Primary Pancreatic Tissue

Met

asta

tic L

esio

ns

Multi-omic molecular comparison of primary versus metastatic pancreatic tumorsGagandeep Brar1, Edik Blais2, R Joseph Bender2, Jonathan Brody3, Davendra Sohal4, Subha Madhavan2,5, Vincent J. Picozzi6, Andrew Eugene Hendifar7,Vincent M. Chung8, David Halverson2, Sameh Mikhail9, Lynn McCormick Matrisian10, Lola Rahib10, Emanuel Petricoin2, Michael J. Pishvaian1,5

1MedStar Washington Hospital Center, Washington, DC; 2Perthera, Inc., Mclean, VA; 3Thomas Jefferson University, Philadelphia, PA; 4Cleveland Clinic, Cleveland, OH;5Georgetown University, Washington, D.C.; 6Virginia Mason Hospital and Medical Center, Seattle, WA; 7Cedars-Sinai Medical Center, Los Angeles, CA; 8City of Hope, Duarte, CA;9Mark H. Zangmeister Cancer Center, Columbus, OH; 10Pancreatic Cancer Action Network, Manhattan Beach, CA

BACKGROUNDMetastatic pancreatic adenocarcinoma is the thirdleading cause of cancer-associated mortality with amedian overall survival of < 12 months. Despiteimproved outcomes with multimodality therapy in earlystage pancreatic cancer, >70% of patients withresectable disease ultimately develop recurrence. It islikely then that circulating metastases are alreadypresent at initial diagnosis. Advances in theunderstanding and multi-omic characterization of solidtumors have allowed for the development of molecularlytargeted therapeutic agents in a number of solidtumors, and we have demonstrated that up to 25% ofpancreatic adenocarcinomas harbor “actionable”molecular alterations.

In an effort to assess the multi-omic changes thatmay occur as a pancreatic cancer metastasizes, wecompared the frequency of genetic, protein, andphosphoprotein alterations from primary vsmetastatic pancreatic tumors and from metastasesof different sites. By focusing on potential actionablegenetic, proteomic, and phosphoproteomic information,we sought to explore whether targeted therapies couldbe tailored to patients at metastatic progression basedon primary surgical material.

SITE-SPECIFIC TRENDS

CONCLUSIONS

PATIENT DEMOGRAPHICS GENOMIC COMPARISONS

Fig. 2 – Geographic distribution of patients whoreceived Perthera Reports for pancreaticadenocarcinoma based on biopsies of the primarytumor (182 blue dots) or of a distant metastasis (323orange dots). We have provided Perthera Reports toeach of the 505 patients in the analysis cohort and theironcologists from 41 different states across the UnitedStates (Fig. 2).

Fig. 6 – Frequency heatmap of lesscommon genomic alterations seen acrossprimary pancreatic tumors and metastaticlesions to the lung, liver, or other areas ofthe abdomen. Hierarchical clustering wasbased on euclidian distance with single linkagefor the percentages shown.

Comprehensive multi-omic (genomic, proteomic, andphosphoproteomic) testing was coordinated by Pertherathrough Perthera Precision MedicineTM for 505 patientswith pancreatic adenocarcinoma as part of the Know YourTumor Initiative in partnership with the Pancreatic CancerAction Network.

We compared proteomic results between metastaticlesions and primary tumors and found significantdifferences (FDR-adjusted q-value < 0.10, Fisher’s exacttest) in TLE3, ERCC1, MET, and TOP1 (Fig. 3).

• Within distant metastases, we observed possible tissue-specific patterns of protein expression with increasedTUBB3 and decreased PTEN in the liver compared tothe lung (FDR-adjusted q-value < 0.10)

• The four most common mutations (Fig. 5A) seen inpancreatic adenocarcinoma were detected at similarfrequencies across primary pancreatic tumors (182),liver lesions (234), lung lesions (34), abdominal lesions(55), and metastases at other sites (52, excluded fromanalyses).

• Highly actionable alterations were detected in 40primary tumors (22%) which was not significantlydifferent (p-value > 0.3, Fisher’s exact test) than the 85metastatic lesions with highly actionable findings (26%).

• No statistically significant differences in specific geneswere observed between primary vs. metastatic lesions,nor across the site of metastasis after correcting formultiple hypotheses.

• The proportion of actionable alterations (includingmutations in the homologous recombination DNA repairpathway) was similar across subgroups.

• The comparison of the molecular characteristics ofprimary vs metastatic pancreatic adenocarcinoma inpatients who have received Perthera Reportsreveals that the molecular characteristics are verysimilar and that actionable alterations are identifiedat the same frequency.

• Our findings support the belief that primarypancreatic cancers metastasize very early and thus:o The role of biomarker-directed therapy for early-

stage pancreatic cancers in lieu of, or inaddition to standard therapy could be furtherevaluated in prospective clinical trials.

o In addition, re-biopsy at recurrence may not benecessary which can decrease patientdiscomfort and anxiety, cost, and delay to nexttreatment.

• This is consistent with previously published reports,where the driver mutations were similar (althoughacknowledging differences in methods)1-5.

• This is unlike the data observed from other solidtumors (e.g. colon, lung, breast cancer) in whichsubstantial molecular discordance/ heterogeneityexists between primary tumors metastases.

