Scientist Stories: Jeffrey Johnson, Studying Host Cellular Remodeling with Quantitative Proteomics

Well, thank you for coming, and thanks for everybody online for coming. It's my pleasure to introduce Jeff Johnson. Jeff got his PhD at Scripps here in John Yates' group, who's one of the leaders in mass spectrometry and data searching, which was a big thing. And then he went and did a postdoc at Genentech, and then a short postdoc at Genentech, and then went to Neville Krogan's lab up at UCSF, where it looks like he was the shortest postdoc I've ever seen. Followed by basically creating a whole fleet of the—

being the director.

He became the director of the scientific— the thermoscientific proteomics facility up there, which I worked with Jeff when he was at UCSF. Well, I worked alongside Jeff when he was at UCSF, and we went from him borrowing our instrument on occasion to building an entire platform. So having done that, having made a small platform recently, I know how hard that is. He's now an associate professor at the Icahn School of Medicine at Mount Sinai.. And so although he's definitely a card-carrying mass spectrometrist, one of the reasons I like his work is that it's not just mass spectrometry for the sake of mass spectrometry. He's actually asking important questions in immunology and cell biology, and I don't think that's as common as I would like it to be. So anyway, Jeff, thank you for being here.

Thank you. Um, can everybody hear me okay on Zoom? I'm unmuted. And can you all hear me okay? Is there— there's no— okay. Yeah, so actually this is my first invited talk in person. I started my lab in August of 2019, so yeah, I'm excited to actually get out of New York City and come visit somewhere. So my lab's interested in in host-pathogen interactions. This is not advancing. Why? There we go. So we're interested in host-virus interactions and specifically in how viruses remodel host cellular pathways for their own benefit. And where we stand today, one of the best tools for studying pathways in a sort of relatively unbiased comprehensive manner is with mass spectrometry-based proteomics. Which is a big part of the skills that I have and what we do in the lab, of course. And we use mass spec to quantify post-translational modifications, protein-protein interactions, and we're starting to, with Sam's help, look at more protein and nucleic acid interactions. And we can sort of estimate what pathways are doing based on all of these data. And so what we generally end up with are these really biologically rich datasets on multidimensional proteins, phosphorylation sites, ubiquitination sites, sometimes there's RNA-seq in there, sometimes there's metabolomics in there. And we use all sorts of bioinformatics methods to look at which pathways are being modulated within that data and then turn to more traditional molecular biology, virology approaches to validate our pathways, to sort of mechanistically define how these pathways get regulated and— whoops— how those pathways are being regulated by viruses and why do viruses regulate those pathways, with the eventual goal of can you identify drug targets that can sort of prevent a virus from being able to remodel a host pathway that it needs to do in order to successfully infect a cell. So I'm going to talk about two different viruses today. Chronologically, HIV was what I worked on when I was at UCSF, and I sort of carried that forward to my lab. But my lab was shut down 6 months after I started, so we had to sort of pivot to SARS-CoV-2 research. And I'll talk a little bit about that and then what we're getting back to doing on HIV that I wanted to do when I first started. So I'll start with the HIV stuff, and I'm going I'm going to talk about this protein phosphatase that gets degraded by HIV B56. So one of the ways that HIV remodels the host cell is that it interacts with host ubiquitin ligase machinery. And HIV has 3 different proteins that bind to the same family of host ubiquitin ligases, and they're called the co-LIN family. So what I'm showing you here is the HIV VIF protein. Protein that binds to the host colon 5 ubiquitin ligase, and these ligases have all sorts of adapters and it has an E2 and a substrate receptor. It's a big multiprotein complex, but I'm just going to call it colon 5 for simplicity. And what we know about VIF, and this is common for many of the other genes I'm going to describe, is that VIF pulls in a host protein called ApoBec3, ApoBec3G usually. And APOBEC3 proteins have really strong antiviral activity against HIV. And so HIV gets around that by encoding VIF, which pulls APOBEC into this ubiquitin ligase complex and it forces ubiquitin onto the APOBEC and targets it for degradation by the proteasome. And that's how it sort of overcomes this APOBEC restriction. That's a very common theme. The VPU protein of HIV does a very similar thing for a protein called BST2 or tetherin. Same thing, it's an antiviral protein. It holds onto viruses as they're trying to bud off of the cell. So HIV gets around it by pulling it into ubiquitin ligase complex, forcing ubiquitin onto it, and removing it from the cell surface, neutralizing its activity. And same thing for VPR. HIV-2 and older lentiviruses have an additional protein called VPX to sort of get idea here. This is the point of contact that HIV makes to make certain remodeling arrangements to the host cell. And what I'm going to talk about here is an additional protein that came up recently, which is a VIF substrate called PP2AB56, and it's a family of proteins that similarly gets ubiquitinylated and degraded by VIF. It's something that I had found in my data and I got scooped on, and it delayed my publication by about 10 