3 of 40

Precision Oncology and Interception is a continuous process throughout a patient’s oncologic care

MSK Confidential — do not distribute

Patient with Cancer

Presurgical Decision-making

Surgical Management

Predictive and Prognostic Treatment Implications

Oncologic Treatment

Cancer Prevention and Surveillance

Surveillance & Survivorship

Genetic Risk Assessment

Family Risk Assessment & Surveillance

4 of 40

Most (not all) Hereditary Cancer predispositions are caused by germline mutations in tumor suppressor genes

Graphic modified from http://cisncancer.org/research/what_we_know/advances/oncogenes.html

MSK Confidential — do not distribute

Brake: tumor suppressor

Gas: Oncogene

Foot on the gas and no brakes

Cancer

Normal genes regulate cell growth, death

First mutation (susceptible, risky)

Second mutation or loss (leads to cancer)

Multiple pathways to DNA damage

We have many systems in place to prevent and/or fix these DNA errors: Homologous Recombination, Base Excision Repair, Nucleotide Excision Repair, Mismatch Repair

5 of 40

Li Fraumeni Syndrome is caused by germline mutations in the TP53 tumor suppressor gene

Graphic Created with BioRender

MSK Confidential — do not distribute

Autosomal Dominant Syndrome: TP53 germline PVs

Diverse risk of cancers in both childhood and adulthood

Core cancers: sarcomas, brain tumors, adrenocortical carcinoma, breast cancer

>70% lifetime risk of cancer in men and >90% in women

6 of 40

Li Fraumeni Syndrome Management

Villani et al. Lancet Oncol 2011

Villani et al. Lancet Oncol 2016

MSK Confidential — do not distribute

Toronto Protocol:

WB-MRI, MR breast, MR brain, abdominal US, biochemical screening (age-dependent)
Initial study looked at 33 TP53 mutation carriers

7% of patients diagnosed with new cancer

Early detection of tumors resulted in improved long-term survival

Graphic Created with BioRender

7 of 40

Li Fraumeni Syndrome vs TP53-related cancer syndrome: Is this a spectrum rather than a single syndrome?

Kratz et al. JAMA Onc 2021

MSK Confidential — do not distribute

Assessed over 3k persons from 1282 families reported in the literature
Evaluated genotype-phenotype correlations by those meeting LFS testing criteria vs those that did not

Hotspot variants found in both groups
Underlined variants are unique to each group

8 of 40

The 5-year CATCH Clinical Experience with TP53: Key Points

Unpublished Data. Do not copy or distribute

MSK Confidential — do not distribute

52% of patients had at least one cancer prior to CATCH enrollment

8% of patients overall were diagnosed with a new cancer via CATCH (similar to Villani et al data)

10% of patients without prior cancer were diagnosed with 1^st cancer via CATCH surveillance: all were diagnosed with early-stage disease that did not require systemic treatment

9 of 40

Most (not all) Hereditary Cancer predispositions are caused by germline mutations in tumor suppressor genes

Graphic modified from http://cisncancer.org/research/what_we_know/advances/oncogenes.html

MSK Confidential — do not distribute

Brake: tumor suppressor

Gas: Oncogene

Foot on the gas and no brakes

Cancer

Normal genes regulate cell growth, death

First mutation (susceptible, risky)

Second mutation or loss (leads to cancer)

Multiple pathways to DNA damage

We have many systems in place to prevent and/or fix these DNA errors: Homologous Recombination, Base Excision Repair, Nucleotide Excision Repair, Mismatch Repair

10 of 40

MMRD and Microsatellite instability (MSI) are hallmarks of LS-associated cancers

Graphic Created in BioRender

MSK Confidential — do not distribute

PCR-based MSI testing and IHC MMR been used for decades as screening for Lynch Syndrome in colorectal and endometrial cancer

11 of 40

MMRD predicts response of solid tumors to PD-1 blockade

Le et al. Science 2017

MSK Confidential — do not distribute

86 patients with 12 different tumor types

48% of patients had Lynch Syndrome

All had prior treatment with POD prior to study enrollment

Tumors determined to be MMRD by IHC or MSI PCR

Objective radiographic response in 53%, complete response in 21%

12 of 40

MSK-IMPACT™ : Integrated Mutation Profiling of Actionable Cancer Targets

Cheng, et al J Mol Diagn. 2015; Middha et al., JCO PO 2017

MSK Confidential — do not distribute

MSISensor: clinically-validated bioinformatics algorithm for MSI analysis that reports percentage of unstable loci as a cumulative score with >96% sensitivity in a pan-cancer cohort

13 of 40

Research Informs more research

MSK Confidential — do not distribute

Patients

Research

Questions:

What is the MSI and underlying LS prevalence in all patients with MSK-IMPACT somatic profiling?
Can we characterize germline results into causative vs. incidental by MSI status?

