Earlier detection of pancreatic cancer

On-demand webinar

Summary:

Early detection of hard-to-treat cancers, such as pancreatic cancer, is critical for improving overall survival. In this session, we will be discussing areas of greatest need, such as the most promising biomarkers, challenges for clinical adoption, and much more.

Earlier Detection of Pancreatic Cancer

Moderator and Speakers:

Giulia Biffi (Cancer Research UK Cambridge Institute, UK)

Laura DeLong Wood (Johns Hopkins University School of Medicine, USA)

Tobiloba Oni (Whitehead Institute, USA)

Video Transcript

00:00 - 00:25: My name is Julia Bifi, I'm a group leader here at the Cancer Research UK Cambridge Institute
00:25 - 00:29: as part of the University of Cambridge, and I also co-lead the pancreatic cancer program
00:29 - 00:36: at the Cancer Research UK Cambridge Institute. I'm very pleased to chair today's virtual
00:36 - 00:42: seminar on earlier detection in pancreatic cancer, co-organized with Abcam. You will
00:42 - 00:48: have the opportunity to ask questions to the speakers. Please do so by using the Q&A chat
00:48 - 00:56: function at the bottom of the screen. Today we are going to hear from clinicians and researchers
00:56 - 01:02: in the field, and now we're going to hear about this from our speakers. We are going
01:02 - 01:08: to move on to our first speaker, Dr. Laura DeLong-Wood. Dr. Wood is an associate professor
01:08 - 01:14: and director of the Division of Gastrointestinal and Liver Pathology in the Department of Pathology
01:14 - 01:20: at Johns Hopkins. Her laboratory leverages next-generation sequencing to characterize
01:20 - 01:25: genetic heterogeneity and clonal evolution in precancerous pancreatic lesions and employs
01:25 - 01:31: three-dimensional organoid culture to interrogate the molecular drivers of pancreatic cancer
01:31 - 01:36: progression. Today she's going to talk about precancerous pancreatic neoplasms, the SORPIMS
01:36 - 01:41: Act. Thanks, Laura, for making time for this, and I'm looking forward to your talk.
01:46 - 01:52: Okay, I always make the mistake and don't unmute myself before I share, so I managed to avoid that.
01:53 - 01:57: Thank you to Julia and to Abcam for the invitation to speak today, and thank you
01:57 - 02:03: for the inspiring start to this meeting. It's a hard act to follow, but it's a good reminder of
02:03 - 02:06: why we're doing what we're doing here, and thank you to the patients and patient advocates who
02:06 - 02:11: shared their stories. So, today I'm going to talk about precancerous pancreatic neoplasms,
02:11 - 02:17: the SORPIMS Act. So, I want to start out by talking about our first look at the whole
02:18 - 02:25: exome of pancreatic cancer. This happened back in 2008. We got the chance to interrogate the
02:25 - 02:29: entire exome, the entire coding region of the genome in pancreatic cancer samples. This was
02:29 - 02:34: done by Burt Vogelstein's group using Sanger sequencing, which is a very challenging way to
02:34 - 02:39: look at the exome. We have much better technologies now, but what that study showed was that the
02:39 - 02:44: pancreatic cancer genome landscape is really dominated by four mountains, four genes that
02:44 - 02:50: are somatically altered in the vast majority of pancreatic cancers, and that is the oncogene K-RAS
02:50 - 02:57: and then three tumor suppressor genes, which are CDKN2A, also known as P16, TP53, and SMAD4.
02:57 - 03:01: Also important to point out in this landscape is there are a lot of hills. There are less
03:01 - 03:05: frequently altered genes that are still significant and still likely contribute to
03:05 - 03:12: tumorigenesis. But that was almost 14 years ago now, so there's been a lot that has happened in
03:12 - 03:17: those 14 years. Several groups have now reported whole exome or whole genome sequencing of
03:17 - 03:21: pancreatic cancers. For this talk, we're focusing pancreatic cancer to mean pancreatic ductal
03:21 - 03:28: adenocarcinoma, or PDAC. These advances were largely possible through the invention and
03:28 - 03:33: expansion of next-generation sequencing, which is far more scalable than traditional approaches,
03:33 - 03:37: and also bioinformatic approaches that allow us to deal with all this data. But now we have
03:37 - 03:42: hundreds and hundreds of exomes and genomes of pancreatic cancers. This includes efforts
03:42 - 03:46: by the International Cancer Genome Consortium, as well as the Cancer Genome Atlas and several
03:46 - 03:52: institutional efforts. And in these hundreds and hundreds of exomes and genomes, the mountains are
03:52 - 03:57: still the mountains. The four most frequently altered genes are still K-RAS, TP53, P16, and
03:57 - 04:03: SMAD4. But we've really clarified the hills in the genome landscape, and these have identified
04:03 - 04:09: several groups that are responsive to specific therapies. However, since the focus of this talk
04:09 - 04:16: is not cancer, but rather earlier detection, I want to ask the question, since these studies
04:16 - 04:20: have given us a static picture of advanced pancreatic cancer, what's the best strategy
04:20 - 04:24: to understand the initiation and progression of pancreatic tumorigenesis, those early stages
04:24 - 04:29: that are important to understand to really enable early detection? There are really two
04:30 - 04:34: broad conceptual options. You can infer this computationally through sequencing advanced
04:34 - 04:40: cancers, or you can directly analyze these precancerous lesions. And my group favors
04:40 - 04:46: the latter approach, where we directly analyze precancerous lesions in human tissue samples.
04:47 - 04:52: Especially as the first speaker, I want to give a brief introduction of what these precancerous
04:52 - 04:57: lesions are. So, the majority of pancreatic cancers, between 80 and 90 percent, arise
04:57 - 05:03: through pancreatic intraepithelial neoplasia, also called PANIN. These are small lesions,
05:03 - 05:08: microscopic, by definition less than five millimeters, and they involve the pancreatic
05:08 - 05:15: ductal system. This small size means they're not visible on most current radiographic approaches.
05:15 - 05:20: They're also not visible in a grossly resected pancreas specimen to harvest for research.
05:21 - 05:26: PANINs are graded based on the morphologic atypia of their lining epithelium.
05:26 - 05:32: Low-grade PANINs are by far much more common, and these have mild atypia. High-grade PANINs
05:32 - 05:37: have more severe atypia. They're thought to be a step before the transition to invasive cancer,
05:37 - 05:40: and these are much less common than the low-grade lesions.
05:41 - 05:47: A smaller but important minority of pancreatic cancers arise through large cystic lesions.
05:47 - 05:52: The most common of this is intraductal papillary mucinous neoplasm, or IPMN.
05:52 - 05:56: These are by definition greater than a centimeter, but often get much larger,
05:56 - 06:01: and this means that they're visible on radiographic imaging, and they often get
06:01 - 06:07: diagnosed incidentally. This opens up a whole clinical quandary of how to do surveillance in
06:07 - 06:12: these patients and when to intervene, because like PANIN, the majority of IPMNs are low-grade.
06:12 - 06:19: They're unlikely to progress, and so there's important decisions to be made about how often
06:19 - 06:25: to screen these patients and when to intervene clinically with surgery. However, from a research
06:25 - 06:29: perspective, the fact that these are often diagnosed prior to transformation or progression
06:29 - 06:35: to invasive cancer means it gives us an important opportunity to study precancerous pancreatic
06:35 - 06:41: neoplasms in human tissue samples. For this talk, I'm going to focus mostly on our work on IPMNs.
06:41 - 06:44: We are starting to apply these approaches in PANINs. I don't have slides on that in this talk,
06:45 - 06:50: but if folks are interested in the question and answer, I'm happy to discuss how the concepts
06:50 - 06:57: apply to PANINs. With that, I want to explain the analogy to the serpent's egg. A couple of years
06:57 - 07:02: ago now, I was giving a talk with my friend and colleague, Jorge Polino, in Portugal on
07:02 - 07:07: precancerous lesions, and he suggested this reference as a title for the talk. I had to admit
07:08 - 07:13: that I didn't get it, and he pointed me to this film, which is by a very famous Scandinavian
07:13 - 07:18: director named Ingmar Bergman. I am deep in the throes of parenthood of little kids, and so the
07:18 - 07:25: closest I get to fine Scandinavian film is Frozen. Again, I hadn't seen the movie, but when Jorge
07:25 - 07:31: brought up the reference, I did some research into it. I found that this is a reference that
07:32 - 07:38: is all over literature and film and art. I found this quote from Shakespeare, which I thought was
07:38 - 07:44: appropriate. Therefore, think him as a serpent's egg, which hatched, would as his kind grow
07:44 - 07:50: mischievous and kill him in the shell. Are we really trying to find the serpent's eggs here?
07:50 - 07:55: These precancerous lesions, they're the eggs. They're by themselves never going to hurt
07:55 - 08:00: anybody. They're not progressing as they are, but once they hatch and become an invasive cancer,
08:00 - 08:05: they have the opportunity for a lot more destruction. I'd like to conceptualize
08:05 - 08:11: early detection of pancreatic cancer as finding the serpent's egg and killing it in the shell.
08:13 - 08:19: Can we do this? We need to find the eggs, and then most importantly, we need to predict which
08:19 - 08:24: ones will hatch. The vast majority of these precancerous lesions won't progress. We need
08:24 - 08:31: to avoid overtreatment and find ways to predict which ones will hatch. With that, I'm going to
08:31 - 08:35: tell you a few different stories about clonal evolution in IPMNs and how that might influence
08:35 - 08:42: early detection. I'm going to tell three brief stories in our time. I'll start with talking
08:42 - 08:48: about the origin of IPMNs. This is work done by Katherine Fisher, who was at the time a graduate
08:48 - 08:54: student in my lab and has since graduated and now works for the FDA. What Kathy wanted to do
08:54 - 09:00: was to assess genetic heterogeneity in IPMNs. To do that, she did multi-region targeted sequencing
09:00 - 09:07: of IPMN driver genes. She took 20 surgically resected IPMNs, 10 that were low-grade and 10
09:07 - 09:12: that had at least a component of high-grade dysplasia. IPMNs, when they're surgically
09:12 - 09:17: resected, they get entirely submitted for histologic examination, which means they get cut
09:17 - 09:21: up, put into tissue blocks, and then once the clinical diagnosis is done, they just live in
09:21 - 09:26: the pathology archives. In the pathology archives, we have entirely submitted IPMNs that we can
09:26 - 09:32: analyze. That's exactly what Kathy did. She did laser capture microdissection of these neoplastic
09:32 - 09:38: cells from all available tissue blocks. This led to 233 total samples from these 20 IPMNs.
09:38 - 09:44: The entire IPMN analyzed separately in tissue blocks. She did targeted sequencing of the entire
09:44 - 09:50: coding region of 15 genes. I put the genes. These include our big four mountains as well as some
09:50 - 09:56: IPMN-specific genes like gene SNR and RNF43, and then several other hills in the
09:56 - 10:03: genome landscape. This is a broad overview of Kathy's data. I'll briefly orient you that
10:03 - 10:09: the groups of bars here, each one of these is an IPMN. The tiny bars are a block or a region from
10:09 - 10:14: that IPMN. Then the low-grade IPMNs are here on the left and the high-grade on the right.
10:15 - 10:20: The rows each represent mutations. We have the different hotspot mutations in KRAS and RNF43 at the top.
10:21 - 10:27: Then the less common mutations that are mostly not in hotspots are grouped together
10:27 - 10:32: for the different genes on the bottom. What I want you to appreciate from this diagram is that
10:32 - 10:36: the low-grades look quite different from the high-grades. When you look at KRAS mutations,
10:37 - 10:42: there are often multiple KRAS mutations in the same IPMN or even in the same tissue block from
10:42 - 10:49: the same IPMN in the low-grade. There's often not a single KRAS mutation that's shared across
10:49 - 10:54: the entire IPMN. This is in contrast to the high-grade samples where there's typically one
10:55 - 11:00: mutation in KRAS and/or RNF43 and it's shared through the entire IPMN.
11:02 - 11:07: Looking at this a different way, this is just the number of KRAS and RNF43 mutations. Again,
11:07 - 11:12: low-grade on the left and high-grade on the right. There's a lot more variability in these peaks just
11:12 - 11:16: with respect to the number of KRAS mutations throughout the IPMN as opposed to on the right
11:16 - 11:22: where it's much more flat in the high-grade regions. I think this sequencing data demonstrates
11:22 - 11:28: that there are more often multiple KRAS and RNF43 mutations in low-grade compared to high-grade
11:28 - 11:33: IPMNs. That's the observation, but what does this mean for the origin of IPMN?
11:34 - 11:39: We first tried to answer this using evolutionary modeling. We collaborated with Rachel Karchin
11:39 - 11:46: using her schism approach. What we found was that the subclones defined by distinct KRAS
11:46 - 11:52: mutations did not share a common ancestor other than the germline, suggesting that these represent
11:53 - 11:58: independent clones. The key caveat to this, though, is this is targeted sequencing data.
11:58 - 12:02: Based on this, we can't exclude that they share a mutation outside of our panel.
12:03 - 12:08: To address this, we did whole exome sequencing on a subset of cases where we took regions with
12:08 - 12:14: discordant KRAS mutations in the same IPMN and sequenced the whole exome. In all of these,
12:14 - 12:20: we found no shared mutations between the paired samples, really convincing us that these regions
