Impact of inherent biases built into proteomic techniques: Proximity labeling and affinity capture compared

Abstract

The characterization of protein–protein interactions (PPIs) is of high value for understanding protein function. Two strategies are popular for identification of PPIs direct from the cellular environment: affinity capture (pulldown) isolates the protein of interest with an immobilized matrix that specifically captures the target and potential partners, whereas in BioID, genetic fusion of biotin ligase facilitates proximity biotinylation, and labeled proteins are isolated with streptavidin. Whilst both methods provide valuable insights, they can reveal distinct PPIs, but the basis for these differences is less obvious. Here, we compare both methods using four different trypanosome proteins as baits: poly(A)-binding proteins PABP1 and PABP2, mRNA export receptor MEX67, and the nucleoporin NUP158. With BioID, we found that the population of candidate interacting proteins decreases with more confined bait protein localization, but the candidate population is less variable with affinity capture. BioID returned more likely false positives, in particular for proteins with less confined localization, and identified low molecular weight proteins less efficiently. Surprisingly, BioID for MEX67 identified exclusively proteins lining the inner channel of the nuclear pore complex (NPC), consistent with the function of MEX67, whereas the entire NPC was isolated by pulldown. Similarly, for NUP158, BioID returned surprisingly few PPIs within NPC outer rings that were by contrast detected with pulldown but instead returned a larger cohort of nuclear proteins. These rather significant differences highlight a clear issue with reliance on a single method to identify PPIs and suggest that BioID and affinity capture are complementary rather than alternative approaches. The characterization of protein–protein interactions (PPIs) is of high value for understanding protein function. Two strategies are popular for identification of PPIs direct from the cellular environment: affinity capture (pulldown) isolates the protein of interest with an immobilized matrix that specifically captures the target and potential partners, whereas in BioID, genetic fusion of biotin ligase facilitates proximity biotinylation, and labeled proteins are isolated with streptavidin. Whilst both methods provide valuable insights, they can reveal distinct PPIs, but the basis for these differences is less obvious. Here, we compare both methods using four different trypanosome proteins as baits: poly(A)-binding proteins PABP1 and PABP2, mRNA export receptor MEX67, and the nucleoporin NUP158. With BioID, we found that the population of candidate interacting proteins decreases with more confined bait protein localization, but the candidate population is less variable with affinity capture. BioID returned more likely false positives, in particular for proteins with less confined localization, and identified low molecular weight proteins less efficiently. Surprisingly, BioID for MEX67 identified exclusively proteins lining the inner channel of the nuclear pore complex (NPC), consistent with the function of MEX67, whereas the entire NPC was isolated by pulldown. Similarly, for NUP158, BioID returned surprisingly few PPIs within NPC outer rings that were by contrast detected with pulldown but instead returned a larger cohort of nuclear proteins. These rather significant differences highlight a clear issue with reliance on a single method to identify PPIs and suggest that BioID and affinity capture are complementary rather than alternative approaches. Most proteins function as part of multisubunit complexes, and identification of protein–protein interactions (PPIs) is valuable for understanding function. Identification of PPIs can uncover a wide range of interactions, which include direct, indirect, static, or dynamic binding. Furthermore, proteins can moonlight and engage in multiple different specific complexes, whereas complex composition can change in a temporal manner and/or a spatial manner. Presently, there are two common methods used to identify PPIs in a cellular context: affinity capture (colloquially pulldown) (1Dunham W.H. Mullin M. Gingras A. Affinity-purification coupled to mass spectrometry: basic principles and strategies.Proteomics. 2012; 12: 1576-1590Crossref PubMed Scopus (230) Google Scholar) and proximity labeling (2Bosch J.A. Chen C. Perrimon N. Proximity-dependent labeling methods for proteomic profiling in living cells: an update.Wiley Inter. Rev. Dev. Biol. 2021; 10: e392Crossref PubMed Scopus (43) Google Scholar). While evidence indicates that both methods provide valuable insights, there are considerable differences between them, both in technical requirements and consequently the interactome revealed. Each method has its adherents and while it would be fallacious to view one or the other approach as superior, it is unclear what choice of method implies in terms of the types of PPIs identified and hence critical assessment of a given dataset. Affinity capture requires cell rupture with the lysate or extract then exposed to a solid phase with a specific affinity to the protein of interest or bait. The solid phase is commonly coupled to an antibody against the bait, or alternatively, the bait is genetically fused to a tag and purified using a solid phase with affinity to the tag. The bait along with copurified interaction partners is then eluted, and this eluate is analyzed. A major disadvantage here is that interactions are determined from the cell extract and not the true cellular environment, and thus, some interactions may be lost, whereas spurious and nonphysiological interactions may occur as a result of membrane breakage and decompartmentalization of protein complexes. Screening of extraction conditions including detergent solubilization, pH, ionic strength, and other parameters is therefore necessary. A variation on this approach is cryomilling, which avoids the use of detergents during lysis, but still disrupts cellular organization (3Obado S.O. Field M.C. Chait B.T. Rout M.P. High-efficiency isolation of nuclear envelope protein complexes from trypanosomes.Met. Mol. Biol. (Clifton, NJ). 