GR 12-1

December 04, 2023

ID11042

To CiteDCA Citation Guide

00:00Good morning, everyone.
00:01As you trickle in, I'm just going
00:04to start presenting this morning's
00:06grand round speaker, Ben Liu.
00:07So Ben Liu is one of our graduates and
00:10he actually exemplifies what we're trying
00:12to do in the Cancer Center in terms
00:14of building a pathway for training.
00:16So as some of you know,
00:19we have a whole list of T30 twos and K
00:22twelves and our goal is that people go
00:24from one program to another to another.
00:26And so Ben, who he graduated
00:30from NYU Medical School,
00:31came here as a resident and
00:32then joined our fellowship.
00:33And as a fellow,
00:35he joined Doctor Herbst T32 and now he's
00:38a trainee on the K12 in Immuno Oncology.
00:42The only problem is that Ben sometimes
00:44does things a little bit backwards.
00:46So he's currently finishing his PhD,
00:48but yet he's a faculty member and this
00:50makes it extremely complicated when
00:52it comes to the paperwork of the K12.
00:54So if there are other trainees in
00:55the room who are hoping to come
00:57up through this pathway,
00:58please do it in the right order.
01:00But all that aside,
01:03OK if you that's true too.
01:06So if you want to go by the chaotic
01:08method then all I can say is I
01:10strongly recommend Doctor David
01:11Heffler as an amazing mentor.
01:13He's done a great job with Ben who
01:16is doing amazing work on actually
01:18three major projects.
01:19One focuses on brain metastasis
01:22in lung cancer,
01:24one on Melanoma and liquid biopsies
01:25and that's the one he's going
01:27to be talking about today.
01:28And the third one is single cell RNA
01:30sequencing studies of glioma patients
01:32treated with anti TIGIT antibodies.
01:34So there are very few people who
01:36can shoulder all of this while being
01:38a chief fellow and AT30T trainee
01:39and AK12 trainee and everything
01:41else that and a dad I think and
01:43everything else that Ben does.
01:44So without further ado,
01:46I'd like to welcome Ben to give
01:48us his presentation.
01:54Thank you so much, Doctor,
01:55for that very kind introduction.
01:58Good morning everyone.
01:59Hope everyone had a very nice Thanksgiving.
02:02I can tell you that I'm very thankful to
02:04be standing up here on the podium today
02:06and for the opportunity to share some
02:08of our work that has been supported by
02:11the Skinspur over the past five years.
02:13I do have to say that I think it speaks
02:15a lot to our cancer centering community
02:17that we're willing to amplify even
02:19junior investigators such as myself.
02:20And I'm incredibly grateful to my mentors,
02:23Dr. Haffler and Dr.
02:24Kluger for nominating me to
02:26represent our team on this project.
02:29So I've titled my talk Immune
02:31Liquid Biopsies,
02:32Remote Learning and Remote Control.
02:35And the topic here is a little bit
02:37different than the liquid biopsies that
02:38I think many of you are familiar with,
02:41which are more tumor centric.
02:42And yeah,
02:43I'm specifically referring to
02:44circulating tumor cell free DNA.
02:49But I think what we're starting to
02:50realize is that these liquid biopsies
02:52are really powerful companion
02:53diagnostics that are really trying
02:55to become game changers in care.
02:57And it's my hope that with additional
02:59work on immune profiling that
03:01these two will start to emerge as
03:04important tools to help us improve
03:07our care for patients with cancer.
03:10And so I have no personal
03:12financial disclosures.
03:12Some data in the presentation
03:14was generated in collaboration
03:16with Repertoire immune medicines.
03:20And just to briefly go over
03:21the structure of my talk,
03:23I'm first going to talk a little
03:24bit about some evidence that we
03:26have that the broader systemic
03:28immune response is really a critical
03:30component to anti tumor immunity.
03:32And then going to review some rationale
03:33and prior work that's been done in this
03:36space of immune profiling in the blood.
03:38And I'm going to talk through
03:39two stories that we have,
03:41one which is published and one which
03:43is being prepared for submission that
03:46really focuses on using the T cell
03:48receptor as a molecular barcode to
03:50help us understand what what is the
03:53relationship between T cells in the
03:55tumor and T cells within the blood.
03:58I'm going to close by discussing a
04:00little bit of our early efforts to try
04:03and translate our biological discoveries
04:05into clinically relevant biomarkers.
04:11And so just to start, as we all know,
04:15immune checkpoint inhibitors have
04:17really revolutionized the way that
04:19we treat patients with cancer.
