Skip to Main Content

Dissecting and Overcoming Different Shades of Cancer Immune Evasion

May 20, 2024

Yale Cancer Center Grand Rounds | May 17, 2024

Presented by: Dr. Benjamin Izar

ID
11696

Transcript

  • 00:00Represent some of the work that my lab
  • 00:02has been up to in the last few years.
  • 00:05And as you will see it's it's quite
  • 00:08different and perhaps different from
  • 00:10what you expect from our prior work.
  • 00:12But it's sort of the next wave that I,
  • 00:14I, I sort of view important in
  • 00:16building on what we have done before.
  • 00:18So today I will talk about really
  • 00:21different shades of cancer immunivation
  • 00:22as you'll see and some strategies that
  • 00:25we're developing to overcome these
  • 00:27from multiple at different angles.
  • 00:30So just very briefly,
  • 00:31you know if I had to summarize sort of
  • 00:33what our lab is doing at this point,
  • 00:35it's really you know doing large
  • 00:37scale genomics to understand
  • 00:39cancer genetics using single cell
  • 00:40genomics and things like that.
  • 00:42But that's really the beginning
  • 00:44really want to use this information
  • 00:45to inform you know mechanisms
  • 00:47that underlie two major areas.
  • 00:49One is immunobiology and the other one
  • 00:52is metastatic organotropism which is
  • 00:54really critical to consider when we use
  • 00:57an immune base or or even other therapies.
  • 01:00The approach,
  • 01:00as you will see in a minute also is
  • 01:03typically so that our questions are
  • 01:05really inspired by clinical problems
  • 01:07and then broken down into models.
  • 01:10Frequently you have to develop
  • 01:11those models or methods to study
  • 01:13these things systematically.
  • 01:14And ultimately,
  • 01:15of course,
  • 01:16the goal is to bring this back to patients.
  • 01:19So I divided the talk into three chapters
  • 01:23that are loosely linked to each other,
  • 01:25as you'll see in a moment.
  • 01:27And I want to start with the first
  • 01:29one because I think it really
  • 01:32exemplifies how we can use information
  • 01:34from large genomic data to inform
  • 01:37precisely what mechanisms and sort
  • 01:39of things we should study in the lab.
  • 01:43So a few years ago we were really
  • 01:45interested in understanding
  • 01:46mechanisms of resistance to immune
  • 01:49checkpoint inhibitors and you know
  • 01:51the details don't matter so much,
  • 01:53but the approach was that we use single
  • 01:55cell RNA sequencing in these patients.
  • 01:58We looked at the cancer cells specifically
  • 02:00and we came up with this you know
  • 02:02signature which we called ICR signature,
  • 02:04immune checkpoint resistant signature
  • 02:05and again the the genes don't matter
  • 02:08so much but you know and and I think
  • 02:10we learned a lot in this study but
  • 02:12we're what we were left with is a
  • 02:13hypothesis that came out of the data
  • 02:15that was that this cancer cell intrinsic
  • 02:18program was somehow conferring T cell
  • 02:21exclusion or poor T cell infiltration
  • 02:24and impaired T cell activity.
  • 02:26But when you have these gene lists
  • 02:28which I know all of you have been at,
  • 02:30at one point in your project,
  • 02:32the question is how do you actually
  • 02:34prioritize what you're going to
  • 02:36look at right and and what not.
  • 02:37And typically what we do and this
  • 02:39is what we had done for this paper,
  • 02:41we focus on something that's plausible
  • 02:43for which there are reagents or other
  • 02:46practical or pragmatic reasons.
  • 02:48But we really wanted to make
  • 02:50this process of how
  • 02:52we validate things a bit more unbiased
  • 02:55and to do that we developed this method
  • 02:58which we dubbed Perturb site seek.
  • 03:00So basically what this method allows
  • 03:02you to do is couple CRISPR CAS 9
  • 03:06perturbations with single cell RNA
  • 03:08and protein profile using the site
  • 03:10seek and a 10X genomics platform.
  • 03:13So what you would get from this method
  • 03:16would be that you could perturb a
  • 03:18gene and then ask well what does it
  • 03:20do to the entire cells transcriptome
  • 03:22and part of the surface proteome.
  • 03:25And the way we use this method to
  • 03:28validate some of those findings that
  • 03:30we had in patients is leveraging
  • 03:32it in in in patient derived models
  • 03:34where we had isolated in addition
  • 03:36to doing the sequencing the cancer
  • 03:38cells from a Melanoma patient and
  • 03:40the tumor infiltrating lymphocytes.
  • 03:42The advantage is that this is a
  • 03:43you know fully autologous system.
  • 03:45It doesn't require any sort of
  • 03:47engineering if you will to do a
  • 03:50Co culture experiments into these
  • 03:52cancer cells we generated.
  • 03:53We generated a library that would
  • 03:55target each of the genes that we had
  • 03:58identified in the signature in patients.
  • 04:00So we could test in one you know pooled
  • 04:03experiment the relevance of any one of
  • 04:06those 248 genes that were in the signature.
  • 04:08And not only that but also we would
  • 04:10get you know the transcriptome
  • 04:11and part of the surface podium.
  • 04:14And the premise here would be that
  • 04:16cells that survive Co culture with
  • 04:19their autologous T cells must harbour a
  • 04:22perturbation that confers that mechanism.
  • 04:24And this is exactly what we did.
  • 04:27And I'm just going to give you this
  • 04:28is of course published at this point.
  • 04:30But what what this would get us is
  • 04:32a sense of the immune fitness and
  • 04:34the phenotype associated with this.
  • 04:36I'm just going to give a snippet of
  • 04:38results because I want to show you how
  • 04:40this helped us inform kind of what we
  • 04:42did next in the last couple of years.
  • 04:44So here we're looking at gene
  • 04:47knockouts on the bottom.
  • 04:48So genes knocked out here associated
  • 04:51with increasing fitness against
  • 04:53autologous T cells or tilts.
  • 04:55And on the Y axis here we
  • 04:57increased immune pressure.
  • 04:58So there's an effector to target ratio of 1
  • 05:01to 1 to the one and four to one and so on.
  • 05:03And as you would expect,
  • 05:05mutations or deletions in all of the
  • 05:06genes that I indicate here with the red,
  • 05:09Red Arrows,
  • 05:10all of those had been known before,
  • 05:12right.
  • 05:13Mutations or deletions in those
  • 05:15genes are strongly associated
  • 05:16with immune evasion and they had
  • 05:18been associated clinically with
  • 05:20resistance to immunotherapy.
  • 05:22So that was really good because
  • 05:23it validated our approach that we
  • 05:25could recover all of those hits.
  • 05:26Basically in one experiment,
  • 05:28we got really interested in another
  • 05:30hit that was less expected.
  • 05:31That was loss of a gene called CD 58.
  • 05:34So I'm going to get,
  • 05:34I'm going to talk about that more.
  • 05:36But on the flip side,
  • 05:37we could also couple how these
  • 05:39perturbations change the phenotype.
  • 05:40And as you can imagine this
  • 05:42is a huge matrix, you know,
  • 05:43perturbation by gene expression by by,
  • 05:45you know, protein profile.
  • 05:46So I'm just going to show you a tiny,
  • 05:48tiny snippet from that which I
  • 05:51want to use to guide you to the
  • 05:53to the experiments that we did.
  • 05:55So the way to read this here is that
  • 05:56on the bottom you have knockouts
  • 05:58of those genes and then on the
  • 06:00Y axis you have a few selected
  • 06:02features that I I want to present.
  • 06:04So for example, if you knockout CD 58,
  • 06:07then there will be no CD 58
  • 06:09protein left in that cell.
  • 06:11Logical.
  • 06:11But what was really interesting was the
  • 06:14observation that cells that lose this gene,
  • 06:17CD 58 had concurrently more
  • 06:21protein by of a gene encoded
  • 06:24by a protein encoded by CD 274,
  • 06:27which of course is PDL 1.
  • 06:28So that seemed like a double whammy.
