Microenvironmental Determinants of Systemic Therapy Response in Kidney Cancer: from human to mouse and back

March 15, 2024

Information

Yale Cancer Center Grand Rounds | March 15, 2024

ID11479

To CiteDCA Citation Guide

00:00Real pleasure to introduce Doctor Ari
00:02Hakimi as today's Grand Round speaker.
00:05He's the an Associate Professor
00:06and Co leader of the Translational
00:08Kidney Cancer program and Memorial
00:10Sloan Kettering Cancer Center. Dr.
00:12Hakimi is a urologic surgeon who's
00:14focused on the care of patients
00:16with urologic malignancies,
00:17especially kidney tumors.
00:18He received his medical degree in
00:20residency training from Einstein
00:22College of Madison and completed
00:24his fellowship in Urologic Oncology,
00:26Oncology and Memorial Sloan
00:27Kettering Cancer Center.
00:29His research really aims to
00:30understand immune infiltration,
00:31inflammation in the tumor microenvironment
00:33in RCC and to identify novel
00:35therapeutic targets to overcome
00:37resistance to systemic therapy.
00:39His studies apply bulk single
00:41cell and spatial RNA sequencing,
00:43flow cytometry and immunogenomic analysis.
00:46Really to understand both patient
00:48samples and a novel immunocompetent
00:50kidney cancer mouse line that his
00:52lab has developed really has been
00:53a game game changer for the study
00:55of kidney cancer and particularly
00:57the immunobiology of kidney cancer.
00:59I followed Ari's work for many,
01:01many years now and to see him go
01:02from what was a rising star in kidney
01:04Cancer Research to now really one of
01:06the world leaders has been a pleasure.
01:08And it's a a body of work that
01:09I've tremendously admired.
01:10So it's really a pleasure to be able to walk.
01:12Welcome Doctor Akimi to grand rounds today.
01:21All right. Thank you so much.
01:22And it's really a pleasure to be here
01:24at Yale and especially I was especially
01:27enthusiastic come here because of David
01:31and I mean talk about rising stars.
01:32David is incredible and he's got great
01:34mentors here with with Harriet and others.
01:36And I think it's it's a real pleasure here.
01:39So that's my only disclosures,
01:40none of which is pertinent
01:42to this talk today.
01:43So I'll talk a little bit about the genetic,
01:46the genomic and genetic background
01:48of kidney cancer,
01:50in particular clear cell renal cell
01:52carcinoma which is the most common and
01:54aggressive form of kidney cancer but
01:55also one of the most immunoresponsive.
01:57We'll talk a little bit about the
01:59role of the micro environment as a
02:02predictive response and really focus
02:03on myeloid compartment which is one
02:06of my lab's interests a little bit
02:08from the genomic determinants of this
02:10and then some of the insights using
02:12both human and mouse strategies to
02:15understand this for for future targeting.
02:17So kidney cancer is about is the
02:206th most common cancer overall or
02:23eighth most common cancer overall,
02:256th most common in men.
02:26There's a 2 to one gender difference
02:28and we know that you know within
02:30the kidney there are many,
02:31many subtypes of kidney cancer and even
02:33if you just took the most common subtypes,
02:36they're very genetically different.
02:37Clear cell represents the most common form.
02:40About 6570% of all kidney tumors are
02:43a clear cell and we know the most
02:45about it from a genetic standpoint.
02:47But we also,
02:48there's also some very intriguing
02:49phenomenon that existed in it,
02:50particularly the amount of the
02:52immune response.
02:53And it's still not clear why these
02:55tumors are so immune infiltrated and
02:56and why they are so responsive to
02:59immunotherapy compared to other tumors
03:01that are much more highly mutated,
03:03for example like melanomas or or
03:05bladder lung cancers where you see
03:06a lot of mutations in those tumors.
03:08And probably you know everyone
03:11references TCGA papers initially in
03:14terms of the fundamental understanding.
03:16And I think some of the takeaways
03:17from this tumor from this analysis
03:19which is one of the first tumors to be
03:21profiled was that it's really dominated
03:22by a few driver mutations related to
03:25tumor suppressors On the 3P locus.
03:28There's not a lot of mutations.
03:29There's some copy number events that
03:31are really fundamental and maybe some
03:33of which are enriched in metastases.
03:35But there's not an obvious clue as
03:37to you know why these tumors retain
03:40such a high immune infiltration.
03:42We also know a little bit about what happens.
03:44Thanks to Seminole work from the
03:46Sanger Institute where they looked
03:48at what were the fundamental events
03:49that that are associated with clear
03:51cell nasal carcinoma development.
03:53And what this paper showed for the
03:56really the first time was that
03:58the loss of chromosome 3P1 arm is
04:02critical to the oncogenesis and then
04:04that's followed by VHL loss whether
04:06it's mutations or or methylation.
04:08But basically 90% of all clear cells
04:11have this and then eventually over time
04:13additional driver mutations are lost and that
04:16leads to different evolutionary subtypes.
04:18And in this paper Samara Trashlik and others
04:22came up with a relatively complex schema.
04:25But you know, fundamentally you can
04:27think about it as the tumors lose
04:29VHL and then they usually acquire
04:31one or two additional hits to form
04:33into sort of different trajectories.
04:35We know that tumors that lose PBM one and
04:38set D2 for example maybe more angiogenic,
04:41they may they may tend to be a
04:43little bit more indolent overall,
04:45maybe possibly more responsive to certain
04:47therapies like including even immunotherapy
04:49although that's not entirely clear.
04:51And then BAP one mutations which occurs
04:53well are are typically associated with
04:55more high grade aggressive proliferative
04:57tumor types and then you can have multiple
05:00clonal drivers which also represent a
05:02very aggressive form of kidney cancer.
05:03So we're starting to get a better
05:05framework for the underlying genomics,
05:07but none of this really has been
05:08shown to be targetable.
05:09So for years we just really started to
05:12understand what was driving kidney cancer,
05:14but we really didn't,
05:15wasn't giving us any further insights,
05:16weren't an oncogene that you
05:17could develop a target to.
05:18And while people are certainly working
05:21on epigenetic regulation and strategies,
05:24it's it's certainly not an obvious pathway
05:27forward in kidney cancer at least right now.
05:29And at the same time,
05:30we also knew clinically that that
05:32you know most of the targeted
05:33therapies were limited to the micro
05:35environment of the of the cancer.
05:37And really we've seen a tremendous
05:39growth in in outcomes and survival
05:41for kidney cancer patients.
05:43But it's all really focusing
05:44on the micro environment.
05:45So even back in,
05:46in the 90s when we were studying IL 2
05:49both in melanomas and kidney cancers,
05:51that was really the only
05:53treatment that seemed to work.
05:54We had tried all the chemotherapies
05:56you can imagine in the 80s,
05:5890s and you would see responses
06:01even 7 to 10% cures,
06:03but with very,
06:04very high toxicity in those
06:06populations and in those patients.
06:08And it wasn't until the advent of
06:10of really by work from Bill Kalin
06:13and others where we recognized the
06:14importance of Hifs that all these
06:16focus on VEGF had come around.
06:17And then it wasn't really until this
06:20notion of immunotherapy came around that
06:23we started to see this big revolution
06:25in terms of survival and outcomes.
06:27But it's all really focused on,
06:28on the micro environment.
06:31We know that you can take some
06:33of these genetic events that I
06:35mentioned earlier and risk stratify
06:37patients a little further.
06:38This is the work that we did several
06:40years ago now looking at the impact
06:43of some of these common mutations
06:45and outcomes for patients that were
06:47receiving VEGF therapy and it was
06:49you know prognostic maybe you could
06:51further stratify patients that
06:52were grouped into clinical risk
06:54groups by by common mutations.