04:21And it's in large part due to work
04:23such that's been done by Doctor
04:24Kruger and many of you out in
04:26the audience and the gold mark,
04:30the gold standard for the potential
04:32that immunotherapies have for treating
04:35patients with cancer remains in
04:38Melanoma amongst other cancer types.
04:40But you can see that these are really
04:43practice changing survival curves
04:44from the Checkmate 067 trial which
04:47was a frontline trial looking at
04:49anti PD one and or anti CTLA for
04:52for patients with advanced Melanoma.
04:55So you can also tell from these curves
04:57that about 50% of patients still fail
04:59to derive long term benefit and I think
05:02it's it caused into question why,
05:05why is that?
05:06What, what are the mechanisms that
05:08are causing them to not be able to
05:12amount of systemic immune response
05:14that can result in tumor rejection
05:16and are there markers that we can use
05:18to try and identify these patients
05:21and are there therapeutic avenues
05:22that we can explore by learning
05:26with this information.
05:28And so a a better understanding of the
05:31fundamental determinants dictating
05:32clinical response are really needed
05:36just to review what our current understanding
05:38of immune checkpoint inhibitors are.
05:40When immune checkpoint inhibitors
05:42were first introduced into a
05:44clinic now over a decade ago,
05:46the the thought was really that these
05:49agents target negative signals within
05:51the local tumor microenvironment and
05:54thereby reinvigorate T cells which
05:56we believe to be the primary factor
06:00immune cells and resulting in tumor
06:02rejection reinvigorating these local
06:05T cells to recognize that tumor.
06:09Well, we've since come to learn though
06:11that at least in part the the potential
06:14for immune checkpoint inhibitors to
06:16mount successful tumor rejection is
06:21the the need to induce immune responses
06:25beyond the local microenvironment.
06:27And several groups including those
06:29here at Yale have identified the tumor
06:32during lymph node for example as one
06:35reservoir for tumor specific stem like
06:38T cells that help to regenerate and
06:41sustain anti tumor immune responses.
06:44This is nicely illustrated in preclinical
06:46models whereby we can block lymphocyte
06:49trafficking and in doing so we see
06:53that anti tumor immunity is really
06:56impaired in the efficacy of immune
06:59checkpoint inhibitors is also limited.
07:00This has been demonstrated by several groups,
07:02including two papers out of groups
07:06from Yale from Nick Joshi's lab and
07:09then also from Marcus Bosenberg
07:10and Richard Favell's lab.
07:15We also know that immune checkpoint
07:18inhibitors not only recruit new T cells
07:22to the local tumor microenvironment,
07:23but that these T cells may have actual
07:26actually be recognizing different antigens.
07:29And we're assessing that based off of
07:31their T cell receptor sequences termed
07:33novel chronotypes here on the right.
07:38And perhaps some of the most exciting
07:40data that's merging is the potential
07:43benefit of immune checkpoint inhibitors
07:45to work in early stage disease even
07:47after the tumor has been removed,
07:49the macroscopic tumor has been removed.
07:51And so these are disease free survival
07:54curves on recent trials that have
07:56explored anti PD one therapy in the
07:58adjuvant and neoadjuvant settings.
08:01And what these data reinforces is that
08:03checkpoint blockade really potentiates
08:05immune surveillance beyond the local
08:07microenvironment and helps to prevent
08:10tumor regrowth and disease recurrence.
08:17So in this setting you know we
08:20really believe that a systemic immune
08:23response is an important contributor
08:26to effective anti tumor immunity and
08:28our underlying hypothesis for this
08:30project was that blood based tumor
08:32related T cells really have distinct
08:35characteristics and can be informative
08:36of local tumor immune microenvironment.
08:40Our translational goal is therefore
08:42to try and identify clinically
08:43relevant biomarkers which can be
08:45obtained non invasively through the
08:47blood to try and assess inform us
08:49on anti tumor immune responses
08:54and so prior work in this arena have
08:57nominated several blood based biomarkers.
08:59However uptake into the clinic
09:01is likely challenged in part
09:02due to the lack of specificity.
09:04So several serum cytokines which we
09:06know to be context dependence are not
09:09widely used or due to inavailability
09:12of certain techniques within our
09:15clinical labs such as the ability to
09:18determine T cell receptor diversity
09:20or clone sizes in clinical labs.
09:25And so our general approach has been to
09:27first take a deep dive and deep look into T
09:30cells within the tumor microenvironments.
09:32And in order to do that we employed
09:34using single cell sequencing.
09:36This is a technique that allows us to
09:40simultaneously characterize both the
09:41gene expression profile of individual
09:43cells and in the case of T cells also
09:46the full length T cell receptor sequence.
