Scientist Stories: Stanley Qi, CRISPR Technologies for Genome Engineering

Good afternoon and welcome to the 4th annual Thomas and Salma Haider Biomedical Breakthrough Lecture. I would like to recognize the planning committee for this event. Dr. Thomas and Salma Haider, generous donors and supporters of the School of Medicine mission. Dr. David Low, Senior Associate Dean for Research and Distinguished Professor of Biomedical Sciences. Dr. Monica Carson, Chair of the Division of Biomedical Sciences, Professor of Biomedical Sciences, and the S. Sue Johnson Presidential Endowed Chair. Dr. Emma Simmons, the Salma Hyder Endowed Chair for the Thomas Hyder Program, Senior Associate Dean for Student Affairs. Edna Yohannes, Executive Director, as well as the development team. Linda Ryman, Chief of Staff and Assistant Dean, as well as the strategic initiative team. I would especially like to recognize Dr. Emma Simmons, Senior Associate Dean of Student Affairs, who's joining us tonight as the Salma Hyder Endowed Chair for the Thomas Hyder Program. A special thanks to Tom and Salma Hyder for having the vision to create and support this event. We would not be here tonight without their ongoing support from the very beginnings of the School of Medicine. This afternoon, we are fortunate to have Dr. Stanley Qi joining us to present on the synthetic genome engineering for genomics and therapeutics. Dr. Qi's lab combines synthetic biology with genome engineering to understand the function of mammalian genomes and develop gene therapy. Based on my brief meeting with Dr. Qi this afternoon, I can assure you that you will enjoy this excellent lecture highlighting some remarkable possibilities in genome engineering. Now it is my pleasure to introduce Dr. Thomas Heider, a world-renowned spine surgeon, philanthropist, entrepreneur, and inventor. He has been a driving force behind the creation of UCR School of Medicine. Dr. Heider established the Thomas Heider Program at UC Riverside School of Medicine, which provides an avenue for qualified and culturally diverse students to enroll in UCR School of Medicine each year. This program stems from the earlier UCR UCLA Thomas Hyder Program in Biomedical Sciences. Part of the branch campus of UCLA Medical School, where students completed their first 2 years of medical school at UCR before entering UCLA for their clinical years and graduating with the medical doctor degree from UCLA. Dr. Hyder is a recipient of the UC Riverside Chancellor's Medal of Honor, the American Medical Association's Pride in the Profession Award. And he was awarded the Outstanding Medical Profession of the Year Award by the United States Congress. He has developed and patented significant medical instrumentation that is used in spine surgery throughout the country. He is a founding member of the American College of Spine Surgery. And Tom and his wife, Salma, created and operate the Children's Spine Foundation, through which they provide free medical care to children at their clinic, the Hyder Spine Center, as well as in hospitals abroad. Dr. Haider has authored numerous medical publications and was appointed by the Governor of California to the Medical Board of California and later to the State Bar Board of Governors. He is currently serving as Associate Clinical Professor at UCR School of Medicine, as well as clinical professor at the UCLA School of Medicine. Without further ado, Dr. Thomas Haider.

Thank you, Dr. Deese. This is a great evening. I did have a chance to talk to Dr. Chi earlier, today, and it was a great conversation. I would like to begin by thanking our Chancellor Wilcox for making all of this possible. And then special thanks to Dr. Deborah Dees, Dean of the School of Medicine. She has provided this opportunity and the, and the mechanism to have these lectures, to present our campus to our visiting professors, and also bring these great lectures to the UCR community and also Riverside. I also would like to thank our planning committee, Dr. David Lowe, Dr. Monica Carson, Edna Johannes. They have put countless hours planning for this lecture, and it really wouldn't be in the shape and form it is now without their help and their guidance. I'm especially honored to introduce Dr. Stanley Chee. Dr. Chee got his bachelor's degree in physics from Tsinghua University in China. He got his master's in physics from UC Berkeley, and then he went on to get his PhD in a combined program with UC Berkeley and UCSF. His PhD was in bioengineering with emphasis on genetic engineering. His advisors were Dr. Adam Arkans and Jennifer Doudna. As most of you may know, Dr. Doudna was awarded the Nobel Prize last year for chemistry. Dr. Qi has been an assistant professor at Stanford University since 2014, and again working on genetic engineering And he has accomplished quite a bit in the past 6, 7 years that he has been at Stanford, and obviously way before that in his graduate education at Berkeley. He has developed the multiplex CRISPR for gene regulation and editing. He has also developed CRISPR for live cell imaging, and also CRISPR as antiviral for COVID-19. That one is called Pac-Man, very appropriate name. Dr. Qi has been involved in over 70 publications in peer-reviewed scientific journals. He has numerous patents, 6 of which I saw that's already issued and 12 that are in provisional status. A lot of his patents revolve around the CRISPR technology and genetic engineering. He has received numerous awards including Director Early Independence Award from NIH, Career Award from National Science Foundation, and Technology Review 35 Innovators Under 35 from MIT. We're so honored to have him here. He— his lecture is going to be fascinating. The possibilities with genetic engineering that he's involved with can solve problems with diabetes, blindness, and even Fountain of Youth, who knows? But please join me in welcoming Dr. Stanley Chee.

