Preoperative Metabolic Signatures of Prostate Cancer Recurrence

Subscriber access provided by LUNDS UNIV

Article

Preoperative Metabolic Signatures of Prostate Cancer Recurrence Following Radical Prostatectomy. Chaevien S. Clendinen, David A. Gaul, Maria Eugenia Monge, Rebecca S Arnold, Arthur S Edison, John A Petros, and Facundo M Fernandez J. Proteome Res., Just Accepted Manuscript • DOI: 10.1021/acs.jproteome.8b00926 • Publication Date (Web): 13 Feb 2019 Downloaded from http://pubs.acs.org on February 14, 2019

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 35 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Journal of Proteome Research

Preoperative Metabolic Signatures of Prostate Cancer Recurrence Following Radical Prostatectomy.

Chaevien S. Clendinen1, David A. Gaul1, María Eugenia Monge2, Rebecca S. Arnold3, Arthur S. Edison4, John A. Petros3,5, Facundo M. Fernández*1 1

School of Chemistry and Biochemistry, Georgia Institute of Technology, Atlanta GA 30332.

2

Centro de Investigaciones en Bionanociencias (CIBION), Consejo Nacional de Investigaciones

Científicas y Técnicas (CONICET), Godoy Cruz 2390, C1425FQD, Ciudad de Buenos Aires, Argentina. 3

Department of Urology, Emory University, Atlanta GA 30308

4

Department of Genetics and Biochemistry and Molecular Biology, Complex Carbohydrate

Research Center, University of Georgia, Athens GA 30602 5

Atlanta VA Medical Center, Atlanta GA 30033

*Corresponding author: Facundo M. Fernández, [email protected], Phone: +1 404-385-4432.

ACS Paragon Plus Environment

Journal of Proteome Research 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Abstract Technological advances in mass spectrometry (MS), liquid chromatography (LC) separations, nuclear magnetic resonance (NMR) spectroscopy, and big data analytics have made possible studying metabolism at an “omics” or systems level. Here, we applied a multi-platform (NMR+LC-MS) metabolomics approach to the study of preoperative metabolic alterations associated with prostate cancer recurrence. Thus far, predicting which patients will recur even after radical prostatectomy has not been possible. Correlation analysis on metabolite abundances detected on serum samples collected prior to surgery from prostate cancer patients (n=40 remission vs. n=40 recurrence) showed significant alterations in a number of pathways, including amino acid metabolism, purine and pyrimidine synthesis, tricarboxylic acid cycle, tryptophan catabolism, glucose, and lactate. Lipidomics experiments indicated higher lipid abundances on recurrent patients for a number of classes that included triglycerides, lysophosphatidylcholines, phosphatidylethanolamines, phosphatidylinositols, diglycerides, acyl carnitines, and ceramides. Machine learning approaches led to the selection of a 20-metabolite panel from a single preoperative blood sample that enabled prediction of recurrence with 92.6% accuracy, 94.4% sensitivity, and 91.9% specificity under cross-validation conditions.

2 ACS Paragon Plus Environment

Page 2 of 35

Page 3 of 35 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Keywords: Prostate cancer, biochemical recurrence, metabolomics, lipidomics, liquid chromatography mass spectrometry, nuclear magnetic resonance spectroscopy.



Introduction In the United States, prostate cancer (PCa) is the most prevalent non-cutaneous solid tumor in men, with approximately 200,000 new cases each year. PCa develops primarily in men over 65, with the average age of 66 at diagnosis.1 The most common treatment for men under 75 years old is radical prostatectomy (RadP)2, but up to 50% of men will still suffer from biochemical recurrence (BCR)3 manifested by detectable serum prostate specific antigen (PSA) levels even after prostate removal.4, 5 Attempts to predict BCR following RadP from clinical, laboratory and radiologic information available prior to surgery have been ongoing for over 20 years. As early as 1994 it was recognized that biopsy Gleason grade and initial PSA were strong predictors of a disease-free outcome following surgery.6 Although preoperative PSA levels alone have been found to be unreliable,7 the preoperative change in PSA over time (PSA velocity) has been successfully correlated with death from prostate cancer following surgery.8 Transcripts for PSA and prostatespecific membrane antigen (PSMA), in either blood or bone marrow, have also been found to be useful, but only moderate predictive power for BCR has been demonstrated.9 A PSA isoform known as 2pPSA has been used to predict final surgical pathology, but its correlation with BCR has not yet been translated to the clinic.9 Histone modifications in patients with low grade tumors do correlate with BCR but are less useful in patients with higher Gleason scores.10 The Cancer of the Prostate Risk Assessment (CAPRA) score, which combines PSA, Gleason sum, clinical tumor grade, percentage of positive biopsies, and patient age, has also been applied to BCR prediction, but has only yielded an area under the curve (AUC) of 0.78-0.81.11, 12 Other online nomograms combining patient age, PSA, primary and secondary Gleason pattern, clinical stage, and percentage of positive biopsy cores have made pre-operative prediction of post-operative outcomes more


