Pathology Grand Rounds, October 17, 2024

Transcript

• 00:01Today, I'd like to, introduce
• 00:03today's grand round speaker is,
• 00:05Valsamo or Elsa Anagnosto.
• 00:08Many of you who've been
• 00:09here for a long time
• 00:10remember Elsa. She was a
• 00:12postdoc in my lab, and
• 00:13then she was a resident
• 00:14in internal medicine here.
• 00:16She did her medical school
• 00:17training, however, in Greece, the
• 00:19National and Capedestrian
• 00:20University of Athens, and then
• 00:22came did her PhD there,
• 00:24then came to do a,
• 00:25postdoctoral fellowship in my lab
• 00:27after which she did internal
• 00:28medicine here at Yale, after
• 00:30which she,
• 00:31became she went to Johns
• 00:33Hopkins. We we lost her
• 00:34on this one. She went
• 00:35to Johns Hopkins for oncology
• 00:36fellowship,
• 00:37and then she's been there
• 00:38rising through the ranks now
• 00:40to the level of associate
• 00:41professor and also the co
• 00:42director of thoracic oncology.
• 00:45She's a very accomplished
• 00:48scientist as well as a
• 00:49doctor. She has over a
• 00:51hundred publications and an h
• 00:52index of forty nine with
• 00:54over fifteen thousand citations. So
• 00:56and that's not to mention
• 00:57all the awards she's won
• 00:58on the way up. So
• 00:59without further ado, I'll let
• 01:00Elsa give the talk.
• 01:02Thank you.
• 01:03Thank you.
• 01:05David?
• 01:08Thank you for the kind
• 01:09introduction. You can hear me
• 01:10okay?
• 01:11Okay. Wonderful.
• 01:12It's hard to beat your
• 01:13h index,
• 01:14David, but I'm working.
• 01:17I'm working on it. It's
• 01:18truly a great pleasure and
• 01:19a great honor to to
• 01:21be here today for pathology
• 01:22grand rounds. As David said,
• 01:24I spent, three wonderful years
• 01:26as a postdoctoral fellow in
• 01:28his lab, and then I
• 01:29continue to come back to
• 01:30this very same auditorium,
• 01:32for medicine,
• 01:34rounds as as an internal
• 01:35medicine
• 01:36resident.
• 01:37And then I left for
• 01:39Hopkins, where I now am
• 01:41for the past several years,
• 01:42and I now co direct
• 01:44the upper aerogestive malignancies
• 01:46program. I also co lead
• 01:48a number of precision oncology
• 01:50efforts at Johns Hopkins,
• 01:52that include our,
• 01:54molecular tumor board, as well
• 01:56as our thoracic oncology precision
• 01:58center,
• 01:59center of excellence.
• 02:01So very briefly I'm going
• 02:03to go over my disclosures.
• 02:06And I wanted to start
• 02:08with a brief layout of
• 02:09my talk today.
• 02:11So we will focus on
• 02:13a nuanced genomic features of
• 02:15immunotherapy
• 02:16response and then our studies
• 02:18in trying to capture
• 02:20evolution of cancer cells together
• 02:22with the tumor microenvironments
• 02:24in the context of
• 02:25immunotherapy.
• 02:27And I'm going to approach
• 02:28this as my lab does
• 02:30under,
• 02:30two angles. First, under the
• 02:32angle of multiomic
• 02:34tissue analysis, and second, under
• 02:36the angle of liquid biopsy
• 02:38analysis. And then, you know,
• 02:40ultimately,
• 02:42talk about how do we
• 02:43bridge these efforts, the, you
• 02:45know, the scientific discoveries, if
• 02:47you will, with,
• 02:48clinical cancer care, and how
• 02:50do we translate these findings
• 02:52into interventions that,
• 02:54actually improve patient outcomes.
• 02:56And so certainly
• 02:58the
• 02:58genomic wiring of response to
• 03:01immunotherapy
• 03:02and
• 03:03in general,
• 03:05immunotherapy responses
• 03:07is a complex phenomenon. It
• 03:09it really
• 03:10depends on the interactions
• 03:11between tumor intrinsic and tumor
• 03:13extrinsic parameters that all happens
• 03:15in the tumor microenvironment.
• 03:17It's actually hard to capture
• 03:18and it's hard to translate
• 03:20into
• 03:21biomarkers of immunotherapy response.
• 03:23So there is a lot
• 03:24of value
• 03:26to actually go beyond concepts
• 03:28like tumor mutation burden, which
• 03:30is a mere reflection of
• 03:31mutations for coding sequence or
• 03:33per megabase,
• 03:35to,
• 03:36refined and nuanced
• 03:38genomic feature analysis and trying
• 03:40to dissect,
• 03:42the genomic landscape of tumors
• 03:44and understand and capture
• 03:47sets of mutations
• 03:48with differential weights within the
• 03:50tumor mutation burden, and some
• 03:51of which I'm going to,
• 03:53to talk to you about
• 03:54today. But what we're showing
• 03:56here as as as a
• 03:58starter, right? So it's it's
• 04:00one of the more established
• 04:01or the most established,
• 04:02biomarkers of response to immunotherapy
• 04:05that has to do with
• 04:05tumor mutation burden that has
• 04:07inherent challenges, and these are
• 04:09both biological
• 04:10and technical as as we're
• 04:12showing here.
• 04:14Nevertheless,
• 04:15this, tumor mutation burden is
• 04:18one of the most,
• 04:20commonly used biomarkers for immunotherapy
• 04:22response if one also, you
• 04:23know, considers together with PDL1
• 04:25expression. And of course, the
• 04:27association of tumor mutation burden
• 04:29with immune checkpoint blockade response
• 04:32has been established and shown
• 04:34in a number of studies.
• 04:35Now you can appreciate the
• 04:36dose
• 04:37dependent relationship between tumor mutation
• 04:40burden and immunotherapy responses is
• 04:42truly exemplified in the context
• 04:44of hypermutated
• 04:45tumors
• 04:46and and and MSI.
• 04:48But as I just told
• 04:49you, it is an imperfect
• 04:51biomarker,
• 04:52and the challenges are both
• 04:53technical
• 04:54and biological. So the question
• 04:56that I'm going to pose,
• 04:57and I'm actually gonna pose
• 04:58a lot of rhetoric questions
• 05:00throughout this talk,
• 05:01is,
• 05:03how do we identify those
• 05:06biologically
• 05:06distinct
• 05:08set of sets of mutations
• 05:10and mutation associated neoantigens
• 05:12within the overall tumor mutation
• 05:14burden?
• 05:15And historically,
• 05:16we have been focusing
• 05:19on sequence based
• 05:21analysis, right, but there's a
• 05:22lot of value in actually,
• 05:25understanding
• 05:26structural
• 05:27genome wide copy number profiles.
• 05:30And of course, there is,
• 05:32there are indications that tumor
• 05:34aneuploidy
• 05:35is,
• 05:36associated with lack of response
• 05:38to,
• 05:40to immunotherapy.
• 05:41But there is a lot
• 05:42of unknowns. So what we're
• 05:44showing you here is
• 05:47an analysis,
• 05:48of a cohort of patients
• 05:50with mesothelioma
• 05:51that received chemoimmunotherapy
• 05:53for which we performed whole
• 05:54exome sequencing and performed genome
• 05:56wide copy number analysis.
• 05:59And, let me orient you
• 06:01a little bit to this
• 06:02figure. So each patient's tumor
• 06:04sample is shown as a
• 06:06row. Each chromosome is shown
• 06:07as a column. And then
• 06:09aggregate measures of tumor aneuploidy
• 06:11are shown on the right
• 06:12hand side. The tumors are
• 06:14ordered,
• 06:15in the following order. So
• 06:17regressing tumors with immunotherapy,
• 06:19responding tumors are shown in
• 06:21the top two thirds, and
• 06:22non responding tumors are shown
• 06:23in the bottom third.
• 06:25And what you'll see here,
• 06:27copy number obviously
• 06:29changes, the copy number gains
• 06:31are shown in red and
• 06:32copy number losses are shown
• 06:33in blue.
• 06:34And what we saw from
• 06:35these analysis was practically that
• 06:37there was an enrichment in
• 06:40tumor aneuploidy features in tumors
• 06:42that are
• 06:43responding to chemoimmunotherapy.
• 06:47So in this context, and
• 06:49again remember this is mesothelioma,
• 06:51so TMB
• 06:52low
• 06:53tumor type, right? So in
• 06:54this context, this was the
• 06:55structural, not the sequence landscape
• 06:57that was informative with respect
• 06:59to immunotherapy response. So we
• 07:00started asking why is that.