06:56But it wasn't really telling us anything
06:58about the underlying immunobiology or
07:01angiogenic biology in these tumors.
07:02It was really more just a
07:04prognostic feature about it.
07:06So with this revolution of
07:09immunotherapies both by themselves
07:11and in combination with VEDF therapy,
07:13we've seen the the survival rate and
07:15the response rates go up dramatically.
07:17So you know the median survival when
07:18when I first started my training
07:20for metastatic kidney cancer was
07:22you know a year and a half or so
07:24and now we're pushing five years
07:25for for a lot of patients and and
07:28potentially curing some patients.
07:30And we there's a real need to
07:32understand why that's the case and
07:34how we can do better because we
07:35know that invariably most patients
07:37despite this high response rate will
07:40eventually develop resistance and you
07:43know understanding why that's the case
07:45requires you know good models to do.
07:47So we know that from an immunotherapy
07:51standpoint that it's not ATMB driven
07:53tumor at least not obviously and
07:55maybe you can break down the types
07:57of mutations a little bit more.
07:58I was just talking to David about
08:00this last night,
08:01but you know there's it's not obvious.
08:04There have been several attempts
08:05to look at TMB as a predictor for
08:07responses and most of the large clinical
08:08trials that have been performed that
08:10have have released their data have
08:11not shown this and certainly David
08:13has been on the forefront of this.
08:14But if you look across some of
08:16the major phase three trials that
08:18have at least released their data,
08:20there's not a really a signal at all
08:22with respect tumor mutation burden.
08:23We've looked in David and others have
08:26looked at whether mutations in PBM one
08:28or loss of nine P which is a common
08:31event in metastatic kidney cancers,
08:33whether that's associated with
08:34immune infiltration patterns.
08:35There may be some signals there.
08:38It's not a clear biomarker though and
08:40and that sort of has been lacking from a
08:42mutational and copy number standpoint.
08:44So I think one of the things that
08:46is unique about kidney cancer is
08:47that you know we've started to look
08:49many years ago now at transcriptomic
08:51predictors because the mutations
08:53are clearly and copy number of
08:55events are clearly not sufficient to
08:57determine who's going to respond at
09:00least from a biomarker standpoint.
09:01And you can just take a very simple
09:04metric of the uniqueness of kidney
09:06cancer and this I just plot out you
09:08know VEGFA and CD8 infiltration.
09:10This was an older slide,
09:11but I like showing it 'cause I think it
09:13shows the uniqueness of kidney cancer,
09:15clear cell,
09:15at least with respect to some of
09:17the micro environmental genes.
09:19And we know that they're just
09:21dominated by high infiltration of CD8
09:24cells and high angiogenic programs.
09:26So the question of course and this
09:27was shown across cancers and it's
09:29really distinct from its from the,
09:31from the normal tissue.
09:32If you look at for example lung cancers,
09:35many,
09:36much of the lung itself is very
09:38mean infiltrated
09:38likely due to smoking or other
09:40other carcinogenic features.
09:41But the kidney itself,
09:43the normal kidney is not
09:44particularly mean infiltrated,
09:45but the tumors are often very
09:47dramatically infiltrated.
09:48So there's something very distinct
09:50about the actual tumor itself
09:52rather than the underlying organ
09:54that it's that's derived from.
09:56This is work we did when when I
09:58was just starting out and we we
10:00you know we we we used immune
10:02deconvolution strategies to show this.
10:04You could take signatures for
10:08T cells or for macrophages,
10:11NK cells etcetera.
10:12And you can start just deconvoluting
10:15bulk RNA sequencing data and start
10:17to try to understand where tumors
10:19or particular samples might might
10:21fall in a spectrum and and you
10:23can use this to also subgroup.
10:25So we use this strategy to kind of
10:27think about tumor micro environmental
10:29subgroups within kidney cancer and
10:30this was our first attempt about
10:32eight or nine years ago to look at you
10:34know whether there's these enriched
10:36groups and whether there's you know
10:38other groups within kidney cancer.
10:40Maybe that would sort of explain why you
10:42see some some really great responses
10:43in the streaming top And we we did
10:46see that we saw you could clearly see
10:48these T cell infiltrated clusters.
10:49You could see at this point we really
10:51didn't think about angiogenesis,
10:52but in retrospect you know you've
10:54seen angiogenic cluster and we'll
10:56talk more about that in a minute.
10:57And you know it did correlate with
11:00certain genetic programs mostly
11:01antigen presenting machinery programs.
11:04So there was an up regulation of antigen
11:07presenting machinery transcript,
11:08but it wasn't clear still from this
11:10point what was actually driving this.
11:12We we looked at genetics,
11:15common mutations and wasn't at
11:17least obvious at the time when
11:18we first did this study,
11:19although that's evolved a bit and
11:21that same approach was applied by
11:23Genentech when they first published
11:25and then analyze the EMOTION trial.
11:27This was the first attempt to
11:29combine VEGF and IO therapies.
11:30They used tizolizumab and bevacizumab,
11:33which is a VEGF monoclonal antibody.
11:36And this,
11:36this trial was negative in terms
11:38of improving the standard of care,
11:40but it was biomarker.
11:41Biomarker Rich and I give a lot of
11:44credit to Genentech for not only doing
11:46phenomenal genomic work but also
11:48making it all publicly available,
11:50which is something that other companies
11:51have have been a little reluctant to do.
11:53So and maybe it was because it
11:55was a negative study,
11:55they were willing to share so much,
11:56but it really was very helpful.
11:58And David McDermott and others from
12:02Boston performed really the first
12:04type of analysis in the context
12:06of systemic therapy to show that
12:08these micro environmental groups
12:10angiogenesis and T cell infiltration
12:12and myeloid programs may stratify
12:14patients into different groups but
12:16also may associate with response.
12:18And one thing I would point out I
12:20think it it it's sort of logical
12:21that a tumor that may have a lot of
12:24T effector cells would respond well
12:25to amitotherapy.
12:26But you know what they also showed
12:29was that myeloid populations as as
12:31determined by again a gene program
12:33we're we're driving resistance also.
12:36So you could be AT effect or
12:38high tumor but if you had a high
12:41myeloid program it could supersede
12:42that impact and and and in fact
12:45actually show you know dramatically
12:46different responses in that context.
12:48So that really sort of laid the groundwork
12:50for not only the micro environment
12:52relevant but also it could it could
12:54predict responses and maybe give us
12:56some insight into into resistance.
12:58We applied that same strategy initially
13:01to just a VEGF cohort only and this was
13:03work that I did with with Bob Moecher,
13:05one of my mentors at at Sloan
13:07Kettering for many years.
13:08And this was just looking at
13:10the first trial that compared
13:12two different VEGF inhibitors.
13:14At the time we had really sunitinib
13:16and this was the first attempt to try
13:19a different strategy or a different
13:21VEGF inhibitor and you know it was
13:23really more to look at tolerability.
13:26There was no difference really in
13:27responses but actually Piszopinib
13:28but this you know showed a better
13:31toxicity profile for patients.
13:32So that became the standard of
13:33care ELISA Memorial for many years
13:35until obviously we developed next
13:37generations and immunotherapies.
13:38But at this time we took a
13:40transcriptomic approach.
13:40So we had microarray data from
13:44Novartis and we had looked at this
13:46question of whether you can identify
13:47subgroups and we found you know 4
13:49transcriptomic subgroups at the time
13:51they tended to really stratify patients.
13:53Again,
13:53all these patients received EDF first line,
13:55so clean,
13:56clean cohort and you could clearly
13:58see a difference in in groups.
14:00And there was a green group here
14:01that was very responsive and a red
14:02group that was very not responsive.