12:20 - 12:24: with different KRAS mutations really were independent clones within the same IPMN.
12:26 - 12:30: We also wanted to see where these clones were located. To do this, we used a technique called
12:30 - 12:35: Base Scope, which is an in-situ hybridization-based technique to look for KRAS hotspot mutations.
12:36 - 12:42: I'll show one example. This is one slide that has low-grade IPMN on it. We'll highlight three
12:42 - 12:51: different regions here, and I'll zoom in on the one on the top left. This block of IPMN had
12:51 - 12:56: two different KRAS mutations identified. If we look at the zoomed-in area on the left,
12:57 - 13:02: it's KRAS G12D, and there are small dots labeled with arrowheads indicating expression of that
13:02 - 13:10: KRAS mutation and a lack of expression of the other KRAS mutation. When we apply this to the
13:10 - 13:15: whole slide, what we saw is that the mutations in KRAS G12D were on the left and G12V were on the
13:15 - 13:21: right. These multiple KRAS mutations are in different cells, and they're spatially separate.
13:24 - 13:28: This part of the study showed that heterogeneity with respect to early driver gene mutations is
13:28 - 13:33: more prevalent in low-grade IPMNs, and these mutations occur in different cells and are
13:33 - 13:37: spatially separate. This suggests a polyclonal origin for at least a subset of IPMNs.
13:39 - 13:43: Now we'll move on to looking at the transition from high-grade IPMN into pancreatic cancer,
13:44 - 13:48: and we'll look at genomic analysis of malignant progression. This is work done by Michael Noe,
13:48 - 13:57: who's a pathologist and postdoc in the lab. What Michael wanted to do was to separately analyze
13:57 - 14:04: IPMNs and adjacent invasive carcinomas. He separately microdissected IPMN and immediately
14:04 - 14:10: adjacent invasive carcinoma and then did whole exome sequencing of the paired samples. He did
14:10 - 14:15: this from 18 patients, and in an additional seven, he took all the rest of the blocks of the IPMN in
14:15 - 14:23: cancer and separately sequenced those with targeted sequencing. For those cases where we had
14:23 - 14:28: many samples to sequence, Michael did evolutionary analysis and showed that in all cases, the IPMNs
14:28 - 14:35: were the precursors to the invasive cancer. When we look at the mutated genes, in particular,
14:35 - 14:41: which ones are shared between the IPMN and cancer versus distinct to one component,
14:41 - 14:47: you can see the top three genes on the list are not unexpected. It's KRAS, TP53, and P16,
14:47 - 14:51: and these are all mostly shared between the IPMN and the invasive carcinoma.
14:51 - 14:59: However, SMAD4 has a different pattern. In multiple cases, we had inactivating mutations
14:59 - 15:05: in SMAD4 that were present in the invasive cancer but absent in the immediately adjacent high-grade
15:05 - 15:11: IPMN. There's one example shown here with the IPMN samples in blue on the left and the cancer
15:11 - 15:17: sample in red on the right. You can see these. They share a KRAS mutation. They share TP53.
15:17 - 15:22: They share P16, but there's this SMAD4 nonsense mutation that's present only in the cancer and
15:22 - 15:28: absent in the immediately adjacent IPMN. We had these cancer-specific mutations in four of the
15:28 - 15:34: IPMN-cancer sample pairs. Maybe even more surprising was that many IPMNs lacked a
15:34 - 15:39: cancer-specific mutation in a known driver, suggesting that in at least a subset of IPMNs,
15:39 - 15:43: that transition to invasive carcinoma is not mediated by a somatic mutation.
15:44 - 15:49: The next most prevalent gene on this list actually had the opposite pattern of mutation. This is a
15:49 - 15:57: gene called RNF43. It's specifically enriched in IPMN samples that have been sequenced in the
15:57 - 16:02: past. In this study, what we found was that the mutations were quite frequently in the IPMN
16:02 - 16:08: sample but not in the immediately adjacent invasive cancer. It had an even more curious
16:09 - 16:14: pattern where you had multiple distinct RNF43 mutations that were in mutually exclusive
16:14 - 16:20: subclones. There's something, for lack of a better term, funky going on with the evolutionary
16:20 - 16:27: pressures on RNF43 that lead to this convergent evolution of multiple distinct mutations in
16:27 - 16:33: non-overlapping subclones and then absence in the associated invasive cancer. The mechanisms for
16:33 - 16:39: this unique mutation pattern are at this time not known. These RNF43 mutations occur in subclones
16:39 - 16:46: that do not give rise to invasive cancer. There's convergent evolution with multiple RNF43 mutations
16:46 - 16:54: in distinct subclones. From this genomic analysis of malignant progression, we've learned that IPMNs
16:54 - 16:59: are precursors to invasive pancreatic cancer. These SMAD4 mutations and mutations in the related gene
17:00 - 17:03: TGF-beta receptor 2 may drive malignant progression in a subset of cases.
17:04 - 17:09: In other cases, this malignant progression is likely not driven by a somatic mutation.
17:10 - 17:15: Finally, these RNF43 mutations are heterogeneous and often limited to subclones that do not invade.
17:17 - 17:23: The last IPMN story I want to tell you today is about our discovery of a novel driver gene
17:23 - 17:29: mutation in IPMNs. This is work done by Kohei Fujikura and Waki Hasada, who are both pathologists
17:29 - 17:35: who did postdocs in the lab. What Kohei and Waki wanted to do was to analyze the transition from
17:35 - 17:40: low-grade to high-grade dysplasia in IPMNs. They did multi-region whole exome sequencing to do this.
17:40 - 17:49: They took 17 resected IPMNs. As pathologists, they selected cases that had very specific tissue
17:49 - 17:54: features in that they required them to have sizable components of both low-grade and high-grade
17:54 - 17:59: dysplasia in the same IPMN. An example of this is shown on the right, where in the same tissue block
17:59 - 18:07: you have regions that are both low-grade and high-grade that are adjacent. Waki and Kohei did
18:08 - 18:12: laser capture microdissection of the neoplastic cells in these low-grade and high-grade regions
18:12 - 18:17: in one to three tissue blocks. This resulted in two to six separate regions per IPMN,
18:17 - 18:22: so 76 total neoplastic regions that were exome sequenced along with a matched normal
18:23 - 18:30: to look in detail at these genomes. When we look at the mutated genes, the first two on the list
18:30 - 18:34: are not surprising: GNAS and KRAS, we've talked already. There's the initiating oncogene mutations
18:34 - 18:42: in IPMN tumorigenesis. RNF43, we just mentioned in the previous section. The fourth gene is
18:42 - 18:47: this gene called KLF4. This has not been frequently or previously described as being
18:47 - 18:53: frequently mutated in pancreatic neoplasms. We found it to be quite frequently mutated in over
18:53 - 18:57: half of our cases. It was enriched in the low-grade components of the IPMNs, which
18:57 - 19:01: is a very curious mutation pattern that hasn't been reported before.
19:03 - 19:10: What is KLF4? This is a transcription factor. One of the most striking things about the mutations
19:10 - 19:14: in this gene is that they occurred in one of two hotspots. They weren't spread evenly throughout
19:14 - 19:22: the gene. They either occurred in codon 409 or codon 411. We wondered, why have these not been
19:22 - 19:26: reported in IPMNs before? We dug into the supplementary material of previously published
19:26 - 19:33: IPMN sequencing studies. We found occasional cases with KLF4 mutations, but previous sequencing
19:33 - 19:38: studies have largely focused on high-grade IPMNs. Because this is enriched in the low-grade,
19:38 - 19:42: we think that's why this hasn't been discovered before. They're very rare in PDAC. We looked at
19:42 - 19:50: the TCGA data and found it in one out of 184 sporadic PDAC cases. These hotspot mutations
19:50 - 19:55: occur in critical residues of the DNA binding domain of this transcription factor. It seems
19:55 - 19:59: likely that they are impacting the function, though we don't know exactly how yet.
20:00 - 20:05: One other interesting point is that these mutations occur at
20:05 - 20:07: this position that was previously reported to be
20:07 - 20:10: frequently mutated in the secretory subtype of meningioma,
20:10 - 20:13: so a certain subtype of brain tumor.
20:13 - 20:15: These hotspot mutations have been
20:15 - 20:17: reported frequently in another tumor type.
20:17 - 20:19: Again, adding to the evidence that these do
20:19 - 20:21: have some functional role.
20:21 - 20:23: But the molecular consequences of
20:23 - 20:27: these mutations in pancreatic tumorigenesis are not known.
20:27 - 20:29: We validated the KLF4 mutations
20:29 - 20:32: in an independent cohort of tissue samples,
20:32 - 20:35: using a safe sequencing system assay
20:35 - 20:36: developed by our collaborators,
20:36 - 20:38: Burt Vogelstein and Anne-Marie Lennon.
20:38 - 20:40: We did find again, and this is
20:40 - 20:44: a completely independent cohort that about over 40 percent,
20:44 - 20:45: close to half of low-grade IPMNs
20:45 - 20:48: had at least one mutation in KLF4.
20:48 - 20:51: Again, highlighting its high prevalence in low-grade IPMNs.
20:51 - 20:54: But it's important to point out that almost 20 percent of
20:54 - 20:57: high-grade IPMNs also had KLF4 mutations,
20:57 - 20:59: suggesting that this is a standalone biomarker for
20:59 - 21:03: low-risk is not really feasible.
21:03 - 21:07: We've learned that these KLF4 hotspot mutations are common in
21:07 - 21:11: IPMNs and often limited to low-grade regions.
21:11 - 21:12: The presence of a KLF4 mutation
21:12 - 21:16: might suggest that the egg is unlikely to hatch.
21:16 - 21:19: Taken together, I hope that these stories have convinced you
21:19 - 21:22: that we need some new conceptual models
21:22 - 21:24: for pre-malignancy in IPMN tumorigenesis.
21:24 - 21:27: We have polyclonal origin in the beginning,
21:27 - 21:30: with multiple independent clones
21:30 - 21:31: with different KRAS mutations
21:31 - 21:33: mixed together in the same IPMN.
21:33 - 21:36: We have KLF4 mutations and low-grade subclones,
21:36 - 21:38: and then more broadly convergent evolution
21:38 - 21:40: with respect to other mutations.
21:40 - 21:41: Then finally, invasion of a clone which
21:41 - 21:45: can be mediated by SMAD4 mutation.
21:46 - 21:48: What are the future directions?
21:48 - 21:50: How can these findings improve the care of IPMN patients?
21:50 - 21:53: Can we predict which eggs are most likely to hatch so that we
21:53 - 21:55: can intervene surgically and
21:55 - 21:57: resect those IPMNs that are high-risk,
21:57 - 21:59: but avoid unnecessary surgery in patients
21:59 - 22:01: whose IPMNs are unlikely to progress?
22:01 - 22:05: Again, we need to improve preoperative prediction of
22:05 - 22:09: IPMN behavior using tissue analysis.
22:09 - 22:11: We have identified a potential
22:11 - 22:14: new genetic biomarker in KLF4.
22:14 - 22:16: But as I mentioned, the prevalence that
22:16 - 22:17: is in high-grade IPMN as well,
22:17 - 22:20: it's never going to work standalone.
22:20 - 22:22: But I think the field is increasingly
22:22 - 22:25: moving toward combinations of genetic markers,
22:25 - 22:28: such as saying that a KLF4 mutant IPMN that is also
22:28 - 22:32: wild-type for the high-grade driver genes
22:32 - 22:36: like TP53 and SMAD4 might suggest that it is low-risk.
22:36 - 22:38: Another important concept is that rather than
22:38 - 22:40: just classifying IPMNs as
22:40 - 22:43: binary mutant versus wild-type for a given gene,
22:43 - 22:46: we can incorporate the number of mutations or
22:46 - 22:48: even the variant allele frequency of those mutations to
22:48 - 22:51: better understand where that IPMN is in
22:51 - 22:54: clonal progression rather than just that,
22:54 - 22:56: again, the binary mutant wild-type,
22:56 - 22:59: which is how most assays are read out at this time.
22:59 - 23:01: Any of these ideas, though,
23:01 - 23:02: I think it's important to point out,
23:02 - 23:03: require testing and validation in
23:03 - 23:05: large cohorts of tissue samples.
23:05 - 23:09: I think that's where the field needs to go.
23:09 - 23:11: With that, I want to close by saying,
23:11 - 23:13: I think we can kill the serpent in the shell.
23:13 - 23:15: The key is predicting which eggs are
23:15 - 23:17: likely to hatch to avoid over-treatment,
23:17 - 23:18: and to develop these strategies,
23:18 - 23:21: we need to study the egg itself.
23:21 - 23:23: Study the human precancerous lesions to
23:23 - 23:25: really understand their behavior.
23:25 - 23:28: I think that molecular screening is going to play
23:28 - 23:31: a major role in the risk stratification in the future.
23:31 - 23:33: With that, I want to acknowledge
23:33 - 23:35: the folks in my laboratory who did this work,
23:35 - 23:37: as well as several collaborators,
23:37 - 23:41: our funding, and I am happy to take questions.
23:41 - 23:44: Thanks, Laura. That was brilliant.
23:44 - 23:46: I will remind you, everyone
23:46 - 23:47: to, if you have questions,
23:47 - 23:49: put them in the Q&A,
23:49 - 23:51: and I will start with some questions first.
23:51 - 23:55: You mentioned that you have
23:55 - 23:58: also some data about PANINs.
23:58 - 24:00: I wonder whether you see some of
24:00 - 24:03: the observations that you see in IPMNs,
24:03 - 24:04: also in different rates of
24:04 - 24:07: PanIN development or
24:07 - 24:09: whether you see something completely different.
24:09 - 24:11: We have done some work with PanIN.
24:11 - 24:13: The challenge with them, as I mentioned in
24:13 - 24:15: the beginning, is that they're microscopic.
24:15 - 24:18: Understanding the full extent of PanINs as
24:18 - 24:20: they microscopically involve the duct system
24:20 - 24:22: is much more challenging than with an IPMN,