  • 06:30You lose something good that
  • 06:32confers immune evasion,
  • 06:34but then you also gain something that
  • 06:35is a Co inhibitory ligand of course.
  • 06:38So what is CD58?
  • 06:39It turns out we actually don't
  • 06:41know that much about it in cancer.
  • 06:44What we do know physiologically is
  • 06:46that it is a Co stimulatory protein
  • 06:48that ligates to CD2 on T cells,
  • 06:51and when it does so,
  • 06:52it can become actually the most
  • 06:55potent Co stimulatory protein.
  • 06:57So loss of this gene of protein,
  • 06:59it's plausible that that could
  • 07:01result in immune evasion in
  • 07:02a number of different ways.
  • 07:04So we sought to validate this
  • 07:06and this is work that has since
  • 07:08then been LED and then published
  • 07:10by my first MDPHD student who
  • 07:12just graduated a few months ago.
  • 07:14So here basically what she did is
  • 07:16she took these Melanoma cells,
  • 07:17she knocked out CD58 and then she Co
  • 07:20cultured the cells with autologous T
  • 07:22cells or engineered T cells at that point.
  • 07:24And as you can see loss of the
  • 07:26gene in fact convert a better
  • 07:27survival of these cancer cells.
  • 07:29And when you rescue these,
  • 07:31the gene, either it's GPI anchored
  • 07:34or transmembrane isoform,
  • 07:35then you rescue the sensitivity
  • 07:38to T cell cleaning.
  • 07:40Furthermore,
  • 07:40we also wanted to demonstrate
  • 07:42that this interaction with CD2
  • 07:45was in fact required for this
  • 07:47immune evasion phenotype and that,
  • 07:49you know,
  • 07:50CD 58 loss didn't confer a loss
  • 07:53of fitness through some other
  • 07:54mechanism that we didn't know.
  • 07:56So to test this we repeated the same
  • 07:58experiment that I show on the left,
  • 08:00only this time we rescued the
  • 08:02knockout cells with a variant of
  • 08:05CD58 harboring and mutation K34A
  • 08:08which is unable to actually to CD2.
  • 08:10And as you can see when you rest,
  • 08:12when we rescue with the mutant,
  • 08:13the cells continue to be resistant
  • 08:15to D cell Co culture suggesting that
  • 08:18this is very specific to that interaction.
  • 08:21So one of the reasons I think this
  • 08:23gene is not well understood is that
  • 08:26there is no known mouse homolog.
  • 08:28So we can't use the, you know,
  • 08:30models that we typically like to use in,
  • 08:33you know, studying, you know,
  • 08:34immunotherapy and so on syngenic model.
  • 08:37So to to you know, circumvent this,
  • 08:39what we did is we used an
  • 08:41immunocompromised mouse that has
  • 08:43transgenic expression of human Illinois 2,
  • 08:46which of course is required
  • 08:48for T cell survival in vivo.
  • 08:50So into these animals we could implant
  • 08:53either the parental or the, you know,
  • 08:56genetically modified cancer cell lines
  • 08:58and then adoptively transfer the mouse
  • 09:00with the patient's own tilts, right.
  • 09:02So we could study this interaction
  • 09:05in vivo and as you can see,
  • 09:07the tumors that had the CD 58 loss
  • 09:10were completely resistant to ACT.
  • 09:12They also had an approximately hundredfold
  • 09:15lower infiltration with T cells,
  • 09:17validating some of the predictions
  • 09:19that we had made in patients.
  • 09:21And all of these effects could be rescued
  • 09:24by RE expressing CD58 in the cancer cell.
  • 09:27So overall this suggested that loss
  • 09:29of CD 58 on the cancer cell conferred
  • 09:31impaired T cell infiltration,
  • 09:33proliferation and resistance to ACT.
  • 09:35So coming back to that interesting
  • 09:38interaction which I mentioned earlier,
  • 09:40this interaction between CD58 and PDL 1.
  • 09:44So we did a very simple experiment
  • 09:48in which we knocked out CD58 and
  • 09:50simply asked how much PDL one is
  • 09:52on the surface of these cells.
  • 09:54And in fact, when we knocked out CD58,
  • 09:56we found that these cells do have more CD 58,
  • 10:00excuse me, PDL one protein on the surface.
  • 10:02And this effect again could be
  • 10:05rescued by RE expressing CD58 itself.
  • 10:08So the question then is,
  • 10:10you know what sort of regulates this
  • 10:12interaction and how do you go about this?
  • 10:15Because we there's no like nothing really
  • 10:17to help us inform of where to even start,
  • 10:20right.
  • 10:20What we could exclude pretty
  • 10:22quickly is that there was no direct
  • 10:24interaction between the proteins.
  • 10:26So there had to be some sort of
  • 10:28mediator that regulates that.
  • 10:30So to do this systematically we
  • 10:32did a genome scale loss of function
  • 10:35screen that would show us or point
  • 10:37us towards genes or proteins that
  • 10:40are required for this interaction.
  • 10:41So the design of the screen was that
  • 10:43we took these Melanoma cell lines that
  • 10:45express CAS nine and then we introduced
  • 10:47the genome scale guide library to
  • 10:49knock on every gene in the genome,
  • 10:51let the cells, you know,
  • 10:53edit for a couple of weeks.
  • 10:55And then we sorted out the CD 58
  • 10:58negative or CD 58 positive cells,
  • 11:00sequenced the guide RNH guide
  • 11:02RNAs in each of these pools.
  • 11:04And the premise here is that
  • 11:06in the CD 58 low pool,
  • 11:08there must be perturbations that are
  • 11:10somehow involved in regulating CD 58.
  • 11:13So when you lose that gene,
  • 11:14you see a reduction in CD58.
  • 11:16And this is precisely what we saw.
  • 11:18So here is a result of that screen.
  • 11:21Reassuringly,
  • 11:21the top hit of the screen was
  • 11:24knockout of CD 58 itself,
  • 11:26knockout CD58.
  • 11:27There will be CD50 negative and a
  • 11:29bunch of others.
  • 11:30The one that really caught our attention,
  • 11:32but we because we saw it
  • 11:34also to physically interact
  • 11:36with CD58IN in a mass spec Co IP screen,
  • 11:38is this gene or protein called
  • 11:41CMTM 6 super interesting?
  • 11:42I know David had actually done
  • 11:45some work that looked at the
  • 11:47prognostic value of this protein,
  • 11:48but it was unclear and I think we
  • 11:50have the answer to why that might
  • 11:52be So what was really interesting,
  • 11:54just, you know, a year or two
  • 11:57before we had made this observation,
  • 11:59there were two Nature papers published
  • 12:02showing that the same gene or protein
  • 12:05CMTM 6 was in fact required for
  • 12:08maintaining PDL one protein on the surface.
  • 12:11So this was a plausible, you know,
  • 12:14logical sort of hit to to go
  • 12:15after and this is what we did.
  • 12:17So when we knockout CMTM 6,
  • 12:19we see a reduction in both CD58 and
  • 12:22PDL one protein surface abundance.
  • 12:24And when you rescue the gene CMTM 6,
  • 12:27you rescue that and you have you you
  • 12:29you bring them back to the baseline.
  • 12:31And to really prove that this
  • 12:32is required for the interaction,
  • 12:34we generated a number of additional
  • 12:36genetic perturbations,
  • 12:37double mutants which we rescued where
  • 12:40we rescued only one gene at a time
  • 12:43where we could in fact demonstrate
  • 12:45that CMTM 6 was required for
  • 12:47mediating this reciprocal interaction.
  • 12:49You know the the issue with these
  • 12:51types of down signals, right,
  • 12:53is always, well,
  • 12:54how do you think about making
  • 12:56a therapy from that,
  • 12:57right?
  • 12:58Because ultimately that's always the goal.
  • 13:00So I'm going to skip a lot of
  • 13:02data that we show in the paper.
  • 13:04But Long story short,
  • 13:06we identify we found that the binding
  • 13:09A sequences on CD58 and PDL one for
  • 13:13CM takes actually CMTM 6 actually
  • 13:16differ and so we imagined that
  • 13:18we could leverage that knowledge.