Thank you so much for the kind introduction, Dr. Thomas Heider, and thank you so much for your generosity in supporting the medical education and the research. It's a true honor really for me to give this lecture. And to be honest, I'm humbled to join such a brilliant group or audience and talk about our research. And today my topic is going to be on the synthetic genome engineering for genomics and therapeutics. As you have heard from Dr. Thomas Heider's introduction, I was trained as a physicist and changed my path to become a synthetic biologist and a bioengineer. As a physicist, I was intrigued to learn about the small things, for example, the atoms, how they compose and constitute matter. And this type of curiosity drove me to think about the essence of life, DNA. Fundamental questions on DNA always puzzle me. For example, shown here is a picture depicting the— what our human genome, how organized they are, and the questions around how they control our inheritance. What is the logic behind the life? And what causes a variant of the DNA to cause disease? These are largely unsolved. The human genome is a complete suite of genes and elements that control genes, and also is a blueprint of life, if you don't mind me calling that. In the past decades, two transformative technologies have emerged to help us unveil its secrets. For example, sequencing. And via sequencing, scientists were able to sift through almost 3 billion base pairs of our genome to find clues that may have a role in our health and disease. Yet this was not enough. To build a causal relationship between the DNA and how they function, and also to possibly correct some mistakes in the DNA, genome engineering emerged. Many likened the genome engineering to a molecular scissor, like shown here. This is true, and as you have kept reading about some news in recent years, CRISPR is a specific type of molecular scissor, and in particular, it's a type of bacterial immunity to fight phages. Appear as such a precision molecule with tremendous power on our genome DNA. And shown here is a 3-dimensional depiction of the Cas9 enzyme and how it pairs with its RNA and binds to a double-stranded DNA. And this is actually how the molecule in the natural case used to find its foes, which are bacteriophages, plasmids. That impose survival danger to the bacteria cells. And harnessing this fundamental mechanism, one single enzyme use one single guide RNA to quickly search and cut DNA. In the past decade, we heard so many exciting development about the gene editing and the base editing. Shown here is gene editing when the enzyme guided by RNA binds to a piece of DNA, it will have a precision cut on the DNA, which can be used to guide how this piece of DNA can be repaired to cause a correction of the mutation, as an example. And another branch goes to base editing by further recruiting an enzyme, for example, an adenosine deaminase, which can precisely mutate the adenine into a guanine, so A to G mutation. All these are very exciting and how change our way to modify DNA. Yet in my talk, I really want to put these exciting technologies in the context of the human genome engineering. As you can see, the human genome is not just a stack of letters, A, T, C, G. It actually spans across a large spectrum from angstrom to more than 10 micrometers in the human nucleus. And gene editing and the base editing technologies give us power to modify the nucleotides. With 3 billion base pairs in the human genome, since it's not just stacking ATCG, and there are also many other layers in the genome that requires regulation. And it's actually this huge scale of organization grants the genome functions that beyond our imagination. And I really like to emphasize the main point of the talk. The genome is not just a ladder, A, T, C, G in order. And in a similar sense, genome engineering is not just editing. It's actually exciting to see two decades of the epigenome research started to look into two large regions of the genome. One is epigenome, meaning the chemical modifications on the DNA, such as DNA methylation. The second is the three-dimensional genome organization, including enhancer-promoter interaction and how the chromosomes form territories. How can we engineer these layers to understand how they play a role in disease and also correct that if something goes wrong? And that's the question that I'm going to talk about in the next session. So here's the general outline of the talk. First, I'm going to talk about how to develop genome engineering toolbox, and our goal here is to expand beyond editing. And the second is how can we use synthetic biology and genome engineering tools for enabling new type of gene therapy. And finally, amid this pandemic, I'd like to share our new method and the thoughts about how we can further develop CRISPR as an antiviral to target not only a single virus but broad spectrum of RNA viruses. And the first topic is about how we can develop a genome engineering toolbox. Beyond editing. As you see, a CRISPR is a molecular scissor. And the main focus here is how to turn a pair of scissors into something that can be a regulator. In 2013, our lab was first to turn the Cas9 nucleus into a nucleus-stabbed molecule termed dCas9. And shown here is we broke the scissors, And this was the beginning of the journey for many RNA-guided applications later. And we generally call it dCas, stands for dead. But what is that? It's nucleus is dead. And the concept applies to many RNA-guided DNA targeting applications. As you see that the Cas9 is a multi-domain molecule, and it has two nuclease domains that can specifically cut DNA upon binding to the DNA molecule. And the first step to generate a nuclease that Cas, so it doesn't cut but still bind, is to mutate these two nuclease domains out. And that was exactly what we did. Of course, there's always a question. If we modify a protein, it may destroy it. And to verify that removing the nuclease functions of the Cas9 and it's still a useful molecule to bind to DNA and guide it by RNA, we performed a series of experiments. And to our pleasure, when we test that dCas system with the nuclease domain knockout, we find it serves as a 20-base pair system to find DNA, pretty much like what it naturally does, but without cutting. And it can target many genes and in many organisms. And I want to give you an overview of the system. And as we continue to work on the developing tools beyond editing, together with other labs, we were able to demonstrate a full family of tools that can be leveraged for RNA-guided targeting or RNA targeting. The concept is using a specific dCas protein, dCas9, dCas12, dCas13 shown here, paired with a cognate target-specific guide RNA. And they can together target any type of cell and then cause specific binding to DNA or to mRNA. And then the game starts. Then we start to think, oh, how can we not modify this piece of DNA but still control the function out of it? And we have a series of development. A brief summary here is we were able to regulate transcription by turning on/off genes, how they interact with RNA polymerase shown on the first left side. And we were able to introduce desired epigenome chemical modifications such as demethylation into a chromatin. And we were able to modify how the DNA are organized inside the nucleus their positioning, and their location. And finally, we also showed this system can be turned into an imager. And a series of papers listed here, so in case you need any reference. I will do a brief introduction into these works just to give you a general picture how a single bioengineering molecule can be used for so many different purposes. And I always liken that to be a Swiss Army knife. I think it's not just scissors, it's a— it actually can do many functions depending on your application. For example, we first developed a technology called CRISPR activation and CRISPR inhibition, stands for CRISPR-A and CRISPR-I. And the basic concept is not to cut DNA, but to target a specific piece of DNA and recruit a very specific type of effector. For example, shown here, it could be a transcriptional repressor or a transcriptional activator. Depending on which effector you recruited to the target gene, we can exert different types of function on this gene. And both are usually the domain that shows very good efficiency. For example, a KRAB domain has been used to silence a target gene on the transcriptional level. And another tripartite inhibitor reported by other labs, VPR, can be used to strongly activate this gene. And with this, we essentially imagining, well, we can insert a lot of switches endogenously into the human genome. And by simply recruiting different effectors, we can turn on and turn off different fragments of the gene. And that should bring tremendous power of our engineering on the genome without changing this DNA sequence. And we are further intrigued by this multiplex capability. For example, the human genome is not only consists of one gene. Instead, there are tens of thousands of genes there. And it's usually the interplay between these genes that cause a function. And we think, oh, how can we activate and repress genes differently. So for example, published in 2015, we developed a system to simultaneously activate a receptor gene and turn off another gene that has a function in the disease, just simply by using an orthogonal pair of RNA-binding proteins, which recruits either activator or repressor. I'm not going to show further data. But please trust me, we were able to simultaneously activate many genes Currently, we try to change that number and we can robustly turn on more than 10 genes in a single cell. I'd like to highlight, besides transcription, this system is powerful for epigenetics. What is epigenetics? It means we are rewriting the chromatin status by demethylation or by histone modifications. We are not modifying the regulation. But we are changing the timing on this regulation. And people call it epigenetic memory. For example, if you think about a transcriptional effector, what it does is to prevent the interaction between an RNA polymerase with DNA. But that effect can be very transient. When you have the effector, it turns off the transcription. But when you no longer have the effector, the transcription comes back. And compared to that, epigenetic modification has a persistent long-term effect, because it not only excludes polymerase, but also it permanently changes the status of the DNA so it no longer interacts with polymerase. And shown here is an example of how we can use a transcriptional repressor called dCas9-CRAB by transiently introducing them into