10 years, but we eventually published something and I work on it still today, partially out of spite, but also because I think it's interesting. There's interesting biology there. So protein phosphatase 2A, phosphatases remove phosphorylation sites from proteins as opposed to kinases that put them on, right? So PP2A is a heterotrimer, has a catalytic subunit structural subunit C, and it has this regulatory B subunit. The A and C subunits are constant. There's one gene that encodes each of them. And then the B subunit swaps in and out for about 60 different options. There's 5 different families of B subunits, and this B56 family that gets degraded by HIV has 5 genes within that family: alpha, beta, delta, gamma. Epsilon. And so the B subunit is really what dictates PP2A activity, PP2A localization, PP2A substrate specificity. It's sort of modulating PP2A activity. And the HIV VIF protein degrades them all. So here I'm showing each of the B56 subunits, and for gamma there's two isoforms that are common, so we did both of them. You can see that as you increase the amount of VIF that you express, you get decreasing amounts of B56. And this is specific for the VIF protein. If you infect cells with HIV and you take away VIF, B56 is no longer degraded, whereas if you take away other HIV proteins, you still get that degradation. So just to compare what we know about APOBEC3 degradation and PP2A/B56 degradation, So like I said, ApoBec3 is ubiquitinylated and degraded using this COLN5 complex, and it's a direct interaction with that complex. ApoBec3 is an interferon-stimulated gene, has very clear antiviral activity. PP2A, on the other hand, what's in common? We know it's ubiquitinylated and degraded by VIF. We know that it requires the same complex, but a direct interaction has never been shown.. But most mysteriously, it's not interferon stimulated. If it has an antiviral activity, it's not clear exactly what it is. And it's not clear if there is a fitness advantage to degrading this protein at all. And so that's kind of an open question in the field and one that we're trying to address. What makes this particularly difficult is that it's been shown that the interface that VIF uses to degrade these two different families of proteins is the same. So genetically, it's really hard to separate them. If you take VIF out of the virus, it will not replicate because APOBEC3 will restrict its growth. So it's really difficult to tease apart how important is B56 degradation, does it provide a fitness advantage? Is B56 specifically expressed in T cells or is it broadly expressed? It's everywhere. So the different subunits are expressed in different places, but most of them are required for mitosis and cell cycle, so some of them are everywhere. Thanks. So one way of trying to get at this fitness advantage question is you can look at different lentiviruses. We're not the first animals to get infected by lentiviruses, so back when I was at UCSF, there was a grad student, Josh Keen, who did a project where he was looking at protein interactions of different VIF proteins from Mehdi-Visna virus, which infects sheep, bovine infectious virus, feline, simian, and human HIV. And he cloned each of these VIF alleles and he expressed them in Jurkat. This is a human T-cell line. And we did our global ubiquitin mass spec quantification on these cells. And Joshua was looking at protein interactions. Actually, this data never got published. We had it in the original publication and they wanted more experiments, so we took it out, which is a shame. But what we found was that, as expected, this is the gamma subunit that I'm showing you here. HIV-1 ubiquitinates the gamma subunit, but it also ubiquitinates pretty extensively. Maydivisnavirus is— or sorry, Maydivisnavirus VIF ubiquitinates the gamma subunit of B56. In terms of evolution, Maydivisna is the furthest away from HIV, which suggests that it is a very conserved function. And keep in mind, we're expressing these in human cells, so that we don't see ubiquitination with these probably just means that you need to have the bovine B56 or the cat B56 in order to see this degradation. But this was actually confirmed earlier this year by another lab that was looking at B56-alpha it was fluorescently tagged, so what you're looking at are FACS flow cytometry histograms. You can see that for Mady Wisna, when you add VIF, you get this shift to the left of the protein being degraded for HIV and for Mady Wisna on the right here. So I think that's the clearest indication that we have so far that this is actually important for the virus. You know, these retroviruses, lentiviruses, have extremely compact genomes, they have overlapping open reading frames. If they don't need to do something, they probably wouldn't keep doing it for many millions of years. So I think this does have a function, and we're trying to understand why does HIV target this thing for degradation. So I'm a fossil proteomics guy, so it kind of makes sense that the next question If you can't really tease it apart genetically or functionally very easily, maybe one way of approaching this question is by looking at what are the proteins that B56 is dephosphorylating and why would those be so detrimental to HIV that it would want to degrade it and keep these things phosphorylated, right? And so we had done a large post-translational modification study of HIV and we had looked at a bunch of different mutants where we deleted VIF or VPR or VPU compared to mock, and we added a couple more conditions where we took a delta VIF virus, which does not degrade B56, and we inhibited it with a classic PP2A inhibitor called ocazolic acid, and we used our global phosphoproteomics method to