then enriched for M6A-modified DNA-RNA hybrids, and then sequenced the template genomic DNA, and using basically a
• 31:05 - 31:16:  peak-calling method, we could find a lot of M6A DNA-RNA hybrids, as well as either with S9.6 antibody,
• 31:16 - 31:22:  which is a monoclonal antibody, or with M6A antibody. But what is really important is that
• 31:22 - 31:28:  when we do a parallel experiment, where we pre-treated the nucleic acid with RNAse,
• 31:28 - 31:37:  those peaks seem to disappear. So, what is also interesting at the genome-wide level,
• 31:37 - 31:45:  we see the distribution of M6A and S9.6 are almost identical, and this falls into about 50%
• 31:45 - 31:51:  in dissonance regions, but we also see them in a lot of other regulatory regions, as well as
• 31:51 - 31:59:  exons and 3' and 5' UTR, and some of it also in introns. So, they really overlap nicely,
• 32:00 - 32:09:  the M6A and S9.6 IP DNA-RNA hybrids. So, we know the genome-wide distribution of these
• 32:10 - 32:16:  hybrids that are M6A modified. You saw this image earlier, that in mitotic cells, you have
• 32:16 - 32:22:  a higher level of M6A immunostaining signal that is RNAse resistant. Does this suggest that
• 32:22 - 32:30:  these hybrids are actually being deposited with M6A during mitosis only, or that they have different dynamics during the cell cycle? To address this question, what we've done is to
• 32:35 - 32:43:  separate those cells based on their DNA content using propidium iodide, either in G0, G1, S-phase,
• 32:43 - 32:50:  or G2/M, and looked at the genomic loci, like LINE-1 and SINE-Alu, that have enrichment of
• 32:50 - 32:59:  DNA-RNA hybrids that are M6A modified, but also as a control we used alpha satellite repeats that
• 32:59 - 33:08:  show poor enrichment or no enrichment of M6A DNA-RNA hybrids. And by qPCR, we could see that
• 33:08 - 33:16:  at least in G0, G1 cells, there is very low level of DNA-RNA hybrid. Then they start to appear
• 33:16 - 33:23:  or accumulate during S-phase, and they max out by G2/M, the stages of the cell cycle.
• 33:23 - 33:27:  But what is really interesting is that just before you re-enter the cell cycle, they are being removed.
• 33:28 - 33:36:  And we also tested intronic sequences using S9.6 and M6A direct qPCR. You can see that
• 33:36 - 33:43:  in G0, G1, similar to repeats, there is a low level of DNA-RNA hybrid, but they seem to accumulate
• 33:43 - 33:49:  more during S-phase. And they also similarly are being removed before you re-enter the cell cycle.
• 33:50 - 33:56:  So, what is actually M6A doing, the biological function of M6A? To test this question,
• 33:56 - 34:03:  we looked at the proteins that either introduce M6A, like Methyl-3, or reader proteins that are
• 34:03 - 34:16:  known to basically bind M6A to interpret what this signal means. So, this is a slot plot done
• 34:16 - 34:24:  in our collaborator, Natalia Groma, in Oxford. We could see basically YTHDF1, YTHDF2, which is
• 34:24 - 34:32:  implicated in the degradation of M6A-modified transcripts, and also HnRNP2B1, which is involved
• 34:32 - 34:42:  in splicing of some microRNAs. So, what we found is that YTHDF2 co-localizes nicely to the mitotic
• 34:42 - 34:48:  chromatin where we saw enrichment of DNA-RNA hybrid. To test this using a different method,
• 34:49 - 34:55:  we immunostained with the monoclonal antibody for DNA-RNA hybrid, S9.6. And as you can see here,
• 34:55 - 35:02:  there is a very nice co-localization of the immunostaining signal of both YTHDF2 and S9.6,
• 35:02 - 35:11:  suggesting that YTHDF2 interacts with DNA-RNA hybrids in vivo. So, to test if this interaction
• 35:11 - 35:18:  is direct, what we've done is to use synthetic DNA-RNA hybrids, whether they are actually M6A
• 35:18 - 35:25:  modified or without. And by EMSA, what we saw is that we confirmed what is known previously
• 35:27 - 35:34:  basically by the Chuan-Hei lab, is that M6A interacts with YTHDF2 if it is methylated.
• 35:34 - 35:42:  But what is also interesting is that DNA-RNA hybrids seem to show more efficient binding of
• 35:42 - 35:51:  YTHDF2 if they are also methylated. So now, YTHDF2 seems to interact directly with DNA-RNA hybrid if
• 35:51 - 35:58:  they are methylated, and also co-localizes with the DNA-RNA hybrid marker in vivo. What's the
• 35:58 - 36:07:  function of YTHDF2? To probe this question, we basically got a CRISPR-Cas9 HAP1 human cell
• 36:07 - 36:18:  in which this YTHDF2…YTH domain, which YTHDF2 binds to…uses to bind m6A, is abolished in the
• 36:18 - 36:23:  knockout. And in this knockout, you can see by immune staining, in the wild type, we see
• 36:23 - 36:29:  co-localization of YTHDF2 to the mitotic chromatin, and this is abolished in the knockout.
• 36:30 - 36:37:  These cells grow slower, similar to what has been reported in mouse YTHDF2 knockout
• 36:37 - 36:44:  neural stem cells. But also, what is really nice to see was that YTHDF2 knockout cell lines,
• 36:44 - 36:51:  they actually show accumulation of DNA-RNA hybrids, both in the repeat regions and also
• 36:51 - 36:57:  the candidate intronic sequences that we looked at. And we confirmed this genome-wide by
• 36:57 - 37:08:  looking at m6A species of the IP DNA-RNA hybrids. So, there is more DNA-RNA hybrids in the knockout
• 37:08 - 37:15:  versus the wild type. This suggests that actually YTHDF2 may be required to keep a lid on the
• 37:15 - 37:24:  amount of DNA-RNA hybrids, so without which you see accumulation of m6A-modified DNA-RNA hybrids.
• 37:24 - 37:33:  So, one of the best-characterized consequences of accumulating DNA-RNA hybrids is that the presence
• 37:33 - 37:39:  of this single-stranded looped-out DNA can be basically vulnerable to genotoxic attack,
• 37:39 - 37:47:  and this eventually leads to DNA damage. To see if the accumulation of DNA-RNA hybrids in YTHDF2
• 37:48 - 37:57:  is indeed causing this, we looked by immunostaining using gammaH2AX antibody.
• 37:57 - 38:05:  And as you can see here, in the knockout, you see accumulation of this gammaH2AX-modified
• 38:06 - 38:13:  chromatin regions, suggesting that there is accumulation of DNA damage. Is this DNA damage
• 38:13 - 38:20:  related to the accumulated DNA-RNA hybrids? To test this, basically, we looked why we've
• 38:20 - 38:27:  transfected those cells that are knockout with a GFP-tagged RNAase H, and we looked
• 38:27 - 38:34:  by immunostaining if this is indeed, these cells show some rescue, and that is indeed the case,
• 38:34 - 38:40:  suggesting that actually the accumulation of this DNA damage in YTHDF2 knockout
• 38:41 - 38:49:  cell lines, both in mice as well as in human HUB1 cells, is that because we see accumulation
• 38:49 - 38:58:  of DNA-RNA hybrids. So, we not only show that the majority of DNA-RNA hybrids have
• 38:59 - 39:06:  m6A, but this is also cell cycle regulated. For example, in G0, G1, you have a very low
• 39:06 - 39:16:  level of DNA-RNA hybrids that may be implicated in regulating events such as splicing. YTHDF2
• 39:16 - 39:23:  is precluded from entering because of an intact nuclear membrane, but as you move to S phase,
• 39:23 - 39:32:  these hybrids start to accumulate or appear. And then by the mitotic cells,
• 39:32 - 39:40:  during mitosis, the nuclear membranes start to basically disappear, the amount of m6A modification
• 39:40 - 39:49:  or m6A deposition maxes out. Somehow, this binding of YTHDF2 leads to the removal of this hybrid before
• 39:49 - 39:58:  you re-enter the cell cycle. So, YTHDF2-mediated degradation of the hybrid may allow the cells
• 39:58 - 39:59:  actually to combat.
• 40:00 - 40:06:  accumulation of R-loops during the cell cycle. If you'd like to know more, this work is now
• 40:06 - 40:16:  being published. And I would like to thank my PI, Alexey Rizov, for his basically supreme
• 40:16 - 40:21:  guidance of this project and helping me and several of our collaborators, Professor Arne
• 40:21 - 40:29:  Blumlein for the mice work and Natalia Gromak for the validation by EMSA and Daniel Krakowski
• 40:29 - 40:35:  for the quantification by MassSpec and several other collaborators. Thank you for listening.
• 40:46 - 40:50:  So, Tony, I have a question. Tony isn't there.
• 40:51 - 40:58:  Trying to undo my video, but it's not working. Did somebody override me from
• 40:58 - 41:06:  Abcam? I think they might have done. Yeah. Okay, good. Thank you. So, yes,
• 41:06 - 41:12:  Sammy, please have the first question. So, a great talk. Very interesting. I just have a
• 41:12 - 41:18:  comment and a question. So, my comment is that, you know, structurally, it isn't clear how a
• 41:18 - 41:26:  RNA-DNA hybrid would bind YTHDF2. You know, there's a lot of crystallography. The RNA fits
• 41:26 - 41:33:  inside a groove, a circuitous groove. It doesn't seem like it would be able to bond and the m6A is
• 41:33 - 41:40:  flipped out to fit into a pocket. An RNA-DNA hybrid would have the m6A locked into a helix
• 41:40 - 41:47:  and the m6A would not be flipped out. And so, I'm sort of not clear how this interaction could
• 41:47 - 41:53:  occur, and I'm wondering if you have any thoughts on that. But I also was just wondering if the m6A,
• 41:53 - 41:58:  if you think the m6A is located with specific types of R-loops. You know, there's R-loops near
• 41:58 - 42:02:  transcription termination sites. There's R-loops near initiation sites. There's also R-loops,
• 42:02 - 42:07:  you know, we've mapped some actually very stable R-loops on certain loci, like in histone genes
• 42:07 - 42:13:  and malat1 and a few other places. Do you have a sense of if there are particular R-loops
• 42:13 - 42:20:  that may be enriched in m6A, which might give an indication of a functional role in gene
• 42:20 - 42:25:  expression? So, you know, I just was wondering if you have thoughts on either of those comments.
• 42:25 - 42:33:  Okay. Thank you so much for these questions. They're really important. In regards to how
• 42:33 - 42:40:  m6A or YTHDF2 can interact with m6A-modified DNA-RNA hybrid, the short answer, we don't know, but
• 42:40 - 42:44:  how we detect them by immunostaining is that basically we denature them using
• 42:44 - 42:52:  hydrochloric acid. So this allows the relaxation of what would be a completely sealed hybrid,
• 42:52 - 43:00:  and that allows the antibody to access. In vitro, we place m6A in this putative motif,
• 43:00 - 43:09:  like DGAC motif, and then when we basically put the antibody together with the synthetic hybrid,
• 43:09 - 43:15:  we see a more preferential binding compared to non-m6A-modified. So...
• 43:17 - 43:22:  Okay. That sounds great. And do you have a sense of which R-loops might be most methylated?
• 43:23 - 43:29:  We find all sorts of R-loops that are being methylated. We know this because in parallel
• 43:29 - 43:41:  experiments, we run the S9-6 IPed R-loops. So our findings suggest that at least almost,
• 43:41 - 43:50:  pretty much almost all of the R-loops, about, I think, 23,000, only about, I think, 211 R-loops
• 43:50 - 43:57:  that are not picked up by the m6A, but are picked up by the S9-6. So suggesting that almost all of
• 43:57 - 44:04:  the R-loops are actually methylated. And what is really interesting, this is dynamic, of course,
• 44:04 - 44:10:  in vivo, this would be a dynamic, but we think actually the modification, this m6A is being
• 44:10 - 44:20:  deposited along the transcription, the so-called transcription. So that's how I think we find them.
• 44:20 - 44:22:  Okay. Thank you. All right.
• 44:23 - 44:26:  Great. Thank you. Unfortunately, we've already, you know, 50 minutes behind,
• 44:26 - 44:33:  so I'm going to stop the questions there. Thank you, Dr. Khadir. I'm now going to have a short
• 44:33 - 44:40:  presentation from Eleanor Walton from Abcam, and then we're going to have the plenary talks. So,
• 44:40 - 44:42:  Eleanor, over to you.
• 44:46 - 44:48:  Thank you, Tony. I'll just share my slides.
• 44:53 - 44:58:  Okay. Yeah. So my name is Eleanor Walton. I work for Abcam, and I am the research area lead for
• 44:58 - 45:03:  epigenetics. So I work with the new product development, and I help oversee and help
• 45:03 - 45:08:  support development of our new antibodies and reagents as we produce them for the epigenetics
• 45:08 - 45:12:  field. So thank you to the speakers who've already gone before me, and I'm really looking forward to the
• 45:12 - 45:17:  talks later. I'm really looking forward to the panel discussion as well. So epigenetics at Abcam,
• 45:17 - 45:22:  I just want to take a couple of minutes to talk about this. So epigenetics has been one of our
• 45:22 - 45:26:  core research areas since, I think, the founding of the company, really. And over this time period,
• 45:26 - 45:32:  we've obviously seen how epigenetics is playing an ever-increasing role in life sciences. And
• 45:32 - 45:37:  here at Abcam, we are committed to providing high-quality tools that help you to research
• 45:37 - 45:41:  these areas. So part of the things we're doing more generally is that we are committed to making
• 45:42 - 45:48:  really highly validated recombinant monoclonal antibodies, and this is because monoclonal
• 45:48 - 45:54:  antibodies have really high reproducibility, so you can have confidence in security of supply.
• 45:55 - 46:00:  We're providing guides and protocols to help you get the best out of the tools we're providing.
• 46:01 - 46:06:  We are engaging researchers to provide platforms to help disseminate new ideas, which hopefully you
• 46:06 - 46:11:  are experiencing and enjoying at the moment during this event. We also collaborate with
• 46:11 - 46:17:  researchers to address areas of unmet need, and a big part of my job is to talk to researchers
• 46:17 - 46:22:  in the field and find out what we're missing and how we can support you better. One of the things
• 46:22 - 46:28:  we're doing is we are increasing, making everything easier to find on our website, so we now have an
• 46:28 - 46:35:  epigenetics resource hub, and we now have some very fancy QR codes, I think, on all the next slides,
• 46:35 - 46:39:  so you can always look at what I'm talking about by using your smartphone to look at those.