cells, we can turn off the cells. And rather than doing that, if we recruit three domains consisting of transcriptional effector and the two demethyltransferase called DMT3 and DMT3L, and even after this effector is decayed, we still see persistent repression. And this repression lasts for months. And in our experiment, we tested 3 months. It was a hard experiment for the student to perform it. And not only that, this type of demethylation, which caused the resilience over a long time, can also be reversed by another epigenetic effector like acetyltransferase, p300, combined with an activator. And as well, we can see that after we apply a transgene activator, we can permanently activate the genes. And this work was done by a postdoc, Muniaki Nakamura, who continued to explore, well, a lot of phenotypes in human cells. For example, differentiation, cell fate commitment, and even aging are controlled by such type of epigenetic switches. And how can we unravel their functions? These tools could be a useful system. With these additions, we are happy to see that CRISPR tools developed for transcriptional and epigenetic regulation However, one remaining question is how can we control the three-dimensional DNA organization in cells? To fulfill this goal, we developed two technologies in recent years. One is called CRISPR Genome Organization, CRISPR-GO, for perturbing how DNA are organized. And second is called live-FISH, for live cell fluorescent in situ hybridization, And using these two methods, we can both perturb and visualize the organization of the DNA in human cells. A few years ago, a postdoc, Haifeng Wang, who is now assistant professor in Tsinghua University, asked me how we can harness CRISPR for relocating DNA in the nucleus. We conceived one design using a chemically inducible system called abscisic acid inducible PyL1 and eBi system. And further, we fuse the ABI to a dCas9 molecule, which goes to the DNA side, and we fuse the PRL1 to another protein called emerin, which is an integral protein of the membrane of our nucleus. And this is equivalent to a mail system. For example, dCas9 now is like a dress label tagged to the DNA, and emerin is like a destination to tell where this shipment should go. And adding the small molecule, we hypothesise this can bring the DNA locally to the nuclear membrane and the wash away will reverse the process. And this was indeed what we observed. And Hai Feng tested the design on many chromosome sites, as you can see here, chromosome 7, X, the exist gene, very important gene, and p10 oncogene. And in all cases, adding the inducer abscisic acid was able to recruit the target DNA to the nuclear membrane, while wash away the abscisic acid reversed that. And mostly the DNA is inside the nucleus. I want to show you some images to convince you I didn't cherry-pick a few images just to show you it's not working robustly. Okay. So we actually tried methods to recruit the human telomeres. You know, telomeres are important control elements on the ends of our genome. And as we get older, the telomere gets shorter. But normally, they're not on the nuclear membrane, as you can see from this 3-dimensional reconstruction. The telomeres on a flat cell, mostly inside the cells, but not on the surface, labeled by green, and telomeres in red. Okay, in the second movie, I'd like to show you after adding the abscisic acid, most telomeres go to the envelope. As you can see from the two-dimensional, they are on the edge, but some seem to be in the middle. But actually, if you look at from the three-dimensional, they are either on the top or on the bottom. So it's really interesting to see most of these telomeres after using the CRISPR-GO recruitment, can be recruited to the nuclear membrane. How many of them? Well, they're like every chromosome has two. There are more than 100 because it's an aneuploid cell line, cancer cell. And we not only find that, we also find if we recruit telomeres to the nuclear membrane, we can dramatically reduce the cell proliferation. Keep in mind, these are cancer cells. They can grow very robustly. But since if we put telomere in the wrong position, it really kills the cell, which is interesting and will be interesting for future studies. And not only that, if you think about the human nucleus, it's a mansion. It has many, many rooms. And this is a beautiful picture depicting many compartments. And there are a lot of questions remaining. Why the nucleus is so complicated? Why is it not just a soup? And for example, we further tested how the DNA can be recruited to other bodies circled here, like Cajal bodies. PML body, and the nuclear speckles. And we were able to recruit many of them. But I'd like to highlight one particular compartment called heterochromatin. For people who studied the chromatin, you know, DNA can be organised into two types of forms. One is the open form called euchromatin. Another is the closed form called heterochromatin. And heterochromatin is highlighted by this particular protein called HP1-alpha. HP stands for heterochromatin protein. This protein is super interesting. It has a particular organization. It has a chromoshadow domain, and also it has a lot of hinge regions. Based on that, the chromoshadow domain can usually dimerize, and also the protein can also oligomerize via its hinge region. Consider this, it's not just a single protein. It can form a condensation if it's reached high enough number. We actually use the CRISPR-GO to recruit that to dCas9 to the DNA. What we did is we recruit a lot of copies of the HP1a to the human chromatin tiled along a chromatin region. We can make this protein reach about a certain threshold. When we got that happened, we actually did see some very interesting things because it starts to mimic the natural human endogenous heterochromatin. A few interesting behaviors can be observed. For example, when we try to recruit these two different chromatin regions separated by several hundred kilobases, and because this system can form a condensation, it can bring together the two DNA into one location. So it can modulate the DADN interaction even though they are far away. The second is when we try to recruit the synthetic heterochromatin on the target DNA, we find the nearby genes can be silenced. This is very consistent with the natural function of heterochromatin, but we were surprised to see even genes that is half megabase away from the target site, were still suppressed. The second function we saw is this heterochromatin, synthetic heterochromatin can compact DNA. For example, this is a real microscope image. Naturally, the chromatin can look more diffused. After adding the small molecule, we start to condensate HP1-alpha on this piece of DNA, and HP1-alpha start to form aggregation and bring together this into a more compact form. And after we wash away this abscisic acid, it reverts back into its relaxed state. But the reversal process is much slower than the recruitment process, suggesting a memory that was created by this heterochromatin protein 1-alpha. Okay, as a last example, I want to point you to CRISPR as a chromatin imager. Besides using that as a regulator, we use CRISPR as a labeler to look into how we can label DNA rather than control them. We chemically synthesized DNA and delivered Cas9 protein in the complex into the human cells. Here I want to mention one interesting observation, which is how How do we know the guide RNA is very stable? Well, the guide RNA in cells is very unstable, to be honest. It's only stable if it binds to the on-target, but it can be very unstable if it's floating around without a target. So that actually inspired us to think, oh, well, this could be bad news for editing, because you cannot get a rich, high concentration, but this is good news for imaging, only the on-target signal will persist while the off-target signal will decay. And by calculating the signal-to-noise ratio, which is on-target to off-target, we actually saw this method can improve the ratio by more than tenfold. This further allowed us to perform very clear, very easy diagnostics on the aneuploid cells. For example, we got these patient samples, from normal patient amniotic fluid cells. And this should be 2 copies, right? Sometimes 4 copies if cells dividing. And indeed, you can see how strong the signal-to-noise ratio is. And when we take samples from the Patau syndrome, a rare disease which patients have 3 copies of the chromosome 13, we indeed saw 3 copies. And keep in mind, this is not by fixing cells. This was performed in living cells. The samples were live tissues from the patients and can still be used for many other functional tests. And without further getting into it about imaging, because I have a more exciting story to share next, I'd like to give you a brief summary. And our work in the past years has been thinking how we can expand beyond editing. We focus on epigenome. We made several tools for transcription and epigenetic modifications. We also made some imaging tool and the control tool to control the 3D genome. And we think these tools could be used broadly to study the general picture of the genome and how that leads to more complex disease. The second part is our recent work and also some many unpublished work. The topic is to develop genome engineering, including regulation and editing, towards in vivo research. I'm sure as a biomedical researcher here, many of you were curious how that could be and if whether like CRISPR is good. And in particular, we are interested in genetic, epigenetic type of gene therapy. And forgive me if I'm too optimistic about the future of the genome engineering, which I called a bright blue sky. But there are definitely some clouds that shadow the sky. There are a few of them. I'm showing two. One is, how can we make CRISPR tools work in vivo? They are not as effective as usually shown for ex vivo, which actually already observed by many researchers and physicians. And the second is, current applications of CRISPR often orientate towards monogenic disease. And to our interest, we think, well, with great tools, there may be more applications that were not imagined before. How can we expand these therapeutics? And so since a few years ago, we start to ask how to make genome engineering suitable for in vivo research. I'm going to give two examples, and both recent works on how we address this bottleneck via bioengineering. The first example is to engineer a mini-sized CRISPR system termed CasMINI. We try to address the payload size of the CRISPR, which is a major bottleneck in the field of genome engineering. Mostly commonly used Cas9 or Cas12 family, I'm listing