quantify relative to the wild-type virus, how do each of these viruses What are the phosphorylation sites that change for each of these different viruses? And so we get this dataset that has, you know, in total it was something like 14,000 phosphorylation sites. Of those 14,000, there are about 1,400 that were changing significantly in one of the comparisons that we were interested in. And we wanted to sort of identify the B56 sites substrates in the context of our biological system. I'm very much a fan of an unbiased approach. Like, there is some data known about B56, but we don't know if that's what's going to control HIV infection. So based on what we know about B56, like I said, it is involved in mitosis and the cell cycle, and it's mutated in cancer. There are quite a number of substrates that are validated B56 substrates that we detected in our study. And this is what that data looks like. So we can sort of use that to guide cutoffs or to sort of filter that dataset of 1,400 phosphorylation sites to figure out which ones are most likely to be B56 substrates. It's not— this is not the most brilliant procedure. I tried a few different things, but it's such a small sort of training set that like something fancy like linear regression or You can't train an AI based on that, right? So we're just using really modest cutoffs to say like minimally what could be a B56 substrate then. So with those filters, we get about 125 phosphorylation sites that are behaving pretty similar to what the known B56 sites are. And this is a point where I get stuck in my research a lot and that Sam and I were just talking about and you're gonna see it again for the SARS thing, is there's this point where you want to be able to functionally test these phosphorylation sites to figure out which ones are important for HIV infection, right? And the current technology is just not there. And so we're trying to explore different ways to sort of systematically test these phosphorylation sites. And I'll talk a little bit about what we see as the future of that, but for now, 125 sites is just too many too many for us to test. So we looked to the PhosphoSitePlus resource, which has annotations of phosphorylation sites and what their known functions are based on the literature. And so you can narrow it down to 11 phosphorylation sites that look like they're PP2A-B56 substrates and that have some annotated function within this database. And keep in mind, I hate this, like you're introducing bias into the system here. You're relying on whatever is in this database, whatever has been published, whatever biological context of the cell cycle or mitosis people have studied before. But realistically, we just can't test that many sites. So this is how we get to our 11 sites. And these are those 11 sites. And you can see that many of them are involved in the cell cycle. Many of them regulate protein-protein interactions. With things like 14-3-3 or HP1-alpha. And so we did screen these. So we screened them in primary CD4 T cells, not the phosphorylation sites, we just knocked out the whole protein and asked like which of these proteins is important for HIV replication in primary CD4 T cells. And the hit that came out of this and that got us kind of excited was a protein called TRIM28. Which is also called CAP1, which is also called TIF1-beta. And in this system, it's going to be a little deceptive here. So these are primary CD4 T cells. They're electroporated with recombinant Cas9 with its guide RNA and tracer RNA. And when you knock out CD4, you get no infection. CD4 is the receptor for HIV. And it's a little deceptive because most host factors that are required for HIV infection don't kill the virus that much. Most of them are in the middle here, things like that we know, like Ledger for CPSF5 or 6, like those are sort of intermediate phenotypes, and the non-targeting controls are up here. It's also hard because HIV sort of plateaus at 10% infection and then tanks in culture. I don't know why, but it's not the best system, but this was sort of the best indicator that we could have to say, okay, maybe this one's worth following up on. And the other reason we got really excited about it is TRIM28 has defined roles in silencing endogenous retro elements and sometimes exogenous retroviruses. And it does so by sort of bringing together KRAB domain-containing zinc finger proteins, which usually recognize sequence elements in endogenous retro elements., and it brings in chromatin silencing factors like SETDB1 that add certain histone marks that silence that region of chromatin. So it kind of made a lot of sense why TRIM28 would be regulated in some way by phosphorylation, and that particular phosphorylation site in the PhosphoSitePlus resource indicates that it breaks the connection between TRIM28 and the chromatin silencing factor. So it made a lot of sense, but again, you're so biased by what you know that— well, I'll show you what happens, and it's not pretty. So we did the experiment, which is sort of like the horrible low-throughput way of testing a phosphorylation site, which is we first made these cell lines that are expressing TRIM28 cDNAs, and we mutated the serine to an alanine. There's another serine residue that gets phosphorylated in TRIM28 that's very important that we don't detect by mass spec, so we just threw that in there just in case and made the double mutant. Ideally, you'd also want to make the phosphomimetic mutants where you mutate serine to glutamate or aspartate, and we did that, but they didn't express as much, and maybe that's the phenotype, but we couldn't get the same level of expression of TRIM28 with those phosphomimetic mutants, so we didn't screen them in the