09:49Now the T cell receptor is
09:51really an essential component to
09:53everything that AT cell can do.
09:55The T cell receptor is what allows
09:57T cells to become activated when
10:00it encounters its cognate antigen.
10:02And the the global diversity of the T
10:07cell repertoire is really really immense.
10:11And so having a high resolution view of
10:13the the sequence is really important.
10:16And when T cells do encounter their
10:17cognate peptides or their androgens,
10:19they become activated and they proliferate
10:21and all of these sister clones are
10:24share the same T cell receptor sequence.
10:27And so in that sense,
10:28the T cell receptor sequence is
10:31really a a useful molecular biomarker
10:33for us to be able to link T cells
10:36that are clonally related within the
10:38tumor and the blood.
10:40And we can then ask the question based
10:42off of his gene expression profile,
10:44how are these cells changing?
10:46What can we learn in these two spaces?
10:51And so in this first portion
10:53of the talk, I'm going to
10:56talk a little bit more about
10:57using TCR as a molecular barcode.
11:00And I'd really like to just acknowledge
11:02Liliana Luca who was a former post
11:04doc in our lab and junior faculty
11:06member in our lab who's now an
11:08independent investigator in France.
11:09She was really an important
11:11architect in driving this project
11:14forward to this initial story.
11:16And so for this initial or the initial
11:20look at using TCR as a barcode,
11:23we performed single cell RNA sequencing
11:25and T cell receptor sequencing
11:27from in blood and tumor from 11
11:29patients with stage 4 Melanoma.
11:31These patients all had mixed histologies
11:34and treatment histories and the the
11:36purpose of this initial look was
11:39to try and assess a global look at
11:42what these clonal related T cells,
11:45global features of these clonal
11:46related T cells.
11:50The way that we went about identifying
11:53tumor T cells which we think are relevant
11:55to the anti tumor immune response was by
11:58looking at how clonal extended they are.
12:00This helps us differentiate T cells
12:02that we may be located within the
12:04tumor but that are not actively
12:06participating in the inter tumor response.
12:09We then link these over into the blood and
12:11we termed for this initial story these cells,
12:14these cloning related but blood
12:15based cells as circulating tumor
12:17infiltrating lymphocytes which
12:18I'll refer to as circulating tills.
12:23And so these circulating tills are a
12:25relatively rare population in the blood.
12:27They are comprised of less than 10%
12:29of our total T cells and you can see
12:32that they're predominantly located
12:33within the CDAT cell compartments.
12:35So what I'm showing is on the right
12:37is a dimensionality reduction plot
12:39of our single cell RNA sequencing
12:42and the circulating tilts are
12:44highlighted in dark green.
12:45You can see that they're predominantly
12:48distributed within the CDAT cell compartment.
12:50These cells are clonal expanded not
12:53only within the tumor but also within
12:56the blood and that interestingly
12:57this population seems to accumulate
12:59over the course of your disease
13:04and so we can perform differential
13:05expression analysis to try and take an
13:08unbiased look at the transcriptional
13:09features of this population.
13:11These circulating tills are the ones that
13:13are located in the right and the all
13:15other blood T cells are located on the
13:17left and we're focusing on CDAT cells.
13:19In this case, what we find is that
13:22they share features of icytotoxicity,
13:24tissue residence, cell migration,
13:28tissue homing and importantly as a A,
13:32it's kind of a A a check.
13:35They they lack features of naive
13:38or memory markers such as CCR 7,
13:40TCF 7 and these are features that
13:43was this is important to us because
13:46it reinforces the fact that these
13:48are cells that have been activated
13:51and are actively participating
13:53in the immune response.
13:58We can also ask the question,
13:59how are these circling tills related to the
14:03features of tumor cells or tumor T cells.
14:07And so the first analysis that we did was we
14:10generated a gene set that is characteristic
14:14of expanded T cells within the tumor.
14:18We then took a look at the expression
14:20of these this expanded tilde gene set
14:24within our circulating till population as
14:26compared with all other blood cells and
14:28we do see that there is a enrichment for
14:30this population or these this gene set.
14:33We can also ask our circulating
14:37tills characteristic of gene sets
14:40of T cells which are specifically
14:43expanded within the the the tumor and
14:47thereby removing genes that may be
14:50just generally associated with clonal
14:52expansion and we again see that there
14:54is an enrichment for this gene set.
14:56One thing that I will point out is
14:59that there there are several hallmark
15:01genes which are have been described as
15:04important features for T cell dysfunction
15:07or tumor exhaustion such as our Co
15:10inhibitory checkpoints such as CTLA 4
15:12Tim 3 which is encoded by the gene HAV
15:15CR2 and then the transcription factor
15:17Tox PD one is also found although not
15:20listed displayed here on this screen.