• 46:39 - 46:46:  But this is a good launch pad for all things epigenetic at Abcam. So one of the things we're
• 46:46 - 46:51:  really proud of is that we have the Abcam Epigenetics Application Guide. It's one of the
• 46:51 - 46:57:  bigger guides we host on our website. It's got a lot of different techniques, and it's a live
• 46:57 - 47:02:  document where we're adding new stuff all the time. So we've just added a new chapter on cut and run
• 47:02 - 47:08:  and cut and tag, and that was done in conjunction with some of our collaborators. And, you know,
• 47:08 - 47:13:  we're really keen that if there is anything you'd like to see or anything you think we should add,
• 47:13 - 47:17:  please, please, please get in contact with us and let us know what you would like to add
• 47:17 - 47:24:  to that guide. Also on our website, you'll find information on several different pathways that
• 47:24 - 47:29:  we've got. I think in the pre-video, there was a lot of links to the posters and stuff that we've
• 47:29 - 47:35:  got going on. So I just wanted to highlight our histone H3 mutation pathway that we've got,
• 47:35 - 47:41:  and this gives a guide to how these mutations are, how they are in cancer. And we've also got
• 47:42 - 47:51:  a series of antibodies to those targets as well. So we've got our H3K27 mutant, our H3K36 mutant,
• 47:51 - 47:57:  and we've also got an antibody to the H3G34, and we've got the W, the V, and the R mutants
• 47:57 - 48:03:  all available for you to use. So as this is the RNA modifications and epitranscriptomic session,
• 48:03 - 48:07:  I just want to highlight some relevant products that are on our catalog. I don't think I need
• 48:07 - 48:12:  to tell anyone in this event what these antibodies, what the targets of these antibodies do,
• 48:13 - 48:17:  but especially as I think they've just been quite well introduced in the previous talk.
• 48:17 - 48:23:  So we have, we're focusing on the readers, the writers, and the erasers. So, you know, for
• 48:23 - 48:28:  example, we've got a lot of the methyl antibodies. So here we've got methyl 3. We've got a lot of the
• 48:28 - 48:34:  erasers. So we've got, here we have FTO. We also have, I think we've got quite a lot of the
• 48:35 - 48:40:  readers. I mean, we've got quite a few of the YTHDF family available now,
• 48:41 - 48:45:  but also we have antibodies to some of the modifications themselves, the modifications themselves.
• 48:45 - 48:51:  Sorry. So here we've just highlighted our acetyl 4C antibody that is available for everybody.
• 48:53 - 48:57:  We are releasing new products all of the time and new antibodies all the time. And that's a
• 48:57 - 49:03:  massive part of my job is helping with the validation of these and making sure they all
• 49:03 - 49:08:  are very, very high quality and well validated and validated in a way in which you as epigenetics
• 49:08 - 49:14:  researchers would expect. We have a new product release page as well, if you wish to have a look
• 49:14 - 49:21:  at that on the website. And also if you have any further questions, we have our email address,
• 49:21 - 49:26:  which is epigenetics@abcam.com. And please, you know, email us if you have any questions you'd
• 49:26 - 49:36:  like to ask. Thank you, Eleanor. Now we've come to the part of the meeting which had plenary talks.
• 49:36 - 49:44:  We have three consecutive plenary talks from what I think are amazing people and leaders in the
• 49:44 - 49:50:  field that really make the field happen. I'm not going to give long introductions to them because
• 49:51 - 49:56:  I don't want to compare anything and I don't want to embarrass them. So I'm just going to
• 49:56 - 50:04:  introduce them very simply. And the first speaker will be Sammy Jeffrey from Cornell University in
• 50:04 - 50:11:  New York. And the title of his talk is Understanding the Function and Dynamics of m6A in the EPIC
• 50:11 - 50:18:  Transcriptome. Over to you, Sammy. All right. So thank you, Tony. Thank you for that introduction.
• 50:18 - 50:22:  So I want to talk now about what does m6A actually do. And we heard a little bit about
• 50:22 - 50:30:  m6A readers. There are five major m6A readers. YTHDF1, YTHDF2, YTHDF3 are preceded usually by the words
• 50:30 - 50:36:  YTH. We just call it DF for short. And it's also it gets confusing for people outside the field
• 50:36 - 50:42:  because they all sound the same. So YTHDF1, 2, and 3 are cytoplasmic readers. YTHDC1 is a nuclear reader.
• 50:42 - 50:47:  And YTHDC2 is a testes type reader. It's very confusing. It doesn't find m6A very well. But,
• 50:47 - 50:52:  you know, it's still often called an m6A reader. So the top four are the major proteins. And if
• 50:52 - 50:56:  you look at them, I've shaded in green the low-complexity domains. They pretty much are
• 50:56 - 51:01:  all just low-complexity domain proteins. They don't have recognizable domains. And the DF
• 51:01 - 51:06:  proteins are very similar to each other. They're paralogs. They evolved recently. They are not
• 51:06 - 51:11:  homologs. At YTHDF1, 2, and 3 in lower organisms, they usually only have one DF protein. So the
• 51:11 - 51:15:  question is, what do they do? And in fact, there's been so much great research that has helped to
• 51:15 - 51:20:  reveal the function of these YTHDF proteins. And the function has really argued that these
• 51:20 - 51:25:  proteins are different, that you have three proteins in the cytosol. They all bind to m6A.
• 51:25 - 51:30:  YTHDF1 regulates translation. YTHDF2 regulates degradation. And YTHDF3 has this interesting
• 51:30 - 51:35:  combination effect of translation and degradation. And the other thing that's fascinating is that
• 51:35 - 51:41:  even though all three of them bind m6A, they bind different transcripts. At least this has been
• 51:41 - 51:45:  reported. So they have different functions, and they bind different transcripts. And this is a
• 51:45 - 51:51:  great example of a Venn diagram where people have tried to do an overlap analysis from CLIP data
• 51:51 - 51:57:  to see which RNAs bind YTHDF1, YTHDF2, and YTHDF3. And you can see, in fact, only a small subset of transcripts
• 51:57 - 52:03:  bind all three. Most transcripts uniquely bind YTHDF1 or YTHDF2. And therefore, it's really
• 52:03 - 52:08:  caused so many people to, when they discover an m6A mapping result, to start doing CLIP experiments
• 52:08 - 52:16:  to figure out which YTHDF protein is binding their RNA. And it's really been supported also
• 52:16 - 52:22:  by these amazing studies where people have done CLIP analysis of different YTH proteins and see
• 52:22 - 52:27:  dramatically different results. And so in the top result, you have YTHDF1 CLIP binding to the HIV
• 52:27 - 52:34:  genome. And you have a site there which binds YTHDF1, and YTHDF2 does not bind it. YTHDF2 binds
• 52:34 - 52:40:  other places. And so these dramatic differences have really animated the field and really sent
• 52:40 - 52:45:  so many researchers in the direction of trying to figure out which reader is binding to which
• 52:45 - 52:54:  m6A site. And it's really been a fascinating and really exciting area of biology because of these
• 52:54 - 52:59:  findings. But, you know, the problem, I think, and I don't want to say problem, but the fascinating
• 52:59 - 53:05:  aspect is if you look at the reader domain, the YTH domain, in YTHDF1 and YTHDF2 and YTHDF3,
• 53:05 - 53:09:  I've overlaid them. And if the residues are identical, I colored them in gray.
• 53:09 - 53:15:  They're exactly identical on the side that faces the m6A RNA. And that's in pink is the m6A RNA.
• 53:15 - 53:19:  It's only when you flip it on the back, you see some changes, but these changes are really minor,
• 53:19 - 53:24:  like serine to threonine, isoleucine to leucine. So the question is, and this was really the
• 53:24 - 53:29:  question that my postdoc, Sarah Zakura, asked, how can these proteins, which look so similar
• 53:29 - 53:35:  throughout their sequence, have such divergent functions and binding properties? And this is
• 53:35 - 53:40:  really, I think, not even a question in m6A. This is a general question because it defies our basic
• 53:41 - 53:44:  understanding of how proteins work. And, you know, when you think of different proteins,
• 53:44 - 53:49:  different kinases, they usually look different. And in fact, these proteins look very, very similar.
• 53:49 - 53:54:  And so how can they, but, you know, there are subtle differences, but how could they have such
• 53:54 - 53:59:  divergent functions? And so when she started to look at this, she realized that actually many of
• 53:59 - 54:06:  those CLIP studies actually had some deep flaws, which had argued that the proteins actually bind
• 54:06 - 54:11:  to different places. So this was actually work done by Brian Cullen, but he reanalyzed the CLIP
• 54:11 - 54:16:  data. And he showed that when you look at the CLIP data that had shown these differential binding
• 54:16 - 54:22:  effects, the reads were on average five to seven nucleotides in length. And if an RNA fragment
• 54:22 - 54:27:  that's bound to YTHDF proteins is just seven nucleotides, it doesn't map to one place in the
• 54:27 - 54:31:  genome. It can map to thousands or hundreds of thousands of places. And so it will only map to the
• 54:31 - 54:38:  first place that it encounters. And so basically, a lot of challenges were raised about the CLIP,
• 54:38 - 54:44:  at least about some of the CLIP datasets, saying that the CLIP datasets were not actually reliable.
• 54:44 - 54:50:  And so that helped us a little bit because it said maybe they aren't as different as we think.
• 54:50 - 54:56:  And so when you look at how the sites were mapped, Sarah also noticed that if you just look at the
• 54:56 - 55:03:  mapping results and you look at the purple versus the pink of YTHDF1 and YTHDF2, this is with a
• 55:03 - 55:08:  high-quality CLIP dataset, you might say, oh, they look identical. But once you put a threshold and
• 55:08 - 55:13:  you say only peaks above that threshold are called, then you get very different calling.
• 55:13 - 55:17:  So even though your eyes might tell you that these are the same results, the data would tell
• 55:17 - 55:21:  you that they're actually different. And so thresholds are actually often discouraged in
• 55:21 - 55:29:  this type of analysis. Instead, Sarah reanalyzed the data where she measured the YTHDF1 and YTHDF2
• 55:29 - 55:35:  reads at every individual m6A site in the genome. And of course, we mapped them at single nucleotide
• 55:35 - 55:41:  precision. And we can just plot on each axis the coverage of YTHDF1, DF2, we've done DF2, DF3,
• 55:41 - 55:45:  we've done every pairwise combination. And basically, instead of finding these unique
• 55:45 - 55:51:  sites, which Sarah had hoped to find, Sarah wanted to really find which sites uniquely found DF1 and
• 55:51 - 55:57:  which sites uniquely found DF2 to understand the basis for that. She instead found that basically,
• 55:57 - 56:02:  there was a linear correlation. And this was a cross-CLIP dataset. She did her own CLIP dataset.
• 56:02 - 56:07:  Every high-quality CLIP dataset from different labs produced the exact same correlation.
• 56:07 - 56:13:  And so what she realized is there are not unique sites bound by DF1 and DF2 and DF3. And that
• 56:13 - 56:18:  really makes a lot of sense based on the structure. But it was disappointing to her because she
• 56:18 - 56:22:  thought she was going to discover some new principles about protein biochemistry. But I
• 56:22 - 56:28:  told her, you know, don't be so depressed. There's still something very exciting, is even though they
• 56:28 - 56:33:  may bind the same RNAs, they have very different functions. And it's very divergent what they do.
• 56:33 - 56:39:  And in fact, if you have one m6A and it could bind to all three, maybe the DF protein that's
• 56:39 - 56:45:  most abundant in the cell might matter. Now, of course, most cells have YTHDF2 as the most abundant
• 56:46 - 56:52:  reader. But nonetheless, there could still be something very fascinating from a protein
• 56:52 - 56:57:  perspective, like how these proteins can mediate different functions. So the translation is
• 56:57 - 57:03:  something that she's very interested in. So she did polysome analysis just to examine the idea
• 57:03 - 57:07:  that these are translational regulators. And translation regulators are famous for being
• 57:07 - 57:13:  enriched in the polysome fractions. PABP1, all these other things that bind to three prime UTRs,
• 57:13 - 57:19:  as was proposed for YTHDF1, are enriched in the polysome fraction. But when she looked at it,
• 57:19 - 57:25:  in fact, she did not find that at all. I mean, there's trace levels, but the vast majority of
• 57:25 - 57:30:  YTHDF1 was associated with ribonucleoproteins and not with the ribosomes. And this was just one
• 57:30 - 57:35:  of many approaches that she used. She just was not able to find it enriched in ribosomes.
• 57:35 - 57:43:  So she then said, OK, if it's not in polysomes, maybe I should reexamine the data that said that
• 57:43 - 57:50:  it regulates translation at all. So translational regulation was established by ribosome profiling,
• 57:50 - 57:54:  where knockdowns were done. And then every transcript was analyzed to see how much less
• 57:54 - 58:00:  translated it was, especially based on the number of m6A sites. And when she looked at
• 58:00 - 58:05:  the data sets, it was very confusing, because there were two replicates. One was clearly a
• 58:05 - 58:10:  knockdown. There was no YTHDF1 mRNA, no ribosome footprints in YTHDF1 mRNA. But the other one was
• 58:10 - 58:15:  actually massively enriched in YTHDF1 ribosome footprint. So it looked like it was actually
• 58:15 - 58:21:  an overexpression rather than a knockdown. So she said, I need to start over. I need to just get
• 58:21 - 58:26:  data sets where I know everything was knockdown. So she knocked down YTHDF1. And what you're seeing
• 58:26 - 58:33:  is the full drop in translation for every mRNA. And we were binning RNAs based on whether they
• 58:33 - 58:38:  have 0 m6A, 1 m6A, 2 to 4, or 5 or more. So every cumulative distribution plot is hundreds or
• 58:38 - 58:44:  thousands of RNA transcripts. And you can see the RNAs are not getting reduced translation. And it
• 58:44 - 58:49:  certainly does not depend on m6A. And certainly, when we knocked down others, like YTHDF2 or YTHDF3,
• 58:49 - 58:54:  we didn't see any effect. In YTHDF3, we saw even a subtle, but this is statistical noise almost
• 58:54 - 58:59:  definitely, a subtle increase in translation, not decrease in translation. So this was not consistent
• 58:59 - 59:02:  with the idea that these proteins promote translation. So then she said, maybe they're
• 59:02 - 59:08:  redundant, and we need to knock out all three of them. But again, when she knocked out all three,
• 59:08 - 59:13:  nothing happened. And so the question is, OK, what do these proteins do? What does YTHDF1 do if it
• 59:13 - 59:17:  doesn't regulate translation? And really, the best hint of what a protein does is to look at
• 59:17 - 59:23:  its binding partners. And so she looked at proteomic data. And this is a BioID study of YTHDF1,
• 59:23 - 59:29:  YTHDF2, and YTHDF3. And what she's looking for is what proteins are uniquely bound to YTHDF1 and
• 59:29 - 59:34:  not bound to YTHDF2, what proteins are bound to YTHDF3 and not bound to YTHDF1, this sort of thing.