24:22 - 24:27: where you have a big hollow wall.
24:27 - 24:29: We've had to develop approaches for doing
24:29 - 24:31: 3D reconstruction of pancreatic tissue,
24:31 - 24:32: which we've done with colleagues
24:32 - 24:34: here in the School of Engineering,
24:34 - 24:36: our collaborator, Denise Wirtz.
24:36 - 24:37: We have, with that,
24:37 - 24:39: been able to do some sequencing of PanINs and
24:39 - 24:42: really understanding their anatomic relationship.
24:42 - 24:45: Those studies have shown that low-grade PanINs
24:45 - 24:47: are insanely common.
24:47 - 24:48: We all have a lot of them,
24:48 - 24:50: and separate PanINs all have
24:50 - 24:52: different KRAS mutations.
24:52 - 24:55: There's a lot of multifocal neoplasia going on.
24:55 - 24:56: Even within one PanIN,
24:56 - 24:59: we have seen a few cases
24:59 - 25:01: with multiple KRAS mutations.
25:01 - 25:05: I think that polyclonal origin also applies to PanIN.
25:05 - 25:09: The challenge with the transition to cancer,
25:09 - 25:11: well, transition to high-grade,
25:11 - 25:12: high-grade PanINs are quite uncommon,
25:12 - 25:17: and so in prospectively harvested normal pancreatic tissue,
25:17 - 25:20: it's hard to just get lucky and get a high-grade PanIN.
25:20 - 25:23: We have a few, and we do see that genetic progression.
25:23 - 25:26: But with respect to the cancer,
25:26 - 25:27: we have not found a case where we
25:27 - 25:31: identified the PanIN that gave rise to the cancer.
25:31 - 25:34: I think that that's a challenge that
25:34 - 25:35: is much, that question is much
25:35 - 25:38: harder to address in PanINs than IPMNs.
25:38 - 25:43: Thank you. We have quite a few questions coming up.
25:43 - 25:48: One question is about KLF4 mutations and whether
25:48 - 25:50: you analyze them in
25:50 - 25:54: circulating cell-free DNA in blood plasma.
25:54 - 25:56: We have not done that.
25:56 - 25:59: I suspect that it will be
25:59 - 26:01: challenging because IPMNs are non-invasive,
26:01 - 26:03: and so by definition,
26:03 - 26:06: they shouldn't be able to get their DNA into
26:06 - 26:10: the circulating blood or plasma.
26:10 - 26:13: I think those plasma-based assays
26:13 - 26:14: are useful for invasive cancers,
26:14 - 26:16: but I think for precursors,
26:16 - 26:18: it's much more challenging because they shouldn't be
26:18 - 26:21: secreting, not secreting, they shouldn't be
26:21 - 26:25: connected to the blood system to get their DNA in there.
26:25 - 26:31: Another question is about how common is
26:31 - 26:35: the polyclonal origin heterogeneic neoplastic
26:35 - 26:37: precursors and other cancer types,
26:37 - 26:41: or is this unique to pancreatic cancer?
26:43 - 26:47: That's testing my knowledge of other tumor types.
26:47 - 26:49: It's a great question.
26:49 - 26:56: I think there's not a ton of work on precancers.
26:56 - 26:57: I will say that there's a lot of work on
26:57 - 27:00: normal tissue and other organs,
27:00 - 27:03: and there's a disturbing number of
27:03 - 27:05: expanded clones even in normal tissue.
27:06 - 27:08: The Sanger Center has been a real leader on this,
27:08 - 27:10: and they did work on eyelid skin,
27:10 - 27:12: they did work on esophagus.
27:12 - 27:15: I think this multifocal neoplasia,
27:15 - 27:17: even in normal tissue,
27:17 - 27:20: is a lot more common than we appreciate.
27:20 - 27:23: But with respect to precursors,
27:23 - 27:26: I don't know the data in other organ systems, I apologize.
27:26 - 27:28: The mutations pattern is typically
27:28 - 27:31: different than the one just described.
27:31 - 27:34: Another question is whether there are concerns of
27:34 - 27:36: sampling error from cyst fluid?
27:36 - 27:40: Yeah, I think that is a really important point.
27:41 - 27:44: Look down on our gastroenterology colleagues,
27:44 - 27:46: they do what they can do,
27:46 - 27:50: but I think because the cysts are often multiloculated,
27:50 - 27:53: meaning they have multiple different regions
27:53 - 27:55: that are variably connected,
27:55 - 27:57: I think that exactly where
27:57 - 27:59: you put the needle in can affect what you sample.
27:59 - 28:01: I think that hasn't been well
28:01 - 28:02: quantified in cyst fluid sequencing studies,
28:02 - 28:06: and it is an important thing to get a sense of.
28:06 - 28:08: In Kathy's paper, we did a bit of that,
28:08 - 28:12: where we serendipitously had a few IPMs,
28:12 - 28:14: where we analyzed the whole tissue that
28:14 - 28:16: had had previous cyst fluid sequencing,
28:16 - 28:20: and we did identify some smaller clones
28:20 - 28:21: that were completely missed by
28:21 - 28:23: the cyst fluid sequencing, presumably due to sampling.
28:23 - 28:26: I think that's super important.
28:26 - 28:28: Especially because you showed this difference
28:28 - 28:31: in spatial location of the different mutations.
28:31 - 28:32: Exactly. I think it all depends
28:32 - 28:34: on where the needle goes in.
28:34 - 28:37: What's the frequency in having
28:37 - 28:40: both low-grade and high-grade within the same patient?
28:40 - 28:41: Yeah.
28:41 - 28:43: I mean, monitor this.
28:44 - 28:46: When you have high-grade,
28:46 - 28:49: there's often an associated low-grade component.
28:49 - 28:51: Because the way that the IPMNs are graded
28:51 - 28:54: is just by the highest grade present.
28:54 - 28:56: When we say a pathologist says it's a high-grade IPMN,
28:56 - 28:59: there's often an associated low-grade component.
28:59 - 29:02: But still, the low-grades are much more common.
29:02 - 29:05: I think any surgical series,
29:05 - 29:08: to get the ground truth from the surgical specimens,
29:08 - 29:10: you have to have taken it out.
29:10 - 29:11: Any surgical series is going to be
29:11 - 29:12: biased by the fact that
29:12 - 29:15: most IPMNs don't get taken out.
29:15 - 29:16: What I've seen with my eyes is
29:16 - 29:18: very enriched for high-grade dysplasia,
29:18 - 29:19: but that's because those are
29:19 - 29:20: the patients who go to surgery.
29:20 - 29:23: Most patients have their IPMNs monitored,
29:23 - 29:24: they don't have worrisome features,
29:24 - 29:27: and so they never come across my microscope.
29:28 - 29:31: Another question is whether there's
29:31 - 29:33: any kind of patient with
29:33 - 29:35: a specific clinical background that has
29:35 - 29:39: higher frequencies of these KLF4 mutations?
29:39 - 29:43: That's another super interesting question.
29:43 - 29:44: I think we don't know that yet
29:44 - 29:48: because our sample size is too small right now.
29:48 - 29:50: They are enriched in the low-grade
29:50 - 29:51: IPMNs versus the high-grade,
29:51 - 29:53: but other features, we don't
29:53 - 29:55: have enough samples to really say that.
29:56 - 29:59: A technical question perhaps is,
29:59 - 30:01: obviously, you don't want to keep poking
30:01 - 30:05: the pancreas of a patient to induce inflammation.
30:05 - 30:09: What would be the logical and how
30:09 - 30:11: frequent it would be possible
30:11 - 30:14: to monitor these patients over time?
30:14 - 30:18: Yeah, that's another really important question.
30:18 - 30:19: I have to defer to,
30:19 - 30:21: there's specific guidelines in
30:21 - 30:22: the gastroenterology community as
30:22 - 30:26: to how often folks should get screened.
30:26 - 30:30: It's typically a combination of imaging,
30:30 - 30:31: whether it's CT or MRI,
30:31 - 30:36: and then endoscopic ultrasound with fine needle aspiration.
30:36 - 30:41: But yeah, I don't want to
30:41 - 30:43: speak to the wrong exact clinical guidelines,
30:43 - 30:45: but there are guidelines existing for that.
30:45 - 30:48: I think another point when we think about
30:48 - 30:50: sampling is that Mike Goggins has done
30:50 - 30:52: great work here at Hopkins on sampling
30:52 - 30:53: pancreatic juice, so rather
30:53 - 30:56: than sticking a needle into the IPMN,
30:56 - 30:58: you give the patient something called secretin that makes
30:58 - 31:00: the pancreas spit out juice into the duodenum,
31:00 - 31:02: and you can suck that up and sequence it.
31:02 - 31:05: I think that might be a nice complementary tool,
31:05 - 31:08: not only for avoiding stabbing the pancreas,
31:08 - 31:10: but also to sample
31:10 - 31:13: the whole pancreas rather than just the IPMN.
31:13 - 31:16: One set of data I didn't mention is that we did
31:16 - 31:17: some work with IPMNs in
31:17 - 31:20: cancers and looked at how often they were related.
31:20 - 31:22: We actually found that about a fifth of the time,
31:22 - 31:24: the pancreatic cancer that arises in
31:24 - 31:27: an IPMN patient is totally independent.
31:27 - 31:29: You won't see those high-risk mutations in
31:29 - 31:30: the IPMN cyst fluid because
31:30 - 31:32: the cancer doesn't communicate with the cyst.
31:32 - 31:35: I think whole pancreas sampling approaches
31:35 - 31:37: in the future are going to be quite important too.
31:37 - 31:41: We're going to end with a very tough question.
31:41 - 31:42: Oh no.
31:42 - 31:44: Although there are way more questions here,
31:44 - 31:45: but hopefully we'll go back to
31:45 - 31:47: them during the panel discussion.
31:47 - 31:49: This is, how can we relate
31:49 - 31:51: patient symptoms to the existence of
31:51 - 31:54: precancerous lesions and so help with
31:54 - 31:57: the earlier diagnosis and if this is at all possible?
31:57 - 32:01: Yeah. I mean, I think with IPMNs it is possible.
32:01 - 32:06: These cysts do, they are most often diagnosed incidentally,
32:06 - 32:08: but they do sometimes cause symptoms.
32:08 - 32:11: I think the challenge with PanINs is that they're quite
32:11 - 32:15: small and they don't often provide symptoms.
32:15 - 32:19: Any screening approach, if the goal is to detect
32:19 - 32:21: a precancerous lesion, it will have to rely
32:21 - 32:24: on biomarkers rather than symptoms,
32:24 - 32:26: because even high-grade PanINs are still so small
32:26 - 32:27: that I think they're unlikely to
32:27 - 32:30: cause physical symptoms in the patients.
32:30 - 32:32: Great. Thank you so much.
32:32 - 32:36: We'll get back to you shortly and I'm very
32:36 - 32:39: excited to introduce our second speaker,
32:39 - 32:42: Dr. Tobilo Baoni.
32:42 - 32:44: He's not only a fantastic scientist,
32:44 - 32:47: he's a great collaborator and friend.
32:48 - 32:51: It's such a pleasure to introduce you.
32:51 - 32:54: Dr. Baoni earned his PhD from Stony Brook University
32:54 - 32:57: and he performed his graduate studies
32:57 - 33:00: with Dave Tuvason at Cold Spring Harbor Laboratory.
33:00 - 33:03: After that, since last year,
33:03 - 33:06: he is a Whitehead Fellow at the Whitehead Institute.
33:06 - 33:08: His lab seeks to uncover the elements of
33:08 - 33:11: effective anti-tumor immunity and develop
33:11 - 33:15: novel antibody-based tools to induce tumor clearance,
33:15 - 33:16: with the goal to develop new treatments
33:16 - 33:18: for pancreatic cancer.
33:18 - 33:20: Thank you so much, Toby.
33:20 - 33:22: Today, he's going to talk about interrogating
33:22 - 33:24: the tumor surface zone for early detection
33:24 - 33:26: in pancreatic cancer.
33:26 - 33:28: Thanks for joining us.
33:29 - 33:32: Thank you very much.
33:32 - 33:36: Thanks, everyone, for —
33:36 - 33:38: well, let me just share my screen.
33:38 - 33:39: Okay, great.
33:39 - 33:40: Thanks, everyone, for being here
33:40 - 33:44: and particularly Abcam for the invitation.
33:44 - 33:46: I'm going to talk about our work,
33:46 - 33:49: particularly starting from when I was in
33:49 - 33:52: Cold Spring Harbor Laboratory with David Tuvason,
33:52 - 33:56: and this is about interrogating the tumor cell surface zone
33:56 - 33:59: for early detection in pancreatic cancer.
33:59 - 34:02: So I'm going to start out by stating the obvious,
34:02 - 34:06: which is that most patients don't make it
34:06 - 34:09: for very long in this disease.
34:09 - 34:12: And within the first year, as you can see here,
34:12 - 34:14: most patients actually die.
34:14 - 34:16: And there's this long tail of patients
34:16 - 34:20: that are living for a long time, as you've heard earlier.
34:20 - 34:22: And the question is, what happens here?
34:22 - 34:24: You know, what's really happening here?
34:24 - 34:27: Well, it turns out that patients that are diagnosed
34:27 - 34:30: while the disease is localized
34:30 - 34:34: are the ones really holding that long line of survival.
34:34 - 34:38: And just to summarize this here,
34:38 - 34:41: most of the cases that you get with pancreatic cancer
34:41 - 34:43: are diagnosed at a point
34:43 - 34:46: when it's either regional or distant,
34:46 - 34:51: which means it's essentially, you know, pretty late.
34:51 - 34:54: And when you look at the survival rate,
34:54 - 34:55: it's very clear that, you know,
34:55 - 34:57: when this disease is diagnosed earlier,
34:57 - 35:00: when the disease is still localized,
35:00 - 35:04: it's very clear that you have a higher survival rate.
35:04 - 35:07: And so given the current therapies that we have
35:07 - 35:11: that are not very effective,
35:11 - 35:14: if all we do is just to shift this back
35:14 - 35:17: and just detect this disease a little earlier,
35:17 - 35:20: we can actually make a tremendous impact in this disease.
35:20 - 35:24: This is without even discovering new therapies.
35:24 - 35:26: And so this underlies the importance
35:26 - 35:30: of really being able to discover new ways
35:30 - 35:33: to detect this disease earlier.
35:33 - 35:38: And this was, you know, the major point of my work,
35:38 - 35:42: you know, when I was at Cold Spring Harbor.
35:42 - 35:45: And it's clear that we currently don't have,
35:45 - 35:47: you know, very good early detection biomarkers.
35:47 - 35:50: Otherwise, I wouldn't be here.
35:50 - 35:53: And importantly, even on top of that,
35:53 - 35:56: the current imaging methods that we have,
35:56 - 35:58: particularly non-invasive ones,
35:58 - 36:01: they're not adequately sensitive.
36:01 - 36:05: And this extends to being able to see
36:05 - 36:08: if you have a small cancer