infection assay. So we had these 5 cell lines and then we actually screened these, I saw no phenotype, but there was still endogenous wild-type TRIM28 there. So then I went and I probably made it more complex than I needed to, which I usually do. So I knocked out TRIM28, knocked out CD4, knocked out, was a non-targeting control, in each of those different cell lines and then infected those cell lines. So now they don't have their endogenous TRIM28, they only have their mutant, and it's dox inducible, so you can turn the mutants on or off. And then we take these cell lines, I infect them with a GFP reporter HIV virus, and essentially no difference at all. There's some differences between these cell lines which could be just clonal differences between the cell lines, but It's a massive amount of work to get this kind of result. I was nonplussed. And so it sort of highlighted some of the drawbacks of this approach for functionally testing phosphorylation sites. It's too low throughput and there's too many caveats to it. One, you have to deplete the endogenous protein. That means you have to be able to knock it out, your CRISPR guide has to go to where it has to go and has to cut the DNA or whatever. You have to express these cDNAs, you have to make these mutants. You're kind of limited to cell lines if you're doing all of this engineering work. You have artificial expression of your cDNAs. These phosphomutations are artificial as well. They might affect the structure of the protein itself. It's just too low throughput. I don't ever want to do that again. And so we've been looking into— these CRISPR-Cas9 base editing approaches to introduce mutations into different phosphorylation sites. And I'm always surprised by— some people know a lot about it, Sam knows everything about it. A lot of people have never heard of it. So I'll just describe it. And so the idea here is you have a Cas9, this is a nickase mutant, so it doesn't induce a double-strand break, it's just a single strand on one side that repairs itself. And it's fused to a nucleotide deaminase enzyme. And these editors, they've been sort of developed in different iterations, but the editors that we have today are really good at defining about a 5-nucleotide editing window. Sorry. Within— no worries. And this edit window is very defined in how far away it is from the protospacer adjacent motif. And so if you're lucky and you can find a PAM that's the right distance away from your editing window, and if you're lucky and you can induce something that— a mutation that a nucleotide deaminase can do, you can mutate a serine to another residue, right? And so the mutations that you're able to do are basically purine to purine or pyrimidine to pyrimidine. Trimming for now. People are trying to work on transversions, but I think they're pretty inefficient. But in effect, what you can do with these editors, you can convert a C to a T or an A to a G, and you can do it on either strand. So you could also do vice versa. You just can't go across that way. And so for TRIM28, for example, the one that we've been able to edit in a 293 cell. [LAUGHTER] The serine codon is encoded by a TCA, and you can, with a C to T mutation, you can mutate it to a leucine residue, which makes it TTA. And so we did this using the cytosine, cytidine base editors, this BE4 Max. We're using NG editors, which are somewhat relaxed PAM requirements, so that you have more options for the guide RNAs that put you in the right distance from the editing window. And then this is Sanger sequencing data, where we amplified in genomic DNA the region of TRIM28. And you can see in untransfected or no guide RNA transfected, this is all dominated by the C here. And then with two different guide RNAs, you have this little blip where you got that mutation to a T. Probably efficient enough in a cell line to sort of dilute and single-cell clone out the mutants that contain that mutation. But what we're hoping to do is to sort of develop that into more of a screening platform so you can, in high throughput, iterate through all of the phosphorylation sites of interest and test them in a pooled format or in an arrayed format or something easier. Even developing a whole new screening platform would be easier than doing that stupid approach I did before. So that's sort of where we stand with HIV. We're a little stuck, to be honest, but that's sort of the reality of the situation. So to summarize where we're at with that, we know that B56 is degraded by VIF. We know that it's conserved across distant lentiviruses, and we are working on cloning the B56s from these different species to see if that actually We can show that B56 degradation is conserved in all of those. We know that B56 induces widespread changes to the phosphoproteome. What I haven't mentioned is that B56 degradation leads to cell cycle arrest by activating the DNA damage response, and so it actually has a huge effect on the phosphoproteome. And I think a lot of times you're always looking for— you have a list, for me I have a list of 125 or 11 phosphorylation sites, and you're always looking for the one that drives your phenotype, and that's always what reviewers want, I can tell you that. But a more realistic thing is that it may not ever be that. It may be that you have 100 sites, maybe 60 are beneficial to the virus and 40 are detrimental, and that you've evolved to degrade that thing because on balance you provide a fitness advantage. And, again, I think being able to, in more high-throughput, iterate through all those things and test them, to understand how they contribute to that, you can get a better idea of is it one or a few number of substrates or is it like a big balanced effect.