• 07:02So what I'm showing you
• 07:03here on the left hand
• 07:04side is an increased number
• 07:06of break points across the
• 07:08genome,
• 07:09again, in tumors that are
• 07:11responding,
• 07:12to to chemoimmunotherapy
• 07:14and a higher degree of
• 07:15homologous recombination
• 07:17deficiency, again, in tumors regressing
• 07:19with immunotherapy, and these were
• 07:20patients that did well,
• 07:22on chemoimmunotherapy,
• 07:24that led us to form,
• 07:27the following hypothesis.
• 07:29We looked at the number
• 07:30of mutations
• 07:31residing in haploid regions of
• 07:33the genome. And this is
• 07:34what I'm showing you on
• 07:35the right hand side. Well,
• 07:37you'll appreciate that
• 07:39tumors harboring a higher number
• 07:41of mutations in haploid regions,
• 07:43again, single copies, right?
• 07:45These are tumors that regress
• 07:47with immunotherapy.
• 07:48That was a very interesting
• 07:49observation and, led to the
• 07:51question,
• 07:52why is that? And so
• 07:53the hypothesis we formed was
• 07:55actually that,
• 07:57these are two these are
• 07:58mutations that are hard to
• 08:00lose.
• 08:00They,
• 08:01reside in in in in
• 08:03single, in single copies. Right?
• 08:05So this could potentially loss
• 08:07of these mutations,
• 08:09in the context of tumor
• 08:10evolution can be legal to
• 08:12the cancer cell, right, as
• 08:13these,
• 08:14exist in linkage with essential
• 08:16genes. So that that that
• 08:17was that was interesting. So
• 08:19then, this led us to
• 08:21form the following hypothesis,
• 08:25and and pose the following
• 08:26question,
• 08:27which is,
• 08:29are there mutations
• 08:30that cannot be lost during
• 08:33selective or under selective pressure
• 08:36of immunotherapy.
• 08:38And, and if this is
• 08:39the case, then these are
• 08:41mutations
• 08:41that cannot, and associated neoantigens,
• 08:44that cannot be edited out
• 08:47during immunotherapy.
• 08:49And practically a way to
• 08:50think about it is,
• 08:52as,
• 08:53having cancer cells being continuously
• 08:56tagged for elimination by the
• 08:58immune system, right? So if
• 09:00this was true, then this
• 09:02is a biologically
• 09:03distinct subset
• 09:05of mutations within the overall
• 09:07TME,
• 09:08which practically confers a fitness
• 09:11disadvantage, if you will,
• 09:13for for the cancer cells.
• 09:14So which ones are these
• 09:15mutations? I showed you mutations
• 09:17residing in haploid regions.
• 09:19But then are these that
• 09:21we're showing here in the
• 09:22middle, as these the only
• 09:23ones,
• 09:26we considered next mutations in
• 09:28multiple copies. Right? So again,
• 09:31you can imagine that these
• 09:32mutations again are hard to
• 09:34be eliminated
• 09:35because,
• 09:36elimination would require multiple copy
• 09:39number events, which is less
• 09:41likely to happen in the
• 09:42context of of tumor evolution.
• 09:44So we termed
• 09:45only copy mutations and mutations
• 09:47and multiple copies as persistent
• 09:50mutations
• 09:51and considered
• 09:52those as a biologically
• 09:54uneditable
• 09:56set of mutations within the
• 09:57overall tumor mutation burden in
• 09:59contrast to loss prone mutations.
• 10:06With respect to previous
• 10:08studies on
• 10:10immunotherapy acquired resistance that we
• 10:12did several years back, where
• 10:14in performing
• 10:15comparative whole exome sequencing analysis
• 10:18of tumors prior to immunotherapy
• 10:20initiation and the time of
• 10:22acquired resistance,
• 10:23we discovered that there is
• 10:24a number of mutations and
• 10:26mutation associated neoantigens
• 10:29that are actually lost at
• 10:30the time of acquired resistance.
• 10:32And I'm showing you an
• 10:33example here. These are IGV
• 10:34plots. You can appreciate,
• 10:37mutant reads at baseline time
• 10:38points and then no mutant
• 10:40reads whatsoever
• 10:41at the resistant
• 10:42at the time of acquired
• 10:43resistance.
• 10:44And so from this we
• 10:45concluded that there is a
• 10:47subset of mutations and mutation
• 10:49associated neoantigens
• 10:50that are actually
• 10:52eliminated
• 10:53at the time of acquired
• 10:54resistance. And this is what,
• 10:55you know, potentially drives acquired
• 10:57resistance
• 10:58in, in this setting.
• 11:00And we then ask the
• 11:01question, what is the mechanism?
• 11:03Right? So if this is
• 11:04the case, what is the
• 11:05mechanism? And of course, you
• 11:06can imagine that the first
• 11:07mechanism is through immune elimination
• 11:09of neoantigen containing
• 11:11cells. These are a subset
• 11:13of the tumor,
• 11:14cell population and then there
• 11:15is outgrowth of the remaining
• 11:17cells. But the second mechanism
• 11:20is one that requires
• 11:21additional genetic events and loss
• 11:24of heterozygosity
• 11:25through chromosomal deletions,
• 11:28that practically
• 11:30the chromosomal region that contains
• 11:32the mutation is lost under
• 11:33selective pressure of immunotherapy
• 11:35followed by selection and expansion
• 11:37of the resistant clone. So
• 11:38then we ask the question,
• 11:39well let's look at mutations
• 11:41that appear to be eliminated
• 11:42at the time of acquired
• 11:43resistance. How are these eliminated?
• 11:46So the interesting finding here,
• 11:48and I'm pointing to a
• 11:49B allele frequency graph, this
• 11:51is chromosome seventeen that we're
• 11:52showing here.
• 11:54You can appreciate
• 11:56loss. So so what is
• 11:58a point five is indicates
• 11:59heterozygosity,
• 12:00and then any deviation,
• 12:03includes,
• 12:04points to,
• 12:05to to to LOH. So
• 12:07you'll appreciate that there is
• 12:08a region here on chromosome
• 12:10seventeen
• 12:11that is lost at the
• 12:12time of acquired resistance, and
• 12:14we this contained,
• 12:16three mutations, mutation associated neoantigens
• 12:19for this case,
• 12:21all of these being clonal.
• 12:23So what we found
• 12:25was that all clonal
• 12:27eliminated neoantigens
• 12:28were actually lost by these
• 12:30additional genetic events,
• 12:32via chromosomal deletions. And of
• 12:34course, the next question we
• 12:36asked is were these
• 12:38mutations and mutation associated neoantigens
• 12:40relevant targets,
• 12:42for, for,
• 12:45for T cells in terms
• 12:46of eliciting anti tumor immune
• 12:48responses. So what we did
• 12:50is we set up these
• 12:51T cell cultures and stimulated
• 12:53T cells from the same
• 12:54patients with neopeptides that we
• 12:56synthesized based on the whole
• 12:58exome sequence data. And this
• 12:59is what we're showing here,
• 13:02at the bottom. And what
• 13:03we found was that every
• 13:04single one of those eliminated
• 13:07neoantigens
• 13:08when we appulsed
• 13:10autologous T cells from the
• 13:11same patients prior to immunotherapy
• 13:13initiation,
• 13:15induced
• 13:16neoantigen
• 13:17specific clonotypic
• 13:19expansions and thus
• 13:21represented
• 13:22viable targets
• 13:24for the immune system and
• 13:26potentially drove the initial anti
• 13:28tumor immune response.
• 13:30So now let's flip the
• 13:31question mutations. If neoantigen elimination
• 13:32is a mechanism of, immunotherapy
• 13:34resistance, what happens with persistent
• 13:35mutations that actually cannot be
• 13:36lost, right, because of their
• 13:38properties,
• 13:39because of the regions that
• 13:47the bees reside on.
• 13:49So,
• 13:50we looked into this very
• 13:51deeply. We first assessed, and
• 13:53this is all TCGA data.
• 13:55We looked at the distribution,
• 13:57of,
• 13:58practically the fraction of the
• 14:00genome and multiple copies and
• 14:01single copies. This is what
• 14:03we're showing up top. And
• 14:04you can appreciate the, the
• 14:06different distributions based on cancer
• 14:08lineage. All the TCGA cancer
• 14:10lineages are shown,
• 14:11at the bottom here. And
• 14:13then, this was followed by
• 14:15an assessment of the fraction
• 14:17of mutations and multiple copies
• 14:18and only copies. And again,
• 14:20you can appreciate the fact
• 14:21that this differs based on
• 14:22the cancer lineage.
• 14:24And then I think it's
• 14:25interesting to consider the distribution
• 14:27of persistent mutations,
• 14:30also
• 14:31considering the background tumor mutation
• 14:32burden, which we're showing here
• 14:34as a gray shaded area.
• 14:36So you'll appreciate that, there
• 14:38is and I'll show you
• 14:39more data on this but
• 14:40practically there is no true
• 14:42association between the overall very
• 14:44crude tumor mutation burden and
• 14:46the number of persistent mutations,
• 14:48you know, across different lineages.