2016; 1411: 67-80Crossref PubMed Scopus (21) Google Scholar, 4LaCava J. Fernandez-Martinez J. Hakhverdyan Z. Rout M.P. Optimized affinity capture of yeast protein complexes.Cold Spring Harb. Protoc. 2016; https://doi.org/10.1101/pdb.prot087932Crossref Scopus (10) Google Scholar). In general, affinity capture requires a lot of optimization for each individual bait, and results tend to be more variable between experiments and/or labs because of minor differences in cell harvesting, lysis, and buffer conditions. In proximity labeling, some of the pitfalls of affinity isolation are avoided by capturing interactions in vivo. Here, the bait is fused to an enzyme that converts a substrate into a reactive radical that is covalently linked to nearby proteins. Modified proteins are purified frequently under extremely stringent conditions. The most commonly used method is BioID and variants. The bait is coupled to BirA∗, a mutant form of biotin ligase from Escherichia coli. Wildtype BirA converts biotin to biotinol-5′-AMP, which is retained by BirA until it is transferred to acetyl-CoA carboxylase. A mutant version, BirA∗, is modified to release biotinol-5′-AMP, causing biotinylation of lysine residues of proteins nearby (5Roux K.J. Kim D.I. Raida M. Burke B. A promiscuous biotin ligase fusion protein identifies proximal and interacting proteins in mammalian cells.J. Cell Biol. 2012; 196: 801-810Crossref PubMed Scopus (1398) Google Scholar, 6Chapman-Smith A. Cronan J.E. In vivo enzymatic protein biotinylation.Biomol. Eng. 1999; 16: 119-125Crossref PubMed Scopus (69) Google Scholar, 7Choi-Rhee E. Schulman H. Cronan J.E. Promiscuous protein biotinylation by Escherichia coli biotin protein ligase.Protein Sci. 2004; 13: 3043-3050Crossref PubMed Scopus (181) Google Scholar). Additional BioID variants have been developed, one of which is TurboID, which exhibits greater biotinylation efficiency (8Branon T.C. Bosch J.A. Sanchez A.D. Udeshi N.D. Svinkina T. Carr S.A. et al.Efficient proximity labeling in living cells and organisms with TurboID.Nat. Biotechnol. 2018; 36: 880-887Crossref PubMed Scopus (723) Google Scholar). TurboID biotinylates within minutes after addition of exogenous biotin, in contrast to other BirA∗ variants (8Branon T.C. Bosch J.A. Sanchez A.D. Udeshi N.D. Svinkina T. Carr S.A. et al.Efficient proximity labeling in living cells and organisms with TurboID.Nat. Biotechnol. 2018; 36: 880-887Crossref PubMed Scopus (723) Google Scholar), but also has some biotinylation activity in the absence of exogenous biotin, leading to an increased labeling radius (9May D.G. Scott K.L. Campos A.R. Roux K.J. Comparative application of BioID and TurboID for protein-proximity biotinylation.Cells. 2020; 9: 1070Crossref PubMed Scopus (76) Google Scholar). Given the distinct underlying principles of these two methods, we have carried out a direct comparison between them, employing four different proteins from trypanosome mRNA metabolism that encompass a variation of location and positional constraints. For affinity capture, we used data previously published by us (10Zoltner M. Krienitz N. Field M.C. Kramer S. Comparative proteomics of the two T. brucei PABPs suggests that PABP2 controls bulk mRNA.PLoS Negl. Trop. Dis. 2018; 12e0006679Crossref PubMed Scopus (20) Google Scholar, 11Obado S.O. M. Chait B.T. et the of the nuclear pore Biol. 2016; PubMed Scopus Google Scholar) and the for one nucleoporin NUP158, for Cell was by cryomilling, and cell PPIs and has been for nuclear pore complexes and other complexes from organisms (3Obado S.O. Field M.C. Chait B.T. Rout M.P. High-efficiency isolation of nuclear envelope protein complexes from trypanosomes.Met. Mol. Biol. (Clifton, NJ). 2016; 1411: 67-80Crossref PubMed Scopus (21) Google Scholar, S. C. A. et trypanosome a a in PubMed Scopus Google Scholar, T. S. et interactome from 2016; Scopus Google Scholar, M. et and of the yeast Mol. Biol. 2016; PubMed Scopus Google Scholar). proteins were as from the and with an (10Zoltner M. Krienitz N. Field M.C. Kramer S. Comparative proteomics of the two T. brucei PABPs suggests that PABP2 controls bulk mRNA.PLoS Negl. Trop. Dis. 2018; 12e0006679Crossref PubMed Scopus (20) Google Scholar, 11Obado S.O. M. Chait B.T. et the of the nuclear pore Biol. 2016; PubMed Scopus Google Scholar). For BioID, we the proteins from fused to the biotin ligase TurboID and used biotinylation, by biotin in the found surprisingly between PPIs identified from affinity capture and With BioID is for poly(A)-binding proteins with false In for the identification of MEX67 BioID was more and exclusively identified NPC that its inner rather than the entire which was isolated by the affinity affinity capture identified most part of the NPC cellular including proteins that are but with the bait, which are the BioID labeling Whilst affinity capture a of PPIs the of lysis, proximity labeling a of protein interactions during the of labeling, which can against of a complex in of dynamic data that the of each method is and hence be as complementary rather than as four trypanosome proteins as bait to BioID with TurboID biotin ligase PABP1 and PABP2 are and by within the under conditions (10Zoltner M. Krienitz N. