SAFER, an Analysis Method of Quantitative ... - ACS Publications

Subscriber access provided by GAZI UNIV

Article

SAFER, an analysis method of quantitative proteomic data, reveals new interactors of the C. elegans autophagic protein LGG-1 zhou Yi, Marion Manil- Ségalen, Laila Sago, Annie Glatigny, Virginie Redeker, Renaud Legouis, and Marie-Hélène Mucchielli-Giorgi J. Proteome Res., Just Accepted Manuscript • DOI: 10.1021/acs.jproteome.5b01158 • Publication Date (Web): 21 Mar 2016 Downloaded from http://pubs.acs.org on March 30, 2016

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Journal of Proteome Research is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Journal of Proteome Research

SAFER, an analysis method of quantitative proteomic data, reveals new interactors of the C. elegans autophagic protein LGG-1 Zhou Yi1‡, Marion Manil- Ségalen1‡, Laila Sago2, Annie Glatigny1, Virginie Redeker2,3, Renaud Legouis1*, Marie-Hélène Mucchielli-Giorgi1,4,* 1

Institute for Integrative Biology of the Cell (I2BC), CEA, CNRS, Univ. Paris‐Sud, Université

Paris‐Saclay, 91198, Gif‐sur‐Yvette cedex, France 2

Service d’Identification et de Caractérisation des Protéines par Spectrométrie de masse

(SICaPS), CNRS, 91198 Gif-sur-Yvette, France. 3

Paris-Saclay Institute of Neuroscience (Neuro-PSI), CNRS, 91198 Gif-sur-Yvette cedex,

France. 4

Sorbonne Universités, UPMC Univ Paris 06, UFR927, F-75005, Paris, France.

KEYWORDS Statistical methodology, label free mass spectrometry, proteomics, autophagy, C. elegans, atg8/LC3

ACS Paragon Plus Environment

1


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 2 of 30

ABSTRACT

Affinity purifications followed by mass spectrometry analysis are used to identify protein-protein interactions. As quantitative proteomic data are noisy, it is necessary to develop statistical methods to eliminate false positives and identify true partners. We present here a novel approach for filtering false interactors, named “SAFER” for mass Spectrometry data Analysis by Filtering of Experimental Replicates, which is based on the reproducibility of the replicates and the foldchange of the protein intensities between bait and control. In order to identify regulators or targets of autophagy, we characterized the interactors of LGG1, an ubiquitin-like protein involved in autophagosome formation in C. elegans. LGG-1 partners were purified by affinity, analyzed by nanoLC-MS/MS mass spectrometry and quantified by a label-free proteomic approach based on the Mass spectrometric Signal Intensity of peptide precursor ions. Because the selection of confident interactions depends on the method used for statistical analysis, we compared SAFER with several statistical tests and different scoring algorithms on this set of data. We show that SAFER recovers high confidence interactors that have been ignored by the other methods, and identified new candidates involved in the autophagy process. We further validated our method on a public dataset and conclude that SAFER notably improves the identification of protein-protein interactors.


2

Page 3 of 30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Introduction Identification of partner of a protein of interest is a common strategy to document regulatory proteins of key biological processes. The commonly used strategy consists in the combination of affinity-based purification of protein complexes, proteolytic digestion of the purified proteins and mass spectrometric (MS) analysis of the proteolytic peptides. Using liquid chromatographic separation combined to tandem MS (LC-MS/MS), this MS analysis has proven to be very efficient to rapidly identify up to hundreds of proteins and perform relative quantification of proteins between samples (see 1,2 for review). Several relative MS quantification strategies can be used3. Various strategies include isotopic labelling of the samples4, either of the proteins at the cell or tissue level (SILAC5), or at the protein level (ICAT6), or at the peptide level (iTRAQ7). Although these approaches appear sensitive, “label-free” MS quantification strategies are increasingly used, mainly because these approaches can be applied to a large variety of proteins and organisms, and are relatively easy and rapid to perform8. Relative label-free MS quantification of the proteins are of particular interest to identify proteins enriched in the biological experiment containing the bait versus the control experiment without the bait. However these label-free relative MS quantification methods require a careful normalization between samples and an efficient MS data validation method in order to select true interactors3,9. Actually, the validation of true interactors within all the identified proteins appear to be an important limitation when considering low abundant proteins or small proteins generating few proteolytic peptides. Two different mass spectrometric label-free quantification approaches can be used to measure and compare protein abundance from two or more experiments. The first uses the mass spectrometric signal intensity of peptide precursor ions (MSI) belonging to a given protein, the


3


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 4 of 30

second the spectral counts (SC), i.e. the total number of peptide sequencing events (MS/MS fragment spectra) identifying peptides for a given protein10. For a statistical analysis, these two approaches result in a different distribution of the values used to measure the abundance of a given protein: a continuous and discrete probability distribution approximates the distribution of the MSI and SC values, respectively. MSI values are considered as more accurate and more sensitive for measuring the relative abundance of low abundant proteins than the SC10. However, MSI data, which can contain missing values and a high number of false positive proteins, requires specific statistical treatments in order to discard false positive partners and enhance confidence identification of true protein partners11. To overcome this drawback of MSI quantification, several scoring and statistical methods have been developed. The t-test12 or modified t-tests13 measures the p-value of the difference between the mean of the control and the biological replicates. Scoring methods such as MiST14, SAINT15 compute a confidence score for each partner protein from MSI values. Their performance depends on the overall experimental process and the nature of the obtained data, i.e. the number of baits, the number of replicates of each bait, the number of control experiments and also the number of preys. In order to identify regulators or targets of autophagy, we choose a proteomic approach including MSI data acquisition to identify interactors of LGG-1 in C. elegans. LGG-1 is an ubiquitin-like protein essential for initiation and biogenesis of the autophagosome, and involved in multiple autophagy processes such as mitophagy16,17 and aggrephagy18. LGG-1 partners were purified by affinity, analyzed by nanoLC-MS/MS mass spectrometry and quantified by a label-free proteomic approach based on the MSI. Because the selection of confident interactions depends on the method used for statistical analysis, we evaluated the three principal commonly used statistical and scoring methods (t-test, SAINT, MIST) for identification of the immuno-purified


4

Page 5 of 30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


partners of LGG-1. We observed that these three methods do not identify the same protein partners of LGG-1. In order to improve the confident identification of true protein partners of LGG-1 in C. elegans and avoid the main drawbacks of the three methods, we optimized a novel approach for filtering false interactors, named “SAFER” for mass Spectrometry data Analysis by Filtering of Experimental Replicates. This method dedicated to relative label-free quantitative proteomic analysis using MSI is based first on the reproducibility of the biological and control replicates, and second on the fold change of the protein intensities between bait and control. SAFER selected potential LGG-1 interactors that have been ignored by the other methods. We further evaluated our SAFER data filtering method on a public label-free quantification dataset (doi:10.1128/MCB.01742-12) and confirmed that SAFER efficiently improves the confidence of identification of affinity purified protein interactors. Finally, we validated the regulatory role of several LGG-1 interactors in the autophagic process, by quantification of the autophagic flux in worms after RNAi against interactors identified only by SAFER. We show that SAFER notably improves the identification of protein-protein interactors and identified with confidence new regulators of the autophagic process.