15:24And so we can see that circulating
15:27tills are are not representative of
15:30features of exhaustion within the tumor
15:33but that there is a good concordance
15:35between a cytotoxicity signature between
15:38this population and the the degree
15:40of cytotoxicity within the tumor.
15:46And we can also ask the question,
15:47are T cells that have been described
15:50to be predictive of response
15:52to immune checkpoint blockade,
15:53are those T cells also found
15:56within the circulation?
15:57So this is work out of Nirha
15:59Cohen's group whereby he generated
16:012 gene signatures of CDAT cells,
16:03one that was enriched in patients who
16:05responded to immune checkpoint blockade,
16:07another which was enriched
16:09in those who were resistance.
16:11And then we took a look to see whether
16:13or not these gene signatures what what
16:16are the global distribution of these cells.
16:19What we find is that the resistance
16:21signature is really only enriched
16:22within T cells which are exclusively
16:25found within the tumor and not found
16:27within circulation whereby the response
16:30signature is found in T cells that
16:33are shared in both compartments.
16:34And I think what this point illustrates
16:37is that the a key component to a
16:42good response to immune checkpoint
16:44blockade is prior systemic priming
16:48of the anti tumor immune response.
16:53And so just to summarize from this
16:54first portion of the talk we've we've
16:56described that circling tills are
16:58enriched with genes and are associated
17:00with clonal expansion specifically
17:02within the tumor and that the degree
17:04of cytotoxicity but not exhaustion
17:06are reflected in circulating tills.
17:07You find that tumor T cells that
17:10are predictive of immunotherapy
17:12response are also shared within the
17:14blood and that hallmark features
17:16of a productive anti tumor immune
17:18response may be reflected in the blood.
17:23So one of the assumptions from this early
17:26work was that the most tumor relevant
17:28or most relevant T cells to the anti
17:31tumor immune response are those that
17:34are most largely clonally expanded.
17:36And around the time that we were
17:37performing this initial work,
17:38there were also groups that had
17:41described that you can use strictly the
17:44transcriptional signature of T cells
17:46to accurately predict whether or not
17:49these T cells were neo oxygen specific,
17:52whether they're truly tumor specific.
17:55One such paper was out of Steve Rosenberg's
17:58group at the National Cancer Institute
18:00whereby he described 2 gene signatures,
18:02one for CD4T cells and one for CD8T
18:06cells that can with high accuracy,
18:09predicts whether or not a given T cell
18:12was likely to be neo Entergen specific.
18:15And so we simply ask the question,
18:17can the transcriptional identification
18:19of tumor specific T cells improve
18:22our understanding of the blood
18:24and tumor relationship.
18:29And so we then apply this transcription
18:32prediction to our own data.
18:33And so these are the CDAT cells that
18:36I had shown in the previous section
18:39and this is all from the tumor and
18:42we've identified those that we think
18:44are likely tumor neo antigen specific.
18:46I apologize about the colouring of
18:49the the graph on the right over here.
18:51But what we can see is that using
18:54our previous definition of expanded
18:55or unexpanded T cells that you
18:57have to trust me on the coloring,
18:59but the vast majority of them are also
19:01predicted to be in the antigen specific.
19:04But I think an important point
19:06is that of the unexpended,
19:07there's also a portion that we were
19:09not capturing before and that are
19:11actually unexpended within the tumor
19:14migraine environment to functionally
19:16confirm that these predicted T cells
19:20do in fact recognize new antigens.
19:23We collaborated with rapid farming
19:25medicines and with data that was
19:27generated by Ruth Haliband's lab as
19:30well to analyze wholexom sequencing
19:34and bulk RNA sequencing to be able to
19:37predict for each individual patients neo
19:40antigens and tumor associated antigens.
19:43We then ran these peptides that were
19:46synthesized in a relatively high
19:48throughput manner against select T cell
19:51receptor sequences and tested for reactivity.
19:56What we find in this data is that the
19:59vast majority of those NEO TCR predicted
20:03T cells account for basically all of
20:06the the T cell receptor sequences that
20:09elicited react functional reactivity
20:11and that the only clonotype that
20:14wasn't that was reactive but was not
20:17predicted to be neo antigen specific.
20:20It was in fact reactive to CMV and
20:23this peptide was included as a negative
20:25control for by the repertoire team.
20:30And so using this approach,
20:31we then analyzed cutaneous 17 patients
20:38with cutaneous Melanoma who are
20:40immunotherapy naive and we chose
20:42to focus on a more biologically
20:44homogeneous cohort to try and really
20:46eliminate any treatment related effects.
