Feature Extraction from Foxo3a Co-IP Samples

Supplementary Results

FT-linearization

The prerequisite for MS1 based quantification is sufficient linearity and dynamic range of the mass spectrometer, reproducible LC separation, and high mass precision. Only then can MS1 features be mapped with high confidence over different LC-MS runs. The mass precision is very high on a FT-experiment (<5 ppm), enabling the matching of MS1 features with high confidence over different runs. Furthermore, it has been shown that measurements on ion-trap instruments can be linearized by relatively simple normalization methods21, making label free quantitative proteomics conceivable. We measured dilution curves of standard digests on our LTQ-FT setup to confirm that ion currents (IC) of peptides are linearly scaled with their quantities. Lysozyme was diluted in ten steps into a constant amount of background proteins (bovine -lactoglobulin, and bovine -casein) in ascending concentrations (0 – 1.3 pmol/ul). The LC-MS runs were aligned and evaluated with the SuperHirn program. The resulting dilution curve was best fitted by a logarithmic linear regression (R2=0.99, supplementary Figure 1a), indicating that the IC scales linearly to the peptide molarity over more than 2 orders of magnitude.

Feature extraction from FoxO3A Co-IP samples

The dilution mixture of the FoxO3ACo-IP samples was measured on a hybrid FT-ICR mass spectrometer (LTQ-FT, Thermo Finnigan) in data dependent acquisition mode (DDA). Data preprocessing and peak detection yielded an average of 4366 ± 465 (n=4) detected MS1 features per run. Following multiple LC-MS alignment, the MasterMap contained 2290 M1 feature profiles with minimally 3 data points. These profiles were subjected tounsupervised clustering (see Fig. S6) and the final clusters were ranked according to their similarity to the expected interaction partnerdilution profile. Profiles were obtained for peptides spanning a wide range of signal intensities showing that also low abundance peptides can be quantified (Fig. S3).

FoxO3A-LY Co-IP samples under growth inhibiting conditions were analyzed by LC-MS (3347 ± 465 extracted MS1 features, n=4) and a MasterMap containing 3400 peptide profiles was created.

Analysis of cross-linked Co-IP samples

A common problem in IP experiments is the loss of weak interactions during the washing steps. One way to stabilize these interactions is chemical cross-linking. The potential increase in sensitivity is compromised by an increased complexity of the MS1 pattern due to variable modifications introduced by the cross-linker, lower cleavage efficiency of trypsin on lysine residues that react with the cross linker, and the presence of non-specifically cross linked proteins.We therefore were interested in determining whether we would still be able to separate specific interactions from contaminants under cross linking conditions and, moreover, if we would be able to identify new binding partners of FoxO3A.

FoxO3A and control cells were lysed in presence of the amino-reactive bi-functional cross-linker DSP45 and data analysis was performed as described for the previous experiment. The extracted protein cluster profiles showed again a bimodal distribution, which was used to separate specific from background proteins (data not shown). In agreement with the previous experiments, all six members of the 14-3-3 family that were seen in the first experiments were clustered together with FoxO3A in two independent experiments (Table S2). In addition we identified in both experiments the heterotrimeric protein phosphatase 2A (PP2A) and confirmed specific association to FoxO3A to three subunits (65 kD regulatory subunit A, 56 kD regulatory subunit B, catalytic subunit C; see Fig. S5). An association of PP2A with its substrate has been reported also for other proteins including Bad, KSR1 and Raf-1that are regulated by PKB phosphorylation dependent binding of 14-3-3 proteins46, 47, which may suggests a more general signaling mechanism wherein stably associated phosphatase PP2A could sustain a certain turnover of regulatory phosphorylation sites that are phosphorylated by active AKT/PKB and thereby controls association with 14-3-3 family members. However in the case of FoxO3A it remains to be tested which phosphorylation sites are indeed affected by associated PP2A.This notwithstanding our data could provide evidence for a molecular mechanism wherein associated PP2A could render FoxO3A activity responsive to changes in PI3K-PKB signaling.