Materials and Methods Worm strains and culture Nematode strains were grown on NGM plates seeded with E. coli strain OP50 and handled as described19. The GFP control strain is the EG1470 (pmyo2::GFP) and GFP::LGG-1 is the DA2123 strain (plgg-1::GFP::LGG-1). Briefly, approximately 200 L1 larvae were grown for 4 days at 20°C on each plate, leading to non-synchronized populations of worms containing embryos, larvae and adults. Pools of 20 culture plates containing similar population of worm


5


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 6 of 30

formed each replicate. All the cultures were grown simultaneously and four biological replicates were collected for each strain.

Immunoprecipitations Total protein extracts, immunoprecipitations and western-blotting were performed simultaneously for each replicate. For protein extraction, animals from 20 fully grown, but not starved, plates were collected, corresponding to approximately 100µl of animal pellet per replicate. Animal pellets were lysed in 20mM Tris HCl pH7.5, 100mM NaCl, 10% Glycérol, 5mM MgCl2, 2mM EDTA, 2mM DTT and Protease inhibitor cocktail 1X (Roche). Glass beads (Sigma, 425-600µm) were added before homogenization by two cycles of 60 sec at 6,000 rpm using Precellys24 instrument (Bertin Technologies). Cell extracts were obtained after centrifugation at 12,000 rpm. Immunoprecipitations were performed in the same buffer using Ademtech PAG beads with 2mg of protein extract (quantification by nanodrop (Thermofisher) and 2µg of rabbit anti-GFP antibody (Invitrogen). Proteins were eluted in 25µl of the provided elution buffer. 20µl were loaded after denaturation and separated on a NuPAGE 4%–12% BisTris gel (Life Technologies). After Coomassie blue staining, each gel lane was sliced in protein bands of about 1cm, in order to remove remaining immunoglobulins and to fractionate the protein samples into four fractions before in-gel digestion and mass spectrometry analysis (Figure S-1). 5 µl of immunoprecipitation were used to perform control western blot (mouse anti GFP 1/1000, Roche, HRP-conjugated secondary anti-mouse antibody, 1/10000, Promega, ECL detection kit, Thermo Scientific).


6

Page 7 of 30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Protein digestion and tryptic peptide preparation In-gel digestion using trypsin was automatically performed using the Progest robot (Genomic Solutions) after washing of the bands, and reduction and alkylation of proteins with 10mM DTT and 55mM Iodoacetamide respectively using standard conditions described previously20. Tryptic digestion was performed overnight at room temperature by addition of 20 µl of 10 ng/µl Porcine Gold Trypsin (Promega) diluted in 25 mM NH4HCO3. Tryptic peptides were extracted first by addition of 20 µl of 60% acetonitrile and 0.1% formic acid and second by addition of 20 µl of 100% acetonitrile. Peptides extracted from the different bands of each gel fraction were pooled, vacuum dried and resuspended in 5% acetonitrile and 0.1% TFA prior to nanoLC-MS/MS mass spectrometry analyses. The same cutting pattern of the SDS-PAGE lane was performed for each IP experiment.

NanoLC-MS/MS analysis and data processing NanoLC-MS/MS analyses were performed with a LTQ Orbitrap Velos mass spectrometer (Thermo Scientific) coupled to the EASY nLC II high performance liquid chromatography (HPLC) system (Proxeon, Thermo Scientific, Waltham, MA). Peptide separation was performed on a reverse phase C18 column (100 µm inner diameter, 5-µm C18 particles, 15-cm length, NTCC-360/100-5) from NikkyoTechnos (NikkyoTechnos Co., Ltd., Tokyo, Japan) with a 5 to 35% solvent B gradient in 40 min, followed by a washing step at 100% solvent B. Solvent A was 0.1% formic acid in water, and solvent B was 0.1% formic acid in 100% acetonitrile. NanoLCMS/MS experiments were conducted in the Data Dependent acquisition mode. Mass range for MS1 was set to 400-2000 m/z. Precursor Mass of the precursor ions was measured in the Orbitrap with a high resolution (60,000 full weight at half maximum). The 20 most intense


7


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 8 of 30

multiple-charged precursor ions (singly charged precursor ions were rejected), above an intensity threshold of 2000 counts and with a 10 s exclusion time, were selected for CID (Collision Induced Dissociation) fragmentation and analysis in the LTQ. Mass window for precursor ion selection was set to 2.0. For CID, normalized collision energy was set to 30.

For protein identification, data were processed with MaxQuant software (v 1.3.0.5) using defaults parameters21. Briefly, raw data were searched against a forward and reversed version of Uniprot-worm database (Uniprot 6239 date 10/10/13, 26834 sequences) using andromeda supplemented with frequently observed contaminants. Carbamidomethylation of cysteines and oxidation of methionines were included as fixed and variable modifications respectively. Enzyme specificity was set to trypsin, allowing a maximum of two missed cleavages. Mass tolerance was set to 20 ppm for monoisotopic precursor ion masses and 0.5 Da for MS/MS fragment monoisotopic masses. Maximum false discovery rates (FDR) were set to 0.01 at both the peptide and protein levels. For label-free quantitation using MaxQuant software (v 1.3.0.5), unique and razor peptides were considered. Briefly, the ion intensity of a given peptide was defined by its mass, its chromatographic elution time and its intensity. For each peptide, chromatographic time alignment and mass comparison are used to compare the peptide intensities from multiple runs and calculate a peptide intensity ratio. Pair-wise peptide intensity ratios are determined only when the corresponding peak is detected in both LC-MS runs. As median values of all peptide intensity ratios of one given protein represent a robust estimate of the protein abundance ratio, the MaxQuant LFQ algorithm calculates these ratios after a normalization procedure using the total peptide ion signals22.