20:50We then applied the NEO TCR 8 and
20:52neo TCR 4 signatures to predict and
20:54identify reactive T cells and then in
20:57similar fashion link them back into
20:58T cells within the blood based off
21:01of their T cell receptor sequences.
21:04In total, we identified about 7000
21:06reactive CDAT cells which again
21:08reinforces that this is a relatively
21:11rare population.
21:15We again find that they're predominantly
21:18CD8 that they're highly expanded and
21:20also have a restricted clonal diversity.
21:23So what I'm showing here on the
21:25right is a linearized metric for
21:27the degree of clonal expansion
21:29within the blood and the tumor.
21:31Matched reactive are the ones that are
21:34predicted to be reactive based off of
21:37their tumor transcriptional signature.
21:39Unreactive are ones that were
21:41unreactive but also found within
21:43the tumor and then also the ones
21:46that were only found in the blood.
21:51We also find that there's a higher
21:54frequency of previously reported
21:56tumor antigen specific TCR sequences
21:59in our reactive population,
22:01and to do this analysis we use
22:04publicly available databases of TCR
22:06sequences that had been annotated
22:09with their functional epitopes
22:14to try and understand in a more specific
22:16manner the transcriptional features
22:18of these reactive T cell population.
22:20We collaborated with Doctor Yuval
22:22Kluger's group and they had developed
22:25a novel computational method for
22:28identifying the differential
22:30abundance of certain populations.
22:33Wes Lewis was a graduate student
22:35in his lab who applied this to our
22:38data set and what we find is that
22:40we can identify a subpopulation
22:42of cells that are differentially
22:44enriched for tumor reactive T cells.
22:49A look at the differential expression
22:51gene signature shows that in the
22:54unmatched and unreactive cells,
22:56there's again an enrichment
22:57for naive and memory markers.
22:59In line with our previous work,
23:01there's a high degree of cytotoxicity
23:03that's both found within our reactive
23:06and our unreactive populations.
23:08There's signs of cell trafficking,
23:10tissue resonance and MK associated markers
23:14and there's one marker in particular
23:16that really stood out to us and this
23:19is the killer cell immunoglobulin
23:20like receptor family which occurred
23:23to DL3 is one of those subtypes here.
23:27And the reason why this is interesting
23:29to us is because this work current
23:32expressing CDAT cells was recently
23:34described in autoimmunity and in
23:37infection as being important mechanism
23:40for restoring peripheral tolerance.
23:42So just a little bit about the CUR receptor.
23:44So they're best understood for
23:47their function and role within NK
23:49cells and they in part a negative
23:55suppression signal upon encounter
23:59with class one Class 2 MHC.
24:05So they're in in fact Co inhibitory
24:08signaling within NK cell within CDAT cells.
24:11So they denote this regulatory like
24:14T cell which is analogous to the live
24:1749 expressing CDAT cells that Harvey
24:20cancers group had described in mice.
24:22But these cells have a high expression
24:24of the transcription factor HELIOS,
24:26and although the mechanism
24:29isn't fully understood,
24:30they can target pathogenic T cells
24:33in autoimmune D infection and kill
24:36them in a contact dependent manner.
24:38And so in essence this is a alternative
24:41mechanism to try and eliminate
24:43hyperinflamed or hyperactive T cells.
24:47The role in tumor immunity is not really
24:50well described or well understood.
24:54So turning back to our data,
24:55we can take a look at gene signatures
24:58that are characteristic of these auto,
25:00these Kerr CD8 regulatory cells
25:03in autoimmunity and compared
25:05to them with our population,
25:06our reactive population in Melanoma.
25:10What we find is that there's a broad
25:12expression of the Kerr family of receptors.
25:14There's also high expression of
25:17cytotoxicity and NK associated genes
25:19in addition to many of the other
25:21features that I pointed out before.
25:23Importantly,
25:23there's a high expression of the
25:25transcription factor HELIOS,
25:26which is thought to be essential
25:28to their regulatory function or
25:30their suppressive function and a
25:32notable absence of Co stimulatory
25:36molecules. You can also see from
25:38the slide that this gene signature
25:41seems to be fairly specific for this
25:44reactive subpop subpopulation reactive
25:45cells as compared with all other
25:48CDAT cells found within the blood.
25:52We can also perform GENESAT
25:54enrichment analysis and we do find a
25:58statistically significant enrichment
25:59for the top 200 genes of human cure
26:03CDAT cells found in autoimmunity.
26:05And we can also ask the question,
26:07does this population or does
26:10this cure CDAT cell population?