36:08 - 36:11: disseminated at some other part of your body.
36:11 - 36:13: And so being able to see this through imaging
36:13 - 36:15: is very important,
36:15 - 36:18: because then it doesn't lead to understaging
36:18 - 36:21: and potentially not being able
36:21 - 36:25: to get the right treatment that you need.
36:25 - 36:27: So to really solve these problems,
36:27 - 36:29: you know, we're really thinking,
36:29 - 36:32: how can we detect pancreatic cancer
36:32 - 36:35: wherever it could be, right?
36:35 - 36:38: And so something that, you know,
36:38 - 36:41: a lot of us that work in this field all know,
36:41 - 36:43: but we kind of take for granted,
36:43 - 36:44: is the fact that, you know,
36:44 - 36:47: genetic aberrations in tumor cells
36:47 - 36:50: actually lead to dysregulation of several processes.
36:50 - 36:53: And this includes expression of particular proteins
36:53 - 36:55: on the cell surface, for example,
36:55 - 37:00: splicing, post-translational modifications.
37:00 - 37:03: And what this leads to is a systematic difference
37:03 - 37:07: between a tumor cell and a normal cell.
37:07 - 37:10: And so being able to see those little differences
37:10 - 37:11: between tumor and normal cells
37:11 - 37:15: actually might give us a way to detect the tumor cells
37:15 - 37:18: anywhere they could be in the body.
37:18 - 37:22: And this was essentially what I set out to do.
37:22 - 37:25: And the goal here was to use antibodies
37:25 - 37:29: because of the exquisite specificity of antibodies.
37:29 - 37:32: We were thinking, well, the antibodies could tell apart
37:32 - 37:35: even the small differences on the cell surface
37:35 - 37:38: of a tumor cell versus a normal cell.
37:38 - 37:40: And if we can use these antibodies
37:40 - 37:42: to either detect the tumor itself
37:42 - 37:45: or even detect some of the antigens
37:45 - 37:48: that these tumor cells secrete into circulation,
37:48 - 37:52: then we could have imaging-based antibody reagents
37:52 - 37:56: and also antibodies that we can use
37:56 - 38:02: as a way to do serological tests for pancreatic cancer.
38:02 - 38:04: And so the way we set out to do this
38:04 - 38:07: was to use a method called the hybridoma method.
38:07 - 38:11: This is a traditional method where you inject the antigen,
38:11 - 38:14: you know, into a mouse or a rat
38:14 - 38:17: and then isolate the antibody-secreting cells,
38:17 - 38:22: fuse them with a myeloma cell to essentially immortalize them,
38:22 - 38:25: and then, you know, select and screen them.
38:25 - 38:28: And when you find one clone that actually binds
38:28 - 38:30: to your original target, then you can expand that.
38:30 - 38:34: And that's your monoclonal antibody.
38:34 - 38:35: And this has been done before, you know,
38:35 - 38:38: actually, you know, in the 1980s.
38:38 - 38:40: And when this was done, you know,
38:40 - 38:43: using homogenized tumor tissue for pancreatic cancer,
38:43 - 38:45: you know, people didn't find it a lot.
38:45 - 38:49: And the reason for this is that pancreatic cancer actually,
38:49 - 38:51: you know, when you take a look at it,
38:51 - 38:54: you see that the tumor cells make a small percentage
38:54 - 38:55: of the whole tumor.
38:55 - 38:59: So the old tumor is, like, you know,
38:59 - 39:01: congested with all this stroma.
39:01 - 39:03: And so if you inject the whole tissue,
39:03 - 39:06: you essentially are injecting a small percentage
39:06 - 39:08: of the tumor cells.
39:08 - 39:10: So the other way is to use cell lines,
39:10 - 39:15: which didn't turn out to also work great.
39:15 - 39:19: But then, you know, years ago, we developed a new system,
39:19 - 39:20: which is the organoid system,
39:20 - 39:25: which allowed us to grow pancreatic cancer in a dish,
39:25 - 39:28: but grow them in this 3D matrix.
39:28 - 39:30: And importantly, it allowed us
39:30 - 39:34: to efficiently grow patient tumors ex vivo.
39:34 - 39:37: And what this does is to help us retain
39:37 - 39:40: the heterogeneity between patients
39:41 - 39:43: and potentially allow us to have a good idea
39:43 - 39:47: of representative shared antigens between patients.
39:48 - 39:51: And so I set out to use a system of organoids
39:51 - 39:55: to develop monoclonal antibodies against PDAC
39:55 - 39:58: to find those special epitopes that tumor cells have
39:58 - 39:59: and normal cells don't have.
40:00 - 40:10: And so I'm going to tell you very quickly about how we produce the antibodies and how we validated them and finally, some applications for them.
40:10 - 40:30: So first, how we produce the antibodies. So essentially what I did was to get, you know, 15 organoids from 15 different patients, and essentially pull them together, extract the membrane fraction, and then immunize rats with this membrane fraction.
40:30 - 40:40: And then I performed the hybridoma method, and then screened these antibodies by cell-based ELISA.
40:40 - 40:56: So, for the screening, we just did cell-based ELISA using some pancreatic cancer cells. And what you see here is like these red circles are the antibodies that actually bound the cells and the black circles are the ones that didn't make it.
40:56 - 41:03: And so we had about 18% of the total clones that were screened that were positive for binding.
41:03 - 41:14: So the next question was, well, do these antibodies really bind to something on the cell surface, because remember the cell surfaces were actually accessible by antibodies.
41:14 - 41:22: And so if it's intracellular, you can't get there; you can't reach it, so you can't use that to detect the tumor cells.
41:22 - 41:40: And so we performed flow cytometry on intact cells, and you see that these red dots actually show really nicely that these antibodies are binding to the tumor cells on the outside, and not on the inside because these tumor cells are intact, they're not
41:40 - 41:43: perforated.
41:43 - 41:58: And so from now on, I'm going to be calling these antibodies TOBi-bodies, and this stands for tumor organoid binding antibodies. And this is by coincidence that it turns out to be my name.
41:58 - 42:08: And so the next question was, fine, we have these antibodies but if we look in tissues of these patients, does it really tell tumors apart from normal.
42:08 - 42:26: So this is just a control IDG control just to show that if you stain this with tumor, and normal, there's essentially, you know, no staining, and the blue you see is just the nucleus, which shows you where the
42:26 - 42:28: cells are.
42:28 - 42:45: And so in TOBi-bodies, you see that there's a clear difference between the tumor and normal from this in patients, which tells very clearly that these TOBi-bodies are distinguishing between the tumor and normal.
42:45 - 42:58: So the next question was, well, what did they bind to? And so we did a bunch of experiments including immunoprecipitation mass spec, so immunoprecipitation with these TOBi-bodies and then mass spectrometry.
42:58 - 43:19: So, I'll just show you two examples of two TOBi-bodies that we performed this assay with, and using mass spec, you see that TOBi 85 here is binding to SLC6A20 on two different cell lines, which tells you that SLC6A20 is the protein that this
43:20 - 43:41: And then the next question was, well, are there other TOBi-bodies that are binding to SLC6A20 or even a closely related antigen SLC5A5.
43:41 - 43:55: And there's actually 10 of them that bind to SLC6A20 or SLC5A5, so what you can see here is that these antibodies either bind only to SLC6A20, or in the case of TOBi 111 here,
43:55 - 43:58: it's binding to both SLC6A20 and SLC5A5.
43:58 - 44:10: So what this means—and this is critical—what this means is that these TOBi-bodies are binding to different epitopes, so it's the same antigen, but they're binding to different parts of that antigen.
44:10 - 44:13: And this is very critical.
44:13 - 44:28: And so, for example, I, you know, I'll show you here that we can actually show very clearly that when you knock out SLC6A20, the TOBi-bodies that are only SLC6A20 reactive, they don't bind anymore.
44:28 - 44:36: Right. So if you take an SLC6A20 antibody that's commercially available, it's completely gone; you knock out SLC6A20 by CRISPR.
44:36 - 44:48: And this mark here shows that it's completely gone. And same with the TOBi-body, so that's SLC6A20 specific, so it's not binding to anything else in the cells—it's only binding to SLC6A20.
44:48 - 44:58: And so, you know, you might be thinking, well, fine, it distinguishes between, you know, tumor versus normal in the pancreas. What about other cells in the body?
44:58 - 45:07: So you can look at normal human tissues from different cells from different tissues, and just ask the questions.
45:07 - 45:14: Well, does this TOBi-body distinguish between these tissues and pancreatic cancer?
45:14 - 45:23: And the interesting thing is that even though you had different antibodies that were binding to SLC6A20,
45:23 - 45:36: you got this really interesting pattern. You had some that are binding to pancreatic cancer, but don't bind to normal pancreas, but then when you look at other tissues.
45:36 - 45:50: You can see that they start stop binding. So in the case of this commercial antibody or even like this TOBi 114, it's binding to, you know, these squamous epithelial cells in the esophagus and also binding in the colon.
45:51 - 46:07: So, if you take TOBi 85 and 89, they also distinguish PDAC from normal pancreas, but they don't bind to, you know, any other tissue in the body. So they have minimal binding to other tissues in the body, while distinguishing between PDAC and normal.
46:08 - 46:13: So, these antibodies are binding to unique epitopes on SLC6A20.
46:13 - 46:19: And so the next question was, well, what about chronic pancreatitis, which is like a disease state?
46:20 - 46:31: So we tested that and it turns out it binds to a majority of PDAC tissues, but not to chronic pancreatitis, not to normal pancreatic cancer tissues.
46:31 - 46:43: Same with TOBi 89. And we've actually tested this on slides we got from Laura and others at Johns Hopkins to see whether they bind to precursor lesions.
46:44 - 46:57: And the summary right now, we're still analyzing the data but the summary is that these TOBi-bodies are very specific for frank PDAC and might be seeing some high-grade lesions.
46:57 - 47:07: And so, what are these epitopes, what are these antibodies really binding to? Well, if you look at the, you know, structure of SLC6A20, it’s just a 2D structure.
47:08 - 47:13: And it has multiple glycans, several glycans, not just one.
47:13 - 47:17: And it also has a disulfide bond that holds together the ID domains.
47:19 - 47:35: And so we thought, well, let's get rid of this glycans, or get rid of this disulfide bond and see whether this is a conformation or epitope that that this antibodies are binding to, or, you know, is this just a glycan but I did bind to this
47:35 - 47:37: sugars and this protein.
47:37 - 47:55: And so I did this experiment where I got rid of the glycans by using this enzyme called PNGase F, and also added reducing agent. And so in this case if you take a commercially available SLC6A20 antibody, it binds everything; you know, you add reducing
47:55 - 48:12: agent, it binds to it; if you get rid of the sugar, it binds to it; it doesn't really matter. Right. But then if you take this TOBi-body, you see very clearly that once you get rid of the disulfide bond and you lose that structure.
48:13 - 48:31: And then when you also remove the glycans, then the binding is minimal—it's reduced. So this tells us that this, you know, TOBi-bodies—and several of them are like this—they're binding to glycan-dependent conformational epitopes on SLC6A20.
48:32 - 48:45: And so, the summary here is we had several TOBi-bodies that we ended up with two tumor-specific TOBi-bodies that bind to epitopes on SLC6A20 that are glycan-dependent.
48:45 - 48:57: And we really think we can now use these TOBi-bodies to specifically detect pancreatic cancer in several assays.
48:58 - 49:14: So, I just want to, you know, really hammer on this point which is that a lot of us think about, you know, biology or even early detection stopping at protein; you know, we look at, you know, protein markers, biomarkers, and all that we do proteomics.
49:15 - 49:30: I think we're in an age now where we should be looking at epitomics—not just proteomics or, you know, just looking at the proteins themselves—because epitopes in cancers are different, and we should be doing global epitomics.
49:30 - 49:45: And one of the things that my lab is doing is actually leading those efforts to do global epitomics to determine what are globally different between cancer and normal cells or cancer and, like, you know, other disease states, particularly by looking at the epitopes.
49:45 - 50:06: Okay, I'm gonna end on this last note of what can we do with these antibodies. First, of course, we can use this for, you know, imaging to detect the cancer, particularly earlier—earlier when it's tiny or when it's still, you know, kind of morphing into
50:07 - 50:09: PDAC.
50:09 - 50:15: And also, we can use this as a way to detect, you know, this cancer is also in the blood.
50:16 - 50:33: So, these TOBi-bodies will be useful to target, you know, different payloads or even, you know, a drug immunotherapy in pancreatic cancer. And finally, we can actually start to use these to my specific TOBi-bodies to interrogate tumor biology
50:33 - 50:38: and ask the question, what are the functions of these epitopes.
50:38 - 50:45: And I'm just going to mention this even a PET approach of using these TOBi-bodies as an imaging reagent.
50:45 - 50:59: And so, we performed labeling of these TOBi-bodies, just labeling this with zirconium-89, and this is with Jason Lewis and Scott Lyon's lab.
51:00 - 51:10: And we can actually create a mouse model where we can inject the pancreas patient-derived organoids into the parenchyma of the pancreas.