8

Page 9 of 30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


RNAi and visualization of the autophagic flux RNAi by feeding was performed using DA2123 (GFP::LGG-1) worms as previously described23. Briefly, L1 larvae are laid onto NGM plates containing IPTG and seeded either with bacteria (E. coli HT115 [DE3]) carrying the empty vector L4440 (pPD129.36) as negative control or with one bacterial clone from the J. Ahringer library (Open Biosystem). We used one specific RNAi clone for each of the twenty six candidates recovered by SAFER and selected randomly for this analysis. Among them six clones, previously described for resulting in a strong phenotype of slow growth, were used as positive controls for RNAi but not analysed for the autophagic flux. L1 larvae expressing GFP::LGG-1 were cultured for four days on bacteria expressing a single dsRNA, until they reach adulthood and they start laying eggs. Total protein extracts were prepared from these populations of adults and embryos as described24, denatured for 10 min at 72°C, and separated on a NuPAGE 4%–12% Bis-Tris gel (Life Technologies). Blots were probed with anti-GFP (1:1,000; Roche) or anti-Tubulin (1:2,000; Sigma) antibodies and revealed using HRP-conjugated antibodies (1:10,000; Promega) and the ECL detection system (Thermo Scientific). Signals were revealed on a Las4000 photoimager (Fuji) and quantified with ImageJ. For each candidate, we performed three experiments with the same RNAi clone. Blots were analysed according to the autophagy community guidelines25,26.

Scoring and statistical methods for the selection of the protein partners Several statistical methods to assign confidence scores to interactions detected by MS experiments have been published. For the treatment of the MSI values, the most current


9


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 10 of 30

statistical test is the Student’s t-test and the two principal commonly used confidence scores developed are SAINT and MIST. The SAINT (Significant Analysis of INTeractome) score15 measures the probability that the interaction between bait and prey is a true interaction. It uses label free quantitative data to construct separate distribution for true and false interaction to derive the probability of bona fide protein-protein interaction. It allows then the selection of interactions specific of the bait. It takes indirectly into account the reproducibility of the experiments. Three features are used to define the MiST score15: the abundance, the reproducibility and the specificity of an interaction between a bait and a prey. The three features were combined into a single composite score by maximizing the variance in the three features space using a standard principal component analysis (PCA). The t-test analysis were performed with the Perseus software (http://www.perseusframework.org). Briefly, results from MaxQuant were cleaned for reverse hits and positive intensity values were converted to logarithm. Signals that were originally zero were imputed with random numbers from a normal distribution, whose mean and standard deviation were chosen to best simulate low abundance values below the noise level (Width = 0.3; Shift = 1.8). Significant interactors were determined by combining t test p-values < 0.05 and Fold Change (FCp) > 2. Contrary to the statistical tests15,27,28, the scoring methods have the advantage to deal with missing values, but their main weakness is that these scores do not follow any standard probability distribution, which makes impossible the calculation of p-values. Therefore, the choice of the threshold to filter out false positives remains empirical. While MiST gives an interaction score for all the preys, SAINT restricts the resulting scores to a subset of candidates. So, to compare the results of SAFER, the t-test and the scoring methods SAINT and MiST, the


10

Page 11 of 30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


number of protein candidates must have the same magnitude. We then chose a scoring threshold giving the same number of candidates as our method when applicable.

Results and Discussion Experiment description To identify new interactors for LGG-1, one of the C. elegans LC3/GABARAP homologs, we performed an immunoprecipitation approach. Total protein extracts from a mixed stage population of worms expressing GFP::LGG-1 or GFP alone were immunoprecipitated using a GFP antibody. These experiments were done in quadruplicates and immuno-precipitates were separated into four fractions by 1D electrophoresis (Figure S-1). Proteins were in-gel digested using trypsin and tryptic peptides were further analyzed by mass spectrometry (Figure 1). The data generated by the nanoLC-MS/MS analysis were further processed for protein identification and protein quantification. Label-free quantification was performed with MaxQuant using the LFQ procedure of intensity determination of the MS signal and normalization 22.

SAFER : a new method for identification of protein partners by filtering experiments To analyze our experimental data, we elaborate SAFER (mass Spectrometry data Analysis by Filtering of Experimental Replicates) a new approach consisting of a multi-step filtering method. SAFER was conceived for experiments in tri or quadruplicates. Its workflow is represented in Figure 2A for quadruplicates and described below. Each filtering step is illustrated on Figure 2B by representative preys of the LGG-1 protein. The first step selects all the proteins that are present in at least 3 biological bait replicates. The second step keeps all the proteins that are absent in most control replicates; that is, have at most one non-zero value among the 4 control


11


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 12 of 30

replicates. Among the proteins discarded by the second filter, the value of the fold change (FCs) is considered. The FCs corresponds to the ratio of the average values of the MSI of the biological bait replicates and of the control replicates. Those having a FCs value greater than 2 are retained. For the calculation of the FCs, the means of the biological and control replicates are computed without taking the values equal to zero into account. Indeed, we consider that a zero value in the experimental replicates is an error since all the values of the other replicates are positives. In the control replicates, taking into account only the two non-zero values allows the selection of a lower number of preys making SAFER more stringent. Note that our FCs values are different from those of the t-test (FCp) that are computed by replacing the zero values by an estimated value. SAFER was conceived for experiments in tri or quadruplicates because it is the minimum number of replicates for a good measure of the fold change. Indeed, if the experiment is composed of three replicates and a value is missing, the variability of measurements cannot be calculated and the value of the Fold Changes can be biased. For triplicates set of data, a supplementary filter is added to this second step: A protein is not selected when the maximum value of the three negative control replicates is higher than the minimum value of the three bait replicates. This filter permits to lessen the effect of one high value among a set of low values of the control on the selection of the protein partners.

Statistical and scoring methods identify different protein partners of LGG-1 We tested and compared the two principal scoring methods presented above (SAINT, MiST), the Student’s t-test and SAFER to select interactors of the C. elegans autophagic protein LGG-1. The four methods gave very different results as shown in Figure 3A. The total of preys selected