  • 13:20So it turns out that CMTM 6 binds
  • 13:24to a specific amino acid domain
  • 13:28on North terminal domain in PDL 1
  • 13:32spanning the amino acids 20 to 32.
  • 13:35So when we scramble that area,
  • 13:36turns out that, you know,
  • 13:38CMTM 6 can no longer bind to PDL 1.
  • 13:41So our we imagined that if we could
  • 13:44selectively disrupt the interaction
  • 13:46between PDL one and CMTM 6 at that site,
  • 13:49it should result in reduction in PDL
  • 13:521 without affecting the levels of CD 58.
  • 13:55And this is precisely what we saw.
  • 13:57So we took these Melanoma cell lines,
  • 13:58we knocked out PDL One,
  • 14:00and then we rescued either the wild
  • 14:02type orph or a orph where we scrambled
  • 14:05that region that is unable to bind CMTM 6.
  • 14:08And as you can see, the WILD type
  • 14:10orph rescues PDL ONE expression,
  • 14:12but the mutant does not.
  • 14:15And then in a Co-op experiment,
  • 14:17we could also directly show that this
  • 14:20variant where we scramble that sequence
  • 14:22is unable to has a significantly
  • 14:25lower binding of Co-op to CMTM 6.
  • 14:28So just to summarize this part,
  • 14:31I hope I was able to show you that
  • 14:33we were able to go from you know
  • 14:36sequencing data to using the right
  • 14:38functional tools to really inform
  • 14:39precisely kind of what to go after.
  • 14:42But you know what we were left
  • 14:44with is actually the question,
  • 14:45how many of these interactions do we miss,
  • 14:47right. Every time you knockout a
  • 14:49gene and you observe A phenotype,
  • 14:52you know how do we know that
  • 14:53that's not mediated through a
  • 14:55number of these interactions.
  • 14:56And the, the the clinical or
  • 14:58therapeutic correlate of that is,
  • 15:00you know giving somebody a single
  • 15:02agent immunotherapy or you know even 2
  • 15:04drugs and asking well what do these,
  • 15:06what does inhibition of these two proteins
  • 15:08do to everything else that's going on,
  • 15:10on the surface or within within the cell.
  • 15:13And this is actually something we're
  • 15:15trying to address systematically.
  • 15:17OK.
  • 15:17So switching gears a little bit and coming
  • 15:20to the second chapter which is a much,
  • 15:22much more recent chapter in the
  • 15:25lab that is leveraging novel
  • 15:28base editing tools to hopefully
  • 15:32improve cell based immunotherapies.
  • 15:35So all of you know that cell based
  • 15:38immunotherapies are now a critical
  • 15:40component of the treatment of many
  • 15:42hematologic malignancies and most
  • 15:44recently there was also an approval
  • 15:46for the the treatment of till
  • 15:49transfer for patients with Melanoma.
  • 15:50And you know the the typical you
  • 15:52know workflow is so that you take
  • 15:54something out of the patient and
  • 15:55you don't do something with the
  • 15:57cells and you you put them back in.
  • 15:58And in the context of CAR T cells,
  • 16:00of course that's taking PBMCS and putting
  • 16:02a CAR into the cells and reinfusing them.
  • 16:05In the context of till therapy,
  • 16:07it is isolating tills
  • 16:09from metastatic lesions,
  • 16:10expand them ex vivo and then
  • 16:12give them back to the patient so
  • 16:13that they have in common.
  • 16:15What they also have in common is the,
  • 16:17the observation that has really
  • 16:19emerged in the last few years is
  • 16:21that there are very specific T cell
  • 16:24features before you put this therapy
  • 16:26into the patient that are strongly
  • 16:28predictive of whether or not that
  • 16:31cell product is going to work.
  • 16:33And and this was published by the
  • 16:35Rosenberg Group A few years ago
  • 16:37and there's nothing shocking about
  • 16:38some of the observation.
  • 16:40But it really is sort of the rationale
  • 16:43for thinking about how we can improve
  • 16:45T cell function itself to build
  • 16:48better cell cell therapies on top.
  • 16:50And of course,
  • 16:51we're not the only ones thinking about.
  • 16:53There's a lot of groups take CAR T
  • 16:55cells until therapies and engineer
  • 16:56them in a number of different ways.
  • 16:59You know,
  • 16:59frequently in the last few years
  • 17:01what people have done is, you know,
  • 17:03for example,
  • 17:03knocking out an inhibitory receptor
  • 17:05such as CTLA 4, right?
  • 17:07And I picked one study from Carl
  • 17:09Jun's group could have picked,
  • 17:11you know,
  • 17:11hundreds of other papers where they
  • 17:13try to improve CAR T cell therapy by
  • 17:16deleting some of these inhibitory receptors.
  • 17:18But the challenge with knocking
  • 17:21out a gene especially in T cells
  • 17:24is that one of the off target
  • 17:27effects of CRISPR CAS 9,
  • 17:28which is due to the double
  • 17:30stranded DNA breaks that
  • 17:32take place is that you actually get
  • 17:34a pretty high rate of aneuploidy.
  • 17:35So 7 to 14% of cells in of T cells in
  • 17:39a pool that you you know engineer with
  • 17:42CRISPR CAS nine will be anemployed.
  • 17:43And of course aneuploidy or chromosomal
  • 17:46instability as I will talk about
  • 17:48later is is a hallmark of cancer.
  • 17:50And that comes with all sorts of concerns.
  • 17:53Of course, what got me really
  • 17:56interested in thinking about how to
  • 17:58improve cell therapies are papers
  • 17:59like the one that I'm showing you
  • 18:01here from the Jonathan Powell group.
  • 18:03So rather than deleting a gene or
  • 18:06over expressing a gene which which
  • 18:08comes with other issues, you know,
  • 18:10they made a really interesting
  • 18:12observation that is they found that
  • 18:15some of their mice had sporadic
  • 18:17mutations in a gene called TSE 2.
  • 18:19The mutation itself really
  • 18:20doesn't matter so much.
  • 18:21But what they were able to show is actually
  • 18:24when you take T cells from the mouse
  • 18:26that has this germline mutation and you
  • 18:29adoptively transfer mice with the wild
  • 18:31type gene that harbor Melanoma tumors,
  • 18:34those mutant T cells are much,
  • 18:36much more potent in eliminating not only
  • 18:39melanomas but they also show this in
  • 18:41context of leukemias and and other diseases.
  • 18:44So what that suggested to us is
  • 18:46that maybe we don't have to,
  • 18:48you know,
  • 18:48take the wheels off the car.
  • 18:50I mean knocking out an entire gene with
  • 18:52all of its unintended consequences,
  • 18:55maybe it's sufficient to introduce very
  • 18:57specific mutations in genes that will
  • 19:00significantly alter T cell function.
  • 19:02And this is the hypothesis that
  • 19:05we sought to to test in this
  • 19:07project that just like in mice,
  • 19:09there must be either naturally a cure
  • 19:12occurring or synthetic protein variants
  • 19:14that may enhance T cell function and
  • 19:17therefore may enable the production of
  • 19:20more effective cell therapies on top.
  • 19:23And the method that you know
  • 19:25would be required to do that,
  • 19:28One of the ways to do that
  • 19:29is using base editors.
  • 19:31So there's different types of base editors.
  • 19:33These are CRISPR CAS 9 dependent base
  • 19:39editors that either citadine or adenosine
  • 19:42deaminase linked proteins that unlike
  • 19:45CRISPR CAS 9 don't introduce double
  • 19:47stranded DNA breaks but rather they
  • 19:50induce deamination events in very
  • 19:53specific windows guided by guide RNAs.
  • 19:56And ultimately in the in the example of
  • 19:59citadine base editors you get C to T
  • 20:01changes and in the context of denizine
  • 20:03base editors you get A to G changes.
  • 20:06So what this allows you to do
  • 20:09is introduce at some specificity
  • 20:11mutations at defined loci in a gene
  • 20:14rather than knocking out the gene.