14:04And then there was these yellow and
14:05blue in the middle and the green
14:07group really had a lot of angiogenic
14:08program and that made sense, right?
14:10You know, if you have a lot of angiogenesis,
14:13it makes sense that you would
14:15respond very well.
14:16But the red group which was #4,
14:18that was the one that didn't respond well.
14:20They were actually the worst but they had
14:22the second highest amount of angiogenesis.
14:23So why why was that the case?
14:24Why didn't they stratify nicely by
14:27by angiogenic program and and when we
14:29compared that group to the other groups,
14:31we could see that it was really
14:32being dominated by a myeloid program.
14:34So despite having high angiogenic
14:37phenotype or transcript,
14:38they were they were reversing the
14:41response based on an infiltration of
14:44myeloid myeloids at least inferred
14:46by by microarray data.
14:49So,
14:50so that suggested that it could
14:53actually be driving a response overall
14:56and the micro environment may be
14:58useful in understanding biomarkers in in,
15:01in metastatic kidney cancer.
15:03And then actually this was something
15:04that we did and at the end and
15:05actually the my fellow at the time,
15:06one of the urology fellows at
15:07the time who was working in,
15:08in my group actually had the
15:10suggestion that we look,
15:11we kind of left them all together
15:13because there's open Evans Sunitnib,
15:14we're sort of both VEGF inhibitors.
15:16But we know that the TKIS target
15:18lots of different kinases,
15:20not they're not super specific
15:22and actually if you look at the
15:25macrophage and angiogenic groups,
15:26you could actually see that
15:28Pizzopinib has quite a different
15:31stratification than tsunitiv.
15:31This suggests to us we we didn't
15:33talk about too much in the paper,
15:35but it it really suggested that the
15:37targets of these TKIS may also actually
15:39be driving some of their responses.
15:40So some of these kinases are present on
15:43on immune cell populations for example.
15:45And the fact that these biomarkers
15:47were actually different with respect
15:49to the different Tkis was something
15:51that that we've now followed up on.
15:53And I think it's a really
15:55interesting finding that he made.
15:56And this just leads back to the
15:58same concept that what what Dave
16:00showed in in this in this beautiful
16:02paper from the from the Genentech
16:04study others have looked at micro
16:06environmental features in other
16:08in other more positive trials.
16:09So this,
16:10this was also sort of a negative,
16:11well not negative but has not has
16:14not been brought forward further.
16:16This was the Javelin 101 study again
16:20avolumab and exitinib again another
16:22combination of PD one and and and VEGF
16:26and they focused on a a lymphocytic
16:31signature identified 26 genes again
16:34specific signature for specific trial.
16:36Some of the challenges of these have been
16:38you know applying it to other data sets.
16:39But again you could see the the
16:42micro environmental features being
16:44associated with response here
16:46and suggesting you know that we
16:48could utilize this as a strategy.
16:50And then there have been subsequent
16:52efforts by by collaborative
16:55groups including Genentech again
16:57to integrate what we know about
16:59mutations and those evolutionary
17:01subtypes I showed you earlier into
17:03and the micro environmental feature.
17:05So if we have the micro environment
17:07and we know the genetics that are
17:09that arise as kidney cancers evolve,
17:12could they could they be sort of
17:14grouped together to form these kind
17:16of different molecular subgroups?
17:18And I think there's been some
17:19attempt to do this.
17:20I think it's improving,
17:22but you know you sort of have the sense
17:24that there are these different angiogenic,
17:26stromal and angiogenic alone.
17:27So some of these may have this
17:29myeloid phenotype that I showed you
17:31earlier and just a purely angiogenic
17:33tumor maybe the purely angiogenic
17:35tumors would respond really well to
17:37VEGF alone and those are tend to
17:39be the less aggressive tumors PBR 1
17:41mutated and then you have the ones
17:43that are myeloid and angiogenic
17:45and those actually don't respond
17:46at all to to VEGF inhibitors and
17:48then you have these proliferative
17:49ones and and other ones.
17:50So we're starting to get a sense
17:52that maybe you can subgroup kidney
17:54cancers into those features.
17:55And then came along this
17:57notion of single cell.
17:59And there have been a series of papers,
18:01one of which David LED,
18:02but that came out from the time
18:03because we had done all of all
18:05this work on bulk RNA sequencing.
18:07And as the fields across
18:09cancers have evolved,
18:10we started utilizing single cell to not only
18:13get a better sense of what was happening,
18:17but also really understand you know
18:18what are the specific features of these of,
18:21of the of the micro environment
18:22in a much more high resolution.
18:23So the advantage of bulk RNAC of
18:25course is that you can do big,
18:26big numbers of samples because
18:28it's relatively inexpensive.
18:30Single cell gives you deep dive,
18:31but often the cohorts were
18:32much more modest in size.
18:34So I think there's constantly
18:35a need to go back and forth.
18:36If you find a signal in one,
18:37you have to validate in the
18:39other so to speak.
18:40And so we we we did this in in
18:44clear cell focusing really on
18:46patients that had received just dual
18:49immunotherapy with PD1 and CTLA 4.
18:52We focused on 6 patients initially.
18:55When we did this together with Ming Lee,
18:57one of my immunology mentors,
18:59Christina Leslie,
18:59who's a computational biologist and a very,
19:01very talented graduate student at the
19:03time who's finishing his post doc at Harvard,
19:06now Shirag Krishna.
19:07And we looked at patients that had
19:10either were were high risk and not
19:14had not received PD one right away or
19:17versus ones that had had Ipinivo. Yeah.
19:21And were eventually went on to surgery.
19:24One of the unique things I I do
19:25as a surgeon is that we're able
19:27to get tissue after treatment and
19:28kidney cancer has evolved so much
19:30so because of the response rates
19:32now to upfront immunotherapy that
19:33most patients if they come in with
19:35metastatic disease will get upfront
19:37systemic therapy and then we're being
19:39asked to operate on them later on.
19:41So that gives us a unique opportunity
19:43to study tissue after treatment,
19:44which is something I think really
19:46unique to kidney cancer amongst many
19:48solid tumors we have this opportunity.
19:49So we were able to utilize that
19:51strategy here and this sort of
19:53gave us a broader sense and this
19:54has been replicated I think by
19:56many other single cell studies
19:57including David's really Seminole
19:58work in this and you can kind of
20:01get a sense of what's happening now.
20:02One of the things that's interesting
20:04about a quirk of single cell is that
20:07you know there's for those of you
20:09familiar with the technology is that
20:10you know there's generally at least if
20:12you do single cell and not single nucleus,
20:14you have to do some sort of
20:16sorting and a lot of the.
20:18Tumor cell populations actually
20:19die die off from that process.
20:20They're very fragile for some ironically
20:23and immune cells will often survive
20:25although you lose neutrophils.
20:27So kind of interesting quirk
20:29of any sort of single
20:31cell study that you do, you lose a
20:33lot of the cancer cell populations.
20:34But we're able to kind of get a
20:35good sense of what's going on in the
20:37immune cell population and you get
20:38a general sense that and this has
20:39been replicated by our flow analysis
20:41over the many years that about 6040
20:44to 60% of the immune compartment is,
20:46is made-up of T cell and you have
20:49a good amount of Tams in this.
20:51And then a whole bunch of other
20:53populations including B cells and K
20:55cells and dendritic cell populations,
20:57but they're really dominated by these T
21:00cell and and and and and Tam populations
21:03and you could further phenotype them
21:04into you know and this has been done.
21:06You know,
21:07everyone's got their own slightly different
21:08way of of phenotyping populations,
21:10but this allows to sort of get a
21:13sense of what's happening, yeah,
21:14in both primary sensitivity and
21:16and primary resistant patients.