Quantitative analysis of protein interaction changes between LY and FCS was also performed for the crosslinked samples. The enrichment factors obtained in the independent purifications using different lysis buffers and sample preparation protocols were reproducible. As before, 14-3-3 proteins were enriched under growth stimulating conditionsalbeit with somewhat lower ratios than before (1.5 – 2.4). In contrast, the newly identified subunitsof PP2A (65 kD regulatory subunit A, catalytic subunit C 56 kD regulatory subunit B) were reduced between in the LY condition, consistent with a role in the dephosphorylation of FoxO3A under growth inhibiting conditions (Table 2).

Supplementary Methods

Standard curve

Stock solutions of bovine -lactoglobulin, bovine -casein and chicken lysozyme (10 mg/ml in water, all Sigma-Aldrich, Switzerland) were diluted in 100 µl NH4HCO3 to a final concentration of 50 mM. Prior to trypsin digestion, proteins were reduced with 5 mM TCEP (Pierce, Rockford, IL) for 30 min at 37ºCand alkylated with 10 mM iodoacetamide (Sigma-Aldirch, Switzerland) for 30 min at room temperature. 1 ug trypsin was added and digestion proceeded overnight at 37°C and the reaction was stopped by acidifying with 1% FA. Peptides were desalted with C18 ultra-micro spin columns (Harvard Apparatus), dried down, and redissolved in 0.1% FA to get a stock concentration of 20 pmol/µl calculated as 100% recovery.

Cross-linking of cell lysates

For post-lysis cross-linking cells were lysed in DSP-TX100 buffer (1% Triton X-100, 40 mM HEPES [pH 7.5], 120 mM NaCl, 1 mM EDTA, 50 mM NaF, 1.5 mM Na3VO4, 0.1 mM PMSF, 3mM DSP (Pierce), 1 protease inhibitor tablet mix (Roche, Basel Switzerland) per 50 ml). Cells were subjected to 10 strokes using a tight douncer and incubated for 30 min. on ice in the presence of 3 mM DSP. The cross-linking reaction was quenched be addition of Tris-HCl (pH 7.5) to a final concentration of 50 mM. All other steps were identical to the Co-IP with TNN buffer.

Supplementary Figures:

Supplementary Figure 1.A: Dilution curve of lysozyme mixed into a background of -lactoglobulin and -casein. Lysozyme was diluted in logarithmic steps (Table S1). The average profile of measured ion currents of lysozyme peptides is best fitted by a logarithmic linear curve (R2= 0.99) indicating a linear response characteristic of the FT signal. B: Dilution experiment of Horse myoglobin in complex sample background. Horse myoglobin was mixed from 4 to 800 fmol into a complex sample background of N-glycocaptured serum peptides. The 6 dilution steps were analyzed by LC-MS and data processing was performed by SuperHirn. The dotted line shows the theoretical profile normalized to the 800 fmol value. Error bars indicate standard error of the mean.

Supplementary Figure 2. Simple ratios between bait and control sample fail to separate interaction partners from contaminant proteins. Ratios were calculated between normalized ion currents of peptides that were identified in control and bait sample for the FCS dataset shown in Table 1. Proteins with less than two quantifiable peptide ratios are indicated as undefined. Red squares show FoxO3A and 14-3-3 protein which were identified as binding partners in Table 1. The reason that some of these binding partners are not found with interpretable ratios is the absence of a peptide signal in the control sample. On the other hand many other contaminant proteins are also not successfully identified on MS1 level in both samples. Consequently, absence of a MS1 feature in a sample does not reliably confirm its specific presence in a compared sample where this feature was identified.

Supplementary Figure 3. Dynamic range of quantified MS1 feature profiles. The plot shows the intensity range of all quantified MS1 feature profiles from the Co-IP experiment performed under TNN-FCS condition (see Table 1). Log10 intensities (100% bait sample) of quantified MS1 feature profiles are ranked by their intensities covering morethan 4 orders of magnitude (black dots). In addition, intensities of MS1 feature profiles of all identified peptides (blue square) and peptides from FoxO3A and its 14-3-3 interaction partners (red diamonds) are shown that have been quantified by MS1 profiling and were identified by CID.