  • 20:16And one of the challenges that that
  • 20:18had existed with these base editors
  • 20:21was the issue that their efficiency,
  • 20:23especially in primary human T cells,
  • 20:25was rather low.
  • 20:27So we This was led by an MDPHD
  • 20:30student in my lab.
  • 20:31Zach Walsh has shown all the
  • 20:33way to your left.
  • 20:34He took it upon himself to try
  • 20:36to improve the efficiency of
  • 20:38these base editors because these
  • 20:40would be the right tools to
  • 20:42really introduce some of these
  • 20:43mutations in a targeted fashion.
  • 20:45And he's done that through a really sort of
  • 20:48smart way of delivering the base editor.
  • 20:51I'm not going to go through all the details,
  • 20:52but suffice it to say, you know,
  • 20:55we're able to achieve extremely high
  • 20:57efficiency, relatively speaking,
  • 20:59between 80% and and 99% with these base
  • 21:03editors introducing very precise variants.
  • 21:06This is a paper that is accepted,
  • 21:08that will be published next week.
  • 21:10But so, you know,
  • 21:11equipped with these methods,
  • 21:13we then imagined, well,
  • 21:14what do we want to actually,
  • 21:16you know, edit in these T cells?
  • 21:17What do you even start?
  • 21:18We we can't do this on the genome scale.
  • 21:20There's too many bases, right?
  • 21:22It would be an impractical experiment.
  • 21:25So what we decided to do is rather
  • 21:27be guided by experiments of nature.
  • 21:30And what I mean with that is,
  • 21:31you know,
  • 21:32there's a lot of variants out there
  • 21:34that are reported to be either
  • 21:37definitively associated with immune,
  • 21:38clinical, immune syndromes,
  • 21:40either autoimmunity or immunodeficiency
  • 21:42and everything in between or variants.
  • 21:45And these are most of them that are
  • 21:47variants of uncertain significance
  • 21:49where there may be an association,
  • 21:51but we don't know exactly because
  • 21:53we can't prove each of them,
  • 21:54you know, emotionally one at a time.
  • 21:56So what we decided to do is put
  • 21:59together sort of a library of 30,000
  • 22:01variants that are out there across
  • 22:04102 genes spanning all major T
  • 22:06cell functions and introduce all
  • 22:08of them with these base editors
  • 22:10in a massively parallel fashion.
  • 22:12And then ask how each of these
  • 22:16variants changes known hallmarks of
  • 22:18T cell mediated anti tumor immunity
  • 22:22including the activation of T cells,
  • 22:24the proliferation, cytochrome production,
  • 22:26long term expansion, persistence, etcetera.
  • 22:29And the the,
  • 22:30the premise here is that we want to
  • 22:33identify variants from this pool
  • 22:35that improve most and perhaps all
  • 22:37of those favorable features that
  • 22:39we know are important to build
  • 22:42good cell therapies on top.
  • 22:44And so just you know a word
  • 22:46about negative control.
  • 22:47So in addition to the
  • 22:48variants that we introduced,
  • 22:49we had a number of
  • 22:51different negative controls.
  • 22:51I think it's always important
  • 22:53to think about these when you
  • 22:54do these large scale screens.
  • 22:55So here's a distribution and
  • 22:57the log fold change of negative
  • 22:59controls that we had introduced.
  • 23:01So these include splice acceptor
  • 23:03and splice donor variants.
  • 23:05Or when you when you mutate these
  • 23:07these splice sites then what you
  • 23:09get is truncated proteins that get
  • 23:11basically knocked out or lack of a
  • 23:13better word or they get truncated
  • 23:15proteins that that are non functional.
  • 23:18So as you can see these are
  • 23:21significantly depleted at a very
  • 23:22high lock fault rate ratio,
  • 23:24while mutations that introduce silent
  • 23:26changes or empty window changes
  • 23:28don't change the distribution at all.
  • 23:30In addition,
  • 23:31another good control for T cells
  • 23:33of course is introducing mutations
  • 23:35that result in disruption of the
  • 23:37CD3 complex because the cells need
  • 23:39the complex to to be activated and
  • 23:41proliferate and do what they do.
  • 23:43And these are also significantly
  • 23:44depleted and and this was highly
  • 23:46consistent between different
  • 23:48donors that we did the screen,
  • 23:50so we did them multiple donors.
  • 23:51So here are a few results from that
  • 23:54screen now and I'm going to show
  • 23:56a couple of very selective ones.
  • 23:58So here we are looking at the lock fold,
  • 24:01a change of genes and the designated
  • 24:04variants that were introduced and the
  • 24:07you know the negative lock 10 FDR.
  • 24:09So the higher you go the most statistically
  • 24:12significant things were in the screen
  • 24:14and again there were many sort of
  • 24:17expected depleted genes such as CD3
  • 24:19or you know row A and what have you.
  • 24:21But what caught our attention was this
  • 24:25number of these mutations that were
  • 24:28enriched meaning they improved T cell
  • 24:31function were found in the PIC three
  • 24:33CD gene and also in the PIC 3R1 gene.
  • 24:36These two genes encode for the two domains
  • 24:39of the immune cells specific PI3K delta.
  • 24:44And this is another way to look at is this
  • 24:46time we're only looking at pick three CD,
  • 24:49this time looking at the amino
  • 24:51acid sequence from left to right
  • 24:53and where each of these mutations
  • 24:55that either enrich on the top or
  • 24:57deplete in the screen on the bottom.
  • 24:59And as you can see,
  • 25:00there's a number of of enriched
  • 25:02variants here that are associated
  • 25:05with a favorable TC cell phenotype,
  • 25:08you know spanning residues 524 to 529.
  • 25:11But then this is also this other variant,
  • 25:14the C416R that was strongly enriched.
  • 25:17And on the flip side,
  • 25:18we have a mutation that that causes
  • 25:20a loss of function of this gene.
  • 25:23When you look at where these
  • 25:25where this all of these gain of
  • 25:27function mutations are located,
  • 25:29they're not anywhere in the kinase domain.
  • 25:31As you might imagine,
  • 25:33since this is the catalytic
  • 25:35sort of subunit of PR3K delta,
  • 25:37they are all aligned and this is
  • 25:39a prediction from alpha fold.
  • 25:41They're all aligned at the interface
  • 25:43between these two gene products
  • 25:46of pick three CD and pick 3R1.
  • 25:49So you know,
  • 25:50we of course then went on to to
  • 25:52validate these observations.
  • 25:53You know one of the expectations
  • 25:55would be that you have higher
  • 25:56output from the PI3K pathways.
  • 25:58So we looked at downstream signalling
  • 26:01at phosphor akt and phosphorus 6 and
  • 26:03we tested many different variants.
  • 26:05But I'm just,
  • 26:06you know highlighting here in
  • 26:07the red box the loss of function
  • 26:09mutation and then the green box
  • 26:10the gain of function mutation.
  • 26:12And as you can see the gain of function,
  • 26:14you see more phosphor AKT and more
  • 26:16phosphorus 6 while it is both of these
  • 26:19are reduced in the loss of function.
  • 26:21In the same vein and again I
  • 26:23highlighted that with the green
  • 26:24boxes the gain of function and in
  • 26:26the red boxes the loss of function.
  • 26:28And this is just a selection of the data.
  • 26:29But we can see that the gain
  • 26:31of function variant
  • 26:32was associated with improved TNF alpha
  • 26:34production and proliferation and so on.
  • 26:36You know an initial validation
  • 26:38of the of the screen.
  • 26:39So now of course the question
  • 26:41is can we use this information
  • 26:42and improve cell therapies?
  • 26:44Can they be better of cell killers?
  • 26:46And to do this we used a simple
  • 26:48coke culture experiment like the one
  • 26:50that I had presented to you earlier.
  • 26:53Only this time we engineered the T
  • 26:55cells to express a very specific T
  • 26:58cell receptor against Nye cell one,
  • 27:00which is a commonly expressed
  • 27:02neo antigen on Melanoma.