21:18And you know this one again we
21:20had epinivo resistant and a mixed
21:22response and a complete response,
21:24complete response patients are always
21:25interesting because why are we operating
21:27on them if they have a complete response?
21:28Well, when I say complete response,
21:30I mean that the tumor's
21:32mass itself is not viable.
21:33It's it's it, it there's a mass there,
21:35we we take it out.
21:36But actually under the microscope,
21:38there's no tumor left.
21:39It's just a conglomerate of
21:40immune cells and fibroblasts.
21:43And so that's kind of an interesting
21:44population to look at because
21:45what's what's residual there.
21:46And there we found these tissue
21:49resident T cell populations that
21:51were very abundant in the in the
21:54in the residual mass of the of the
21:56of that of that kidney even though
21:58it was there was no tumor left.
22:00And then we found within the
22:01patients that were not responding
22:03it really you know there was T cells
22:05there but it was really dominated
22:06by specific Tam populations.
22:08This was a primary resistant patient.
22:09He had had big tumor, big lymph nodes.
22:11We gave him a trial with Epinivo to
22:13see if that would help and it didn't.
22:14So we still end up operating on him,
22:16no response in the tumor whatsoever.
22:18And you could see this was really
22:20a Tam dominated tumor type.
22:21So it started giving us insights
22:22that really this may be associated
22:24again small numbers.
22:25So you really have to start building
22:27that out and you can develop
22:28signatures which is what we did.
22:29We actually took the single cell genes
22:32and from the different clusters and
22:34overlaid them on some of these genomic
22:37signatures that have been published.
22:38The javelin when I mentioned the the,
22:40the,
22:40the genomic ones from Genentech and we
22:43started saying like what are the actual
22:45populations that they're capturing.
22:47You get a better sense that there
22:49are some dominant Tam populations
22:51and that some of these populations
22:53may be potentially targetable.
22:54And I'll talk about that in in a
22:57moment. But I want to bring your
22:59attention to some of these adenosine
23:01signatures that were were published
23:03from an HUAR inhibitor which is
23:05something that has been shown to
23:07potentially shift Tam phenotypes.
23:08So, so this was sort of an interesting
23:10way for us to look at the data and
23:12you can develop signatures based on
23:13the single cell data and compare them
23:16to existing signatures to see if you
23:19could further stratify patients and
23:21responses across different different data
23:23sets And and we were able to do that.
23:25And then the question is also, well,
23:27is there a relationship between the
23:29underlying micro genetic micro environment
23:32and these specific immune micro environments?
23:35So I I showed you again on bulk,
23:37maybe there's these different
23:38sub classifications,
23:39but we also know there's a lot of
23:41heterogeneity in kidney tumors, right.
23:42So we know that if you took a kidney
23:44tumor and you sequence different regions,
23:46Charlie Swanton showed this many years
23:48ago in a famous paper New England Journal,
23:49intratumal heterogeneity exists.
23:51Does that same thing apply to
23:53the micro environment as well?
23:54And that's something of of course
23:56if you're going to develop a
23:57biomarker or suggest something,
23:59you have to think about that.
24:00And this is work that we did in
24:02collaboration with Illumina where we
24:03really thought about this question of,
24:05OK,
24:05now we have a good sense of what's
24:06going on in the micro environment.
24:08We have a good sense of what's
24:09going on in the underlying
24:10genomics and and how does that,
24:11how does that relate to the individual tumor.
24:13And one of the reasons why clinically
24:15that's interesting is 'cause if you
24:17look at at at data sets where the the
24:19primary tumor's still in place in
24:21the with the patient with metastatic
24:22disease and they get immunotherapy,
24:24often the Mets will respond well,
24:25but the primary tumors maybe
24:27only shrink modestly, right.
24:28So there's not a,
24:29there's not that same level of response.
24:31And one hypothesis is that,
24:32well,
24:33it's because the primary tumor is
24:35more clonally diverse and the Mets
24:36is just a a clone that was able to
24:38metastasize out that was selected for.
24:40So when you get a response in the
24:42Mets maybe it's because there's just
24:44a clone that's really responsive,
24:45but the primary may only have that
24:47clone in in part of it and sort
24:48of been our rationalization for
24:49continuing to operate on these
24:51patients because I tell them well
24:52a you know I like operating.
24:55But more more fundamentally it's
24:57actually because we think that you
24:59know we're we're removing the diverse,
25:00the biological diversity of them Even
25:01if they've had a good response up front,
25:03the chance for them to develop persistence
25:05down the road may come from from the primary.
25:08And so we tried to look at this with
25:10multi regional sequencing again
25:12relatively modest numbers but we
25:14we utilized the combinations of DNA
25:16and RNA and and TCR and different
25:18things within within tumors that
25:20had been exposed to
25:22immunotherapies as part of a
25:24trial that we ran and and others.
25:26And so we were able to look at the
25:28question of whether overall immune
25:30diversity is associated with,
25:32I'm sorry, overall genetic diversity
25:34is associated with particular micro
25:36micro environmental phenotypes.
25:37Indeed, we found at least in this
25:39study that if you were a very high lead
25:41diverse tumor from a genomic sample,
25:43you were more likely to be
25:44a myeloid high tumor,
25:45which was interesting and and vice
25:48versa with respect to some of the
25:51antigen presenting machinery genes.
25:52So the ITH tumors were actually
25:55lower with respect to the APM genes
25:57and actually if you took a specific
25:59tumor and you actually marked out
26:02the immune infiltration patterns,
26:04you could start seeing evolution
26:05within that same tumor.
26:06So this is an example of a of
26:08a tumor that was I TH high.
26:09It had a lot of intratumoral heterogeneity.
26:11It was AP Bear Monsanti 2 kind of
26:14micro environment or evolutionary
26:16subtype and we were able to look
26:18at individually in different
26:19regions of this tumor to see.
26:20We found that some of the regions
26:22were very T cell infiltrated at
26:23least by RNA and some of them
26:25were very mild and infiltrated.
26:26And we were actually able to
26:27track down like what were the
26:28individual genetic events that were
26:30occurring as this tumor developed.
26:31And you could see that, you know,
26:32as the tumor developed,
26:34there was loss of HLA and maybe some
26:36CDK into A&amp;B loss which has been
26:38associated with a more immune desert
26:41or less immune infiltrated micro.
26:43So within the same tumor itself,
26:44you could see this evolution and
26:46that was correlating with the micro
26:48environmental features suggesting
26:49that there's this constant interplay
26:50And I think David has shown this and
26:52others have suggested this constant
26:54interplay between the underlying Genoma
26:56architecture of a tumor and what's actually,
26:58you know,
26:59underlying the response micro
27:00environmentally in that tumor.
27:01Obviously we don't fully tease
27:03out the mechanism here at all,
27:04but it begs the question that there's
27:07an opportunity here to to explore this
27:10and what about in in the metastatic question.
27:11So that was another
27:12question we had in the lab.
27:14So I've showed you everything in terms
27:15of treatment response potentially.
27:17But we also wanted to know are are
27:19these micro environmental groups
27:21also predicting or or associating
27:23with development of metastas,
27:24which is a different question,
27:25right.
27:26You know you could have a micro
27:27environment that's really
27:28important for treatment response,
27:29but it may not be associated
27:31with metastatic development.
27:32So,
27:33so David had had somewhat hinted at
27:36this with his work with with Ellie
27:38and Tony and others at Dana Farber
27:41and they showed it you know very
27:43elegantly in this paper that looked
27:45our paper focused on the advanced
27:47disease and and Ibidevo treated.
