The Determinants of Effective Anti-Tumor Immunity in Kidney Cancer

December 14, 2022

Information

Yale Cancer Center Grand Rounds | December 13, 2022

Presented by: Dr. David Braun

ID9280

To CiteDCA Citation Guide

00:00And so it's my great pleasure.
00:02I'm Marcus Bosenberg,
00:03I'm one of the Co leaders of the cancer
00:06immunology program and we're actually
00:08have our full House of a program
00:11Co leaders in person as we speak.
00:15So today's grand rounds speaker will
00:17be David Braun who's an assistant
00:19professor in the Department of Medicine
00:21and also has appointments in pathology
00:23and urology and is a Lewis Goodman
00:25and Alfred Gilman, Yale scholar.
00:27So he is also a member of.
00:31The Center for Molecular and Cellular
00:33Oncology that Marcus Musham leads at
00:36the Yellow Cancer Center and received
00:38his PhD in computational biology at
00:41NYU and his MD at Mount Sinai and was
00:44a resident in Boston and a fellow at
00:47Dana Farber prior to coming here.
00:50And we're very glad he's come to Yale.
00:52So David, as you will see,
00:54is covering a remarkably broad set of
00:57things related to cancer immunology,
00:59really focused initially on renal
01:01cell carcinoma.
01:02But you know,
01:03we'll see where that goes and
01:04we're super happy to have him here
01:06and have them associated with the
01:08cancer analogy program.
01:09And David, without further ado,
01:11we'll have to start.
01:13That's perfect.
01:17Thank you so much for the incredibly
01:19kind introduction and for that the
01:20chance to speak to you all here today
01:22and all the people on zoom as well.
01:24And So what I'm going to talk about today
01:26is some of the determinants of effective
01:28antitumor immunity in kidney cancer.
01:29And as Marcus mentioned,
01:30a lot of this is relevant specifically
01:32for kidney cancer and I see patients
01:34with kidney cancer at smilow.
01:35And so it's relevant for these,
01:37it's relevant for these
01:39patients specifically.
01:40But I hope as well that this could be used
01:42as a as a broader model of human tumor
01:44immunology that we might be able to learn.
01:46Uh, learn particularly neurologic
01:47mechanisms that might be applicable
01:50to other cancer types as well.
01:52Here my disclosure is not relevant
01:54for today's talk and so I usually
01:56like to start off with just a couple
01:57of patients just to highlight the
01:59challenges within kidney cancer.
02:00This is going to be true for a lot
02:02of solid tumors and particularly in
02:04the immune therapy era.
02:05So the first patient,
02:06these are two of my patients
02:07from the last couple of years.
02:08The first patient was a
02:0968 year old gentleman,
02:10had a fairly common age for kidney cancer,
02:12had metastatic clear cell kidney
02:15cancer that's the most common.
02:17Thank you.
02:18The most common histologic diagnosis of
02:21kidney cancer and had pretty widespread
02:23metastatic disease throughout the lungs.
02:25Lungs actually had a brain metastasis
02:27and so received a standard first
02:29line combination of checkpoint
02:30inhibitors and the volume animac.
02:32And for those that don't
02:33look at CC's every day,
02:34the primary tumor is outlined in red and
02:36we can see compared to prior to therapy,
02:39there's a tremendous shrinkage
02:40of that primary.
02:41There's basically resolution
02:42of all metastatic disease.
02:43The the primary,
02:44the residual primary was resected.
02:46And so this patient is free of disease,
02:48potentially cured of their disease has
02:51now been off therapy for over a year.
02:53That's in sharp contrast
02:54to a very similar patient,
02:56very similar demographic received
02:58the exact same therapy for the
03:00exact histologic diagnosis.
03:01But unfortunately the tumors
03:02did not respond to therapy.
03:04They all grew and despite this
03:05and subsequent lines of therapy,
03:07the patient passed away within
03:09eight months of diagnosis.
03:10And so it's it's cases like these,
03:12these extreme response phenotypes,
03:13the sort of promise of long standing
03:16durable responses and resistance
03:17that sort of drives the questions
03:19in in my lab and there's some
03:22peculiarities to kidney cancer.
03:23That on the scientific front has also
03:24been really interesting and fascinating.
03:26So historically we think of CDA T cell
03:29infiltration as being a positive thing,
03:32positive prognostic thing.
03:33They're main factor self
03:34or anti tumor immunity.
03:36So having a lot of them in the
03:37tumor is positive though historical
03:39exception is really kidney cancer
03:41where over time of the past 20
03:43years or so having a high degree
03:45of CD T cell infiltration has been
03:48associated with a worse prognosis
03:49really in contrast to just about every
03:52other solid tumor type.
03:54Further when we think of what our
03:55tumor types that typically responsive
03:57to new checkpoint inhibitors,
03:58we think of tumor types
03:59that are on the far end,
04:01the right end of this mutation
04:02spectrum that makes a lot of sense.
04:04We have tumors that have high
04:07mutation burdens, lots of neoantigens,
04:08so lots of potential antigenic
04:10targets for the immune system.
04:12They're potentially more likely to respond.
04:14So Melanoma, non small cell lung cancer,
04:16urothelial cancer,
04:17MSI colon cancer and then we see right
04:19in the middle is clear cell kidney
04:21cancer with a modest mutation burden.
04:23Pretty similar to glioblastoma or
04:25pancreatic cancer, ovarian cancer.
04:27Yet both historically and contemporarily,
04:29it's responsive to immune therapy.
04:31And so my hope is we might be able
04:32to learn a little bit about why,
04:34if we can figure out why this
04:35is responsive to immunotherapy,
04:37maybe we can apply those lessons elsewhere.
04:38And then also for those patients who
04:40aren't lucky to benefit from immune therapy,
04:43the current forms of immune therapy,
04:44can we understand mechanisms of
04:46resistance that can guide rational
04:48combinations of future therapies and so
04:51the framework that our lab really uses to.
04:54Answer these questions,
04:55what are the infiltrating immune cells
04:56and what are the antigenic targets
04:58in kidney cancer is a pretty simple
05:00framework and I've sort of outlined
05:02it here where we have a tumor cell,
05:04the kidney cancer cell,
05:05yeah,
05:05interacting with an infiltrating T cell
05:07and that takes place in the context
05:10of a heterogeneous microenvironment.
05:11And so with this really basic
05:14sort of worldview,
05:15we can begin to ask focus questions
05:17and these are the really the
05:19questions that end up guiding a
05:20lot of the projects in our lab.
05:22So the first is what are the genetic
05:24alterations in kidney cancer,
05:25how do they potentially impact immune
05:27infiltration into the tumor and
05:29ultimately how do those intersect or
05:31interplay to impact therapeutic response.
05:33What are those other immune cells
05:35with immune cells within the tumor
05:37microenvironment? How do they?
05:40Interact with T cells and impact
05:43cell phenotype.
05:44And finally when everything goes
05:45right and T cells are capable CD,
05:47T cells are capable of recognizing
05:49the tumor and eliminating it.
05:51What is it that it's recognizing.
05:52And so we know that at the heart of
05:54this interaction of antigen specific
05:56immunity is the tumor cells presenting
05:59antigenic peptides and MHC Class 1
06:01molecules being recognized by the
06:03cognate T cell receptor and for for
06:05this for kidney cancer and for most
06:07of their I would say solid tumors,
06:09we don't actually know what
06:11those antigens are.
06:12We know sometimes for high
06:13mutation burden tumors that.
06:14Can be classic in the antigens,
06:16but for things that are
06:17modest mutation burdens,
06:18tumors like kidney cancer,
06:20it's much less clear.
06:21And so these are the sort of three
06:23fundamental areas that the lab is
06:25currently working on and we'll
06:26kind of step through each one,
06:27maybe talking a little bit about
06:29some prior work over
06:30the last couple of years when I was
06:32back in in Boston and then some
06:33ongoing efforts now in the lab.
06:35So the first is really what are the
06:37mutations in kidney cancer that
06:39might impact immune infiltration
06:41and ultimately response to therapy.
06:43And I would say broadly,
06:44we use a lot of classic genomic techniques,
06:46whole exome sequencing,
06:47RNA sequencing to really approach
06:49these questions.
06:50And so a lot of the motivation
06:51for this came from an early study
06:53just a few years ago from Ellie
06:54van Allen's group at Dana Farber,
06:56where he looked at a small phase one
06:58trial of nivolumab in kidney cancer,
07:00the checkmate O 9 trial and it was
07:02only about 35 patients that had genomic
07:04data available, but for those 35.
07:06Patients asked a pretty simple question,
07:08what are the the mutations that are
07:11recurrent in kidney cancer among those
07:1335 patients and out of those recurrent
07:15mutations and that's on the X axis.
07:17And now those recurrent mutations
07:19which are actually significantly
07:20impacting response to therapy
07:22and that's on the Y axis.
