| Metadata |
| datasetIdentifier | PASS00912 |
| datasetType | SRM |
| submitter | Kenneth Williams <kenneth.williams@yale.edu> |
| submitter_organization | Yale University School of Medicine |
| lab_head_full_name | Chirag R. Parikh |
| lab_head_email | chirag.parikh@yale.edu |
| lab_head_organization | Yale University School of Medicine |
| lab_head_country | United States |
| datasetTag | TUPA0002 |
| datasetTitle | Use of a Targeted Urine Proteome Assay (TUPA) to Identify Protein Biomarkers of Delayed Recovery After Kidney Transplant |
| publicReleaseDate | 2017-07-22 00:00:00 |
| finalizedDate | 2016-07-22 11:26:16 |
| summary | Purpose: Development of delayed graft function (DGF) following kidney transplant is associated with poor outcomes. An ability to distinguish patients, promptly after surgery, with DGF versus immediate graft function (IGF) based on the differential expression of urine protein biomarkers may facilitate treatment and the research needed to improve prognosis. The purpose of this study was to use a Targeted Urine Proteome Assay (TUPA), which was described in Cantley et al, 2016 (PMID: 26220717, PMCID in progress), to identify protein biomarkers of delayed recovery from kidney transplant.
Experimental Design: Potential biomarkers were identified by using the TUPA Multiple Reaction Monitoring (MRM) assay to interrogate the relative IGF/DGF levels of expression of up to 167 proteins in urine from 21 DGF, 15 SGF (slow graft function), and 16 IGF patients. An iterative Random Forest analysis approach evaluated the relative importance of each biomarker, which was then used to identify an optimum biomarker panel that provides the maximum sensitivity and specificity with the least number of biomarkers.
Conclusions and clinical relevance: With a sensitivity of 82.6%, specificity of 77.4% and AUC of 0.891, the “Top 4” proteins have significant ability to identify DGF. This panel represents an important step towards identifying DGF at an early stage so that more effective treatments can be developed to improve long term graft outcomes. |
| contributors | Kenneth R. Williams
Christopher M. Colangelo
Lin Hou
Lisa Chung
Justin Belcher
Thomas Abbott
Jennifer L. Cantley
Isaac Hall
Hongyu Zhao
Lloyd G. Cantley
Chirag R. Parikh |
| publication | Williams,KR,Colangelo,CM,Hou,L,Chung,L,Belcher,J, Abbott,T, Cantley,JL,Hall,I,Zhao,H,Cantley,LG, and Parikh,CR, Use of a Targeted Urine Proteome Assay (TUPA) to Identify Protein Biomarkers of Delayed Recovery After Kidney Transplant, Proteomics Clinical Applications, in preparation |
| growth | |
| treatment | Patient Cohorts
Patients from the multicenter kidney transplant cohort were categorized as either immediate graft function (IGF), slow graft function (SGF), or delayed graft function (DGF). IGF patients were those whose serum creatinine levels decreased by at least 70% within 7 days following their kidney transplant while those patients requiring dialysis during the first 7 days were defined as having DGF. Unfortunately, there are no well-established, universally accepted quantitative criteria for dialyzing patients post-operatively. Common indications for dialysis during the first 7 post-operative days included hyperkalemia, volume overload and acidosis. SGF were defined as patients whose serum creatinine levels did not fall by 70% within 7 days but who also did not require dialysis during that period of time. The 52 patients in this study included 16 IGF, 15 SGF, and 21 DGF that had been enrolled at three sites. To better assess for possible markers associated with recovery from injury, all patients were required to have urinary NGAL levels >400 ng/mL, reflecting significant structural injury, at 0 or 6 hrs following transplant. |
| extraction | Urine Sample Preparation
A fresh 10-ml urine sample was collected 12-18 hours post-transplant from each patient via Foley catheter tubing. Given the intention of identifying markers of recovery and repair, the 12-18 hour time frame was chosen as opposed to immediately post-operative. Urine samples were immediately refrigerated and then centrifuged at 5000 x g for 10 minutes at +4°C. Aliquots of 1 ml of supernatant were subsequently stored within 6 hours of collection in cryovials at -80°C. No additives or protease inhibitors were utilized. All analyses were performed on frozen aliquots that had not undergone any additional freeze-thaw cycles.