27:49But David's work was performing single
27:52cell sequencing on early locally
27:54advanced in metastatic tumors and
27:56at least in this work he showed this
27:58evidence of T cell exhaustion but
28:00also this shift in the macrophage
28:01polarity as tumors became more
28:03aggressive, more metastatic.
28:04So suggesting to us and others that
28:06you know maybe some of these Tams and
28:09and myeloid populations that were so,
28:11so, so driving responses are also
28:14associated with metastatic development.
28:16And so for this we again utilize
28:19our strategy with with going
28:21to clinical trials and and we
28:23worked with this adjuvant study.
28:25So this was one of the series of
28:27negative studies unfortunately,
28:28but again the benefit of having a lot
28:30of genomic data looking at whether
28:32giving it a VEGF inhibitor monotherapy.
28:34Again Pezopinib in this case was
28:36associated with better outcomes
28:37in patients with high risk non
28:39metastatic kidney cancer.
28:39All these patients in this trial
28:41had advanced kidney cancers.
28:42They had a high risk of relapse but
28:47but you know standard at the time
28:48was just to observe them and so there
28:50was a series of trials to see if you
28:52gave a VEGF inhibitor whether that
28:54actually was improved their survival.
28:57The vast majority of the studies were
28:59were -1 was sort of positive but no
29:01one has really adopted it and now
29:03we've moved on to immunotherapy but at
29:05this time this was a very interesting study.
29:07So we we compared we had microarray
29:09yet again from Novartis,
29:10we compared the the again all high risk.
29:13So they're they're you're sort
29:14of controlling for the potential
29:16tumor confounding features right.
29:17They're all high risk patients and
29:18we compared the ones that relapsed
29:20versus the ones that didn't.
29:21This is work that one of our one of
29:23our fellows LED and who who's now at
29:26Rochester with a surgeon scientist
29:28track their great guy Phil Rippolt
29:30with with Lynn Von from my lab who's
29:33a senior member now and we compared
29:35the the tumors that record versus
29:36didn't and we used an unbiased you know
29:38whole genome approach with with this.
29:39And so of course you saw things
29:42like EMT and mtor,
29:44which made sense to us because
29:45those are obviously very known
29:47and relevant oncogenic processes
29:48that that promote metastases.
29:49But we also saw a lot of immune
29:52inflammatory genes in particular Illinois
29:546 and Jack in Stat 3 kind of stood out
29:57to us as being a driver of metastasis.
30:00Again,
30:00we applied the same single cell
30:02strategies what we had done before to
30:04see you know what are these myeloid
30:06inflammation and Illinois 6 pathways,
30:09what are they really converging on?
30:10And and and indeed it it showed a
30:14real enrichment in some of these tan
30:16populations that we had defined a
30:18little bit better with the single cell data.
30:21And suggesting that if you were a tumor
30:23that was myeloid high or adenosine high,
30:25similar overlapping signatures,
30:27you are more likely to develop metastasis.
30:30Again controlling for other features,
30:32all of the clinical and pathologic features.
30:34These are completely independent programs
30:37and you could show that if I mean if
30:40you were a AMSK inflammatory signature,
30:42which was a gene signature we developed
30:44from from the micro RAY data strongly
30:46overlapping with the myeloid signature
30:48from Genentech and others that you could
30:51take all these high risk patients.
30:52And really I mean it's pretty amazing
30:54to see curves like this separate out.
30:56But again all of these patients,
30:58this is a myeloid load tumor intermediate
31:00and then very high and you could
31:02see their survival probability.
31:03And then we were able to replicate this
31:05in multiple other data sets including
31:07from one of our former fellows who's
31:09at Moffett now and and again showing
31:11that if you were controlling for
31:12all these high risk features from
31:14a clinical pathologic standpoint,
31:16you you could still stratify patients
31:17by the risk of recurrence in that
31:20and it didn't seem to be associated
31:22much with the T cell response.
31:24So what was driving metastas is
31:26at least in this data set,
31:27but again it has been valid in others,
31:29was not really AT cell driven process.
31:31What was at least from a micro
31:34environmental standpoint promoting
31:35metastasis was was independent seen.
31:37We tried all the different T cell
31:38signatures that have been shown.
31:40We looked at IHC scores,
31:41we had IHC from CD8 infiltration
31:44patterns here.
31:44We were able to see if they were inflamed
31:46or excludeded and we really didn't see me.
31:48Maybe there's a signal with the
31:50desert that those are the tumors
31:52that have no T cells at all,
31:53but it wasn't clear at least that was
31:55that wasn't the clear driver here.
31:57It was really much more of the myeloid
31:59and Tam phenotypes and actually the
32:01angiogenic tumors that were low
32:03were also similarly associated,
32:05not quite as clean of a signal,
32:06but it's certainly it looks like if
32:08you're a low angiogenic tumor you're
32:10you're much more likely to recur.
32:11So this suggested that these micro
32:13environmental features also might
32:15be useful for adjuvant strategies
32:17and indeed a lot of the work now
32:19that's going forward in some of the
32:21newer adjuvant trials are factoring
32:23in things like these transcriptonic
32:25signatures into risk adapted models.
32:26And I think you know the future of
32:28course would would would incorporate
32:30some of these into selecting not only
32:32who's going to recur or not but maybe
32:34whether you give them a combination
32:36strategy or a single agent drug.
32:37But how can we test this.
32:38So ultimately you know I I show a lot of
32:41like nice kind of genomic correlative
32:44work and maybe some evolution of how
32:46we think about kidney cancer from a
32:49micro environmental and genomic standpoint,
32:50but really how do we test this.
32:51And so this is the challenge that I faced.
32:53I was kind of writing all these
32:55papers and thinking about this a lot
32:56and getting advice from mentors and
32:58everyone kept saying well you got to,
32:59you got to,
33:00you got to find a model that that works.
33:02And so you know we didn't have a
33:04lot of good models at the time and
33:06I'm going to talk about one we've
33:07we've developed a
33:08second which which I think is
33:09maybe even better, but I'm going
33:10to talk about the first one today.
33:12So. So you know there are some
33:15genetic models in kidney cancer.
33:17The, the one that was was used
33:19really until this time was really
33:21the Renka model which is for those
33:23of you familiar it spontaneously
33:25arose in a valve C mouse which is
33:28immunocompetent mouse and it was called
33:30the renal cortical Adam carcinoma.
33:31Back then we we really had very
33:33limited understanding but it has
33:35been profiled now and we know
33:36that it's not a VHL driven tumor.
33:38So VHL as I showed you earlier is is
33:40the fundamental event in in clear
33:42cell you have to have a VHL mutation
33:43to be a clear cell really and and so,
33:48so it was used forever.
33:50GEM models of course we know the
33:52genetics so why can't we just use
33:54gems and there are there are many
33:55many GEM models you can probably
33:57find six or seven out there.
33:59They're often from mixed mixed backgrounds.
34:01They have a very long tumor initiation time,
34:04very low lower tumor petitrins
34:06compared to other gems and they're
34:09very hard to transplant and often
34:10they develop cystic renal failure
34:12because when you knock out even in
34:14a KSP specific or Cree specific
34:17kidney specific fashion you will,
34:19you'll often develop cystic renal
34:20failure which is what you see in in
34:22people with hereditary kidney cancers.
34:24Many of them especially with
34:25VHL will develop many,
34:26many cysts in their kidney
34:27in addition to tumors.
34:28So,
34:29so it's hard to use those models.
34:31So we teamed up with Scott Lowe
34:33together and Scott Lowe is for those
34:36of you not familiar with him is very,
34:38very compass scientists at Memorial
34:39and does a lot of mouse engineering
34:42beautiful ways and we used an
34:43electroporation strategy at the time
34:45which would allowed us to deliver guides.