May I ask a question?

Yeah.

It's like as the phosphorylation is very specific to amino acids like tyrosine or threonine as you have seen here, but is this the dephosphorylation as easy Is it specific to particular B phosphorylation at particular amino acid?

PP2A? Yeah. PP2A is a serine-threonine phosphatase. I don't know that it removes tyrosine phosphorylation, but I would assume probably not. Those are usually encoded by different enzymes. Okay, I have one question.

Does HIV have any proteins which is phosphorylated?

It does, yeah, and actually it's kind of surprising. There's a protein— so many of them are phosphorylated, although it's not clear if those phosphorylation sites are required for infection in any way. We haven't been able to show that. There's a lot known about B56, how it binds to its substrates. There's a sequence motif, it's LXXIXE, and HIV matrix protein contains that domain. We mutated that domain and you can't make virus with it, so it's actually structurally important for the virus. But it's possible that there's some sort of viral phosphorylation, dephosphorylation that's happening that's involving PP2A. VIF gets expressed late during infection, so I haven't shown the life cycle, but HIV infects the cell, reverse transcribes, integrates into the genome. That's sort of the early phase., and then it makes its RNA, makes its protein, and that's the late phase. So VIF only gets made after integration, so it's possible that there's some B56 event in early infection that's important, but that you have to turn that thing off late during infection. So we're always trying to think about that in terms of the viral proteins, like you said. There's something going on there, but it's not clear exactly what.

I wanted to ask you about this result with TRIM28 in stem cells that lead to arrest in proliferation.

In your gut cells, that also happened? An arrest in proliferation? Yes. With the TRIM28 mutation? Yes. I mean, I didn't see that. I guess— well, no, because I grew up the cells for the assay, so they were still growing. But the knockout also— The knockout grew fine for me. In different cell lines, we've seen different things. In some cell lines, when you knock out TRIM28, you get interferon activation, and that sort of slows things down, which could be proliferation, or it could just be they're not happy with the interferon. In my Jurkat cells, when I knocked it out— well, I intentionally put the cDNAs in first and then knocked out the endogenous second, so I always had a TRIM28 there in my TURCAT lines. Whereas if you're starting with the cell line, you take TRIM28 away first, I don't know how that would behave. Yeah, TRIM28 does a lot of things. Every time I talk about it, people are like, "Oh, I know TRIM28." I'm like, "Well, it doesn't do anything for AGV." All right, so I'm going to move on to SARS-CoV-2. Then, and I'll talk about P38 MAP kinase pathway and why SARS-CoV-2 loves it so much. So like I said, when I started my lab, 6 months in, the pandemic came, and I didn't have a mass spectrometer at that time, so I couldn't really do a whole lot. But I was very fortunate that my former mentor, Nevan Krogan, sort of built together this big collaborative team to sort of use the tools, the phosphoproteomics tools that I had developed at UCSF and sort of repurpose them for SARS-CoV-2. And so I didn't generate the data for this. I didn't do the infections for this. What I did was help with the data analysis and then we sort of picked our pathway of interest and followed up on it and then spun it off into our own story. But the point I'm trying to make here is this was a huge effort with many, many different people. And I, my part was actually pretty small for the first study. So what this study was, this was very early in the pandemic. We didn't know ACE2 was the receptor. We didn't have ACE2 expressing cells. So we did this in Vero cells. I'm very defensive about it, 'cause like they're the worst cell line to work in. They don't have an interferon system competent, but this is what we had. So we infected, zero cells with SARS-CoV-2 in a time course biological triplicate, and we did global phosphoproteomics and global protein abundance analysis. And we have a bioinformatics method based on what we know about kinase substrate interactions. You can estimate kinases that are up or downregulated based on phosphoproteomics data. And so we found that a number of kinases that regulate activation of the p38 pathway are being upregulated during SARS-CoV-2 infection. It's not totally surprising. A lot of viruses turn on MAP kinase pathways. They're very important for cytokines and inflammation. And a lot of times we think about cytokine pathways as being the JAK-STAT pathway. But if you go backwards over evolutionary time, MAP kinases were some of the earliest kinases that transduced signals from cell surface receptors down to transcription factors and their effectors. And sort of the JAK/STAT pathway got written on top of that pathway. Those two pathways interact quite a lot more than I think most of us appreciate. So MAP kinase pathways are split into sort of four parallel pathways. There probably are some interactions across them, but they're somewhat siloed.. And the p38 pathway is one arm of this, right? And so we kind of expected this. So that HIV, or that SARS-CoV-2 would turn on the p38 pathway is not surprising. SARS-CoV-2 infection leads to a lot of inflammation, a lot of cytokines. They bind the receptors, they turn on p38. And it's also not surprising that if you inhibit p38 with this SB-203580 inhibitor that you would see a reduction in cytokine levels. This is RT-qPCR of IL-6, but we have a whole Luminex panel of proteins as well, and this holds up. It's really not that surprising. What was surprising was that if you inhibit p38, you also inhibit viral replication. You don't see that for most viruses. Most viruses do turn on p38, but if you inhibit it, It'll block the cytokine production, and in animals it might reduce disease, but it doesn't directly affect viral replication. So that was sort of where the cell study with Nevan's lab ended, and then this is where my lab sort of picked it up and started working on it on our own.