07:23And we can see there's only one loss
07:26of function mutations in the Pfaff
07:28complex member PBR one was associated
07:30with improved response and we see
07:32the bottom also improve survival in
07:34this small cohort of 35 patients.
07:37So really building off of this
07:39initial funding that we were through
07:41partnership with Bristol-Myers able
07:43to sequence not only the phase
07:45one data but the tumors from the
07:47phase two and phase three trials
07:49of nivolumab in kidney cancer.
07:51And looking specifically at the phase
07:53three trial which is the Checkmate O25 trial,
07:55we were able to confirm that yes,
07:57loss of function mutations and BM
07:59one were associated with improved
08:00response and in this case progression
08:02free and overall survival with
08:04nivolumab with immune therapy though
08:05we can see the effect size is fairly.
08:07Modest.
08:08So it's really, you know,
08:09it's something that's there,
08:10but it's certainly not the whole picture.
08:13And so this was a very focused
08:15sort of validation question,
08:16but we really want to look at
08:17this much more comprehensively.
08:19And So what we did was again in
08:21partnership with Bristol-Myers,
08:22we're fortunate to have access
08:23to tumors from the phase one,
08:25phase two and phase three trials
08:27of nivolumab in kidney cancer.
08:29The phase three trial is really
08:31the pivotal trial that led to the
08:33first checkpoint inhibitor approval
08:34within kidney cancer.
08:35And we're lucky to benefit as
08:36well from that phase three trial
08:38also having a control arm and mtor
08:40inhibitor of your elymus.
08:41And so we could see if there's
08:42anything that we find that might be
08:44associated with response or resistance.
08:45Is that something that's specific for
08:47immune therapy or is that something
08:49that's perhaps more prognostic
08:50that we see with old therapies,
08:51including one with an independent
08:53mechanism of action like an mtor inhibitor.
08:55So we performed whole exome sequencing,
08:57RNA sequencing and CD8 immunofluorescence
08:59to really look at the immune
09:01infiltration of the broad immune
09:03phenotype of these tumors.
09:04At the time this was the largest
09:06study of advanced kidney genetic
09:08study of advanced kidney cancer.
09:10The TCG which was really a foundational work
09:12really skews towards earlier stage tumors.
09:14Only around 10 to 15% of the kidney
09:17cancer tumors in TCG are advanced stage.
09:18And so by sequencing this we can really
09:20get an understanding of first the genetic
09:22landscape of advanced kidney cancer and
09:24there's some interesting findings in here,
09:26some enrichment and clinically
09:28unfavorable aggressive mutations
09:29and copy number variants,
09:31things like NF2 mutations, a loss of.
09:34Nine P 21.3,
09:35but really our our primary question
09:37was how does this ultimately
09:39impact response to immune therapy?
09:41And so the first thing we did was look
09:43at some of the classic markers of somatic
09:46alteration burden that is present in
09:48that we associate other tumor types
09:50in other tumor types response with.
09:52So we know that there's a Histology
09:54agnostic approval for the immunotherapy
09:56drug pembrolizumab based purely on
09:58mutation burden and we know that in many
10:00tumor types there's an association with
10:02high mutation burden response to therapy.
10:04And so for kidney cancer,
10:05we looked at total mutation burden,
10:07we inferred,
10:09neoantigen load we inferred.
10:11You want to drive from frameshift insertion,
10:13insertion deletions which create new
10:15open reading frames and no metric
10:17of somatic alteration burden was
10:18at all associated with response.
10:20So again this is in sharp contrast
10:22to Melanoma and non small cell
10:24lung cancer and bladder cancer.
10:25Here mutation burden really does not
10:28associate or predict response to therapy.
10:30When we then looked across each
10:32individual recurrent mutation
10:33and tried to see which one might
10:35be associated with resistance or
10:37response in this much expanded cohort,
10:39again we found only one in
10:41this pooled analysis.
10:42So again it was only PBR on one which
10:44is a very common mutation president
10:45perhaps up to 30 to 40% of of kidney
10:48cancer tumors that was associated with
10:50improved response and overall survival.
10:52And again here we really could
10:53benefit and see that this response,
10:55this impact on response and survival
10:56was unique to the patients tree
10:58with immune therapy and was not
11:00seen in patients treated with the
11:01control arm with an M Tor inhibitor.
11:04So that's uh mutations.
11:06What about the immune landscape and
11:08how that might impact kidney cancer?
11:10We know that prognostically having a lot
11:12of CDT cells might be a negative thing,
11:14but how does that impact response
11:16to immunotherapy?
11:17And so the first thing we did
11:19was characterize these tumors
11:20broadly into 3 immune phenotypes.
11:21And these are types you might be familiar
11:23with the classic immune infiltrated
11:25where there's lots of CDT cells,
11:26immune deserts,
11:27where there's a positive CDT cells
11:29and then in Munich excluded tumors
11:31where there's made perhaps lots of
11:33CDT cells lining up to the tumor.
11:34Urgent but really unable to infiltrate
11:36the tumor center and have a factor
11:38function that's a potential mechanism
11:40resistance and you know other solid
11:41tumor types including a common mechanism
11:43of resistance in bladder cancer.
11:45And so when we looked at
11:47our kidney cancer tumors,
11:48our advanced kidney cancer tumors,
11:49the first thing we observe
11:50was that immune occlusion is
11:52really not a common phenotype.
11:53It's not looking to be a predominant
11:56mechanism of resistance in in kidney cancer.
11:58We see here only about 5% of
12:00these tumors are Munich excluded.
12:02This is in contrast to some like
12:03bladder cancer where up to 50%
12:05of metastatic bladder cancers.
12:06Would have this immune exclusion phenotype.
12:09The other thing we can observe
12:10is by and large these are heavily
12:12CD8 infiltrated tumors.
12:13About 3/4 of these tumors are
12:15highly infiltrated by CDT cells.
12:17But again in contrast to a lot
12:19of other solid tumors,
12:20we're having a lot of CD T cells
12:23might positively impact response.
12:25Here it had no impact on response
12:26and survival and we can see that for
12:29patients I'm showing here treated
12:30with immune therapy that regardless
12:32of whether you had an immune excluded
12:34tumor and infiltrated tumor or desert tumor,
12:36all of those had roughly the same
12:38response to therapy and survival.
12:40And so looking at genetics alone
12:42you know didn't yield much.
12:44Looking at the immune phenotype
12:46alone really doesn't tell us which
12:48tumors are responsive to therapy.
12:50Is there perhaps some interaction
12:52or interplay between them and
12:53so the first thing we did was.
12:55Really,
12:55just look at among the infiltrated
12:57and non infiltrated tumors.
12:59Are there different mutational landscapes?
13:00Are there different driver mutations that
13:02might be present in one or the other?
13:04And the answer was yes.
13:05And here again it was only one that it
13:07was actually the immune desert tumors,
13:09the ones that lack CD infiltration
13:11that were really enriched for these
13:13clinically favorable CDA T cells that
13:15nearly half of those immune desert
13:17tumors had mutations in people on one,
13:19whereas less than 1/4 of the
13:21immune infiltrated ones did.
13:22So my mutations,
13:23we have enrichment of clinically favorable
13:25PR1 mutations within desert tumors.
13:27How about within infiltrated tumors,
13:31particularly looking at copy
13:32number alterations?
13:33And here there was a a different picture.
13:35There was actually a lot more copy
13:37number alterations within these
13:39infiltrated tumors potentially
13:40indicating these might be perhaps
13:42more chromosomally unstable having
13:43a higher copy number burden in
13:45general than non infiltrated tumors.
13:47And so we took a similar approach.
13:50We looked systematically which copy
13:52number alteration was associated
13:53with was increased and infiltrated
13:55tumors and that's on the X axis.
13:58And then out of those infiltrated tumors
13:59which copy number alteration might be
14:01associated with altered response or
14:03survival either positively or negatively.
14:05And that's on the Y axis.
14:06And again only one came out
14:08deletions of nine P 21.3 which
14:11contain genes like CDKN 2A,
14:13CDKN 2B M tap loss of function.
14:15The loss single copy loss of nine
14:18P 21.3 was associated with worse
14:21survival and worse response.
14:23And looking at whether this effect
14:25would really specific to NTP one
14:27treatment or a broad prognostic effect,
14:29we could see that really loss of nine
14:31P 21.3 within these infiltrated tumors
14:33was associated with worse progression.
14:35Green overall survival
14:36really only with anti PD,
14:38one treatment with immune therapy on
14:40the left and not with mtor inhibition
14:42of control arm shown on the right.
14:46So what is it that's impacting this response?
14:48Well, how is 9 P 21.3 actually acting to,
14:52to lessen response to immune therapy?
14:55That remains an open question.
14:56We we took a a first look at
14:58least some activated pathways by
15:00integrating the RAC data and really
15:03looking at which pathways might be
15:05enriched in these nine P 21.3 tumors.
15:07And there's some potentially initial hits.