51:10 - 51:25: And when we do this, we're able to actually measure the tumor that are derived from this using small animal ultrasound really nicely; it's about 3.4 millimeters in diameter.
51:25 - 51:41: And we can also use these TOBi-bodies that are zirconium-labeled to see these tumors. So, I mean this is not surprising; you know, the tumors are big enough to be seen by this ultrasound. You can also use these TOBi-bodies to see it.
51:41 - 52:02: But the critical question was, well, what if it's a tiny, tiny tumor? And so, thankfully, Koji Miyabayashi, who was at a lab at the time, created this really nice model where he was able to inject the pancreas patient-derived organoids into the duct of the
52:03 - 52:20: pancreas, and what he does is it creates this multi-nodal lesions that are spanning the length of the pancreas. So you have multi-nodal lesions that are in different places in the pancreas but then essentially spanning the length—and that's what
52:20 - 52:24: this red circles correspond to.
52:24 - 52:37: Can you see it? And these lesions are less than one millimeter. And so, we really wanted to test the ability of these TOBi-bodies to see it in this format.
52:37 - 52:47: When you do it this way, you don't see these lesions using the ultrasound. And so, you know, it's almost like there's no tumor there, right?
52:47 - 53:01: And when you use these TOBi-bodies, particularly TOBi 89, you see really nicely that you're able to delineate these lesions really nicely, and that's why you see these multi-nodal lesions across the length of the pancreas—you could actually see the
53:01 - 53:08: pancreas, you know, using these lesions as a proximal way to look at it.
53:08 - 53:19: And so, this tells us really nicely that these TOBi-bodies could be useful as a PET imaging reagent for pancreatic cancer and have the potential to detect when the pancreas.
53:19 - 53:31: When this pancreatic cancer has not left the pancreas, and maybe when they're still in that stage where we can come for, you know, better survival for the patients.
53:31 - 53:45: And so we're humanizing these TOBi-bodies for clinical PET imaging at the moment, and we're, you know, really evaluating also the potential of the TOBi-bodies to drive therapeutic targeting in preclinical models.
53:46 - 53:56: And finally, you know, we're trying to ask whether these TOBi-bodies can really detect, you know, the specific epitopes in the blood.
53:56 - 54:02: This is not an easy part, but it's something that we're very keen to get going.
54:02 - 54:21: And I just want to end on this note that this detection work is important; it's essential. And it's going to take multiple approaches, you know, being able to actually stratify patient populations, and reach them, and actually, you know, use these
54:22 - 54:40: approaches to really get accurate imaging readouts, and, you know, it speaks to the work that, you know, Laurie's doing and many other people are doing to really identify people who have higher risk.
54:40 - 55:02: Identify early lesions that have the potential to really morph into invasive PDAC, or even use novel serological tests that you're going to hear about, hopefully, to stratify patients and really get a patient population that can be used to then perform
55:02 - 55:21: more, you know, accurate imaging tests. And so on this note, you know, I want to say that this is what we need to really transform pancreatic cancer from that disease that is like, you know, uniformly lethal to a disease that has a much higher survival rate,
55:21 - 55:37: without changing anything on the therapeutic side. So thank you very much for listening, and I just want to thank my lab, and obviously the tourism lab, where I did most of this work. I want to thank my clinical collaborators, including Laura, who you've heard from,
55:37 - 55:47: and also the Lewis lab, the Yale lab, and the Lyon's lab at Castman Carver. So on that note, thank you very much, and I'll take questions.
55:47 - 55:49: Thank you, Toby.
55:49 - 56:01: Very compelling data; less compelling is the story about the TOBi-bodies being called TOBi-bodies just by coincidence.
56:01 - 56:13: So we have already a couple of questions in the chat. So you mentioned that TOBi 85 and TOBi 89 might also be binding to high-grade lesions.
56:13 - 56:25: So did you look at high-grade IPMN or PanIN3, and did you look at how the binding of these TOBi-bodies changes with pancreatic progression?
56:25 - 56:27: Yeah.
56:27 - 56:31: We looked at that, and it's data we're analyzing now.
56:31 - 56:38: So, we see it in high-grade lesions—not in all of them, and some of.
56:38 - 56:50: And, you know, we've looked at IPMNs, and we've looked at, you know, MCNs, we've looked at multiple lesions.
56:50 - 57:00: And very even PanINs, we've looked at PanINs. You don't get a lot of high-grade PanINs that we see.
57:00 - 57:11: We see a lot of low-grade PanINs, but then you don't see binding to them. So we don't see binding to any of the low-grade PanINs.
57:11 - 57:28: We see binding to some high-grade PanINs but not all of them, and so we're still trying to, you know, get the full data and actually analyze it in a way where we can, you know, give it one shot summary of that data, but it's very clear that it's marking
57:28 - 57:33: certain high-grade lesions but not others.
57:33 - 57:53: And so I'm kind of also rephrasing one of the questions here: can you elaborate a little bit about the challenges of developing a serological assay using these TOBi-bodies, especially considering their binding to conformational epitopes rather
57:53 - 57:55: than just marking protein abundance?
57:55 - 58:03: Yes, I think that last part is really what does it. So,
58:03 - 58:07: these epitopes are.
58:07 - 58:20: So, so these epitopes are conformational, and the conformation I have to is determined by a disulfide bond, right? And then the, and it's glycan-dependent.
58:20 - 58:38: So the question is, how stable are these epitopes, particularly in the ways the samples are stored at the moment? So, you know, we've been trying to develop these assays, and what we see when we use fresh samples, we see nice signals when we use samples that have
58:38 - 59:08: been stored for like ages, then the signal is not so great. And so how can we reall address this and test it, in a very robust way if you don’t have tones of samples to really test this on. But if you’re using archival samples then we don’t know how stable those epitopes are so we’re using mouse models to answer those questions
59:08 - 59:22: But the other important things is actually that it turns out that when you run a lot of these assays, you need two antibodies—one that captures it and one that detects it, right?
59:22 - 59:39: And so, it turns out that when you look at some of the commercial antibodies, they actually bind to different epitopes. So, in some studies have tried to do this work, and they see that it's not working
59:39 - 59:59: very well, but it's because they're using antibodies that bind to unique episodes, actually. So, what we've learned from this is that the epitopes matter so much more than just the protein, and we're trying to figure out how to best go forward with this without compromising, you know, the accuracy of the assay.
60:00 - 60:15: So do you envision perhaps for the serological assays combining multiple TOBi-bodies and other antibodies available in kind of like a mix to enrich for conformational epitopes that are indeed from the cancer?
60:15 - 60:28: Yeah, that's what we're trying to do. We're trying to essentially test, you know, which TOBi-bodies are compatible together and which ones are able to selectively detect pancreatic cancer.
60:29 - 60:42: A question in the chat is whether you have an idea what cancer changes or genetic mutation causes the expression of these specific epitopes.
60:43 - 60:45: Fantastic question.
60:45 - 61:02: So it turns out that, you know, the reason why I got really interested in epitopes is that, you know, I thought, well, you know, this would really read out for changes in the state of the cells and not necessarily, it might not have to do with a particular
61:03 - 61:13: mutation. And it turns out that when we look at the SLC6A20, it's not mutated, right? But when we found out that this is glycan-dependent.
61:16 - 61:30: And this now drives the thought that, well, you know, glycosylation is something that we know is dysregulated in cancers, and glycosylation is not templated, which means, you know, it's not just based on a template where you can read a sequence
61:30 - 61:44: of glycosylation mark; glycosylation instead is based on the state of the cell, and it's determined by, you know, the substrates, the levels of the glycosyltransferases, and so much more.
61:44 - 61:50: So, this is sort of a readout for the state of the cell.
61:50 - 61:58: And that's what we think has happened here; so it's not so much of a specific mutation driving this.
61:58 - 62:07: But it's something that has to do more with the expression of certain glycosyltransferases in the cell.
62:07 - 62:25: Perhaps the last question is whether you can speculate about the resolution for the imaging studies and whether you will be able to detect micrometastases or even going at the single-cell level with these antibodies.
62:25 - 62:28: That's a terrific question.
62:29 - 62:48: We don't know, you know, precisely the resolution at the moment, but we know that it detects less than one millimeter lesions—that's all we know at the moment. But in the future, we will do those experiments, particularly just even in the mouse,
62:48 - 62:58: and ask the question, you know, if you see the limited dilution series of cells.
62:58 - 63:04: You know, what do you get? You know, are you able to—when are you able to see this?
63:04 - 63:10: And when can you really start to see the signal with these antibodies?
63:10 - 63:19: So that's a critical part that actually, you know, we'll be interested to do in the next couple of years.
63:19 - 63:20: Great.
63:20 - 63:26: Thank you again, Toby. We will get back to you during the panel discussion and address the other questions.
63:26 - 63:29: So let's move on to the next speaker.
63:29 - 64:05: Dr. Scott Ferguson. Dr. Ferguson received his PhD from the University of Buffalo. In 2019, he joined Dr. Weiss's group in the Harvard MGH center. As a postdoctural fellow he developed single extracellular vesicles analytics for early cancer detection. He continues now with this line of research in an early cancer detection company. Today he’s going to talk about Single-EV analysis of mutated proteins allows detection of stage 1 pancreatic cancer. Thanks Scott, the floor is yours.
64:05 - 64:07: Thank you for that introduction.
64:07 - 64:12: So, my first slide here is a slide I think Toby also was just sharing.
64:12 - 64:29: I'm really driving home the point of, if we are able to detect pancreatic cancer earlier, our ability to potentially save lives is greatly improved and one other little piece of information I've overlaid here is just unfortunately how many of the
64:29 - 64:47: diagnoses currently are occurring in the later stages of disease. So people who have already advanced to a Stage 3 or Stage 4—these are about 80% of diagnoses, whereas localized Stage 1 where maybe we can do a curative surgical resection is probably
64:47 - 64:58: less than one in five patients. So if we could even just flip that ratio around, the number of lives we'd be saving would be dramatic.
64:58 - 65:10: So, we know that in order to save lives, we have to detect pancreatic cancer earlier. But how, how much earlier, and how can we even detect such small tumors?
65:10 - 65:19: So these were the questions in our minds, me and my advisor Dr. Weiss, when I started my postdoc three years ago.
65:19 - 65:34: And the answer that we thought would maybe be the most promising answer is extracellular vesicles as the biomarker of choice, and specifically interrogating these extracellular vesicles at the single vesicle level.
65:34 - 65:47: In order to gain as much sensitivity as possible over some of the other analytical assays that are available for looking at these EVs as a liquid biopsy or a biomarker tool.
65:47 - 66:03: So, an overview of this talk. First of all, why do we think extracellular vesicles, and more specifically single extracellular vesicle approaches, are most promising for early cancer detection and specifically in the context of pancreatic cancer?
66:03 - 66:16: So we have a few slides with the theoretical underpinnings and modeling work that we did to sort of establish feasibility here before we moved into looking at some of these clinical samples that we had available to us.
66:16 - 66:36: So, I won't go too deep into the math, but I will provide a little bit of that work, and then troubleshooting and method development, and then some of our results of using these new methods in order to detect Stage 1 pancreatic cancer, and then I'll finish with future directions.
66:36 - 66:59: So, the first question of why extracellular vesicle analysis for early cancer detection. So if you're unfamiliar with extracellular vesicles or exosomes, these are small 100-nanometer-sized vesicles that cells, all cells, but especially cancer cells, are releasing tons of these into the body.
67:00 - 67:17: They're used to communicate between cells, but then as a function of that, they also contain a lot of information about the cell that produced them. So they have different proteins, different RNAs, and different molecular cargo and payloads that can represent the status of the cell that is making them.
67:17 - 67:33: So cancer cells may harbor mutated proteins like the ones that we've been hearing about, like a mutated KRAS. Some of those mutated KRAS proteins might be shuttled into those extracellular vesicles, and then when the cells release these vesicles and they end up in the bloodstream,
67:33 - 67:50: our question is can we look at them, find them in the blood, and detect on them the presence of mutated proteins that would be indicative of cancer or potentially precancerous, you know, driver mutations that warrant further investigation.