• 59:34 - 59:40:  So these are pairwise comparisons of YTHDF1 and YTHDF2. This is a complicated plot. But basically,
• 59:40 - 59:46:  the colored circles represent highly statistically significant interactors. And in orange,
• 59:46 - 59:50:  if it's found in both, when it's green, it's sometimes found in both, but just slightly
• 59:50 - 59:58:  below the 95% confidence interval. If it's gray, it has a low p-value and a non-confident
• 59:58 - 60:00:  interactor. So these are the interactors.
• 60:00 - 60:05:  It's based on proteomics. And it was a perfectly linear correlation for each of these proteins.
• 60:05 - 60:10:  And the top interactors for every one of them were the RNA degradation machinery. So CNOT,
• 60:10 - 60:17:  the PAN proteins, LSMs, all of these degradation proteins were amongst the top enriched proteins
• 60:17 - 60:25:  for all of them, not just YTHDF2, which we had expected, but all of them. And also,
• 60:25 - 60:32:  a lot of these putative m6A binding proteins, they're actually binders of the YTHDF2 proteins,
• 60:32 - 60:39:  FMRP, IGF2BPs. They also come out as binders. I can't say that they're direct, but they're
• 60:39 - 60:45:  physically associated through a direct or indirect interaction with the YTHDFs. So if all these proteins
• 60:45 - 60:49:  are finding the degradation machinery, maybe they have something to do with degradation.
• 60:50 - 60:56:  So Sarah reanalyzed the localization with P-bodies. And this is YTHDF1 localized with EDC4,
• 60:57 - 61:02:  YTHDF2 localized with EDC4, which is a P-body marker, every time she got co-localization.
• 61:02 - 61:07:  And I should say multiple independent studies where people have done P-body proteomics
• 61:07 - 61:11:  have shown the same thing, that all of these proteins are found in P-bodies.
• 61:11 - 61:16:  There was a paper by Xiaowei Zhuang where she did some immunofluorescence of the YTH proteins.
• 61:16 - 61:20:  This was in the supplement. She had three P-bodies, didn't quantify it, but commented that
• 61:20 - 61:25:  YTHDF2 is in the P-bodies, but not YTHDF1. It was not a quantified result. But the problem is YTHDF1 is
• 61:25 - 61:31:  lowly expressed. And so it's often hard to see it unless you normalize the co-localization for
• 61:31 - 61:36:  expression. When you normalize it for expression, then it looks like it is co-localized. So the
• 61:36 - 61:40:  question then becomes, okay, if they're all associated with P-bodies, all associated with
• 61:40 - 61:45:  degradation, maybe the function is actually to degrade mRNAs. And so what she did here,
• 61:45 - 61:52:  she knocked out YTHDF1, and then she looked to see if the RNAs are now increased in expression.
• 61:52 - 61:57:  But in fact, they barely changed. YTHDF1 did not. So the theory didn't seem to work.
• 61:58 - 62:03:  She knocked out YTHDF2, and as had been published, mRNAs actually become more stable, slightly,
• 62:03 - 62:09:  when you knock out YTHDF2. You knock out YTHDF3, nothing really happens. But then when
• 62:09 - 62:15:  she knocked out all three of them, big effect. Then she got this massive increase in RNA stability.
• 62:15 - 62:19:  So what she realized is that these proteins can compensate for each other. Now they aren't
• 62:19 - 62:24:  perfect at compensating because YTHDF2 is so highly expressed that the other ones can't really
• 62:24 - 62:30:  compensate when you lose YTHDF2. But when you lose YTHDF1 and YTHDF3, the others can compensate.
• 62:30 - 62:35:  And when you knock down all three, then it really matters. So it's important to realize these are not
• 62:35 - 62:39:  equally abundant proteins, so you can't simply compare them. Now we looked in leukemia cells where
• 62:39 - 62:44:  Camille Krantz had been trying to figure out which reader mediates the effects of m6A on
• 62:44 - 62:48:  leukemia. And as you may know, when you knock down m6A from leukemia cells, they start differentiating.
• 62:48 - 62:52:  But then when you knock out each individual reader, you don't see the differentiation
• 62:52 - 62:58:  marker, which is CD14. And so what we found is if we knock down all three of them, this is in red,
• 62:59 - 63:04:  then you start seeing this differentiation marker. And you also see specific transcripts that are
• 63:04 - 63:10:  thought to be inhibiting a differentiation, I'm sorry, promoting differentiation, go up when you,
• 63:10 - 63:13:  not when you knock them down individually, but when you knock out all three of them.
• 63:13 - 63:18:  And so what we realized is that there were serious problems with the CLIP datasets. Some of them have
• 63:18 - 63:25:  been retracted or whatever, but the influence on the field has certainly been very pronounced.
• 63:25 - 63:31:  And the ribosome profiling datasets, not reproducible, and I'm not even sure if both
• 63:31 - 63:36:  of them were knockdowns. And so, because they don't, there's massive amounts of YTHDF1 in one
• 63:36 - 63:42:  of those datasets. So what we also realized is that all m6A sites are equivalent. And it looks
• 63:42 - 63:46:  like the DF proteins redundantly mediate degradation with the most important thing being
• 63:46 - 63:50:  the number of sites. So this resolves a lot of the questions because you would wonder if you had one
• 63:50 - 63:55:  m6A site and you have so many different readers, how do you get any molecular sense out of what's
• 63:55 - 64:00:  going on? The answer is because they don't do different things. They actually do very, very
• 64:00 - 64:04:  similar things. Now, they could regulate translation because so many of the m6A proteins are
• 64:04 - 64:09:  translational regulators. And so you can indirectly affect translation. And I should also mention,
• 64:09 - 64:14:  when you knock down these proteins, you induce eIF2-alpha phosphorylation in some cell lines.
• 64:14 - 64:19:  And this has been reported. And so you can get global repression of translation just through
• 64:19 - 64:24:  the stress response in certain cell types. Now, I'm not saying that the DF proteins are identical.
• 64:24 - 64:28:  They could have subtle differences, maybe differences in phase separation or fiber
• 64:28 - 64:34:  formation. They have very similar phosphorylation sites, slightly different. But the key thing is,
• 64:34 - 64:39:  I think if there are going to be differences, and my results are only valid in the cell lines
• 64:39 - 64:45:  we tested, if there are different effects, it's going to be very important to figure out a mechanism
• 64:45 - 64:50:  because the proteins are so identical. People will really have to show that it's this one serine to
• 64:50 - 64:55:  threonine substitution or isoleucine or whatever to leucine that's really causing a dramatic shift
• 64:55 - 64:59:  in the binding. And we're going to really need some mechanism. Now, I should mention, just
• 64:59 - 65:04:  recently, Jacob Hanna published perhaps a much better paper than ours where he did this in
• 65:04 - 65:09:  embryonic stem cells. And what he showed is in embryonic stem cells, any one of the DF proteins,
• 65:09 - 65:13:  you only need one of them, will allow the embryonic stem cells to differentiate. But in
• 65:13 - 65:20:  gametes and sperm cells, there's no YTHDF1 and YTHDF3 or barely any YTHDF1 and YTHDF3. So if YTHDF2
• 65:20 - 65:25:  will cause a big phenotype in those tissues. And basically, the argument was you'll get
• 65:25 - 65:32:  phenotypes depending on which tissue has the inability to compensate based on the other YTHDF proteins.
• 65:32 - 65:36:  So really beautiful mouse genetics. I encourage you to look at that. It's a very different perspective.
• 65:36 - 65:40: I want to thank the people who did this work. And I can take any questions.
• 65:40 - 65:44: Great. Thank you, Sami.
• 65:48 - 65:49: Questions.
• 65:54 - 66:05: I've got a question from Thomas Chelminski. Did you look at the stability of P-bodies upon triple knockdown? Yes, we did. So I think this is an important question from Tomas.
• 66:14 – 67:07: So it's possible that the P-bodies are dependent on the YTHDF proteins, and they're certainly found in the DF proteins. When we knock out either metal 1 or all three YTHDF proteins, we don't see any clear change in the P-bodies. Now, I should say you do other things like puromycin and cycloheximide. P-bodies really get very morphologically different. But the P-bodies do not change. We also don't see changes in stress granules either. So I think the P-bodies are composed of many things. The YTHDF proteins are just one of them. Thank you. Another question from an anonymous attendee, which says, have you used sequencing platforms based on direct RNA sequencing, such as Oxford Nanopore for m6A? What do you think about that approach?
• 67:07 - 67:10: Well, so I should say I think there's some leaders in this field, and they're not me.
• 67:12 - 67:20: And they include people like Ava Novoa, and I forget the very last name of her. I think you might know her last name. I think, Tony, you mentioned to me that you have some interest in Nanopore. I think other people have an interest in Nanopore. It's complicated. So basically, the idea with Nanopore is that when the Nanopore encounters the m6A, it gives a possibility of a sequencing error. And that sequencing error or the enrichment of sequencing errors at specific adenosines is really the mark. We have not found, when we do it, that it's super easy to see it. But Ava and other people have been able to get this effect, to be able to see it. But I think that's the future. I think the future is Nanopore. I think the pores are getting better so that the error rate is even higher. And now, regarding the m6A levels, this will be on bioRxiv very shortly and hopefully published soon, both the SCARPIT method and the m6A method. So you should be able to see that probably within the month of November. I think, in the interest of time, I think we need to move on, I'm afraid, Sami. Okay. Now, we're going to move on to a secondary talk, which is from Richard Gregory from Harvard Medical School in Boston. And the title of his talk is Metal 1 Mediated m7G tRNA Modification Drives Oncogenic Transformation. Richard.
• 68:51 - 69:23: Okay. So hopefully you can see those slides. Thanks, Tony, for the introduction and also for the opportunity to be part of this meeting. It's been great so far, and I'm really privileged to be a part of this. What I'd like to do today is present an unpublished work where we've elucidated an oncogenic function of the methyltransferase Metal 1 and elucidated the downstream molecular and cellular mechanisms for this oncogenic transformation.
• 69:23 – 70:13: So, just to put this in context, one of the things that I find most appealing about this field is that unlike DNA methylation or modification, which is restricted to one or two distinct types of modification, when you consider RNA, there's over 150 known RNA modifications that occur on each of the four bases. Some of these modifications are quite simple, comprising just a simple methyl mark. Others are much more elaborate and require a cascade of different enzymes to deposit these marks. The function of many of these still remains to be determined, and certainly there's a lot to be learned in terms of how these modifications impact different diseases, and our focus has really been on understanding the role in cancer.
• 70:13 - 70:32: So, some of these modifications can affect Watson-Crick base pairing. Others, like the one I'm going to talk about today, N7-methylguanosine or m7G, affects other properties of the nucleotide. In this case, it affects Hoogsteen base pairing and the ability of RNAs to form quadruplexes.
• 70:32 – 71:15: You know, you can see many of these modifications listed here. These are deposited, as I alluded to, by highly specific enzymes, and it's emerging that these modifications are important for regulated gene expression. And when we look at the epitranscriptome, of course, it would focus on mRNA, shown here, but I would argue that apart from m6A, many of these other marks are still…the true impact of these marks in controlling gene expression remains to be determined. And you can also notice that there are relatively few known modifications of messenger RNAs. In contrast, I think we need to consider RNAs more generally and certainly consider tRNAs.
• 71:15 - 71:19: And so, tRNAs are the most heavily modified RNA in the cell, subjected to potentially 50 or more distinct modifications, and these occur at characteristic sequence or structure motifs within the tRNA, and some of those are shown here.
• 71:30 - 71:48: It's emerging that these modifications are dynamic and can offer an additional layer of gene regulation control, and as we're starting to appreciate, dysregulation of these pathways contributes to human disease. And my lab is focused primarily in the context of cancer.
• 71:48 - 72:38: So, what we're talking about today is m7G that's catalyzed by a heterodimeric complex. This is the yeast complex shown here. Metal 1 is the catalytic subunit. It has an essential cofactor called WDR4. And together, this heterodimeric protein complex modifies guanosine at position G46 in a variable loop of tRNAs, in a subset of tRNAs. And this modification is known from work in various model organisms as being required for tRNA stability. And so, tRNAs that contain m7G at position G46 are able to form this base triple with nucleotides in the D-loop of the tRNA, which helps the folding of the tertiary structure of the tRNA. And so, these tRNAs are more stable.
• 72:38 - 73:15: And without modification, these tRNAs are degraded by a so-called rapid tRNA decay pathway. So, it's known, as I mentioned, from working yeast and other model organisms that m7G stabilizes the tertiary structure of tRNAs. And we developed a methyl RIP-seq using an antibody to m7G to pull down all the modified RNAs and then perform RNA sequencing, as well as a chemical method that we refer to as TRAC-seq, tRNA reduction and cleavage, which is a chemical-based method that allows mapping of m7G and nucleotide resolution throughout the transcriptome.
• 73:15 - 73:33: And so, we previously developed these methods, and we applied it to the mammalian transcriptome. In this context, we looked at mammalian mouse embryonic stem cells and were able to define the subset of tRNAs that contain m7G for the first time at nucleotide resolution.
• 73:33 - 73:37: And we were also able to knock out Metal 1. And in contrast to work that had been done in yeast, where this modification is non-essential, what we found in mouse embryonic stem cells is that there's a clear role for Metal 1 in ES cell biology. When you knock out Metal 1, ES cells have a defect in self-renewal and proliferation, and they also have defects in lineage choice and differentiation. So, this work was recently published, and today what I'm going to focus on is this question of what is the role of Metal 1 in cancer.
• 74:03 – 74:48: One of the reasons we're particularly interested in this is that if you look at the Cancer Genome Atlas, the so-called TCGA, you can see that in certain cancer types with high frequency, the Metal 1 gene, the locus that encodes Metal 1 is amplified. And this is particularly true for sarcomas and glioblastomas, but also a wide variety of other different cancer types. Again, looking at TCGA data, if you look at the mRNA levels from RNA-seq, this is a panel of different tumor types in TCGA, you see that there are two in particular cancer types that have especially high levels of Metal 1. And this, again, corresponds to the sarcomas as well as the glioblastomas.
• 74:48 - 75:45: So, this is looking at that Metal 1 mRNA level. We've done immunohistochemistry, looking at tumor sections, tumor microarrays, and this is the glioblastoma tumor stained here with the Metal 1 antibody. You see there's very strong staining in the tumor. There's barely any detectable staining in the normal brain tissue. And as you can see, that's quantified here. So, at the protein level, Metal 1 expression is highly upregulated in patients, in tumors, glioblastoma samples, compared to normal brain tissue. Furthermore, when you look at the patients with high levels of Metal 1 expression compared to those with low levels, you see that Metal 1 is a poor prognosis factor. So, patients with high levels of Metal 1, shown here, have a much worse prognosis, both glioblastoma as well as sarcoma, and several other cancer types.