So in that result, how do you know that IL-6 phenotype isn't related to the virus, if you have lower amounts of virus, so that means less infection of IL-6, cytokines, blah, blah, blah.

Yeah, it's really hard to tease those two things apart, but I think, well, I don't know, what's the question again?

So you said, oh, there's decreased cytokine production, right, with that inhibitor.

Yeah, right. But there is decreased— Oh, I see what you're saying. So if you block viral replication, you have less virus, you have less cytokines, Yeah, I guess, yeah, it could be that, but I think based on what we know, like if you treat with TNF-alpha to turn on inflammation and then you titrate in a P38 inhibitor, you would see the same thing. I think based on previous data that we've seen for P38 in cytokine profiles, I would have expected that. I'm sure the virus contributes even more to it, but if you take the virus away from there, I think that would still stand.— you're right, that experiment doesn't show it super clearly. So I think the second phenotype is what sort of interested us. I'm always interested in, you know, what pathways does a virus need in order to replicate? So I'm sort of expanding the p38 pathway here. It has 4 p38 kinases. Alpha and beta are the most well-described. Alpha by far is the the most well described. And then delta and gamma are sort of the non-canonical p38s. And then there's some downstream effectors. So we developed an siRNA screening method. This is using ACE2 A549 cells. We transfect with siRNA 48 hours, infect with SARS-CoV-2 36 hours, fix the cells, stain them with SARS-CoV-2, counterstain with DAPI, and then we're reading out on a We got Selega, which is basically a flow cytometer microscope, right? And so on the right here, what you're seeing is what the data, what the Selega is seeing. And with the non-targeting control, you see lots of infection. This is our positive control, which is an siRNA targeting the SARS-CoV-2 nucleoprotein itself, and you get very low infection there. What's the timing for SARS-CoV-2 replication? —replication cycle? It takes about 24 hours, but by that point— This is one amplification. It's probably a little bit more than that because it's very asynchronous. So by 24 hours, most of the things have replicated once, but some have replicated once already. You know what I mean? So by the 36-hour time point, you've probably got on average 2-ish. So this is the siRNA screening data for the p38s and down. And what I want to draw your attention to here is p38 beta, which had a really strong phenotype. If you knock it down, you get hardly any infection. It's almost as much as knocking down the SARS-CoV-2 N protein itself. And some other things had effects. Gamma, this was very reproducible actually, but P38 gamma knockdown always increased infection. And then for the downstream substrates, MSK2 had a sort of intermediate effect between the control and P38 beta. So P38 beta does signal through MSK2, so it's partial that— it's possible that partially some of the phenotype is through— I'm pointing to MSK2, but probably not all of the phenotype. Phenotype comes from that. And we did validate by Western blots. I'm not showing them here. We did plaque assays for p38 beta for all of our hits, actually. I'm just showing p38 beta. It's just another measure of infectivity, and in this case you can see about a 3-log reduction in the amount of infectious virus that you get out when you inhibit p38 beta, when you knock it down. So that was pretty surprising because P38 alpha is the one where all the inhibitors are developed against. It's the one that everybody studies. There is a P38 beta knockout mouse, and what this paper about P38 beta says is basically it doesn't do anything in inflammation. If you knock it out of mice, they're viable, they develop fine, they still develop sensitivity to LPS and TNF, and they still induce all of the downstream effects of inflammation that go through p38. So it's often presumed that p38 beta doesn't do anything and that alpha does everything. And what we had seen here was that alpha doesn't do anything and beta does everything, which is interesting and kind of worthy of following up on. [Speaker:JOSHUA_ELKINGTON] So again, I'm the phosphoproteomics guy, and so what we wanted to do here was to sort of perturb the system in as many ways as we could and collect phosphoproteomics data and using what we know about P38 substrates to extract new P38 substrates within this context of infection. And so we have a bunch of different conditions here with different inhibitors, different concentrations, different time points, and then we have different knockdown conditions. Unfortunately, there's only P38 alpha inhibitors available. There's no beta inhibitors available, so we can't exactly target P38 beta with an inhibitor itself. And so, from here, it's going to look very similar to the previous talk, actually, where we have this big dataset, we want to know where are the P38 substrates, In this case, there are a lot of P38 substrates that are annotated. With this data, I could do it a little bit fancier, where you have this dendrogram tree on the left here, and you can basically cut the tree in order of decreasing height. And for each cluster that you get out of cutting that tree, you can look for the p-value of enrichment for P38 substrates. And so basically what that says is that when you cut it at a certain point, there's one cluster here that has the most significant enrichment of p38 substrates. And once you reach this point, you've cut that cluster in half, and then your p-value goes up. So this is sort of your optimal p38 cluster, and that's this one at the bottom here. Realistically, if you have eyeballs, you could probably pick it out yourself, but it's kind of nice to have an unbiased approach to do it. So this is that cluster. I think we're doing a pretty good job of picking out P38 substrates, so we find all the ones that should be there. The ones with stars are the known P38 substrates. This is TRIM28, which is a P38 substrate. Many of these sites are on P38 substrates at different phosphorylation sites from what was known before, and many of them are RNA-binding proteins, which is a favorite target of P38. P38. And then the other way that we know that we're doing a good job of enriching for this is that we worked with Tomer Yaron and Luke Cantley's lab, and they've done this crazy study where they cloned and expressed and purified every human kinase and put it onto a peptide library and derived in vitro the kinase