a few here, have sizes larger than 1,000 amino acids. That is a fairly large protein. Which is not very helpful for using a clinically relevant adenovirus for delivery because AAV usually have limited payload size of 4.7 kb. And also, with more and more applications of the mRNA and lipid nanoparticle delivery, it's becoming clear that these LNPs deliver smaller payloads better. So, it's imperative to define CRISPR systems that is as small as possible, yet as effective as possible? And to answer this question, Xiaoxue Xu, who is a postdoc in my lab, started to look into this question. And she first looked at the gold standard gene editor named Cas9 here. As you see, the structure, which I also showed before, is a large protein. It has 1,368 amino acids. And this size restricts it from easy delivery. In 2018, excitingly, Dr. Jennifer Doudna's lab reported a characterization of a naturally occurring smaller CRISPR system. At the moment, they termed Cas14, and later the system was also termed Cas12f from phylogenetic profiling. And it remained unclear if this Cas14 system can be used for gene editing in human cells. But just look at its size, 529 amino acids, which is more than half smaller than the Cas9. More than that, intriguingly, like Cas9 is a single monomer binding to DNA, Cas14 form a dimer and together they bind to a piece of DNA. That could also explain why its size is so small, yet it's still functional. But since it were not known if the system could work in human, we set out to test the system. So Xiaoshu first generated a nuclease-dead version following our previous rationale to see if we fuse this dCas12F by mutating the nuclease domains to an activator, can we really bind to DNA in human cell and activate an easy gene like a reporter GFP? And to our disappointment, it didn't work. As you see from the diagram with a non-targeting guide, No GFP, but with a targeting guide, we also saw 0% of cells could be activated. That seems to be insufficient for gene activation in human cells. It's known that many CRISPRs are evolved in bacterial context, and when we move them to the human context, they are not ready yet. If you think about that, it's actually quite reasonable because bacterial genome is 1,000 times smaller than human genome, and evolution only evolve sufficient function. They do not necessarily need to make them much better. Conceivably, if we can improve their targeting ability by 1,000 times, it should work. That's the simple math, may not be the case, but it shows you how this may become a rationale. We think, okay, if evolution has not done this yet, can we via engineering to make that work? First, we turn our attention to the guide RNA. And looking at this guide RNA secondary structure, we made a few changes. We eliminated some transcriptional terminators. We also modified at the interface between CRISPR RNA and the tracrRNA. And to our pleasure, it worked and improved. But to our disappointment, it didn't work too well. The best one only showed 3.6-fold improvement. It's helpful, but not sufficient. Then we turned our attention to protein engineering. Similar rationale, this protein has not been evolved yet good enough to target human DNA. And human DNA has histones, a lot of structure, and we need to make them better. So to do this, Joshua look at the structure. And by the way, I want to mention that by that moment when we work on this project, there was not yet a three-dimensional structure available. So what we did was we simulated using computational algorithms to predict what it might look like, how they might interact with DNA. And then we choose the amino acid within 1.5 nanometer distance, so fairly close to the DNA, and so they may interact with DNA. And furthermore, we made another hypothesis. To make this protein binding to DNA better, we need to increase the interaction by converting some negatively charged amino acids into positively charged. Why? Because DNA is negatively charged. If we can make the protein more positively charged, it may enhance the interaction and make it more efficient and faster. That's indeed what we did. If you look at this big panel of the mutations, we indeed find some mutations can dramatically enhance the activity. This was truly rewarding. And in particular, a single mutation at the position 143 was able to improve the efficiency by 123-fold. And many other mutations also have beneficial effects, suggesting the original hypothesis was correct. The protein was suboptimal, and we need human engineering to make them better. And furthermore, Xiaoshu combined these double, triple, quadruple mutations in multiple rounds. And for each round, she see better and better mutants. With better and better activation. And at the end of round 4, you can see some mutants already start to show more than 1,000-fold improvement over the original system, which was fully very exciting. Like I said, we need 1,000-fold improvement, right? So question is now, how about the system work in human cells? We actually also compare that to another wild-type system, Cas12a. And the question is, Cas12a is a gold standard, and it's actually known to be a very strong enzyme to activate. So, but it's much larger, it's actually 2 kB larger as shown here as an activator. And we choose them because Cas12a and Cas mini share the same PAM sequence, which is called protospacer adjacent motif, which is a sequence absolutely required for Cas molecule to bind, then we can design a base-pairing guide RNA to bind to a nearby region. So they can finally form a complex. As you can see from this data, when we try to activate a mammalian copy of GFP, the Cas mini outperformed Cas12a. We further tested on a few endogenous genes, and shown here is interferon gamma and the human beta globulin gene. And for each case, the magenta bars, the performance of Cas12a activator, and the blue bars is the performance of the Cas mini activator. My conclusion is they are comparable, which is good news. And you might also ask, well, you engineered a more efficient enzyme, but it may be less specific, right? That is a dilemma and a trade-off. And to verify it's still specific, we perform whole genome RNA sequencing in cells. We want to ask, besides the GFP gene, can we also non-specifically activate other genes as off-targets? We don't want to see that. So we measured all the genes and compare a targeting guide to a non-targeting guide. On the left is Cas mini activator, on the right is Cas12 activator. You can see they are fairly comparable. And also, this specificity is, to my experience, is very comparable to the dCas9 systems. So our conclusion is, yeah, specificity is good. So the next question is, how does this system work for gene editing? Given that Cas14 was really a new case, we asked whether Cas mini can be used for editing and whether it can overcome the limitations of the wild-type enzyme. We chose to mutate a few sites guided by the protein engineering. And we find using the improved system, we indeed can perform gene editing for insertion, deletion modifications called indels at different locations of the gene. And we also characterized what the editing pattern— we performed deep sequencing. And as you can see, some versions of the Cas mini showed a broader editing window on the target, which is useful for gene knockout. The Cas mini editing window in my mind is larger than Cas9. As you can see, even as big as 20 base pairs region, which is great news for knocking out a gene. And compared to that, Cas9 usually has 1 to 2 base pair editing window. And recently, people are excited about the base editing, as already described briefly, which converts a single base pair. To test whether the Cas mini can be used for base editing, we fuse the dCas mini to a pair of engineered and nozyme deaminase reported by David Liu's lab from Harvard. The fusion system showed robust single base pair A to G mutation at different genes at different sites. And further, we characterized which window can allow base editing A to G, and we saw a very narrow window that allowed that to happen. With the work on CasMINI is still going on as we try to demonstrate its use for in vivo applications, I'd like to show another example how we can combine by engineering for in vivo applications. The project is to develop a novel gene therapy for retinal regeneration. And Lucy Guo is an MD-PhD resident here. And when she came to my lab, she asked me a question. Stanley, can your lab use your tools to treat some complex disease like glaucoma? And she's an ophthalmologist and cares about glaucoma, which is an eye condition that damages optic nerves. Usually caused by high pressure, but also it's a leading cause of blindness for people over age of 60. As people develop glaucoma, it's a progressive phenotype. They start to see narrow and narrow and ultimately lose sight. The concept here is since it's not a single gene disease, it's due to a systemic problem. We need to turn on a few genes that's beneficial for the neuron regeneration and reprogramming. We need to choose a system to activate multiple genes. That was what she did. She chose these three genes, Oct4, Sox2, and the Klf4. These are Yamanaka factor genes. Biology told us if you turn on these genes, they are becoming more pluripotent. The previous work showed retina can also reprogram better to generate more neurons. Then Lucy generated a single construct encoding a CRISPR array and Cas12a. And try to activate these three genes. Well, you might ask, why do you choose a Cas12a activator? Why not a Cas9 activator? The reason is Cas12a is a superb system to activate many genes by simply processing a long CRISPR array into smaller fragments, while Cas9 cannot do that at all. So what Lucy did was injecting this into the newborn mice. After 3 weeks, harvest the retina and then try to see whether we can activate these three genes. And I'm not sure if you often read this retina immunostaining. And from DAPI staining, we can see all the cells, and the GFP is used as a marker for delivery into the different layers of the retina neurons. And the red color HA tag labels which cell actually can express the Cas12a system. And the KL4 is example for immunostaining expression level of this gene. As you can see, they are very poor colocalization. We were disappointed by that. And not only KL4, but for other genes, Alt4 and SAF2, we see almost no colocalization. That suggests there is a big gap between in vitro and in vivo applications. A good standard CRISPR system that works very nicely in vitro may work poorly in vivo. How can we fill the gap? Again, we turn to protein engineering. For this system, we chose again