14 of 40

MSI is Associated with Lynch Syndrome Pan-Cancer

Latham, Srinivasan et al., JCO 2019

MSK Confidential — do not distribute

Assessed >15K solid tumors for MSI status covering >50 cancer subtypes

Determined underlying LS status across the pan-cancer cohort

**50% of LS patients with MSI-H/I tumors had cancers other than CRC/EC

15 of 40

MSI is Associated with Lynch Syndrome Pan-Cancer

Latham, Srinivasan et al., JCO 2019

MSK Confidential — do not distribute

Assessed tumor mutational signatures in all LS patients

In pts with MSI tumors, MMR-D tumor signature was present in 88% of tumors

In pts with MSS tumors, 89% did NOT have an MMR-D tumor signature

16 of 40

MSI is Associated with Lynch Syndrome Pan-Cancer

Latham, Srinivasan et al., JCO 2019

MSK Confidential — do not distribute

Majority of germline mutations in the MSI groups were in MLH1/MSH2

Majority in MSS group were found in PMS2/MSH6

= prevalence of LS in general population

17 of 40

Research informs patient care

MSK Confidential — do not distribute

Research

Patients

Lynch Syndrome is a Pan-cancer syndrome

An MSI-H tumor, regardless of tumor origin should prompt consideration for underlying Lynch Syndrome

18 of 40

Lynch Syndrome Surveillance and Risk-reduction Recommendations

Adapted from NCCN v2. 2023

MSK Confidential — do not distribute

Cancer Type	Screening/Risk-reduction
Colorectal	Colonoscopy every 1-2 yr
Endometrial	EMB every 1-2 years; Surgery: Hys/BSO
Ovarian	Surgery: Hys/BSO
Renal pelvis/ureter	Urinalysis every year RUS if FH
Gastric	EGD every 2-4 years
Small bowel	EGD every 2-4 years (CTE/MRE in high risk)
Pancreas	Only if FH, MRI/EGD research setting

colonoscopy is the only proven effective cancer screening for Lynch Syndrome patients despite this being a pan-cancer syndrome

19 of 40

Patient adherence to surveillance and risk-reduction strategies varies

Mittendorf et al., Hered Cancer Clin Pract. 2019

Stoffel et al. Am J Gastroentol 2010

MSK Confidential — do not distribute

Per patient adherence over observation period (%)

Recommended Procedure

Mittendorf et al. 2019: adherence to colonoscopy recommendations 81%. Adherence to other recommendations varied significantly. No difference when comparing patients with prior cancer history to those without

Stoffel et al. 2010: Personal and/or family history of CRC improved colonoscopy adherence (85% vs. 60%)

20 of 40

Patient adherence to surveillance and risk-reduction strategies varies: RR Hysterectomy uptake is low

. Seppala et al., Eur J Cancer 2021

MSK Confidential — do not distribute

Data from the PLSD: Risk-reducing uptake was low and varied by causative MMR gene

21 of 40

Can we further improve surveillance protocols for hereditary cancer syndromes with cfDNA assays?