• 75:45 - 76:30: So, the reason that Metal 1, at least in some cases, is overexpressed is that, as I mentioned, the locus itself is amplified. And so, this is just looking in more detail at that. This is the sarcoma sample, soft tissue sarcomas, which includes a variety of different tumor types in TCGA data set. Each column here is a different chromosome, and each row here is a different tumor sample. And shown here in red is DNA copy number gain, and shown in blue is DNA copy number loss. And when you just sort these tumors based on relative Metal 1 expression, shown here, what you can notice is that there's a correlation with a region of chromosome 12 that is amplified.
• 76:30 - 76:57: So, this is the region shown here. And you can also, what you can also notice is that by sorting these tumor samples, these sarcomas, based on Metal 1 expression, you see that there's this blue group that particularly correlates with Metal 1 high levels of expression plus amplification of the locus. And it turns out that this corresponds to, in particular, liposarcomas. So, about 70 percent of all liposarcomas have amplification of the Metal 1 locus.
• 76:57 - 77:39: Similarly, in glioblastoma, about 16 percent of all glioblastoma samples have amplification of the Metal 1 locus. And this, again, corresponds to this subset of tumors with very high levels of Metal 1 expression. So, the question is, could Metal 1 be an oncogenic driver, or is it simply just co-amplified with neighboring known oncogenes? So, Metal 1 is in close proximity on chromosome 12 to known oncogenes, including CDK4, MDM2, and a variety of other known oncogenes. So, we wanted to ask the question, could Metal 1 itself be an oncogenic driver and promote tumorigenesis?
• 77:39 - 78:12: : So, to test this, we performed a classical transformation assay where we took non-transformed mouse embryonic fibroblasts (MEFs), and we overexpressed Metal 1 together with its cofactor, WDR4. We can express the wild-type version of the complex, or we could also express the catalytic mutant of the Metal 1 complex. And what you can see here is that, indeed, when we overexpress Metal 1, we see a very strong effect on colony formation in soft agar. So, this is with the wild-type complex.
• 78:12 - 78:50: You see the very large colonies form, and this is not true with the catalytic mutant, so you require catalytic activity for this oncogenic transformation phenotype. That's just quantified here, colony number shown there. And then we performed a wide variety of other cancer cell biological assays, and in every case, when we overexpress Metal 1 with its cofactor, the wild-type complex, we see a positive output. So, basically, in every assay that we measure, Metal 1 promotes oncogenic transformation and, therefore, behaves as a bona fide oncogene.
• 78:50 - 79:32: We see primarily the effects on MEFs, on their growth, is through proliferation. This is in, I don't have time to show the data, but we see an acceleration of the cell cycle when we overexpress Metal 1. We don't see any effect on apoptosis or senescence, so this is really a pathway that's driving accelerated cell cycle to promote tumorigenesis. We take these MEFs that have been transduced with the Metal 1-expressing virus, we put them into a xenograft mouse model subcutaneously, and you see that, again, these cells that are transduced with a wild-type complex are capable of forming tumors in these mouse xenografts.
• 79:34 - 80:55: So, we wanted to start to explore the underlying molecular mechanism for this, and so we applied the method I previously mentioned, the TRAC-seq method, to find out which tRNAs are methylated and what is the effect of Metal 1 on the subset of tRNAs that are methylated. And so, just a little bit of detail here, so this method requires we isolate the small RNAs, less than 200 nucleotides, we treat these RNAs with recombinant AlkB, which removes most methyl marks but preserves m7G, so this then allows the more efficient subsequent sequencing of these tRNAs that would otherwise be inhibited by the multiple methyl marks on the tRNA. We treat the RNAs with sodium borohydride, which reacts with the modified m7G, and then we subsequently treat with aniline, which cleaves at a basic site, and then we can prepare cDNA libraries and sequence them using Illumina high-throughput sequencing, and this method works very reliably and precisely, and you can see just two examples here. So, you see a drop-off of reads. This top track is just looking at sequences of the untreated sample, the RNA. The bottom track here shows the chemically treated sample, and you can see there's a fall-off of reads that occurs exactly at the modified guanosine position here.
• 80:55 – 81:22: This works for this tRNA, another example here, and we're able to apply this method to identify the subset of tRNAs, about 26 tRNAs, that are modified in mouse embryonic fibroblasts. This modification occurs at this guanosine, in this characteristic sequence motif, and it always occurs in this variable loop of the subset of tRNAs that contain the mark.
• 81:24 - 82:00: When we apply TRAC-seq to the control MEFs, I just showed the data, we did the same thing for MEFs overexpressing wild-type Metal 1 or catalytic mutants, and what we observed is it's the same subset of tRNAs that are modified, regardless of the condition we tested. So, it's not that when we overexpress Metal 1, we're getting new substrates for the enzyme complex, but we do see an increase in the proportion of m7G to G. So, we are getting increased levels of m7G, but it's not occurring at new positions within the tRNA.
• 82:00 – 82:42: And so, what we interpret this to mean is that, oh, furthermore, we see that when we overexpress Metal 1, as we would expect, we see increased stability and abundance of the m7G-modified tRNAs. And so, taken together, what we interpret this to mean is that, basically, for any given tRNA, the stoichiometry of modification may not be 100 percent, and by having extra Metal 1 expression, that modification becomes more prevalent for that particular tRNA. And that's shown and reflected here. When we look at the change in cleavage score, which is reflecting m7G modification, there's a positive correlation with the fold change in tRNA abundance.
• 82:42 – 83:22: And this is especially true for these two tRNAs, and pay attention to arginine TCT, because that turns out to be one of the most important tRNAs in this context. So, just to confirm, when we overexpress Metal 1, this is a northern blot, we see increased abundance of this tRNA, shown here, when we overexpress Metal 1. And, furthermore, when we use biotinylated oligos to pull down this particular tRNA, and then use mass spectrometry to look at the relative modification frequency, we see that there's indeed an increase in m7G occurring on this particular tRNA.
• 83:24 – 84:11: So, the next question is, we're getting an increased abundance of tRNAs, so what is the effect on translation? And so, we performed ribosome profiling, as Sami just mentioned. The approach looks for the abundance of ribosome-protected fragments, and it allows us to measure translation quantitatively across the transcriptome. And what we found is that when we overexpress Metal 1, we see the translation efficiency of several hundred mRNAs goes either up or down. But we next ask the question, of these mRNAs with altered translation efficiency, are there any particular codons that are enriched in the subset of mRNAs that become more efficiently translated or less efficiently translated?
• 84:12 - 84:17: And what you can appreciate from this is that there are certain codons that are overrepresented in the mRNAs that become more efficiently translated. And again, I'll just point your attention to this one, arginine AGA, which is, of course, decoded by the m7G-modified arginine TCT tRNA that I just showed you is one of the most upregulated and abundant when you overexpress Metal 1. Furthermore, when we do gene ontology analysis to see which pathways are preferentially affected when we overexpress Metal 1, and what we find is that those mRNAs are very highly enriched in controlling cell cycle. And furthermore, when we look again at codon enrichment of this subset, so these are genes whose translation efficiency is changing, and they belong to this gene ontology of the cell cycle, you can see that there's only, strikingly, there's only one codon that is enriched, particularly enriched in this subset of mRNAs, and this is arginine AGA.
• 85:15 - 86:05: So this got us thinking that perhaps this particular tRNA may be especially important in cancer, and so we wanted to look in patient samples to see whether this tRNA is dysregulated compared to normal tissue. And so we again turned to the Cancer Genome Atlas, and as you're probably aware, the TCGA has mRNA expression profiling for normal versus tumor, that's the poly-A plus RNA-seq, and for many tumors there's also corresponding small RNA-seq. So that has been analyzed for the microRNA expression in normal versus tumors, but we were able to gain access to the raw sequencing data from the small RNA-seq and extract tRNA sequence reads bioinformatically from those datasets, and then we simply plotted tRNAs, the change in expression of these tRNAs in normal, in tumor samples compared to normal.
• 86:05 - 87:12: This is the list of cancer types we looked at, and what you can see right at the top of this list, the tRNA that's most highly up-regulated in tumors compared to normal is arginine TCT. This up-regulation of arginine TCT is, at least in some cases, mediated, we think, through increased levels of Metal 1, so there's a positive correlation shown here between levels of Metal 1 mRNA by RNA-seq and levels of this arginine TCT tRNA by small RNA-seq, and most strikingly, just by looking at expression of this single tRNA in normal versus tumors, you can use this as a prognostic marker. So basically, patients, this is shown here for sarcoma, this is true for many other cancer types, patients with high levels of arginine TCT have a poorer prognosis than patients with a lower level of this tRNA.
• 87:12 - 87:40: So we wanted to test the hypothesis that perhaps this tRNA may be mediating many of the downstream effects of Metal 1 oncogenic transformation. So we wanted to test this by overexpressing this particular tRNA, and we did this by cloning the genomic sequence into a promoterless plasmid, so this overexpression of arginine TCT is driven by the endogenous promoter from this tRNA.
• 87:40 – 88:21: You see that we can achieve overexpression of the tRNA. We also tested whether or not this tRNA is charged, whether it contains an amino acid, and indeed it does. So this is the overexpressed tRNA. When we treat with carbamoyl phosphate, this breaks the peptide bond, sorry, the bond of tRNA, and you can see that the tRNA migrates slower when the amino acid is released. So this is a charged tRNA. It's also a functional tRNA. So we tested using this reporter system whether or not overexpressing Metal 1 can preferentially boost translation of the reporter that contains AGA codons compared to one that doesn't.
• 88:21 - 89:06: So for this purpose, we used this dual luciferase reporter. So firefly luciferase has 20 arginine codons, none of which are this AGA type. Ranilla, in the normal vector, two of the 16 arginines are encoded by AGA codons. So what we did is we converted the other 14 arginine codons into the AGA type shown here, and what we saw is when we overexpress the tRNA, it preferentially enhances the translation of this reporter shown here compared to this one. So we know that the tRNA is somewhat limiting in the maps, and when we overexpress it, we can get enhanced translation of the reporter.
• 89:06 - 89:47: We next tested whether this can contribute to cellular transformation. The more you can see here is that when we overexpress Metal 1, indeed, looking at colony formation in soft agar, we see an increased number of colonies that's mediated by overexpression of the tRNA. Perhaps even more compelling, we collaborated with Costas Slepian in Cambridge, UK, and in this AML model, human AML model, he showed that when you overexpress Metal 1, it makes these, in these xenograft experiments, makes these leukemia cells more aggressive, as shown by bioluminescence imaging here.
• 89:47 - 90:17: This is, again, dependent on the catalytic activity. When he overexpressed the catalytic mutant, there was no effect, but most strikingly, he could phenocopy the effects of Metal 1 overexpression just by overexpressing this arginine TCT tRNA. That's shown here, and it's also reflected in the overall survival of these mice. So mice that are transduced with a virus to express either Metal 1 or the tRNA have a poor survival compared to the control cells.
• 90:17 - 90:49: So finally, we just wanted to explore what effect this might have on the proteome, and so we use proteomics to look at the effect on protein abundance when we overexpress the wild-type Metal 1, or when we overexpress the tRNA. What you can notice here is that there's a very nice positive correlation and correspondence between the proteins that have become upregulated when we overexpress Metal 1, compared to those that become upregulated when we overexpress tRNA.
• 90:49 – 91:28: Furthermore, these proteins are highly enriched in the AGA type of arginine, or AGA-encoded arginine, and when we do gene ontology on this subset of proteins, again, cell cycle is the top gene ontology. So this, you know, potentially very nicely explains the cellular phenotype at the molecular level. And then we are also able to look at the cell cycle of the protein, at the molecular level. And then we are also able to confirm the overexpression of some of these targets at the protein level by Western blot, and then showed that the mRNA of these genes doesn't change when we perform qRT-PCR.
• 91:28 - 92:01: So this leads me to the model. What I've shown so far is that in many cancer types, you have elevated levels of the Metal 1 complex. This then leads to increased modification of certain tRNAs, in particular, arginine TCT. This then leads to a remodeling of the translatable pool of mRNAs such that those that are enriched in arginine AGA become more efficiently translated, perhaps at the expense of other mRNAs that are depleted in this codon.
• 92:01 – 92:43: And this is at least in part what's responsible for driving this oncogenesis. So in the spirit of this meeting, you know, we're talking about from bench to clinic, this potentially raises the possibility that Metal 1 is an appealing target for cancer therapeutics. And so we've tested this in a variety of different contexts. But I'm just showing one example here. This is a knockdown in a glioblastoma cell line that's amplified for Metal 1. And you can see that when we knock down Metal 1 with short hairpin RNA and perform xenograft experiments, you can completely suppress tumorigenesis by depleting these glioblastoma cells of Metal 1.
• 92:43 – 93:06: And again, we performed tRNA sequencing. And what we saw is the expected effect on the tRNAs. So these tRNAs become less abundant and destabilized, as shown here by tRNA-seq. It's also confirmed here by northern blot. So this m7G modified tRNAs become less abundant when you knock down Metal 1. The non-modified tRNAs are unaffected by Metal 1 depletion.
• 93:07 - 93:34: And then we see a global change in translation, perhaps not surprisingly, because one of the tRNAs that we know is the initiator IMAT tRNA, which is, of course, required for translation of pretty much every protein. And you can see that by metabolic labeling, we see decreased global translation. And then this is also reflected in general down regulation at the proteome level, shown here by SILAC-based proteomics.
• 93:34 - 94:04: So to summarize, what I've shown today is we've identified the oncogenic role of Metal 1, increased Metal 1 expression and amplification occurs in a large variety of different tumor types, in particular in glioblastoma and sarcomas. Metal 1 is an oncogene, is capable of promoting transformation and tumorigenesis. And we would argue that this, the effect on depleting Metal 1 highlights this methyltransferase as a future drug target for cancer therapeutics.
• 94:04 – 94:38: So just to acknowledge the people that did the work, Esteban, a really terrific postdoc in my lab, did all of the pretty much all of the bench work I showed today. A really talented bioinformatician, Qi, shown here, he did pretty much all of the analysis that I showed you. And some of this work has been done in collaboration. Yangxi helped us with the mass spec, RNA mass spec. Frank Slack helped us with the xenograft experiments, as well as, as I mentioned, Kostas has been a great collaborator on some of this work. I have a couple of disclosures here.
• 94:38 - 94:50: I’m Co-founder, SAB member of 287 Therapeutics, as well as Epitoire. And we received some sponsored research from Sanofi and Astellas. So thank you very much for your attention. And I would like some questions.
• 94:50 - 94:58: Great. Thank you, Richard. Great talk. I've got some questions here from the audience. Let's try some of those. How can Metal 1 WDR4 alone be able to transform wild-type MEFs when all bona fide oncogenes, RAS, MYC, PME1A, on their own induce INC4A, ARF, LOCUS, and downstream PVT3RB to cause senescence or apoptosis?