12

Page 13 of 30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


by the four methods is equal to 393 and only 22 proteins are common. Among them, we found the proteins ATG-7, an autophagic protein which is known to interact with the homologs of LGG-1and is necessary for its post-translational modifications and localization to the autophagosome29. Moreover, the percentage of intersection between two methods (percentage of proteins identified by two methods among those identified by at least one of them) is often low30 (Figure 3A): It is included between 17% for the intersection between MiST and the t-test and 42% for the intersection between the t-test and SAFER. Many reasons can explain the differences between the results obtained with the two scoring methods and the t-test. One of the main reason is that the t-test is more stringent than the other methods: it gives a small and incomplete list of confident protein partners contrary to the other methods that select a significant number of false positives, different from one method to the other. For this reason, their overlap is very low and none of them can result in a complete and confident identification of true protein partners. Another reason is that the SAINT scoring method does not take explicitly the reproducibility of the replicates into account. It can thus lead to select some preys which exhibit a high variance of the biological replicates or with two zero of four MSI values in the biological replicates (as exemplified by the C. elegans preys DEB-1 and RPL-31 in the Figure 2B). A third reason is the preponderant contribution of the reproducibility of the replicates in the MiST score. Then it selects preys having a low variance of the biological replicates although their values of MSI are low (as exemplified by prey UBC-20 in the Figure 2B). A fourth reason is that the t-test measures the p-value of the difference between the mean of the control and of the biological replicates. It is not applicable when the distribution of the MSI values is not a Gaussian. Moreover, if the threshold of the p-value, measuring the false positive rate, is fixed at 5%, the false negative rate can then be high. To minimize both the false positive


13


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 14 of 30

rate and false negative rate, modified t-tests were proposed and used13. But mostly, these tests should not be employed because all the conditions of their application are not satisfied; particularly the distribution of the p-values is not a Gaussian. The principal limitation of the t-test and of the modified t-test is that they require the calculation of the variances of the biological and of the control replicates that still needs at least three values. This is impossible since the datasets contain inevitably values equal to zero. To overcome this limitation, an estimation of the missing values can be performed22 from the positive values in the three or four control or biological replicates and from the distribution of the values of all preys in the experiment compared to the distributions of the other replicated. This estimation leads to aberrant results when one or several missing values are estimated within the replicates. Some preys can also be selected with non-zero values in the control (as exemplified by results of MSI for prey HAF-4 in Figure 2B). A last reason is that SAFER does not take the variance into account and thus selects some proteins that are not retained by the other methods. So, it selects proteins absent in one of the biological bait replicates, which are removed by the other statistical methods despite their high MSI values (as exemplified by prey LARP-1 in the Figure 2B). Moreover, despite MSI values of the controls different from zero, proteins having a fold change (FCs) more than 2 are selected by SAFER contrary to the other methods (as exemplified by prey RACK-1 in the Figure 2B).

Among the LGG-1 partners identified by SAINT or MiST or the Student’s t-test, the percentage of those selected by SAFER is always higher than those identified by SAINT, MiST or the Student’s t-test (Figure 3 part B). As an example, among the 85 LGG-1 partners selected with the t-test, 96% are found with SAFER, when only 54% and 47% are found with SAINT and MiST.


14

Page 15 of 30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


These results show that SAFER presents both the advantages of SAINT, MIST and of the Student’s test: it takes into account both the values of the control condition, the reproducibility of the biological replicates (bait and control) and can deal with the null values in the control condition. So, it is an accurate strategy for the selection of bait protein partners from experiments including a control condition and replicates of the bait and of the control. Thus, with SAFER we get most of the proteins selected specifically by each method. However SAFER selects protein partners not found with SAINT, MIST and of the Student’s ttest: 56 partners of LGG-1 (Figure 3A, Table S-1). Some of them are orthologs of proteins in the CRAPome bank31 that lists contaminants in H. sapiens, S. cerevisiae, E. coli. But, in some cases, they can be true preys having a real biological interaction with the bait protein. So it is very important not to discard them, as it would be the case with a method that systematically filters the contaminants, such as the one reported by Morris et al.32.

Validation of SAFER on Public Data To test our method, we looked in the public database PRIDE33 for dataset containing GFP control samples and baits in tri or quadruplicates. To be consistent with our data, we searched for an experiment done with the same instrument and analyzed with the same software (LTQ Orbitrap velos, Maxquant). We choose the project PXD00017234 whose purpose is to determine the subunit composition and stoichiometry of the human SET1/MLL complexes by using labelfree quantitative mass spectrometry. Among its different baits in triplicates, we choose the proteins ASH2L, RBBP5, WDR82 because they compose the conserved core of the complexes.


15


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 16 of 30

All the conclusions obtained on our experiment on LGG-1, are confirmed on the proteins ASH2L, RBBP5, WDR82: (1) as for LGG-1, the four methods gave also very different results as shown in Figure 3A. The number of partners identified by the four methods is globally low: 69 for ASH2L, 4 for RBBP5 and 135 for WDR82 among respectively 435, 195 and 711 partners identified by at least one method. (2) The percentage of overlap between two methods is low. For example for RBBP5, it is equal to 3% between the t-test and MIST. The higher percentage of overlap is obtained between SAFER and the t-test for ASH2 (72 %), RBBP5 (77%) and WDR82 (78%). (3) Among the proteins identified with SAINT or MiST or the student-test, the percentage of overlap between themselves is always weaker than with SAFER, except for the MIST results on the protein RBBP5 (Figure 3B, columns SAFER). SAFER permits then to get the most of the proteins selected specifically by each method. To check whether SAFER efficiently selects validated interactors of ASH2L, RBBP5 and WDR82, we searched in the BioGRID database35 for interactors that have been independently reported (Table S-2). We then identified 47 interactors of ASH2L, RBBP5, WDR82 that define 72 interactions present both in BioGRID database and in the raw data from project PXD000172. Among them, 50 interactions were selected by SAFER, 41 by the t-test, 31 by SAINT, 32 by MiST and only 17 are selected by the four methods. This result validates the accuracy of SAFER to identify the most pertinent interactors.

SAFER selects protein partners of LGG-1 not detected by other statistical/scoring methods With SAFER, we have highlighted 56 novel potential interactors of LGG-1, not detected by the other statistical/scoring methods. To validate these interactions, co-immunoprecipitations or colocalization experiments were envisaged. Unfortunately antibodies or tagged constructs were


16

Page 17 of 30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


not often available for these candidates. It was therefore not possible to test them rapidly. LGG-1 interactors could be either regulators of autophagy or substrates degraded by this process. We analyzed whether the proteins identified specifically by SAFER could be regulators of the autophagic process. We then quantified the autophagic flux in worms after inactivation of selected candidates by RNAi25. To do so, we analyzed 20 of the potential interactors. On western blot, we observed the variation of the autophagic flux between the different candidates (Figure 4A). We analyzed two distinct pieces of information: (1) the total amount of GFP proteins (GFP::LGG-1 and the cleaved GFP, degradation product, indicating a correct fusion between autophagosome et lysosome)(Figure 4B) and (2) the ratio of the degradation product among the total GFP proteins (Figure 4C). This quantification method is widely used among autophagy community and allows a global quantification of the autophagic flux25,26. Four of the tested candidates present a modified autophagy. In T09A5.11, VF13D21L.3 and Y53G8AL.2 worms the ratio GFP/total GFP is increased (Figure 4C), indicating a probable increase of the flux, while in cyc-1 worms, only the total amount of GFP and GFP::LGG-1 is increased (Figure 4B). These results indicate that some of the candidates selected only by SAFER are effectively regulators of the autophagic flux. Moreover, among the interactors identified specifically by SAFER, some have already been described in global approaches studying autophagy (Table 1). In 2014, Wild et al. recapitulates all the yet known interactors of the yeast Atg8 and mammalian LC3/GABARAP36. Other teams choose to purify autophagic structures in order to identify proteins enclosed in autophagosomes, putative targets of the autophagic process37,38. Among the proteins identified only by SAFER, six have been co-purified with autophagic structures in other species (AP in Table 1) and about one third have also been implicated in autophagic cell death in drosophila39.