similar as CasMINI, we choose amino acids that is close to the DNA backbone within 1 nanometer. Also, we mutate most negatively charged amino acid into positive charge to further enhance its activity. As you see, we choose this size and our hypothesis is by making a stronger protein, it may reduce the required molecular concentration. And we believe that is one of the limiting factors in vivo, because we cannot deliver too many molecules in vivo into cells. And to our pleasure, this protein engineering was also rewarding. And really, a lot of CRISPR systems can be engineered, which is very interesting. And for example, one mutation can activate genes by 600-fold compared to a wild-type system only shows 16-fold. And by combining them, we saw more than 1900-fold activation. And this is more than 100-fold improvement. And Lucy likes this name, Very Good Cas12a. That's what she called this Bath mutant, and VG Cas12. I don't know if you like it, but that's her favorite name. Okay, so we further use this Very Good Cas12 to see, well, does that improve gene editing? And each mice has two eyes. And we inject a wild-type Cas12 and also a VG Cas12 into each eye, so we can rule out the genetic background. After the gene editing, we dissect the retina and stain for the marker, particularly look at this reporter in these transgenic mice. And we compared 5 mice side by side, left eye to right eye. As you can see, the yellow bars show the editing efficiency for wild-type and the blue bar shows so VG-Cas12, and the lower expression means stronger gene knockout. So in all cases, the very good one outperforms the wild type. The next question, how can we activate 3 genes altogether? Again, we use the same setup as before, and now look at how we can activate these genes. Look at HA gene here. Among the red cells, many of them also show standing for SOX2 codemodulation. Not only that, it also works for KR4 and ALT4. Very exciting. The engineered version works much, much better in vivo. And further, we look at what happens if we really activate SOX2, ALT4, and KR4. And when we compare this to controls, wild-type system which doesn't work, and another system called enhanced Cas12a developed by a different lab, we find a lot of new neurons regenerated inside the inner plexiform layer and also the ganglion cell layer. As you know, these two layers are most important in talking to the brain to confer vision. Generating neurons there likely indicate there is a neuronal regeneration. Furthermore, we also showed if we only activate a single gene using the very good Cas12, we cannot trigger the similar phenotype. The characterization shows with the VG-Cas12, about 25% on average neurons were regenerated and migrated to this ganglion cell layer. With this, I'd like to summarize this part, which is why I find this exciting. Because first, in vivo research is hard, and mostly people focus on a single gene. Here, we find a system that can allow us to study multi-genes. And it doesn't have gene damage because we turn gene activation. Of course, if you use active editing system, you can also modify genes permanently. And it has an easy delivery using either AAV or in vivo electroporation. And it may provide some solutions to starting more complex disease like glaucoma or retinitis pigmentosa, but also may applicable to other organs. I'd like to use the final 5 minutes to talk about our CRISPR antivirals for the broad-spectrum RNA viruses. And I hope this will inspire some conversation. And also, if you find it interesting, there could be further collaboration. Our question into this is, can we develop a universal antiviral that targets thousands of viruses? Just for your reference, Most vaccines or antivirals can only target very limited number of viruses, so they are not universal enough. When we look at the coronavirus, I was surprised to see, wow, there are so many coronaviruses and many of them infect animals. A few of them do infect humans. For example, besides SARS-CoV-2, we know SARS and MERS caused a big trouble for the human society in the past two decades. There are also other viruses, 229E, OC43 cause cold symptoms. It's very common, but it's there, right? They may evolve to become new dangers for us. But first of all, we think animal-inhabiting virus may become infectious to humans someday, and we do need a method to quickly respond to emerging virus and also to counter the rapid evolution of the virus. To do this, a student named Timsey Abbot started to tackle the question. He chose a specific CRISPR named Cas13, which is an RNA-guided RNA targeting system, and he imagined that could be a universal antiviral. And what does Cas13 do is it will use a guide RNA and then mirror base pairing to find its target RNA and then cut it because it's an RNA nuclease. The sixth protein is small, relatively small, not super small, but also fairly efficient and specific as shown by previous characterization. And the team can see the pipeline to be 3 steps. First, whenever there is an outbreak of the virus, in one day, we all know the sequence. And by bioinformatically aligning the sequence, we should know what sequence is concerned. And second is we predefined the— we call PAKMAN, stands for prophylactic antiviral CRISPR. We predefined PAKMAN compositions, in particular, the guide RNAs that can bind to the highly conserved regions of the virus and quickly verify them using human patient samples like organoids. And then we use predefined molecules combined with medical device like a nebulizer let the patient inhale into the airway, and in that, the molecules could enter the respiratory cell layer and get rid of the virus and cause their degradation. This would serve two roles. One is to greatly reduce the titer of the virus to reduce the lethality. Second is to buy us enough time so even a normal person can develop enough immunity to catch up in such an environment. And in particular, we look into targeting many viruses, coronaviruses. As you can see, the SARS-CoV-2 keep evolving, and we already have many variants here. And there are also other coronaviruses. They enter the cells using different receptors. Regardless, after they get into the cells, they release their genome RNA and become a replication machine copying machine to make many, many copies and ultimately destroy their target cell and release. And we think CRISPR is an interesting solution because it doesn't target the surface layer. Therefore, it can just become universal because it targets the inner layer because all the virus share a common mechanism. And particularly, we use the guide RNA to target the viral genome RNA, which also binds to the subgenomic mRNA to get rid of both protein production and also the genome RNA. It's actually two benefits there. We hope that would work. I want to show you some data here. We designed many guide RNA to target to a particular gene named nucleocapsid, which is highly conserved. We further defined lung epithelial cell culture delivered with CRISPR., and in fact, we source code two virus. After 1, 2, 3 days, we measure how much virus still there. We hope nothing. And after 1 day, we saw using a best guide, we see 96%. Using a paired guide, we see 97% inhibition, which is very efficient. This was the first test. Then we went on to test different variants of the coronavirus, of the SARS-CoV-2 here. As you can see, they have very consistent higher inhibition. I was really curious about the Delta virus strain, and that is a big trouble these days. Unfortunately, our collaborator, Catherine Bleach, here is still working on culturing them, but we will try to test on Delta virus soon. But also not only SARS-CoV-2, we also test other coronaviruses just using the same design. And again, we saw very strong inhibition, 97% using a single guide, or for OC43, we achieved even 99%. And I want to show you some— seeing is believing, right? So when we're not using a good guide, non-targeting guide, cells indeed died, because red cells mean cells with CRISPR Pac-Man, but green cells mean dead cells. And when we use a targeting guide, the cell grows dense. No dead cells showed up as in 4 or 5 days. This is all live cell imaging. Not only that, the CRISPR allowed preventing us against multi-infection. For example, when we use two virus to infect the same cell at the same time, we actually see very strong inhibition on both, and both showed 99.46%, very exciting, which I don't think any antiviral can easily do that because a lot of antiviral was highly specific. And furthermore, we combine the system with nebulizer, like generate a mist. What we did for ex vivo is we use this mist to get onto each well of the cells. So this mist only for 10 seconds, mimicking how we can breathe in this inhaler system. And further, we use patient-derived lung epithelial cells and differentiate into organelle, deliver them with virus, and then using LNP mRNA to treat that. And delivery efficiency into this epithelial culture is 62%. And we see a 92% reduction of viral infection after 3 days. And as a final data I'd like to share is there are many ongoing work on defining small molecules, but they target very different pathways from the CRISPR. CRISPR degrades RNA, but they may block entry or block replication. So we hypothesize these things may work together. Indeed, when we combine remdesivir analog, which is the analog for the nucleoside for RNA replication, when we combine them, while individual ones showed 100-fold or 200-fold inhibition, either by drug or by our Pac-Man, combining them shows almost 4 logs of the reduction. I talked with some virologists, and they believe 3-4 logs of reduction is important for initiating clinical trials. So we're also excited to see how synergistically they work together. And with this, I'd like to summarize my whole talk. I hope I explained the picture that CRISPR has many species, and we can harvest natural ones, we can also engineer new ones. And we can use them for many different functions, from transcription to epigenome, to imaging, to 3D genome, to antiviral, and to some other interesting developments. And this could fit into a big picture of the genome. And here's my favourite analogy. It's a Swiss Army knife. And we need a toolkit, not just a single tool. With this, I'd like to thank the lab who really contributed to the progress and data, and our nice collaborators over the years to support, supply knowledge that we didn't know at all. And also the funding from different agencies. At the end, I'd like to really thank Dr. Thomas Hader and Mrs. Hader for sponsoring this lecture, and this gives me an opportunity to talk about our work. Thank you so much.