15:09There's certainly more angiogenesis
15:12and hypoxia on those tumors.
15:14There's definitely more increased
15:16mtor signaling.
15:17And so at least some initial hints
15:18as to how these might be associated.
15:20But there's a lot of mechanistic
15:22work that still needs to be done.
15:23And so the initial model we put
15:26forward for this is that yes,
15:27in theory CD and infiltration should be
15:29associated with better response to PD,
15:31one therapy just like in other tumors.
15:34But what we have here is overlying
15:35the genetics of the tumor,
15:36these non infiltrators.
15:38Infiltrate tumors enriched for loss
15:39of function P brown one mutations,
15:41these clinically favorable mutations
15:43that bring the response rates up and
15:45the infiltrated tumors are enriched
15:47for these clinically unfavorable
15:49deletions of NI P 21.3 and again
15:51dragging the response rates down.
15:53And so this is the work that was done
15:55now published a couple of years ago.
15:57And while at the time this was
15:59a large sequencing effort,
16:00it was about 454 tumors that
16:02underwent whole exome sequencing.
16:03It turns out that that is only enough to
16:06capture you essentially fairly common.
16:08Mutations and kidney cancer that might
16:10be associated with immune infiltration
16:11or response that you actually need much,
16:13much larger numbers to really saturate
16:14and really get a better sense of
16:16the full landscape of of genetic
16:18alterations within kidney cancer.
16:20And so in efforts that we're leading
16:22together with Allie Van Allen's lab,
16:24we've put together in our cohort
16:25of just about 2000 patients that
16:28were treated with immune therapy.
16:30This is from a series of phase three
16:32trials including that checkmate O25 trial,
16:34but other more modern combination
16:36therapies phase three trials of.
16:38Pure immune checkpoint inhibitors like
16:40the volume Applebaum lab or immune
16:43therapy plus antiangiogenic inhibitors.
16:45And the reason we're doing this is sort
16:47of demonstrated by this simulated power
16:49calculation which I've shown here,
16:50which is basically for the number
16:52of patients we have in our cohort,
16:54which is shown on the X axis,
16:56what frequency of mutation are
16:57we actually powered to detect.
16:59So if we look at our original paper
17:01from a couple of years ago now,
17:03we were actually powered to detect
17:04exactly what we found things that are
17:07really quite common in in this case.
17:08Number one,
17:09mutations was present in nearly
17:11half of responsive
17:12patients and a little bit less than
17:131/4 of non responsive patients,
17:15very, very common mutations.
17:16That's all that we were power to detect.
17:19Now that we have a much more substantial
17:21cohort of over 2000 out of which about
17:231500 were treated with immune therapy,
17:25we're now powered to detect a much
17:27broader range of mutations that
17:29might impact response or resistance,
17:31things that might be present
17:32in as low as 5% of responders.
17:34And so really getting a much broader
17:36land idea of the genetic alterations
17:37and how they might be associated with.
17:39Resistance and so our our sort of
17:41questions driving this project are
17:43one just about kidney cancer genetics.
17:45What are the long tail of mutations,
17:47we know the the most common
17:49mutations from TGA,
17:50but what are the long tail of
17:51driver mutations and kidney cancer?
17:53Do those fall within common pathways
17:55that might actually lead us to better
17:57understand kidney cancer biology?
17:58What is the connection between the somatic
18:00alterations and immune infiltration?
18:02We saw some interactions between PT
18:04Barnum one or deletions of nine P 21.3,
18:07but again are there others?
18:08And then finally how do these intersect?
18:10Or interplay to ultimately impact response.
18:14So that's a large ongoing project,
18:16but I think the use of whole exome
18:19sequencing and an RNA sequencing
18:21is really applicable to answer a
18:23number of other focus questions.
18:25And I think as we think about these,
18:27it's also important.
18:28We're really obviously tumor focused,
18:30but also to integrate what's happening
18:32in the host immunity as well,
18:33for instance, soluble circulating,
18:35soluble factors in the plasma or
18:37circulating immune cells as well.
18:39And so just a little bit of a hint
18:41of some of the things that we've
18:43been working on over the past year.
18:45One is a partnership with Random McKay
18:47at sorry that should say UCSD who ran
18:49a phase two trial of a another immune
18:52therapy drug atezolizumab plus bevacizumab.
18:55This was actually non clear cell
18:56kidney cancer less common variants.
18:58And what we could we could see by
19:00looking at circulating factors by plasma
19:02cytokines that is actually a highly
19:05correlated module of inflammatory
19:06cytokines that are present in a
19:09variety of these patients with non
19:11clear cell disease and what we call
19:13the systemic inflammatory module.
19:15This was actually associated with
19:16worse response and worse survival
19:18within these patients.
19:22We've talked mostly about the genetics,
19:23but we know that the RNA sequencing can also
19:25be leveraged to really understand some of
19:27the molecular subtypes of kidney cancer.
19:29There was really nice work done from
19:32the Genentech group and work that was
19:34initially led by Bob Motzer and and
19:36Brian Reaney where they broke down kidney
19:39tumors from a phase three trial and kidney
19:41cancer into different molecular subtypes.
19:43And actually we're able to see
19:45differential response to therapies was
19:47actually predictive of whether a patient
19:49would respond to therapy A or therapy.
19:51Which is a really exciting sort of idea,
19:53biomarker driven selection of of
19:55therapy for patients with kidney cancer.
19:58However that was from patients treated
20:00with drugs that are not FDA approved.
20:02It was overall a negative phase three trial.
20:04And so in work led by Renee Maria
20:06Saliby in my lab we've actually used
20:09a random forest model to now classify
20:11tumors in a FDA approved regiment,
20:14a value map plus axitinib and actually
20:16look at whether these are associated
20:18with response or resistance in a really
20:20FDA approved. Measurement really.
20:21Can you use this for for treatment selection?
20:23The answer is probably not.
20:25And finally,
20:26we know that some patients as I showed
20:28not just respond to immune therapy
20:30but really have exceptional response.
20:32They really have long term durable
20:34response that goes on for years
20:36or tremendous tumor shrinkage.
20:37And so how can we learn what might
20:39drive not just responsive theory
20:40but exceptional response.
20:42And this is together with Suchat
20:43Shukla's lab at MD Anderson.
20:45We've partnered to look at a handful
20:47of these exceptional responders both
20:48from courts we have but also again
20:50in partnership with industry and are
20:52able to identify certain features,
20:54the presence of high clonal
20:55neoantigens and actually.
20:56A higher proportion of tertiary lymphoid
20:58structures they release seem to be associated
21:00with these exceptional responders,
21:01ones that really have response
21:03that lasts for years.
21:04And so overall,
21:05our lab is really focused on using
21:07a lot of these classic genomic and
21:10transcriptomic tools to understand
21:12response resistance to therapy.
21:14But we know these are sort of broad tools,
21:17classic genomic tools that to
21:18understand really what's happening
21:19in the tumor microenvironment and
21:21the tremendous heterogeneity both
21:22in T cell phenotypes but also in
21:24other cells with immune system,
21:26we need finer tools and that
21:27we've heavily relied on single
21:29cell RNA sequencing for this.
21:31And so our past work really asked a
21:33pretty basic question independent
21:34of therapy which was as you advance
21:37along disease stage as you go from
21:39a relatively normal kidney or at
21:41least non malignant kidney to
21:42early stage kidney cancer.
21:43To locally advanced kidney cancer,
21:45to metastatic kidney cancer,
21:47how does the immune microenvironment change?
21:49How do the T cells change?
21:50How do the myeloid cells change and are
21:52there any interactions between them?
21:54And to do this,
21:55we prospectively collected fresh tumor
21:57specimens from different patients with
21:59either early stage locally advanced or
22:01metastatic disease and perform single
22:02cell RNA in T cell TCR sequencing.
22:04Overall, we had a pretty good balance.
22:07We sequence about 165,000 cells from
22:09a little over a dozen patients heavily
22:11skewed towards sequencing the immune.
22:13Uh, the immune compartment
22:15of the microenvironment.
22:16And so now armed with this data set,
22:18we can begin to ask questions what are
22:20the T cell compartment look like and how
22:22does that evolve with progressive with
22:24advancing disease stage and ask similar
22:26questions of the myeloid compartment.
22:29For the T cell compartment,
22:31we can see heavy infiltration by CDT cells.
22:33Largely there's a huge component of
22:35terminally exhausted CD8T cells,
22:37but we see a variety of T cell
22:39phenotypes ranging from resident
22:40memory like cells to classic T regs.
22:42And so when we classify these
22:44different cell populations,
22:46these T cell clusters and organize
22:49them about organize them in a way
22:51to see which might be increase
22:53in advanced disease stage,
22:55we begin to see highlighted in red
22:57that there are few T cell clusters,
22:58few T cell populations.
23:00There really seems to be enriched
23:02in more advanced disease.
23:04Now, at least for me,
23:05it's a little bit unwieldy to look at
23:06so many different cell populations.
