Published OnlineFirst December 17, 2019; DOI: 10.1158/1078-0432.CCR-19-1496

CLINICAL CANCER RESEARCH | PRECISION MEDICINE AND IMAGING

The Proteomic Landscape of Pancreatic Ductal Adenocarcinoma Liver Metastases Identifies Molecular Subtypes and Associations with Clinical Response A C Henry C.-H. Law1, Dragana Lagundzin1, Emalie J. Clement1, Fangfang Qiao1, Zachary S. Wagner1, Kimiko L. Krieger1, Diane Costanzo-Garvey2, Thomas C. Caffrey1, Jean L. Grem3, Dominick J. DiMaio2, Paul M. Grandgenett1, Leah M. Cook2, Kurt W. Fisher2, Fang Yu4, Michael A. Hollingsworth1, and Nicholas T. Woods1

ABSTRACT ◥ Purpose: Pancreatic ductal adenocarcinoma (PDAC) is a highly (i) metabolic; (ii) progenitor-like; (iii) proliferative; and (iv) metastatic disease that can be separated into distinct subtypes based inflammatory. PDAC risk factors of alcohol and tobacco con- on molecular signatures. Identifying PDAC subtype-specific ther- sumption correlate with subtype classifications. Enhanced sur- apeutic vulnerabilities is necessary to develop precision medicine vival is observed in FOLFIRINOX treated metabolic and pro- approaches to treat PDAC. genitor-like subtypes compared with the proliferative and inflam- Experimental Design: A total of 56 PDAC liver metastases were matory subtypes. In addition, TYMP, PDCD6IP, ERAP1, and obtained from the UNMC Rapid Autopsy Program and analyzed STMN showed significant association with patient survival in a with quantitative proteomics. PDAC subtypes were identified by subtype-specific manner. Gemcitabine-induced alterations in the principal component analysis based on expression profiling. proteome identify , such as serine hydroxymethyltrans- Proteomic subtypes were further characterized by the associated ferase 1, associated with drug resistance. clinical information, including but not limited to survival analysis, Conclusions: These data demonstrate that proteomic analysis drug treatment response, and smoking and drinking status. of clinical PDAC liver metastases can identify molecular signa- Results: Over 3,960 proteins were identified and used tures unique to disease subtypes and point to opportunities for to delineate four distinct PDAC microenvironment subtypes: therapeutic development to improve the treatment of PDAC.

Introduction metastatic lesions within 5 years of resection (3, 4). While the analysis of the primary tumor facilitates our understanding of the Pancreatic ductal adenocarcinoma (PDAC) is among the most molecular etiology of PDAC (5), characterization of metastatic lethal of cancers, with a 5-year survival rate of 8.5% and a cancer lesions has the potential to improve clinical interventions that mortality rate projected to outpace both breast and colon cancer in address the main cause of cancer mortality. Thus, understanding the coming years (1). The poor survival of patients with PDAC is the underlying molecular features of metastatic PDAC is necessary associated with the highly metastatic nature of this disease. to develop effective therapeutic interventions that improve patient Approximately 80% of these patients develop liver metastases, but survival. other common sites include the lung and peritoneum, with mul- Identifying cancer subtypes has the potential to improve patient tiple organ involvement often observed during the end stages of outcomes because subtype can be associated with treatment response. this disease (2). The contribution of disseminated disease to This has been most effectively employed for breast cancer (6), where lethality in PDAC is exemplified by the fact that, among patients expression profiling using the PAM50 gene expression predictor with early-stage and resected lesions, 60%–70% will present with outperforms IHC classification methods for its prognostic and pre- dictive ability (7). Several studies have characterized primary PDAC subtypes based on transcriptional profiling (5, 8–11). Using laser 1Eppley Institute for Research in Cancer, Fred & Pamela Buffett Cancer Center, capture microdissection (LCM), Collisson and colleagues classified University of Nebraska Medical Center, Omaha, Nebraska. 2Department of PDAC tumors into three subtypes (exocrine-like, classical, and quasi- Pathology and Microbiology, College of Medicine, University of Nebraska mesenchymal; ref. 8). Moffitt and colleagues classified PDAC into two 3 Medical Center, Omaha Nebraska. Internal Medicine, Division of Hematology tumor subtypes (classical and basal-like) and two stroma subtypes Oncology, University of Nebraska Medical Center, Omaha Nebraska. 4Depart- ment of Biostatistics, College of Public Health, University of Nebraska Medical (normal and activated) using virtual microdissection (9). Using tran- fi Center, Omaha Nebraska. scriptional pro les of intact whole-tumor tissues, Bailey and colleagues classified four subtypes of PDAC (squamous, immunogenic, pancre- Note: Supplementary data for this article are available at Clinical Cancer Research Online (http://clincancerres.aacrjournals.org/). atic progenitor, and ADEX), in which the squamous subtype is associated with significantly shorter survival than the other sub- Corresponding Author: Nicholas T. Woods, University of Nebraska Medical types (10). Puleo and colleagues also evaluated the transcriptome of Center, BCC9.12.395, 986805 Nebraska Medical Center, Omaha, NE 68198. fi fi fi Phone: 402-559-2248; Fax: 402-559-4651; E-mail: [email protected] PDAC formalin- xed paraf n-embedded (FFPE) and identi ed basal- like and classical subtypes (11). The consensus across these studies is Clin Cancer Res 2020;XX:XX–XX that there are two predominant PDAC tumor cell subtypes: classical doi: 10.1158/1078-0432.CCR-19-1496 and squamous, following the proposed harmonized nomencla- 2019 American Association for Cancer Research. ture (12). To date, PDAC subtyping is based on RNA expression

AACRJournals.org | OF1

Downloaded from clincancerres.aacrjournals.org on October 2, 2021. © 2019 American Association for Cancer Research. Published OnlineFirst December 17, 2019; DOI: 10.1158/1078-0432.CCR-19-1496

Law et al.