  • 27:03So we can really test the specificity
  • 27:06and then we either use the native T
  • 27:08cells or we introduced one of a number
  • 27:11of variants that we had identified in
  • 27:13the screen and then Co culture them.
  • 27:15And as you can see again the gain
  • 27:17of function in green was strongly
  • 27:19associated with a higher degree of Poly
  • 27:22functionality here summarized as the
  • 27:23fraction of cells that expressed TNF alpha,
  • 27:26renzon B and IL 2 while the loss
  • 27:28of function showed a reduction.
  • 27:30And then this also translated in
  • 27:33improved license of Melanoma cells.
  • 27:35So here I think you can see my cursor.
  • 27:37Yep.
  • 27:37So here is the gain of function
  • 27:39variant and we're looking at the
  • 27:41number of surviving Melanoma cells.
  • 27:42As you can see that strongly reduced
  • 27:44the loss of function does not
  • 27:47enhance the activity of the T cells.
  • 27:50What was really gratifying and
  • 27:52this sort of closing the loop
  • 27:53to the first part of my talk,
  • 27:55we were also able to show that this
  • 27:57gain of function variant was able to
  • 28:01overcome resistance from CD58 loss.
  • 28:02So we did all of these coke culture
  • 28:04experiments, all repeated them,
  • 28:06only this time we knocked out CD58 and
  • 28:09then the coke culture and the C416R,
  • 28:12the game function variant and T cells
  • 28:14was in fact able to almost completely
  • 28:16radical themselves and you know,
  • 28:18without the labouring the point too much.
  • 28:20We also tested the same strategy in
  • 28:22a number of different CAR T cells and
  • 28:25we find the exact same thing whether
  • 28:27you use the CD9 CAR or CD22 CAR
  • 28:29against different leukemia models.
  • 28:31Introducing these variants in the T
  • 28:33cell that is the basis for making that
  • 28:36product improved their functionality
  • 28:38and their ability to lyse these.
  • 28:40Looking as summarise this portion
  • 28:44of the presentation,
  • 28:45Hope was able to show you that we
  • 28:48are now able to base edit primary
  • 28:51human T cells with a high efficiency
  • 28:54that unbiased discovery of variants
  • 28:57from a from a big pool of variants
  • 28:59may be able to identify those that
  • 29:01improve T cell function and those
  • 29:04perhaps could be used to improve cell
  • 29:06therapies broadly in the future.
  • 29:13OK. So now I'm going to switch to a
  • 29:18somewhat different area of the of the
  • 29:20lab or work in the lab that we're doing.
  • 29:23But you know the common theme is that we are
  • 29:26interested in what causes immune evasion
  • 29:28and and lack of response to immunotherapies.
  • 29:31And the the reason we got into this,
  • 29:34again it's a clinical one.
  • 29:35As you as all of you know,
  • 29:37brain metastasis are a common
  • 29:40problem across cancers,
  • 29:41but very common in Melanoma.
  • 29:43In fact, the incidence is probably as high as
  • 29:4675% in patients who have advanced disease.
  • 29:50And while the combination of
  • 29:52immunotherapies are you know,
  • 29:53showing efficacy against brain metastasis,
  • 29:57there's still a lot of work to do.
  • 29:59You know one of the reasons for that
  • 30:00is that those regimens are very toxic
  • 30:03and and despite the activity in some
  • 30:05patients we we still see you know
  • 30:07forms of Immunivision that that seem
  • 30:09to be pretty distinct in the brain.
  • 30:11So the the you know the the motivation
  • 30:14was really to study a brain metastasis
  • 30:16but I'll show you how that sort of got us
  • 30:19into this field of chromosomal instability.
  • 30:21So a couple years ago we we published a
  • 30:25paper in this was led by Jana Johannes
  • 30:27and Yiping postdocs in my lab where we
  • 30:31asked a simple question in patients,
  • 30:33what is the difference between an
  • 30:35untreated brain metastasis and an
  • 30:37untreated extracranial metastasis.
  • 30:39What we didn't want is any sort of
  • 30:41therapeutic intervention in between.
  • 30:42We're really interested just in
  • 30:44the salient biology which has been
  • 30:47quite poorly described actually in
  • 30:49patients compared to other you know
  • 30:51areas in Melanoma at least.
  • 30:53And I just want to point out a couple
  • 30:55of sort of results from this paper.
  • 30:57The first one is when we compare Melanoma
  • 31:01brain Mets MBM versus extracranial Mets ECM,
  • 31:04we found that the brain Mets were
  • 31:06had a higher fraction of the genome
  • 31:09altered FGA and that is a surrogate or a
  • 31:13process called chromosomal instability.
  • 31:15What is chromosomal instability?
  • 31:16It is a a a hallmark of cancer.
  • 31:19It's rather broadly seen across
  • 31:22almost every solid tumor.
  • 31:24And one of the ways by which chromosomal
  • 31:27instability can arise is through errors
  • 31:30that cancer cells make during anaphase,
  • 31:32where they don't segregate chromosomes
  • 31:34properly so that one of the data
  • 31:37cell you know is left with more and
  • 31:39the other one with less material.
  • 31:41The end product of this is aneuploidy,
  • 31:43right, and chromosomal civility,
  • 31:44sort of the perpetual dynamic
  • 31:46process that gives rise to that.
  • 31:48The extra material in the one of the
  • 31:51data cells is frequently packaged,
  • 31:52if it survives,
  • 31:54is frequently packaged in so-called
  • 31:56micronuclei.
  • 31:56So one of the ways to quantify chromosomal
  • 32:00instability more functionally beyond
  • 32:02just genomics is actually look at
  • 32:05the frequency of those micronuclei.
  • 32:07And this is what we did.
  • 32:08This is in the same study where we
  • 32:10had cell lines that were derived from
  • 32:12either a brain or an extracranial
  • 32:14metastasis from the same individual.
  • 32:17We enumerated the rate of micronuclei
  • 32:19and as you can see the one from the
  • 32:21brain in fact had more micronuclei
  • 32:23compared to the one that came
  • 32:24from a lymph node in this case.
  • 32:26And when we put these cells back into
  • 32:29animals in immunocompromised mice,
  • 32:30those cells in fact are more likely to
  • 32:34cause brain metastasis in the mouse than
  • 32:36those that come from outside the brain.
  • 32:38The second sort of a key result
  • 32:40from the study when we looked at the
  • 32:43microenvironment was the observation
  • 32:44that brain that appeared to have a much
  • 32:47more rhotomogenic myelod compartment
  • 32:48as you can see both from the single
  • 32:51cell data to the on the left here.
  • 32:53And then also we validated this
  • 32:55in two independent patient cohorts
  • 32:57by Multiplex immunofluorescence.
  • 32:59So this is all in in Melanoma.
  • 33:02The question then of course is you know
  • 33:04what about other common cancers that
  • 33:07frequently you know metastasize to the brain,
  • 33:09The most common one in terms of
  • 33:11prevalence is non small cell lung cancer.
  • 33:13So naturally we're interested in
  • 33:15asking do some of these concepts also
  • 33:17apply to non small cell lung cancer
  • 33:19And the answer is and this is sort of
  • 33:21in an analogous study that we're that
  • 33:23we are trying to publish right now.
  • 33:26We also asked the same question in
  • 33:28in that disease and this time though
  • 33:30we had a lot more data.
  • 33:32This time we could leverage data from
  • 33:35the MSK impact cohort where we had
  • 33:37genomic data that was linked with the
  • 33:40location where the of the of the disease.
  • 33:43So there's lots of primary tumors and
  • 33:45then a bunch of different metastatic
  • 33:46sites and then all the way on the right
  • 33:49here you see again brain metastasis have
  • 33:51the highest fraction of genome altered,
  • 33:53again a surrogate for
  • 33:55chromosomal instability.
  • 33:56We went on to validate this and
  • 33:58a couple of additional published
  • 34:00cohorts that are out there as well
  • 34:03as in a very large cohort of nearly
  • 34:069500 patients where we had whole
  • 34:08exome and RNA sequencing Through
  • 34:09an industry collaboration,
  • 34:11we find the exact same observation
  • 34:13that brain Mets are more unstable
  • 34:14than extracranial Mets which are
  • 34:16more unstable than the primary tumor.