• 14:50This is shown here as
• 14:51well where you can appreciate
• 14:53that differential correlation
• 14:55between
• 14:56the overall tumor mutation burden,
• 14:57persistent tumor mutation burden across
• 15:00different lineages.
• 15:01And then by employing a
• 15:02quantile approach, we ask the
• 15:04question, if we were to
• 15:06reclassify
• 15:07based on persistent tumor mutation
• 15:09burden, what would we find?
• 15:11And it was interesting in
• 15:12the sense that we saw
• 15:13an average,
• 15:14reclassification
• 15:15rate of thirty three percent.
• 15:19Differential
• 15:20reclassification
• 15:21of tumors based on this
• 15:23biologically
• 15:24distinct subset of mutations within
• 15:26the overall TMB.
• 15:28So if this all holds
• 15:29true, right, so we should
• 15:31actually see the clinical benefit
• 15:33of patients with tumors harboring
• 15:35a higher density of persistent
• 15:37mutations
• 15:38with respect to response to
• 15:39immunotherapy,
• 15:40so we took
• 15:41close to five fifty patients
• 15:43across lineages, this is what
• 15:44we're showing you here. And
• 15:46in every
• 15:47single cohort, we looked at
• 15:49what we found
• 15:51was that persistent tumor mutation
• 15:53burden better differentiated
• 15:55responding from non responding tumors
• 15:57across lineages. You can appreciate
• 15:59melanoma, head and neck, mesothelioma,
• 16:01and non small cell lung
• 16:02cancer. I just wanted to
• 16:04show you an example here.
• 16:05This is, a Circos plot
• 16:07representing a patient's genome. This
• 16:09was a patient with melanoma.
• 16:11Well, the copy number
• 16:14profiles are shown in the
• 16:15outer ring. We're showing LOH
• 16:18as, you know, those blue
• 16:19bars in the inner ring.
• 16:21And then in the
• 16:23middle ring and then in
• 16:24the inner ring, what we're
• 16:25showing,
• 16:26are the mutations with persistent
• 16:29mutations being the yellow dots,
• 16:31on the inner ring.
• 16:33This was a patient that
• 16:34if one looked at the
• 16:35overall tumor mutation burden, they
• 16:37were,
• 16:38not as high, but then,
• 16:40they wouldn't be classified as,
• 16:43you know, high
• 16:44tumor with a high TMB.
• 16:45But if one actually looks
• 16:47at the persistent mutation density,
• 16:49they were among the highest
• 16:50ones. And this is indeed
• 16:51a patient that did very,
• 16:53very
• 16:54well. Again, highlighting the importance
• 16:56to actually focus on what
• 16:57is biologically relevant,
• 16:59right, within the overall tumor
• 17:00mutation burden.
• 17:02We asked additional questions in
• 17:04terms of, you know, if
• 17:05all of hypothesis,
• 17:06held true, then persistent mutations
• 17:08shouldn't be eliminated in the
• 17:10context of of tumor evolution.
• 17:11And this is,
• 17:13what we did here where
• 17:14we're practically biopsying tumors prior
• 17:16to immunotherapy and then again
• 17:18at the time of resistance.
• 17:20And I'll draw your attention
• 17:21to,
• 17:22the blue bars here. So
• 17:23these are the,
• 17:26the persistent mutations that are
• 17:28maintained
• 17:29compared to what is lost,
• 17:31which are shown, which is
• 17:33shown in the pale orange,
• 17:37boxes,
• 17:38here
• 17:38on the graph. So you
• 17:39can appreciate practically that the
• 17:41mutations that are predominantly
• 17:42lost are loss prone mutations.
• 17:45So persistent mutations persist, and
• 17:47this was a nice,
• 17:49validation, if you will, of
• 17:51our hypothesis.
• 17:52And certainly mutations,
• 17:54certainly tumors with a high
• 17:56density of persistent mutations, these
• 17:58had a very inflamed
• 18:00TME phenotype. You can appreciate
• 18:02here in terms of enrichment
• 18:03and interferon gamma response,
• 18:06gene sets that was actually
• 18:08more pronounced
• 18:09as patients received immune checkpoint
• 18:11blockade.
• 18:12Certainly,
• 18:14more pronounced
• 18:15in terms of tumors that
• 18:17harbored a traditional high tumor
• 18:19mutation burden.
• 18:21So
• 18:22again,
• 18:24to highlight
• 18:25the importance of going past
• 18:26tumor mutation burden, but I
• 18:28think a realization in the
• 18:29clinic
• 18:30is that resistance emerges, right?
• 18:33And so there is patients
• 18:34with primary resistance and there
• 18:36is,
• 18:37certainly patients that initially respond,
• 18:40but acquired resistance emerges. And
• 18:42this is a sizable fraction,
• 18:44right? So
• 18:46one of her focus has
• 18:47been to
• 18:48identify
• 18:50ways to potentially
• 18:52circumvent immunotherapy
• 18:53resistance and,
• 18:55focus on, the example that
• 18:57I'm showing you here today
• 18:59is from
• 19:01multiomic
• 19:02integrative
• 19:03whole exome serial transcriptome
• 19:06and TCR sequencing,
• 19:08from a cohort of non
• 19:10small cell lung cancer patients,
• 19:11all with metastatic disease,
• 19:14treated in a randomized,
• 19:16trial
• 19:18by radiation
• 19:20plus pembrolizumab
• 19:21versus pembrolizumab
• 19:22alone.
• 19:23And in doing these integrative
• 19:25analysis, what we found out,
• 19:26which was very surprising to
• 19:28us at the beginning, was
• 19:29what I'm showing you here
• 19:30at the bottom, which is
• 19:32practically that patients with low
• 19:34tumor mutation burden, patients with
• 19:36no PD L1 tumors, rather,
• 19:38with no PD L1 expression,
• 19:40and tumors,
• 19:42that harbored Wnt mutations,
• 19:45which again, you know, these
• 19:46are typically,
• 19:48associated with a cold tumor
• 19:49microenvironment.
• 19:51These were actually the patients
• 19:53that did well
• 19:55in
• 19:56the radiotherapy,
• 19:57in the combined radiotherapy
• 19:58plus
• 19:59immunotherapy
• 20:00arm. And here, this is
• 20:03indicated here as the SBRT
• 20:05arm. So that was initially
• 20:07surprising, right? So why would
• 20:08we see responses in immune
• 20:09cold tumors?
• 20:12So we started looking further,
• 20:14into this and of course
• 20:15there is a lot of,
• 20:18preclinical for the most part
• 20:19and some,
• 20:21clinical evidence in terms of
• 20:22the immunostimulatory
• 20:24effect of radiation in terms
• 20:26of priming,
• 20:27anti tumor,
• 20:28and boosting anti tumor immune
• 20:30responses. And of course, this
• 20:33conceptually
• 20:34happens both at
• 20:36the site of radiation as
• 20:37well as at abscopal sites,
• 20:39right? This is the abscopal
• 20:40effect of radiotherapy.
• 20:43And this is precisely
• 20:44what we saw
• 20:46here where we performed
• 20:49serial RNA seq analysis
• 20:52of
• 20:53tumors
• 20:54that were not irradiated.
• 20:56These were abscopal sites.
• 20:58And what we found was
• 21:00that in tumors that harbored
• 21:02low tumor mutation burden or
• 21:04were PDL1
• 21:06null, there was a significant
• 21:08upregulation of adaptive immunity gene
• 21:10sets.
• 21:12And
• 21:13this was certainly
• 21:15the case for the SBRT
• 21:16arm compared to the control
• 21:18arm. And you can appreciate
• 21:20the leading edges that we're
• 21:21showing here, which are very
• 21:22clear,
• 21:24in via's, putatively
• 21:26immunologically
• 21:27cold tumors.
• 21:29And, you know,
• 21:31the
• 21:33additional evidence supporting this is
• 21:36now looking in an orthogonal
• 21:37manner
• 21:38at TCR Vbeta sequencing, both
• 21:40in intratumoral repertoires as well
• 21:42as in peripheral blood compartment.
• 21:44And what we found there
• 21:46was that there was a
• 21:47significant
• 21:52enrichment
• 21:52of new that we're showing
• 21:54in A as well
• 21:56as preexisting
• 21:59TCR
• 22:00chronotypic expansions
• 22:02in the SB or T
• 22:03arm. Again, we're showing here
• 22:04in blue. And that was
• 22:06independent of PD L1
• 22:08and, TMB
• 22:10status that we're showing in
• 22:11the middle panels. And what
• 22:12you'll appreciate in looking at
• 22:13the bottom panels here, these
• 22:14are two representative examples
• 22:16of patients in the SBRT
• 22:18arm with,
• 22:19TMB low or PD L1
• 22:21null tumors. And you can
• 22:22appreciate in the volcano plots
• 22:24the enrichment
• 22:25of,
• 22:26TCR clones
• 22:28in different,
• 22:29in different compartments.