17


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 18 of 30

These examples of proteins identified both by SAFER and biological experiments as putative autophagic-related proteins strengthens our analysis and the idea that these proteins could be true interactors of LGG-1 that would have been discarded using other filtering/scoring methods. One can raise the question of the presence of several ribosomal proteins in the selected preys, but it has been demonstrated in yeast that ribosomes are selectively degraded by autophagy in a process named ribophagy40,41. The ribosomal proteins that we identified could so be considered as potential targets of autophagy.

Conclusions We propose here a method consisting in a multi-step filtering (SAFER). We demonstrated on a public dataset that SAFER is the most suitable protocol to analyse Affinity Purification data in order to identify, with high confidence, the protein partners of a given protein. Its main benefit is to allow the selection of proteins that are removed discarded by the other commonly used statistical methods: (1) proteins absent in one of the biological replicates but abundant in the others and (2) proteins which have positive values in the control and a high difference between the biological condition and the control. We cannot totally exclude that some of these proteins are false positives but their values in the control and biological replicates do not support this hypothesis. SAFER outperforms the other statistical tests and scoring strategies, since it abrogates various defects of these methods. Because SAFER does not give a confidence measure to the selected proteins, we recommend to associate to them the most effective confidence score calculated on all the preys depending on each type of experience32. SAFER recovered potential partners of the C. elegans protein LGG-1, whose homologs have already been identified by autophagy screens in other species. Moreover, it reveals new potential


18

Page 19 of 30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


interactors of LGG-1 that were not selected by the other tested methods and that are effectively regulators of the autophagic flux.

FIGURES Figure 1. Schematic representation of the experimental approach

Figure 2. (A) Workflow of SAFER (mass Spectrometry data Analysis by Filtering of Experimental Replicates) indicating the three filters used. exp: experiments, ctrl: controls. (B) Example of preys of protein LGG-1, picked to illustrate each filtering step of SAFER. The 5

filtering is performed on MSI intensities, here expressed in 10 , from MS/MS experiments with four replicates for the control (bait= GFP) and four for the experimental condition (bait= GFP::LGG-1). The first filter removes preys highlighted in light grey. The second filter followed by the filter on the Fold Change (FCs) removes those in dark grey. Green lines correspond to preys with values in the controls and with a FCs higher than 2. The light green lines correspond to preys selected because they have less than two positive values in the control replicates and so they do not require any filtering on the FCs. Results obtain by SAFER are compared with those obtain by three other methods: t-test, performed by perseus, SAINT and MiST. The + indicates that the prey is selected, the - that the prey is removed. Except for ATG-3 and ATG-7, all the examples have been picked randomly. Figure 3. (A) Venn diagram of the candidates identified by SAFER (green) and the three tested methods: t-test (yellow) (p-value < 5% and FCP>2 where FCP is the fold-change (experiment vs control) calculated by Perseus by estimation of the missing values), SAINT (blue) and MiST


19


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 20 of 30

(red). The cut-offs of the two scores SAINT and MiST were chosen to obtain a number of candidates close to that of SAFER (see Material and Methods). Four sets of data were analysed: our experimental data based on LGG-1 potential interactors and three public data set based on ASH2L, RBBP5 and WDR82 potential interactors. The total number of selected candidates for a given test is written besides the diagrams. (B) Percentages of overlap between each methods. For each line, the percentages indicate the fraction of the proteins also identified by the other methods. For example, 96% of the potential partners of LGG-1 identified by the t-test, are also selected by SAFER. Data obtained by the SAFER method are highlighted in green. Figure 4. Visualization of the autophagic flux for some SAFER specific candidates after inactivation by RNAi. (A) Western blot analysis of GFP::LGG-1 in adult and embryos using anti-GFP antibodies. The cleaved GFP forms correspond to the GFP degradation products in the autophagolysosome. The localizations during the autophagic flux of the different GFP populations detected on the blot is schematized (right panel). (B-C) The autophagic flux was analyzed by quantifying the cytosolic GFP::LGG-1 and the cleaved GFP, specific of a proteolysis in the autophagolysosome. Normalization to tubulin of the total GFP signal (GFP::LGG-1 + GFP) revealed the total amount of the autophagic marker. (B) GFP/total GFP ratio indicates possible modification of the autophagic flux. (C) Plain black line: mean. Dotted black lines: upper and lower limits of a 90% confidence interval of the mean. The candidates out of these limits are colored in black.


20

Page 21 of 30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Table 1. Partners of the protein LGG-1 identified in other autophagy screens and selected only by SAFER. AC-MS: Affinity Capture Mass spectrometry, AP: Autophagosomal structures purifications, GI: Genetic interaction.

Sequence Name (Wormbase)

Gene Name

Identification method

Homologue identified

C34E10.6

atp-2

AP

ATP5B (Human)

F17C11.9

eef-1G

AP

EEF1G (Human)

F32H2.5

fasn-1

AP

FASN (Human)

K04D7.1

rack-1

AP

GNB2L1 (Human)

R05G6.7

vdac-1

T09A5.11

ostb-1

AC-MS AP

T19B4.3 Y22D7AL.5

hsp-60

VDAC1 (Human)

GI

WBP1 (Yeast)

GI

APT2 (Yeast)

AP

HSPD1 (Human)

References 37

38

38

38

36,37,42

36,43

36,43

37

AUTHOR INFORMATION

Corresponding Authors * [email protected], (33) 1 69 15 45 44 * [email protected], (33) 1 69 82 46 27


21


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 22 of 30

Author Contributions ‡These authors contributed equally.

Funding Sources IFR 115 supported the experiments. The Legouis group is supported by the Agence National de la Recherche (project EAT, ANR-12-BSV2-018) and the Association pour la Recherche contre le Cancer (SFI20111203826). M.M.-S. is a recipient of fellowships from the Ligue Nationale contre le Cancer.