Thank you so much, Dr. Qi, for such an excellent presentation. It certainly gives us a lot of hope for the future and it opened doors to much-needed therapeutics. I'd like to ask Dr. David Lowe, Senior Associate Dean for Research, to facilitate the discussion, our question-answer session.

Thank you, Dean Deese. So if people want to ask questions, you could raise your hand or put something into the chat, and I can really forward the questions here to Dr. Chee. Maybe let me start off with a question. Thank you for a great lecture. It was really fun to watch all these tools and applying them. So I had a kind of a technical question is that It was interesting to hear the test comparing in vitro versus in vivo application. And so I wondered whether or not one of the technical issues is like, if you were to try and enter this into more stably inherited or like even germline, could you potentially even make somebody genetically immune to every coronavirus? Is that— have you tested something like that, putting something into the germline?

No, we haven't done that. There is a big barrier for the ethical side about performing editing in the germ cells of the human. And as we still— there are a lot of things to understand how modifying a gene or a few genes will play a role in the long term of the human's life. And it's not the best medical process to do that. And furthermore, any mutations introduced could also be inherited by the offspring of the individual. So that also creates a much longer influence on the human society. As such, the current gene editing technologies mostly referring to somatic cell engineering because even In the worst case, something goes wrong, it only affects the individuals who receive this treatment, but it won't go too long into the later generations. I think we still need years and years of work to fully understand the performance and the mapping of every single step of this type of system in the human body, and also over decades of the tracking to see what's happening. Before we can commit to germline editing.