23:08And So what we did was brought more broadly,
23:10classify them just using standard
23:12hierarchical clustering.
23:13We can construct this dendrogram
23:15where transcriptionally related
23:16groups of cells are near each other
23:19on this dendrogram and once that
23:20transcriptionally different are far apart.
23:22And what we can see is now instead of,
23:24you know, 18 or 19 different clusters,
23:27we can see really 2 broad groups
23:29in red T cell CD.
23:30She sells that broadly have
23:32markers of T cell exhaustion,
23:33expression of talks, high expression,
23:36multiple inhibitory receptors
23:38and then everything else.
23:40All of the other T cells
23:41which are shown in blue,
23:42the non exhausted T cells.
23:43And now with this much more
23:45simplified definition,
23:46we can see a pretty striking pattern
23:49that terminally exhausted or exhausted
23:51CD8T cells progressively increase
23:53with advancing disease stage.
23:54They're essentially absent
23:55a normal kidney president,
23:57very low levels in early stage disease
23:59and progressively increasing more.
24:00Against disease stages and that
24:03we by contrast see relatively few
24:05non exhausted CD T cells with
24:08an advanced disease stage.
24:10So that's the T cell compartment.
24:11We have this progressive exhaustion
24:14with advancing disease stage.
24:15What about the myeloid compartment?
24:18And for the myeloid compartment,
24:19it's often harder to put these
24:21cells into discrete buckets.
24:22You know,
24:22for T cells were labeled them as either
24:25AT RAG or a CDA T cell that's exhausted.
24:27You put them in some of these
24:29discrete buckets or clusters.
24:30Myeloid cells as we know can exist
24:31much more along a phenotypic spectrum
24:33and so for this sort of analysis,
24:35for for this sort of continuous phenotype
24:37using a trajectory inference analysis.
24:40Is a really nice approach.
24:41It doesn't force you to put
24:42things into discrete buckets.
24:43It allows cells to exist on
24:46a phenotypic spectrum.
24:47And so when we do that for our myeloid
24:49cells we see a actually this interesting,
24:51this nice interesting branching pattern,
24:53which I think recapitulates a
24:54lot of sort of known biology.
24:55We have classic monocytes at the
24:57root and then branching either into
24:59non classical monocytes on the left
25:02or into macrophages on the right.
25:04And if we look at where these
25:06individual cells are coming from,
25:07those myeloid cells that are
25:09present in normal kidney.
25:10I should say normal with a caveat.
25:13They're adjacent non malignant kidney.
25:14So it's from a cancer patient
25:16might not be totally normal,
25:17but we see that they're largely classical
25:19monocytes and non classical monocytes,
25:20very few macrophages in
25:22these non malignant kidneys.
25:24Now if we look at myeloid cells
25:26from different tumor types,
25:27we see a very different pattern.
25:28The first thing that might catch your eyes,
25:30there's many more macrophages
25:31that's the right-hand branch and
25:33across different disease stages
25:34there's just a lot more macrophages
25:36than there are normal kidney.
25:38But if we actually hone in on on.
25:40That right branch,
25:41we see again a different pattern
25:44between early stage disease,
25:45locally advanced and metastatic
25:46and early stage disease.
25:48Those myeloid cells,
25:49those macrophages are heavily clustering.
25:52Relatively early along that branch,
25:54relatively early in that bifurcation.
25:56By contrast,
25:57locally advanced tumors are kind of
25:59spread throughout that branch or spread
26:01throughout what we call pseudo time.
26:02And if you look at the
26:03metastatic tumors at the bottom,
26:05those macrophages are all the
26:06way at the end of that branch,
26:08all the way at the end of pseudo time.
26:10And So what are the genes and
26:12gene programs that are really
26:14driving these trajectories?
26:15It looks like a a switch from a
26:17more pro inflammatory state to
26:18a more ah as they use the term.
26:20But an M2 like state,
26:21an imperfect term but a a more immune
26:24suppressive or pro tumorigenic state
26:25that we see if we look at signatures
26:27of a pro inflammatory signature
26:29those are really those peak and are
26:32really high early on in pseudo time
26:34at that right hand branch where
26:36those early stage macrophages are.
26:37And by contrast we look in an
26:39anti-inflammatory signature that
26:40really peaks later corresponding to
26:42where those metastatic macrophages
26:44are if we look at individual.
26:45Means prone flammatory genes,
26:46aisle 1, beta TNF, aisle 6.
26:49Those are all relatively absent in those
26:52metastatic macrophages outlined in pink.
26:54Whereas if we look at genes that
26:57are typically associated with
26:58this more M2 like phenotype,
27:00things like C163 of the folate receptor,
27:03those are significantly enriched expression
27:05in those metastatic macrophages.
27:07I should say not shown here those
27:09really do express high levels of
27:11complement genes as well and and
27:12trimmed 2 which has been described
27:14by other groups,
27:14these trimmed 2 positive macrophages.
27:18And so we've looked independently at T
27:20cells independently and myeloid cells.
27:22The natural question is,
27:23are those independent events or are
27:25they actually talking to one another?
27:26And to begin to look at this,
27:28we inferred cell cell interactions
27:30using the transcriptomic data.
27:32And the idea is fairly simple.
27:33We use a tool called cell phone DB,
27:35and the idea is if one group of cells
27:36is expressing a ligand and another
27:38group of cells is expressing the known
27:41receptor complex for that ligand,
27:42you might infer that they're
27:44interacting or talking to one another.
27:45And by randomly permuting the labels,
27:47you can actually get some statistics.
27:48And say, is this something that
27:50we expect more than by chance?
27:51And what's shown here is a heat map
27:53of the number of interactions between
27:55different cell populations and this
27:56couple different patterns you might see.
27:58There's the darker blue area that's an
28:01area of relatively low number of cell
28:03cell interactions between different
28:04cell types and that's between usually
28:07between different cell populations.
28:08In red in the upper left corner,
28:10you see a lot of interactions mostly
28:12between different myeloid cell populations.
28:14And then outlined in Black was a
28:16particular area that caught our eye,
28:18which are.
28:19High number of interactions between T
28:20cells and myeloid cells and when we zoom
28:23in and look on exactly what populations,
28:25it's actually these M2 like these trim
28:272 positive macrophages interacting
28:29with these terminally exhausted
28:30CD T cells at a high degree.
28:32So these are inferred interactions.
28:34We obviously have to make sure
28:36they're actually present in
28:37the same in the same sample.
28:38They have to be in the same
28:39tumor to physically interact.
28:40And so we'll look at the relative
28:42proportion of these different populations.
28:43We can again see they're
28:45really highly correlated.
28:46That's a strong correlation
28:47between the presence of these CDT.
28:49Cells and these tumor associated
28:51macrophage populations.
28:52And So what are these interactions?
28:55They appear to be interactions
28:56that are bidirectional and really
28:58support these different phenotypes.
28:59So these tumor associated
29:01macrophages are producing ligands
29:02for inhibitory receptors on T cells,
29:04things we already know about and
29:06target like PD1 and PDL 2 for PD one,
29:09but things we don't yet target
29:10and kidney cancer,
29:11things like PVR and actin 2 for
29:13TIGIT ligands for Tim three and
29:15other inhibitory checkpoints.
29:17But it's not all myeloid
29:18cells inhibiting T cells,
29:20those T cells terminally exhausted
29:21CD T cells are also producing
29:23factors like myth or.
29:25Uh,
29:25CS,
29:25CSF one that support this more
29:27M2 like polarization and so it's
29:30really a bidirectional sort of
29:32immune circuit that seems to be
29:34present in advanced disease stages.
29:36This is of course all inferred
29:38from gene expression.
29:39So can we gain a little bit more
29:41confidence that this is this is true,
29:42this is real and we have a couple
29:44of different validation exercise
29:45I'll briefly go through.
29:46One is are these actually expressed
29:47at the surface of the protein level
29:49and these different populations and
29:50we can use flow cytometry to look at
29:52that are these actually present in the
29:54same tumors and physical proximity?
29:55In situ within the tumor itself
29:57and we can use
29:58Multiplex immunofluorescence and
29:59then ultimately is is actually true
30:01not just in this small discovery
30:03cohort but also more broadly in in
30:06other other larger patient cohorts.
30:08And so briefly to step through this is
30:10in partnership with Arlene Sharps lab.
30:11We looked at these different terminal
30:14exhaust C8T cell populations and CD
30:16163 positive macrophage populations
30:18and ask the question do the T cells
30:20express the ligands we think they
30:22should or the receptors we think they
30:24should and do the macrophages express?
30:26Well, I guess we think they should
30:27and the answer was yes,
30:28they do and they're higher
30:30in advanced disease stages.
30:31Partnering with Sabina Signer Ideas Lab
30:34performing Multiplex immunofluorescence,
30:35we looked at whether these myeloid cells,
30:37these CD 163 positive macrophages
30:39were actually physically interacting
30:41in advanced tumors.