Page 4 of 35

Page 5 of 35 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


accessible to patients.13 However, their use of single-institution data has considerably limited their generalizability and applicability.14 Metabolomics, the rapidly growing “omics” field examining metabolites in complex systems, provides sensitive and dynamic phenotypic patterns15 that more closely reflect cellular and molecular changes during cancer onset and progression.16 The metabolome is the total collection of small molecules in an organism with molecular weights lower than ~1500 Da. This includes endogenous molecules that are biosynthesized in “primary metabolism”, specialized “secondary metabolite” signaling molecules, molecules derived from diet or environmental exposure (the exposome), and molecules derived from microbial cells and host-microbiome interactions. No single analytical method can adequately profile all metabolites in a single experiment because of the vast chemical diversity of the metabolome and its large dynamic range (mM to fM).17 The most commonly used metabolomics technologies are Nuclear Magnetic Resonance (NMR) and Liquid Chromatography Mass Spectrometry (LC-MS). Though not often used together, partially due to the complexity of the generated datasets, notable benefits to fusing data from MS and NMR platforms have been reported.18-20 Literature evidence of detectable metabolic alterations associated with PCa21-28 suggests metabolomics could enable a more robust prediction of BCR following RadP than existing approaches. Metabolomics studies have already shown alterations in citrate, polyamine, choline, glycerophospholipids, and lactate levels during PCa progression.29, 30 Tissue sarcosine levels have also been shown to increase during PCa metastasis, but differences in urine levels were much less marked,31 with the role of sarcosine still being debated.23, 32-34 In previous work, we showed that serum metabolomic signatures are sensitive to the presence of PCa with improved accuracy, sensitivity, and specificity compared to the PSA test.24 Here, we apply a non-targeted multi-



platform metabolomics approach that fuses MS and NMR datasets to obtain metabolic fingerprints useful for understanding BCR and for distinguishing patients who will experience recurrence from those who would remain in remission following RadP. Over 600 metabolites were identified with different confidence levels as being altered in patients that recur, including numerous lipids. We provide strong evidence that patients’ metabolic phenotypes reflecting PCa-associated systemic metabolic changes can be used as robust BCR predictors.

Materials and Methods Chemicals Optima® LC-MS grade acetonitrile (ACN), Optima® LC-MS grade methanol (MeOH), Optima® LC-MS grade formic Acid (H2CO2), and ammonium hydroxide (NH4OH) (20-30%) were purchased from Fisher Chemical (Suwanee, GA, USA). Chromasolv® (Fluka) LC-MS grade MeOH and Burdick & Jackson LC-MS grade isopropanol (IPA) were purchased from Honeywell (NJ, USA) and used for sample preparation. Ammonium acetate (NH4OAc) was purchased from Sigma-Aldrich (St. Louis, MO, USA). Ammonium formate (NH4HCO2) was purchased from Fluka (Milwaukee, WI, USA). Ultrapure water with 18.2 MΩ cm resistivity (Thermo Scientific Barnstead Nanopure UV Ultrapure water system, Marietta, OH, USA) was used to prepare mobile phases. Quality control (QC) serum samples were from Sigma-Aldrich. D2O and 4,4-dimethyl-4silapentane-1-sulfonic acid (DSS) for NMR samples was obtained from Cambridge Isotope Laboratories (Andover, MA, USA).


Page 6 of 35

Page 7 of 35 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Serum Collection After Emory Institutional Review Board (IRB) approval, patients were consented and blood collected in SST Serum Separation Tubes: Hemogard (Fisher Scientific). Serum was separated by centrifugation at 1300 rcf for 20 min at 4 °C. Serum aliquots were stored in liquid nitrogen until removed for analysis. Information on male patients was compiled between 2000 and 2015 for advanced prostate cancer. The cohort consisted of a total of 80 male subjects: 10 African Americans and 70 Caucasian Americans. The average age was 57 for both groups. All patients had long-term post-operative follow up documenting the presence or absence of cancer recurrence. The median follow up time period was 50 months in patients with remission (range 41-80). Forty men had no recurrence after surgery, while 40 men had recurrence. Metadata on collected samples, including preoperative PSA levels, Gleason scores, pathological stages and TNM stages of all patients in this study are given in Table 1 and Figure S1.