Thank you. That's a great, great point to make. Here's a question. What made you want to go into genome engineering?

I was— when I first became transferred from physics to biology, I was intrigued by DNA biology. And there are a lot of interesting knowledge, actually, genetics, molecular biology is about biology of DNA. However, when we talk about DNA inside the human cells, our knowledge is limited due to our inability to modify them directly in the human cells. A lot of previous work needs to be performed in simpler organisms like yeast, or extracting out, modify them, starting with the going on. So how can we really greatly accelerate the research by orders of magnitude requires such tools. And second reason is my personal choice of the synthetic biology. As my understanding of the synthetic biology is how we can engineer functions into cells and create novel functions and study the whole cell as a system. That requires us to not only modify a single gene, but also control the genome as a network, as a program. And that requires different ways to control them. That further inspired our interest to this dCas system to control it rather than adding it.

Thank you. So here's another question. So the engineered forms of these vector molecules, you have better the effector function in vivo and longer duration. Is there a potential that over the longer period that there could be complications from the longer-term effector activity?

In many cases, the effector is a transcriptional controller. Our observation is, the effect will dissipate over time. So it actually does not create super long effect, but that would be suitable for, let's say, stem cell differentiation because stem cell differentiation only occurs in a few weeks, doesn't need to be years. But in terms of epigenome editing, for example, permanently modifying a demethylation mark to silence a chromatin, Those may have long-term effects. As the field of epigenetics trying to answer the question like, oh yeah, what happens if we modify it over years and how that change the behavior of cells, how that change behavior of the organ? I think these tools should provide an entry point. I will be very interested to learn about that biology first. Before claiming everything, anything here.

Maybe I can ask a question in the opposite direction. Is that when you were describing what is essentially like a molecular immunization, what would be the potential benefit long-term? I mean, would we need a booster inhalation of a mist every year, every 6 months, you know, that sort of thing? What would be the persistence of the kind of molecular immunization?

I have to say that we imagine that the system is— it doesn't need to be used repeatedly again and again. It shouldn't become a daily use or weekly use. The reason is there is only a particular season of the infection that is very severe to the community. And the point is, the patient, the person is exposed to the viral risk is only restricted to a period of time. If we can protect that period, that's probably enough. Second is our human immunity will kick in. For example, when virus infect our body and in the early stage of infection, we use our inhaler system to get rid of the 99% of the viral load. That may buy us time for our immunity to develop, like antibodies can be generated. While the patient will not suffer too much from cytokine storm or from the bad damage from the viral infection. That is the concept here. In case of the flu, I didn't show the data here. We also tested on influenza A, certain flus. It also worked. And in those cases, since the flu is not that lethal, we think this smaller dose of this could become either prevention or become early-stage treatment.