30:42With these terminal exhausted
30:43T cells and again in multiple
30:45multiple metastatic tumors,
30:47we can see evidence for
30:49insight 2 interactions.
30:50And then finally looking
30:51at other external cohorts,
30:53we first looked at a previously
30:55published saitov cohort from Burn
30:57Bodenmiller's Group and re analyze
30:59that data to specifically look for
31:01our T cell and myeloid populations,
31:03our CD163 myeloid populations.
31:04And again could see this pattern
31:06where the proportion of these cells
31:08increase with advancing disease stage.
31:09And finally we derived a gene
31:12expression signature representing that
31:13interaction between those myeloid
31:14cells and T cells and use that gene
31:17signature to interrogate the TCA
31:19and again found this pattern of.
31:21Increasing signature of interaction
31:23with advancing disease stage.
31:24Now that we have this RNA seek signature,
31:27we actually ask the question is
31:29this interaction signature actually
31:30associated with worse outcomes with
31:33worse survival. And the answer was yes.
31:35If we look at the TCG a data having
31:37a high expression of this Tam to
31:39T cell interaction signature was
31:41really associated with worse
31:43overall survival overall.
31:45But again we have to be cautious
31:46that might just be because it's
31:47enriched in stage four disease.
31:49And so if we look specifically
31:50at those patients with stage
31:52four disease in the TCG a,
31:53again we see the same.
31:54Effect having a high number,
31:55higher number of those interactions
31:57associated with worse overall survival.
31:59And if we look at again our initial cohort,
32:01our checkmate cohort that we previously
32:03reported on again having a high expression
32:06of those that interaction signatures
32:08associated with a horse survival.
32:10And so the model we would put forward would
32:11be that with advancing disease stage we
32:13have this progressive T cell exhaustion,
32:15this switch to more M2 like this
32:18anti-inflammatory macrophage
32:18population and that critically in
32:20advanced disease states that they're
32:22really talking to one another,
32:23they're interacting in a way
32:24that we hope is therapeutically.
32:25Marketable.
32:27So up until now I've talked broadly about
32:30kidney cancer as if it's one entity,
32:32but I've been sort of misleading you.
32:35It's actually many,
32:35many different diseases.
32:36And So what we've been talking about is
32:38really clear cell kidney cancer which
32:40is shown here histologically which is
32:42looks clear under the microscope
32:43where it gets its name.
32:45But actually this is a host of over 20
32:47different disease with more entities
32:49being described each WHO update
32:51and there's a huge proportion of,
32:53I apologize, should be about
32:5525% non clear cell which is.
32:57So it was bad to be labeled by
32:58what you're not, but these are
33:00variant histologies of clear cell.
33:01The more common ones are papillary
33:04chromophobe accounts for about 5%
33:06rarer types including translocation
33:08and then hereditary forms including
33:10FH deficient really aggressive
33:12disease biology that often affects
33:14people in their 30s.
33:15And so while all of our efforts has
33:17been have been really focused on clear
33:19cell kidney cancer and that's where a
33:21lot of the clinical data is as well.
33:22We know that these non clear solver
33:25variants really aren't unmet
33:26clinical and scientific need.
33:28We really need to understand their biology.
33:29And how to treat them because most of
33:31their treatment is really extrapolated
33:33from our experience in clear cell.
33:34And so our first sort of approach to
33:36this is really in the chromophobe space.
33:39And so chromophobe is really are
33:40along a spectrum of these renal
33:42oncocytic neoplasms that range
33:44from pretty benign tumors,
33:45renal oncocytoma which never metastasize,
33:49I should say never virtually
33:51never metastasize.
33:52They really have limited
33:55genetic genomic alterations.
33:56Chromophobe is on the other end, are they?
33:58They are true malignant disease.
34:00They have multiple whole chromosome
34:02losses and then there's also these
34:04emerging entities in between low
34:06and high grade unconscious tumors
34:08which have variable potential to
34:10actually invade and metastasize.
34:13And so one of the key things
34:15about these chromophobe tumors is
34:17unlike clear cell kidney cancer,
34:19these don't do well with immune therapy.
34:21And so there's always going to be exceptions.
34:23But both in in clinical trial data
34:25of immune therapies where they've
34:27included chromophore patients,
34:29the response rate is typically less than 10%.
34:31And if we look at these chromophobe tumors,
34:33this is our own data partnership with the
34:36international Metastatic Database consortium,
34:38Danny hangs group at University
34:40of Alberta and we looked at now
34:42hundreds of patients treated
34:43with new checkpoint inhibitors.
34:44Real world data that are the clear cell
34:46or non or chromophobe and the chromophobe
34:48patients really don't do well here.
34:50This is in sharp contrast to in other
34:54treatment varieties chromophores typically
34:55shape a better disease prognosis and so
34:57really this is an area of unmet need.
34:59Why aren't chromophobe tumors
35:01responding to immune therapy.
35:03And so to begin to look at this and
35:05again these are rare tumor types we we
35:07were able to identify a handful of patients,
35:10about five patients that really
35:12represent this disease spectrum
35:13and again perform single cell.
35:15Kinda sequencing to look at the
35:17immune microenvironment and also the
35:18tumor and stromal components as well.
35:20And really our focus questions
35:22were why aren't these responding
35:23to immune therapies and going in.
35:25We had a couple of hypotheses.
35:27So one is maybe it's just a lack
35:29of immune
35:29infiltration. If you don't
35:31have cells there to begin with,
35:32then it's hard to get any immune response.
35:34The second is perhaps they're exhausted
35:37in ways that don't rely on PD1.
35:39So perhaps these have some severely
35:41exhausted or dysfunctional program that are
35:44really incapable of being reinvigorated
35:45by our current immune therapies.
35:48And the last possibility is maybe
35:50those are just bystander T cells
35:52where they're not actually infiltrated
35:54by tumor specific T cells that are
35:56required for true anti tumor efficacy.
35:58And so with these sort of focused
36:00hypotheses we can begin to look
36:02at these different areas and see
36:03which of them actually are.
36:05Able to chromophobe tumors.
36:06So the first is immuno filtration.
36:08Here we did very basic CD 45
36:11immunohistochemistry just to
36:12look at the immune infiltration,
36:14broad immune infiltration of these
36:16tumors and what we can see is on
36:18the right you see clear cell and the
36:20immunohistochemical stain for CD45.
36:22These are really heavily immune
36:24infiltrated T cells,
36:25immune infiltrated tumors as we
36:26saw in our previous study.
36:28By contrast these oncocytic neoplasms,
36:31oncocytoma is low grade oncocytic tumors,
36:33chromophobe tumors,
36:33these have really low degree of immunity.
36:36Centration and we can see that sort of
36:39characterized statistically on the right.
36:41And so that seems to be one problem.
36:43They just aren't a lot of immune cells.
36:44So we're going to improve immune
36:46responsiveness of these tumors.
36:47One will be driving,
36:49driving immune cells actually
36:51into the tumor itself.
36:53The second is,
36:53are these are these cells that are
36:56there just too exhausted to respond.
36:58And to begin to look at this weekend,
37:00look,
37:00turn to our single cell data and look
37:02to the CDA T cell populations and saw
37:04that they express markers of immune
37:06exhaustion and the transfer is no,
37:07those were not exhausted T cells.
37:09So if we look at clear cell which on the
37:11left versus chromophobe on the right,
37:12we see as we know the the ones in the
37:14clear cell tumors are often exhausted
37:17terminally exhausted CDT cells.
37:18That was not the case for chromophobe
37:20tumors and if we look now at the TCGA.
37:22Data,
37:23we see a very similar pattern that
37:25chromophobe tumors have relatively
37:27low levels of exhaustion markers,
37:29expression of exhaustion markers
37:30that's compared to clear cell disease.
37:32And so these are not exhausted T cells.
37:35They seem actually like they're cytotoxic.
37:36They they seem like they should
37:38be functional.
37:38The ones that are there,
37:39why aren't they actually doing the job.
37:41The last part is maybe they're not
37:43tumor specific and so there's ways
37:45to formally prove this by actually
37:47TCR sequencing these,
37:48reconstructing those TCR's and and
37:50testing for antitumor specificity.
37:51That's what needs to ultimately be done.
37:53Our first take out is is to use
37:55the single cell TCR data that we
37:57have to try to infer specifics and
37:58we did this in two different ways.
38:00One is by taking those TCR's and
38:03mapping them to known to TCR's.
38:06I've known viral specificity.
38:07Those are usually specific for
38:09things like CMB or EB or flu,
38:12common viruses that have
38:13nothing to do with these tumors.
38:14And we can see that the the T cells
38:16and chromophobe kidney cancer
38:18mapped out a much more significant
38:20degree to these viral specific.
38:22They have a viral specificity so much
38:25more likely to be bystander T cells.