67:51 - 68:09: So these vesicles, as compared to other potential liquid biopsy markers, we think have a number of advantages. So, for example, circulating cancer cells, especially at early times, are very rare. They're unlikely to be present in the blood at a level which we can actually detect.
68:09 - 68:27: And similarly, with cell-free DNA, usually the DNA also, as Laura had mentioned, doesn't really get into the bloodstream, especially if it's not an invasive cancer and especially if there's no necrotic regions or areas where, you know, the cells are dying to release the DNA into the blood.
68:28 - 68:39: Whereas EVs, on the other hand, are produced in quite great quantity—where an individual cell is releasing hundreds of these vesicles every day.
68:40 - 68:54: So we can, you know, do an assay where we try and collect all of these EVs and analyze them in bulk, or we could try to look at them individually under a microscope, and we think that that latter approach is a lot more promising.
68:55 - 69:16: So if you think, for example, about a bulk approach where maybe your antibody has very high specificity for your protein of interest, but also maybe a very low level of off-target binding to a really highly abundant protein that might also be in the blood, this drives up the level of noise in your assay really considerably.
69:17 - 69:23: So if you're sort of looking for that needle in a haystack, you probably won't find it with this bulk approach.
69:25 - 69:33: Whereas if you are able to sort of separate everything out and see things one by one, maybe you can actually see that signal still.
69:34 - 69:56: Also, if you are using multiple antibodies and you can find, you know, the presence of potentially a couple of cancer markers on one single EV, the coincidence of those two things together can also drive up the specificity and confidence you have in making certain diagnoses or prognoses for patients.
69:57 - 70:14: So just to sort of illustrate what that last slide was trying to convey: here on the left, I don't know if anyone's ever done one of those color blindness tests, but in the healthy, it's a totally grayed out, everything centered, you know, red, green, blue, gray box.
70:15 - 70:23: And then on the right, for a bulk assay, there's been one little click to the right for red and green.
70:24 - 70:28: And I don't know if you can see the difference, but there are actually two different colors there.
70:29 - 70:43: Whereas if we break those boxes down into 130 little boxes and do the same percentage shift and just make one of them red and one of them green, I think it really can jump out at us what we're looking at and able to see.
70:44 - 70:53: So this is sort of the basic idea behind looking at these vesicles at the molecular level.
70:54 - 71:05: So now I will talk about some of the math that we did to give ourselves some confidence that actually we would be able to see these vesicles in circulation.
71:06 - 71:14: So if we can take a blood sample and, you know, look for the vesicles and look for potentially tumor-associated vesicles, are they really there?
71:15 - 71:16: How much blood would we even need?
71:16 - 71:24: And at what size of a tumor are we able to detect these vesicles that are originating from the tumor cells?
71:26 - 71:32: So in order to do this, we first need to understand sort of the kinetics of those vesicles in the bloodstream.
71:33 - 71:36: How fast do they distribute in the blood?
71:37 - 71:38: And then how long do they stick around in the blood?
71:39 - 71:55: So for this, we do a pharmacokinetic model where we looked at data in the literature where people have sort of just administered exogenously some cancer EVs, and we're able to fit the kinetics of how those stay in the blood pretty well.
71:56 - 72:13: And then the next thing that we have to do in order to connect this all together and see, you know, how many EVs do we expect in the blood as a function of tumor volumes is look in preclinical models where mice have different, you know, sort of spontaneous or implanted tumors.
72:14 - 72:21: And researchers have measured the amounts of vesicles in those animals.
72:22 - 72:38: So we can model the size of the tumor as it grows, as well as the increases in vesicles that are directly a result of those increases in tumor size and then connect these two pieces of information together in order to estimate basically a tumor shed rate.
72:38 - 72:43: So how fast are tumors shedding EVs into circulation?
72:44 - 72:49: And at what size tumor do we have a specific concentration of EVs in the blood?
72:50 - 73:00: One thing that we found from doing this work is that the range in EVs as a function of tumor size is very variable.
73:01 - 73:11: So some of the tumors shed EVs into circulation at a very low rate, whereas others shed the EVs into circulation quite highly.
73:13 - 73:27: So just sort of looking at this, right, so in one breast cancer model, this PYMT, the accumulation of the vesicles per milliliter of blood as a function of tumor volume skyrockets.
73:27 - 73:31: They're releasing tons of vesicles. This would be really promising.
73:32 - 73:44: Whereas some of the other models, such as one of the PDAC models, a KIC spontaneous cancer model in mice, was releasing very few EVs into circulation.
73:44 - 73:47: So we have a pretty huge range here.
73:48 - 73:59: But we are able to then scale these mice shed rates into humans, and then we also can validate these model predictions with what's been observed with clinical data.
74:00 - 74:07: And we were pretty happy to see that the model was aligning well with later stage human disease.
74:08 - 74:16: So these next two figures may be a little dense, so I'll try and walk everyone through briefly with these.
74:17 - 74:30: But the question here is, now that we have maybe a basic understanding of what size tumor correlates with what concentration of these potential biomarkers in blood that have come from a tumor,
74:31 - 74:39: is what is the current detection limit for these assays that we already have for looking at the extracellular vesicles?
74:40 - 74:47: So something like a Western blot or ELISA, where we do a bulk sort of screen for maybe some of the markers on those vesicles,
74:48 - 74:54: is probably only able to detect tumors in the about 10 to 100 centimeter cubed range.
74:54 - 74:59: So these are already quite large tumors—tumors that would be visible by imaging already.
75:00 - 75:06: And therefore, you know, those are not sensitive enough of an assay using extracellular vesicles as a marker.
75:07 - 75:17: However, down toward the bottom of this left-hand panel, there are other newer assays which are highly sensitive for extracellular vesicles.
75:17 - 75:20: That are looking at these vesicles at the single-cell level.
75:21 - 75:27: So C here was the first generation of single exosome or single extracellular vesicle analysis.
75:28 - 75:41: And just by math, the theoretical limit of detection is extremely small tumors down to, you know, 0.0001 centimeters cubed,
75:41 - 75:50: where already the model predicts there's a handful at least of these tumor-originating EVs per each milliliter of blood.
75:51 - 76:00: On the right is another chart which sort of takes into consideration that just because an extracellular vesicle has come from a tumor cell,
76:01 - 76:05: it doesn't mean that it necessarily has a tumor marker on it.
76:06 - 76:12: It might not be distinguishable from other EVs in the blood that are coming from healthy tissue.
76:13 - 76:22: And so, you know, depending on how able we are to discriminate which biomarkers we have available to us to interrogate on these EVs
76:23 - 76:27: also comes into play with how feasible the detection is.
76:28 - 76:34: So the question we still sort of had then is, you know, there's tons of EVs in blood.
76:35 - 76:43: Actually, there's probably something like a billion or more EVs per each milliliter of blood, mostly coming from healthy tissue.
76:44 - 76:50: And at early, early stages, where the tumor is maybe less than the size of a match head,
76:51 - 76:57: tumor EVs are probably outnumbered about a half million to one by healthy EVs.
76:58 - 77:01: And can we really actually detect these in clinical samples?
77:02 - 77:07: And so that was our question. The math seemed to suggest that it was at least feasible.
77:08 - 77:14: And so we tried to develop methods toward being able to do just that.
77:14 - 77:22: One of the reasons, too, we think these blood-based liquid biopsies would be super beneficial is, you know,
77:23 - 77:29: there's going to be a number of challenges still with this line of work, such as, you know, what are the best biomarkers?
77:30 - 77:34: If we can find something in the blood, but we can't even see it by imaging, how does that really help us?
77:35 - 77:40: But in terms of opening up monitoring space to see, you know, if these things are increasing over time
77:41 - 77:46: or if there's, you know, more markers are coming on board over time, we think we actually do have time.
77:47 - 77:52: So most tumors are not actually growing all that fast—or at least all that quickly.
77:53 - 77:57: So for a sort of, you know, slower-growing tumor, something that's doubling every year,
77:58 - 78:06: the red line vertically towards the right-hand side of this graph is sort of the time from, you know,
78:06 - 78:09: initiation of the tumor to when it would be visible by imaging,
78:10 - 78:14: which is it's already been maybe growing or seeded 30-something years ago.
78:15 - 78:24: And the limit of detection from the modeling for the tumor EVs would be about 17 years earlier than what we detect by imaging.
78:25 - 78:35: So this window of potentially years to look at what's happening in someone just by taking blood tests is potentially very promising.
78:37 - 78:44: So moving on to the method that we developed, we take the samples.
78:45 - 78:47: So we take 100 microliters of plasma.
78:48 - 78:52: We concentrate and purify EVs using size exclusion chromatography.
78:53 - 78:58: And we label all of those EVs with a protein-reactive dye so we're able to see all of the vesicles,
78:59 - 79:03: clean those up, and then we also label in solution with fluorescent antibodies.
79:03 - 79:12: Mostly this work is using fluorescent antibodies directed against cancer driver mutation proteins such as KRAS G12D,
79:13 - 79:17: where we have specific antibodies and mutant p53,
79:18 - 79:22: where there are now available specific antibodies that detect just those mutant forms of the protein
79:23 - 79:26: and do not detect the wild-type forms of the protein.
79:27 - 79:35: Then we clean these again on another size exclusion column so that we can remove all of the nonspecific signal from the unbound antibodies,
79:36 - 79:39: attach the EVs onto glass slides,
79:40 - 79:48: and use standard fluorescent microscopy to interrogate the percent of EVs that are positive for these cancer markers.
79:50 - 79:53: So a little overview of how this looks.
79:53 - 79:56: On the left-hand side here, you see some green circles.
79:57 - 80:00: So these are just all of the EVs coming from a sample.
80:00 - 80:19: They stain very positively with this TFP-reactive dye. And then we can look in other channels for some of these cancer proteins and quite clearly see which of them are positive and which of them are negative for those specific proteins.
80:19 - 80:42: And so this was really exciting for us; we had never really before been able to detect KRAS G12D in EV samples, even though we knew they'd come from, say for example, a patient who had a G12D mutation.
80:42 - 80:48: Being able to see them in the single EVs, at least in cell lines, we were excited to move forward into some of our clinical samples.
80:48 - 81:03: So, getting back to an earlier point I was making, we also looked not just at a bunch of our pancreatic cancer cell lines, but also some patient-derived xenograft cell lines that Mass General Hospital had in its cell bank,
81:04 - 81:17: to see what percent of them had at least these panel of cancer markers that we were using in this early study. And in this case, about half of all of those vesicles that came from the cells were marker negative.
81:17 - 81:31: So right off the bat, there's a lot of room here, I think, for coming up with new early cancer markers that could hopefully cover more of the EVs. We're already, you know, not able to detect at least half of these.
81:32 - 81:55: But when we move into real clinical samples, first of all, we started off with later stage disease, and we see, compared to our healthy controls, that the percent of EVs in late-stage PDAC that are standing for mutant p53 or a mutant KRAS are, you know, well-separated over our control samples.
81:55 - 82:24: This is also true when we included other markers like MUC1 or EGFR. And then, nicely, we see pretty good specificity. So when we had the next-gen sequencing data to be able to separate, if we knew it had, say, for example, a G12D mutation or no KRAS mutation,
82:26 - 82:49: we're also able to see the specificity in looking at the percent of EVs that have those, and the blood is also agreeing with those NGS results.
82:49 - 83:05: And so sort of the final exciting data that we had is we moved into our early-stage one PDAC samples. So just to sort of underscore how hard it is in clinic to detect these really early-stage one PDACs,
83:05 - 83:15: in our bank of samples, we had over 3,000 samples.
83:16 - 83:26: And these are the only 16 samples in 3,000 that were able to be diagnosed at Stage 1 at MGH. So with that in mind, it was really exciting that we saw in 15 of 16 of these, just by looking for mutant KRAS or mutant p53,
83:26 - 83:39: we were able to separate them out from our healthy controls.
83:39 - 83:43: So furthermore, the smallest detected sample that was positive corresponded to a tumor volume of just 14 millimeters cubed.
83:43 - 83:53: So we then had, you know, some pretty good estimations of our limit of detection and some of the variability of healthy controls, and we ran simulations with that information.