MSK Confidential — do not distribute

22 of 40

Circulating tumor DNA is cancer DNA that is shed into peripheral circulation

Images adapted and modified: https://www.bcgsc.ca/news/genome-sequencing-helps-prioritize-cancer-treatment-options & https://www.diacarta.com/products/direct-cfdna-test

MSK Confidential — do not distribute

Genetic alterations from tumor are shed into periphery = ctDNA

These genetic alterations can be detected in blood

23 of 40

There are many commercially-available cfDNA tests that assess different genetic changes

MSK Confidential — do not distribute

In cancer patients: monitor for treatment response, assess for cancer recurrence

Different assays may be better suited for different uses

M. Cisneros-Villanueva et al., 2021

24 of 40

Key Studies for MCD testing in cancer-unaffected pts: PATHFINDER

MSK Confidential — do not distribute

Schrag et al., 2023

Prospective study of pts >50 years old without known cancer

Enrolled >6K participants in one year for cfDNA collection using a methylation assay

Primary outcome: time to/extent of diagnostic testing required

Sensitivity: 29%�Specificity: 99%

PPV: 38%

NPV: 99%

25 of 40

Key Studies for MCD testing in cancer-unaffected pts: PATHFINDER

MSK Confidential — do not distribute

Schrag et al., 2023

Took ~2 months to diagnosis in TP cases, but ~6 months in FP cases

Most resolved with labs and/or imaging

9% of TP required surgery for a diagnosis; 2% (n=1) FP cases

What about those with genetic risk for cancer?

26 of 40

Early Cancer Detection in Li-Fraumeni Syndrome with Cell-free DNA (Wong et al., 2024)

�Wong, Lou et al, Cancer Discov 2024

Latham et al., Cancer Discov 2024

MSK Confidential — do not distribute

Cancer signal in 81.6% of samples from LFS patients with active cancer when using ANY combination of assay type

Early stage disease: decreases to 50% detection
PPV of 54% overall

A Multi-modal cfDNA analysis is needed to optimize cancer detection

27 of 40

Early Cancer Detection in Li-Fraumeni Syndrome with Cell-free DNA (Wong et al., 2024)

�Wong, Lou et al, Cancer Discov 2024

Latham et al., Cancer Discov 2024

MSK Confidential — do not distribute

Cancer signal in 81.6% of samples from LFS patients with active cancer when using ANY combination of assay type
Early stage disease: decreases to 50% detection
PPV of 54% overall

What may we be able to leverage in other high-risk populations like Lynch Syndrome? Is there a pan-cancer signal that can be detected in blood?

A Multi-modal cfDNA analysis is needed to optimize cancer detection

28 of 40

MSI is associated with Lynch Syndrome Pan-cancer. Can we leverage that for cfDNA?

Latham, Srinivasan et al., JCO 2019

MSK Confidential — do not distribute

colonoscopy is the only proven effective cancer screening for Lynch Syndrome patients despite this being a pan-cancer syndrome

MSI-ctDNA may be a novel non-invasive approach for early detection & interception in Lynch Syndrome given the nearly universal MSI signal.

29 of 40

MSK-ACCESS is our institutional NYS-approved ctDNA assay

�Brannon et al., Nat Comm 2021

MSK Confidential — do not distribute

MSI detection was crucial in MSK-ACCESS design:

>25,000 tumors previously sequenced by MSK-IMPACT were analyzed, with >1,000 microsatellite sites captured to select the 165 most informative MSI loci
Regions are sequenced to ultra-deep coverage (~20,000x)
Enables reliable mutation calling to 0.1% allele fraction within the 129 cancer-associated genes captured

30 of 40

MSI calling in ctDNA via ADMIE: Allelic Distance-based MSI Estimator

Figures Courtesy of Dr. Mike Berger

MSK Confidential — do not distribute

Bioinformatics machine-learning tool to classify cfDNA samples based on distribution of absolute distances in allelic representation across all loci

Can we quantify that difference at MSI loci?= distance metric

The higher the difference in tumor vs normal= more likely MSI-H

In cfDNA, distance metric is also dependent on tumor fraction in cfDNA

31 of 40

MSK ACCESS collections from patients with known MSI tumors with or without underlying Lynch Syndrome

Unpublished Data. Do not copy or distribute

MSK Confidential — do not distribute

32 of 40

MSIDetect for MSI-ctDNA calling (Broad)

Utilizes ultra low-pass WGS
~23 million MS-loci simultaneously assessed
Distribution of read deviations from the reference at each locus is assessed
Log-likelihood ratio of each read that supports a tumor as being MSI is calculated
Detected all MSI cases with as low as 0.1% MSI DNA based on mixing studies

In contrast to ultra deep targeted sequencing, at the Broad, Gaddy Getz and group utilized an alternate strategy for MSI detection from cfDNA.