• 22:31So
• 22:32these findings
• 22:34are important
• 22:35as we are constantly trying
• 22:37to identify
• 22:38ways to
• 22:40sensitize
• 22:41cold tumors or resensitize
• 22:43tumors that,
• 22:44initially regressed with immunotherapy
• 22:47but are now resistant.
• 22:49And, and, you know, radiotherapy
• 22:51has certainly,
• 22:53shown promise,
• 22:54but we
• 22:56haven't been able to,
• 22:58incorporate
• 22:59as there is clinically at
• 23:01least as,
• 23:03there is specific patients that
• 23:05can benefit from this approach.
• 23:07Now
• 23:08what happens
• 23:10if we actually do not
• 23:11know
• 23:12the mechanism
• 23:13of resistance, right? And
• 23:15I presented to you some
• 23:17of the genomic features that,
• 23:19confer or associated with response
• 23:22and resistance to immunotherapy,
• 23:24potentially ways to think about
• 23:26resensitizing
• 23:27tumors. But
• 23:28how about when we actually
• 23:29do not know
• 23:30what the mechanism of resistance
• 23:32is? Can we still capture
• 23:35response and resistance and can
• 23:36we actually intervene? And so
• 23:38that's where liquid biopsies are
• 23:39coming in, which is the
• 23:41second major focus of my
• 23:43lab currently.
• 23:44So, just by means of
• 23:46introduction,
• 23:47liquid biopsies are certainly emerging
• 23:49as potentially powerful approaches to
• 23:51capture tumor burden, to monitor
• 23:53tumor evolution,
• 23:56across, I would argue, across
• 23:58the cancer care continuum. And
• 24:00certainly there is a lot
• 24:02of promise
• 24:03as ways and approaches to
• 24:05capture minimal residual disease,
• 24:07as well as
• 24:09therapeutic
• 24:10response in the metastatic setting.
• 24:12So before talking about the
• 24:14opportunities
• 24:15of liquid biopsies, I wanted
• 24:17to take a step back
• 24:18and review the
• 24:20compendium of alterations that are
• 24:22detected by liquid biopsy approaches.
• 24:22And, I'm
• 24:27listing those here. And of
• 24:28course, what you'll appreciate is
• 24:30that,
• 24:31we can reliably
• 24:32detect not only sequence alterations,
• 24:35but also larger structural
• 24:38alterations in cell free DNA,
• 24:41including,
• 24:43physical properties of DNA fragments
• 24:46that have to do with
• 24:47fragment sizes,
• 24:48as well as cell free
• 24:50DNA and motifs. And I'm
• 24:51going to show you some
• 24:52data
• 24:53later on.
• 24:55But certainly, liquid biopsy approaches
• 24:57come with challenges. And one
• 24:59of the challenges has to
• 25:00do it's inherent to the
• 25:01scarcity of the analyte itself.
• 25:03Right? So mutant DNA
• 25:05that is tumor derived is
• 25:07present in very, very small
• 25:08quantities in the circulation. So
• 25:10technically,
• 25:11it has been difficult to
• 25:13detect,
• 25:15tumor derived mutant molecules, mutant
• 25:17DNA molecules.
• 25:19But certainly over the past
• 25:21years with advances
• 25:24in technology, advances in bioinformatic
• 25:27approaches,
• 25:28we have been able to
• 25:29reliably
• 25:30detect
• 25:31at least mutations in the
• 25:33circulation of patients with cancer.
• 25:35And I just wanted to
• 25:36go over
• 25:37two main approaches
• 25:39that we're
• 25:41employing and you can appreciate
• 25:42here on the left hand
• 25:43side, this is a tumor
• 25:45informed
• 25:45NGS, a
• 25:47PCR enriched,
• 25:49approach where you can appreciate
• 25:51that, there is
• 25:53personalized profiling
• 25:54by either targeted NGS,
• 25:57or whole exome sequencing or
• 25:58whole genome sequencing of the
• 25:59tumor,
• 26:00followed by
• 26:02specific
• 26:03development of personalized
• 26:05assay per patient.
• 26:06And this is contrasted to
• 26:08what we're showing here on
• 26:09the right, which is a
• 26:10tumor agnostic approach.
• 26:13For the most part, these
• 26:14approaches
• 26:15rely on hybrid capture, next
• 26:17generation sequencing,
• 26:18that is done either alone
• 26:20or with matched
• 26:21white blood cell sequencing,
• 26:23or, together with methylation.
• 26:25These have differential limits of
• 26:27detection.
• 26:28But I think we're very
• 26:29excited to see the development
• 26:31of tumor informed approaches
• 26:33that have actually,
• 26:36have pushed the limit of
• 26:38detection currently
• 26:39to
• 26:41ten or hopefully,
• 26:43less than ten parts per
• 26:45million, which I think is
• 26:46great,
• 26:47in terms of analytical performance.
• 26:50So technical issues are not
• 26:52the only
• 26:56biological limitations and biological challenges
• 26:58as well. And I just
• 26:59wanted to highlight here that
• 27:01the majority of cell free
• 27:02DNA that is released in
• 27:04the circulation is actually non
• 27:05tumor derived.
• 27:06And so any blood based
• 27:08NGS approach has the potential
• 27:10to be contaminated and confounded
• 27:12by mutations
• 27:13that arise from clonally expanded
• 27:15hematopoietic cells or CH or,
• 27:19as we call those.
• 27:21And
• 27:22to show you the importance
• 27:23of this,
• 27:26we here are plotting
• 27:28mutations by lineage,
• 27:31in
• 27:32P53 and then SMARCA4
• 27:34at the bottom. And you'll
• 27:36appreciate here that if one
• 27:38employs a plasma only next
• 27:40generation sequencing approach, a number
• 27:42of mutations that are detected
• 27:44in fall in these genes
• 27:45that are cancer drivers, right,
• 27:48can,
• 27:49be derived from white blood
• 27:51cells,
• 27:52and be clonal hematopoiesis.
• 27:53In origin, these are the
• 27:54ones that are shown in
• 27:56the, orange, in the LALAPA
• 27:58plot as as as orange.
• 27:59And then to show you
• 28:00the clinical importance of this,
• 28:02if all the variants,
• 28:04are considered, and this we're
• 28:05showing this on the far
• 28:06right,
• 28:07again, the tumor derived and
• 28:09then the clonal lymphopoiesis derived,
• 28:11then practically any prognostication,
• 28:14of outcomes, and this is
• 28:16for patients with non small
• 28:17cell lung cancer in the
• 28:18metastatic setting, is obliterated, right?
• 28:20So you really need to
• 28:22understand the lineage and distinguish,
• 28:25tumor derived mutations from clonal
• 28:27lymphopoiesis mutations to understand,
• 28:30the
• 28:32patterns,
• 28:33if you will, and associations
• 28:35with outcomes.
• 28:38Now if one wanted to
• 28:41do this, there is few
• 28:42ways, and,
• 28:44I'll show you in a
• 28:45little bit what it is
• 28:46we've used in the context
• 28:47of a clinical trial.
• 28:49One could perform matched white
• 28:51blood cell sequencing and then
• 28:52practically use that sequence data
• 28:54to filter out mutations derived
• 28:56from clonal hematopoiesis.
• 28:57But clinically, we don't do
• 28:59that, right? So we do
• 29:01plasma only sequencing. So would
• 29:02we still be able
• 29:04to tell
• 29:05the difference origin,
• 29:07cellular origin wise with respect
• 29:10to clonal hematopoiesis versus tumor
• 29:11derived mutations? This is where
• 29:14machine learning approaches come in,
• 29:15and this is one of
• 29:16the efforts that
• 29:18is coming out from my
• 29:20lab, where you'll appreciate here,
• 29:22we're integrating
• 29:23patient level, gene level, mutation
• 29:26level, and fragment
• 29:28level features,
• 29:30by machine learning to generate
• 29:32a model that predicts cellular
• 29:34origin
• 29:34from of mutations detected by
• 29:37plasma only hybrid capture next
• 29:39generation sequencing. You can appreciate
• 29:41the AUC here,
• 29:43which which which is, reasonable.
• 29:46And this was, this held,
• 29:50at this VAUC held at
• 29:51this level in an independent
• 29:54validation setting. And I wanted
• 29:55to show you the clinical
• 29:56utility of this, right, which
• 29:58we're showing here across lineages.
• 30:00So we looked at
• 30:01external datasets of patients with
• 30:03breast cancer, non small cell
• 30:05lung cancer, and prostate cancer.
• 30:07And, you can appreciate the
• 30:08accuracy with respect to cellular
• 30:10origin
• 30:11of the variant with the
• 30:13model. This is the orange
• 30:15bars and without the model.