34:47We focused on a actually interesting
34:49combination of genes which are not
34:51super common in kidney cancer,
34:52but they define and I'll show you
34:54in a minute some of the myeloid
34:55phenotypes which is one of the
34:57reasons why we focus on it is there
34:59also happen to be a very nice gem
35:00from Ian Frus group in in Germany.
35:02It's time that utilize VHRB and
35:05P53 and showed very nicely a gem
35:07that which we had been using in the
35:09lab for many years and I'll show
35:10you some of that work in a minute.
35:12But we had utilized this strategy
35:14because a because Scott had utilizing
35:16P53 in a lot of different tumour
35:18models and he had very good guides
35:20for and very good strategies,
35:21but also because of the genetic
35:23combination and also because of
35:25the myeloid phenotypes.
35:26This was sort of our strategy at
35:28the time and this was not trivial
35:30because we had VHL as the backbone
35:32and that that that makes cells
35:34very tricky to add additional guys.
35:36For some reason when when you
35:38when you knockout VHL
35:39in vitro the cells don't
35:41tolerate it very well.
35:42They they senesce they they.
35:44And so this was about a year and
35:46a half of work that Lynn did.
35:48And eventually though we were able
35:49to develop a tumor that we could
35:51transplant and we were able to
35:53show transcriptomically that this
35:54matched into that myeloid high
35:56group that that Bob Moitzer and
35:58others from Genentech had shown
35:59to be critical for that cluster.
36:01The the, the very aggressive one
36:03that seems to be resistant to to
36:05vegif and I can show you here
36:07you know this was that cluster
36:08here it's P53 enriched 30%.
36:10Again P53 is not commonly seen
36:12in localized kidney cancer,
36:14but if you look at metastatic
36:16cohorts it's about between 10
36:17and 30% of those will have them.
36:19And so this is really the
36:21stromal proliferative cluster.
36:22When you do flow cytology analysis on it,
36:27they they're very macrophage and rich
36:29tumors and they have AP 53 program.
36:32When you look at transcriptomics,
36:33it's very similar to that group.
36:35And as I mentioned there was this
36:36adenosine signature that we saw which
36:38overlap with the myeloid sector.
36:39This,
36:40this actually came from this paper that
36:42was published in from UCSF from Fong ET al.
36:46In combination with with Corvis
36:48which is a biotech company.
36:50And they had developed an adenosine
36:532 receptor blockade therapy for
36:56patients with metastatic kidney
36:58cancer that were treatment refractory
36:59and they had developed a signature
37:01which which was very much overlapping
37:02with this myeloid signature.
37:04So this gave us the thought of OK,
37:05well we showed that this myeloid program is
37:08so important for metastasis development.
37:10Maybe if we targeted this adenosine
37:12pathway this could abrogate metastasis.
37:14And So what Lynn did in her model
37:17when she developed it was was
37:19stored sort of started testing
37:21CP1444 is this adenosine inhibitor.
37:22And we were able to show that you
37:24get this dramatic abrogation of
37:25metastasis doesn't fully control it,
37:27but the the number of Mets and the
37:30development of Mets is abrogated
37:32quite dramatically using a myeloid
37:33depletion strategy or a specific
37:35myeloid depletion strategy.
37:36It doesn't deplete all Tams,
37:37but it does shift the phenotype of
37:39some of the Tams quite dramatically
37:41suggesting perhaps that this could be
37:43a strategy and utilizing a mouse model.
37:46In the background of all this,
37:47we've also you know really been
37:50thinking about how to utilize the
37:53micro environment to study resistance.
37:56And so again we had to go back to
37:58a mouse model and I I showed this
38:00engineering model which we developed.
38:01But we're in the background of all this.
38:02For many years we had been utilizing
38:04this gem and we we utilize this
38:07gem from Ian Frue ET all this.
38:09Again this was a KSP, sorry Cree,
38:12ERT 2 KSP 1 Flocks mouse that had lost VHL,
38:17PT3 and RB.
38:18They develop pretty nice clear cell
38:21tumors and we utilize this model to to
38:23actually study some of these questions.
38:26We want to understand what's
38:28happening with both PD1 and VEGF
38:31therapies alone and in combination.
38:33Again I showed you the the real
38:35the relevance of this from a
38:37predicting response standpoint.
38:38And so we first looked at these
38:40tumors genomic this is this is
38:43unpublished data but hopefully will
38:45be submitted in the next few months.
38:47So, so we we started looking at these
38:49KVPR tumors that had from this Ian
38:51fruit model again validating the fact
38:52that they are really representative
38:54of this myeloid high tumor that I had
38:57shown you earlier from from Bob's work.
38:58This is again showing RNA sequencing
39:00from from from many tumors that we
39:02have from these mice and that they
39:04overlap very nicely with the the myeloid
39:06high PV 3 driven tumors in in human.
39:11And we started treating these mice
39:13and really focusing on a combination
39:14strategy which I think is near to dear
39:17Harriet from her from work in Melanoma.
39:19But we we utilized linvanib and and PD
39:22one in combination for a few reasons.
39:24One we were very interested in levanim
39:25and PD one comma that that has the
39:27highest overall response rate.
39:28If you look at the clinical trials,
39:30it's about 75% of patients will have
39:33a first line response which is really,
39:34really incredible and and also we
39:39know that Lenva has potentially a
39:41lot of micro environmental targets
39:43beyond just veg F So so we're very
39:46interested in this question and we
39:47utilized a sort of a mouse clinical
39:49trial from this work.
39:50We also included though ACSF 1 inhibitor
39:54BLZ 945 to see if if we just broadly
39:56depleting Tam's would be helpful.
39:58And I should note that CSF one and
40:00our inhibitors have been have been
40:02a DUD in the clinic.
40:03Primarily because they they tend
40:05to deplete lots of Tams and Tams
40:06can be good and they can be bad.
40:08So we don't really we we weren't
40:10really didn't really know what to
40:11expect here and I'll just show some
40:13of that data and we did single
40:14cell on pretty much all of these
40:17mice that we developed tumors from
40:19in different categories.
40:20And we and we this to this,
40:21this model actually is quite
40:23sensitive to linvatinib and actually
40:25the combination is is pretty
40:26dramatically responsive here.
40:27But they don't respond at all to
40:30PD1 and actually CSF ONE inhibitors
40:31don't really do anything at all.
40:32So we were then also able to take
40:36early responders and resistant and and
40:38actually start comparing them as well.
40:40So we can look at the impacts on
40:42the micro environment from these
40:44different treatment strategies alone
40:45in combination and in resistance and
40:48in sensitivity which is which is
40:50really I think something you want to do.
40:51And we could take single cell data
40:53from this same strategy that we
40:54applied before and start looking at
40:56the differences between responders
40:57and non responders between ones that
40:59are in combination or or alone and
41:02get a sense of what's really driving us.
41:04You know one of the interesting
41:06things about this single cell data
41:07set was that we actually had a
41:09lot of neutrophil populations and
41:11we don't really know their role.
41:12I mean there's been some really nice
41:14work from Taha Murgoob and Jed Walchak
41:16recently on the on on on neutrophils
41:18roles in in immunotherapy strategies.
41:22So it's an emergency,
41:22but I'm not going to talk about
41:23that too much today,
41:24but it was a striking finding here
41:26and we could further subtype the
41:28macrophage clusters within here
41:29and understand what's happening.
41:31It's a, it's a very Tam dominated tumor
41:33type as I mentioned with the P53 and RB.
41:37And we can actually further stratify
41:39them and to understand what's actually
41:41happening in the context of both
41:43therapeutic sensitivity and resistance.
41:45And you can start to see that certain Tams
41:46are associated with response and certain
41:48Tams are associated with resistance.