26:12Does it represent distinct differentiation
26:15state or is it part of a continuum
26:18within clonally related T cells?
26:20And I performed pseudo time trajectory
26:22analysis here which attempts
26:24to try and order biologically
26:27related cells along a continuum.
26:31And what we find is that there seems
26:33to be a branch trajectory here and
26:36that in unsupervised analysis we also
26:41find that Helios which is encoded by
26:45the gene IKC F2 also came up as one of
26:48the most differentially expressed an
26:51associated genes along the trajectory.
26:56We can also ask the question,
26:57so if we think that these are
27:00regulatory cells within the blood,
27:01do they maintain their transcriptional
27:03state within the tumor?
27:04And in essence we're trying to
27:06understand what might there be,
27:07what might be their role within
27:10the tumor microenvironment.
27:11And So what we can do is we can
27:13trace these cells based off of
27:14their T cell receptor sequences
27:16back into the tumor and look at
27:18the transcriptional profile.
27:19And what we see is that this
27:21Kerr CD8T cell transcriptional
27:22profile is largely maintained
27:24within these sister clones within
27:26the tumor microenvironment.
27:30We have ongoing work in collaboration
27:32with Doctor Marcello Distasio and the
27:35Department of Pathology to try and
27:37better characterize these histologically
27:40using spatial multiomic analysis
27:42in the tumor micro environment.
27:48And so just to conclude from
27:49the 2nd portion of the talk,
27:51we've demonstrated that transcriptional
27:53signatures can identify a subset
27:55of tumor reactive T cells which
27:58are not clonally expanded.
27:59Differential abundance techniques
28:01can help us identify subpopulation
28:03of these reactive T cells which
28:07largely resemble Kerr CD8 regulatory
28:09T cells and that these Kerr CD8T
28:11cells seem to represent a distinct
28:13differentiation state which is preserved
28:15in the tumor micro environment.
28:21And so an important question for us
28:23is what is the clinical relevance
28:26of this T cell population And
28:28because we can't perform single cell
28:30sequencing on all of our our patients,
28:31we really wanted to move towards
28:34markers that could be assessed
28:36within the chemical laboratory.
28:38I'm specifically referring
28:39to using flow cytometry.
28:41And So what we wanted to do is
28:43to move from transcriptional
28:45features over to protein level
28:47cell surface features which are pre
28:51conventionally used in flow cytometry.
28:53And in order to do so,
28:55we collaborated with Doctor Steve
28:57Moss Group and Yuan Shin Chan and
29:00Ji Ping Wang are post docs and
29:01graduate students in his lab who
29:03primarily worked on this project.
29:04And we asked them to see whether or
29:08not they can construct A classifier
29:11that is limited to clinical
29:13variables and also genes that are
29:16specifically associated with protein
29:19cell surface proteins.
29:21We also restrict ourselves to
29:24genes which are known to correlate
29:26both at the transcriptional level
29:29and also the protein level,
29:30and so they use a they construct
29:32A LASSO logistic regression model
29:34which can accurately predict or
29:36classify cells as being likely
29:38within our subpopulation or not.
29:41And when we then apply this back
29:43into our single cell data set,
29:45what we find is that this tumor reactive
29:49or CDAT cell population seems to
29:52be associated with the poor survival.
29:54And what we did here was we simply
29:56split our cohort into a high
29:59expressing group and a low expressing
30:01group using a median cut point.
30:04And what I can tell you is that
30:06it doesn't matter whether or not
30:08these patients were immunotherapy
30:10naive or immunotherapy resistance.
30:12The mere presence of these cells seem to
30:15be associated with worst overall survival.
30:22Moving towards applying our classifier,
30:25applying these features to flow cytometry,
30:28we then asked them whether they can
30:31construct a hierarchy of these the
30:35these genes and protein markers in
30:37order for us to be able to develop
30:40combinations of markers that we
30:41can assess on flow cytometry.
30:45And so the first use single
30:47cell data that we had generated,
30:50but this includes protein level
30:52expression that from site seek data.
30:55And what I'm showing you here is
30:57that the the expression of KRD one
31:01as an example that the the site seek
31:05expression is relatively similar to
31:07what we would see on flow cytometry.
31:11We then constructed a decision
31:13tree model which allows us to
31:15assign a hierarchy and summarizes
31:17a combination of markers.
31:20And using this model and this,
31:22these, this combination of markers,
31:24we can accurately classify
31:27cells 91% of the time.
31:31And with the caveat that this is still
31:33ongoing work and that we have short
31:36interval fall for this exploratory cohort,
31:38we do see an early trend in
31:40separation curves that is in line
31:42with what we were seeing before.