83:53 - 84:09: And that supported, just with the method that we were able to develop over the course of my postdoc, that about 70% of PDAC tumors would be detectable.
84:10 - 84:18: And then for a tumor doubling every six months, this translates to a window of about six years earlier detection than what would be realized by normal imaging that's being used currently.
84:19 - 84:25: So some of the questions we still have in mind here are, you know, is this really a clinically feasible approach?
84:26 - 84:32: So is size exclusion chromatography, especially when we're doing this two times over, going to be feasible?
84:33 - 84:40: Or should we, you know, start to figure out more clinically relevant, faster isolation protocols?
84:41 - 84:46: And then the main thing will be how many additional markers and what kind of multiplexing could be available towards this to increase our detection power.
84:47 - 85:06: And then I think eventually these kinds of blood tests can probably be used in combination with other blood-based early detection strategies that a lot of companies are moving towards, such as, you know, multi-omic next-gen sequencing and proteomic studies.
85:07 - 85:12: So I thank everyone for their attention and happy to take questions.
85:13 - 85:15: Thank you very much, Scott.
85:16 - 85:18: I'm going to start with the questions.
85:19 - 85:39: So the first question is, so obviously working on pancreatic cancer, I was intrigued in looking at the two models that you showed, the KPC and KIC, and the fact that they had pretty different levels of shedding in terms of EVs as per tumor burden.
85:40 - 85:59: And I was wondering whether you can speculate on the reason for these differences, and also whether you looked at subgroups of EVs, because my understanding is that not all EVs have all of these cancer markers, and whether you looked at differences in abundance of these different subsets.
86:00 - 86:02: So yeah, very great questions.
86:03 - 86:16: In terms of the reasons for variability in the EV shed rate, I think it's probably foremost related to actually just the intrinsic EV production and shedding of the different tumor cells.
86:17 - 86:21: We know that KRAS mutation, for example, does drive up EV production.
86:22 - 86:31: So as soon as there's a KRAS mutation on board, there's about a twofold increase, at least, associated with that, just in the amount of EVs being produced.
86:33 - 86:38: And then just across all sorts of different cancer types, this can vary about 100-fold.
86:38 - 86:41: So we're not totally sure all of the reasons for that.
86:42 - 87:04: But then I'm sure there's also probably other things that haven't been explored yet that relate to how the vasculature is related to the tumor and how much interface they have there, because a lot of EVs are probably just ending up sort of interstitial and not even extravasating into the blood flow, which is what we're looking at.
87:05 - 87:14: And then did you look at particular subsets of these EVs in terms of this shedding, or you were looking at the overall population?
87:15 - 87:26: So here, for those models that we were able to use to drive some of our feasibility, we didn't have any of that available to us to ask that question.
87:26 - 87:34: It would be a really interesting question for my end as well, to understand which EV populations are actually ending up in blood.
87:35 - 87:45: And then you talked about stage one pancreatic cancer, and we will get back to that during the panel session because I have some questions that overlap for all the speakers.
87:46 - 88:00: But are you planning to look perhaps at patients that had a resection and whether you could use this strategy to follow metastatic burden and recurrence?
88:01 - 88:08: Yeah, so I think we do have a lot of that clinical samples available to us at MGH for our collaborators.
88:09 - 88:15: And I do think that those are sort of other future directions that we didn't get into just yet.
88:15 - 88:26: But yeah, understanding also what others have been pointing out, you know, the IPMNs or these other precancerous things that can lead to PDAC.
88:27 - 88:35: You know, how do we, can we detect those? Do we have biomarkers for those that are also on the EVs that can maybe end up in blood?
88:36 - 88:41: I think it's a pretty tricky question right now because it's a little bit of a black box inside of the body.
88:42 - 88:48: But as we've heard from the first two speakers, there's a lot of really great work going into having, you know, answers to those questions.
88:49 - 89:02: And I think once that really is known, plugging that into sort of an EV or blood-based test is going to hopefully really accelerate the ability to detect earlier cancers.
89:03 - 89:09: So one question coming through in the chat is whether you also looked at patients with pancreatitis.
89:11 - 89:22: So it's interesting, we did actually look not only with like younger healthy controls, we had another separate section where we did age-matched healthy controls and pancreatitis.
89:23 - 89:39: And I will say it will become very important to also include additional biomarkers, not only mutated proteins, because by the time we started to look at blood samples from like 50, 60-year-old patients or patients with pancreatitis,
89:40 - 89:50: many of them have also KRAS mutations and the ability then to say, okay, so this is a KRAS mutant positive blood sample, but is it actually cancerous?
89:50 - 89:53: I think that, you know, is going to be the hard work moving forward.
89:54 - 89:58: Great. Thank you very much. We will get back to you shortly.
89:59 - 90:10: Before our panel discussion, we are going to have a short talk by Dr. Mike Prater, who is a Senior Commercial Strategist for Oncology at Abcam based here in Cambridge
90:11 - 90:15: and oversees global collaborations and engages with oncology researchers to address their needs.
90:16 - 90:20: Mike, thanks for organizing this webinar. Please go ahead.
90:21 - 90:37: Great. Thank you, Julia. I'm going to spend a few minutes highlighting how our reagents are supporting early detection research and introduce our multiplex immunoassay FirePlex that allows sensitive detection of microRNAs and secreted protein biomarkers.
90:38 - 90:51: So, the main focus of, the main goal of my role is to ensure that oncology researchers have the tools that they need to advance oncology research, both now and in the future.
90:52 - 91:05: And a key part of that is tracking emerging trends and targets within the field, but also importantly, speaking to researchers to understand common pain points and needs that we can help address.
91:05 - 91:26: All of those insights are then fed into our development pipelines, and we're also establishing collaborations and partnerships with researchers in both academia and industry to integrate our reagents into novel technologies and into diagnostic assays.
91:27 - 91:50: Abcam is well known for recombinant antibodies, but we're also investing heavily in in-house production of other product lines, such as bioactive proteins, knockout cell lines, and also ELISA kits, both singleplex and multiplex.
91:51 - 92:07: Within the early detection space, our reagents are supporting research into the early stages of cancer initiation and transformation, and the key targets involved in that process.
92:08 - 92:23: In the tumor exosome space, our antibody pairs are being used to capture and detect tumor-specific markers on exosomes and EVs, as Scott demonstrated beautifully in his talk.
92:24 - 92:35: Similarly, our recombinant antibodies are being used to detect CTCs using tumor-specific biomarkers, such as antigen receptor variant 7.
92:36 - 92:50: I'm going to spend a short while just introducing our multiplex immunoassay technology called FirePlex that can be used to assay both microRNAs and serum protein biomarkers.
92:51 - 93:04: In the paper I've highlighted here, the authors combined a traditional serum biomarker, CA19-9, together with three microRNA biomarkers.
93:05 - 93:16: When those combined with CA19-9 gave a more sensitive and specific test to detect early-stage pancreatic cancer.
93:17 - 93:22: The FirePlex assay comes in a range of different formats.
93:23 - 93:28: There's the microRNA format and the protein immunoassay.
93:29 - 93:47: It's available in pre-built off-the-shelf panels with established, published microRNAs and secreted protein biomarkers together in the same panel to things like inflammatory pathways, cytokines, chemokines.
93:48 - 93:59: There are high-throughput plate options available, and there's also a bespoke service available for building bespoke panels with novel targets.
94:00 - 94:06: There are also separate formats for flow cytometers and high-content images.
94:07 - 94:12: The technology itself is a fluorescence-based multiplexing approach.
94:12 - 94:16: The inner label is used to quantify the analytes.
94:17 - 94:27: On the end of the hydrogel particle, there's a unique fluorescent barcode, which is unique to each analyte detected by the antibody.
94:28 - 94:35: And in this way, multiple analytes can be detected simultaneously from very low sample volumes.
94:35 - 94:56: Each antibody pair that's used within this assay is validated heavily for specificity, that the signature is specific to each analyte being detected, and for sensitivity, having a linear dynamic range.
94:59 - 95:14: The microRNA assay can be used for a wide range of biosamples. It could also be used for tissue samples as well to compare and correlate microRNA biomarkers between tissue and fluid samples.
95:16 - 95:42: I'd like to highlight a talk that's available on our website by Dr. Rüdiger Greinert, who used our microRNA FirePlex assay to detect microRNA biomarkers associated with the very early stages of melanoma transformation, and also used the assay in a clinical setting to detect microRNA biomarkers that predict whether patients will respond to immunotherapy.
95:44 - 96:06: So as well as supporting fundamental research, increasingly, our recombinant monoclonal antibodies are being used by partner companies within diagnostic assays in the clinic that are being used to directly benefit patients and help identify patients who will respond to target therapies
96:07 - 96:21: We're really happy to support research projects and to hear from oncology researchers to share insights and recommendations for specific projects, and to help shape our development pipelines and the new products that we're making.
96:22 - 96:29: If that's of interest, please do give me an email at either of the addresses shown here.
96:30 - 96:41: And finally, I'd just like to mention that if you missed the earlier sessions in this series, the first was the epigenetic landscape of AML, and the second was overcoming ovarian cancer resistance.
96:42 - 96:46: Both of these will be available soon on demand on our website.
96:46 - 96:50: So with that, I'll hand back to Giulia to start the panel discussion.
96:53 - 96:55: Thank you very much, Mike.
96:59 - 97:04: So I will ask the speakers to please turn on their videos and unmute themselves.
97:05 - 97:12: Thank you very much and remind everyone that you're welcome to post your questions in the Q&A chat option.
97:13 - 97:16: So I'm going to start with the question for Laura and Toby.
97:17 - 97:33: You envision an analysis of precancerous lesions, potentially combining your approaches looking at genetic profiling and protein analysis with the Tobii bodies.
97:34 - 97:41: Yeah, I mean, I guess I can go first, but I think multi-omic profiling of these precancerous lesions is really the way to go.
97:42 - 97:49: We certainly, you know, my background is in genomics, so we started there, but I think our genomic studies have certainly shown that it's not all mutations.
97:50 - 97:59: Mutations are important, but especially for that transition from low-grade dysplasia to high-grade dysplasia or high-grade to cancer, mutations are not going to be 100% sensitive.
97:59 - 98:03: Like SMAD4 gets some, but you're never going to get all of them with just mutations.
98:04 - 98:09: And you'll need other DNA features, proteomic features, maybe even serum and microenvironment features.
98:10 - 98:18: So I think collaborating with folks with different expertise to profile different analytes in the most state-of-the-art way is going to be the way to go.
98:19 - 98:22: Yeah, I second that.
98:23 - 98:30: Yeah, I think it's not just important to do, I think it's necessary.
98:31 - 98:41: I think that those approaches, combining them, will give us the most accurate sensitivity and specificity, which is quite important, right?
98:42 - 98:57: So, you know, and to talk more about the contrast between genomics and actually looking at other things like proteomics and epitomics and, you know, EV analysis and all that,
98:58 - 99:09: I think this is really important, particularly, you know, when I think about the fact that despite the fact that you have, you know, some different mutations in different patients,
99:09 - 99:17: when we take our Tobii bodies and actually profile all these patients that have different mutations, you find those epitopes there.
99:18 - 99:26: So for some reason, the different mutations are leading to similar state changes that have these epitopes.
99:26 - 99:44: And so not only just the genomic aspect, but the epitomics aspect and even epigenetics and all these other aspects will really give us a clearer way to go about the error detection stuff.
99:45 - 99:52: And for Scott, I was looking at your slide that 50% of EVs do not have cancer markers.
99:52 - 100:00: And I was then, you know, thinking about Tobii's work and whether you, there is a potential combination of your approaches in terms of looking at EVs and Tobii bodies.
100:00 - 100:05: And I would like both of you to comment on that.
100:05 - 100:12: I think, yeah, when I was listening to Tobii's presentation, it was something that I had sort of envisioned earlier on in my postdoc too, which is if we preconceive the marker