Urine samples were “cleaned-up” prior to dual enzyme digestion and LC-MRM analysis by using a chloroform/methanol precipitation procedure that began by adding 400 µl of methanol to 100 µl of each urine sample. After vortexing, 100 µl of chloroform was added followed by 300 µl of water. Samples were vortexed and spun at 14,000g for 1 minute. The top aqueous layer was removed and an additional 400 µl of methanol was added. After a 2 minute spin at 14,000g, the methanol was removed carefully without disturbing the pellet that was then dried in a Speedvac. |
| separation | LC-MRM with the Targeted Urine Proteome Assay (TUPA)
Each MRM scan was carried out by loading 1 µg (5 μl of 0.2 µg/μl) sample together with either 500 or 625 fmol SIS peptides (as indicated in the sample description column in the accompanying listing of files) that had been spot synthesized by JPT Peptide Technologies. Samples were loaded onto a 180 μm x 20 mm 5 μm Symmetry C18 nanoACQUITY trapping column with 2% acetonitrile (ACN)/0.1% formic acid (FA) at 15 μl per min for 3 min. After trapping, a 2-40% 60 min linear ACN/0.1% FA gradient was run at a flow rate of 500 nl/min with a 75 μm x 150 mm 1.7 μm BEH130 C18 nanoAcquity column. As indicated in the accompanying listing of files, each of the 52 samples was analyzed non-sequentially in triplicate. |
| digestion | Dual Enzyme Digestion
Each of the chloroform/methanol precipitated and dried samples were dissolved in 25 µl 8 M urea, 0.4M ammonium bicarbonate. The pH was verified to be ~8.0 and the proteins were reduced by adding 2.5 µl 45 mM dithiothreitol followed by 20 min incubation at 37şC and were then alkylated by adding 2.5 µl of 100 mM iodoacetamide followed by 20 min incubation in the dark at room temperature. After adding water to reduce the urea concentration to 2 M, the samples were digested first by adding 5 µl of 1 mg/ml Lysyl endopeptidase and incubating for 5 hours at 37°C and then by adding 5µl of 1 mg/ml trypsin and incubating at 37°C for 16 hours. Amino acid analysis was performed as described below on aliquots of each digest so that equal amounts of each sample could be subjected to LC-MRM analysis. |
| acquisition | Extended Multiple Reaction Monitoring (xMRM)
As described previously in Cantley et al (2016), TUPA uses a scheduled LC-MRM assay that was developed using a previously described workflow [23] that was carried out on an AB SCIEX 5500 Q-TRAP to potentially quantify the level of expression of 167 proteins in human urine based on 896 peptide transitions in TUPA that were determined to be free of interference (Supporting Table 2 ) and that include two transitions each from 224 naturally occurring “light” and matching “heavy” stable isotope labelled internal peptide standards (SIS). Except where indicated in Supporting Table 2, the MRM method used peak windows of 5 min and a cycle time of 2.5 sec. In the case of hydrophilic peptides that elute before 20 minutes, the peak window was extended to 400 seconds to allow for the higher variability in the retention times of these peptides. The xMRM assay continuously monitors the primary (1) transition while the secondary transition (2) is only monitored when the primary transition reaches above the threshold value that was set at a value of 200 to be above the noise level. Additionally, the xMRM assay was run with the “timeslip” feature that extends the scheduled MRM window by an additional half window (2.5 minutes) if the peak area threshold is greater than 200 at the end of the first window. All transitions were weighted equally. The resulting LC-xMRM data was processed and quantitated using AB Sciex Multiquant™ 2.1 software (research version). The SignalFinder™ 2 algorithm (research version) was used to integrate and score peak groups. |
| informatics | Statistical Analyses
The MRM data is reported as the ratio of the relative level of expression of each protein in the TUPA assay relative to the internal standard in IGF versus DGF samples. The relative level of expression of each protein was calculated by first determining the light to heavy peak area ratio for each transition pair in each of the three technical replicates for each sample. These data were then used to calculate the median transition peak area ratio for each sample. The protein fold-change in each biological sample was then calculated from the mean of the “light”/”heavy” intensity ratios for each of the transitions that were used to interrogate the level of expression of each protein. The protein fold-changes for each group of DGF, SGF, and IGF samples were then calculated from the average protein fold-changes for the respective biological samples. To control the false discovery rate the p values for the protein fold-changes were adjusted using Benjamini-Hochberg's approach.
Coefficient of Variations (CVs) were calculated at the transition level by first filtering out heavy transition peaks with S/N<5 and light transition peaks with S/N<3. For those sets of technical replicates with two or three quantified peaks the CVs were calculated across the two or three technical runs for each transition, for each biological sample. The resulting CVs were then averaged over the biological samples within the DGF, SGF, and IGF groups to determine an average technical level CV.
Random Forest (RF) analyses were carried out by generating 10,000 random trees using an R-script based on the approach described by Breiman (2001). To begin each RF analysis, 25 of the 37 (IGF + DGF) samples were randomly selected as the training set with the remaining 12 samples then serving as the test set. This procedure was repeated 10x and the resulting average and standard deviation data for the correctly and incorrectly classified samples in the test sets were then compiled and used to calculate the corresponding sensitivity and specificity. To estimate their relative importance, the average Mean Decrease in Accuracy (MDA) of each transition was determined for the 10 RF analyses and the protein level MDA was calculated from the mean of the transition level MDAs that were used to interrogate the level of expression of each protein. The RF algorithm calculates MDAs by permuting the values of each transition and then determining how much the permutation decreases the accuracy of the resulting classification. Hence, for unimportant transitions, the permutation should have little to no effect on the resulting classification and the resulting MDA values will be small. In contrast, permuting important transitions should significantly decrease the accuracy of the resulting classification and result in larger MDA values. The average AUCs (Area Under the Curves) for the Receiver Operator Characteristic (ROC) curves were calculated with the performance function in the ROCR package from the mean of the 10 RF analyses that were carried out for each set of biomarkers. |
| instruments | AB Sciex 5500 QTRAP |
| species | Human |
| massModifications | C+57.021464 |