38:29The other approach we used
38:30was to take signatures,
38:31gene expression signatures defined by
38:33both Kathy who's group but also Steve
38:35Rosenberg's group at the NCI that
38:37are signatures of tumor specificity
38:39including neoantigen specificity
38:41and see what that expression looks
38:44like in these different tumor types.
38:46And we can see for clear cell kidney
38:48cancer in red that those have a
38:50high degree of tumor specificity
38:52signature and for chromophobe
38:54that's substantially less.
38:55And so overall this is our first
38:57sort of attempt to really.
38:59Characterize what is an uncommon
39:00and rare tumor type and really try
39:02to understand it's immune biology.
39:04It looks like it has a lack of
39:06immune infiltration.
39:06It looks like the T cells that are
39:09there are probably fully functional
39:11but lack of tumor specificity.
39:13So that's the work we've we've
39:15largely done the work that was
39:16published last year and the work
39:18on on chromophobe tumors that's
39:20been really over the past year.
39:22But now what we want to do is
39:23move from just characterizing
39:24the disease biology to really
39:26understanding how these different
39:28tumor microenvironments might impact
39:29response or resistance if therapy.
39:31And ultimately how can we
39:33functionally evaluate these,
39:34how can we actually go from a
39:36laundry list of potential sell
39:37sell interactions to ones that
39:38actually might be therapeutically
39:40targetable in the clinic and
39:41that's really what the focus is.
39:43We've started this process,
39:44but over the next year.
39:46The sort of idea behind this
39:48is is shown here.
39:49This is a perspective we publish a
39:51couple of years ago now which is
39:53to really try to integrate these
39:54tumor biopsies for fresh tissue
39:56collection for single cell and
39:58sequencing into clinical trials.
39:59And obviously this is expensive,
40:01this is technically difficult
40:03this feasibility issues.
40:05But if you could do this even
40:06for a handful of patients,
40:08a small discovery cohort and really go
40:09into a lot of depth for small number
40:11of patients then you can actually
40:12learn some lessons like we did in
40:14our our prior work and try to then use more.
40:17Conventional tools,
40:18standard exome sequencing,
40:20RNA sequencing,
40:21immunofluorescence,
40:21amnestic chemistry,
40:23then try to apply that to a larger
40:26validation cohort.
40:27And so this is our our attempt,
40:30this is our basic schema that we try
40:31to take patients who are responsive
40:33and non responsive to immune therapy.
40:34We try to get biopsies before treatment
40:36as much as possible is often challenging.
40:38We try to get biopsies on treatment
40:41or at least that progression,
40:43very variable success on that.
40:46And then to perform single song
40:47RNA sequencing to really uncover
40:48what are the cell type differences,
40:50cellular composition differences,
40:51phenotypic differences and ultimately
40:53what are the differences in cell
40:55cell interactions and so we've.
40:57Uh Bin lucky again.
40:58This is through partnership with a
40:59number of academic collaborators,
41:01but also industry collaborators
41:02preparing to a number of phase two
41:04and phase three trials that we've
41:06been able to collect fresh tissue
41:07from a total of 96 tumors that
41:09were treated with immune therapy.
41:11And we've performed enzymatic
41:13association single cell RNA sequencing
41:14on these on these tumors really to
41:17understand what is the difference
41:18in the immune landscape between
41:20responsive and nonresponsive tumors.
41:22And we've put through a partnership
41:24with AKOYA began to look at
41:25what is the orientation of the.
41:27The physical location of these tumor types,
41:29these different immune populations in space
41:31using these high dimensional platforms.
41:33This is an example of of one of
41:35our tumors from a responsive
41:37patient showing actually a high,
41:39high number of traditional
41:41lymphoid structures.
41:42And so is this actually feasible,
41:44are we actually able to collect
41:45these cryopreserved specimens
41:46from different clinical trials
41:48and get viable cells out of this?
41:50So our first attempt at this was
41:51on a small number of patients
41:53was just on 13 patients.
41:55This is a collaboration with with
41:57Kathy still with MM Atkins at
41:59Georgetown and with Kelly St who
42:01runs a computational group at USC
42:03where we looked at the small number
42:06of cryopreserved tumors and said,
42:08are we able to get viable cells out of this?
42:10And the short answer was yes,
42:12that we're able to get actually really good.
42:13Representation of both tumor cells,
42:15immune cells and also stromal components
42:18and actually even with this small
42:20cord of end up being about 13 samples
42:22that were suitable for analysis.
42:24After passing quality control,
42:25we can actually begin to see are the
42:28differences in immune microenvironment
42:30between responsive and resistant tumors.
42:32And so again this is a
42:34trajectory inference analysis,
42:34this time it's for T cells.
42:36And again we see a branching structure,
42:38but here are fairly interesting one,
42:40one that starts with a root of naive
42:42T cells and branches either into
42:44terminally exhausted CD8T cells.
42:45Those are the same ones we saw in
42:48our our prior work across disease
42:50stages or to these still having
42:52an exhaustion phenotype but these
42:53slam F7 positive CD8T cells.
42:57And we look,
42:57when we look specifically at which
42:59immune populations are associated with
43:01resistance or with altered survival,
43:03it turns out that the slam of seven
43:05positive CDT cells again in this
43:07very small cohort only 13 patients,
43:09but associated with progressive
43:11disease and with worse progression
43:13free and overall survival.
43:15And so this is our sort of initial
43:1813 patients.
43:18We're now parsing through the sequencing
43:20data from our our 96 patients to
43:23really get a better handle on what are
43:25the different human populations that
43:27might exhibit this sort of behavior.
43:29But we also have to move
43:30beyond immune profiling.
43:31We might get a sense of what are the immune
43:33populations that are are relevant
43:34and what are the immune interactions,
43:35which could play a role.
43:37But we actually have to to nominate
43:38individual targets for the clinic.
43:40We actually have to be able to test this.
43:41And then kidney cancer is a
43:43unique opportunity to do this.
43:45Part of it is just purely practical.
43:46These are enormous tumors.
43:47You can have a 6 1/2 centimeter
43:49tumor that's a stage one tumor and
43:51it's not uncommon for these tumors
43:53to extend to exceed 10 centimeters.
43:55And so there's just lots of
43:56material to be able to extract
43:58individual immune populations.
43:59Individual tumor populations and
44:02actually functionally test which
44:03interactions might actually play a
44:05role and we've begun to do this.
44:07We've been able to take these primary
44:10tumors and I should mention Cat sudakin
44:12in the lab is really spearheaded this
44:15process of identifying patients as
44:17in collaboration with Mike Hurwitz
44:19who runs the Gu tumor bank with Debo
44:21Adeniran and Peter Humphrey and
44:23Pathology and Pat Kenny and Urology
44:25where we're able to routinely on
44:27just about every nephrectomy that's
44:28done at the at the hospital.
44:29Here,
44:30collect fresh tumor for this sort of work.
44:34And so we're able to extract both
44:35immune cells and tumor cells and for
44:37a subset of patients we're actually
44:39able to to grow out tumor cell lines.
44:41And so this is really a valuable
44:43resource for thinking about autologous
44:45coculture experiments with T cells.
44:47And so our sort of overall plan for
44:49functional validation is sort of two phases.
44:51One is a more reductionist approach
44:53and one is one that maybe perhaps
44:56recapitulates the contacts of the
44:573D micro environment a bit better.
45:00The reductions to approach.
45:01Again this is led by a few people
45:03in the lab soaky.
45:03Uh Katrina and Hannah is to basically
45:06break this down into individual
45:08cell populations.
45:09So to associate those tumors
45:11into single cell suspension,
45:12isolate individual cell populations
45:14of interest,
45:15coculture just those populations of
45:17interest with a therapeutic drug with
45:19an inhibitor of a particular interaction.
45:21And then be able to measure
45:23changes in T cell function,
45:25basic flow cytometry for intracellular
45:28cytokine production collaboration assays,
45:30expression of cytotoxicity markers like
45:32granzyme and then we begun to implement.
45:35These model antigen systems
45:36where we have now T cells that
45:38we engineer with a specific TCR,
45:40TCR against NY,
45:41so one or against WT1 and we have
45:43tumor cells that express those
45:45antigens and express luciferase
45:47and we can actually now test have
45:48a a model antigen system for for
45:51testing these impact on cytotoxicity.
45:54Again,
45:54there's limitations of a reductionist model.
45:57And so in work that's done in
45:59collaboration actually now with
46:00AstraZeneca and and a lot of
46:02mentorship from from Marcus here,
46:03we also have begun to implement these
46:05tumor fragment models where we try
46:07to recapitulate the 3D microenvironment.
46:09We actually cut the tumor
46:10into these various fragments.
46:12We embed them in a collagen matrix.
46:14We float that in media where we
46:16can add various
46:16perturbations and we can see that over
46:18the course of its short-term culture,
46:21three to five days,
46:22we really can recapitulate the Histology,
46:24the architecture of these clear cell.
