Systems pharmacology dissection of cholesterol regulation reveals determinants of large pharmacodynamic variability between cell lines
In individuals, heterogeneous drug response phenotypes result from a complex interplay of dose, drug specificity, genetic background, and environmental factors, thus challenging our understanding of the underlying processes and optimal use of drugs in the clinic. Here, we use mass spectrometry-based quantification of molecular response phenotypes and logic modeling to explain drug response differences in a panel of cell lines. We apply this approach to cellular cholesterol regulation, a biological process with high clinical relevance. From the quantified molecular phenotypes elicited by various targeted pharmacologic or genetic treatments, we generated cell-line-specific models that quantified the processes beneath the idiotypic intracellular drug responses. The models revealed that in addition to drug uptake and metabolism further cellular processes displayed significant pharmacodynamic response variability between the cell lines, resulting in cell-line-specific drug response phenotypes. This study demonstrates the importance of integrating different types of quantitative systems-level molecular measurements with modeling to understand the effect of pharmacological perturbations on complex biological processes.
Sample Processing Protocol
Experiment: The experiments were performed in the following manner: Cells were seeded into 6-well dishes at a density that the cells reached 90% confluence when harvested after 2-3 days. For phosphopeptides enrichment a 15cm dish was used to cultivate cells. The cells were treated one day after seeding with drugs or siRNAs. For the drug treatment, the drugs were dissolved prior in DMSO (Atorvastatin, GW3965, T090137) or EtOH (25-Hydroxycholesterol) and (0.1% v/v) DMSO and EtOH was added to a separate well as control treatments. Hence, in total three negative controls were performed for the drug perturbations: untreated, 0.1% DMSO treated, and 0.1% EtOH treated cells. Cells treated with LPDS were washed twice with warm phosphate-buffered saline before changing to medium containing 10% LPDS instead of FBS. The cells were harvested 48h after adding the drug. For the siRNA treatment, SilencerSelect® siRNAs were dissolved prior in H2O and transfected using Lipofectamine RNAiMax according to the manufacturer protocol to obtain an end concentration of 5nM. As control samples, cells were treated with an siRNA targeting no coding gene (called sNeg9) or cells were mock transfected with transfection reagent and water instead of siRNA (mock). Hence, three negative controls were performed per experiment: untreated, mock, sNeg9. SiRNA transfected cells were harvested 72h post transfection. Samples from the same cell line were harvested together (blocking) but the order of harvesting the different wells was changed between different biological replicate to not introduce a possibly systematic bias. All treatments resulted in comparable cell growth except for HEK293 cells treated with 10uM atorvastatin and HEK cells treated with LPDS + 5µM atorvastatin. In addition, the HEK293 cells treated with LPDS + 1µM atorvastatin in 15cm dishes for phosphoenrichment were detached after 48h. These samples were therefore not harvested. Proteomic sample processing: Cells were harvested for proteomics measurements by washing once with ice-cold phosphate-buffered saline, scraping off the cells, aspirating the phosphate buffer-saline and freezing the cell pellet in liquid nitrogen. Cells were lysed using 8 M Urea in 100 mM Ammonium Bicarbonate with the help of sonication for 10 min . The lysate was reduced using 2.5 mM tris(2-carboxyethyl)phosphine (TCEP) for 30min at 37°C and alkylated using 40mM Iodacetamide for 45 min at 25°C in the dark. The protein amount was measured using the Bicinchoninic acid (BCA) assay and 60 µg protein was digested with LysC (1:100) for 4 h and Trypsin (1:75) over night. Samples were diluted to 6 M and 1.5 M Urea in 100mM Ammonium Bicarbonate using 100 mM Ammonium Bicarbonate for digestion with LysC or Trypsin respectively. The digestion was stopped by adding TFA until a pH~2-3 was reached. The digested peptides were desalted using C18-columns, washed with 2% acetonitrile and 0.1% trifluoroacetic in H2O, eluted with 50% acetonitrile and 0.1% trifluoroacetic acid in H2O and subsequently dried in a speedvac. The dried peptides were dissolved in 2% acetonitrile and 0.1 formic acid in H2O and iRT peptides were added to the sample. Phosphoproteomic sample processing: Cells were harvested for proteomics measurements by washing once with ice-cold phosphate-buffered saline, scraping off the cells and freezing the cell pellet in liquid nitrogen. Cells were lysed and digested using 8 M Urea in 100 mM Ammonium Bicarbonate and sonication. The lysate was reduced using 5 mM TCEP for 30 min at 37°C and alkylated using 10 mM Iodacetamide for 45 min at 25°C in the dark. The protein amount was measured using a BCA Assay and 1mg proteins was used for digestion with LysC (1:150) for 4h and Trypsin (1:75) over night. The digested peptides were purified using C18-columns, washed with 2% acetonitrile and 0.1% trifluoroacetic acid in H2O, eluted with 50% acetonitrile and 0.1% trifluoroacetic acid in H2O and subsequently dried in a speedvac. The dried peptides were then dissolved in loading buffer for phosphoenrichment (6% trifluoroacetic acid and 80% ACN in H2O) and incubated for 60min with 1.25 mg TiO2 beads. The beads were washed twice with loading buffer, twice with buffer C (80% acetonitrile, 0.1% trifluoroacetic acid in H2O), twice with buffer D (50% acetonitrile and 0.1% trifluoroacetic acid in H2O) and then twice with buffer E (0.1% trifluoroacetic acid in H2O). The phosphopeptides were eluted with 0.3 M Ammonium hydroxide pH 10.5 and afterwards re-acidified immediately to pH 2-3. Afterwards the phosphopeptides were purified on C18-columns as described before and after drying dissolved in 2% acetonitrile and 0.1% formic acid in H2O and iRT peptides were added to the sample.