MSIDetect does not detect indels at particular loci, but classifies samples by combining information from all sequenced ~23 million MS-loci simultaneously utilizing low depth

(~0.5X) whole-genome sequencing (WGS) data. Even though many MS-loci will not be covered, the vast number of loci enable data aggregation, generating an “MSI mutation

signature” in cfDNA. Using 60 MSI and 200 MSS cell-line WGS, the distribution of read deviations from the reference for each locus was assessed and the log-likelihood ratio (LLR) of each read that supports a tumor as MSI was calcuated (Figure 4b). We then sum the LLR of all reads to obtain the MSIDetect score for MSI classification. We evaluated LLR score performance on low-tumor DNA fractions by mixing DNA of MSI and MSS cell-lines with normal DNA in different fractions, ranging from

100% to 0.1% MSI DNA, along with only normal controls. With WGS of mixed DNA and controls

(depth of ~0.5X), MSIDetect detected all MSI cases with as low as 0.1% MSI DNA in the

mixture (Figure 4c).

We applied MSIDetect to 50 cfDNA samples first using 22 MSS and 7 MSI samples as a training

set to determine MSI thresholds. Utilizing this threshold, we analyzed MSI and MSS cases with

detection of 23/26 MSI cases and 4/4 MSS cases. The cancer DNA fraction of cfDNA estimated

using copy number alterations were below the detection level in most of these cases by using

CNVs (<3% cancer DNA ).17

Figure 4. MSIDetect

A. Read length

distributions of MSI and

MSS. B. A schematic

description of the MSI

classification

process. C. MSIDetect

score values for

different fractions of

mixing the DNA of MSI

cell-lines with noncancerous

DNA and for

real cfDNA form colon

cases with and without

active MSI tumors using

WGS with 0.5x coverage.

33 of 40

Goal: MSIPredict development

To develop MSIPredict, a highly sensitive/specific MSI-cfDNA assay for the detection of MSI tumors in Lynch syndrome patients that utilizes both deep-targeted (MSI-ADMIE) and ultra-low-pass whole genome sequencing (MSIDetect) for MSI detection.

0.5x WGS

MSIDetect

20,000x 165 loci MSI-ADMIE

0.5x WGS + 20,000x 165 loci

MSIPredict

34 of 40

MSI-ctDNA calling and average VAF from genotyping

Average VAF may be used as a surrogate for quantifying ct-DNA

In samples with higher average VAF, improved MSI-ctDNA detection was observed

35 of 40

MSI-ctDNA calling and average VAF according to clinical disease status

36 of 40

MSK-ADMIE & MSIDetect: MSI-ctDNA calling and �clinical disease status

37 of 40

Notable Case

Unpublished Data. Do not copy or distribute

MSK Confidential — do not distribute

MSH6+

38 of 40

Notable Case

Unpublished Data. Do not copy or distribute

MSK Confidential — do not distribute

MSH6+

ADMIE

IMPACT (prostate)

MSK-ACCESS

Tumor mutational profile is different- initially filtered as possible background noise

39 of 40

Final Thoughts

MSK Confidential — do not distribute

Hereditary cancer syndromes are complex and surveillance recommendations evolve as more data is generated

Surveillance burnout is real- we need to be thinking of ways to improve adherence

cfDNA is promising technology for high-risk populations. It will likely require a multi-modal approach

40 of 40

Acknowledgements

MSK Confidential — do not distribute

Zsofia Stadler
Mike Berger
Ken Offit
Jennifer Mueller
Ying Liu
Gopa Iyer
Dave Solit
Rob Sidlow
Karmelina Charalambous

American Cancer Society Clinician Scientist Development Grant
The Society of MSK Research Grant
Starr Cancer Consortium
Mark Foundation
Robert and Kate Niehaus Center for Inherited Genomics
Romeo Milio Lynch Syndrome Foundation

General Internal Medicine Service Attendings & Clinical Genetics Service Attendings and Genetic Counselors

Mentors, Collaborators, and Research Team Members

Research Funding

Kanika Arora
Kelsey Breen
Megha Ranganathan
Matilde Borio
Jinru Shia
Arnie Markowitz
Fei Ye
Lora Ellenson
Lisa Milli