  • 34:18So.
  • 34:18So clearly this process seems to be
  • 34:22important in conferring a sort of
  • 34:25aggressive phenotype and perhaps
  • 34:27also in modulating the immune
  • 34:29environment in an unfavorable way.
  • 34:31But it's kind of an obscure
  • 34:33concept to to study, right,
  • 34:36because it's such a perpetual
  • 34:38dynamic process.
  • 34:39How do you,
  • 34:40how do you go about actually studying that?
  • 34:42I think the so the first question
  • 34:44that we as ourselves is well what
  • 34:46is a good model to use to study this
  • 34:49and what is a good model that we we
  • 34:52could sort of easily identify, right.
  • 34:55So the way we approach this problem is
  • 34:58we looked at public data again TCGAACR genie,
  • 35:01CP tag and we asked which subsets
  • 35:04of lung non small cell lung cancer
  • 35:08are particularly chromosomally
  • 35:10unstable and are defined by very
  • 35:13distinct genomic subsets.
  • 35:14And it turns out that one particular
  • 35:17mutation or loss in a gene called LKB
  • 35:20one also known as STK 11 was across
  • 35:23the board associated with a higher
  • 35:25rate of chromosomal instability.
  • 35:27What you also need to know about
  • 35:30this particular subset of non small
  • 35:32cell lung cancer which is common
  • 35:34is that these patients virtually
  • 35:37never respond to immunotherapy.
  • 35:39So this is from a paper published
  • 35:41from MD Anderson where they
  • 35:42looked at patients with or without
  • 35:44mutations or deletions in SDK 11.
  • 35:46And as you can see here in red,
  • 35:48these patients do extremely poorly
  • 35:50in response to PD1 inhibition.
  • 35:53And lastly,
  • 35:54this particular subset happens
  • 35:56to also more frequently be
  • 35:59associated with brain metastasis.
  • 36:01So you know, with this information,
  • 36:03we believe that this particular subset is a
  • 36:07really good archetypical sinhi chromosomally,
  • 36:10chromosomal instability,
  • 36:11high disease to study some of those concepts
  • 36:14that that we want to understand better.
  • 36:17And this is what Lindsay,
  • 36:18another MDPHD student in my Lac lab,
  • 36:21took upon herself a couple of
  • 36:23years ago and she started with a
  • 36:25couple of very simple experiments.
  • 36:26We brought in a few human cell
  • 36:29lines which are shown here and
  • 36:30two of them are LKB 1 deficient,
  • 36:32the other one LKB 1 proficient.
  • 36:35And we did imaging and enumerated
  • 36:37the rates of these micronuclear.
  • 36:40I'm showing you here 2 exemplary ones.
  • 36:42And as we would predict from
  • 36:44the genomic data,
  • 36:45the LKB 1 deficient subset in fact
  • 36:47had more of these micronuclear,
  • 36:49suggesting that there are in fact
  • 36:51more chromosomally unstable.
  • 36:52We also got 2 cell lines from Quoc
  • 36:56Wong at NYU where he had, you know,
  • 36:59established the the KP model,
  • 37:01you know which have the K Ras
  • 37:03mutation and P53 mutant.
  • 37:04But on top of that they had developed
  • 37:06a model with deletion of LKB one
  • 37:08and derived A syngenic cell lines.
  • 37:10So we got those into the lab as well
  • 37:13and find that the LKB 1 deficient line
  • 37:15in fact was more chromosomally unstable.
  • 37:18So this seems to be shared between
  • 37:20both the human and the available
  • 37:22best available mouse models.
  • 37:24So now coming back for a second
  • 37:27to these micronuclei,
  • 37:28as I mentioned earlier they are
  • 37:31you know extra material that are
  • 37:33engulfed in these mini nuclei.
  • 37:35But the the,
  • 37:36the envelope of these micronuclei
  • 37:37is very rupture prone.
  • 37:39So the DNA within those is released
  • 37:41at some rate into the cytosol which
  • 37:44of course is an absolute no go
  • 37:46when it comes to you know normal
  • 37:48immunity where you know our bodies
  • 37:50and this is highly conserved are
  • 37:52trained to sense DNA in the cytosol
  • 37:54from all sorts of infections for
  • 37:56example and respond to that.
  • 37:58And and you know cells do that very
  • 38:01efficiently through several pathways,
  • 38:02perhaps the most important one
  • 38:04being the C gas sting pathway.
  • 38:06So here C gas senses cytosolic DNA
  • 38:11and well OK lights off convert,
  • 38:15convert this DNA to C gam which binds
  • 38:17to sting and ultimately triggers a
  • 38:20cascade that typically will result in
  • 38:22the production of type 1 interferons
  • 38:25which of course a very potent
  • 38:27antiviral and anti tumor activity.
  • 38:30Now this is only true when you
  • 38:33activate the pathway very briefly,
  • 38:36but this is not really what
  • 38:38we see in cancer, right?
  • 38:39Because chromosomal instability is perpetual,
  • 38:42this pathway is tonically activated.
  • 38:44And it turns out that when you do that,
  • 38:47you actually flip the
  • 38:48entire pathway on its head.
  • 38:50And and this is what I want to
  • 38:52demonstrate in in in the next few slides
  • 38:54and how we might be able to use this
  • 38:57information to target this process.
  • 38:59So when you tonically activate the pathway,
  • 39:01you in fact see less type 1
  • 39:03inference production and you see a
  • 39:05more aggressive phenotype sort of
  • 39:07flipping the pathway on its head.
  • 39:09So we wanted to test this hypothesis and
  • 39:11we did a couple of simple experiments.
  • 39:13First, we asked you know are LKB 1
  • 39:17deficient cells in fact less capable
  • 39:19of activating type 1 interferon
  • 39:21related pathways And this is what
  • 39:24I'm showing you here.
  • 39:25We have these cell lines that we
  • 39:27stimulated with double stranded
  • 39:28DNA as sort of surrogate for
  • 39:30chromosomal instability.
  • 39:31And then we looked at a couple
  • 39:33of key downstream nodes from in
  • 39:35the C gasting pathway,
  • 39:36phosphor TBK one and phosphor IR 3.
  • 39:39As you can see the LKB 1 proficient
  • 39:41Syn low cell lines are able to
  • 39:44do that rather efficiently while
  • 39:46the LKB 1 deficient lines do not.
  • 39:48This results in a significant
  • 39:50impairment in the LKB 1 deficient
  • 39:52lines with respect to a couple
  • 39:55of important Type 1 interference,
  • 39:56I'm just showing you a couple selected ones.
  • 39:59So suggesting that these,
  • 40:00the Syn high state in fact confers
  • 40:04impaired production of type 1 interference.
  • 40:06So how can we now prove that it is
  • 40:09sin that drives this impairment?
  • 40:11So one way to do this is by
  • 40:14modulating either up or down the
  • 40:16rate of chromosomal instability.
  • 40:18And one way to do this is by over
  • 40:21expressing a variety of different
  • 40:23genetic constructs which had
  • 40:25been previously established.
  • 40:26So in this example we can take a cell
  • 40:30line that is highly chromosomally
  • 40:32unstable at baseline LKB 1 deficient
  • 40:34and over express a gene called MCAC,
  • 40:37also known as KF2C,
  • 40:39which improves the segregation
  • 40:41fidelity that cells have when
  • 40:43they undergo A chromosome.
  • 40:45Segregation in so many words reduce the
  • 40:47number of errors that these cells make
  • 40:50and therefore on a population level,
  • 40:52the rate of chromosomal instability,
  • 40:54which is again measured here as the
  • 40:56number of frequency of micronuclei.
  • 40:58So when you suppress chromosomal instability,
  • 41:01that alone is sufficient to rescue
  • 41:04the ability of these cells to
  • 41:06again produce type 1 interference.
  • 41:09You can do the converse experiment
  • 41:11where you take a cell line that is
  • 41:13relatively chromosomally stable and
  • 41:15express them with a different construct
  • 41:17that makes them more unstable,
  • 41:18as shown here again as a measure
  • 41:21of micronuclei.