• 95:21 - 95:53: Yes, thank you for that question. And I would like this opportunity to clarify. So when I say wild-type MEFs, these are MEFs that have been immortalized. So you're right that if you were just to put oncogenes into truly wild-type MEFs, it would induce. But, you know, this is still a valid assay of transformation. So basically, the short answer is these are not completely wild-type MEFs. They're immortalized but not transformed. And so we can transform them by putting in Metal 1 WDR4. Great. Thank you, Richard.
• 95:53 - 96:01: The next question is from Nibedita Lenka. Was the oncogenic effect of MEFs transient or permanent?
• 96:01 - 96:25: So we haven't done an experiment that you might be alluding to where we initiate transformation and then turn off Metal 1 expression to see whether these cells are truly addicted to persistent Metal 1 expression. I believe that our collaborator Costas has done some experiments to address that. But we haven't looked at that ourselves yet.
• 96:25 - 96:49: We haven't looked at that ourselves yet. Thank you. Another question. Is the correlation with Metal 1 ARGTCT expression present upon knockdown of Metal 1 WDR4 in sarcomas, gliomas that have Metal 1 amplifications, amplicons?
• 96:49 - 96:54: I'm not sure I totally get the question. But so is that in the patient sample or in the cells?
• 96:54 - 96:59:  If we knock down METTL1 in the cancer cells, we do see decreased expression of this tRNA.
• 97:03 - 97:06:  And we also see a correlation in patient samples. But of course,
• 97:06 - 97:08:  we wouldn't be able to do any manipulation in that context.
• 97:11 - 97:14:  I think that gets to the question. All right.
• 97:15 - 97:17:  Exactly. We can't. We can't.
• 97:17 - 97:23:  Did I miss the point there, Tony? That's right. Yeah. I think you did answer it.
• 97:25 - 97:35:  So this is from Matthew Levine. Do AGA ARG on molecules need to be
• 97:35 - 97:39:  methylated by METTL1 to prevent tumorigenesis?
• 97:39 - 97:53:  Do AGA ARG on molecules that need... He's saying does the overexpression
• 97:53 - 97:57:  experiments require those tRNAs to also get methylated?
• 98:00 - 98:03:  You mean when I overexpress the tRNA, does it become methylated?
• 98:03 - 98:04:  Yes.
• 98:10 - 98:14:  I'm not sure. We haven't really looked at that. But we would anticipate that it should, right?
• 98:14 - 98:17:  But yeah, we haven't actually looked at that.
• 98:17 - 98:22:  Is he saying how can overexpression rescue if it doesn't also get methylated?
• 98:23 - 98:27:  I mean, that's right. So we expect it does. But we don't have direct demonstration that
• 98:27 - 98:31:  it is methylated. But there's no reason to think that it wouldn't be. But we should look at that.
• 98:33 - 98:38:  Great. I think we need to move on, I'm afraid, because of time. Thank you, Richard.
• 98:38 - 98:45:  So now we move on to the final privilege talk from Xuan He from the University of Chicago.
• 98:46 - 98:51:  And the title of his talk is tRNA methylation in gene expression regulation
• 98:51 - 98:53:  in cancer therapies. Xuan.
• 98:55 - 99:00:  All right. Thanks, Tony, and the Abcam team for this opportunity.
• 99:00 - 99:04:  So here's my outline. I'm going to give you two stories, one on transcriptional regulation,
• 99:04 - 99:10:  another on post-transcriptional regulation. And I believe I don't need to introduce m6A.
• 99:11 - 99:18:  There are plenty of introductions already. So we started with the discovery of the erasers,
• 99:18 - 99:24:  and we and others characterized the writers. And we and others pretty much characterized
• 99:24 - 99:31:  most of the readers, which provides sort of a pathway for understanding the consequence of m6A
• 99:31 - 99:38:  on messenger RNA regulation. So in the last, I would say, five, six years, most of the fields
• 99:38 - 99:45:  been focusing on this post-transcriptional regulation studies. And we started wondering
• 99:45 - 99:52:  about the potential regulation of transcription or m6A deposition to the nuclear non-coding RNAs,
• 99:52 - 99:54:  the potential consequence for chromatin regulation.
• 99:56 - 99:59:  So I'm just trying to quickly summarize. Here's the results we show.
• 100:00 - 100:04:  So the published early this year already, that if you knock out the methyltransferase,
• 100:04 - 100:09:  you clearly see more open chromatin and up-regulation of the global transcription. And this work
• 100:09 - 100:15:  was done by Dr. Jun Liu. And what she showed, essentially, is that the same methyltransferase
• 100:15 - 100:21:  complex that put on the methylation on the pre-messenger RNA can also mediate methylation
• 100:21 - 100:27:  of this, we call it the chromatin-associated regulatory RNAs. And the methylation serves
• 100:27 - 100:32:  as an active mark in most ES cells, in cancer cells is different, depending on the cell
• 100:32 - 100:40:  type. And these really help to decay this RNA and suppress the expression of YTHDC1,
• 100:40 - 100:48:  serve as a reader to decay a portion, not all of this, in particular repeats.
• 100:48 - 100:55:  So this is my first story about the demethylase. It's an interesting system. We reported the
• 100:55 - 101:01:  first RNA demethylase in 2011. This is RNA demethylase, the first one. Everybody agreed. And we studied the
• 101:01 - 101:09:  demethylase FTO, m6A. And there, there's lots of confusions. And a work done by Jeffrey
• 101:09 - 101:17:  Lab in 2017 and later on showed FTO is quite potent at mediating cap-m6Am methylation and
• 101:17 - 101:23:  concluded that m6A is not the substrate of FTO. I think we addressed with this paper
• 101:23 - 101:31:  showing that FTO can work on both m6A and m6Am. And here's another work published just online
• 101:31 - 101:39:  in the presence of NADPH. You can really boost m6A demethylation activity in vitro and instead
• 101:39 - 101:44:  of the cell by something like 8 to 10 folds. So this is quite potent in m6A demethylation.
• 101:45 - 101:49:  In cancer cells, it's pretty much established that the messenger m6A, which is a system we
• 101:49 - 101:59:  studied previously, is a substrate of FTO. In certain leukemia cells, 40% of total m6A can
• 101:59 - 102:04:  be demethylated by FTO in the cytoplasm. And there's lots of work on this.
• 102:05 - 102:12:  The question for the physiological substrate of FTO is still a mystery, right? The knockout of
• 102:12 - 102:18:  mice display quite severe phenotype or partial embryo lethality. Actually, most of the mice in
• 102:18 - 102:26:  our hands die before birth. And most of them die, you know, if they were born, they would die
• 102:26 - 102:31:  within a month, at least in our hands. So there's quite a significant portion of embryo lethal
• 102:31 - 102:37:  phenotype. And after they're born, there's a gross retaliation and a postnatal lethality,
• 102:39 - 102:44:  etc., etc. But if you look at m6Am methyltransferase, which we created
• 102:45 - 102:51:  in our collaborator's mouse model, the knockout mice are fine. You don't see much effect on
• 102:51 - 102:59:  viability, fertility. Not 14 knockout mice, again, you do not see much effect on the fertility
• 102:59 - 103:05:  and viability. So although m6Am could be demethylated by FTO, but it's definitely not
• 103:05 - 103:14:  these substrates to explain this physiological phenotype. Now, if you deep, deeper, this is
• 103:14 - 103:22:  the work done by Chengqi Yi at Peking University, in which he sequenced m6A and m6Am from mouse to
• 103:22 - 103:28:  human tissues. So if, and perform the correlation studies, if you have a high FTO, you would expect
• 103:29 - 103:36:  a low m6A. And that was not exactly the case. I did not really see sort of the correlation expected
• 103:36 - 103:41:  for messenger RNA in many of these mammogram tissues, suggesting the messenger RNA m6A
• 103:42 - 103:46:  in the primary tissue cases might not be the main substrate of FTO.
• 103:47 - 103:54:  So that brought us back to the methyltransferases METTL3 and METTL14. We knew that this thing has severe
• 103:54 - 104:01:  phenotype if you knock out of the mice died very early embryo stage. And now we show that the
• 104:01 - 104:06:  poly-A RNAs, the repeats, and the chromatin-associated regulatory RNAs could be the main
• 104:06 - 104:11:  substrates of methyltransferase. We wondered could demethylation occur to some of these
• 104:11 - 104:20:  substrates that tune gene expression. So a senior student, Dr. Jiangbo Wei just graduated,
• 104:20 - 104:27:  or going to graduate. So he, what he showed is that indeed, if you look at these chromatin
• 104:27 - 104:33:  associated RNAs, when you knock out FTO, you'll see a quite significant increase of m6A that
• 104:33 - 104:42:  occurs to both intergenic, but also pre-mRNA portions. If you perform sequencing, you realize
• 104:42 - 104:49:  that the main changes that tend to occur is the repeats. You gain m6A peaks mostly on repeats.
• 104:49 - 104:55:  When you gain these m6A peaks, you see reduced stability, again, confirming mouse ES cells, m6A
• 104:55 - 105:02:  serve, and that can mark destabilize these repeats and CARAs. Here's another analysis.
• 105:02 - 105:10:  All of the repeats are loose, and some of these other repeats tend to change as well. But LINE
• 105:10 - 105:16:  one seems to be a dominant one, but there's other repeats could serve as FTO substrates.
• 105:17 - 105:24:  So what this suggests is that FTO actually works on chromatin-associated RNAs, and a major group
• 105:24 - 105:31:  of the substrates are these repeats. To confirm this, we performed m6A profiling, and you can
• 105:31 - 105:38:  clearly see it at the five prime, towards five prime m6A size tend to be the substrates of
• 105:38 - 105:43:  FTO in this demethylation. And if you knock out, you'll see gain of m6A on this LINE,
• 105:43 - 105:49:  and you'll see reduced abundance in there. They co-localize, and if you put on FTO, you do
• 105:49 - 105:58:  see some binding of L1 RNA. Now, besides knockout, you can also use FTO inhibitors.
• 105:58 - 106:04:  When you apply FTO inhibitors, you see similar increase of m6A and a reduction of L1 abundance,
• 106:04 - 106:10:  confirming this is a demethylation activity-dependent, and not all repeats or L1s seem
• 106:10 - 106:15:  to be subjected to the young LINE one families tend to be regulated through this pathway.
• 106:16 - 106:23:  What's the biological consequence? Well, if you knock out FTO, you'll gain m6A, and that leads
• 106:23 - 106:32:  to decay of L1, and that turns to lead to close chromatin, all right? And here, if you use ASO
• 106:32 - 106:40:  to inhibit LINE one, you see reduced chromatin, and you can recover that with a FTO delivered
• 106:41 - 106:48:  by a Cas13 system to L1, you can partially rescue that chromatin state change.
• 106:49 - 106:55:  One experiment I like that John Boyd did was careful kinetic studies. So he applied FTO
• 106:55 - 107:02:  inhibitor, and what he observed is a gradual increase of m6A, and followed by the LINE one
• 107:02 - 107:08:  abundance change, all right? And then that followed with LINE one abundance change,
• 107:08 - 107:17:  the chromatin closed state, all right? Confirming the sort of sequence of this is probably the m6A
• 107:17 - 107:21:  change, it leads to abundance change of L1, and subsequent chromatin state change.
• 107:23 - 107:30:  So that's sort of the current model that FTO works on this m6A, and YTHDC1, again, YTHDC1
• 107:30 - 107:36:  has lots of interesting roles, even our L1, one role is to mediate decay of some of these
• 107:36 - 107:44:  m6A modified LINE ones, it really depends on where m6A is on L1, and DC1 seems to have
• 107:44 - 107:50:  different roles. What's a biological, sort of physiological consequence, right? If you monitor
• 107:50 - 107:55:  LINE one during mouse ES cell differentiation, you clearly see an increase upon differentiation,
• 107:55 - 108:03:  and with the FTO knockout, you accumulate m6A, and the L1 level cannot accumulate in mouse ES cells,
• 108:03 - 108:09:  and that leads to increased differentiation, and reduced proliferation, and you can
• 108:09 - 108:20:  partially recapitulate this by inhibiting the L1 RNA with ASO. You can also bring in a site-specific
• 108:20 - 108:27:  m6A FTO demethylation system, and here you have the Y-type, and then the knockout case,
• 108:28 - 108:33:  and now if you mediate, so in this knockout case, you accumulate m6A because there's a knockout of
• 108:33 - 108:38:  FTO, but if you bring in the systems to specifically work on L1, you can partially
• 108:39 - 108:46:  rescue this sort of phenotype, same is true for proliferation, but not an active FTO control,
• 108:47 - 108:53:  suggesting a large portion of this differentiation proliferation phenotype
• 108:54 - 109:03:  contributed through m6A on LINE one RNA. Previously, we looked at the correlation
• 109:03 - 109:10:  from messenger RNA m6A with FTO level, but if you look at L1 m6A and L1 abundance correlation with
• 109:11 - 109:16:  FTO, you actually observe a quite consistent correlation, the high
• 109:16 - 109:20:  FTO expression in certain tissues, you generally see low m6A level and high
• 109:21 - 109:29:  LINE one abundance, so this is consistent across mouse and human tissue that FTO seems
• 109:29 - 109:36:  to be the demethylase in the primary tissues of the L1 m6A, and that correlation is not so obvious
• 109:36 - 109:45:  for ALKBH5, but ALKBH5 also works on CAR RNAs, but looks like the main sort of demethylase for
• 109:45 - 109:52:  at least the LINE one is FTO. Now, if you look at the membrane tissues, the highest FTO expression
• 109:52 - 110:00:  is cerebellum, so we look at the cerebellum, and when we knock out the FTO, we see L1 m6A level
• 110:00 - 110:08:  increase in the cerebellum, and we also see reduced L1 RNA abundance consistent from sort
• 110:08 - 110:15:  of a tissue, knock out the tissue part. Lastly, to study the developmental phenotype
• 110:15 - 110:21:  with peaked oocyte, it's just a lot easier for us at this moment. If you knock out FTO,
• 110:21 - 110:30:  you see about 50% reduction of mature oocyte, and what you observe is a reduction of L1 level
• 110:30 - 110:36:  because it will accumulate m6A, right? The L1 RNA level is reduced by almost 50%,
• 110:37 - 110:46:  and that correlates with reduced chromatin openness. We think this explains partially
• 110:46 - 110:52:  the defect of oocyte development. Now, if you look at the other m6Am cases,
• 110:53 - 110:57:  METTL14, TCAF knockout, they do not show these phenotypes.