Associated Content Supporting Information. Figure S-1: Gel slicing illustration Table S-1: List of LGG-1 interactors identified only by SAFER Table S-2: Comparison of MiSt, SAINT, t-test and SAFER on interactors for ASH2L, RBBP5 and WDR82

Acknowledgments We would like to thank Céline Largeau and Céline Jenzer for technical help with RNAi experiments.


22

Page 23 of 30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


REFERENCES (1)

(2)

(3)

(4)

(5)

(6)

(7)

(8) (9)

(10)

(11)

(12) (13)

(14)

Miteva, Y. V.; Budayeva, H. G.; Cristea, I. M. Proteomics-based methods for discovery, quantification, and validation of protein-protein interactions. Anal. Chem. 2013, 85 (2), 749–768. Nesvizhskii, A. I. Computational and informatics strategies for identification of specific protein interaction partners in affinity purification mass spectrometry experiments. Proteomics 2012, 12 (10), 1639–1655. Bantscheff, M.; Lemeer, S.; Savitski, M. M.; Kuster, B. Quantitative mass spectrometry in proteomics: critical review update from 2007 to the present. Anal. Bioanal. Chem. 2012, 404 (4), 939–965. Trinkle-Mulcahy, L. Resolving protein interactions and complexes by affinity purification followed by label-based quantitative mass spectrometry. Proteomics 2012, 12 (10), 1623– 1638. Ong, S.-E.; Blagoev, B.; Kratchmarova, I.; Kristensen, D. B.; Steen, H.; Pandey, A.; Mann, M. Stable isotope labeling by amino acids in cell culture, SILAC, as a simple and accurate approach to expression proteomics. Mol. Cell. Proteomics MCP 2002, 1 (5), 376– 386. Gygi, S. P.; Rist, B.; Gerber, S. A.; Turecek, F.; Gelb, M. H.; Aebersold, R. Quantitative analysis of complex protein mixtures using isotope-coded affinity tags. Nat. Biotechnol. 1999, 17 (10), 994–999. Ross, P. L.; Huang, Y. N.; Marchese, J. N.; Williamson, B.; Parker, K.; Hattan, S.; Khainovski, N.; Pillai, S.; Dey, S.; Daniels, S.; et al. Multiplexed protein quantitation in Saccharomyces cerevisiae using amine-reactive isobaric tagging reagents. Mol. Cell. Proteomics MCP 2004, 3 (12), 1154–1169. Tate, S.; Larsen, B.; Bonner, R.; Gingras, A.-C. Label-free quantitative proteomics trends for protein-protein interactions. J. Proteomics 2013, 81, 91–101. Keilhauer, E. C.; Hein, M. Y.; Mann, M. Accurate protein complex retrieval by affinity enrichment mass spectrometry (AE-MS) rather than affinity purification mass spectrometry (AP-MS). Mol. Cell. Proteomics MCP 2015, 14 (1), 120–135. Patel, V. J.; Thalassinos, K.; Slade, S. E.; Connolly, J. B.; Crombie, A.; Murrell, J. C.; Scrivens, J. H. A comparison of labeling and label-free mass spectrometry-based proteomics approaches. J. Proteome Res. 2009, 8 (7), 3752–3759. Webb-Robertson, B.-J. M.; Wiberg, H. K.; Matzke, M. M.; Brown, J. N.; Wang, J.; McDermott, J. E.; Smith, R. D.; Rodland, K. D.; Metz, T. O.; Pounds, J. G.; et al. Review, evaluation, and discussion of the challenges of missing value imputation for mass spectrometry-based label-free global proteomics. J. Proteome Res. 2015, 14 (5), 1993– 2001. Student. The Probable Error of a Mean. Biometrika 1908, 6 (1), 1–25. Benjamini, Y.; Hochberg, Y. Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. J. R. Stat. Soc. Ser. B Methodol. 1995, 57 (1), 289–300. Sowa, M. E.; Bennett, E. J.; Gygi, S. P.; Harper, J. W. Defining the human deubiquitinating enzyme interaction landscape. Cell 2009, 138 (2), 389–403.


23


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 24 of 30

(15) Choi, H.; Larsen, B.; Lin, Z.-Y.; Breitkreutz, A.; Mellacheruvu, D.; Fermin, D.; Qin, Z. S.; Tyers, M.; Gingras, A.-C.; Nesvizhskii, A. I. SAINT: probabilistic scoring of affinity purification-mass spectrometry data. Nat. Methods 2011, 8 (1), 70–73. (16) Al Rawi, S.; Louvet-Vallée, S.; Djeddi, A.; Sachse, M.; Culetto, E.; Hajjar, C.; Boyd, L.; Legouis, R.; Galy, V. Postfertilization autophagy of sperm organelles prevents paternal mitochondrial DNA transmission. Science 2011, 334 (6059), 1144–1147. (17) Manil-Ségalen, M.; Lefebvre, C.; Jenzer, C.; Trichet, M.; Boulogne, C.; SatiatJeunemaitre, B.; Legouis, R. The C. elegans LC3 acts downstream of GABARAP to degrade autophagosomes by interacting with the HOPS subunit VPS39. Dev. Cell 2014, 28 (1), 43–55. (18) Tian, Y.; Li, Z.; Hu, W.; Ren, H.; Tian, E.; Zhao, Y.; Lu, Q.; Huang, X.; Yang, P.; Li, X.; et al. C. elegans screen identifies autophagy genes specific to multicellular organisms. Cell 2010, 141 (6), 1042–1055. (19) Brenner, S. The genetics of Caenorhabditis elegans. Genetics 1974, 77 (1), 71–94. (20) Redeker, V.; Pemberton, S.; Bienvenut, W.; Bousset, L.; Melki, R. Identification of protein interfaces between α-synuclein, the principal component of Lewy bodies in Parkinson disease, and the molecular chaperones human Hsc70 and the yeast Ssa1p. J. Biol. Chem. 2012, 287 (39), 32630–32639. (21) Cox, J.; Mann, M. MaxQuant enables high peptide identification rates, individualized p.p.b.-range mass accuracies and proteome-wide protein quantification. Nat. Biotechnol. 2008, 26 (12), 1367–1372. (22) Cox, J.; Hein, M. Y.; Luber, C. A.; Paron, I.; Nagaraj, N.; Mann, M. Accurate proteomewide label-free quantification by delayed normalization and maximal peptide ratio extraction, termed MaxLFQ. Mol. Cell. Proteomics MCP 2014, 13 (9), 2513–2526. (23) Kamath, R. S.; Ahringer, J. Genome-wide RNAi screening in Caenorhabditis elegans. Methods San Diego Calif 2003, 30 (4), 313–321. (24) Alberti, A.; Michelet, X.; Djeddi, A.; Legouis, R. The autophagosomal protein LGG-2 acts synergistically with LGG-1 in dauer formation and longevity in C. elegans. Autophagy 2010, 6 (5), 622–633. (25) Klionsky, D. J.; Abdelmohsen, K.; Abe, A.; Abedin, M. J.; Abeliovich, H.; Acevedo Arozena, A.; Adachi, H.; Adams, C. M.; Adams, P. D.; Adeli, K.; et al. Guidelines for the use and interpretation of assays for monitoring autophagy (3rd edition). Autophagy 2016, 12 (1), 1–222. (26) Jenzer, C.; Simionato, E.; Legouis, R. Tools and methods to analyze autophagy in C. elegans. Methods San Diego Calif 2015, 75, 162–171. (27) Karpievitch, Y. V.; Taverner, T.; Adkins, J. N.; Callister, S. J.; Anderson, G. A.; Smith, R. D.; Dabney, A. R. Normalization of peak intensities in bottom-up MS-based proteomics using singular value decomposition. Bioinforma. Oxf. Engl. 2009, 25 (19), 2573–2580. (28) Clough, T.; Key, M.; Ott, I.; Ragg, S.; Schadow, G.; Vitek, O. Protein quantification in label-free LC-MS experiments. J. Proteome Res. 2009, 8 (11), 5275–5284. (29) Nakatogawa, H. Two ubiquitin-like conjugation systems that mediate membrane formation during autophagy. Essays Biochem. 2013, 55, 39–50. (30) Pu, S.; Vlasblom, J.; Turinsky, A.; Marcon, E.; Phanse, S.; Trimble, S. S.; Olsen, J.; Greenblatt, J.; Emili, A.; Wodak, S. J. Extracting high confidence protein interactions from affinity purification data: at the crossroads. J. Proteomics 2015, 118, 63–80.