Thanks. So you were talking earlier about reducing the size of the effector proteins. Is there a theoretical goal in terms of how truly small you can make the The Cas protein?

A GFP is the size of 1 kb, and many genes is around the size of 1 kb. I will be happier if the protein further become 1 kb. Of course, if it's even further smaller, sure, as long as the function is still there. Smaller always has some benefits, but the additional benefits introduced by becoming even smaller may dissipate quickly. Depending on the need. For example, our current imagination of the Cas mini is because the protein is only 1.5 kb, AAV has 4.7 kb size, we still have 3 kb size, and we can encode a lot of other things using the 3 kb. For example, putting additional gene or putting multiple guide RNAs to modify multiple genes. We do believe those are benefits, but if needed, maybe it's possible to make even smaller because there's an even smaller CRISPR, and also there are some other protein engineering can possibly do down there.

Not to the point where you can synthesize some small organic, you know, you know, cage or something like that. Anyway, so I know this is a, this is an ongoing question for everybody who uses this. It's like, is it, do you have a systematic way to design your guide RNAs?

When I first entered this CRISPR field, I admit that I overlooked the guide RNA design question. But the more I work in this field, the more I feel core to protein engineering is your guide design. Human gene has so— there are so many genes, right? And even on the same gene, there are so many sequences that 20-base pairs guide can bind to. And the number is just huge. But which guide would you use actually is the major question determining success versus failures, not only for editing, even more true for gene regulation. Because when we try to modify the epigenetic marks, demethylation, we need to target the enzyme to a region that cell cares about. For example, a region, if we methylate, will have a function. Not a region, even if you mess with it, doesn't matter. That requires us to understand the epigenome landscape of the gene, of the cell. For example, I will use this example. When people grow up, many people develop lactose intolerance because there is a gene for metabolizing lactose called epigenetic silenced. If you want to activate that, we cannot just design a random guide to work. Very likely it doesn't work. We need a method to target to this gene to a region that causes its silencing and to reverse its effect. This can be done by sequencing and other people's database. And also, there are emerging computational tools to help your design life easier. It's not 100% accurate, but it's better than just gambling. Thanks.

I see two related questions here. One about— Basically, they're about delivery systems. So one was about specialized approach for this nebulizer you showed that mist for lung delivery, but also let's say we're talking about delivering to tumor cells or other kinds of targets. Are you developing more specialized delivery systems for this kind of therapeutic?

Right. Not only us, but many labs are working on this, especially for delivery lab. For example, trying to engineer the surface protein of the AAV viral particle will steer its preference. For LNP, their preferred location is liver, and trying to de-target the liver to muscle, to brain may help their efficient delivery. And there is a second layer of the specificity, which is by using cell-type-specific control elements. For example, a promoter that will only be turned on in the right type of cell but not in the wrong type of cell. With those layers, we may also make sure even sometimes these particles go to a wrong type of cell, but they never expressing these gene editors, but only in the right type of cells, they will express the molecule and edit the genome. Activate the gene. So it will be interesting to continue to explore on that front.

Great, thank you. I'm sure there's a lot of different companies eager to find nice technological solutions to that question. So, well, I think we're kind of toward the end of our time here. I thank you so much for your great answers to the the thoughtful questions from our audience. I don't know if there's any other comments or questions. Oh, here's another. Oh, here's a quick one. Well, maybe I don't know how quick. So you did touch on this with some of the attacking the chromatin and so on.

Could you—

do you have any other comments on how your toolkit to interacts with the three-dimensional structure of the chromosomes, especially in this enhancer-promoter interaction, the folding of the chromosome and the activation of gene regulation of genes. Any more comments on that?

Yes. In terms of this, this is actually a really important layer of the regulation happening inside cells and also a unique power of CRISPR to target these enhancers and promoters. And for people who are not starting genetics, promoters control genes, but enhancers is like a fueling station for the promoter to work. These two need to come together to turn on the gene. And for us, we've been trying a few things. One is using the CRISPR silencing. We do find that simultaneously silencing both promoter and enhancer is the most effective way to silence the gene. If you only silence promoter, enhancer is still coming there, right? So it's still not completely silenced. But this is used as a technology. But in another case, we find that's become a method to probe the function of the enhancers. Because the fundamental question in enhancers is how do we know an enhancer is an enhancer? And DNA may come close together. But the they may not have the function. And by using the technology to probe them and measuring the outcome on gene regulation, we may indeed know not only these two genes come together, but also they really have a functional output. So that's something that I'm very excited about. And I want to comment that CRISPR can be used as a very large-scale technology to not only study one site at a time, it can be used to study tens of thousands of the genomic sites in parallel. Because designing guides and synthesizing guides is not a challenge, not a very challenging task these days. And in terms of studying enhancers, promoters, we do have work. I haven't included data to perturb tens of thousands of the sites for enhancers to probe which site is indeed an enhancer and what is their contribution. I hope to share the story in the future.

I guess the next step is to have artificial intelligence design a whole program out of that, right?

If we have enough data, yes. Machine learning and big data would be helpful to predict so that we'll save a lot of time and money to perform my lab experiments.

Yeah. Great. Well, thank you so much. This has really been a great discussion. Maybe I'll hand this back to Dean Deese or Dr. Heider to maybe thank you again and close out the session.

Again, Dr. Chee, thank you so much for an excellent presentation. It leaves us with a lot to think about, and I know that many of the biomedical science faculty are here, We're going to be contacting you and hopefully opening a path for some collaboration in the future. We look forward to having you here at UCR School of Medicine in person. Dr. Haider, would you like to say a few words before we end?

Sure. This was a wonderful talk. It wants me to go back to graduate school, Dr. Chee. This is such an exciting time and you gave it in such a concise way, in such an understandable way. It's a great topic but very difficult, but you made it so clear that I hope everybody that watched this understood it. I know I did and I'm sure from the way the presentation was that everybody connects with it in one way or another. Thank you so much.

Thank you, Dean Diaz, and also Dr. Thomas Heider. Yeah, it's actually our hope that some of these tools will someday no longer just be a research tool but also really change people's lives and bring hope to some challenging disease. And that actually is the ultimate goal, hopefully soon, but in the next 10 years, hopefully.

Well, thanks everyone for attending, and more to come next year. Have a good night.