31:44That is those who have a higher
31:47proportion of this subpopulation
31:49seem to have worse clinical outcome.
31:53And to validate both our transcriptional
31:56data and also our protein level data,
31:59we are collaborating.
32:00We established A collaboration
32:02with Doctor Benjamin Fairfax
32:04at the University of Oxford.
32:05He's a Melanoma oncologist who has
32:08generated bulk RNA sequencing data
32:11and also flow cytometry data from
32:14over 200 patients with Melanoma
32:16prior to treatment also on treatment.
32:19And so we're looking forward
32:22to seeing those results.
32:26And so just to summarize,
32:29we believe that the induction of
32:31systemic immunity is really a
32:32critical component to successful
32:34anti tumor immune responses,
32:35but that clinical biomarkers
32:37which allow us to profile on this
32:41population remains an unmet need.
32:45We use single cell technologies
32:46to try and provide insights into
32:49the relationship between T cells
32:50within the tumor and those within
32:53the blood and that we've identified
32:55a subpopulation of tumor reactive
32:57Cur CD8 regulatory T cells which
32:59may actually suppress anti tumor
33:01immunity and negatively correlate
33:03with clinical outcome.
33:05I think this is largely exploratory,
33:07but you know potentially if we
33:09can identify this cell population
33:11within a clinical cohort,
33:14we may be able to
33:17explore a new therapeutic Ave.
33:19for targeting these cells.
33:23And so with that I'd like to just
33:24take a moment to acknowledge all the
33:26people who've made this work possible.
33:28I think first and foremost we need
33:29to acknowledge the patients and
33:31families who are very generous in
33:33donating their tissue and blood.
33:35But also I'd like to thank them and
33:37acknowledge them just for the motivation
33:38that they provide all of us for the work
33:40that we do in the clinic, in the lab.
33:43And also like to thank my mentors Dr.
33:45Hathor and Dr.
33:46Kruger for their unending support
33:49and really the opportunity to
33:51perform this research in addition
33:53to members of the halfway lab.
33:55So Liliana Luca had mentioned before is
33:59a independent investigator in France.
34:01Pierre, Paul and Nick were also essential and
34:04instrumental in generating data on the study.
34:08Our collaborators both internally
34:09here at Yale and also externally.
34:12So I'm Doctor Yuval Kluger's group
34:14and Wes Lewis's Wes Lewis for their
34:16work on the Differential Abundance
34:18analysis and Steve Ma Yuan Shin Chen
34:21G Ping Wang for their work on our
34:24constructing biomarker classifiers.
34:26As I mentioned before,
34:28Martello Distasio,
34:29we have an ongoing collaboration
34:31to explore the spatial orientation
34:33of the cell population and Doctor
34:35Benjamin Fairfax is a collaborator
34:39who's going to help us explore this
34:43population in a larger cohort.
34:45I'd also like to acknowledge our
34:47collaborators at Repertoire Immune Medicines,
34:49in addition to the Yale Skins Board whose
34:53support has really made this effort feasible,
34:57and also to the core facilities here at Yale.
35:00And a personal thank you to both
35:01David Braun and David Schoenfeld,
35:03who unfortunately couldn't
35:04be here in person today.
35:06But they were incredibly generous and
35:08help with their thoughts and also with
35:10their time in helping prepare for this
35:12presentation and also for my funding sources,
35:14the T32 and the K12,
35:15as Harry had mentioned before.
35:18OK.
35:18I'd be happy to take any questions.
35:26Thank you, Ben, for a
35:28terrific talk. Any questions?
35:33So while people, so we have a few online.
35:36Oh yeah, let's do that. Yeah. So
35:40I don't know if people want to unmute,
35:42but I see that SRIVATAM,
35:46has there been an effort to isolate
35:48and phenotypically characterize
35:49these current CDAT cells?
35:50I'm curious to understand the
35:52uncommon state of CDAT cells.
35:54So yes, there has been work in
35:56other contexts to do that and
35:57I didn't show the data today,
35:59but we have also done that
36:01in Melanoma and have largely
36:03validated the the protein level
36:06immunophenotypes of these cells.
36:10The next question is from Marcus Bosenberg.
36:14He has do you have a hypothesis as
36:16to how Cur CDA regulatory T cells
36:19negatively affect anti cancer
36:21immune responses and outcome.
36:24You know I think the mechanism
36:26for these Cur CDA T cells is still
36:29really not fully understood.
36:31The hypothesis has kind of demonstrated
36:34here on this side or our hypothesis
36:37is that they're somehow impacting
36:40tumor antigen specific CDAT cells
36:42in the tumor micro environment.