Supplementary Figure 4. Transfer of MS2 information across aligned MS1 features compensates MS2 under-sampling. The reason why some peptide escape identification can be that low intensity precursor ions do not give rise to an interpretable MS2 spectrum or they are not selected by the mass spectrometer for MS2 analysis Areplicate FoxO3ACo-IP preparedunder TNN-buffer conditions was analyzed four times by LC-MS/MS. SuperHirn was used to create a MasterMap by two separate MS1 features annotation processes: using only performed MS2 scans (shotgun MS2 sequencing) or projecting peptide identifications across aligned features (MS2 continuation). A: The percentage of how many times a MS1 features has been identified in every LC/MS run by conventional shotgun MS2 sequencing (blue) or MS2 continuation (red) is shown. B: MS1 features are plotted by their m/z (y axis) and Tr (x axis) as gray dots. Identified features (peptide probability > 0.9) are colored according to how many times they have been sequenced in every LC-MS run by either conventional shotgun MS2 sequencing (left) or MS2 continuation aligned matched features (right). Even though MS2 transfer over aligned MS1 features is much more reproducible than the shotgun approach peak extraction and alignment can fail for the following reasons: They can deviate significantly from the expected isotopic pattern, which can lead to missed or wrong assignments or, in complex mixtures, overlapping signals can lead to wrong assignments of monoisotopic peaks or charge states. Furthermore weak signals can be pushed below detection limit by ion suppression effects. Even high intensity peaks can be missed for peptides that show a large variation in retention time; these features can escape alignment between different runs and introduces gaps into the profiles.

Supplementary Figure 5 The heterotrimeric protein phosphatase PP2A was identified as FoxO3A interaction partner under cross-linking conditions. Normalized protein profiles from a 5 point dilution series wereobtained from post-lysis DSP cross-linked extracts of cells grown at 10% FCS. Besides 14-3-3 proteins, the heterotrimeric protein phosphatase 2A is specifically enriched in the bait sample.

Supplementary Figure 6K-means clustering of feature profiles from a dilution series. Feature profiles were subjected to K-means clustering analysis and the consensus profiles of the obtained clusters centers are shown (gray). Color highlighted are the cluster consensus profiles of features from contaminants (blue), proteins which are partially enriched (green) and enriched proteins such as the FoxO3A and its interaction partners (red).

Supplementary Tables:

Supplementary Table 1. Dilution scheme of the standard protein mix. Background proteins were a mix of digests from bovine -lactoglobulin and bovine -casein at constant concentration of 500 fmol. Lysozyme is diluted in 10 steps from 0-1200 fmol. Molarities are calculated from the digested protein amounts without correcting for losses.

Supplementary Table 2. Identified interaction partners of FoxO3A using DSP cross-linking in FCS condition. Protein profiles were scored according to similarity to the target cluster profile. Assigned profile probability values reflect the likelihood for a true similarity with the target profile. p > 0.9 was used as significance threshold. Only proteins with proteotypic peptides are shown. Number of members indicate peptide identifications with complete quantifiable dilution profiles.

a Number of MS1 features. b Number of identified peptides. c Increase in MS1 features by inclusion list and LTQ data. d Increase in identified peptides by PMM, e quantified enriched peptides in a replicate experiment, #scoring from replicate experiment.

Supplementary Table 3. Growth state specific changes in the FoxO3A phosphorylation pattern. A dilution mix was created from Co-IP samples collected under LY and FCS conditions and profiles of FoxO3A peptides containing phosphorylation sites and their modified counterparts were extracted. An enrichment factor was subsequently determined to quantify the relative abundance changes of each peptide compared to the average protein profile of FoxO3A. TR and indicate the retention time and molecular mass differences between the phospho-peptide and its unmodified counterpart.

a cross-linker (DSP or non cross-linked (TNN) Co-IP experiment, b/c enrichment factor of the phospho-peptide and its non modified counterpart respectively, d Mascot score from the phospho-peptide search.

1