motif for every kinase. And then they basically took the motifs of all of the phosphosites in our cluster and looked for enrichment of those motifs, and what you're getting out here are MK2, 5, and 3, which are directly downstream of p38 alpha and beta. And then the CAM kinase is actually phosphorylate the same motifs, which I didn't know, and it's kind of interesting. And so we took those sites and we screened them with our siRNA screening platform, and this is not what I expected to see at all. The known substrates here are in this sort of candy stripe, I guess. So the most famous substrates are down here, HspB1 and RBM7, and then TRIM28 is here. For the most part, these guys don't seem to do a lot. They might reduce infection a little bit, and maybe the combination of them is contributing something, but by far the overwhelming effect of most of these substrates is that when you knock them knock them down, you get more infection. Which is kind of counterintuitive, but if you think about it, these are RNA-binding proteins. A lot of times RNA-binding proteins bind to their substrates in a phosphorylation-specific manner, so when they're phosphorylated, they let go of the RNA, and if you knock them down, they let go of the RNA, so you would see the sort of same effect as phosphorylation. That's sort of how I rationalize this, Again, we're looking at knocking down the whole protein. We're not perturbing individual phosphorylation sites, and I think that's the next step that we're trying to get to, which is to understand of these p38 phosphorylation sites, which ones are so important for SARS-CoV-2 infection. Then the other interesting finding from this was that we found phosphorylation sites on SARS-CoV-2 nucleoprotein itself that were sensitive to p38 inhibition. And so these sites, when you treat with the p38 inhibitor, this is a short treatment that doesn't affect viral replication yet. It's like a 4-hour treatment, I think. So the nucleoprotein levels are constant, but these phosphorylation sites went away really rapidly with that p38 inhibitor. Interestingly, there's been a lot of talk about other phosphorylation sites within the serine-arginine-rich domain. This guy, Nick Heaton, has shown that they're regulated by SRPK kinases. And it's interesting that when you turn these ones off, these ones come on. So there might be some mutual exclusivity in terms of where a nuclear protein gets phosphorylated. And the other piece of information I'm showing here is from the Nextstrain resource. Entropy is basically the degree of non-conservation. And so the SR sites in this region, they're actually not super conserved, whereas these sites that we're finding at the extreme N-terminus are pretty well conserved across all the SARS-CoV-2 strains that have been sequenced so far. At the level of phosphorylation, we don't know. I actually have seen some data that the beta strains don't seem to phosphorylate those sites as much as all the other strains. So there's something interesting going on there. But we were able to make a recombinant virus using— I forgot to acknowledge the guy— Luis Martinez-Sobrino made these recombinant SARS-CoV-2 viruses where we took our 4 phosphorylation sites and mutated them to alanines and then made the recombinant viruses and infected cells with them. And you're getting about a log reduction in the titers with these recombinant viruses, which similar to the— I had talked about MSK2 having an intermediate effect. It doesn't fully explain p38. I think this is similar where this is probably not everything that p38 is doing, but it could be some of the contribution to why SARS-1 cisphosphorylated so much.. And the other piece of information we have here is we know that p38, if you take it away, you get interferon activation. I haven't talked about that very much, but we are very interested to see that when you make this N4A mutant virus, you get much stronger upregulation of ISGs. MX1 is sort of our canonical one that we like to block for, but if you look at other ones, they do the same. So we think that somehow these phosphorylation sites are protecting probably viral RNA from being sensed and turning on the interferon pathway in those cells. There's a lot more to be done there. So I'll just summarize there. With SARS-CoV-2, we know that p38 is activated by SARS-CoV-2 infection. You know that p38 beta and not alpha is required for SARS-CoV-2 infection. Many of the p38 substrates that we identified through our studies appear to be antiviral, and we found a number of SARS-CoV-2 N-sites that are sensitive to p38 inhibitor, and that if you mutate those sites, you get less replication, you get more interferon activation. And then the last little tidbit I want to leave you with is this concept of analog-sensitive kinases that you may or may not be familiar with, but because there's not a P38 beta inhibitor, it just makes things very annoying for us. But there is this technology that was developed by Kevan Shokat at UCSF that's based on this idea that most kinase active sites have what he calls a gatekeeper residue that prevents— bulky analog sensitive— bulky ATP analogs from fitting inside the substrate binding pocket. And so what he has shown is that for most kinases, if you mutate this gatekeeper residue to a smaller residue like an alanine or a glycine, it opens up the binding pocket and allows you to use an ATP analog like one in APP1 that then fits only into the kinase that you've mutated. And what's really, really nice about this is that this particular threonine residue in p38 alpha and p38 beta has a perfect PAM right in the right distance to base edit it and to make exactly this mutation to an alanine. And so that's what we're working on now is making these analog-sensitive p38 alpha and beta alleles. And then we can inhibit with an inhibitor rather than knocking down for 2 days, infecting for a day and a half, and reading out a phenotype. Phenotype that. So this is my lab, and most of the SARS-CoV-2 work I showed is from Christina Higgins, who's shared between my lab and Ben Tenuver's lab. Trevor works on HIV. Boris is the base editing guy in my lab, and all the mass spec data that you've seen was generated by Andrew Courland, who's a really phenomenal technician All the cell line engineering and most of the cloning you've seen was done by Priti Adhikari, who's also excellent. A lot of the work was done with my colleagues at Mount Sinai and people that I used to work with at UCSF and I continue to work with even though I'm probably not supposed to according to reviewers.