Translational Relevance Materials and Methods Ethics statement Pancreatic ductal adenocarcinoma is a deadly disease with a Investigators obtained informed consent for each patient enrolled in propensity to metastasize even at the earliest detectable stage. the UNMC Rapid Autopsy Program (IRB #091-01). This study was Effective treatment strategies must address metastatic disease, conducted in accordance with the ethical guidelines established by the which requires a better understanding of the underlying molecular Declaration of Helsinki. features of this disease. Stratifying PDAC into distinct microen- vironment subtypes based on proteomic signatures has the poten- Sample preparation tial to identify subtype-specific treatment vulnerabilities that could The frozen PDAC tissues and the corresponding tumor-adjacent improve patient outcomes. Utilizing a quantitative mass spectrom- tissues were available from the RAP in UNMC. For each of the samples, etry approach, we identify four unique metastatic PDAC micro- 5 mg of the frozen tissue was ground into a fine powder with a liquid environment subtypes and demonstrate subtype-specific vulner- nitrogen-cooled mortar and pestle. The ground tissue was then lysed abilities using patient treatment data, and identify 52 proteins that with 1 mL of RIPA buffer (25 mmol/L Tris-HCl pH 7.6, 150 mmol/L exhibit subtype-specific correlations with patient survival. The NaCl, 1% NP-40, 1% sodium deoxycholate, 0.1% SDS) and was frozen classification system and the protein expression signatures in 80C until further used. The albumin and IgG contents in the described here provide a basis to facilitate the design and imple- protein lysates were first depleted with the Pierce Top 2 Abundant mentation of subtype-specific PDAC treatment strategies. Protein Depletion Spin Columns and labeled with TMT reagents per the manufacturer's instructions. Detailed procedures were listed in the Supporting Information. profiles, but orthogonal methodologies, such as proteomics, have the LC/MS-MS and bioinformatics analysis potential to reveal additional features that could improve the func- tional characterization of PDAC subtypes. The mass spectrometry data was acquired on a Dionex Nano There have been comparative proteomics studies on PDAC that Ultimate 3000 coupled with an Orbitrap Fusion Lumos. The fractions m have identified proteins from tissue, plasma, pancreatic juice, cyst collected from the high-pH separation were resuspended in 20 L fluid, and urine associated with this disease. These efforts illustrate the of 0.1% formic acid. Two microliters of each fraction was injected into the system for tandem mass spectrometry analysis. The MS and potential of applying proteomics approaches to improve early detec- n tion and treatment of PDAC based on single proteins (13, 14). IHC MS spectra collected from the experiment were searched against the evaluating expressions of only KRT81 and HNF1A proteins has been homo sapiens protein sequence database (downloaded in 10/2017, used to stratify PDAC tumors as either classical, quasi-mesenchymal, 42252 entries) and the respective decoy database with Sequest HT in or exocrine-like (15). However, prior to this study, proteomics-based the Proteome Discoverer 2.2 pipeline. The reporter ion ratios of these approaches have not been performed at a scale that supports PDAC proteins were exported from the Proteome Discoverer and the P values subtype classification based on proteome quantification. Further were calculated with the Wilcoxon-signed rank test using R. The exploration of the PDAC tumor proteome along with detailed clinical software packages used in the postdatabase search analysis are listed in fi records could improve the diagnosis and treatment of this cancer. the Supporting Information. Mass spectrometry data les have been The overarching goal of this project was to evaluate the proteome deposited to the ProteomeXchange Consortium via the PRIDE (16) of PDAC liver metastases to distinguish unique subtypes from partner repository with the project accessions: PXD012173 and clinical samples. As a proof-of-principle that PDAC microenviron- PXD015492. ment subtypes can be delineated using proteomics, we have devel- oped and validated a classification system using quantitative pro- Gemcitabine treatment of MIA PaCa-2 cells teomics data from 68 tissue samples in total from the rapid autopsy The human PDAC cell lines MIA PaCa-2 and Panc 10.05 was program (RAP) managed by the UNMC Pancreas SPORE. This obtained from the ATCC and were cultured per the manufacturer's proteomics analysis identified over 3,960 proteins and quantitative instructions. Briefly, the cells were grown in DMEM supplemented profiling of the 916 of these proteins was used to delineate four with 10% FBS, 2.5% horse serum, amphotericin B, and penicillin– distinct subtypes of PDAC liver metastases, which share many streptomycin (Corning) in 5% CO2 atmosphere at 37 C. Gemcitabine- molecular signatures with transcriptomic-derived subtypes. The conditioned MIA PaCa-2 cells were generated by incubation with proposed proteomic-based subtyping system showed a significant 10 nmol/L gemcitabine (Selleckchem) freshly diluted in DMSO for association with patients’ alcohol and tobacco exposure. In addi- 6 days. The cell lysates were collected with RIPA buffer (25 mmol/L tion, a survival advantage was observed in the metabolic and Tris-HCl pH 7.6, 150 mmol/L NaCl, 1% NP-40, 1% sodium deox- progenitor-like subtypes treated with FOLFIRINOX (5-fluoroura- ycholate, 0.1% SDS) and probed for MTHFD1 and SHMT1 by Western cil, leucovorin, irinotecan, and oxaliplatin) and gemcitabine com- blotting (detailed conditions in Supporting Information). pared with gemcitabine, but this survival benefitwasnotobserved in the inflammatory and proliferative subtypes. The serine hydro- Gemcitabine cytotoxicity assay xymethyltransferase (SHMT1), a metabolic enzyme involved in Two unique shRNA constructs targeting SHMT1 single carbon metabolism, was identified as a mediator of gemci- (TRCN000034766 and TRCN000034767; shSHMT1 #1 and #2) and tabine resistance. In addition, 52 protein expression profiles a nontargeting scrambled control (shScr) in pLKO.1 were used in this were found to correlate with patients’ survival in a subtype- study (Sigma). The lentiviral supernatant was produced by calcium specific manner. These data demonstrate the clinical relevance of phosphate transfection into 293FT cells, as described previously (17), this proteomics classification model and illustrate its potential for and used to transduce MIA PaCa-2 and PANC 10.05 cells. The the development of therapeutic strategies to target PDAC liver transduced cells were selected with puromycin for 5 days before use metastases. in cytotoxicity assays. Knockdown of SHMT1 was confirmed by

OF2 Clin Cancer Res; 2020 CLINICAL CANCER RESEARCH

Downloaded from clincancerres.aacrjournals.org on October 2, 2021. © 2019 American Association for Cancer Research. Published OnlineFirst December 17, 2019; DOI: 10.1158/1078-0432.CCR-19-1496

Proteomics of Pancreatic Ductal Adenocarcinoma Metastases

Western blotting. For the gemcitabine cytotoxicity assay, 1,000 cells collected at the same stage of disease under a standardized proce- were aliquoted into each well in a 96-well plate. The cells were dure (18). The metastatic tissue proteome was determined from a incubated with gemcitabine at different concentrations for 3 days. cohort of 59 patients [56 PDAC, 3 pancreatic neuroendocrine tumors Cell viability was determined by CellTiter-Glo (Promega), and (PanNET)] that were annotated for tumor stage at diagnosis, gender, the luminescent signal was measured by FLUOstar Optima (BMG age, overall survival (OS) calculated from the day of diagnosis, Labtech). EC50(s) were estimated with GraphPad Prism 7 from the metastatic involvement, and PDAC risk factors of alcohol and tobacco exported data (GraphPad). consumption (Table 1; Supplementary Table S1). These samples were randomly divided into seven batches and differentially labeled with isotopic tags using the 10-plex TMT kit (Fig. 1A; Supplementary Results Table S2). The reference mix of all 59 samples was tagged with the Acquisition of the PDAC liver metastases proteome TMT126 label, which serves as a common reference for quantitation This project aimed to explore proteomic variance in PDAC liver across all seven batches (Supplementary Table S2). The overall analysis metastases from patient samples collected by the UNMC Pancreas identified 30,811 peptides mapping to 3,960 proteins in which 1,842 SPORE Rapid Autopsy Program. This Program ensures all samples are were quantified and 916 were quantified with at least 5 peptides across

A Reference PDAC Liver control TMT Labeled metastatic tissue Extract Tryptic proteins digest Mix 127N 127C 128N 128C 126

High pH reversed 130N 130C 129N 129C 131 phase fractionation

MS3 MS2 MS1 Quantitation Peptide seq. Precursor ion

m/z m/z m/z Mass spectrometry

B 30,811 Peptides (1% FDR) C 10 3,960 Proteins (1% FDR) PanNET 1,842 Quantified proteins 0 PDAC 916 with ≥5 Peptides PC2 (22%) -10 for quantitation -20 -20 0 20 40 PC1 (48%)

D Tumor- E Inflammatory 20 adjacent 20 tissue Progenitor-like 0 Liver PC2 (12%) 0 -20 metastases PC2 (12%) -40 0 40 Metabolic PC1 (43%) -20 Proliferative

-40 -40 -20 02040 PC1 (26%)

Figure 1. A, Proteomics workflow for the analysis of the PDAC liver metastases proteome. B, Overview of the proteomics data (FDR). C, Score plot of the multivariate analysis of PDAC and PanNET proteomes. D, Score plot of the multivariate analysis of the liver metastases and the tumor-adjacent tissue proteome. E, Score plot of the multivariate analysis of the liver metastases proteome.

AACRJournals.org Clin Cancer Res; 2020 OF3

Downloaded from clincancerres.aacrjournals.org on October 2, 2021. © 2019 American Association for Cancer Research. Published OnlineFirst December 17, 2019; DOI: 10.1158/1078-0432.CCR-19-1496

Law et al.