  • 41:22And that alone is sufficient to reduce
  • 41:25their ability to produce type 1 interference,
  • 41:28suggesting that it is really
  • 41:30sin that's driving the ability
  • 41:32or inability of these cells to properly
  • 41:34signal through this through this
  • 41:36pathway and produce type 1 interference.
  • 41:38The other way to to approach this of course
  • 41:41is to imagine either deleting genetically
  • 41:44or pharmacologically inhibiting C gas.
  • 41:47And the rationale here is
  • 41:48if you don't have C gas,
  • 41:50then you won't be able to tonically
  • 41:53activate the pathway downstream.
  • 41:55And by at least temporarily relieving
  • 41:57the tonic activation through genetic
  • 41:59deletion of pharmacological inhibition,
  • 42:02you might be able to allow this thing to
  • 42:04come back to an equilibrium where you can
  • 42:07leverage its physiological function which
  • 42:09is producing these important cytokines.
  • 42:11So we tested this in a number
  • 42:13of different ways.
  • 42:14One is by deleting Cgas and then
  • 42:17stimulating the cells with the with
  • 42:19its natural product that is CGAM O,
  • 42:22we delete Cgas.
  • 42:23We let the cells sort of relax for
  • 42:25a week or two and then we ask was
  • 42:27that enough to bring the pathway to
  • 42:29an equilibrium and stimulate it and
  • 42:31show that they're now again able
  • 42:33to produce type 1 interference.
  • 42:35And this is precisely what we find.
  • 42:37So after a few days of of of
  • 42:40of of deleting C gas,
  • 42:42genetically stimulating the cells with C
  • 42:44gaps or stimulating sting in this case,
  • 42:47we see that these cells regained their
  • 42:49ability to produce Type 1 inference.
  • 42:51And this is both true in the human as
  • 42:53well as in the mouse models that we use.
  • 42:56And lastly we,
  • 42:57we tested these concepts also in vivo.
  • 43:00So here we use the KL,
  • 43:02the LKB 1 deficient line and
  • 43:05implanted them into B6 mice.
  • 43:07We treated these animals either
  • 43:09with with isotype control or with
  • 43:11an anti PD1 antibody and just like
  • 43:13in patients there is absolutely
  • 43:15no response to PD1 inhibition and
  • 43:18these LKB 1 deficient Synthi tumors.
  • 43:21Deleting Cgas alone was sufficient
  • 43:23to partly reduce the growth rate of
  • 43:27these tumors but also significantly
  • 43:29sensitized them to PD1 inhibition.
  • 43:32The converse experiment we also did
  • 43:35is taking the KP cell line which
  • 43:38are relatively sensitive to PD1
  • 43:39inhibition as you can see here and
  • 43:41make them more chromosomally unstable.
  • 43:44And this time we show that sin
  • 43:46elevating sin alone is sufficient to
  • 43:49render them resistant to immunotherapy.
  • 43:51This is work that we have done
  • 43:54in collaboration with Sam Bakum,
  • 43:56the other Bakum and and Chrissy
  • 43:59Hong a postdoc in his lap.
  • 44:03Of course we're now interested in moving
  • 44:05these concepts closer to patients.
  • 44:07The problem is that there
  • 44:09is no known, you know,
  • 44:11soluble human selective CS inhibitor.
  • 44:13So we work with our medicinal chemists
  • 44:16to actually develop such an inhibitor.
  • 44:19And what I'm showing you here is the REDUCT
  • 44:22redacted structure of that compound.
  • 44:24And on the bottom,
  • 44:25the functional assay that we use to
  • 44:27determine the activity of the compound.
  • 44:28We're looking at the levels of
  • 44:30CGAMP where we treated the cell
  • 44:32lines that are we've you know I've
  • 44:33shown you throughout this talk with
  • 44:35this with this new compound and we
  • 44:38see that we can pretty potently
  • 44:40suppress the production of of C gas
  • 44:42meaning the activity of C gas.
  • 44:44And just treating the cells with
  • 44:46these with the C gas inhibitor alone
  • 44:50results in reconstitution of these set
  • 44:52of these cell lines to phosphorylate
  • 44:55TBK one and IR three more efficiently
  • 44:58and ultimately result in an improved
  • 45:01ability to produce these important cytokines.
  • 45:05So with that, I want to summarize
  • 45:07this last part of my talk.
  • 45:09Hope I was able to show you that,
  • 45:11you know not all metastases
  • 45:13are created equal.
  • 45:14Brain meds are quite distinct genomically.
  • 45:17That tonic activation of the C gas
  • 45:20thing pathway through sin is is bad
  • 45:23and results in suppression of Type
  • 45:251 interference that can be rescued
  • 45:28through genetic modulation of sin
  • 45:30or inhibition or deletion of C gas.
  • 45:32And that we are, you know,
  • 45:33very interested in moving this into the
  • 45:37clinic using AC gas inhibitor of course.
  • 45:39And I,
  • 45:40you know,
  • 45:41mentioned the people who have
  • 45:42done the work throughout,
  • 45:44but this is their entire lab.
  • 45:45I want to think and of course
  • 45:47all of the collaborators,
  • 45:49both nationally and international
  • 45:50collaborators who we are lucky to work
  • 45:53with and the funding sources that support
  • 45:55this work happy to take questions.
  • 46:06Thank you Ben for a wonderful talk.
  • 46:08I think you're very deserving of the plaque.
  • 46:11While people think what they want to ask.
  • 46:12I'm going to ask you a question
  • 46:14about the CD2 CD 58 access.
  • 46:17If you up regulate CD2,
  • 46:19do you get reciprocal up regulation
  • 46:21of CD 58 because that might
  • 46:22be another way to approach.
  • 46:24If you up regulate CD2 because
  • 46:26you were focused on the tumor
  • 46:28cell with CD 58, but can you
  • 46:29manipulate it via the T cell?
  • 46:31Yes. So we we did the other way around also.
  • 46:34So we when we I guess I should talk here.
  • 46:37So so we did the other way
  • 46:39around where we knocked out CD2
  • 46:41and rescued it in the T cells.
  • 46:44The interesting part there is of course
  • 46:46deletion of CD2 is basically leads to non
  • 46:50responsiveness irrespective of CD 58 status.
  • 46:52But the interesting observation and and and
  • 46:54that might be an artifact of the system is
  • 46:57when you over express CD2 through an ORPH,
  • 47:00then you see actually suppression of CD 58.
  • 47:05So we can't just change the
  • 47:06T cell in that case. Yeah.
  • 47:10All right. Questions.
  • 47:11And I'm sure there's some
  • 47:12online there 50 something people
  • 47:14online I saw some have left bones. I didn't
  • 47:18think the microphones working.
  • 47:21OK Chen. Yeah.
  • 47:24I guess I can click the. Yes,
  • 47:26I can click the very interesting
  • 47:29work. I have a question about the third part.
  • 47:32You have to show those result
  • 47:35in primary tumor setting.
  • 47:37Have you looked at in metastas
  • 47:39setting in brain metastas setting?
  • 47:42No, because we don't have
  • 47:44the right model for it.
  • 47:45And I think that everyone who works on
  • 47:49trying to understand brain metastasis,
  • 47:51this was sort of how we
  • 47:53got interested in this.
  • 47:54One of the challenges is that
  • 47:56to my knowledge at least,
  • 47:58there isn't a good immunocompetent
  • 48:01model that has a sufficiently high
  • 48:03rate of developing brain metastasis
  • 48:05in a number of different ways,
  • 48:07at least through sort of natural
  • 48:09sort of metastatic routes.
  • 48:10So we haven't been able to
  • 48:12actually directly study this in
  • 48:14brain metastasis for that reason.
  • 48:17So I assume some of the young
  • 48:19models are LKB. One loss,
  • 48:21maybe some of those could be positive,
  • 48:26Yeah,
  • 48:30yeah. But even putting the KL the the
  • 48:34mouse cell lines that I've shown earlier,
  • 48:37even if you do an LV injection
  • 48:40or an intracarotic injection,
  • 48:42they don't see the brain for some
  • 48:45reason which we don't really understand.