Yeah, I'll leave it there. After infected by the SARS-CoV-2, you turn on the PE38, right? Is this turned on at the early stage of the infection or because of the late stage because of the cytokine and then activated?

Yeah, it is very early.

If I can find my mouse here.

—because of some virus protein turns on the kinase? Yeah, we don't know exactly how it's getting turned on. But if you look at the time course here, our 0-hour infection was actually we allowed binding to occur before we washed the virus away. So actually at 0 hours, you see activation of that pathway really fast. So it probably has something to do with binding the ACE2 receptor. Somehow modulating signaling with the receptor. Most receptors are somehow hooked up to a MAP kinase pathway, so I think it probably gets turned on very, very early that way, and then when you start making cytokines, it gets turned on even more at sort of this feed-forward effect, right? Do any other coronaviruses get back to hijack these pathways? Yeah, they do, and they do it through different ways. So one of them I forget which one it was. It was my department chair who did it, and I should probably know it. But other coronaviruses do turn on p38 using— I think it was the envelope protein turns it on, but it's not required for that virus.

I think it was HCoE229E. I have a question. Yeah. I was surprised too when you said the P30 and beta knockout mouse had no phenotype. What about alpha and gamma, and have they made like double depletion, triple depletion?

Like, what's your take on that? Yeah, delta and gamma haven't been studied very much at all as far as I know. And we didn't really see a— well, we saw a phenotype with gamma that went the other way, but we haven't really looked into it. But the alpha-beta thing I have looked into a lot, and almost everybody studies alpha. The only guy who studies beta is Simon Arthur, and when he looked at it, he saw nothing. I did see one study where they made an alpha-beta double knockout, and it had more of an effect than you would expect doing alpha alone or beta alone. It wasn't a knockout mouse because you can't knock out alpha in a mouse, but there's— It's embryonic lethal? It's embryonic lethal, yeah. But beta is not? Beta is not. Beta is fine. It's very strange, but I think, you know, the systems that Simon Arthur has tested for p38 beta were mostly like bacterial stress. Like LPS is a pretty generic bacterial stress. It wasn't a viral infection that he was throwing on there. So actually for my grant, I got him to provide a letter of support saying we would infect those mice with a mouse-adapted SARS-CoV-2. So if I get money to do it, I will do it. I'm not ready to pull the trigger on it just yet though. Sure. Thank you. Cool. Thank you very much.