80% of the samples (Fig. 1B; Supplementary Tables S3 and S4). The set (GSN) and lumican (LUM) between the proliferative and of 916 proteins were used in the multivariate analysis (Fig. 1B; the inflammatory subtypes that only resulted in a slight difference Supplementary Table S5). These 916 proteins represent a broad array between the signature scores of the normal and the activated stroma of cellular functions across the proteome, including extracellular subtype between these two proteomic subtypes (Fig. 2B and D). matrix organization, protein processing and transport, translation, This demonstrates the proteomic PDAC subtyping method incor- glycolytic processes, NADPH metabolism, cell migration, immune porates extracellular protein expression, which may not be captured response, fibronectin binding, and cell homeostasis determined by by transcriptomic approaches. Spatial Analysis of Functional Enrichment (SAFE) (Supplementary There exists a debate whether the exocrine-like and ADEX classi- Fig. S1; Supplementary Table S6; ref. 19). fications represent unique subtypes or contamination from acinar cells Partial least squares-discriminant analysis (PLS-DA) effectively present in the tumor microenvironment (5, 12). However, the exis- distinguished PDAC from PanNET liver metastases with high confi- tence of patient-derived cell lines and propagated xenografts that are dence (Fisher probability ¼ 3.1 10 5; Fig. 1C). We further tested the classified as exocrine-like suggest they may represent unique sub- methodology by comparing the proteomic signatures of 9 PDAC liver types (15, 22). Furthermore, different cellular components of the metastases against matched tumor-adjacent uninvolved liver. Princi- PDAC tumor microenvironment affect a range of cancer phenotypes pal component analysis (PCA) effectively separated the liver metas- including metastatic potential and treatment efficacy and contribute to tases and the uninvolved liver tissue into two distinct categories the overall tumor microenvironment (23, 24). Like our study, Bailey (Fig. 1D). All pairs of liver metastases and tumor-adjacent tissue and colleagues also used the whole tumor for their analysis without proteomes were well separated on the score plot (Supplementary virtual or physical microdissection. Therefore, we chose to include Fig. S2A). In the corresponding PLS-DA model, the ROC curves the cross-comparison to the Bailey and colleagues classification differentiated both PDAC and PanNET tumors from the correspond- system. The proliferative and inflammatory subtypes share signa- ing tumor-adjacent liver tissues with high sensitivity and specificity tures associated with the squamous/quasi-mesenchymal subtype [area under the curve, AUC(PDAC) ¼ 1; AUC(PanNET) ¼ 1; (Fig. 2A, E,andF). However, our proteomic classification system Supplementary Fig. S2B and C]. These analyses demonstrate that the can further subcategorize the squamous subtype into two distinct underlying quantitative proteomics data provide sufficient sensitivity subtypes (inflammatory and proliferative). The progenitor-like and specificity to differentiate two distinct types of pancreatic cancer subtype nomenclature was used because of the similarity to the metastases to the liver as well as distinguish tumor tissue from the Bailey and colleagues progenitor subtype. There is also an associ- adjacent uninvolved liver. ation between the metabolic and the ADEX/exocrine-like subtypes, To explore the variance in liver metastases proteome, we con- which would not be attributed to acinar cell contamination because structed a PCA model with hierarchical clustering using the 916 this analysis used liver metastases. The metabolic association with quantified proteins (Fig. 1E; Supplementary Table S5). The sum of the ADEX subtype (Bailey and colleagues) may provide further squared error analysis revealed that the intragroup variance was best support to an exocrine-like subtype of PDAC, but we cannot rule explained when the samples were divided into four major subtypes and out the possibility that it is a byproduct of signals from normal liver three protein clusters (Supplementary Fig. S3A and S3B). The four cells. However, liver tissue adjacent to metabolic tumors exhibits protein subtypes identified were as follows: (i) metabolic (n ¼ 7); (ii) unique protein expression signatures that can differentiate these progenitor-like (n ¼ 21); (iii) proliferative (n ¼ 11); and (iv) inflam- tissues (Figs. 1D; Supplementary Fig. S2). matory (n ¼ 17; Fig. 1E). Subtype nomenclature is based on protein The consistencies between our proteomics-based and the transcrip- enrichments for each subtype or their relation to previously described tomic-based classification systems suggest that many of the same transcriptional subtypes. The robustness of the model was evaluated by transcriptional signatures found in primary tumors can be found by the 7-fold cross-validation operation built-in in SIMCA 15. The proteomics in the liver metastases. This is supported by a recent study goodness of fit(R2X) and the predictability (Q2) of this unsupervised that determined RNA signatures from metastatic tissue obtained from PCA model were 0.63 and 0.42, which was comparable with other PCA a variety of anatomic sites can be used to identify PDAC subtypes (25). models in the literature (20, 21). The coefficients for each of the To confi rm this, we performed quantitative proteomics on nine proteins in the corresponding supervised PLS-DA model were listed in matched primary tumor samples to evaluate PDAC subtype conser- Supplementary Table S7. The x2 analysis showed that batch effects did vation with metastases. The PLS-DA model had R2X, R2Y, and Q2 of not impact subtype classifications (Supplementary Fig. S3C). 0.695, 0.99, and 0.401, respectively (Supplementary Fig. S5). This indicates PDAC molecular subtypes were generally conserved between Overview of the classification scheme the primary tumor and the liver metastases in matched samples. We identified significant correlations between the subtypes iden- The protein expression profiles observed in PDAC liver metastases tified by proteomics and those identified by Moffitt and colleagues distinguish three groups of similarly expressed proteins (protein [x2 test, P (tumor) ¼ 8.92 10 6, P (stroma) ¼ 2.06 10 3], cluster 1–3) associated with a range of biological processes (Supple- Collisson and colleagues (P ¼ 2.10 10 5), and Bailey and mentary Fig. S6A and S6B). The metabolic and progenitor-like sub- colleagues (P ¼ 8.85 10 8; Fig. 2A; Supplementary Fig. S4A– types are characterized by metabolism-related proteins in protein S4D). In comparison with the Moffitt and colleagues classification cluster 1, enriched with proteins in the ethanol oxidation pathways, system, three representative proteins with the highest weight factor mitochondrial fatty acid b-oxidation, and retinoic acid signaling (ALDH2, IDH1, and TST) associated with the classical tumor pathways (Supplementary Fig. S6C; Supplementary Fig. S7; Supple- signature exhibit higher expression in the metabolic and the pro- mentary Table S8; Supplementary Information). Even though both the genitor-like subtypes than the other two proteomic subtypes, while metabolic and the progenitor-like subtypes were characterized by the inflammatory subtype showed a high expression of basal-like protein cluster 1, the metabolic subtype exhibits higher expression of tumor signature (ANXA1, ANXA3, and ITGA2), resulting in these proteins, such as those associated with signaling by retinoic acid higher signature scores (Fig. 2B and C). There were significant (Supplementary Fig. S6D). The proliferative subtype proteome is differences between the expression of extracellular proteins like enriched with ribonucleoproteins and Cajal body proteins in protein

OF4 Clin Cancer Res; 2020 CLINICAL CANCER RESEARCH

Downloaded from clincancerres.aacrjournals.org on October 2, 2021. © 2019 American Association for Cancer Research. Published OnlineFirst December 17, 2019; DOI: 10.1158/1078-0432.CCR-19-1496

Proteomics of Pancreatic Ductal Adenocarcinoma Metastases

A Metabolic Progenitor-like Proliferative Inflammatory

Tumor Moffitt et al: stroma Collisson et al: Bailey et al: Protein cluster 1 Protein cluster 2 Protein cluster 3

Moffitt et al: Classical tumor Basal-like tumor Normal stroma Activated stroma Low stroma Unclassified

Collisson et al: Exocrine Classical Bailey et al: ADEX Immunogenic Progenitor Squamous Quasi-mesenchymal

B Classical tumor Basal-like tumor Normal stroma Activated stroma

ALDH2 ANXA1 TAGLN THBS2

IDH1 ANXA3 C7 LUM Protein name TST ITGA2 GSN POSTN

−1.0 −0.5 0 1.00.5 −1.2 −0.8 −0.4 0 0.80.4 −1.2 −0.6 00.6 −1.2 −0.6 0 0.6

Log2 (relative expression) Metabolic Progenitor-like Proliferative Inflammatory

C Metabolic Progenitor-like Proliferative Inflammatory D Metabolic Progenitor-like Proliferative Inflammatory 0.8 1.0 1.0 0.6 0.8 0.8 0.8 0.8 0.8 0.6 0.5 0.4 0.6 0.6 0.4 0.4 0.4 0.4 0.0 0.2 0.4 0.4 0.2 0.0 Signature score 0.2 Signature score -0.5 0.0 0.0 0.0 0.0 0.0 -0.2 CB CB CBCB LNA LNA LNA LNA Moffitt et al (tumor): C = Classical B = Basal-like Moffitt et al (stroma): L = Low N = Normal A = Activated

E Metabolic Progenitor-like Proliferative Inflammatory F Metabolic Progenitor-like Proliferative Inflammatory 0.5 0.6 0.6 0.6 0.8 0.6 0.6 0.4 0.4 0.4 0.4 0.4 0.4 0.2 0.4 0.2 0.3 0.3 0.2 0.0 0.0 Signature score Signature score -0.2 0.2 0.2 0.2 0.0 0.2 CEQ CEQ CEQ CEQ AIPS AIPS AIPS AIPS Collisson et al: C = Classical E = Exocrine-like Bailey et al: A = ADEX I = Immunogenic Q = Quasi-mesenchymal P = Progenitor S = Squamous

Figure 2. A, Mapping of the proteomic subtypes to the Moffitt et al., Collisson et al., and Bailey et al. classification schemes for each of the samples. The missing data in the ribbon above the heatmap indicate the signatures scores for these samples did not reach the threshold to accurately assign a corresponding transcriptomic subtype. The details of the x2 test are shown in Supplementary Fig. S4. Heatmap showing the association between protein expression and the proteomic subtypes. The red and blue colors in each pixel indicate protein up- and downregulation, respectively. B, Representative signature gene expressions in the Moffitt et al. classification scheme across the four proteomic subtypes. C–F, The signature scores of the four proteomics subtypes in the Moffitt et al., Collisson et al., and Bailey et al. classification systems.