OPERATIONAL SUMMARY

Fig. 1 – Operational overview of patientswith pancreatic adenocarcinoma referredwho received a Perthera Report. Informedconsent was obtained to coordinate multi-omictesting and incorporate medical and treatmenthistory into each Perthera Report.

PROTEOMIC & PHOSPHOPROTEOMIC COMPARISONS

Fig. 5 – Comparative frequencies of genomicalterations between metastatic lesions and primarypancreatic tumors. Each point represents an individualgenomic marker with size corresponding to the inverselog-transformed unadjusted p-value. Labels are shown forgenes with a modest trend toward significance(unadjusted p-value < 0.05).

RESULTS

A

B

82 Pre-Consent§ 7 in process§ 3 on hold

227 Closed Pre-Biopsy§ 34 patient changed mind§ 34 doctor unsupportive§ 6 unable to biopsy§ 2 cost/insurance§ 20 health concerns§ 82 deceased

226 Closed Pre-Consent§ 41 patient changed mind§ 27 doctor unsupportive§ 14 unable to biopsy§ 10 cost/insurance§ 16 health concerns§ 17 deceased

72 Closed Pre-Report§ 1 patient changed mind§ 2 doctor unsupportive§ 5 health concerns§ 51 deceased

142 Pre-Biopsy§ 10 in process§ 38 on hold

177 Pre-Report§ 40 in testing§ 7 in analysis§ 29 on hold

979 Patients Biopsied(78% of consented patients)

1254 Patients Consented(84% of referrals)

1490 Referrals(all pancreatic cancer subtypes)

838 Perthera Reports(86% of biopsied patients)

505 Patient Reports in Analysis Cohort(pancreatic adenocarcinoma)

METHODS• Tissue sent for CAP/CLIA-accredited testing:o NGS (Foundation Medicine, Inc.)o Immunohistochemistry (NeoGenomics)

• Perthera AITM was used to assess each patient’smulti-omic profile in the context of their treatmenthistory. Treatment options were then reviewed by amedical review panel in a cloud-based tumor boardand actionable treatment options were presented tothe patient and physician in the Perthera Report.

• Definition of actionability – literature supports ahigh response rate in patients with a specificmolecular abnormality (any cancer type) orpossible implication of response to therapy,based on mechanism or pathway.

• Treatment options regularly included specific clinicaltrials and off-label therapies.

Primary (n = 182) Metastatic (n = 323)Gender

Male 96 (53%) 172 (53%)Female 86 (47%) 151 (47%)

Age< 65 years old 101 (56%) 184 (57%)

>= 65 years old 81 (44%) 139 (43%)Ethnicity

African American 2 (2%) 6 (2%)Ashkenazi Jewish 3 (2%) 2 (1%)

Asian 8 (4%) 10 (3%)Caucasian 73 (40%) 122 (38%)Hispanic 9 (5%) 4 (1%)

Other 4 (2%) 4 (1%)Not Available 83 (45%) 175 (54%)

Table 1 – Baseline characteristics of analysis cohort.

Fig. 3 – Comparative frequencies of proteinor phosphoprotein positivity betweenprimary and metastatic tumors. Each pointrepresents an individual proteomic biomarkerwith size corresponding to the inverse log-transformed FDR-adjusted q-value and redindicating trends toward significant differences(q-value < 0.1).

Fig. 4 – Comparative frequencies of protein orphosphoprotein positivity between liver and lungmetastatic lesions in patients with pancreaticadenocarcinoma. Each point represents anindividual proteomic biomarker with sizecorresponding to the inverse log-transformed FDR-adjusted q-value and red indicating trends towardsignificant differences (q-value < 0.1).

% Mutated

Genom

ic Marker

Biopsied Lesion

Primary Pancreatic Tissue

Met

asta

tic L

esio

ns

Primary Pancreatic Tissue

Met

asta

tic L

esio

ns

Metastatic Lung Lesions

Met

asta

tic L

iver

Les

ions

%

%

Frequently altered genes assessed by NGS:• Common PDAC drivers: KRAS, TP53, CDKN2A, SMAD4• DNA repair: BRCA1/2, ATM, PALB2, MUTYH, BAP1, BRIP1,

CHEK2, FANCA• Epigenetic regulators: ARID1A, SMARCA4, KDM6A, KMT2D• RTKs: ERBB2, NTRK1/2/3, ALK, ROS1, FGFR1/2/3• Cell cycle: CCND1/2/3, CCNE1, CDK4/6, MYC• Kinases: STK11, BRAF , PIK3CA, MAP2K4, AKT1/2/3

Proteins assessed by IHC:• Chemopredictive: TS, RRM1, ERCC1, MGMT, TUBB3, TLE3,

TOP1, PGP• Immunotherapy: PD-1, PD-L1, MLH1, PMS2, MSH2, MSH6• RTK-related: PTEN, MET, ERBB2• Phosphoprotein: phospho-AKT

PrimaryAbdomen LiverLung

References1) Yachida S, et al. Nature. 2010 Oct 28;467(7319)2) Campbell PJ et al. Nature. 2010;467:1109–1113.3) Iacobuzio-Donahue CA. Gut. 2012 Jul;61(7):1085-944) Jones S, et al. PNAS 2008;105:4283–4288.5) Navin N, et al. Genome Res. 2010;20:68–80.