24

Page 25 of 30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


(31) Mellacheruvu, D.; Wright, Z.; Couzens, A. L.; Lambert, J.-P.; St-Denis, N. A.; Li, T.; Miteva, Y. V.; Hauri, S.; Sardiu, M. E.; Low, T. Y.; et al. The CRAPome: a contaminant repository for affinity purification-mass spectrometry data. Nat. Methods 2013, 10 (8), 730–736. (32) Morris, J. H.; Knudsen, G. M.; Verschueren, E.; Johnson, J. R.; Cimermancic, P.; Greninger, A. L.; Pico, A. R. Affinity purification-mass spectrometry and network analysis to understand protein-protein interactions. Nat. Protoc. 2014, 9 (11), 2539–2554. (33) Jones, P.; Côté, R. G.; Martens, L.; Quinn, A. F.; Taylor, C. F.; Derache, W.; Hermjakob, H.; Apweiler, R. PRIDE: a public repository of protein and peptide identifications for the proteomics community. Nucleic Acids Res. 2006, 34 (Database issue), D659–D663. (34) van Nuland, R.; Smits, A. H.; Pallaki, P.; Jansen, P. W. T. C.; Vermeulen, M.; Timmers, H. T. M. Quantitative dissection and stoichiometry determination of the human SET1/MLL histone methyltransferase complexes. Mol. Cell. Biol. 2013, 33 (10), 2067– 2077. (35) Stark, C.; Breitkreutz, B.-J.; Reguly, T.; Boucher, L.; Breitkreutz, A.; Tyers, M. BioGRID: a general repository for interaction datasets. Nucleic Acids Res. 2006, 34 (Database issue), D535–D539. (36) Wild, P.; McEwan, D. G.; Dikic, I. The LC3 interactome at a glance. J. Cell Sci. 2014, 127 (Pt 1), 3–9. (37) Gao, W.; Kang, J. H.; Liao, Y.; Ding, W.-X.; Gambotto, A. A.; Watkins, S. C.; Liu, Y.-J.; Stolz, D. B.; Yin, X.-M. Biochemical isolation and characterization of the tubulovesicular LC3-positive autophagosomal compartment. J. Biol. Chem. 2010, 285 (2), 1371–1383. (38) Dengjel, J.; Høyer-Hansen, M.; Nielsen, M. O.; Eisenberg, T.; Harder, L. M.; Schandorff, S.; Farkas, T.; Kirkegaard, T.; Becker, A. C.; Schroeder, S.; et al. Identification of autophagosome-associated proteins and regulators by quantitative proteomic analysis and genetic screens. Mol. Cell. Proteomics MCP 2012, 11 (3), M111.014035. (39) Martin, D. N.; Balgley, B.; Dutta, S.; Chen, J.; Rudnick, P.; Cranford, J.; Kantartzis, S.; DeVoe, D. L.; Lee, C.; Baehrecke, E. H. Proteomic analysis of steroid-triggered autophagic programmed cell death during Drosophila development. Cell Death Differ. 2007, 14 (5), 916–923. (40) Kraft, C.; Deplazes, A.; Sohrmann, M.; Peter, M. Mature ribosomes are selectively degraded upon starvation by an autophagy pathway requiring the Ubp3p/Bre5p ubiquitin protease. Nat. Cell Biol. 2008, 10 (5), 602–610. (41) Ossareh-Nazari, B.; Bonizec, M.; Cohen, M.; Dokudovskaya, S.; Delalande, F.; Schaeffer, C.; Van Dorsselaer, A.; Dargemont, C. Cdc48 and Ufd3, new partners of the ubiquitin protease Ubp3, are required for ribophagy. EMBO Rep. 2010, 11 (7), 548–554. (42) Ewing, R. M.; Chu, P.; Elisma, F.; Li, H.; Taylor, P.; Climie, S.; McBroom-Cerajewski, L.; Robinson, M. D.; O’Connor, L.; Li, M.; et al. Large-scale mapping of human proteinprotein interactions by mass spectrometry. Mol. Syst. Biol. 2007, 3, 89. (43) Costanzo, M.; Baryshnikova, A.; Bellay, J.; Kim, Y.; Spear, E. D.; Sevier, C. S.; Ding, H.; Koh, J. L. Y.; Toufighi, K.; Mostafavi, S.; et al. The genetic landscape of a cell. Science 2010, 327 (5964), 425–431.