100:18 - 100:23: we're going to go for, it's a very sort of biased approach. Whereas if you do sort of an approach where you can, you know, find maybe new antibodies, things that are binding
100:27 - 100:33: to specific epitopes, I think this kind of an approach with the EVs would be fantastic because we have no idea what kind of glycosylation epitopes are on the EVs as well. And I think
100:40 - 100:46: the ability to find highly specific cancer epitopes on EVs is a totally underexplored area.
100:49 - 101:02: Yeah. And to add to that, you know, Ralph's work and work from your lab is something that I've followed really closely since I actually started. An aspect of EVs obviously is also,
101:03 - 101:20: and this also extends to other circulating biomarkers, is that we need to start thinking about where are we sampling these biomarkers from, right? So you can take some from the
101:20 - 101:35: blood, right? Maybe sampling also from the pancreatic fluid would be, the pancreatic juice would be very important. And, you know, maybe we can look at like, you know, feces
101:35 - 101:47: analysis. I mean, we have to, or urine analysis, we have to be comprehensive. And so I think just looking at the blood is easy because there are all these like, you know, IqoVol samples,
101:47 - 102:02: but I think we need to look beyond that. And also, you know, when I also think about EVs and other aspects is that, you know, EVs also contain other markers. So you have micro
102:02 - 102:09: RNAs there and all these other species of macromolecules that we can actually look at. And the last
102:09 - 102:23: part is that obviously, you know, we think about EVs coming from the cancer cells, but EVs also come from other, you know, cells in your body, right? And so by understanding the
102:23 - 102:35: contribution, actually, you know, by understanding the cell surface profiling of the cancer cells and knowing where, like, you know, are these EVs coming from the cancer cells and not stratifying
102:35 - 102:45: by that, you can actually enrich for a population that can make the signal on these assays even better. And so this is some of the ideas that we should start thinking about to make this
102:45 - 102:56: approach better. Yeah. And in fact, following up from that, I was thinking about, I mean, Scott, you talked about looking at stage one pancreatic cancer, but whether you envision
102:56 - 103:09: the possibility to looking at patients with IPMNs or pancreatic cysts and whether you think that either looking in the blood or as Toby was mentioned, looking at directly the cyst fluid,
103:09 - 103:22: you would be able to have the sensitivity to identify what you're looking for in the pre-cancer setting. So yeah, I know some of the work in our lab using other assays,
103:22 - 103:32: sort of more clinically advanced protocols, we're now shifting toward doing different, you know, the pancreatic juice or things that might have more concentrated a signal than the
103:32 - 103:44: blood. So I don't see why we wouldn't try to move toward that as well. And then I think also, you know, if we don't necessarily have great biomarker or perfect biomarker to say,
103:44 - 103:53: you know, at this early stage, this is something that will develop into something that's more aggressive later on. At the very least, we might be able to come up with assays that are sensitive
103:53 - 104:03: enough to look at the dynamics to say, okay, well, we don't know, but this is increasing quite rapidly. What we'll keep looking or, you know, these other sorts of metrics that if you're
104:03 - 104:17: doing longitudinal studies might come out and strike you as important. So I think that's also another avenue. So Laura, you will look, when you look at, do your genetic sequencing analysis
104:17 - 104:28: looking at the cyst fluid, you're looking, you're not arranging for EVs, but would you think that maybe that will give you a, you know, looking for cancer markers and then sequencing those EVs
104:28 - 104:36: will give you a higher noise signal to noise ratio?
104:37 - 104:41: Hey, I honestly don't know the answer to that in cyst fluid. I think Scott probably knows better
104:41 - 104:48: than I do kind of how, what the impact of the EVs would be. I do think that having the kind of
104:48 - 104:59: barcoding error suppression techniques in the sequencing, you can, especially in fresh cyst fluid samples have fairly high confidence in the directly isolated DNA, even mutations
104:59 - 105:04: at very low prevalence. I think to me, the bigger issue as, as I think it was Giulia brought up
105:04 - 105:09: earlier is just the sampling of it. You're, you're dependent on where the gastroenterologist takes
105:09 - 105:14: the needle. And if you don't sample the component with the highest grade, then the best assay in
105:14 - 105:20: the universe is not going to find mutated DNA that's not there. Yeah. So I want, I want to end
105:20 - 105:29: with a few philosophical questions that we had in a previous early detection webinar organized
105:29 - 105:34: by the program. And the first one would be whether there's anything available at the
105:34 - 105:42: moment for high-risk patients with pancreatitis or BRCA mutations and anything available in terms
105:42 - 105:47: of earlier detection or people that had the familiar history of pancreatic cancer.
105:48 - 105:52: Yeah, I mean, I guess I can, I can take this one first. There's, you know, several
105:52 - 106:00: multi-institutional screening studies. There's the CAPS, which is Mike Goggins' study that's
106:00 - 106:06: based at Hopkins, but has other institutions. And Diane Simeone has the, the PRECEDE consortium
106:06 - 106:12: that's focused on, I think mostly with patients with family history. That's, you know, they're,
106:12 - 106:16: they're screening programs, but also obviously gathering clinical data to try to inform future
106:17 - 106:24: recommendations. And they're focused on imaging, you know, EUS as well as CT and MRI, but also
106:24 - 106:32: involve harvesting various biosamples, you know, like blood, like cyst fluid, like pancreatic juice.
106:32 - 106:36: And so I think, so basically the number has moved to the States from the UK.
106:38 - 106:43: I don't, yeah, I, I confess, I don't know the landscape in the UK. I know that CAPS is,
106:43 - 106:47: is multinational. I think there's, there's folks from other countries as well. But,
106:51 - 106:55: but I think one of the bigger challenges is defining exactly who is at risk. Like we know
106:55 - 107:00: the family history question and we know that people with cysts, but the majority of pancreatic
107:00 - 107:08: cancers don't arise in those two groups. And so I think we, the math on population screening,
107:08 - 107:12: I think is very challenging in pancreatic cancer because the incidence is relatively low.
107:12 - 107:19: And the next step is big, right? But even biopsying the pancreas is like a big clinical step. And so
107:19 - 107:24: even a perfect, almost perfect clinical test will give you too many false positives. And so you need
107:24 - 107:31: to enrich the population. And so what do you do beyond cysts, beyond family history? I know
107:32 - 107:36: there's a big study that honor bond mitrans or s-char you're doing with nuance of diabetes.
107:36 - 107:41: So I think defining those higher risk groups is going to be super important.
107:42 - 107:46: Yeah, I have nothing to add to that. I, I was going to mention the onset diabetes and
107:46 - 107:51: just mentioned it. So I think, yeah, finding those high-risk groups and then enriching them for,
107:51 - 107:58: you know, more intense, more accurate tests, then I think that makes sense.
108:00 - 108:10: Great. And then we're going to end with the question, which is how far away are we from being
108:10 - 108:15: able to offer a reliable early test for pancreatic cancer? And what is going to win?
108:15 - 108:20: It's going to be imaging. It's going to be bio. Are you going to hold us to this?
108:20 - 108:24: Well, I'm here to ask the tough questions. If we're wrong, we're going to get fired in five
108:24 - 108:32: years. No, well, you're allowed to, to speculate and you know, make a prediction and maybe we can,
108:33 - 108:36: we can have a little prize for you if you get close to the prediction.
108:40 - 108:42: Realistically, what are we looking at?
108:44 - 108:47: I think it, oh, go ahead, Toby. No, I was just going to say, you know, you would,
108:47 - 108:53: you would have a much better prediction than, than we do. You're more proximal to,
108:53 - 109:00: to the clinical side. So, yeah, I mean, I think, and this is based on, you know, our work on IPMNs
109:00 - 109:06: as well as some of our work in, in PANNs, the precursors are so super common that we're never
109:06 - 109:11: going to be able to use KRAS mutation alone, because I think everybody's walking around with
109:11 - 109:16: 50 PANNs at least, you know, and they're like crazy, even young people, they're just so common.
109:17 - 109:22: So I think what we really need are biomarkers for pre-cancers that are likely to progress. And
109:22 - 109:26: whether that's morphologic high-grade dysplasia, or whether that's some other marker that's not
109:26 - 109:30: related to morphology, that is what we need. And I think mutations will be part of it,
109:30 - 109:37: because SMAD4 mutation, P53 mutation are fairly specific for progression, but their sensitivity
109:37 - 109:44: stinks. And so I think it's never going to be only DNA. And so we're going to need
109:44 - 109:47: other complementary markers, whether it's Tobii bodies, whether it's EVs,
109:47 - 109:55: especially if we're detecting it in more, and I want to say like proximate samples,
109:55 - 110:00: like if you're doing cyst fluid, if you're doing juice, then KRAS mutation is never going to work.
110:00 - 110:03: But if you're doing blood, maybe KRAS mutation will work, because you shouldn't have
110:03 - 110:09: mutations in your blood if you don't have an invasive cancer. So it's possible that
110:09 - 110:14: combining more proximate analyses with things in the blood could also be helpful.
110:15 - 110:19: Much as I want to say it's going to be a mutation, it's not. It's not going to be
110:19- 110:22: only mutation, it's going to need to be multi-dimensional, I think.
110:23 - 110:27: Maybe it's going to be Tobii bodies and sequencing and EVs all together.
110:27 - 110:36: Yeah, I really think it cannot be just mutations, because we know just beyond Lara's work,
110:36 - 110:45: that you have this clonal hematopoiesis that happens, and even in the blood you do see
110:45 - 110:54: cells that are mutant, and they might have, you know, maybe not the pancreatic cancer
110:55 - 111:00: mutations, but you know some of them do have KRAS mutation. But these mutations, you know,
111:00 - 111:07: happen as you grow older, and so I just don't think that will be specific enough.
111:07 - 111:17: So I think, you know, as Lara said, it will be a combination of approaches, and it's hard to say
111:17 - 111:24: exactly what platform it will be, but for sure, you know, you're going to have to look at a protein
111:25 - 111:31: kind of marker, at least to enrich, or the enrichment step, I think it's so important.
111:31 - 111:36: And when it comes to imaging, you know, obviously they're just like, you know,
111:36 - 111:46: you can use EUS, but then that doesn't help you see, you know, maps as much. And so I really think
111:46 - 111:52: in terms of the nitty-gritty, like let's really look at what's going on, I think,
111:52 - 111:58: you know, even a PET image might really come to the fore. I might be wrong about that,
111:58 - 112:05: but I think once we get good reagents that can detect epitopes in a really specific way, I think
112:05 - 112:11: there's no reason why that can't come to the fore. But I think that happens only after you've
112:11 - 112:19: enriched for populations. Yeah, I really like the way Toby just framed that, which is that
112:19 - 112:23: there's almost two questions. There's one is how do you enrich your population, and then two is
112:23 - 112:28: how do you detect in that population? And currently our strategies for enriching are
112:28 - 112:32: solely clinical, right? It's do you have a family history? Do you have a cyst? Do you have
112:32 - 112:39: nuance of diabetes? But maybe we need fancier molecular ways to enrich that population.
112:39 - 112:43: And I think that's something that I think we're, all of us are focusing a lot on detecting the
112:43 - 112:48: cancers, but maybe we also need to shift to say what are the things that make you higher risk,
112:48 - 112:53: and can we define our screening population there? I think the other point that Toby made that I want
112:53 - 112:59: to echo is that imaging is going to be super important. I'm a molecular biologist. I don't
112:59 - 113:05: really understand how CTs work, but I know there's amazing work in particular from Elliot Fishman
113:05 - 113:14: here at Hopkins in doing cinematic rendering of CTs. He partnered with movie people in California,
113:14 - 113:18: and I forget if it was Pixar or who it was, I apologize, but they make these gorgeous renderings
113:18 - 113:23: of the CT scans, and you can just see so much more. And I think even applying AI to those,
113:23 - 113:28: I think there's a lot of, you know, opportunities for improvement of the scans we have. And then,
113:28 - 113:34: as Toby mentioned, adding in kind of targeted agents that are based on the molecular differences
113:34 - 113:38: between tumor cells and normal cells, I think that will almost certainly be an important piece.
113:40 - 113:47: Brilliant. Great conclusion to our virtual conference. Thanks to the speakers, to Mike
113:47 - 113:57: Prater, and to our patients, and thank you everyone for joining.

Earlier detection of pancreatic cancer

On-demand webinar

Summary:

Moderator and Speakers:

Video Transcript

You may be interested in...