• 30:17And,
• 30:18we truly,
• 30:19you know, we are really
• 30:21not doing well,
• 30:22in plasma only next generation
• 30:24sequencing even
• 30:26after exclusion of what I
• 30:28call the clonal hematopoiesis
• 30:30blacklists, right? So these
• 30:31genes that are prototypically and
• 30:33canonically associated with clonal hematopoiesis.
• 30:36So in addition to matched
• 30:38white blood cell sequencing, there's
• 30:39bioinformatic approaches that can be
• 30:41helpful here, and that's something
• 30:42that we're further developing. Now,
• 30:44having considered,
• 30:46all the challenges, or at
• 30:47least some of the challenges,
• 30:48let's talk about the opportunities,
• 30:51of liquid biopsies.
• 30:53And here I'm showing you
• 30:54the potential implementation
• 30:56of liquid biopsies in the
• 30:58early stage setting that we're
• 30:59showing on the,
• 31:01left hand side,
• 31:03compared to the metastatic setting
• 31:05that we're showing on the
• 31:06right hand side. I do
• 31:07want to talk about the
• 31:08early stage setting as this
• 31:09is,
• 31:11lineages and a lot of
• 31:12malignancies. But
• 31:14for lung cancer, that is
• 31:14my, area, my focus area,
• 31:15if you will, as
• 31:24we have far too many
• 31:26therapeutic options and sequences of
• 31:28therapies, but but but we
• 31:30don't really know, you know,
• 31:31how to,
• 31:33how to go about it.
• 31:34So so truly an unmet
• 31:36need,
• 31:37in the early stage
• 31:39paradigm. So let me start
• 31:41by,
• 31:42going over
• 31:44what it is previous studies
• 31:46in the
• 31:47early stage setting. And so
• 31:49I think one of the
• 31:50challenges we talk about, technical
• 31:51challenges,
• 31:53has been the fact that
• 31:55we do know that ctDNA
• 31:57MRD is prognostic,
• 31:59right?
• 32:00But then we don't have
• 32:02that much evidence that it
• 32:03is predictive
• 32:05in,
• 32:05for patients with early stage
• 32:07non small cell lung cancer.
• 32:09And then the added challenge
• 32:10here is the fact that
• 32:11the clinical sensitivity
• 32:13of current ctDNA MRD approaches
• 32:16is not where we want
• 32:17it to be for clinical
• 32:18implementation. Again, what does clinical
• 32:20sensitivity mean? It's the fraction
• 32:22of patients that are ctDNA
• 32:24that that recur that are
• 32:25also ctDNA
• 32:26MRD positive. So what I'm
• 32:28showing you here, this is
• 32:29from, analysis from the TRACERx
• 32:31initiative, from, Chris Abosch and
• 32:33Charlie Swanton,
• 32:34where you can appreciate the
• 32:36clinical sensitivity with a tumor
• 32:38informed assay of about forty
• 32:39nine percent. They presented,
• 32:42full update on this at
• 32:43ESMO this year with a
• 32:45slightly
• 32:46better clinical sensitivity with,
• 32:49a tumor informed whole genome
• 32:50sequencing
• 32:51based assay.
• 32:53But then again, as I
• 32:54said, in lung cancer, we
• 32:56do not know if ctDNA
• 32:58MRD is predictive. So this
• 33:00here are results from the
• 33:01EMPOWER ten study. Well, you'll
• 33:03appreciate that atezolizumab,
• 33:06post surgery and post adjuvant
• 33:07chemotherapy,
• 33:08was actually,
• 33:11effective independent of ctDNA status.
• 33:14It did seem to,
• 33:17lead into a prolonged time
• 33:19to ctDNA
• 33:20conversion. But again, not a
• 33:22clear predictive role.
• 33:24And then what it is
• 33:24we know in terms of
• 33:25neoadjuvantly
• 33:26treated,
• 33:28a non small cell lung
• 33:29cancer, we certainly know that
• 33:30ctDNA,
• 33:31at least here for CheckMate
• 33:33eight sixteen, at the time
• 33:34of surgery,
• 33:37was associated
• 33:38with
• 33:39pathologic
• 33:41response. And then there's follow-up
• 33:42studies,
• 33:44that have shown that if
• 33:45we track ctDNA MRD after
• 33:47surgery, this also
• 33:49is associated with better outcomes.
• 33:51But,
• 33:53we still,
• 33:54there's still so many unknowns.
• 33:56And one of
• 33:58the areas that I think
• 34:00ctDNA MRD can be very,
• 34:02very
• 34:03informative
• 34:04is actually in, or ctDNA
• 34:06status rather, determined at different
• 34:08time points, is dissecting
• 34:11a pathologic
• 34:12response after neoadjuvant
• 34:14immunotherapy or immuno chemotherapy
• 34:16for patients with early stage
• 34:17cancer. So this is a
• 34:19recent study that we finished
• 34:20not too long ago, where,
• 34:22and published not too long
• 34:23ago, which you'll appreciate. I
• 34:25just wanna
• 34:26draw your attention here
• 34:28to the bottom right.
• 34:30These are all patients
• 34:32with resectable
• 34:33esophageal
• 34:34cancer that received immuno chemoradiation
• 34:37in the neoadjuvant
• 34:38setting, and they
• 34:40then went to surgery.
• 34:42And these are patients down
• 34:44here that did not attain
• 34:45a past CR. So the
• 34:47tumor didn't regress fully.
• 34:49But then if within this
• 34:51set of patients,
• 34:53if you actually,
• 34:55classify
• 34:56these
• 34:57based on
• 34:58ctDNA
• 34:59status, what you'll see is
• 35:01very clear cut that,
• 35:03ctDNA
• 35:05dynamics versus
• 35:07undetectable status can truly, within
• 35:08this heterogeneous group of non
• 35:08path CRs can actually identify
• 35:09the patients that do well,
• 35:12right?
• 35:12And these are the ones
• 35:14with undetectable ctDNA
• 35:20compared to the ones that
• 35:22actually would need something,
• 35:25following
• 35:26surgery. So I think this
• 35:28is a true opportunity
• 35:29for ctDNA MRD approaches in
• 35:31terms of,
• 35:33being used in addition to
• 35:35path
• 35:36responses, which now are, you
• 35:38know, what we
• 35:39consider a credible
• 35:41endpoint of immunotherapy response for
• 35:43lung cancer in the neoadjuvant
• 35:44setting.
• 35:45And this is something that
• 35:46actually gave us pause.
• 35:48It was so, so, so,
• 35:52gratifying to see this because
• 35:54the next question we asked
• 35:56is,
• 35:56well great, so
• 35:58we
• 35:59identify
• 36:00these, you know, kind of
• 36:01ctDNA
• 36:02trends,
• 36:04but
• 36:05do
• 36:06they mirror or what's the
• 36:07association
• 36:08with anti tumor immune responses?
• 36:10So practically here, what we're
• 36:12showing you
• 36:13is two patients,
• 36:15where,
• 36:16that have very different
• 36:18ctDNA dynamics during neoadjuvant
• 36:20therapy and in the post
• 36:22op period, well, you can
• 36:23appreciate,
• 36:24circulating tumor fraction regression with
• 36:27the first one and persistence
• 36:28with the second one. And
• 36:30then when we actually pulsed
• 36:32autologous T cells from the
• 36:34same patients,
• 36:36what we saw, at the
• 36:37same time points,
• 36:39what we saw was TCR
• 36:41clonotypic expansions in the first
• 36:43one and nothing going on
• 36:44in the second one. And
• 36:46so this was the first
• 36:47evidence, at least that I
• 36:48had seen, that actually the
• 36:50antitumor immune responses
• 36:52mirror
• 36:53ctDNA
• 36:54dynamics,
• 36:56in, for, for, you know,
• 36:57in the context of, in
• 36:59the context of this trial.
• 37:02So
• 37:03certainly a lot of evidence
• 37:05and, you know, we don't
• 37:06have time to review
• 37:07all of the studies that
• 37:08are out there and there
• 37:09is a number,
• 37:11of them, you know,
• 37:12that data that we've generated
• 37:14and others.
• 37:15But there is a clinical
• 37:16question here and a clinical
• 37:18challenge.
• 37:19And this is the following.
• 37:20How do we navigate the
• 37:22expanding
• 37:23therapy
• 37:24landscape for patients with early
• 37:26stage known small cell lung
• 37:28cancer, right? So we need
• 37:29approaches to stratify patients. We
• 37:30need to personalized,
• 37:32you know,
• 37:33neoadju and perioperative therapy. How
• 37:36do we do that?
• 37:37And can ctDNA
• 37:39approaches be helpful here? So
• 37:41in order to answer these
• 37:42questions, especially for patients with
• 37:45non small cell lung cancer,
• 37:47and answer questions like,
• 37:50what is the clinical sensitivity
• 37:51of tumor informed ctDNA MRD
• 37:54in the context of newer,
• 37:56more sensitive
• 37:58MRD assays?