• 110:59 - 111:04:  I think that we can pretty much conclude that internal m6Am repeats RNA, such as L1,
• 111:04 - 111:12:  are the main substrates of primary tissues or cells, and we don't exclude the presence of other
• 111:12 - 111:18:  substrates, such as messenger RNA m6A in certain tissues or other repeats. In cancer cell lines,
• 111:18 - 111:26:  the m6A internal of messenger RNA and potentially repeats are the primary substrates of FTO,
• 111:27 - 111:33:  and we currently, although we confirmed that the cap-m6Am are the FTO substrates,
• 111:33 - 111:40:  but they so far have not, after three years of study, not show any notable phenotype,
• 111:40 - 111:43:  and there was some really interesting work from Harvard Medical School suggesting this may be
• 111:43 - 111:46:  involved in immune response that were meant to be elucidated.
• 111:48 - 111:52:  All right, so that's my first transcriptional story. I will get on to the second story that
• 111:52 - 111:57:  has translational potential, and that's the translation upregulation by YTHDF1, DF2,
• 111:57 - 112:04:  and when we tether DF1, we as a reporter, we clearly see upregulation of translation. Now,
• 112:04 - 112:10:  the DF1, DF2, DF3 knockout mice individually have very different phenotypes genetically. They're
• 112:10 - 112:16:  different. However, in certain cells, if you knock out a DF1, DF3. Sometimes you do
• 112:16 - 112:22:  not see translation differences, and I raise the question, do these guys really promote translation?
• 112:23 - 112:29:  Again, the knockout mice are fine. If you have a DF2 knockout mice, you can't rescue with DF1 or
• 112:29 - 112:35:  DF3 because they're very different pathways, right? Even they affect stability we showed
• 112:35 - 112:42:  previously, but through different pathways, so we're going to sort of publish this in a year or
• 112:42 - 112:49:  later. But specifically on DF1, the knockout mice are sort of a developmental find. DF3 mice,
• 112:49 - 112:54:  no problem. They're learning on the memory problems I showed previously. So if you tether
• 112:54 - 112:59:  a reporter in these neurons, you do not see difference, but when you give a stimulation,
• 112:59 - 113:04:  when you have a DF1, you see upregulation of translation. Again, when you look at nascent
• 113:04 - 113:09:  protein synthesis in these neurons with DF1, without DF1, not much difference, but when you
• 113:09 - 113:15:  add a stimulation with DF1, you'll see upregulation of translation. So we concluded that in certain
• 113:15 - 113:19:  cancer cells, this pathway is hijacked to constitutively promote translation, but in
• 113:19 - 113:25:  primary cells, tissues, this might be a stimulation-induced system. So three years ago,
• 113:25 - 113:32:  we have four different systems confirming this. Another system came out just this week
• 113:33 - 113:39:  from cancer cells. DF1 behaves similarly in another system, not as traumatic as DF3. So here's a case,
• 113:39 - 113:46:  if you knock out YTHDF3, you'll pretty much shut down the breast cancer to brain metastasis. Why?
• 113:46 - 113:53:  There's about 10, 20 proteins. Their translation, not their transcript level, their translation
• 113:53 - 113:59:  are coordinated through upregulation through this YTHDF3 pathway. So then tumor cells in the brain
• 113:59 - 114:05:  microenvironment, they need this sort of stimulation-induced translation, upregulation
• 114:05 - 114:14:  in order for breast cancer to cross the brain-blood barrier. Another system we published already,
• 114:14 - 114:19:  I'm just going to go through quickly, is this tumor immunology system. Again, I showed in many
• 114:19 - 114:27:  talks previously, this work was done by Michelle and Dali. If you take a YTHDF1 knockout mice and
• 114:27 - 114:38:  add mouse melanoma tumors in this sort of immunotherapy model, the Y-type mice
• 114:38 - 114:43:  can be killed, the tumor grows rapidly, the Y-type mice getting killed, but the knockout mice are
• 114:43 - 114:49:  very potent in controlling tumor growth in the system. And we concluded this is related to T-cell.
• 114:49 - 114:55:  It's definitely an immuno effect, but the T-cell cannot really explain this major difference.
• 114:55 - 115:01:  And that led us to look into dendritic cells. So basically, in these experiments,
• 115:01 - 115:06:  Michelle confirmed that if you isolate the dendritic cells from YTHDF1 knockout mice,
• 115:06 - 115:12:  they're much more potent in eliciting the interferon gamma response in priming,
• 115:12 - 115:19:  activating T-cells. And if you make a dendritic cell specific YTHDF1 knockout,
• 115:19 - 115:24:  that can recapitulate a whole body knockout, indicating that the majority of the effect seems to come
• 115:24 - 115:30:  from dendritic cells. What's going on in the system? If you take a Y-type knockout mice and
• 115:30 - 115:35:  isolate the dendritic cells, there's not much mRNA difference, but you see this major translation
• 115:35 - 115:41:  difference. These are targets of YTHDF1. When you knock them out, the translation is downregulated.
• 115:41 - 115:47:  These are m6A methylated, and they belong to lysosomal, phagosomal pathways. So these are
• 115:47 - 115:53:  three sort of pathways that are important in dendritic cells for antigen presentation
• 115:54 - 116:00:  and presentation. And as you can see, these are lysosomal pathways. These are m6A methylated.
• 116:00 - 116:07:  These pathways tend to be mostly m6A methylated. They tend to be almost exclusively YTHDF1
• 116:07 - 116:13:  targeted, but not these guys. Talking about transcripts in the Y-type cells, their translations
• 116:13 - 116:19:  are upregulated compared to the knockout. And we believe in the knockout case, these are proteases.
• 116:19 - 116:24:  They digest antigens. In the knockout case, you have a lot more antigens that can be presented
• 116:24 - 116:31:  on the dendritic cell surface to prime and activate T cells. Confirming that, we evaluated
• 116:31 - 116:38:  the individual proteases in this list. In the knockout versus Y-type dendritic cells, you clearly
• 116:38 - 116:45:  see the translation, the protein level differences. And you can also bring in a protease inhibitor.
• 116:45 - 116:52:  You can also have similar tumor control. And I know this is the translational
• 116:54 - 116:58:  symposium, so I'm going to get into some of the translational work on published.
• 116:58 - 117:04:  Now, this is still the published work. This is a melanoma PDO, anti-PD-1
• 117:05 - 117:12:  immunotherapy model. 40% of mice can be controlled through this anti-PD-1 sort of
• 117:12 - 117:20:  immunotherapy for their melanoma progression. But if you combine with YTHDF1 whole-body knockout,
• 117:20 - 117:27:  you get complete control of tumor growth. And we validated the sort of with a human system.
• 117:27 - 117:33:  Actually, we did take human dendritic cells. I showed that when we knocked on YTHDF1,
• 117:33 - 117:41:  they're more potent at priming T cells. And here, there are 22 colorectal cancer biopsy samples.
• 117:41 - 117:47:  You can clearly see an active correlation between YTHDF1 level and the T-cell infiltration,
• 117:47 - 117:55:  confirming YTHDF1 play an active role in terms of host response to tumor.
• 117:56 - 118:03:  Now, this presents a very nice target for medicinal chemistry and drug development.
• 118:03 - 118:07:  Just like bromodomains, there's reader proteins, there's crystal structures we
• 118:07 - 118:14:  solved previously. The first of the type and a clear pocket with something fitting nicely.
• 118:14 - 118:21:  So we went out, did the, now me, Michelle has her own lab in collaboration with Dr. Cheng Luo.
• 118:21 - 118:27:  We did ourselves. We helped went on, did the screen and here's some natural product. This is actually
• 118:27 - 118:32:  a dose well tolerated by humans, well enough to be tolerated by humans at five micromolar.
• 118:33 - 118:38:  And if you apply the small molecule in the tumor model, you can clearly see tumor control.
• 118:40 - 118:50:  And you basically enhance the dendritic cell cross-presentation capacity of the mouse system.
• 118:52 - 118:59:  You can do more experiments and show that the small molecule can increase T cell infiltration
• 118:59 - 119:07:  to the tumor site. And again, this standard anti-PD-L1 model, you can see this is a
• 119:08 - 119:14:  standard control with anti-PD-L1. And this is a combination with the small molecule. You can
• 119:14 - 119:20:  greatly reduce tumor growth with a small molecule targeting just the YTHDF1.
• 119:22 - 119:27:  What about the effect on the dendritic cell transcription versus translation? Here's the
• 119:27 - 119:33:  transcription or something's up, something's down, nothing jumps out of, you know,
• 119:35 - 119:40:  nothing jumps out. But when you look at the translation, you clearly see a down regulation
• 119:40 - 119:47:  of translation when you treat it with a small molecule. So again, these are belonging to the
• 119:47 - 119:54:  level. So more pathways are consistent with the knockout of YTHDF1 knockout mouse models.
• 119:54 - 119:59:  So I hope I sort of
• 120:00 - 120:04:  give you two stories on a lot of the experimental data,
• 120:05 - 120:09:  indicating that, indeed, there's a demethylation. It's quite
• 120:09 - 120:14:  significant, right? It tunes the repeats. And also, these
• 120:14 - 120:18:  reader proteins have quite interesting roles. And we
• 120:18 - 120:22:  touched upon the YTHDF1. There are other systems in this
• 120:22 - 120:28:  sort of m6A factor families also affect the immunotherapy. In
• 120:28 - 120:32:  a way, I always say there are two hallmarks of m6A biology.
• 120:32 - 120:35:  One is really the stem cell differentiation, tissue
• 120:35 - 120:38:  development. Another is a sort of a response to signaling
• 120:38 - 120:43:  stress and stimulation. And the immunology seems to be a sort
• 120:43 - 120:49:  of goldmine that remains to be explored in this RNA
• 120:49 - 120:54:  modification field. All right, I think I sort of caught out the
• 120:54 - 120:59:  students who did most of the work. JiangBo did the FTO work
• 120:59 - 121:03:  in collaboration with Xianbin. And the initial chromatin
• 121:03 - 121:08:  regulation was done by Jun Liu and Xiaoyang. And the YTHDF1, YTHDF2,
• 121:08 - 121:12:  YTHDF3 knockout were previously done by Hailin Shi.
• 121:12 - 121:16:  And the immunotherapy part was done by Michelle Dali, and
• 121:16 - 121:20:  lately with Cheng Luo and other collaborators. So I guess this
• 121:20 - 121:25:  is my conflict of interest statement and funding sources.
• 121:26 - 121:27:  And I'd be happy to answer any questions.
• 121:29 - 121:33:  Can any of the m6A interacting proteins, including YTH family,
• 121:34 - 121:39:  interact with methylated LINE 1 RNAs and play a role in their
• 121:39 - 121:40:  regulation?
• 121:41 - 121:46:  The YTHDC1 interacted with LINE 1 in different ways. We're
• 121:46 - 121:51:  still trying to figure out the portion of the m6A on LINE 1
• 121:52 - 121:57:  bonded by DC1 to mediate decay. And another portion of m6A on
• 121:57 - 122:01:  LINE 1 bonded by DC1, this is from my previous postdoc work,
• 122:01 - 122:06:  so I can't say too much, but that seems to regulate chromatin
• 122:06 - 122:10:  state through a different mechanism. It's not just the
• 122:10 - 122:15:  decay. So how DC1 has different roles, in one case, it's the
• 122:15 - 122:18:  decays. In other cases, it seems to serve as a scaffold
• 122:18 - 122:23:  effect. I currently don't know. Our hypothesis is related to the
• 122:23 - 122:26:  post-translational modification on DC1.
• 122:28 - 122:32:  Hi, Xuan. I have a comment and a question. So for the
• 122:32 - 122:37:  translation part, I noticed even from Sami's proteomic studies,
• 122:37 - 122:42:  so the YTH proteins, one of their interaction proteins, TOPO3B.
• 122:42 - 122:47:  So what's Fisher Group and Wei Dong Wang Group from NIA, they
• 122:47 - 122:51:  have shown that TOPO isomerase, TOPO3B, functions as RNA TOPO
• 122:51 - 122:56:  isomerase, regulating translation. Maybe you guys can kind of work
• 122:56 - 123:01:  on see if DC1, 2, 3 can, by interacting with that RNA TOPO
• 123:01 - 123:03:  isomerase, regulate translation. But they have
• 123:03 - 123:09:  already nicely shown that TOPO3B regulates certain transcript
• 123:09 - 123:14:  translations. It's not global, but some messenger RNAs probably
• 123:14 - 123:18:  require that TOPO isomerase for translation, for efficient
• 123:18 - 123:19:  translation.
• 123:19 - 123:22:  I think it's pretty clear, if you look at a lot of the data
• 123:22 - 123:27:  out there, YTHDF1 binds RNA and recruits this RNA to the ribosome.
• 123:28 - 123:33:  And also, when you have a nascent RNA come from nucleus to
• 123:33 - 123:38:  cytoplasm, it goes to YTHDF3, YTHDF1, and then YTHDF2. It's a different
• 123:38 - 123:43:  sequence. The YTHDF1, YTHDF3 not only play important roles in m6A
• 123:44 - 123:47:  regulation, it seems to also play roles in the entire sort of
• 123:47 - 123:54:  granule formation along the metabolism homeostasis pathway.
• 123:54 - 123:59:  Yeah, so I have a question on the first part of your talk.
• 123:59 - 124:04:  It's amazing. So we observed the same thing with the METTL14
• 124:04 - 124:10:  knockout. So whenever we do a ChIP experiment, we do the
• 124:10 - 124:14:  sonication, we start from some observations before we saw your
• 124:14 - 124:18:  paper, the science paper, we noticed the knockout mouse
• 124:18 - 124:22:  embryonic stem cells very easily to be fragmented. So the same
• 124:22 - 124:26:  condition we sonicate Y-type cells, the knockout cells, the
• 124:26 - 124:30:  fragment size is much smaller, very easy to get smaller. Until
• 124:30 - 124:33:  I saw your science paper, I kind of know, because of the
• 124:33 - 124:39:  chromatin is so open, that's easy to be fragmented. So then they
• 124:39 - 124:44:  come to the question of, to what extent do you think is the
• 124:44 - 124:48:  chromatin-associated RNA that is contributing to the chromatin?
• 124:48 - 124:53:  This is a great question. Yeah, I don't have time to... Yeah, sorry.
• 124:53 - 124:56:  Do you have more to ask? Or is it?
• 124:56 - 124:59:  No, just the chromatin-associated RNA or LINE one.
