Proteomic Profiling of Lung Adenocarcinoma Indicates Heightened DNA Repair, Antioxidant

Proteomic Profiling of Lung Adenocarcinoma Indicates Heightened DNA Repair, Antioxidant Mechanisms and Identifies LASP1 as a Potential Negative Predictor of Survival

Johannes F. Fahrmann1*, Dmitry Grapov2*, Brett S. Phinney3, Carol Stroble4, Brian C. DeFelice1, William Rom5, David R. Gandara4, Yanhong Zang6, Oliver Fiehn1,7, Harvey Pass8, Suzanne Miyamoto4#

*EQUAL CONTRIBUTORS

1 University of California, Davis Genome Center, Davis CA

2 CDS Creative Data Solutions, Ballwin, MO

3 Genome Center Proteomics Core Facility, Davis CA

4 Division of Hematology and Oncology, Department of Internal Medicine, School of Medicine, University of California, Davis Medical Center, Sacramento, CA 91817

5 Division of Thoracic Surgery, Department of Cardiothoracic Surgery, Langone Medical Center, New York University, New York City, NY, USA

6Department of Pathology and Laboratory Medicine, University of California, Davis Medical Center, Sacramento CA 95817

7 Department of Biochemistry, Faculty of Sciences, King Abdulaziz University, Jeddah21589, Saudi-Arabia

8Division of Pulmonary, Critical Care, and Sleep, NYU School of Medicine, New York, NY, USA

Running Title: Proteomic Perturbations in Early Stage Lung Adenocarcinoma

#Corresponding Author:

Suzanne Miyamoto

Division of Hematology and Oncology, Department of Internal Medicine

4501 X Street, Suite 3016, Sacramento, CA 95817

Phone: 916-734-3769

Fax: 916-734-5356

Supplemental Materials and Methods

1a.LC-MS/MS Analysis

Samples were block randomized into 8 blocks of 4 normal and 4 tumor samples in each block. Three technical replicates were acquired for each sample resulting in 192 total LC-MS/MS analyses. LC separation was done on a Waters Nano Acquity UHPLC (Waters Corporation) with a Proxeon nanospray source.

The digested peptides were reconstituted in 2% acetonitrile/0.1% trifluoroacetic acid and roughly 3µg of each sample was loaded onto a 100 micron x 25 mm Magic C18 100Å 5U reverse phase trap where they were desalted online before being separated on a 75 micron x 150 mm Magic C18 200Å 3U reverse phase column. Peptides were eluted using a gradient of 0.1% formic acid (A) and 100% acetonitrile (B) with a flow rate of 300nL/min. A 120 minute gradient was ran with 5% to 35% B over 100 minutes, 35% to 80% B over 8 minutes, 80% B for 1 minute, 80% to 5% B over 1 minute, and finally held at 5% B for 10 minutes. Each of the gradients was followed by a 1h column wash.

Mass spectra was collected on an Orbitrap Q Exactive Plus mass spectrometer (Thermo Fisher Scientific) in a data-dependent mode with one MS precursor scan followed by 15 MS/MS scans. A dynamic exclusion of 15 seconds was used. MS spectra were acquired with a resolution of 70,000 and a target of 1 × 106 ions or a maximum injection time of 30ms. MS/MS spectra were acquired with a resolution of 17,500 and a target of 5 × 104 ions or a maximum injection time of 50ms. Peptide fragmentation was performed using higher-energy collision dissociation (HCD) with a normalized collision energy (NCE) value of 27. Unassigned charge states as well as +1 and ions >+5 were excluded from MS/MS fragmentation.