AACRJournals.org Clin Cancer Res; 2020 OF5

Downloaded from clincancerres.aacrjournals.org on October 2, 2021. © 2019 American Association for Cancer Research. Published OnlineFirst December 17, 2019; DOI: 10.1158/1078-0432.CCR-19-1496

Law et al.

cluster 2 that are associated with translation, cell proliferation, results support the concept of PDAC subtype–specific response to and telomere maintenance in cancer cells (Supplementary therapy. Fig. S6C and S6E; Supplementary Tables S8 and S9; Supplementary Patient survival based on treatment(s) with gemcitabine, FOL- Information; ref. 26). The inflammatory subtype is characterized by FIRINOX/FOLFOX, abraxane/paclitaxel, tarceva/erlotinib, and protein cluster 3 and is enriched for proteins related to pentose radiation in each PDAC subtype was also evaluated (Supplementary phosphate pathway, adaptive immune response, complement acti- Fig. S10). Patients could be grouped in multiple treatments because vation, IL8 production, and extracellular fibril organization (Sup- their inclusion was dictated by whether they received the indicated plementary Figs. S6C, S6F, and S6G; Supplementary Tables S9 and therapy or not, and patient treatments were variable. This analysis S10; Supplementary Information). These pathways and processes indicated that patients with the progenitor-like subtype treated with are known to create an immunosuppressive and chemoresistant gemcitabine have a significant increase in survival compared with environment that supports tumor growth (27–29). The gene set cases that do not receive this treatment (P ¼ 0.001).Notably,a enrichment analysis on the average protein expression of each significant increase in survival probability was not observed in the subtype in the Reactome pathways also agreed with the biology progenitor-like subtype for any of the other treatments evaluated. In described above (Supplementary Table S11). Together, these data addition, the trends observed in the metabolic subtype following demonstrate that coexpressed proteins participate in cancer- abraxane/paclitaxel treatment suggest a negative correlation associated pathways that are differentially represented across PDAC between receiving this treatment and survival probability (P ¼ subtypes identified by proteomics. 0.008; Supplementary Fig. S10). It will be important to evaluate patient outcomes in response to individual therapies stratified by Proteomic signatures of PDAC liver metastases associate with subtype in a much larger dataset to identify significant trends that clinical features would support the implementation of personalized treatments Health and lifestyle data are collected for each RAP donor, based on PDAC subtypes. including diabetes, alcohol use, and tobacco use, which are risk factors for the development of PDAC (30). There are no significant Association between individual protein expression and survival differences in the neoplastic cellularity across the four proteomics is subtype-dependent subtypes in the tumors analyzed by pathologic review (Fig. 3A), Protein expression patterns in PDAC liver metastases also have suggesting this variable did not affect subtype classification. In the potential to reveal new associations with clinical outcomes or addition, the incidence of diabetes is distributed at expected rates identify novel therapeutic targets (14, 31). With the understanding across each of the four proteomics-defined subtypes and does not that PDAC is not a singular disease, we hypothesized that associa- appear to influence subtype classification (Fig. 3B). However, there tions between protein expression and patient survival might display are significantly more patients than expected with the proliferative subtype-specific characteristics. Partial least squares (PLS) analysis subtype that reported a history of alcohol use (hypergeometric test, was first used to evaluate the association between the number of P ¼ 0.025), and there are no patients with the inflammatory subtype survival days after diagnosis with protein expression from then that reported alcohol use (P ¼ 0.002; Fig. 3C). With regard to entire 56-patient cohort usedinthisstudy.Thisidentified 52 tobacco, the metabolic subtype includes more patients who report proteins associated with either increased or decreased survival tobacco use than expected (P ¼ 0.009), while the inflammatory probability (Fig. 4A). The Kaplan–Meier survival curves were subtype included significantly fewer than expected tobacco users (P plotted using the upper and lower tertiles (Supplementary ¼ 0.014; Fig. 3D). Table S12), and the log-rank test was used to determine significant We grouped the proliferative and the inflammatory subtypes differences in survival probability between the two groups. Proteins because combined they correspond to the squamous subtype in the with P 0.05 were cross-referenced with the PLS survival model as Bailey and colleagues classification system (Fig. 2A), which was internal validation. A total of 32 proteins with elevated expression demonstrated to be associated with shorter survival than the other demonstrated a significant association with increased survival transcriptionally defined subtypes in that study (10). Because the (Supplementary Fig. S11). An additional 20 proteins demonstrated patients in the RAP cohort analyzed in this study were not treated an inverse correlation between expression and survival probability with a standard set of chemotherapeutics, we focused our evaluation (Supplementary Fig. S12). Some of these proteins include thymidine of patient survival on three treatment groups: (i) untreated: did phosphorylase (TYMP; Fig. 4B), programmed cell death 6-inter- not receive either gemcitabine or FOLFIRINOX; (ii) gemcitabine: acting protein (PDCD6IP; Fig. 4C), 1 (STMN1; Fig. 4D), received at least gemcitabine; and (iii) FOLFIRINOX þ Gem: received and endoplasmic reticulum aminopeptidase 1 (ERAP1; Fig. 4E), FOLFIRINOX followed by at least gemcitabine. OS is calculated from which are known to be associated with cancer phenotypes (details the day of diagnosis. Across all subtypes, OS for gemcitabine-treated discussed in Supplementary Information; Supplementary Fig. S14; patients was 271.5 days and 336 days for FOLFIRINOX þ Gem (HR, refs. 32–39). 2.18; 95% confidence interval (CI), 1.57–3.01; P ¼ 0.02; Fig. 3E and F). We found that many of these 52 proteins display subtype-specific The OS for patients classified with proliferative and inflammatory expression patterns (Fig. 4F). Because the proteomics subtype model subtypes treated with gemcitabine was 258 days and 288 days for (R2X ¼ 65%) better explained the variance observed in the proteome FOLFIRINOX þ Gem (HR, 1.57; 95% CI, 1.02–2.41; P ¼ 0.29; Fig. 3G than the survival prediction model (R2X ¼ 48%), we hypothesized and H). However, OS for patients classified with the metabolic that a better survival regression model could be built based on and progenitor-like subtypes treated with gemcitabine was 286.5 days individual subtypes. Figure 4G–J shows the loading plot of the same and 401.5 days for FOLFIRINOX þ Gem (HR, 3.37; 95% CI, 1.02–5.90; 52 proteins in Fig. 4A for the individual subtype survival regression P ¼ 0.03; Fig. 3I and J). These data indicate that the metabolic and model. The variance captured by the regressed survival time (Q2Y) progenitor-like subtypes, but not the proliferative and inflammatory increased from 24% to 38%–84%. TYMP is highly correlated with subtypes, display a decreased risk of death when FOLFIRINOX is given survival in the progenitor-like subtype (Spearman correlation coeffi- in addition to gemcitabine as part of the treatment course. These cient, r ¼ 0.62), but this correlation is not as prominent in the other

OF6 Clin Cancer Res; 2020 CLINICAL CANCER RESEARCH

Downloaded from clincancerres.aacrjournals.org on October 2, 2021. © 2019 American Association for Cancer Research. Published OnlineFirst December 17, 2019; DOI: 10.1158/1078-0432.CCR-19-1496

Proteomics of Pancreatic Ductal Adenocarcinoma Metastases

A B 50 50 40 40 30 30 20 20

Neoplastic 10 10 diabetes (%) Patients with cellularity (%) 0 0

Metabolic Metabolic ProliferativeInflammatory All patients Proliferative Progenitor-like Progenitor-likeInflammatory C P = 0.025 0.002 D P = 0.009 0.014 60 100 50 80 40 60 30 (%) (%) 40 20 20 Patients with Patients with 10 alcohol history 0 tobacco history 0

Metabolic All patientsMetabolic All patients Proliferative ProliferativeInflammatory Progenitor-likeInflammatory Progenitor-like

E 1.0 Untreated F (n = 7) HR (95% CI) P 0.8 Gemcitabine n Gemcitabine 2.18 (1.57–3.01) 0.02 0.6 ( = 28) FOLFIRINOX 0.4 +Gem (n = 21) Untreated 9.82 (5.66–17.03) <0.01 All patients 0.2

survival probability 0.0 Favors FOLFIRINOX+Gem 0 500 1,000 1,500 2,000 1 2 4 8 163264128 Survival days HR G H 1.0 Untreated (n = 1) HR (95% CI) P 0.8 Gemcitabine n Gemcitabine 1.57 (1.02–2.41) 0.29 0.6 ( = 14) FOLFIRINOX 0.4 +Gem (n = 13) Untreated 29.86 (8.29–107.59) 0.01

inflammatory 0.2 Proliferative and survival probability 0.0 Favors FOLFIRINOX+Gem 0 500 1,000 1,500 2,000 1 2 4 8 16 32 64 128 Survival days HR

I 1.0 Untreated J (n = 6) HR (95% CI) P 0.8 Gemcitabine n Gemcitabine 3.37 (1.92–5.90) 0.03 0.6 ( = 14) FOLFIRINOX 0.4 +Gem (n = 8) Untreated 13.28 (6.34–27.78) <0.01 0.2 progenitor-like Metabolic and

survival probability 0.0 Favors FOLFIRINOX+Gem 0 500 1,000 1,500 2,000 1 2 4 8 163264128 Survival days HR

Figure 3. A, Percentage neoplastic cellularity in each of the four proteomics subtypes determined by pathologic review of H&E-stained slides. The t test P values were all >0.05 in all combinations of subtypes. The distribution of patients with a history of diabetes (B), alcohol consumption (C), and tobacco use (D) across different proteomics subtypes. Hypergeometric P < 0.05 are displayed above the bars. Survival analysis of patient treatment groups based on proteomic subtypes. Kaplan–Meier curves of all patients (E), combined proliferative and inflammatory subtypes (G), and combined metabolic and progenitor subtypes (I). Forest plots of the Cox proportional regression adjusted HRs and the corresponding P values of all patients (F), combined proliferative and inflammatory subtypes (H), and combined metabolic and progenitor subtypes (J). Survival days were calculated from the day of diagnosis.