46:25Consumers and preserve a lot of
46:27the immune microenvironment both
46:29T cell and myeloid components.
46:30And so with this we can actually
46:31use this to to actually function
46:33and validate some of these things.
46:34We've done some toy experiments,
46:35I'm showing one where we've added
46:37low dose or higher dose IL two and
46:39can show that we can impact the
46:40T cells that are there just as a
46:42initial sort of proof of concept.
46:43But now together with AstraZeneca
46:45really looking at some of these
46:47interactions that we found in our
46:49original cancer cell paper TIGIT
46:51and others and seeing whether
46:52inhibition of those inhibitory
46:54interactions might actually impact.
46:56Tumor killing and cell function.
46:59And so that second piece is really the
47:01tumor microenvironment and how that
47:02changes with advancing disease stage and
47:04now how we can use that to understand
47:06response and resistance to therapy.
47:07The final aspect of our lab focuses
47:09on is really trying to identify
47:11what are the relevant antigens in
47:12kidney cancer and how we might
47:14be able to therapeutically target
47:15them with antigen directed therapy.
47:17And in order to do this,
47:18in my mind we need sort of two pieces.
47:20We need a toolkit of experimental
47:21toolkit to do this and the other
47:23is we actually need the samples.
47:24And so the experimental tool toolkit
47:26is I would say both computational
47:28and physical tools.
47:29Computational tools includes
47:31better antigen prediction,
47:32the ability to infer not just
47:35neoantigens but things like
47:36expression endogenous retroviruses.
47:39Physical tools for antigen detection.
47:41These are immuno epidemics where we
47:43can actually immunoprecipitate off
47:45Class 1 molecules from tumor cells,
47:47elute peptide and and actually physically
47:49detect the presence of individual peptides.
47:52This was done in collaboration with
47:54Steve Carr's group at the Broad.
47:56And then TCR tools and these are
47:57now pretty established tools,
47:59tools for single cell TCR sequencing
48:00where we know the full alpha,
48:01beta paired TT cell sequence,
48:04but also the ability to then reconstruct
48:07them in primary healthy T cells and
48:09actually probe their specificity.
48:11So we do have a good toolkit for
48:13antigen discovery now we need the
48:15samples and I think clinical trials,
48:17particularly early phase clinical trials
48:19are really wonderful platform to be
48:21able to do this really in-depth analysis.
48:23And so I'm going to give a little
48:24bit of a vignette of work.
48:26That's wrapping up from my time
48:27at Dana Farber.
48:28This is a clinical trial,
48:29a phase one trial that led together
48:31with Tony Sherry and Patrick OTT of
48:33neoantigen vaccination in kidney
48:34cancer that I think has some
48:36interesting findings in of itself,
48:37but really also serves I think,
48:39as a platform for answering some
48:41of these questions.
48:42So this was a a trial that took
48:44stage three or stage four patients
48:46with kidney cancer.
48:47They had to be fully resected
48:49at the time of
48:50surgery. So they had no evidence of disease
48:52and these were clear cell only and we
48:54treated 5 patients, vaccine or local.
48:56Pipeline maps or ctla 4 inhibitor
48:58given at the vaccine site or vaccine
49:00alone and the basic process we
49:02would take their tumor and then
49:04normal cells would take their blood,
49:06perform whole exome sequencing
49:07and RNA sequencing on the tumor.
49:09We'd identify tumor specific mutations,
49:11we ensure that they're actually
49:13expressed that the RNA level
49:15we'd use some of our tools that I
49:17just described to infer what are
49:19likely to be HLA binding peptide,
49:21so likely to be antigens and then we'd
49:24actually get together in a spin a room.
49:26And turn it over zoom,
49:28but we get together as an epitope
49:30selection board to actually pick
49:31which of the mutations we want
49:33to target and design synthetic
49:34long peptides anywhere between 20
49:36to 25 more typically in order to
49:39actually physically synthesize them.
49:41We then partnered with a GMP
49:44peptide manufacturer to actually
49:45synthesize these peptides up to 20
49:48representing different neoantigens,
49:50pulled that together with an immune
49:52adjuvant Poly ICLC and then that
49:54delivered that to the patient.
49:56We treated 9 patients overall.
49:59It's the sort of standard demographics
50:01you'd expect for kidney cancer.
50:03Majority were stage three.
50:04There were a couple of stage
50:06four patients as well.
50:07And in every patient we were able to to
50:09identify enough mutations to actually target.
50:11So median of 13 unique mutations
50:13were targeted per patient
50:14with 15 different peptides.
50:16Kidney tumors have a lot more frameshift
50:17mutations than a lot of other tumor types.
50:19So we're able to target a
50:20lot of frame shifts.
50:21I think interestingly,
50:22we're actually able to target in the
50:24majority of patients actually driver
50:26mutations within kidney cancer.
50:27And it turns out those when we
50:29look back ends up being the most
50:31immunogenic immunogenic peptides,
50:33the ones that represent driver mutations.
50:37And I hesitate to like talk about
50:39any sort of clinical data here just
50:42because it's it's nine patients,
50:43but they're,
50:44yeah,
50:44at least encouraging that was certainly safe.
50:47No one,
50:47everyone did well on the trial and
50:49there have been no clinical relapses,
50:51I would say in this population probably
50:53at this point somewhere around 1/3
50:55third to half might have relapsed.
50:57And so the fact that there have been no
50:59disease recurrences is at least encouraging.
51:02But again,
51:02a big part of this was really the
51:04Biospecimen collection and these
51:05were really generous patients that
51:07went through a lot for this trial.
51:08And so for each of these patients and
51:10they went through the vaccination itself,
51:12which were five priming doses of
51:14the course of three weeks and two
51:16booster shots at week 12 and week 20.
51:17They had multiple skin biopsies
51:19prior to and after vaccination to
51:21look at whether they infiltrated
51:22immune populations within the skin.
51:24We obviously have the tumor tissue itself,
51:26but we need lots of circulating
51:27blood cells as well to perform immune
51:29monitoring and so we perform leukapheresis.
51:32Often required a central line placement
51:34before and after treatment and weeks here and
51:36week 16 and pretty regular 200ML blood draws,
51:39regular 200ML blood draws,
51:41really look at what are the circulating
51:45immune populations and our questions,
51:47we're really trying to look kind of end
51:49to end what's happening at the skin as
51:50we move to the circulating immune system.
51:52Ultimately are we getting tumor reactivity,
51:54it's nice to get reactivity
51:55against the vaccine itself,
51:56but it's actually not impacting
51:58tumor reactivity.
51:59We haven't done too much.
52:00We haven't actually been helpful.
52:02And so just to briefly walk through this,
52:03this is what a typical this
52:05actually patient one.
52:06So this is the prior vaccine scars and
52:08what what the vaccine site looks like
52:11two to three days after vaccination
52:13we can perform enzymatic dissociation,
52:16CD45 isolation and single cell RNA
52:19sequencing identifying really high
52:21populations of myeloid and lymphoid cells.
52:24And this is work in progress,
52:26but we actually see some fairly
52:28interesting changes in both the
52:29myeloid cell and T cell population.
52:31I would say predominantly it happens.
52:32With vaccination,
52:33we're not seeing huge differences
52:35with the addition of epilimnion map.
52:37Moving on to the circulating immune system,
52:39really the workhorse for this
52:41was interferon gamma Ellie spots.
52:42These are taking peripheral T cells out
52:44out of a patient peripheral blood cells,
52:46putting them into a dish and stimulating
52:48them with the same vaccine peptides
52:49and seeing whether those T cells
52:51release interferon gamma as a marker
52:53of antigen reactivity and would see
52:54it week one prior to vaccination.
52:56Basically none of the neoantigen pools,
52:58the 1st 4 rows had any reactivity,
53:01but we get pretty strong reactivity
53:04with vaccination and that when we do.
53:07Close cytometry and it's Cellular said
53:08to kind of standing we actually see that
53:10these are not largely polyfunctional,
53:12that they T cells not only produce interferon
53:16gamma but things like aisle 2 and TNF.
53:19And finally,
53:20moving beyond just vaccine reactivity,
53:22are we getting actually tumor reactivity.
53:23And So what we can do is again take
53:25some of these post vaccine T cells,
53:27stimulate them with one of our vaccine
53:29peptides in this case against a driver
53:32mutation PR one and then coculture with
53:34that same patients autologous tumor
53:35and see whether those PBR one specific
53:38T cells actually recognize tumor.
53:41And the answer is, is yes,
53:42that we are able to actually
53:44get tumor reactivity.
53:44It's not for all patients,
53:45but for the majority of patients
53:47we're able to get evidence of tumor.
53:49The activity with vaccination and
53:50so really this is our our first
53:52sort of attempt at an antigen
53:54directed therapy and kidney cancer,
53:55but I think neoantigens are
53:57a good place to start.
53:58But I think those are clearly
53:59not going to be the whole story.