• 38:00Is
• 38:01tumor informed ctDNA
• 38:03MRD feasible? Is tumor uninformed
• 38:05ctDNA
• 38:06MRD,
• 38:09sensitive
• 38:10enough? When do we actually
• 38:11measure ctDNA MRD? Do we
• 38:14measure once? Do we measure
• 38:15multiple times? So Charu Agarwal,
• 38:17she's at UPenn, and myself,
• 38:19we put together this clinical
• 38:20trial. This is a biomarker
• 38:23observational
• 38:24trial where you'll appreciate that
• 38:26patients with,
• 38:28early stage non small cell
• 38:30lung cancer are enrolled, and
• 38:31then there's a dense collection
• 38:33of both tissue and serial
• 38:35blood samples,
• 38:37with primary endpoints
• 38:39of the trial being
• 38:41to determine the clinical sensitivity
• 38:44of ctDNA
• 38:45MRD at this landmark MRD
• 38:47time point. And we think
• 38:48that this is truly going
• 38:50to be data that can
• 38:52inform
• 38:52ctDNA,
• 38:55adaptive
• 38:56or guided clinical trial design
• 38:58in the early stage setting.
• 39:01The trial is called Tic
• 39:02Tac Toe and this is
• 39:03going
• 39:04to make sense
• 39:05with the following slide because
• 39:07that's the vision, right? So
• 39:09the vision is, at least
• 39:10my vision, is that
• 39:13the therapy landscape
• 39:14of resectable non small cell
• 39:17lung cancer,
• 39:18in terms of surgery first,
• 39:19in terms of neoadjuvant first,
• 39:21in terms of neoadjuvant perioperative
• 39:24is converted to a tic
• 39:25tac toe board,
• 39:26where practically ctDNA
• 39:29status is the is the
• 39:30opening move. And then every,
• 39:33additional, every player
• 39:35move,
• 39:37represents additional strategic decisions that
• 39:40are made based on on
• 39:41informed based on ctDNA.
• 39:43So what you have at
• 39:44the at the end is
• 39:45alignment of the markers, and
• 39:47this is a winning strategy
• 39:48for our patients. So we
• 39:49hope to to achieve that,
• 39:52with with the tic tac
• 39:53toe,
• 39:54trial. Now kind of moving
• 39:56towards the the last part
• 39:59of my talk, there's certainly
• 40:00a lot that, a lot
• 40:02more, if you will, that
• 40:03we know in terms of
• 40:04clinical value
• 40:05of ctDNA molecular response assessments
• 40:07in the metastatic setting.
• 40:09This is something that we've
• 40:10been studying over the past
• 40:11ten years,
• 40:13or so. I've been studying
• 40:14this since I was a
• 40:15medical oncology fellow.
• 40:17I think,
• 40:18again, rhetoric question, why do
• 40:20we need
• 40:21liquid biopsies in terms of
• 40:22informing molecular response assessments
• 40:25and clinical response for immunotherapy?
• 40:27We talked about this at
• 40:27the beginning. We don't have
• 40:29great biomarkers, right,
• 40:31that,
• 40:34help us understand
• 40:36immunotherapy response and resistance. And
• 40:38of course, imaging
• 40:39in the context of immunotherapy
• 40:41is problematic.
• 40:42Right? So it doesn't capture
• 40:44the unique timing or nature
• 40:46of the anti tumor immune
• 40:47response. So we truly need
• 40:48molecularly informed approaches to do
• 40:50this. So,
• 40:52we've been, again, as I
• 40:53said, over the past several
• 40:54years, generating data
• 40:56on this. And we've employed,
• 41:00ultra sensitive
• 41:02hybrid capture next generation sequencing
• 41:04approaches
• 41:07to identify and monitor mutations
• 41:10for the most part in
• 41:11the circulation of patients with
• 41:12metastatic non small cell lung
• 41:14cancer receiving immunotherapy.
• 41:16And I think collectively as
• 41:17a field, we converge on
• 41:19the patterns that are shown
• 41:21here. Well you'll appreciate a
• 41:22pattern of ctDNA
• 41:24response where there is elimination
• 41:26of a tumor,
• 41:27derived mutation. This is a
• 41:29p53
• 41:29splice site here that we
• 41:31see eliminated. And this is
• 41:32a patient that has attained
• 41:34long progression free and overall
• 41:36survival
• 41:37with immunotherapy,
• 41:38immune checkpoint blockade in particular,
• 41:40and this is to be
• 41:41contrasted with a patient,
• 41:43with ctDNA
• 41:44progression where you'll appreciate that
• 41:47mutation levels do not change,
• 41:49rather rise over time. And
• 41:51this is a pattern of
• 41:52molecular
• 41:54disease progression.
• 41:56Now,
• 41:57we don't have the answers
• 41:58in the metastatic, we don't
• 41:59at least have all the
• 42:00answers in the metastatic
• 42:02setting either. So how do
• 42:03we compute tumor fraction? Historically,
• 42:05we've been going after mutations.
• 42:07What I'm showing you here
• 42:08at the bottom is,
• 42:10our ability to,
• 42:12derive plasma aneuploidy
• 42:13from off and on target,
• 42:16sequencing reads from hybrid capture
• 42:18NGS. And what you'll appreciate
• 42:19here is practically convergence
• 42:22of,
• 42:23of plasma anploidy to euploid
• 42:26state with therapy, and this
• 42:27was
• 42:29a patient that attained initial
• 42:31response with immunotherapy. So there
• 42:33is value to actually considering
• 42:34plasma aneuploidy approaches
• 42:36together with mutation trends in
• 42:39a tumor agnostic fashion or
• 42:40a tumor informed fashion that
• 42:42we're showing on the left
• 42:43and right respectively.
• 42:45And does it make a
• 42:46difference in terms of how
• 42:48we determine molecular response? And
• 42:50I'm going to
• 42:51go back to our initial
• 42:53discussion about determination of cellular
• 42:55origin. So it truly makes
• 42:57a difference
• 42:58based on what types of
• 42:59mutations are considered.
• 43:01And so here, what you'll
• 43:03appreciate, we're showing a tumor
• 43:05agnostic white blood cell informed
• 43:07approach
• 43:08contrasted to a plasma only
• 43:10approach. Well, plasma only approach
• 43:11is
• 43:12every plasma,
• 43:14has detectable ctDNA,
• 43:16but
• 43:17but not all mutations are
• 43:18tumor derived. And that introduces
• 43:19a problem when we're actually
• 43:21asked to to make a
• 43:22call in terms of molecular
• 43:24response because there's persistence of
• 43:25mutations, yet they're not tumor
• 43:27derived.
• 43:27And then if one focuses
• 43:29on a tumor informed approach,
• 43:31what you'll appreciate that we're
• 43:32showing here in the middle,
• 43:33what you'll appreciate is that
• 43:34we're left with too few
• 43:35mutations. It's very stringent, right?
• 43:38And then if you look
• 43:39on the right hand side,
• 43:41what is the added value
• 43:42of the plasma aneuploidy approach?
• 43:44For the most part, mutation
• 43:46trends really track well with
• 43:48plasma aneuploidy,
• 43:49derived, tumor fraction estimation with
• 43:50the exception of tumors with,
• 43:58aneuploidy, which
• 44:00we may capture
• 44:01the minority, but it's worth
• 44:03considering.
• 44:05So
• 44:07how do we, having done,
• 44:09you know, having
• 44:11cleaned up perhaps,
• 44:13a cellular origin of variants,
• 44:15how do we define molecular
• 44:17response? And as a field,
• 44:19we've been focusing on landmark
• 44:22molecular response assessments. We've been
• 44:24looking at complete clearance of
• 44:26ctDNA. We've been looking at
• 44:28reduction past ninety percent,
• 44:30seventy five percent, fifty percent.
• 44:32So what do we use?
• 44:34Here you'll appreciate, we've done
• 44:36a very careful consideration of
• 44:38sensitivity and specificity of different
• 44:40definitions of ctDNA molecular responses.
• 44:43This is again immunotherapy treated
• 44:45non small cell lung cancer.
• 44:46And what you'll appreciate is
• 44:48that landmark
• 44:50ctDNA,
• 44:52molecular response assessment, this is
• 44:54assessment three to nine weeks
• 44:56on therapy,
• 44:57actually generates,
• 44:59good sensitivity and specificity
• 45:01and allows us
• 45:03to detect
• 45:04a higher number to call
• 45:06molecular responses practically in a
• 45:08higher number of individuals.