AACRJournals.org Clin Cancer Res; 2020 OF7

Downloaded from clincancerres.aacrjournals.org on October 2, 2021. © 2019 American Association for Cancer Research. Published OnlineFirst December 17, 2019; DOI: 10.1158/1078-0432.CCR-19-1496

Law et al.

A TYMP B TYMP C PDCD6IP High, n = 16 High, n = 16 0.2 Low, n = 15 Low, n = 15 P = 0.02 P = 3.4 × 10–3

0.15 PDCD6IP 0.4 0.6 0.8 1.0 0.8 0.6 0.4

0.1 0.2

0.0 0.2 0.2 1.0 0.8 0.6 0.0 0.4 0.0 0.0 Survival probability 0 200 400 600 800 1,000 Survival probability 0 200 400 600 800 1,000 0.05 Survival days Survival days

0 D STMN1 E ERAP1 High, n = 16 High, n = 19 PC2 (11%) Low, n = 15 Low, n = 18 −0.05 STMN1 P = 0.01 P = 0.02

−0.1

0.4 0.6 0.8 1.0 0.8 0.6 0.4 0.4 0.6 0.8 1.0 0.8 0.6 0.4 ERAP1

−0.15

0.0 0.2 0.2 0.0 0.0 0.2 0.2 0.0 Survival probability −0.1 −0.05 0 0.05 0.1 0.15 Survival probability 0 500 1,000 1,500 2,000 0 500 1,000 1,500 2,000 PC1 (39%) Survival days Survival days

F G K 1,000 Metabolic, r = 0.14 TYMP PDCD6IP Progenitor-like, r = 0.62 0.04 Proliferative, r = 0.33 Inflammatory, r = 0.23 500 Metabolic Progenitor-like Proliferative Inflammatory 0 TYMP ERAP1 APRT Metabolic STMN1 Survival days HSD17B12 PC2 (2.3%) −0.04 OLA1 DDT −0.04 0 0.04 −1 01 NCEH1 TYMP Expression CLYBL PC1 (97%) HSD11B1 PYGB 1,000 GFPT1 H TYMP L Metabolic, r = 0.82 ATP5F1 0.1 Progenitor-like, r = 0.10 PPA1 GAPDH Proliferative, r = 0.17 DBI Inflammatory, r = 0.20 APOC1 0 500 ATPIF1 PDCD6IP EML2 STMN1 Survival days ACTR2 PC2 (29%) −0.1 ERAP1 ARPC4 Progenitor-like CAPN2 − − −0.5 −0.25 000.25 .5 ACTB 0.08 0.04 0 0.04 0.08 0.12 VASP PC1 (63%) PDCD6IP Expression CAPNS1 MVP I M 1,000 Metabolic, r = −0.10 STMN1 PDCD6IP r − KRT9 Progenitor-like, = 0.55 0.04 r PPIA TYMP Proliferative, = 0.07 GARS Inflammatory, r = −0.01 RPL10A 0 500 FKBP4 RPS10 − PDCD6IP 0.04 STMN1 PC2 (11%) Proliferative AGR2 ERAP1 Survival days ANXA4 −0.08 SLC4A1 −0.08 −0.04 0 0.04 −21−1 02 LMAN2 PC1 (86%) STMN1 Expression PEPD PSMA4 PSMA1 J N 1,000 Metabolic, r = -0.39 ERAP1 0.08 TYMP Progenitor-like, r = −0.33 PRDX5 ETHE1 0.04 STMN1 Proliferative, r = −0.49 CP Inflammatory, r = −0.04 YWHAB 0 500 LGALS3 LGALS1 −0.04 PDCD6IP

HNMT Survival days PC2 (16%)

Inflammatory ERAP1 HMGB2 −0.08 NASP CLU −0.08 −0.04 0 0.04 −10−0.5 01.5 HBB PC1 (75%) ERAP1 Expression HBD

Figure 4. A, Loading plot of the multivariate analysis in survival days versus the proteome. The size and intensity of the red color in each dot correlates with the variance of importance (VIP) in the model. B–E, Kaplan–Meier curves and log-rank test P values of representative survival markers, TYMP, PDCD6IP, STMN1, and ERAP1. F, Heatmap of the median expression of the 52 potential survival markers in the four proteomic subtypes. Color of protein name indicates classification: metabolism, black; signal transduction and rearrangement, orange; protein synthesis, blue; protein transport and synthesis, pink; peptidase activity, green; redox homeostasis, purple; other, brown. G–J, Loading plots of the multivariate analysis in survival days versus the proteome. K–N, Scatter plots of survival days versus expression of TYMP, PDCD6IP, STMN1, and ERAP1. Spearman correlation coefficients for the individual subtypes are depicted on the right-hand side of the graphs.

OF8 Clin Cancer Res; 2020 CLINICAL CANCER RESEARCH

Downloaded from clincancerres.aacrjournals.org on October 2, 2021. © 2019 American Association for Cancer Research. Published OnlineFirst December 17, 2019; DOI: 10.1158/1078-0432.CCR-19-1496