41:50So even though if you just deplete all Tams,
41:52you would actually lose that effect.
41:54But actually if you understood which
41:55Tams which tumor associated macrophages
41:56are actually associated with response,
41:58these Angio hide Tams,
41:59Tams that are producing angiogenic genes
42:01which maybe have been reflected by some of
42:03those Angio bulk RNA sequencing data earlier,
42:05they're actually associated with response.
42:06Whereas other Tams,
42:07maybe those myeloid high Angio
42:09hide tumors that don't respond are
42:11actually associated with resistance.
42:12So now we can start getting further into
42:15the phenotypes of these Tams and and
42:17the context of treatment strategies.
42:18And for the sake of time,
42:20I won't talk about the neutrophils,
42:21but it is another story and we
42:24can actually use neutrophils to
42:26associate again further responses.
42:27So perhaps there's a a major role for them
42:29and I don't have time to talk about it.
42:31But then then of course you have
42:32to go back to the human right,
42:33because I showed you something in mouse.
42:34But are there analogous populations
42:36in human post treatment, right,
42:38Because if you're going to,
42:39you can cure lots of mice,
42:40but you don't know if if those,
42:42if those populations are are
42:45relevant to human biology and that's
42:47a major challenge for immuno,
42:48immuno genomic studies or immunologies
42:51immunotherapy related studies because
42:53of course the mouse and the human micro
42:56environments are can be quite different.
42:58So that requires post treatment tissue.
43:01So how do you do that?
43:01So that's you know some of the beauty
43:04about going back and forth in in my group.
43:07So this is work that was led by
43:09Steven Reese who's graduating from
43:11our program this year.
43:12He's spent a year with me in the lab
43:14and a whole whole including a lot of
43:17very talented pathologists and research
43:19pathologists and of course Christina
43:20Leslie from the Computational Biology.
43:23And we and we we took all these patients
43:24that we've been operating on over the years.
43:26Now we started to define them
43:28into categories right of of early
43:30response of a complete response,
43:32partial response and and no response.
43:34These are patients that Pat operates
43:36on all the time and and I operate on
43:39quite a bit and these are patients
43:40that that are post immunotherapy
43:42now which is a new new frontier.
43:44A lot of our surgery now is now in the
43:46in the post IO space that gives you a very,
43:48very unique opportunity to actually
43:50study what's happening both within
43:52and across tumors.
43:53And so this this is allows us to
43:56assemble cohorts of patients that have
43:58been exposed to IO therapy alone to
44:00IOTKI therapy and we have some with
44:04just TKLO but that's really not done anymore.
44:06So,
44:07so essentially we can look at
44:09responses both defined
44:10clinically, radiographically but
44:11also pathologically and understand
44:13are there what are the populations
44:15in human and are they analogous to
44:17the mouse of course which is you know
44:19something that I you know really want
44:21to do what we really want to focus on.
44:25And so you can start defining
44:26this different ways.
44:27This hasn't been,
44:28there's not an official Canon
44:30here on how to do this.
44:31It's sort of been adopted a lot from
44:33the Melanoma literature about how
44:35to define true pathologic response.
44:37A lot of us have looked at you
44:38know complete response where
44:39there's no residual viable tumor,
44:40RVT, residual viable tumor or
44:42near complete response.
44:43And those patients actually if you
44:45if you take their their kidneys out
44:47we we showed and and we'll show in in
44:48our paper that you know they have really,
44:50really durable responses you know many,
44:52many years without even off therapy.
44:54So that's a it's a good biomarker for
44:56how they're going to do down the road.
44:57And then you can have these partial
44:59responses where they have this in
45:00between and you have these non responses
45:01where there's really no treatment response.
45:03You can see it all within the
45:05tumor and you could do single cell
45:07sequencing on these cohorts and
45:09start to get obviously in the human.
45:10You still see a lot more T cells as
45:12I showed you earlier from the work
45:13that we did and what David has done.
45:15But you can see these tan populations
45:17here and then you can overlay work that.
45:19Andrew Corners,
45:20who's an MD medical oncology fellow
45:23working with Ming Lee has done
45:26a lot of this work now.
45:27And we can start seeing what are
45:29the differences between the IO only
45:31and the IOTKI and the untreated
45:32populations in terms of the single cell,
45:34again post treatment populations
45:36and we can start focusing on some
45:39of these same populations that we
45:41saw that and then we can actually
45:44overlay the mouse Tam signatures
45:46onto these populations to see other
45:48analogous populations and are they
45:50associated with resistance and
45:51response both within the tumors
45:53and across the different regions.
45:54And that's sort of where we're focusing now.
45:56And so we can further substratify
45:58the Tams just like I showed you in
46:00the mouse and to see are the TKII
46:02iOS really depleting some of these.
46:04This is just looking at them broadly
46:06without looking at resistance.
46:07But you can start seeing that that the
46:09the different populations are being
46:11affected in different ways by the
46:13different treatments in maybe similar ways,
46:15but I'm sure different
46:16ways as well as the mouse.
46:18But hopefully that will help us hone
46:19in on what are the most relevant
46:21targets to try in the mouse.
46:23And you can see this if you focus
46:24on one of the M0 signatures from
46:25the mouse that I showed you,
46:27that Cape, that GEM mouse.
46:28You can see that there's clearly
46:31differences in terms of the
46:33TKIO combination patients
46:34and in in the upper tail and the
46:36lower tail on this platter actually
46:38associating with resistance and response.
46:39So you get a sense that maybe
46:41these populations are relevant
46:43across you know species.
46:44So that I think is sort of
46:46where we're headed. So overall
46:50my conclusions really are that RNA
46:52signatures and immune response
46:53are really the the useful ones.
46:54Clinically I showed you sort of
46:56the genetic recap of kidney cancers
46:58and how it might relate to some
47:00of these microviomal feature.
47:01But that's really what we have from
47:03a from a predictive and prognostic
47:06standpoint and maybe it'll help us
47:08select adjuvant treatment strategies
47:09down the road for patients,
47:11certainly pick the high risk patients
47:13a little bit better perhaps the the
47:16phenotype seems to be enriched in
47:17the in the map in the metastatic
47:19setting and particularly post IO.
47:22And and maybe this this cross
47:24analysis will allow us to prioritize
47:26targets to test pre clinically and
47:28then hopefully bring them out to
47:30the to the clinic now that more
47:31and more companies are interested
47:33in in targeting tan populations
47:35with different inhibitors.
47:37And I want to obviously thank my funding
47:39and of course, members of my lab,
47:41Ming Lee's lab, the urology department,
47:43Christina Leslie from computational biology
47:45and all the medical colleges I work with,
47:47particularly Doctor Mozer,
47:48who's been a wonderful mentor to me for many,
47:50many years.
47:52Thank you so much.
47:53And I'll have you answer any questions.
48:09Thanks, Ari.
48:09That was a real Tour de force.
48:12I have a question about the
48:14complexity of your clustering.
48:15So I I noted that you had 21 clusters of
48:19myeloid cells in one of your figures,
48:22I believe it was one of the mouse figures.
48:25Yeah. Well so that's sort of the art and the
48:30the dark art of of single cell sequencing.
48:32You could, you can cluster any
48:34way you want and you can set your
48:36parameters quite differently. So yeah,
48:38and this is all all my Lloyd populations,
48:42you're absolutely right.
48:42So one of the things we do then of course is,
48:45is then go back with my immunology
48:48colleagues and actually start
48:49to think about what are the,
48:50what are what are really representing
48:52unique populations versus just
48:54slicing and dicing single cell data
48:56in more and more complex ways.
48:58And we try to validate them by
49:01flow and to look at really the
49:02the dominant populations there.