Data Processing Protocol
Proteomics: The data was analyzed using a pipeline configured on the iPortal platform in the lab (Kunszt et al., 2015). The raw SWATH wiff files were converted using ProteoWizard (version 3.0.5533) to profile mzXML files (Kessner et al., 2008). The extraction of the data was performed using the OpenSWATH workflow (Röst et al., 2014) and the combined human assay library (Rosenberger et al., 2014). An m/z fragment ion extraction window of 0.05 Th, an RT extraction window of 600 s and a set of 10 different scores were used. To match features between runs, detected features were aligned using an spline regression with a target assay FDR of 0.01 (Rost et al., 2016). The aligned peaks were allowed to be within 3 standard deviations or 60s after retention time alignment. For runs where no peak was identified the area was requantified using the single shortest Path method (Rost et al., 2016). The data was then processed using the R/Bionductor package SWATH2stats (Blattmann et al., 2016). Precursors had to pass an m-score threshold of 1E-05 in at least 20% of the 291 files to be selected for further analysis. These threshold resulted in an estimated precursor FDR of 0.0025, peptide FDR of 0.002745 and protein FDR of 0.0140 (using an estimated fraction of false targets (FFT) or π0-value of 0.6 for estimating the FDR). In total 24’266 peptides and 111 decoy peptides passed this stringent threshold. Subsequently, only proteotypic peptides and the 7 peptides with the highest signal per protein were selected for quantitative analysis. This resulted in a data matrix containing 4.5 106 peakgroup intensities, from which 78% of peakgroups had an m-score of < 0.01. The data was normalized using a local total intensity normalization within a retention time window of 10 min and analyzed for differential expression using mapDIA v1.2.1 (Teo et al., 2015). Differential expression was tested using an independent study design with the settings of selecting a minimum correlation of 0.25, a standard deviation factor of 2, between 3-5 fragments per peptide and at least one peptide per protein. Phosphoproteomics: In total 16 different data-dependent files were used to create a common phospholibrary for the different cell lines. For this the data was searched using XTandem, OMssa and Comet using a Parent mass error of 50 ppm, a fragment mass error of 0.04 Da, 1 missed cleavage was allowed and as modification Carbadmidoemthyl on cystein as a static and phospho on serine, threonine and tyrosine and oxidation on methionine as a variable modification was added. A 0.01 FDR cutoff on iprophet-peptide FDR was used to control for false identifications. For each annotated spectra a false localization score was calculated using Luciphor2 (Fermin et al., 2013) and only annotations with a false rate of lower than 0.01 were used. A SWATH assay spectral library was generated as described before using a distance of 2 min to separate adjacent peaks and TPP (Schubert et al., 2015). This resulted in a SWATH-assay library containing 5275 different phosphopeptides and proteotypic phosphopeptides mapping exclusively to different 1978 Swissprot protein identifiers. The extraction of the data was performed using the OpenSWATH workflow (Röst et al., 2014) as described above. The data was then processed using the R/Bionductor package SWATH2stats (Blattmann et al., 2016). Precursors had to pass an m-score threshold of 0.01 in 3 biological replicates of one condition to be selected for further analysis. These threshold resulted in an estimated precursor and peptide FDR of 0.0179 (using an estimated fraction of false targets (FFT) or π0-value of 0.45 for estimating the FDR). In total 2209 peptides and 88 decoy peptides passed this stringent threshold. This resulted in a data matrix containing 1.0 105 peakgroup intensities, from which 61% of peakgroups had an m-score of < 0.01. The data was then normalized using a total intensity normalization and analyzed for differential expression using mapDIA v2.4.1 (Teo et al., 2015). Differential expression was tested using an independent study design with the settings of selecting a minimum correlation of 0.1, a standard deviation factor of 2, between 3-5 fragments per peptide and at least one peptide per protein.
Peter Blattmann, ETH Zurich
Ruedi Aebersold, 1) Department of Biology, Institute of Molecular Systems Biology, ETH Zurich, Switzerland 2) Faculty of Science, University of Zurich, Zurich, Switzerland ( lab head )
Blattmann P, Henriques D, Zimmermann M, Frommelt F, Sauer U, Saez-Rodriguez J, Aebersold R. Systems Pharmacology Dissection of Cholesterol Regulation Reveals Determinants of Large Pharmacodynamic Variability between Cell Lines. Cell Syst. 2017 5(6):604-619.e7 PubMed: 29226804