25

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44

Figure 1

Worm culture Journal of Proteome ResearchPage 26 of 30 Control: GFP 4 replicates 20 plates = 1 replicate

Experiment: GFP::LGG-1 4 replicates

Total proteins extraction In parallel for all replicates

Anti-GFP immunoprecipitation SDS-PAGE fractionation R1 R2 R3

Trypsin digestion NanoLC-MS/MS analysis Protein identification and quantification with MaxQuant Selection of candidates by MSdata filtering

ACS Paragon Plusof Environment Validation SAFER

Page 27 of 30

A

B

DATA Filter 1: Experiment

control

bait

Protein present in 2

Protein absent in ≥3 ctrl replicates

experiment

2

3

4

1

2

3

4

DEB-1

0

0

0

0

0

4.6

0

9.1

UBC-20

0

0

0

0

0

0

0

RPL-31

0

0

0

0

0

29.7

HAF-4

0

0

3.7

11.9

13.3

RPL-7

27.5

22.8

26.6

48.5

RACK-1

43.5

89.3

82.9

LARP-1

0

0

ACS-5

0

ATG-3 ATG-7

method SAFER

t-test

SAINT

MiST

+∞

-

-

+

-

2.4

+∞

-

-

-

+

0

47.4

+∞

-

-

+

+

16.8

15.8

8.1

1.7

-

+

-

+

22.3

59.8

64.5

96.9

1.9

-

-

-

-

115.7

212.6

148.8

140.2

176.4

2.1

+

-

-

-

13,3

11,5

0

45,7

57,8

34,6

3.7

+

-

-

-

0

0

19.5

7.9

16.4

14.4

11.2

0.6

+

-

-

-

0

0

0

0

0

14.7

8.3

4.9

+∞

+

-

+

+

0

0

0

0

5

131.9

113.7

44.8

+∞

+

+

+

+


Figure 2

FCS

1

prey

selected preys

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42

A

LGG-1

ASH2L

n=190 89 18

15 87

19 22

24 0

n=85

n=234

2 16

RBBP5

n=234

n=190 56

n=190

n=234 18

68 8

10 1

51

21

12

23

34

2

81 0

7 19

13

35

n=460

n=90 7

n=79

1 4

3 0

2

n=83

n=338

0

1

48

0

WDR82

n=90 n=210

1

69

40

Page 28 of 30

n=460 14

173 7

29 7

9

20

2

16

50

98

4

1

36

SAFER

135

25

n=438

43

t-test

118

SAINT

2

MiST

B LGG1

ASH2L SAINT

MiST SAFER t-test

SAINT

WDR82

t-test

SAINT

t-test

100%

54%

47%

96%

100%

58%

50%

89%

100%

66%

6%

90%

100%

59%

55%

90%

SAINT

24%

100%

34%

42%

52%

100%

53%

59%

70%

100%

10%

78%

77%

100%

53%

88%

MiST

21%

34%

100%

39%

45%

53%

100%

53%

6%

9%

100%

7%

52%

39%

100%

56%

SAFER

43%

42%

39%

100%

79%

59%

45%

100%

83%

69%

7%

100%

86%

65%

56%

100%

Figure 3

MiST SAFER t-test

RBBP5


MiST SAFER t-test

SAINT

MiST SAFER

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42

total GFP/tubulin 0.0 0.0

0

tubulin

Figure 4 C

1.0 1.0

1

0.5 0.5

0.5 5

0 0

0

ACS Paragon Plus Environment moyenne+ex*1,96 moyenne+ex*1,96

55

Y53G8AL.2

VF13D21L.3

T09A5.11

R05G6.7

F17C11.9

B0491.5

ucr-2.2

rack-1

mtch-1

lmp-1

lec-3

larp-1

inf-1

his-24

gly-8

fgt-1

cyc-1

atp-2

anc-1

acs-5

AUTOPHAGIC FLUX

Autophagosome

moyenne-ec*1,64 moyenne-ec*1,64

cleaved GFP

ecart-type ecart-type

35 25

moyenne+ec*1,64 moyenne+ec*1,64

total/tub total/tub

GFP/total GFP

GFP::LGG-1

Moyenne Moyenne

1.5

moyenne+ex*1,96 L4440 moyenne+ex*1,96 L4440 L4440 moyenne-ec*1,96 acs-5 moyenne-ec*1,96 acs-5 acs-5 anc-1 anc-1 anc-1 atp-2 atp-2 atp-2 cyc-1 cyc-1 cyc-1 fgt-1 fgt-1 fgt-1 gly-8 gly-8 gly-8 his-24 his-24 his-24 inf-1 inf-1 inf-1 larp-1 larp-1 larp-1 lec-3 lec-3 lec-3 lmp-1 lmp-1 lmp-1 mtch-1 mtch-1 mtch-1 rack-1 rack-1 rack-1 ucr-2.2 ucr-2.2 ucr-2.2 B0491.5 B0491.5 B0491.5 F17C11.9 F17C11.9 F17C11.9 R05G6.7 R05G6.7 R05G6.7 T09A5.11 T09A5.11 T09A5.11 VF13D21L.3 VF13D12L.3 VF13D12L.3 Y53G8AL.2 Y53G8AL.2 Y53G8AL.2

1.5 1.5

1 1

1

moyenne+ec*1,64 moyenne+ec*1,64 cleaved moyenne-ec*1,64 moyenne-ec*1,64

B

Moyenne Moyenne ecart-type ecart-type

kDa 55

L4440 L4440 L4440 acs-5 acs-5 acs-5 anc-1 anc-1 anc-1 atp-2 atp-2 atp-2 cyc-1 cyc-1 cyc-1 fgt-1 fgt-1 fgt-1 gly-8 gly-8 gly-8 his-24 his-24 his-24 inf-1 inf-1 inf-1 larp-1 larp-1 larp-1 lec-3 lec-3 lec-3 lmp-1 lmp-1 lmp-1 mtch-1 mtch-1 mtch-1 rack-1 rack-1 rack-1 ucr-2.2 ucr-2.2 ucr-2.2 B0491.5 B0491.5 B0491.5 F17C11.9 F17C11.9 F17C11.9 R05G6.7 R05G6.7 R05G6.7 T09A5.11 T09A5.11 T09A5.11 VF13D12L.3 VF13D21L.3 VF13D12L.3 Y53G8AL.2 Y53G8AL.2 Y53G8AL.2

A

L4440

Page 29 of 30 Journal of Proteome Research

Phagophore

Autolysosome

5

GFP/total GFP/total

4

4

3

3

2

2


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

1057x423mm (72 x 72 DPI)


Page 30 of 30

SAFER, an Analysis Method of Quantitative ... - ACS Publications

Recommend Documents