49:03So it's just that this is just
49:05sort of an early iteration of of
49:08what would be real clustering.
49:09No, I get it and it's,
49:10it is really complicated
49:12before treatment on treatment.
49:13But my other question is spatially
49:14are some of the clusters uniquely
49:16positioned in a certain area of the
49:18large tumours that's in humans,
49:20I guess is where I'm interested.
49:21Yeah, yeah.
49:21So,
49:22so in that same cohort that I
49:23showed you that we're doing,
49:25we're working with Heartland Jackson
49:28who's at who's in Toronto who's
49:31spatial mass cytometry kind of person.
49:33We developed from the single cell
49:36data a series of of populations
49:38really relying on most of the human.
49:40So we we took a conglomerate of
49:42the different single cell studies
49:43that David has done and others have
49:45done and kind of come up with like
49:47a meta analysis of what are the
49:49key markers of the different Tam
49:51populations to reduce it down to
49:53maybe five or six that might you know,
49:56you tag a Tam by you know CD 68 or
49:58something else and then you can add
50:00a few additional markers and then
50:01look at the spatial orientation
50:03in these contexts.
50:04So we're we're taking all these
50:06regions both within tumors and
50:08across tumors and and reducing it
50:10down to probably a core.
50:11And I think ultimately from
50:12a biomarker standpoint,
50:13you want to kind of just be able to
50:15choose a particular Tam or particular
50:17T cell that would be relevant and
50:18just reduce it to a couple quick
50:20stains that a pathologist could
50:22hopefully do as opposed to have to do
50:24fancy and very expensive sequencing.
50:26So absolutely thinking about the same
50:28same questions that you bring up.
50:30Thank you.
50:31I guess I haven't
50:35not. I got an unrelated 1
50:37unrelated the mouse cell line.
50:39So thank you for sending us and sharing that.
50:41Of course, the cell line with us.
50:43You're planning on making additional ones
50:44with different genetic proteins. Yeah.
50:46So that's real community service, yes.
50:48Yeah. So, so we, yeah, we are doing,
50:50we have a VHLBA P1 CD can to be,
50:52which is a more common combination and
50:55that one is a sarcomatoid tumor perfectly
50:58responds well to CTLA 4 very nicely.
51:03So that one yeah we'll be hopefully
51:06that's that papers you know we're
51:08finishing up this that work but
51:09that I think that will be something
51:11that people find more exciting just
51:13because of the common genetics.
51:15In fact when I presented this one
51:17initially Bill very Bill Kalin very,
51:19very astutely said you know that's
51:21not a very common genetic study.
51:23I'm like yeah but that's it's a
51:25common one for the for the bad
51:27tumors that don't respond well.
51:28So so this one is is a much more
51:31common genetic subtype and I it's a
51:34challenge of doing anything engineering.
51:35We tried of course all the the more
51:38common mutations as has Bill and
51:40others you know Crisp bring out PBR
51:43and what the tumors just don't grow
51:45well and they're very hard so the the
51:47nice clear cells are hard to engineer.
51:50The bad ones that don't look super
51:52clear cell but have they retain the CA
51:559 and hip one at least don't don't look
51:57you know those are the ones that grow.
51:59It's it's a challenge of any
52:01syngeneic system.
52:02So that's why you know you can rely on
52:04gems but gems are just hard to to treat.
52:06So limitation of the field
52:11sure.
52:15Thank you. So so really great work in
52:17terms of single cell transcriptomics and
52:20even profiling and site off and such.
52:22But ultimately the biomarker should be
52:25translated into clinic and should be
52:27easily performed and reputable and cheap.
52:29So how do you envision translating these,
52:32yeah, into the clinic? That's great.
52:34So a couple couple ways.
52:35And I think we're thinking about
52:36this a few different ways.
52:37So one of course is where obviously
52:41reducing it to a few markers that might
52:43stand for the most relevant populations
52:45and maybe it's a combination of Tams,
52:47maybe some neutrophils and some CDA
52:49populations that might ultimately come
52:51up with a very straightforward IHC panel.
52:53The other thing of course is that what
52:55a lot of people are thinking about
52:57is digital pathology and sort of AI.
52:59So if you can define groups of tumors
53:02transcriptionally and then you you
53:04put it into some model where you have
53:06the scan slide scanned in and and put
53:08through a machine learning platform.
53:10You could maybe even digitally say
53:11this is this tumor has this feature
53:13even though the pathologist has
53:15no idea what they're seeing,
53:16but the the model does.
53:18And so for that,
53:18you know and I know a lot of
53:20people are working on this,
53:20but we you know for that same
53:22Novartis study where I showed you,
53:23we showed that the myeloid phenotype
53:25is associated with recurrence.
53:27We're working with group at Dartmouth
53:28that has a machine learning model.
53:30We have all the transcriptomic,
53:31We have the the slides sent from Novartis
53:33which is like 12 terabytes of data.
53:35They high,
53:36high resolution scanned all the slides
53:38from that trial and we've given them
53:40the micro environmental subgroups
53:41and strategies and they're trying to
53:43figure out if they can do that and
53:44they have already done it on the TCGS.
53:46So you can kind of replicate it because
53:48that that might be a way to to say,
53:49OK,
53:50now you've put it through an AI
53:52system and they say this tumor
53:54is this thing and this patient's
53:56going to recur much more or this
53:57patient might respond better.
53:58So that that's that's a strategy that I
54:00think a lot of us are thinking about.
54:02Great question for the
54:08community service and making
54:09these mouse models and thank you
54:11for sharing with us as well.
54:12I'm curious from the immuno
54:15oncology metabolism world,
54:17you know we talked a lot about the obesity
54:19paradox in Melanoma and lung cancer.
54:21Do patients with obesity respond
54:23better to immunotherapy.
54:25So I'm wondering if you know knowing that
54:28RCC actually is associated with obesity.
54:30I'm I'm wondering if you can speak
54:32to your mouse models if if you've
54:34observed any potential difference
54:35in the response to immunotherapy
54:37in in these models because it in
54:39mice with obesity and if not maybe
54:41we can collaborate that. Yeah.
54:42So. So it's a great question and
54:45something that's very near and dear.
54:47So.
54:47So I I do have funding through the DoD
54:49to look at obesity in kidney cancer
54:51and these models and we've we've
54:53been utilizing the transplantation
54:55models we so we do we did the GEM
54:58model first we we did fat feed them
55:00they're they're hard to to feed
55:02because of the mixed background.
55:04So they they gain weight not as nicely
55:06as if they were a clean genotype but
55:08we do observe earlier onset tumors
55:10in those in those mice but then so
55:12genetically when we implant them and
55:14that's not trivial by the way if
55:16you're going to do us an orthotopic
55:18transplantation model in a fat
55:20mouse because just like humans they
55:21develop a lot of perinephric fat.
55:23So if you try to open up the mouse and
55:25inject it it's like a **** show part
55:27of my French but but but essentially
55:30it's very challenging so so we've
55:32been doing but we do see that the
55:34tumors grow faster in obese which
55:36sort of makes sense because we know
55:39that obesity associated with with
55:41better development of kidney cancers
55:43but in it does suggest in human at
55:46least they seem to be less aggressive.
55:48So we're we're now trying to understand
55:50immunologically what's going on but
55:52I would I think we're talking soon
55:53so I'm happy to talk more about that
55:55Rachel but I think it would be a
55:57really it's a really cool area and
55:59we're we're we're focused on on Trem
56:012 macrophages which has been shown to
56:03be associated with lipid their the
56:05lipid associated macrophages and and
56:06it's associated with more aggressive tumors.
56:08So we can talk more about that,
56:10but definitely something that we're,
56:11we're thinking a lot about.
56:17All right.