Proteomics of Pancreatic Ductal Adenocarcinoma Metastases

subtypes (r ¼ 0.14–0.33, Fig. 4K). PDCD6IP and STMN1 showed a by restricting dTMP pools, we evaluated the cell-cycle profile and fi ¼ fi – signi cant correlation with survival in the metabolic (r 0.82) and the observed a signi cant increase in S-phase and a decrease in G2 Min progenitor-like (r ¼0.55) subtypes, respectively (Fig. 4L and M). SHMT1 knockdown cells compared with control (Fig. 5H; Supple- The expression of ERAP1 inversely correlated with survival in the mentary Fig. S16). Overall SHMT1 expression, regardless of gemci- metabolic, progenitor-like, and proliferative subtypes (r ¼0.49 to tabine treatment, is higher in the metabolic and progenitor-like 0.33), but not in the inflammatory subtype (r ¼0.04, Fig. 4N). subtypes compared with proliferative or inflammatory subtypes These results demonstrate that subtype-based regression models are (Fig. 5I). This may indicate that individual PDAC subtypes could be better suited for identifying proteins associated with patient survival. more resistant to gemcitabine based on the inherent expression levels of the enzymes regulating the folic acid cycle. Gemcitabine treatment represses SHMT1 to promote drug resistance To validate the proteomic differences observed in our dataset Discussion are biologically meaningful, we evaluated the changes in the pro- Molecular subtyping of cancer can improve therapeutic out- teome associated with gemcitabine treatment. PDAC is a refractory comes by stratifying distinct subtypes of cancer into treatment cancer that readily develops resistance to chemotherapy, including groups based on their predicted response characteristics. There gemcitabine (40, 41). With death as the common endpoint, it is have been several different approaches to subtype PDAC using assumed that protein expression changes associated with gemcita- transcriptomics that prove this is not a singular disease and that bine resistance could be distinguished in the proteomic data from specific subtypes may exhibit unique response profiles to therapies. the RAP samples. Among the cohort of 56 PDAC donor samples, Furthermore, because of the metastatic nature of this disease clinical there are 6 donors that were treatment-na€ve and 9 donors treated subtyping of PDAC should incorporate metastatic characterization only with gemcitabine. A total of 63 proteins are upregulated and 44 to address the primary cause of patient mortality. This study proteins are downregulated significantly in the gemcitabine treat- represents the first proteomics-based subtype classification system ment group compared with the treatment-na€ve group (Fig. 5A; forPDACusinglivermetastasesthatcouldprovidethebasisfor Supplementary Fig. S15A; and Supplementary Table S13). These improving clinical therapy of this disease. differentially regulated proteins influenced by gemcitabine treat- The high-dimensional data obtained in this proteomics study ment have the potential to identify PDAC mechanisms of drug provides the ability to discern both complementary and unique resistance. PDAC subtypes. Previous PDAC proteomics studies have not been The folic acid cycle proteins cytosolic C-1-tetrahydrofolate synthase amenable in comparison with established PDAC subtypes due to (MTHFD1) and SHMT1 were significantly downregulated in PDAC the small number of tumors analyzed or the limited repertoire of liver metastases from patients treated with gemcitabine compared with proteins identifiedforanalysis(43).Therefore,thisisthefirst treatment-na€ve samples (Fig. 5B). MTHFD1 catalyzes the hydrolysis proteomics analysis of clinical PDAC samples that overcomes the of 5,10-methenyltetrahydrofolate into 10-formyltetrahydrofolate, limitations of previous studies to support a cross-comparison with while SHMT1 catalyzes the conversion of tetrahydrofolate (THF) and transcriptomic-based approaches for determining PDAC subtypes. the amino acid serine into 5,10-methenyltetrahydrofolate, the sub- The comparison of the proteomics-based subtypes with transcrip- strate required by MTHFD1 (Fig. 5C). The regulation of nucleotide tomic-based subtyping efforts by Moffitt and colleagues, Collisson pools, such as dCTP, is a mechanism of gemcitabine resistance in and colleagues, and Bailey and colleagues identifies significant PDAC cell lines (40). The folic acid cycle generates and recycles the concordance between these studies. However, the additional strat- metabolites required for the conversion of deoxyuridine monopho- ification of the squamous subtype into both the proliferative and sphate (dUMP) to deoxythymidine monophosphate (dTMP), neces- inflammatory subtypes suggests proteomics could be used as a sary to support DNA synthesis. Thymidylate synthase (TYMS) con- complementary method to identify additional PDAC subtypes. verts dUMP to dTMP and is inhibited by both 5-fluorouracil and Because the squamous subtype has been defined as a more aggres- gemcitabine metabolites leading to defects in DNA replication. In sive PDAC tumor, further stratification based on proteomics has the addition, TYMS expression is correlated with gemcitabine resis- potential to identify subtype-specific features that could impact tance (42). This evidence suggests gemcitabine-mediated regulation clinical response. of the folic acid pathway is important for the development of drug In our analysis, multiple predictive models were built to correlate resistance. the proteomics data with the clinical metadata, including gender, Both MTHFD1 and SHMT1 are downregulated in MIA PaCa-2 age, stage at diagnosis, number of primary or metastatic sites, PDAC cells conditioned with 10 nmol/L gemcitabine for 6 days history of diabetes, and treatment(s) administered. These models (Fig. 5D and E). To determine whether inhibition of SHMT1 expres- showed no significant correlation with the PDAC subtypes. How- sion is associated with PDAC response to gemcitabine, we established ever, PDAC subtypes exhibit a significant correlation with reported stable MIA PaCa-2 and Panc 10.05 cell lines with targeted knockdown PDAC modifiable risk factors of alcohol and tobacco usage, sug- of SHMT1 using two different shRNA constructs (Supplementary gesting these variables may influence the molecular pathogenesis of Fig. S15B and S15C). In MIA PaCa-2 cells, the EC50 of gemcitabine PDAC metastasis. Additional experiments are required to deter- increased from 2.7 nmol/L in the control cells to 17.9–19.9 nmol/L in mine the influence of alcohol and tobacco use on both the primary the SHMT1 knockdown cells (Fig. 5F; Supplementary Fig. S15D). PDAC and liver proteomes to delineate how these factors influence fi Similarly, Panc 10.05 displayed an increase in the gemcitabine EC50 subtype-speci c selection because it could impact patient response from 1.6 mmol/L in control to 9.4 mmol/L and 7.3 mmol/L in the to therapy. SHMT1 knockdown cells (Fig. 5G; Supplementary Fig. S15E), sug- Precision medicine approaches that exploit the unique molecular gesting that the reduced expression of this protein observed in vulnerabilities of PDAC subtypes could be envisioned to provide a gemcitabine-treated patients could act as a mechanism of drug resis- more robust clinical response. Surgical removal and FOLFIRINOX tance. Because the depletion of SHMT1 could prevent DNA synthesis chemotherapy are common treatment strategies for PDAC (44).

AACRJournals.org Clin Cancer Res; 2020 OF9

Downloaded from clincancerres.aacrjournals.org on October 2, 2021. © 2019 American Association for Cancer Research. Published OnlineFirst December 17, 2019; DOI: 10.1158/1078-0432.CCR-19-1496

Law et al.

A No Gemcitabine B treatment treated

P = 5.80 × 10-3 P = 0.0377 0.40.4 0.60.6

0.2 0.3

0 0 0

63 Proteins 0 upregulated -0.2 -0.3 −0.4-0.4 -0.6 -0.6 −0.6 (SHMT1 expression) 2 (MTHFD1 expression) 2 −0.8-0.8 -0.9 No treatment Gemcitabine- No treatment Gemcitabine-

treated Log treated Log No Gemcitabine No Gemcitabine

44 Proteins treatment treated treatment treated downregulated Subtype D C Folate 6 Days RFC/PCFT MIA PaCa-2 DMSO DNA DHF dTMP Synthesis THF DHFR 6 Days TYMS MTHFD1 10 nmol/L Gemcitabine SHMT1 dUMP Methylene-THF F MIA PaCa-2 P EC50 (nmol/L) shScr 2.726 E DMSO Gemcitabine -6 Replicate 123 123 shSHMT1 - #1 17.92 6.30 × 10 MTHFD1 shSHMT1 - #2 19.89 8.43 × 10-5 SHMT1

Actin G Panc 10.05 P EC (mmol/L) P = 0.0285 P = 0.0111 50 1.01 0.2 0.2 shScr 1.645 0.75 0.15 shSHMT1 - #1 9.364 0.0075 0.50.5 0.10.1 shSHMT1 - #2 7.268 0.0030 0.25 0.05 expression expression

0 0 0.0 SHMT1 Relative MTHFD1 Relative No treatment Gemcitabine- 0.0 No treatment Gemcitabine- DMSO Gemcitabinetreated DMSO Gemcitabinetreated I P = 2.10 × 10 -3 H P = 3.34 × 10 -7 P P = 0.0276 = 0.0223 P = 6.45 × 10 -6 P -4 P 50 = 2.28 × 10 = 0.00213 1 P = 2.72 × 10-8 P = 1.53 × 10-10 40 P 0.5 = 0.04 30 0 20 % of Cells −

10 SHMT1 Expression 0.5

0 G1 S G2 1 shScr shSHMT1 - #1 shSHMT1 - #2 Metabolic Proliferative Inflammatory Progenitor-like

Figure 5. A, Heatmap showing the expression of the 107 protein markers altered in gemcitabine-treated patients. The ribbon below the heatmap showed the proteomic subtype of the patients. Green, blue, red, and yellow represent the metabolic, progenitor-like, proliferative, and inflammatory subtypes, respectively. B, Bar chart showing the expression of MTHFD1 and SHMT1 in samples from nontreated and gemcitabine-treated patients. The t test P values are displayed at the top of the figure. C, MTHFD1 and SHMT1 involvement in the folic acid cycle. D, The experimental setup for generating gemcitabine-conditioned MIA PaCa-2 cells. E, Western blot and corresponding quantitative analysis of MTHFD1 and SHMT1 expression in MIA PaCa-2 cells without and with gemcitabine treatment. The t test P values are displayed at

the top of the figure. The EC50 and the corresponding F test P values for control shScr and two independent shRNAs targeting SHMT1 in MIA PaCa-2 (F) and Panc 10.05 (G) cells treated with gemcitabine. H, Cell-cycle analysis of shScr and shSHMT1 MIA PaCa-2 cell. I, Bar chart showing the expression of SHMT1 across different PDAC subtypes. t test P values are displayed in the figure.

OF10 Clin Cancer Res; 2020 CLINICAL CANCER RESEARCH

Downloaded from clincancerres.aacrjournals.org on October 2, 2021. © 2019 American Association for Cancer Research. Published OnlineFirst December 17, 2019; DOI: 10.1158/1078-0432.CCR-19-1496

Proteomics of Pancreatic Ductal Adenocarcinoma Metastases

Table 1. Demographics of patients grouped according to the proteomics subtype.