36:45I'm currently in the process of setting up
36:48assays to try and assess this functionally,
36:53but my guess would be and it's
36:55also possible that they're
36:57they're impacting CD4T cells,
36:59which is a more direct link from
37:01the autoimmunity literature.
37:02But we're first going to explore the
37:06CDA component because of this negative
37:09in fact impact that we see in tumors.
37:12And then the last question was are
37:13these cells called regulatory based
37:15on their transcriptional features.
37:16So this cell population was as I
37:21mentioned described both in mice and
37:24also in human autoimmunity infection
37:27because they are able to actually
37:30functionally kill autoreactive T cells.
37:35And so it's not simply just based
37:37off of transcriptional features,
37:38although our data is certainly inferring
37:40from the transcriptional expression.
37:45I actually have a follow
37:46up question to Marcus's.
37:47Do you think that these are positive
37:50CDAT cells might stick and you can
37:53revert them to the per negative
37:56being A tag even because there
37:57are antibodies that have been actually
38:00given to humans that do that.
38:02Yeah. So I think the NKG 2DA
38:06antibodies which impact the the kind
38:10of analogous Co stimulatory molecule,
38:12not the Co inhibitory molecule have
38:14been tried and I don't think the
38:16data has been all that great for it.
38:19But in terms of the plasticity
38:21of this cell type,
38:22I think especially based off of
38:25our the trajectory analysis,
38:27I think that it is interesting to try and
38:29explore how plastic the cell population is.
38:31It does seem like there's a branch
38:33differentiation trajectory,
38:34but we just don't understand quite yet how
38:38these cells are really being generated,
38:40what it under what context and to
38:44really truly demonstrate their function, just
38:52speak up. Yeah.
38:54So with this model in mind, when you
38:57look at cohorts that are receiving IO,
38:59there's a relationship between force
39:02survival and curve positive cell
39:05strength compared to if you look at
39:07cohorts that aren't receiving IO.
39:09Yeah. So the the P value remains about
39:12the same actually in both cohorts.
39:15And I think that that that's
39:18a really interesting point.
39:20But whether there's a subpopation of
39:23patients where this is a primary,
39:25I mean secondary resistance
39:27mechanism I think is worth exploring
39:31wonderful talk. The these care
39:34suppressor cells that Mark
39:36Davis identified a really hot
39:38issue in human immunology now.
39:39But just looking at the slide again,
39:41do you think the tumor reactive T
39:43cells may express the log in for
39:44digit CD155 and we looked at that,
39:48are you talking about the, the regulatory,
39:50so the cures in cells or the no,
39:53well, the ones on the left,
39:54well the the regulatory cells that
39:56care positive expressed digit,
39:58yes, they do. The ligand is
40:01CD155 and what we're learning
40:02about what CD155 engagement
40:04does to cancer cells as per PPG.
40:08I'm just wondering if CD155IS
40:11expressed on the tumor reactive
40:12T cells, we looked at that.
40:15I don't, I mean I'm I'm inferring
40:16we haven't directly looked at that,
40:18but I'm inferring and I'd be guessing that
40:22the expression if present is very low,
40:24but I think it's definitely worth exploring.
40:27I, I think the effects of anti
40:28tiggering anti PD one on the cell
40:30population for example is something
40:32that we can take a look at. Absolutely.
40:36Another question, have you done
40:37some of the analysis in other
40:39tumor types in the Lumpsor study?
40:42Do you see the same new cell
40:46population of CEA tumor?
40:48So I haven't so So for those
40:50who aren't able to hear online,
40:52Doctor Cleaver asks whether
40:54I've also found this cell
40:56population in other tumor types.
40:57I haven't specifically
40:58looked under this lens.
41:00You know, one of the the reasons why I
41:04think maybe if this population is real,
41:07it may not be as well described is because
41:10it has a large transcriptional overlap
41:12with other cytotoxic CDAT cell populations.
41:15And so we really need to look
41:18carefully The enrichment cures,
41:20for example, can be expressed
41:23on just activated CDAT cells.
41:26And so we really need to look a little
41:28bit more carefully at some of the
41:30other markers like it Grows or HELIOS.
41:37There is some more question online.
41:38So from the inside why
41:40cure CDAT cell products,
41:42can you prevent cure CDT cell product?
41:45I mean, I I apologize, I'm not,
41:48I'm not sure I fully understand.
41:51I'm not sure if you mean AT cell product
41:54that's targeting cure CDAT cells,
41:57not sure if you're available
41:59to unmute and discuss.
42:01But I I I do think that exploring a
42:05therapy that would target or impact
42:08these cells would be of interest. OK.
42:14Thank you everyone. Thank you.