• 124:59 - 125:03:  Right, right, right. That's a great question. The more we dig
• 125:03 - 125:07:  into this, the more we found this is complicated, it's
• 125:07 - 125:10:  really depends on which system dominates, right? So there are
• 125:11 - 125:15:  at least three layers of regulation. One layer through
• 125:15 - 125:18:  chromatin-associated regulatory RNAs, the enhancer RNAs, repeats,
• 125:18 - 125:22:  and promoter-associated RNA. The second layer is messenger RNA,
• 125:22 - 125:25:  because lots of these histone modifiers, chromatin
• 125:25 - 125:29:  remodelers, messenger RNA can be m6A methylated. So in certain
• 125:29 - 125:33:  cancers, that seems to be the dominating factor, right? And
• 125:33 - 125:38:  but also the pre-mRNA m6A can also have an effect through
• 125:38 - 125:44:  things like R-loop. So, and so there's three effects in mouse
• 125:44 - 125:49:  ES cells consistently, we realize it turns to be chromatin regulation
• 125:49 - 125:54:  turns to be from the nuclear effect. But in cancer is really
• 125:54 - 125:58:  depends on which cancer can be mixed or one factor can dominate
• 125:58 - 126:01:  the other effect. FTO is the same thing. If you look at the
• 126:01 - 126:04:  mouse ES cells, the LINE system seems to dominate, but if you switch
• 126:04 - 126:10:  to a melanoma system, some of the EZH, the demethylase of K27
• 126:10 - 126:14:  methylation, those seem to dominate in the chromatin
• 126:14 - 126:20:  regulation. And the YTHDC1 in sort of a primary cell versus mouse
• 126:20 - 126:24:  ES cells versus cancer also can have very different functions.
• 126:24 - 126:25:  Yeah, thank you.
• 126:27 - 126:30:  Great. Thank you. I think I'm going to stop the discussion.
• 126:30 - 126:34:  There. Thank you, Xuan. In the interest of time, we have now
• 126:35 - 126:42:  we're not going to go into the session on discussion of the
• 126:42 - 126:49:  field, which will include the three speakers. So if you can
• 126:49 - 126:59:  put your thank you videos on. So I'm going to be asking you
• 126:59 - 127:04:  questions. And if anybody wants to send some questions in, we
• 127:04 - 127:07:  might have time to ask those too. But to be honest, we're
• 127:07 - 127:11:  quite late. So we may not, but still, if you have a pressing
• 127:11 - 127:17:  question, please ask it. So first of all, I'd like to say
• 127:17 - 127:20:  it's great to see you all together because we don't meet
• 127:21 - 127:24:  anything other than virtually nowadays. And to be honest,
• 127:24 - 127:29:  actually, it's the first time we've met, I think since last
• 127:29 - 127:33:  year at the Storm Therapeutics meeting. And I was looking at
• 127:33 - 127:42:  some photos and I found this one here, which was really, you
• 127:42 - 127:45:  know, showing that we were all together here at this meeting
• 127:46 - 127:52:  as a family, along with François Fuchs and Eric Mischka. I
• 127:52 - 127:53:  thought it was a good meeting.
• 127:55 - 127:57:  That's a dramatic one, Tony.
• 127:57 - 127:58:  Is it what?
• 127:58 - 127:59:  So you're going to show the Sammy one.
• 128:01 - 128:08:  But then I found this picture here. Which, you know, Sammy,
• 128:08 - 128:11:  I'm not quite sure why we were collectively trying to throttle
• 128:11 - 128:16:  you. But I think it just shows that we're just a family, a
• 128:16 - 128:19:  normal family when we meet, you know, stuff happens.
• 128:20 - 128:23:  I know I was bruised, I was bruised for days.
• 128:26 - 128:32:  Well, it was fun. And thank you for attending. So now let's have
• 128:32 - 128:38:  some fun with questions. I have a few questions here. So the
• 128:38 - 128:43:  first one is, will there be more RNA modifications as abundant
• 128:43 - 128:51:  and as important as m6A? Anybody wants to? Maybe Schwan starts
• 128:51 - 128:51:  first?
• 128:53 - 128:56:  Well, Sammy can go first. I think he's raising hands.
• 128:56 - 129:01:  Well, I was just I was just gonna say it's it's I think m6A
• 129:01 - 129:07:  is special in terms of it as an internal modification. It's so
• 129:07 - 129:10:  abundant, and so prevalent, and was discovered way back in the
• 129:10 - 129:15:  1970s. I just don't think anything is going to achieve
• 129:15 - 129:19:  that level of abundance. You know, there's a lot of other
• 129:19 - 129:23:  modifications. And now with SCARLET, we're pretty confident.
• 129:24 - 129:26:  First of all, I think we were confident about pseudouridine to
• 129:26 - 129:31:  begin with. But surprisingly, 5-methylcytosine actually does seem to
• 129:31 - 129:36:  exist in mRNA. And that had been controversial. But there are
• 129:36 - 129:40:  specific sites where we can actually see a 5-methylcytosine spot.
• 129:40 - 129:48:  And I'm now a believer in other modifications, especially 5-methylcytosine.
• 129:49 - 129:51:  Some of the other ones we haven't yet been able to validate
• 129:51 - 129:57:  at least as being abundant. But I think what I think the
• 129:57 - 130:00:  question should not be, is there going to be something as
• 130:00 - 130:04:  abundant as m6A? But are there any others? And I think there
• 130:04 - 130:08:  will be a lot of modifications expressed on specific mRNAs that
• 130:08 - 130:14:  have uniquely acquired a tRNA-like shape to co-opt a tRNA
• 130:14 - 130:17:  enzyme. But I don't think it'll be as general as m6A.
• 130:18 - 130:20:  Great, thanks, Sami. Anybody else?
• 130:22 - 130:28:  I think that's good. Sami agreed there are other
• 130:28 - 130:33:  modifications now. I think that there are two answers to this
• 130:33 - 130:37:  question. Another answer would be that we have to look at the
• 130:37 - 130:42:  tissues differently from these cancer cell lines. And in fact,
• 130:43 - 130:48:  in some of these tissues, we see certain RNA modifications really
• 130:48 - 130:52:  abundant, thousands of them. Or in certain tissues, we see 6,000
• 130:52 - 130:58:  sites. So that's not quite there with m6A, which is in some of
• 130:58 - 131:02:  these tissues 20,000 to 30,000 sites, but in the same order of
• 131:02 - 131:07:  magnitude. So I think going into the real physiological system,
• 131:07 - 131:11:  you'll see a big difference. In some cases, in HeLa cells, we
• 131:11 - 131:16:  see 45 sites, but in some of these tissues, 6,000 sites for
• 131:16 - 131:19:  certain modifications. That's one thing. The second thing is,
• 131:20 - 131:23:  you know, we have a center here to develop sequencing methods.
• 131:23 - 131:27:  And just in the last year, we have methods that we can map pretty
• 131:27 - 131:29:  much every modification quantitatively. We haven't
• 131:29 - 131:32:  published anything. Out of that effort, there are lots of
• 131:32 - 131:36:  surprises. So Tony, I will say that there are probably two other
• 131:36 - 131:42:  modifications in our hands that are not approaching the m6A
• 131:42 - 131:46:  abundance, but in the same ballpark, it's probably a few
• 131:46 - 131:51:  folds less. But again, I agree with Sami to a certain extent. I
• 131:51 - 131:56:  think if you look at the type of function of m6A, it's just so
• 131:56 - 131:58:  broad. I think still the biological community
• 131:58 - 132:03:  underappreciates how broad m6A regulates. And any modification
• 132:03 - 132:08:  shows 20% to 30% of that breadth is like a major histone mark or
• 132:08 - 132:14:  something like that, right? So I think functionally, should be a
• 132:14 - 132:18:  couple, two, three, quite important RNA modifications
• 132:18 - 132:19:  internally.
• 132:20 - 132:24:  Yeah, I would just say, I think what the field needs is less
• 132:24 - 132:29:  controversy and more definitive data. And we just see over and
• 132:29 - 132:31:  over again, these papers claiming modifications, then
• 132:31 - 132:34:  someone else publishes a paper saying they're not, and then so
• 132:34 - 132:37:  on and so forth. It just goes back and forth. I think the
• 132:37 - 132:40:  SCARLET technique or the SCARPIT technique that we use is just
• 132:40 - 132:44:  unambiguous. You can see the modification and you get the
• 132:44 - 132:50:  stoichiometry. And I think the field has really been damaged in
• 132:50 - 132:54:  the eyes of other RNA community members because of all these,
• 132:54 - 133:00:  you know, problems that have happened. And the broad claims
• 133:00 - 133:03:  that have been made and it needs to stop. The goal now is if
• 133:03 - 133:06:  there's a modification, it has to be documented in maybe five
• 133:06 - 133:11:  or six mRNAs exactly demonstrate that it is absolutely there. And
• 133:11 - 133:14:  it's at this stoichiometry, and you can see it biochemically, not
• 133:14 - 133:17:  through a mapping method, because the mapping methods, you
• 133:17 - 133:20:  know, I've written about it, Mark Helm has written about it,
• 133:20 - 133:25:  are so prone to artifacts that you have to validate results. And
• 133:25 - 133:28:  you know, SCARLET was great, they did 15 mRNAs. In my opinion,
• 133:28 - 133:31:  there's no reason why people can't use SCARLET because Tao
• 133:31 - 133:35:  did it very successfully on mRNAs, not on high
• 133:35 - 133:39:  abundance, on mRNAs, but the SCARPIT will make it easier.
• 133:39 - 133:44:  Great, thank you. Anybody else? Anything else on that one?
• 133:44 - 133:47:  Well, just let me add one brief comment. I agree with everything
• 133:47 - 133:49:  that's been said. But I think that when you talk about
• 133:49 - 133:53:  abundance, you know, with m6A being as prevalent because it
• 133:53 - 133:57:  exists on thousands of mRNAs, and you know, many of these global
• 133:57 - 134:00:  detection methods like mass spec can pick that up, right. But if
• 134:00 - 134:04:  you're talking about maybe a modification that occurs on one
• 134:04 - 134:08:  mRNA, or one type of mRNA, or a small family of mRNAs, that
• 134:08 - 134:11:  could still be like 100% stoichiometry, but you wouldn't
• 134:11 - 134:15:  detect that when you look globally. So you know, and you
• 134:15 - 134:18:  know, SCARLET is a helpful method, but at the same time,
• 134:18 - 134:21:  it's a validation method, it's not discovery, right? So you're
• 134:21 - 134:24:  not going to find any new modifications using SCARLET. So
• 134:24 - 134:27:  I think, you know, when we talk about abundance, we have to
• 134:27 - 134:31:  clarify, are we talking about abundance of mRNAs, or are we
• 134:31 - 134:36:  talking about abundance on the message or abundance throughout
• 134:36 - 134:37:  the transcriptome?
• 134:38 - 134:38:  Yeah.
• 134:40 - 134:44:  I think in that regard, what the field needs is the bisulfite
• 134:44 - 134:50:  type of gold standard. I think in the next year, we're going to
• 134:50 - 134:50:  show a few.
• 134:54 - 134:59:  And nanopore might actually be very useful just for finding
• 134:59 - 135:04:  sites that have abnormal currents. And as more nanopores,
• 135:04 - 135:07:  different types of pores become available, and greater
• 135:07 - 135:10:  sensitivity occurs, which is recently, you know, recently now
• 135:10 - 135:14:  the nanopores are much more accurate for modifications. That
• 135:14 - 135:18:  might be a way to discover some. And then SCARLET could be used
• 135:18 - 135:20:  to figure out exactly what modification is there.
• 135:21 - 135:21:  Yeah.
• 135:24 - 135:28:  Great, guys. That's good. Good conversation there. I'm gonna
• 135:28 - 135:33:  ask a question about the clinical side of things. So we
• 135:33 - 135:38:  know that enzymes that modify RNA are demonstrated to have an
• 135:38 - 135:41:  importance in cancer. But will there be other diseases that
• 135:42 - 135:44:  these RNA modifying enzymes will be important in?
• 135:48 - 135:51:  I think this is the complicated question. You know, I think one
• 135:51 - 135:56:  of the things that's going to be really helpful is as these drugs
• 135:56 - 135:59:  become more available, the metal 3 inhibitors in particular,
• 135:59 - 136:02:  and you know, STORM has a patent, and you know, they've
• 136:02 - 136:05:  revealed a lot of their chemical matter. As those
• 136:05 - 136:09:  compounds become available, I think it will be a lot easier to
• 136:09 - 136:12:  just do screening. And you know, obviously, it may be hard to
• 136:12 - 136:15:  de novo pick a good, good thing. But some lab that's studying,
• 136:16 - 136:19:  you know, glomerulonephritis and has a great model, they
• 136:19 - 136:23:  might throw in their inhibitor and might find it. So I think
• 136:24 - 136:27:  the small molecule inhibitors will be a big step forward as
• 136:27 - 136:28:  they become commercially available.
• 136:28 - 136:33:  I mean, we are collaborating with other labs to look at other
• 136:33 - 136:38:  diseases through the molecules that we have in STORM. But I'm
• 136:38 - 136:41:  thinking in general, do you think there's going to be more
• 136:41 - 136:44:  evidence at the moment, you know, for example, Richard
• 136:44 - 136:48:  showed us some evidence over an amplification where we're
• 136:48 - 136:50:  thinking, okay, this is overexpression, but actually is
• 136:50 - 136:54:  amplification underneath that. And therefore, it's like more
• 136:54 - 136:59:  genetic evidence, perhaps. Do you think there'll be more
• 137:00 - 137:00:  evidence?
• 137:01 - 137:05:  Right. So Tony, I think the genetics is really the key. I
• 137:05 - 137:09:  think a lot of these functionality things, you know, does this work
• 137:09 - 137:12:  on this work on that genetics, you know, end of the day is
• 137:12 - 137:15:  genetics, right? We kind of all about chemistry, molecular
• 137:15 - 137:19:  biology, nice, SCARLET, whatever. But genetics, right,
• 137:19 - 137:21:  everything goes back to genetics, knock out the model,
• 137:21 - 137:25:  what's going on with that pathway. So I think there's at
• 137:25 - 137:29:  least one area that I probably will start to think about, you
• 137:29 - 137:32:  know, being busy with other senses, really the, you know, if
• 137:32 - 137:36:  you think about m6A, the hallmark of m6A, as I said, is
• 137:36 - 137:38:  a cell differentiation development and signaling
• 137:38 - 137:41:  stimulation. I think cell therapy will be a really
• 137:41 - 137:47:  interesting area, particularly stem cell therapy. m6A is so
• 137:47 - 137:50:  important in cell differentiation. For instance,
• 137:51 - 137:55:  you know, we recently mapped hematopoietic stem cell
• 137:55 - 137:57:  differentiation based on evolution quantitatively, you'll
• 137:57 - 138:01:  be surprised the groupie factor, you know, there's 10s of
• 138:01 - 138:05:  transcription factors, they're m6A modified, that seems to really
• 138:05 - 138:12:  guide cell differentiation fate. So the simple applications would
• 138:12 - 138:15:  be to produce a ton of stem cells, hematopoietic stem cells
• 138:15 - 138:19:  or knee stem cells for tissue repair, regeneration, CRISPR
• 138:19 - 138:23:  systems, if you look at anemia, right, those CRISPR people,
• 138:24 - 138:26:  right, they have to generate a hematopoietic stem cell, right,
• 138:27 - 138:29:  and eventually to be applied clinically, how do you
• 138:29 - 138:33:  stabilize, you know, hematopoietic stem cell, expand
• 138:33 - 138:37:  them, right, m6A plays a key role there. But then going forward,
• 138:37 - 138:41:  think about generating different types of immune cells or other
• 138:41 - 138:46:  cell types, I think there got to be lots of opportunity.