All patients Metabolic Progenitor-like Proliferative Inflammatory

Number of patients 56 7 21 11 17 Gender Male 39 6 10 9 14 Female 17 1 11 2 3 Age (SEM) 66.2 (1.5) 64.1 (5.5) 66.6 (2.7) 68.3 (3.8) 65.1 (1.9) Survival days (SEM) 333.1 (41.6) 283.4 (88.1) 362.8 (94.7) 252.5 (47.8) 369.0 (56.4) Stage at diagnosis IB 1 0 1 0 0 IIA 1 0 1 0 0 IIB 11 2 2 2 5 III 5 0 1 1 3 IV 37 5 15 8 9

Surgical removal is typically available to only 10%–15% of patients unique subtypes. The subtype-specific associations with response to with PDAC (45). FOLFIRINOX can improve patient survival, but chemotherapy observed in this study support the notion that the because of its toxicity FOLFIRINOX cannot be universally applied to unique features of each PDAC subtype should be incorporated at all all patients with PDAC (44, 46). While gemcitabine is still widely used levels of therapeutic development. in the clinic, FOLFIRINOX treatment as an adjuvant therapy following resection increases disease-free survival to 21.6 months compared with Disclosure of Potential Conflicts of Interest 12.8 months for gemcitabine at the cost of increased adverse No potential conflicts of interest were disclosed. effects (47). It is possible that the benefits of FOLFIRINOX or gemcitabine are restricted to certain PDAC subtypes. Recently, the ’ COMPASS trial also determined that patients with the basal subtype Authors Contributions are less likely to respond to first-line chemotherapy (48). Similarly, our Conception and design: H.C.-H. Law, M.A. Hollingsworth, N.T. Woods Development of methodology: H.C.-H. Law, D. Lagundzin, N.T. Woods analysis indicates the metabolic and progenitor-like subtypes display Acquisition of data (provided animals, acquired and managed patients, provided an increase in survival time in response to FOLFIRINOX þ Gem facilities, etc.): H.C.-H. Law, D. Lagundzin, F. Qiao, Z.S. Wagner, D. Costanzo- compared with gemcitabine, but this is not observed in the prolifer- Garvey, T.C. Caffrey, J.L. Grem, D.J. DiMaio, P.M. Grandgenett, L.M. Cook, ative and the inflammatory subtypes. In addition, the progenitor-like M.A. Hollingsworth, N.T. Woods subtype showed a significant benefit when the patients were treated Analysis and interpretation of data (e.g., statistical analysis, biostatistics, with gemcitabine and/or capecitabine. Although this analysis was computational analysis): H.C.-H. Law, D. Lagundzin, K.W. Fisher, F. Yu, M.A. Hollingsworth, N.T. Woods based on a small sample size, it is a proof-of-concept for the person- Writing, review, and/or revision of the manuscript: H.C.-H. Law, E.J. Clement, alized treatment of PDAC based on proteomic signatures using K.L. Krieger, T.C. Caffrey, K.W. Fisher, F. Yu, M.A. Hollingsworth, N.T. Woods traditional chemotherapeutics that could be readily implemented in Administrative, technical, or material support (i.e., reporting or organizing data, the clinic. constructing databases): H.C.-H. Law, F. Qiao, Z.S. Wagner, T.C. Caffrey, Ultimately, PDAC subtyping must be accomplished on clinically K.W. Fisher, N.T. Woods obtainable tissues to inform first-line cancer treatment. Two recent Study supervision: H.C.-H. Law, D. Lagundzin, N.T. Woods studies have demonstrated the transcriptomics-based PDAC subtyp- ing can be performed on percutaneous core biopsies (25, 48). However, Acknowledgments in both studies, the average time to return results based on RNA The authors thank the patients and their families for their participation in the sequencing was approximately 35–39 days, which could be used to UNMC Pancreatic SPORE Rapid Autopsy Program. The authors thank the UNMC fi Mass Spectrometry and Proteomics core facility for project support, the UNMC Flow inform second, but not rst-line therapy. For RNA-based subtyping, Cytometry core facility, and Dr. Jennifer Black and Dr. Amar Natarajan for their NanoString is a platform that could be used on a subset of transcripts helpful suggestions. This work was supported by the NIH grant numbers without the time-intensive steps required for RNA-seq library prep P20GM121316, P30CA036727, and 1P50CA127297. and data analysis (49). Proteomics could also provide a complemen- tary rapid assay platform that could be completed in several days (50). The costs of publication of this article were defrayed in part by the payment of page Moving forward, it will be important to establish protocols to obtain charges. This article must therefore be hereby marked advertisement in accordance and subtype PDAC samples in a clinically meaningful timeframe to with 18 U.S.C. Section 1734 solely to indicate this fact. inform first-line therapeutic decisions. This study provides further evidence that PDAC is not a single Received May 7, 2019; revised October 19, 2019; accepted December 11, 2019; disease and that quantitative proteomics can be used to delineate published first December 17, 2019.

References 1. Duggan MA, Anderson WF, Altekruse S, Penberthy L, Sherman ME. The 2. Iacobuzio-Donahue CA, Fu B, Yachida S, Luo M, Abe H, Henderson surveillance, epidemiology, and end results (SEER) program and pathology: CM, et al. DPC4 gene status of the primary carcinoma correlates with toward strengthening the critical relationship. Am J Surg Pathol 2016;40:e94– patterns of failure in patients with pancreatic cancer. J Clin Oncol 2009; e102. 27:1806–13.

AACRJournals.org Clin Cancer Res; 2020 OF11

Downloaded from clincancerres.aacrjournals.org on October 2, 2021. © 2019 American Association for Cancer Research. Published OnlineFirst December 17, 2019; DOI: 10.1158/1078-0432.CCR-19-1496

Law et al.