1b.Protein Identification

Raw mass spectrometry files were converted to mzML using MSConvert and peptide spectrum matching (PSM) was performed using MS-GF+ (version 1.0072) and Myrimatch (version 2.2.140) against the NCBI Refseq human database (72,090 sequences, October 2014) plus 110 additional sequences of common laboratory contaminants from the cRAP protein database ( plus an equal number of reverse decoy sequences. Enzyme specificity was set to Trypsin/P with dynamic modifications of deamidation of NQ and oxidation of M. A static modification was set to Carbamidomethyl C. Precursor mass tolerance was set to 10 ppm for Myrimatch and 20 ppm for MS-GF+, fragment ion tolerance was set to 20 ppm for both search engines. Myrimatch and MS-GF+ identifications were combined and PSM’s were assembled into proteins using IDPicker 3. Proteins were assembled on the gene level and IdPicker3 parameters were set to the following. Max Q value = 0.005, minimum peptides per gene = 2, minimum spectra per gene = 4. Minimum spectra per match =1, Max Protein Groups per Peptide =10. These parameters resulted in 79,775 distinct peptide matches corresponding to 4,512,957 spectra assembled into 6,796 gene groups and 10,712 protein groups. A decoy false discovery rate was calculated on the protein, peptide and spectrum level. Protein FDR= 6.6%, Peptide FDR = 5%, Spectrum FDR = 0.29%.

1c. Immunohistochemistry of TMA sections

Briefly, slides were deparaffinized and pretreated with Heat-Induced-Epitope-Retrieval solution (pH 9) (DAKO). Slides were preincubated in Dual Endogenous Enzyme Block(DEEB) blocking solution in a DAKO AutostainerLink48. Polyclonal anti-human HYOU1 (1:100, HPA049296, SIGMA), EPRS (1:100, HPA030052, SIGMA) and LASP1 (1:500, HPA012072, SIGMA) were incubated for 30 minutes and then washed, which was followed by detection by EDL solution (DAKO EnVision+ Dual Link System-HRP). Hematoxylin was used as the counterstain.

1d. Statistical Analysis

Replicate precision as relative standard deviation (RSD) was compared between protein spectral counts and area under the curve measurements. Protein count data displayed lower median RSD for the majority of measured proteins and was selected for all subsequent analyses. Count data for 10712 proteins was evaluated for missing or zero values, replicate precision, protein sequence coverage and annotation to know genes. Robust measurements were retained based on median counts > 0 for control or malignant tissue and protein sequence coverage > 60% were used to limit interpretation to only the most abundant, robust and reproducible proteomic measurements. Analysis of replicate precision was used to filter multiple measurements mapping to the same gene to select variables with lowest RSD. Data normalization strategies were compared based on RSD. Mean normalized count data (each sample count expressed as a ratio to the total counts per sample) yielded lowest RSD and was selected for further analyses. Based on the outlined strategy, 799 high quality protein measurements or 14% of the initially measured features were selected for further statistical and multivariate analyses.

Orthogonal partial least squares discriminant analysis (O-PLS-DA) multivariate models were used to identify the top 10% of all protein discriminants between tumor and control tissues. O-PLS-DA modeling was conducted on shifted natural logarithm transformed age, gender and packs adjusted mean normalized protein counts. Double cross-validation was used for model optimization and validation. One third of the sample set (based on patients) was held out for a test set. The remaining two thirds of the data were used for feature selection based model optimization. Filter based feature selection on full model variable loadings was used to identify the optimal proteomic feature set based on Monte Carlo cross-validated (MCCV) (100 replications) model area under the receiver operator characteristic curve (AUC), sensitivity (Sens) and specificity (Spec) (19).

A Gaussian graphical model protein-protein empirical network was calculated for O-PLS-DA selected top discriminants (n= 16) as previously described [16]. Partial correlations were calculated using q-order partial correlation analysis of shifted natural logarithm transformed covariate adjusted and mean normalized protein counts [16]. Conditionally independent protein-protein associations were identified based on false-discovery rate (FDR) adjusted p-value <0.05.

Supplemental Tables

Table S2. Summary of changes for significantly (pFDR <0.05) differential proteins between cancer and control tissue.

Table S3. OPLSDA Model Validation Statistics.

Table S4. Associations between LASP1 and lymphovascular infiltration and overall survival.

Table S5. Patient Characteristics for Oncomine Okayama Lung Study.

Table S2. Summary of changes for significantly (pFDR <0.05) differential proteins between cancer and control tissue.

Gene ID / Gene Name / Tumora / Normala / Fold Changeb / p-valuec / Ratiod / Percent (%)e
SPTB / spectrin, beta, erythrocytic / 2.82 +/- 2.6 / 10.7 +/- 9.3 / 0.3 / 6.7E-07 / / 76

a Values represent covariate adjusted mean spectral counts ± StDev