• 138:47 - 138:51:  And just to add in following on the genetic theme, I think, you
• 138:51 - 138:53:  know, if you look at the developmental disorders that are
• 138:53 - 138:56:  in children that are caused by mutations in some of these same
• 138:56 - 139:01:  genes, right, so WDR4 leads to microcephaly and dwarfism, for
• 139:01 - 139:04:  example. So, and that's true of many of these RNA modifying
• 139:04 - 139:07:  enzymes, they tend to impact brain development in particular.
• 139:08 - 139:10:  So I think that maybe looking at some of the neurodevelopmental
• 139:10 - 139:13:  disorders, Alzheimer's disease, and things like that, that's
• 139:13 - 139:17:  where these modifications may be particularly important. And then,
• 139:17 - 139:21:  you know, again, Xuan published a paper and others have published
• 139:21 - 139:25:  papers on GWAS studies and QTLs, right, looking at mapping m6A
• 139:25 - 139:30:  across human transcriptomes. And I think that they're going to be
• 139:30 - 139:33:  traits that are linked with maybe even individual m6A
• 139:33 - 139:37:  modification sites. So I think that that's going to guide, you
• 139:37 - 139:39:  know, this area as well in the future.
• 139:40 - 139:44:  Right, actually, I just add to Richard's comment, a major goal
• 139:44 - 139:48:  is to have a quantity of methods that you can really map this
• 139:48 - 139:53:  QTL. There's one example, just following Richard's thing, in
• 139:53 - 139:56:  pancreatic cancer, there's some patient samples, there's one
• 139:56 - 139:59:  mutation site that gets rid of GGACU.
• 140:00 - 140:04:  Consensus sequence cannot be methylated and contributes to pancreatic cancer, right?
• 140:04 - 140:09:  So the question is, how many of these are out there in brains and elsewhere? And, you know,
• 140:09 - 140:13:  really, I think, again, genetics is really the key. And then once we have
• 140:13 - 140:16:  enabling methods that combine with genetics.
• 140:19 - 140:25:  Yeah, I think one thing to think about is, if, especially in the context of m6A,
• 140:25 - 140:29:  if m6A abnormalities are indeed associated with other diseases,
• 140:30 - 140:34:  would you want to inhibit metal-3 in order to treat them? You know, especially if you're
• 140:34 - 140:38:  dealing with a childhood neurodevelopmental disorder, that might, you might imagine
• 140:38 - 140:44:  continuous treatment, that might be toxic. You know, one of the great results from STORM
• 140:44 - 140:50:  is how they can do such a short treatment, just five days, and then withdraw the metal-3 inhibitor,
• 140:50 - 140:56:  but then get effects that last several weeks as they induce, you know, systemic differentiation
• 140:56 - 141:03:  of their leukemia cells. So I think even if metal-3 abnormalities or m6A is abnormal
• 141:03 - 141:08:  in certain diseases, it may not be feasible to do continuous metal-3 inhibition.
• 141:09 - 141:15:  It may be in certain diseases, and the AML data is beautiful from STORM, and it really shows how
• 141:15 - 141:18:  you don't have to get toxicity because you can do a short treatment.
• 141:19 - 141:27:  I mean, neurological diseases will be approachable, but maybe childhood diseases not. So I think there
• 141:27 - 141:34:  are areas that, other than cancer, that will probably be useful, I think, for these molecules.
• 141:37 - 141:40:  I'm going to ask a question which is not really addressed very much,
• 141:42 - 141:47:  to throw in something new. Is RNA methylation potentially heritable, and how?
• 141:48 - 141:54:  I think this is a question for Xuan.
• 141:56 - 142:01:  I mean, so, I mean, is there any reason to believe that they are?
• 142:01 - 142:06:  And if there were, are there any mechanisms that you can think of that could mediate them?
• 142:08 - 142:13:  I think if m6A regulates chromatin state, right, the indirect way would be through the chromatin
• 142:13 - 142:20:  state. And I think we have at least two examples unpublished, again, from my previous co-worker's
• 142:20 - 142:24:  lab, hopefully will come out, which, you know, I saw the data clearly there.
• 142:25 - 142:31:  The other part is a maternal effect. That part I'm not so sure yet, right? You see some of the
• 142:31 - 142:39:  microRNA inheritable, I think that's from Israel and some other work, right? It's a maternal effect.
• 142:40 - 142:49:  So there, I think it requires some experts to look at these m6A regulators from a maternal
• 142:49 - 142:54:  act. Do they have a heritable effect? I think through a chromatin state,
• 142:54 - 142:59:  I'm pretty sure the answer would be yes, but then requires a lot more sort of work.
• 143:00 - 143:09:  Yeah. So another question here is regarding the regulation of these enzymes and these
• 143:09 - 143:17:  modifications. So there are many ways that regulation could happen. How do you think,
• 143:18 - 143:22:  can you predict ways in which these enzymes will be regulated by reversible modifications on the
• 143:23 - 143:29:  enzymes, your expression levels clearly, subcellular localization, can that regulate
• 143:29 - 143:36:  things, post-translational modifications, especially what would be your favorite way of regulation,
• 143:36 - 143:41:  prediction-wise? You know, I think that's such an important question, but I, you know, to me,
• 143:42 - 143:49:  what I feel we need is very clear examples where the writing of m6A is indeed regulated.
• 143:49 - 143:54:  You know, where you can say, without a doubt, we do this treatment and there's more m6A. And
• 143:54 - 144:00:  if we can't replicate it, we can't study it, right? So we don't have super great examples.
• 144:00 - 144:08:  I think some DNA damage examples have been working in our lab. The effects are not super great. And
• 144:08 - 144:12:  I think it may be getting the right time point, but you need to know for sure what your model
• 144:12 - 144:18:  system is for regulated m6A. But, you know, then I think the question becomes, is this massive
• 144:18 - 144:23:  writer complex, which has so many proteins of unknown function, do those individual proteins
• 144:23 - 144:30:  integrate signals in some way? Like does VIRMA integrate signals, ZC3H13? I think that, you know,
• 144:30 - 144:35:  it's completely unclear why do you have a massive writer complex rather than a simple enzyme?
• 144:35 - 144:39:  And it seems like there must be some role for these enzymes to play in regulating
• 144:40 - 144:43:  and it seems like there must be some role for these proteins.
• 144:44 - 144:50:  But to me, I think it's the two clear steps. I need an example. And now that we have better
• 144:50 - 144:55:  assays for quantifying m6A and site-specific sites, like I want a specific site, like
• 144:55 - 145:04:  these five m6A sites will go from 5% to 40%, then I have a clear readout. But I don't have that 4%
• 145:04 - 145:12:  to 40% change for me to even study this because I haven't been able to reproduce that. And so,
• 145:12 - 145:16:  but that being said, I think there's a lot of avenues, but I think a clear readout is important
• 145:16 - 145:24:  too. And I think, you know, it's a little bit more complicated than we can appreciate right now,
• 145:24 - 145:28:  because we know that METTL3 is an oncogene. Several labs have shown that, right? And we know
• 145:28 - 145:33:  that METTL3 is upregulated in cancer. They're not necessarily amplified, but it is upregulated.
• 145:34 - 145:39:  And perhaps getting back to Simon's question about, or point about whether that leads to
• 145:39 - 145:44:  more m6A, that's probably less proven at this point. But I think it's also true to say that,
• 145:44 - 145:50:  you know, it's not necessarily just a balance between the level of m6A mediated by the
• 145:50 - 145:56:  demethylases and the methyl transferases, because, you know, you overexpress METTL3, you can
• 145:56 - 146:02:  drive germination, but you might expect that if you knocked out ALKBH5, for example, you would
• 146:02 - 146:08:  also get more methylation and perhaps phenocopy what you see with METTL3 overexpression,
• 146:08 - 146:13:  but you don't always see that, or perhaps even rarely see that. And of course, there could be
• 146:13 - 146:18:  redundancy with FTO and ALKBH5, but I think that there's maybe some other functions of METTL3
• 146:18 - 146:24:  that is perhaps in addition to its m6A catalytic activity. And, you know, we published
• 146:24 - 146:29:  one example of that in the past, but I think that until we really go back and look, like Sami said,
• 146:29 - 146:35:  at the levels of the modification itself, and I think that this has implications for, you know,
• 146:35 - 146:39:  Storm and other companies that have inhibitors, like, you know, which subset of patients are you
• 146:39 - 146:44:  going to treat? Because it's going to be hard to identify which ones are going to be more responsive
• 146:44 - 146:51:  or less responsive without, you know, biomarkers. So I think that, you know, it's more complicated
• 146:51 - 146:57:  than we can currently understand. But, you know, I also did, I did also show one example where having
• 146:57 - 147:04:  excess METTL1 does indeed lead to extra m7G modification, right? And so, you know,
• 147:04 - 147:08:  there's a clear example of having too much enzyme can have pathological consequences.
• 147:10 - 147:15:  Yeah, you know, I think the ALKBH5 is such an interesting example. Remember in the ALKBH5
• 147:15 - 147:21:  knockout, at least morphologically, the animals look normal. They seem to run around. They have
• 147:21 - 147:27:  major testes defects, but they don't seem to have major phenotypes elsewhere. I'm not saying that
• 147:27 - 147:31:  they don't have endophenotypes, but I'm just saying at least in terms of major phenotypes.
• 147:31 - 147:39:  So ALKBH5 is not the opposite. ALKBH5 knockout is not the opposite of METTL3. And there may
• 147:39 - 147:46:  be very selective mRNAs that are preferred by ALKBH5. But, you know, why would that be?
• 147:46 - 147:51:  What is the basis for selectivity? Does ALKBH5 have, I mean, clearly, in my opinion, if you want
• 147:51 - 147:56:  to study ALKBH5, testes is the place to do it because that's where you get a clear, unambiguous
• 147:56 - 148:02:  phenotype. But if it has phenotypes elsewhere, I have to say, we've never seen a change in m6A,
• 148:02 - 148:06:  but, you know, maybe the wrong cell types we're not looking at, but we're not, we might not be
• 148:06 - 148:12:  looking at the right cell types when we knock out ALKBH5. But I think ALKBH5 is not so obvious.
• 148:12 - 148:16:  It's not just the, it's not simply the opposite of METTL3. It's probably specialized. It
• 148:16 - 148:22:  probably has specific roles. It may be on specific RNAs. Right. And I think, you know, and so that's
• 148:22 - 148:27:  basically agreeing with what I said. And I think that also, if you go back to Xuan's paper,
• 148:27 - 148:37:  and we've recapitulated this in my lab, it's like, right. So you're knocking out Xuan's
• 148:37 - 148:43:  intermediate, you see a very, very minor increase in m6A level. Whereas if you overexpress it in the
• 148:43 - 148:48:  same cell, then you see a decrease, a more substantial decrease. So I think, you know,
• 148:48 - 148:54:  we need to know where is this equilibrium and that might vary in different cell contexts,
• 148:54 - 149:03:  different cell types. So I probably would jump in a little bit. I think, I think we should
• 149:03 - 149:11:  come back to the original comment we made that m6A is so important. Out of several hundred or
• 149:11 - 149:16:  one to 200 different cell types, it's essential, critical for differentiation, right? You have
• 149:16 - 149:22:  such an important mark for almost every tissue development. It's hardly impossible to
• 149:22 - 149:28:  imagine it's not integrated into the entire developmental signaling, right? So I think
• 149:28 - 149:34:  we're just scratching the surface. I think all this phosphorylation, arginine methylation, all
• 149:34 - 149:41:  these signaling, you know, impacts transcription translation at the same time impacts m6A,
• 149:42 - 149:47:  you know, along the same sort of regulation landscape. How does that work, right? It's
• 149:47 - 149:54:  something I think the field needs to figure out in the next 10, 15 years. One area that I think
• 149:55 - 150:00:  the RNA biologists tend to work on RNA and then the chromatin people tend to work on chromatin,
• 150:00 - 150:05:  but it would be nice that, you know, NIH is somewhat a different tissue type developmental
• 150:05 - 150:11:  stage. You know, a simple experiment, just look at a MET3, MET14, ChIP-seq, right? Where they bind,
• 150:11 - 150:16:  right? What kind of RNA is the impact, right? And, and, you know, there's lots of unpublished
• 150:16 - 150:20:  results suggesting they're different, right? They impact chromatin differently,
• 150:20 - 150:26:  they impact messenger RNA differently, and you can correlate with the outcome of the methylation.
• 150:27 - 150:32:  So again, I said, you know, as I said, you know, you need a really good technology,
• 150:32 - 150:38:  which you can really look at the difference of this. And, you know, we generate this map,
• 150:38 - 150:45:  and they're very different along senses like a MYC or STAT or, you know, it's clearly going from
• 150:45 - 150:52:  nothing to, you know, 100% or 80% methylation in some cases and a different developmental stage.
• 150:54 - 150:57:  I think I suspect that that's just at the tip of the iceberg.
• 150:58 - 151:04:  Yeah, agree with that one. So you, thank you very much, everybody. It's been great talking to you,
• 151:04 - 151:08:  and I'm going to pass you back to Eleanor for a final goodbye.
• 151:10 - 151:10:  Thanks, Tony.
• 151:13 - 151:18:  Thank you, everybody. That was awesome. And Richard, yes, you do definitely, definitely
• 151:18 - 151:22:  win the best answer to that question from Abcam's perspective. Thank you for that.
• 151:22 - 151:27:  So yes, thank you ever so much to everybody. Thank you to all the speakers today, to Frankie,
• 151:27 - 151:33:  to Abdul, to Sammy, to Richard, to Chuan, and thank you, Tony, for moderating. You've been
• 151:33 - 151:38:  absolutely fabulous. Thank you to everyone who's partaken, who's asked questions, and thank you for
• 151:38 - 151:42:  everyone who's stuck around to the end. Lastly, we'd just like to point out that this is part
• 151:42 - 151:46:  of a series, and the next one is going to be polycomb and chromatin remodeling, and that's
• 151:47 - 151:53:  next Monday on the 9th of November. So again, thank you to everyone who's, yeah, thank you to
• 151:53 - 151:56:  the speakers. Thank you for joining, and yeah, that's been absolutely fabulous. Thank you.
• 151:58 - 151:58:  See you guys.
• 151:58 - 152:00:  Bye, everybody. Take care.
• 152:00 - 152:01:  Bye.