• 45:10And this is translated,
• 45:13into
• 45:14clinical benefit, as you can
• 45:15appreciate here. I'm just going
• 45:16to throw, draw your attention
• 45:17in the middle, the survival
• 45:19course where patients with a
• 45:20molecular response do well with
• 45:22respect to progression free and
• 45:23overall survival. And then if
• 45:25one does,
• 45:27if one adjusts for clinical
• 45:28covariates as well as lines
• 45:30of therapies
• 45:31or whatnot,
• 45:32certainly molecular response is an
• 45:33independent,
• 45:35predictor
• 45:36of progression free and overall
• 45:38survival.
• 45:39Now, are we done with
• 45:41developing methodologies?
• 45:42And I have so much
• 45:43to tell you. This is
• 45:44such an exciting field, but
• 45:45I'm going to try to
• 45:47speed up
• 45:48a little bit.
• 45:50Here's where we are actually
• 45:52developing new methodologies
• 45:54where instead of mutations, right,
• 45:56historically we've been tracking mutations.
• 45:57But there's so much wealth
• 45:59of features and data in
• 46:00cell free DNA.
• 46:02So here we're integrating fragmentation
• 46:04patterns across the genome together
• 46:07with plasma aneuploidy
• 46:08and integrating those features by
• 46:10machine learning to practically make
• 46:12a call in terms of
• 46:13cancer versus not cancer.
• 46:15And then once this call
• 46:16is made, then we can
• 46:17determine, you know, molecular response,
• 46:20at different time points. And
• 46:22the beauty of this, you'll
• 46:23appreciate here, if you remember
• 46:24the previous slide that I
• 46:25showed you, this is an
• 46:27even more,
• 46:29highly sensitive and specific
• 46:32approach. And the beauty of
• 46:33this is also that our
• 46:34input material,
• 46:36is one ml of plasma,
• 46:38one nanogram of cell free
• 46:40DNA.
• 46:41So,
• 46:42a truly exciting field, you
• 46:43know, to continue,
• 46:45to continue to explore here.
• 46:49But where are we with
• 46:51respect to clinical implementation of
• 46:53all this, right? So if
• 46:54we wanted to clinically implement
• 46:56and guide therapeutic decision making,
• 46:58are we there yet?
• 47:00We weren't a few years
• 47:01back, and so that's precisely
• 47:03the reason that we put
• 47:04together a clinical trial. This
• 47:06was the first,
• 47:07clinical trial
• 47:09in patients with metastatic non
• 47:11small cell lung cancer that
• 47:12were candidates for first line
• 47:13immunotherapy
• 47:15that was put together to
• 47:16answer three specific questions.
• 47:19What is molecular, ctDNA molecular
• 47:21response? When does it happen?
• 47:22And importantly,
• 47:24what is
• 47:25the concordance
• 47:26with radiographic
• 47:27response, the sensitivity question, right?
• 47:29So we felt that we
• 47:30needed to answer these questions
• 47:32before
• 47:33designing a ctDNA adaptive clinical
• 47:36trial. So this was, BR
• 47:38thirty six, the BR thirty
• 47:39six clinical trial. The first
• 47:40stage I'm going to talk
• 47:41to you about, was published
• 47:43not too long ago last
• 47:45year. And now we're,
• 47:47we are enrolling patients for
• 47:49stage II. So in this
• 47:50trial, very quickly,
• 47:52we employed
• 47:53a
• 47:54tumor agnostic white blood cell,
• 47:57informed
• 47:57approach. This was hybrid capture
• 47:59NGS, and you'll appreciate, you
• 48:01know, different,
• 48:02the patients and time points,
• 48:05here and, you know, the
• 48:06landscape of mutations
• 48:08detected.
• 48:09And these are the patterns
• 48:11that,
• 48:12we
• 48:13identified.
• 48:14So clearance of ctDNA as
• 48:16cycle two of pembrolizumab.
• 48:18There were few patients that
• 48:19cleared a little bit later.
• 48:21A very different pattern, you
• 48:23know,
• 48:24with that of molecular disease
• 48:25progression where there's persistence of
• 48:27ctDNA
• 48:28throughout cycle two and cycle
• 48:30three. So molecular response in
• 48:32this context was determined as
• 48:34ctDNA
• 48:35clearance
• 48:36three weeks, on the third
• 48:38cycle, this is six weeks,
• 48:39on pembrolizumab for patients with
• 48:41metastatic non small cell lung
• 48:42cancer.
• 48:43And this is a lot
• 48:45of work,
• 48:46put together in a single
• 48:47slide, but these are the
• 48:48main findings of BR36,
• 48:50the first stage. So we
• 48:51met with the primary endpoint.
• 48:52The sensitivity of molecular response
• 48:54for RECIST best overall response
• 48:56was eighty two percent with
• 48:57a specificity of seventy five.
• 49:00And what is obviously important
• 49:02is that molecular
• 49:04response
• 49:05better
• 49:06separated
• 49:07patients
• 49:08with longer progression free and
• 49:10overall survival.
• 49:12And, you know, if you
• 49:13look at the swimmers plot
• 49:15on
• 49:16the left hand side, it's
• 49:17actually quite interesting because there's
• 49:18patients
• 49:19here in terms of the
• 49:20colors, molecular responders are red,
• 49:22molecular,
• 49:25molecular progressors are red, molecular
• 49:27responders are blue. So there's
• 49:28patients with radiographic partial response
• 49:31that actually did not attain
• 49:33molecular response. These didn't do
• 49:35well, in terms of they
• 49:36had a very short lived
• 49:37PFS.
• 49:38High heterogeneity
• 49:40in the stable disease category.
• 49:41And then even within patients
• 49:43with progressive disease, we had
• 49:44few cases with pseudo progression,
• 49:46where, imaging
• 49:48actually didn't get it right
• 49:49at all.
• 49:51Here's now stage two. So
• 49:53this is,
• 49:55learning
• 49:55from stage one of BR36.
• 49:58We move to a phase
• 50:00two, three
• 50:01ctDNA
• 50:02interventional
• 50:03trial design. This, we're very
• 50:04excited. This trial is now
• 50:06open and enrolling in the
• 50:07United States and Canada.
• 50:09Where practically, you'll appreciate based
• 50:11on the,
• 50:12trial design that,
• 50:13patients
• 50:15that are eligible for first
• 50:16line pembrolizumab
• 50:18and tumors have to be
• 50:19PD L1 fifty percent and
• 50:21above,
• 50:21that six weeks into pembrolizumab
• 50:24therapy are ctDNA positive and
• 50:26resist non PD or have
• 50:28progressive disease for which, though,
• 50:29the physicians would continue to
• 50:31administer
• 50:32pembrolizumab for clinical benefit, these
• 50:34are the ones that are
• 50:35randomized
• 50:36to either continue on pembrolizumab
• 50:38or
• 50:39do a step up step
• 50:41up approach,
• 50:45has,
• 50:49chemotherapy and
• 50:53I'm not sure if I
• 50:54can mute people from here,
• 50:56but I can I can
• 50:57try?
• 50:58So so so practically,
• 51:00this this trial design is
• 51:02going to truly,
• 51:03enable enable us
• 51:06understand the value of ctDNA
• 51:08molecular responses and early endpoint
• 51:10of immunotherapy response, and we
• 51:12hope we get regulatory,
• 51:13interest, after completion of this
• 51:16trial.
• 51:16So,
• 51:17getting towards the end, I
• 51:19wanted to close the ctDNA
• 51:21section by,
• 51:23bringing up this slide on
• 51:25how do we practically bridge
• 51:27scientific discoveries with clinical cancer
• 51:29care through ctDNA adaptive trials.
• 51:31We talked about the low
• 51:32hanging fruit, you know, the
• 51:33quick fail approaches that I'm
• 51:35showing you here on the
• 51:36bottom
• 51:37right. But you can imagine
• 51:38that we could,
• 51:40based on
• 51:41sustained ctDNA
• 51:42elimination, we could deescalate
• 51:44therapy.
• 51:45Or ctDNA
• 51:46dynamics and or can allow
• 51:48us to pivot, right? So
• 51:49what we're showing
• 51:51in the top right and
• 51:52bottom left allow us to
• 51:54early to to to switch
• 51:56therapies, and there is a
• 51:57role here for ctDNA molecular
• 51:59response in terms of,
• 52:01drug approvals, right,
• 52:03and and readout of of,
• 52:06experimental therapeutics.
• 52:08So again, I think this
• 52:10was a long journey, you
• 52:11know, starting with tissue multiomics
• 52:13and again, highlight the importance,
• 52:16of going past tumor mutation
• 52:18burden and truly understand the
• 52:19underlying biology,
• 52:21focus on
• 52:23subsets of mutations within the
• 52:24overall tumor mutation burden that
• 52:26are biologically
• 52:28distinct and potentially carry differential
• 52:29immunogenicity,
• 52:31weights,
• 52:33all the way down to
• 52:35studying tumor evolution via liquid
• 52:37biopsy,
• 52:38approaches and making therapeutic decisions
• 52:41based on on on trends
• 52:42on liquid biopsies.