  • 48:48But what was the second part?
  • 48:49You had another part to your question?
  • 48:52That's yes.
  • 48:53So, so I think it's just a limitation
  • 48:56right now in terms of the available models.
  • 48:59Since
  • 48:59I'm sitting next to Chen, I'll take
  • 49:01advantage to ask a quick question.
  • 49:04So the CMTM 6 in the first part
  • 49:06of the talk is interesting,
  • 49:08but the binding region that you identified
  • 49:10is in the extracellular domain.
  • 49:12I thought it was a stabilizer
  • 49:13binding to the cytoplasmic domain
  • 49:15and the transmembrane protein, but
  • 49:16it's actually it's actually the
  • 49:18extracellular domain and it has two
  • 49:20extracellular domains which we we we
  • 49:23we also mutated and and shown in the
  • 49:25paper that both extracellular domains
  • 49:27of CMTM 6 are required for stabilizing
  • 49:31both PDL one as well as CD 58.
  • 49:33So it's clearly the extracellular
  • 49:35domains that are necessary.
  • 49:36It's a stabilization phenomenon
  • 49:38that is a function of CMTM 6.
  • 49:40So the function of CMTM 6,
  • 49:42I didn't get to the into this
  • 49:44in detail is it's it shuttles,
  • 49:46it shuttles its cargo through
  • 49:49through the recycling endosome.
  • 49:51So typically when proteins are bound
  • 49:53to CMTM 6, they are shuttled back and
  • 49:56recycled back to the cell membrane.
  • 49:59When you lose, I'm shown this in the paper.
  • 50:01When you delete CMTM 6,
  • 50:02you can show that they're more
  • 50:05preferentially lysosomally degraded.
  • 50:19Oh, thank you. Very fantastic talk.
  • 50:22I'm extremely interested in the
  • 50:24the third part of your talk,
  • 50:26the LKP one seems like it's leading to
  • 50:30leads to the chromosome's instability.
  • 50:32But however in my impression now
  • 50:36we think chromosome instability
  • 50:37leads to more like mutation burden,
  • 50:40which is a good thing for you know
  • 50:43response to to immunotherapy.
  • 50:45But in your story seems like it's
  • 50:47on the other way and also you know
  • 50:51some point that we know the PD1
  • 50:53blocking antibody is approved for
  • 50:55all cancer which has the MSI high,
  • 50:58you know your tumors.
  • 50:59So I wonder if you have any
  • 51:01comments between on this.
  • 51:02And also my second question is when you I,
  • 51:07I know you're planning to put
  • 51:09the C gas inhibitor on clinic,
  • 51:12what's your,
  • 51:13what's the approach in your mind
  • 51:15because now we know there's a lot
  • 51:18of steam possibly agonist those
  • 51:20of it has been tested in clinic.
  • 51:22So seems like you know,
  • 51:25I I don't know if you ever discussed with
  • 51:28Thomas Kiosky because he's a you know,
  • 51:33yeah thing guy.
  • 51:35So
  • 51:36I am very happy that you asked both of
  • 51:38these questions because I think it's
  • 51:40really important to clarify a couple
  • 51:42of things and maybe I went too fast.
  • 51:44So to the first point, yes,
  • 51:47TMB to some extent is associated weekly,
  • 51:51but it is associated with
  • 51:53response to checkpoint inhibition.
  • 51:55What I was talking about here is not TMB,
  • 51:59it's not, you know the number
  • 52:01of non synonymous,
  • 52:02you know mutations throughout the genome.
  • 52:04When we talked about chromosomal instability,
  • 52:06a different form of genome instability that
  • 52:10is rather characterized by large changes,
  • 52:12large structural changes,
  • 52:14loss of chromosome arms or gains
  • 52:17of entire chromosomes, etc.
  • 52:19What we were actually able to show,
  • 52:21and that was on the slide but I
  • 52:23went over it probably too quickly,
  • 52:25is that in this particular case of LKB 1,
  • 52:28the mutation is not associated
  • 52:30with a difference in TMB.
  • 52:32It's only associated with a difference
  • 52:35in the rate of chromosomal instability,
  • 52:38at least the the genetic surrogate
  • 52:39of it in in the PCGA cohort.
  • 52:42So I think these are two different forms
  • 52:44of genome instability that are very,
  • 52:46very different. OK.
  • 52:49And then the second part is you
  • 52:52know with respect to sting,
  • 52:55those structures don't work so
  • 52:57well in patients, right?
  • 52:58That's the reality,
  • 52:59right.
  • 52:59The response rate to a sting
  • 53:02agonist is relatively low.
  • 53:03There are and I I had some of them myself,
  • 53:07there are some patients who are
  • 53:09exquisitely responsive to them,
  • 53:11but the vast majority of patients do not.
  • 53:14And I think that the work that I
  • 53:17presented here actually supports why that
  • 53:20might be right because when you have
  • 53:23this tonic activation of the pathway,
  • 53:25as I said,
  • 53:26you sort of you flip the pathway
  • 53:27bit on its head.
  • 53:28And so if you throw a sting agonist on there,
  • 53:31you know it probably doesn't do
  • 53:32very much right to the at least
  • 53:34to the cancer cell and perhaps
  • 53:36it might make the problem worse,
  • 53:37right if if you stimulating the wrong clade,
  • 53:40if you will of the downstream signaling.
  • 53:43What I'm suggesting is that we
  • 53:45might be able to, you know,
  • 53:47rescue some of these in agonists,
  • 53:49but they need to be combined
  • 53:52in the proper way.
  • 53:54So you could imagine that if you,
  • 53:56you know, inhibit sea gas right in
  • 53:58a in a cell line or in a patient,
  • 54:01if you will,
  • 54:02then you allow this pathway to come
  • 54:04back to an equilibrium, you know,
  • 54:06and then coming in with the sting
  • 54:08agonist might be more fruitful.
  • 54:09So I don't think these are
  • 54:11mutually exclusive strategies,
  • 54:12but I think that,
  • 54:13you know,
  • 54:14the work suggests that just
  • 54:15throwing sting agonists on things
  • 54:17will is unlikely to be beneficial.
  • 54:28David, differential output.
  • 54:34How's my treatment quoted? How's
  • 54:36the self burgering out? Atomic CS activity.
  • 54:45Fortunately, somebody figured
  • 54:46that out. It wasn't me,
  • 54:48but it was Sam's lab.
  • 54:50They published a a paper last
  • 54:52year in Nature that where they
  • 54:54show that tonic activity of sting
  • 54:58preferentially induces ER stress.
  • 55:01And through ER stress you
  • 55:03basically you know a tonic ER,
  • 55:06ER stress in this case resulted in the,
  • 55:08in the, you know,
  • 55:10chromatostatic clade.
  • 55:11And they've shown beautifully
  • 55:13in in vitro models,
  • 55:14in fibroblast models that it doesn't
  • 55:16take very much of tonic activation
  • 55:18to get to that point where you
  • 55:20know if you stimulate once you get
  • 55:22a really nice Type 1 interferon.
  • 55:23But even after two or three
  • 55:25stimulations over a couple of days
  • 55:27you very quickly sort of you know,
  • 55:29tacky for LAX on that pathway and
  • 55:31you start activating the the bad
  • 55:33counterpart which is in this case
  • 55:34the chronic ER stress activity.
  • 55:37OK.
  • 55:38I'm going to give the last
  • 55:39question to Ben and Lou.
  • 55:42Perfect.
  • 55:46OK,
  • 55:53we we did not, we actually looked at that.
  • 55:55That was one of my students
  • 55:56just said I'm going to throw it
  • 55:58on the on the flow cytometer.
  • 55:59We don't see, we don't see changes and
  • 56:01see the 58 level depending on Syn.
  • 56:05Ben, thank you for a wonderful talk and be
  • 56:10around for a few minutes if anyone
  • 56:11has extra questions. Thank you.