54:01And kidney cancer we know that
54:02there's not an association between
54:04as I showed high neoantigens and
54:05and response to therapy.
54:06So we have to look beyond this
54:09initial neoantigen focused
54:11world and really look at other
54:12sources of antigens as well.
54:14And very briefly this is work large
54:16in collaboration with Bill Kaylan's
54:17Group and Steve Carr's group.
54:19We're using the same cohort of patients,
54:21the same tumors to actually look
54:23at endogenous retroviruses as
54:24potential antigenic targets.
54:26These are ones that are aberrantly
54:27expressed in a few different tumor types,
54:29but specifically kidney cancer has a high
54:31expression of these endogenous retroviruses.
54:33So we can again use our computational
54:36tools to predict antigens,
54:37potential ER derived antigens and use
54:39mass spec based approach to actually
54:41physically detect those antigens.
54:43And in this first patient,
54:45this patient 110 from our original trial,
54:47we see that there were seven ERV.
54:49Derived peptides that were present
54:51on tumor but on a normal normal
54:53tissue and when we take one of those,
54:55the one highlighted in pink and
54:57actually test those for reactivity in
54:59peripheral blood cells from that patient,
55:01we can see that those that patients
55:03actually capable of mounting a low
55:05level but a response to that peptide.
55:07So just a initial proof of concept
55:09that these RV's can actually be
55:10antigenic and now we can actually do
55:12this much more systematically look
55:14across all patients and all of their
55:16endogenous retroviruses that are
55:18presented and look for antigenicity.
55:20Again, these are very focused approaches.
55:22These are specific hypothesis,
55:25neoantigen.
55:26Uh or endogenous retroviruses,
55:27the last thing we want to do is ultimately
55:29build systems and collaborate with
55:31with groups that are interested in
55:33more broad antigen discovery efforts
55:34for things that we're not thinking of.
55:36And so we recently entered a
55:38partnership with Remedy Bio,
55:39a biotech company based in Ireland,
55:41which has a a novel platform,
55:43a nano reactor platform that actually
55:45allows you to coculture individual
55:47T cells and tumor cells within each
55:49of these wells,
55:50but actually measure which of those
55:52wells are reactive to tumors,
55:54use a pneumatic system to extract viable.
55:56These cells tumor reactive T cells
55:57and be able to sequence their TCR,
56:00so really be able to understand
56:02much more systematically what is
56:04the repertoire of tumor reactive
56:05T cells in kidney cancer.
56:07And so overall our kind of hope with
56:09this branch of the lab is really to
56:11move beyond our classic tools for
56:13immunomodulation to add the steering
56:15wheels rather than only looking at
56:17the inhibitory checkpoints or the
56:19the sort of go signals for your
56:21immune cells to actually be able
56:22to add a component of an antigen
56:24directed therapy really focus on
56:26HLA restricted antigens.
56:27And that's where the model for the lab,
56:29it's been a busy but a great year.
56:31That's been a wonderful, a wonderful time.
56:33I felt incredibly welcome here
56:35at Yale and been lucky to have
56:37remarkably energetic and and and kind
56:38group of people joined the lab and
56:40really focus on sort of this model
56:42that we start with the patient.
56:43We try to learn things from their tumor,
56:46from their immune system.
56:47We have a lot to go
56:49and a lot of open questions about sell sell
56:52interactions and about antigenic targets,
56:53but always with an eye to try to
56:55bring that into improved diagnostics,
56:57actually improve therapeutics.
56:58And try to bring that back into early
57:00phase trials like I showed with our
57:02neoantigen trial and then to continually
57:04iterate to try to get a little bit
57:06better each time that we do this.
57:07And so with that,
57:08thank you again for the opportunity
57:10to speak and a lot of people,
57:11both my lab and collaborators,
57:13but most importantly the patients
57:14and their families.
57:15And this time I'm happy
57:17to take some questions.
57:25Alright, just one question.
57:33In general, most of them are.
57:39There must be some difference. Responded.
57:45Yeah.
57:48The CDA looks great based on
57:51responding to non response.
57:55Within infiltrated tumors,
57:56it's a good question.
57:58I think that's where our sort
58:00of larger collection of this 90
58:02single cell sequence of 96 tumors
58:04will I think be very helpful.
58:05If I were to answer this six months ago,
58:08I would have said it's it's
58:09going to be impacted largely by
58:10the myeloid component as well.
58:12And I think that's still is
58:13probably true that we kind of
58:15showed in our original study that
58:16even though we're we're thinking
58:17that we're measuring CDT cells,
58:19likely what we're actually capturing is
58:21interactions between those terminally
58:22exhausted CDT cells and the myeloid.
58:24Component and that.
58:25Historically,
58:25we've only targeted 11 branch of that.
58:28We've only targeted the T cell compartment
58:29and not the myeloid compartment.
58:31I think that's going to be
58:32one big piece of it.
58:33The second piece which was we
58:35weren't expecting to find is this
58:36particular phenotype of slam of
58:38seven positive CDT cells that
58:39requires a lot of validation
58:41both that they're actually real,
58:42but then that they have a functional role.
58:44I think that's going to be the
58:45other sort of component.
58:46Are there different actually the cell
58:48phone even though they're broadly
58:49infiltrated by similar numbers of CDT cells,
58:51are those CDT cells of a different
58:53phenotype that actually might be perturbed?
58:55In some way and one nice thing is
58:57there are even you know FDA approved
59:00antibodies like elotuzumab for slim F7.
59:01So one can see actually a pathway those
59:04end up being true to to the clinic.
59:06So that's that's kind of ongoing work.
59:08Now actually a rotation student in
59:09the lab is putting some of seven
59:11into some of these T cells and we're
59:13actually seeing whether this impacts
59:15cytokine production proliferation,
59:16tumor killing.
59:20Yeah. Ohh. And. 27.
59:33Thank you.
59:39Yeah, it's a great question.
59:40So the question was just about the
59:43stromal component fibroblasts and
59:44other stromal and kidney can't
59:46strictly anthelion cells as well which
59:48these are heavily vascular tumors.
59:50I would say our first study
59:52really didn't we weren't we,
59:53we really didn't look at it at
59:54all because our protocol really
59:56enriched for immune cells.
59:57I think now with not only the chromophobe,
59:59the Chromophobe project,
01:00:00but also this these you know larger 96
01:00:04samples we actually much more broadly.
01:00:06Capture cancer,
01:00:07associated fibroblasts and epithelial cells.
01:00:09I will say that first 13 patients we
01:00:11didn't see any that were specifically
01:00:12associated with response or
01:00:14resistance in this very broad look.
01:00:15But that obviously doesn't mean
01:00:16they're not important actually driving
01:00:18either T cell or myeloid biology.
01:00:20And so that's something that I think
01:00:21we need to look into in more depth,
01:00:22but we don't have,
01:00:23we don't know quite yet but
01:00:25actually now I think.
01:00:26Yeah.
01:00:26Now we actually have the tools that
01:00:27I would be able to look at it.
01:00:33You know.
01:00:36Really the only type that sold?
01:00:39Obesity and commonly treated
01:00:41with cycling inhibitors.
01:00:42I'm wondering if you have any hints
01:00:44in your data as to the role of
01:00:47metabolism in the micro environment.
01:00:48Yeah, it's a great question.
01:00:49The shortening I'll give is,
01:00:50is not yet, but I'd love to be
01:00:52able to support it and look at
01:00:53it because it's some fascinating
01:00:55parts about kidney cancer as well.
01:00:56So there's something where even
01:00:58though it's you're more likely
01:01:00to get it the the incidence
01:01:01is higher in obese patients.
01:01:03Those patients who are obese who
01:01:05have metastatic disease do better.
01:01:06Something called the obesity
01:01:07paradox within kidney cancer.
01:01:09And we know that there's some
01:01:10hints that these are in general
01:01:12are really metabolically active.
01:01:13There's really excellent work from
01:01:14Jeff Rathmell and Jim Rathman's
01:01:16group that looked at what are the the
01:01:17primary consumers of for instance
01:01:19glucose and the micro environment.
01:01:20And since that's not the tumor cells,
01:01:22it turns out it's mostly the
01:01:23myeloid compartment that's a primary
01:01:25drive the primary consumer.
01:01:26But T cells are are still consuming a lot.
01:01:29How those actually ultimately I
01:01:30think impact the function of T cells,
01:01:32I think we haven't looked at all,
01:01:34but it would be great to be able to
01:01:36explore especially with some of these
01:01:37models where we're you know have them in.
01:01:39Really nutrient rich,
01:01:40metabolically favorable conditions and xvo
01:01:43actually would be nice to recapitulate
01:01:45some of the nutrient limitations
01:01:47that are present in the tumor itself.
01:01:49Yes.
01:01:51Well, thank you,
01:01:52David.
01:01:53I thank you all for also coming
01:01:55here in person and we'll look
01:01:57forward to grand rounds next week.
01:01:59Thanks so much.