• 52:44And again, what I wanted
• 52:45to to finish with is
• 52:46is truly
• 52:47a call for,
• 52:49more clinical trials that actually
• 52:51test these concepts, right, to
• 52:53truly understand
• 52:54the clinical utility.
• 52:56So let me end by
• 52:58acknowledging the great work of
• 52:59folks in my lab. These
• 53:01are the folks that we're
• 53:02showing here on the left
• 53:03hand side and of course
• 53:04in thoracic oncology, Julie Brahmer
• 53:06and others.
• 53:07And, our collaborators, the BR36
• 53:10investigators
• 53:11and our funding sources.
• 53:12And then my last slide
• 53:14is I went down to
• 53:15memory lane.
• 53:17So I wanted to thank
• 53:18David,
• 53:19for the invitation
• 53:21to return to Yale, give
• 53:22this talk, meet with everybody.
• 53:24It was again,
• 53:25a great pleasure. You can
• 53:27see this is,
• 53:30this photo here is a
• 53:31lab group photo. Right? So
• 53:33I think it's
• 53:34back,
• 53:36in, I wanna say, twenty
• 53:37ten or around twenty ten
• 53:39or so. I'm here in
• 53:40the back.
• 53:41And this is, David and
• 53:43I, reconnecting together with my
• 53:45Hopkins mentor, Victor Vapulescu, at
• 53:47AACR,
• 53:48this year.
• 53:50And again, thank you for
• 53:51these wonderful years at,
• 53:54at Yale, and I will
• 53:55always, you know, go back
• 53:56and remember the wonderful conversations,
• 53:58you know, that we had
• 53:59on, on science. Great pleasure
• 54:00again to be here.
• 54:01Thank you. And happy to
• 54:03take questions.
• 54:12We have about five minutes
• 54:13for questions.
• 54:15Yes. Kurt. There are two
• 54:17two questions. So the first
• 54:18one is very technical. How
• 54:19do you manage to make
• 54:21a specific DNA library with
• 54:22one nano brand or or
• 54:24cell free DNA?
• 54:29So so, you know, I
• 54:30mean, you may have obviously
• 54:31less than one genomic equivalent.
• 54:33Right? But then in terms
• 54:34of the technical approach, you
• 54:36know,
• 54:37it's a low,
• 54:39it's it's it's a, you
• 54:40know, pretty standard genomic library,
• 54:43prep,
• 54:44and a very,
• 54:46so in terms of the
• 54:47PCR amplification four cycles, so
• 54:49you're going for low
• 54:50level
• 54:51coverage but it's across the
• 54:53genome, right, and because then
• 54:55the secret isn't the bioinformatics.
• 54:57Right? Then we segment the
• 54:58genome in bins and and
• 54:59we look at fragmentation patterns,
• 55:00I think five hundred thirty
• 55:02or so bins. And so
• 55:03the more features then, you
• 55:05know, if the feature space
• 55:06is expanded, then this is
• 55:08what actually gives you the
• 55:09power
• 55:10to to to detect tumor
• 55:11fraction. Exactly.
• 55:15Exactly.
• 55:15And another question I noted
• 55:17for the fragmentomics
• 55:18pattern, you you use cell
• 55:20free DNA, not ctDNA.
• 55:29Yeah. No. Thank you.
• 55:31I think it is for
• 55:32different purposes. Right? So as
• 55:34as, you know, we all
• 55:35discussed the majority of cell
• 55:36free DNA is now coming
• 55:37from
• 55:38tumor cells, right?
• 55:40So you can train machine
• 55:41learning models
• 55:43to practically tell the difference
• 55:44between cancer and non cancer,
• 55:46right?
• 55:47But then there is additional
• 55:48features in cell free DNA
• 55:50that I think can be
• 55:51very informative
• 55:53in determining lineage
• 55:54of non cancer cells. And
• 55:56this is something that, you
• 55:57know, we're very excited to
• 55:59study. I think I was
• 56:00discussing with Sam earlier on,
• 56:03where you could potentially,
• 56:04based on,
• 56:06transcription factor binding sites and
• 56:08coverage at these sites, you
• 56:09could potentially determine lineage
• 56:12of non cancer cells, imagine
• 56:14immune cells. Right? So that's
• 56:15something that we're working on.
• 56:19Please. Yeah. You mentioned that
• 56:21in that case, like, the
• 56:22entry didn't get the disease
• 56:24depression, but CDMA doesn't come
• 56:26into depression. In this cancer,
• 56:27we have to mention a
• 56:29valuable CTBMA
• 56:30modification.
• 56:47No.
• 56:49The the this is this
• 56:50is a great question. And,
• 56:51you know, we come across
• 56:52this comes across to us
• 56:53as well in the molecular
• 56:54tumor board, right, where, you
• 56:56know, there is mutations or
• 56:57or there is tumor fraction
• 56:58depending on what assay, what
• 56:59commercial vendor, right, you're you're
• 57:02you're working with.
• 57:03The first comment I wanna
• 57:04make here is that not
• 57:05all mutations are tumor derived,
• 57:06right, so this is a
• 57:07mutation based approach. It's genotyping
• 57:09assay, and then you see
• 57:11an ATM mutation, chances are
• 57:14that that it's not coming
• 57:15from,
• 57:16the tumor, right, so this
• 57:17could be derived from clonal
• 57:19hematopoiesis,
• 57:20or clonally expanded hematopoietic
• 57:22cells. So that's one point
• 57:23that I want to make.
• 57:24The second though is let's
• 57:26assume it's all tumor derived,
• 57:27right? And, you know, there
• 57:28is evidence that there is
• 57:29a little bit, you know,
• 57:30like the tumor fraction is
• 57:32right at the threshold, a
• 57:33little bit above the threshold.
• 57:34What do you do with
• 57:35it? Well, that's where we
• 57:36need to do the clinical
• 57:37trials, I think,
• 57:40to show the clinical utility.
• 57:42Because one of the things
• 57:43that, you know, has been
• 57:44challenging
• 57:45is that
• 57:46precisely what it is you're
• 57:48saying, and we see that
• 57:49in the context of,
• 57:52oncogen driven non small cell
• 57:53lung cancer where we may
• 57:54detect, you know, an EGFR
• 57:56mutation after initiation of,
• 57:58EGFR targeted therapy,
• 58:01but should we intervene
• 58:02if imaging is not,
• 58:04doesn't show over disease progression
• 58:06or should we wait? Is
• 58:07there a survival benefit if
• 58:08we actually intervene earlier?
• 58:11I don't think there is
• 58:12necessarily convincing evidence in the
• 58:14oncogen driven space, but but
• 58:16this is precisely the question
• 58:17that we're asking in the
• 58:18immunotherapy
• 58:19trial that I showed you.
• 58:21But this is also, I
• 58:22think, augmented by the fact
• 58:23that in the immunotherapy context,
• 58:25imaging doesn't get it right.
• 58:29Hey. I get to ask
• 58:30a question. Yes.
• 58:31So one of the things
• 58:32that's interesting about ctDNA is
• 58:34some people are using enrichment
• 58:36strategies by immunocrecipitating
• 58:37histones. Mhmm. Can you comment
• 58:39on that?
• 58:40Yeah.
• 58:42So
• 58:44nothing of what I didn't
• 58:45show you any methylation,
• 58:47you know, based approaches.
• 58:51So there are enrichment strategies.
• 58:53I think I haven't seen,
• 58:55David, convincing
• 58:57data in terms of the
• 58:58clinical sensitivity of
• 59:00these approaches.
• 59:02Right. So
• 59:04when I think about,
• 59:05you know, methylation based approaches,
• 59:07it really helps,
• 59:09in determining lineage
• 59:10more so. I haven't seen
• 59:12convincing evidence. Of course, if
• 59:13you look at cancer early
• 59:14detection, you know, I think,
• 59:17the the sensitivity of these
• 59:18approaches is actually very low.
• 59:20We don't need to. We
• 59:21can do without that. So
• 59:22so we think so. So
• 59:24I think the
• 59:25and that's my personal opinion,
• 59:26right? But the more features
• 59:28you integrate for now whole
• 59:30genome sequence data,
• 59:32the lower that limit of
• 59:33detection is going to be.
• 59:34Hence, you know, all the
• 59:35efforts, right, looking at, you
• 59:37know, we started with mutations,
• 59:39oligo mutation, then we went
• 59:40to methylation, about a thousand
• 59:42sites or whatnot. Now we're
• 59:44going potentially, we're expanding to
• 59:46several thousands
• 59:48of of features, and I
• 59:49think this is what is
• 59:50is really going to push
• 59:51down the limit of detection.
• 59:53K. Any other questions?
• 59:56K. Thank you, Austin.
• 59:58Thank you.