Sample Preparation De-identified frozen serum samples were prepared in three randomized batches for protein precipitation. QC samples were prepared together with patient serum samples. All samples were thawed on ice. Individual samples were split into two subsets. (1) Ice-chilled 100% MeOH was added to the first serum sample subset in a 2:1 (MeOH:serum) volume ratio, vortex mixed for 1 minute and incubated in ice for 20 minutes. (2) Chilled 100% IPA was added to the second serum sample subset in a 3:1 (IPA:serum) volume ratio and vortex mixed for 1 minute. All samples were centrifuged in a refrigerated Beckman Coulter microfuge® 20R centrifuge at 15,000 rpm for 20 minutes. The supernatant was dried in a Labconco Refrigerated Centrivap concentrator with a chamber temperature setting at -4 ºC. Pooled samples were made by combining 1.5 µL of each



sample. MeOH-precipitated samples were split in two fractions (900 µL for NMR and 150 µL for Hydrophilic Interaction Liquid Chromatography (HILIC) experiments). For NMR experiments, dried sample pellets were re-suspended in deuterated sodium phosphate buffer with 0.1mM DSS in D2O and transferred to 5 mm NMR tubes. Samples designated for reverse phase liquid chromatography (RPLC)-MS experiments were re-suspended in H2O:MeOH 80:20 v/v. Samples to be analyzed using HILIC-MS were therefore reconstituted in 90% MeOH.

Reverse Phase Liquid Chromatography High Resolution Mass Spectrometry Samples were analyzed using a Dionex Ultimate 3000 LC system coupled to a Q-Exactive HF Orbitrap mass spectrometer with a resolving power of 240,000 FWHM and equipped with a Heated Electrospray Ionization (HESI) probe (260 °C). Samples were analyzed in both positive (+ve) and negative (-ve) ionization modes in technical duplicates. Compounds were separated using a Waters ACQUITY BEH C18 column (2.1 x 50 mm length, 1.7 µm particle size) with a column temperature of 40 °C and a flow rate of 400 µL min-1. The ultra-high performance liquid chromatography (UHPLC) analysis strategy was adapted from Damen et. al.35 Mobile phase A (60% ACN: 40% 10 mM NH4HCO2 in H2O with 0.1% H2CO2) and B (10% ACN: 90% IPA 10 mM NH4HCO2 with 0.1% H2CO2) were initially 60:40, respectively. The gradient method continued as follows: 0-1 min 55% A: 1-1.1 min 50% A; 1.15.0 min 45% A; 5.0-5.1 min 30% A; 5.1-8.0 min 1%A; 8.0-8.1 min 60% A; and 8.1-9.5 min held at 60% A. A 4-min wash run was done after every sample. Full MS spectra were acquired on all samples. Data-dependent acquisition (DDA) MS/MS was performed only on pooled samples.


Page 8 of 35

Page 9 of 35 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Hydrophilic Interaction Liquid Chromatography High Resolution Mass Spectrometry The HESI source was operated at a temperature of 260 °C. Samples were analyzed in both ionization modes. A Waters XBridgeTM BEH HILIC (2.1 x 75mm, 2.5 µm particle size) column was used with a temperature of 50 °C. Mobile phase A (5% ACN: 95% 10 mM NH4OAc in H2O with 0.05% NH4OH) and B (100% ACN with 0.05% NH4OH) were initially 5% and 95%, respectively at a flow rate of 300 µL min-1. The gradient method continued as follows: 0-7.0 min. 63% A; 7.0-7.2 min 5% A with an increased flow rate of 500 µL min-1; 7.2-9.9 min 5% A with a decreased flow rate of 300 µL min-1; and 9.9-10.0 min held at 5% A. Full MS spectra were acquired on all samples in technical triplicates. DDA MS/MS was performed only on pooled samples.

NMR Experiments All NMR experiments were performed on a Bruker Avance III HD 600 MHz system equipped with a triple resonance 5 mm cryoprobe and a chilled SampleJet sample changer. Onedimensional NOESY spectra with presaturation were collected with a 90° pulse, 2 sec relaxation delay, 0.1 sec mixing time, and a 2.72 sec acquisition time. A spectral width of 20 ppm (12019.23 Hz) was used. HSQC and HSQC-TOCSY were collected on the pooled sample and used to make tentative metabolite IDs. HSQC and HSQC-TOCSY were collected with a 90° pulse, 1.5 sec relaxation delay, a 0.15 sec acquisition time. Spectral widths for 13C and 1H were 199.9950 ppm (30182.070 Hz) and 12.0166 ppm (7211.539 Hz), respectively. HSQC was collected with 16 scans and 1024 increments. HSQC-TOCSY was collected with 16 scans and 1024 increments and a 0.09 second mixing time and a MLEV spin-lock sequence.



Data Processing and Analysis A total of 5 datasets [HILIC-MS (+ve), HILIC-MS (-ve), RP LC-MS (+ve), RP LC-MS (ve), and 1H NMR] were acquired for each sample. Background signals from sample and mobile phase blanks, and MS features with no detectable isotopic peaks were removed from the datasets as well as features with less than 0.08 min chromatographic peak widths. All LC-MS data were aligned, peak picked, de-isotoped, adduct analyzed, and normalized using the Progenesis QI software package (Nonlinear Dynamics, Newcastle, UK). The Progenesis QI database search was used to obtain a list of tentative IDs based on accurate mass (

Preoperative Metabolic Signatures of Prostate Cancer Recurrence

Recommend Documents