3. DiMagno EP, Reber HA, Tempero MA. AGA technical review on the epide- 28. David JM, Dominguez C, Hamilton DH, Palena C. The IL-8/IL-8R axis: a double miology, diagnosis, and treatment of pancreatic ductal adenocarcinoma. agent in tumor immune resistance. Vaccines 2016;4:pii: E22. Am Gastroenterol Assoc Gastroenterol 1999;117:1464–84. 29. Khawar IA, Park JK, Jung ES, Lee MA, Chang S, Kuh HJ. Three dimensional 4. Trede M, Schwall G, Saeger HD. Survival after pancreatoduodenectomy. 118 mixed-cell spheroids mimic stroma-mediated chemoresistance and invasive consecutive resections without an operative mortality. Ann Surg 1990;211: migration in hepatocellular carcinoma. Neoplasia 2018;20:800–12. 447–58. 30. Becker AE, Hernandez YG, Frucht H, Lucas AL. Pancreatic ductal adenocar- 5. Cancer Genome Atlas Research Network. Electronic address aadhe, Cancer cinoma: risk factors, screening, and early detection. World J Gastroenterol 2014; Genome Atlas Research N. integrated genomic characterization of pancreatic 20:11182–98. ductal adenocarcinoma. Cancer Cell 2017;32:185–203. 31. Hu D, Ansari D, Pawøowski K, Zhou Q, Sasor A, Welinder C, et al. Proteomic 6. Lehmann BD, Bauer JA, Chen X, Sanders ME, Chakravarthy AB, Shyr Y, et al. analyses identify prognostic biomarkers for pancreatic ductal adenocarcinoma. Identification of human triple-negative breast cancer subtypes and preclinical Oncotarget 2018;9:9789–807. models for selection of targeted therapies. J Clin Invest 2011;121:2750–67. 32. Won HS, Lee MA, Chung ES, Kim DG, You YK, Hong TH, et al. Comparison of 7. Prat A, Parker JS, Fan C, Perou CM. PAM50 assay and the three-gene model for thymidine phosphorylase expression and prognostic factors in gallbladder and identifying the major and clinically relevant molecular subtypes of breast cancer. bile duct cancer. BMC Cancer 2010;10:564. Breast Cancer Res Treat 2012;135:301–6. 33. Marangoni E, Laurent C, Coussy F, El-Botty R, Chateau-Joubert S, Servely JL, 8. Collisson EA, Sadanandam A, Olson P, Gibb WJ, Truitt M, Gu S, et al. Subtypes et al. Capecitabine efficacy is correlated with TYMP and RB1 expression in PDX of pancreatic ductal adenocarcinoma and their differing responses to therapy. established from triple-negative breast cancers. Clin Cancer Res 2018;24:2605– Nat Med 2011;17:500–3. 15. 9. Moffitt RA, Marayati R, Flate EL, Volmar KE, Loeza SG, Hoadley KA, et al. 34. Monypenny J, Milewicz H, Flores-Borja F, Weitsman G, Cheung A, Chowdhury Virtual microdissection identifies distinct tumor- and stroma-specific subtypes R, et al. ALIX regulates tumor-mediated immunosuppression by controlling of pancreatic ductal adenocarcinoma. Nat Genet 2015;47:1168–78. EGFR activity and PD-L1 presentation. Cell Rep 2018;24:630–41. 10. Bailey P, Chang DK, Nones K, Johns AL, Patch AM, Gingras MC, et al. 35. Duijvesz D, Burnum-Johnson KE, Gritsenko MA, Hoogland AM, Vredenbregt- Genomic analyses identify molecular subtypes of pancreatic cancer. Nature van den Berg MS, Willemsen R, et al. Proteomic profiling of exosomes leads to the 2016;531:47–52. identification of novel biomarkers for prostate cancer. PloS One 2013;8:e82589. 11. Puleo F, Nicolle R, Blum Y, Cros J, Marisa L, Demetter P, et al. Stratification of 36. Hashemi M, Yousefi J, Hashemi SM, Amininia S, Ebrahimi M, Taheri M, et al. pancreatic ductal adenocarcinomas based on tumor and microenvironment Association between programmed cell death 6 interacting protein insertion/ features. Gastroenterology 2018;155:1999–2013. deletion polymorphism and the risk of breast cancer in a sample of iranian 12. Collisson EA, Bailey P, Chang DK, Biankin AV. Molecular subtypes of pancreatic population. Dis Markers 2015;2015:854621. cancer. Nat Rev Gastroenterol Hepatol 2019;16:207–20. 37. Kuang XY, Chen L, Zhang ZJ, Liu YR, Zheng YZ, Ling H, et al. Stathmin and 13. Chen R, Dawson DW, Pan S, Ottenhof NA, de Wilde RF, Wolfgang CL, et al. phospho-stathmin protein signature is associated with survival outcomes of Proteins associated with pancreatic cancer survival in patients with resectable breast cancer patients. Oncotarget 2015;6:22227–38. pancreatic ductal adenocarcinoma. Lab Invest 2015;95:43–55. 38. Sun R, Liu Z, Wang L, Lv W, Liu J, Ding C, et al. Overexpression of stathmin is 14. Ansari D, Toren W, Zhou Q, Hu D, Andersson R. Proteomic and genomic resistant to paclitaxel treatment in patients with non-small cell lung cancer. profiling of pancreatic cancer. Cell Biol Toxicol 2019;35:333–43. Tumour Biol 2015;36:7195–204. 15. Noll EM, Eisen C, Stenzinger A, Espinet E, Muckenhuber A, Klein C, et al. 39. Cifaldi L, Romania P, Falco M, Lorenzi S, Meazza R, Petrini S, et al. ERAP1 CYP3A5 mediates basal and acquired therapy resistance in different subtypes of regulates natural killer cell function by controlling the engagement of inhibitory pancreatic ductal adenocarcinoma. Nat Med 2016;22:278–87. receptors. Cancer Res 2015;75:824–34. 16. Vizcaino JA, Deutsch EW, Wang R, Csordas A, Reisinger F, Rios D, et al. 40. Shukla SK, Purohit V, Mehla K, Gunda V, Chaika NV, Vernucci E, et al. ProteomeXchange provides globally coordinated proteomics data submission MUC1 and HIF-1alpha signaling crosstalk induces anabolic glucose metab- and dissemination. Nat Biotechnol 2014;32:223–6. olism to impart gemcitabine resistance to pancreatic cancer. Cancer cell 2017; 17. Hu W-F, Krieger KL, Lagundzin D, Li X, Cheung RS, Taniguchi T, et al. CTDP1 32:71–87. regulates breast cancer survival and DNA repair through BRCT-specific inter- 41. Vaz AP, Ponnusamy MP, Seshacharyulu P, Batra SK. A concise review on the actions with FANCI. Cell Death Discov 2019;5:105. current understanding of pancreatic cancer stem cells. J Cancer Stem Cell Res 18. Haridas D, Chakraborty S, Ponnusamy MP, Lakshmanan I, Rachagani S, Cruz E, 2014;2:e1004. et al. Pathobiological implications of MUC16 expression in pancreatic cancer. 42. Komori S, Osada S, Mori R, Matsui S, Sanada Y, Tomita H, et al. Contribution of PloS One 2011;6:e26839. thymidylate synthase to gemcitabine therapy for advanced pancreatic cancer. 19. Baryshnikova A. Spatial analysis of functional enrichment (SAFE) in large Pancreas 2010;39:1284–92. biological networks. Methods Mol Biol 2018;1819:249–68. 43. Cintas C, Douche T, Therville N, Arcucci S, Ramos-Delgado F, Basset C, et al. 20. Silva E, Souchelnytskyi S, Kasuga K, Eklund A, Grunewald J, Wheelock AM. Signal-targeted therapies and resistance mechanisms in pancreatic cancer: future Quantitative intact proteomics investigations of alveolar macrophages in sar- developments reside in proteomics. Cancers 2018;10:pii: E174. coidosis. Eur Respir J 2013;41:1331–9. 44. Tong H, Fan Z, Liu B, Lu T. The benefits of modified FOLFIRINOX for advanced 21. Lundstrom SL, Zhang B, Rutishauser D, Aarsland D, Zubarev RA. SpotLight pancreatic cancer and its induced adverse events: a systematic review and meta- Proteomics: uncovering the hidden blood proteome improves diagnostic power analysis. Sci Rep 2018;8:8666. of proteomics. Sci Rep 2017;7:41929. 45. Sahin IH, Askan G, Hu ZI, O'Reilly EM. Immunotherapy in pancreatic ductal 22. Knudsen ES, Balaji U, Mannakee B, Vail P, Eslinger C, Moxom C, et al. Pancreatic adenocarcinoma: an emerging entity? Ann Oncol 2017;28:2950–61. cancer cell lines as patient-derived avatars: genetic characterisation and func- 46. Conroy T, Desseigne F, Ychou M, Bouche O, Guimbaud R, Becouarn Y, et al. tional utility. Gut 2018;67:508–20. FOLFIRINOX versus gemcitabine for metastatic pancreatic cancer. N Engl J Med 23. Northcott JM, Dean IS, Mouw JK, Weaver VM. Feeling stress: the mechanics of 2011;364:1817–25. cancer progression and aggression. Front Cell Dev Biol 2018;6:17-. 47. Conroy T, Hammel P, Hebbar M, Ben Abdelghani M, Wei AC, Raoul JL, et al. 24. Liu Q, Liao Q, Zhao Y. Chemotherapy and tumor microenvironment of FOLFIRINOX or gemcitabine as adjuvant therapy for pancreatic cancer. N Engl J pancreatic cancer. Cancer Cell Int 2017;17:68-. Med 2018;379:2395–406. 25. Aguirre AJ, Nowak JA, Camarda ND, Moffitt RA, Ghazani AA, Hazar-Rethinam 48. Aung KL, Fischer SE, Denroche RE, Jang GH, Dodd A, Creighton S, et al. M, et al. Real-time genomic characterization of advanced pancreatic cancer to Genomics-driven precision medicine for advanced pancreatic cancer: early enable precision medicine. Cancer Discov 2018;8:1096–111. results from the COMPASS trial. Clin Cancer Res 2018;24:1344–54. 26. Hebert MD. Signals controlling Cajal body assembly and function. Int J Biochem 49. Brant R, Sharpe A, Liptrot T, Dry JR, Harrington EA, Barrett JC, et al. Clinically Cell Biol 2013;45:1314–7. viable gene expression assays with potential for predicting benefit from MEK 27. Cheah MT, Chen JY, Sahoo D, Contreras-Trujillo H, Volkmer AK, Scheeren FA, Inhibitors. Clin Cancer Res 2017;23:1471–80. et al. CD14-expressing cancer cells establish the inflammatory and proliferative 50. Doll S, Kriegmair MC, Santos A, Wierer M, Coscia F, Neil HM, et al. Rapid tumor microenvironment in bladder cancer. Proc Nat Acad Sci U S A 2015;112: proteomic analysis for solid tumors reveals LSD1 as a drug target in an end-stage 4725–30. cancer patient. Mol Oncol 2018;12:1296–307.

OF12 Clin Cancer Res; 2020 CLINICAL CANCER RESEARCH

Downloaded from clincancerres.aacrjournals.org on October 2, 2021. © 2019 American Association for Cancer Research. Published OnlineFirst December 17, 2019; DOI: 10.1158/1078-0432.CCR-19-1496

The Proteomic Landscape of Pancreatic Ductal Adenocarcinoma Liver Metastases Identifies Molecular Subtypes and Associations with Clinical Response

Henry C.-H. Law, Dragana Lagundzin, Emalie J. Clement, et al.

Clin Cancer Res Published OnlineFirst December 17, 2019.

Updated version Access the most recent version of this article at: doi:10.1158/1078-0432.CCR-19-1496

Supplementary Access the most recent supplemental material at: Material http://clincancerres.aacrjournals.org/content/suppl/2019/12/17/1078-0432.CCR-19-1496.DC1

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://clincancerres.aacrjournals.org/content/early/2020/01/22/1078-0432.CCR-19-1496. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.

Downloaded from clincancerres.aacrjournals.org on October 2, 2021. © 2019 American Association for Cancer Research.