In silico phosphoproteomic study: discrepancy of DCA effectiveness as a possible pyruvate dehydrogenase kinase inhibitor in the treatment of breast and ovarian cancers
Abdelrahman El-Sayed (Mentor: Dr. Karen Ross, Department of Biochemistry and Molecular & Cellular Biology, Georgetown University)
September 1st, 2015, 3:20-3:40pm, Room 1300, Harris Building
High-throughput proteomics initiatives such as the HUPO Human Proteome Project and the Clinical Proteomic Tumor Analysis Consortium (CPTAC) are generating a wealth of experimental data on the proteins and protein post-translational modifications (PTMs) present in healthy and diseased human tissues. As a proof of principle, we analyzed multiple data types from the CPTAC/TCGA project together with Protein Ontology (PRO) representation of proteoforms to explore phosphorylation of pyruvate dehydrogenase (PDHA1) and its regulation in breast and ovarian cancer tumor samples.
Pyruvate dehydrogenase complex (PDC) is a complex of three enzymes that convert pyruvate into acetyl-CoA which may then be used in the citric acid cycle to carry out cellular respiration. E1 is the most important component of the complex which is a heterotetramer of two E1α subunits (PDHA1 and PDHA2) and two E1β subunits. In eukaryotes, PDHA1 is tightly regulated by its own specific pyruvate dehydrogenase kinases (PDK1-4) and pyruvate dehydrogenase phosphatases (PDP1, PDP2), deactivating and activating it respectively. PDKs can phosphorylate a serine residue on PDHA1 subunit at three possible sites (Ser 232, 293 and 300, referred to as Site 3, Site 1, and Site 2 respectively). Restoration of PDHA1 activity by inhibiting its phosphorylation and subsequently restoring oxidative phosphorylation could have anti-cancer effect leading to apoptosis.
The drug dichloro-acetate (DCA) inhibits PDKs binding to the lipoyl (E2) domain of the PDC, interfering with PDK access to its substrate. PDK2 unlike PDK3 is highly susceptible to inhibition by DCA. In vivo and in-vitro models have shown that DCA is more effective against some cancer types such as colo-rectal and ovarian cancer if compared with some others like breast cancer. In our study we explored whether this discrepancy could be explained by differing importance of PDHA1 regulators in different cancer types.
We started our study with 108 breast cancer and 69 ovarian cancer samples. Data on PDHA1 regulator was collected from different bioinformatics resources. mRNA expression levels based on TCGA data were obtained from the c-bioPortal database, protein and phosphoproteome (phosphorylated proteins) abundances from CPTAC. We used the Protein Ontology (PRO) to represent the proteoforms and the relationships between them. Based on the literature, we defined three proteoforms where each individual site is phosphorylated and phosphorylation on the other two sites is unknown. Relying on the fact that site 1 and site 2 are close together, we used the peptides observed by CPTAC that span both sites to define 3 additional proteoforms describing combinations of phosphorylation sites.
We clustered our samples based on the abundance of four PDHA1 regulators (PDK1-3 and PDP1). From our cluster analysis we observed that clusters with the highest PDK1 and PDK3 levels have the highest relative abundance of the doubly phosphorylated peptide. We also performed multi-variate linear regression analysis and we found that breast cancer samples have two statistically significant models in which PDK3 has the largest positive statistically significant coefficients. PDK2 also has a statistically significant coefficient but surprisingly the coefficients were negative. On the other hand, we noticed generally that PDK2 has the highest correlation coefficients with ovarian cancer peptide phosphorylation whereas PDK3 has mostly negative ones.
We concluded from our analysis that PDK3 is the major contributor to the abundance of doubly phosphorylated peptide and PDP1 may play a role in regulating abundance of the pS300 form in breast cancer. On the other hand, when we performed a similar analysis on the ovarian data, we did not see these association with PDK3 and PDP1; instead we found a possible association with the pS293 form and PDK2 abundance. Given that PDK2 is more susceptible to inhibition by DCA than PDK3, this may help explain the observations about DCA effectiveness in breast and ovarian cancer.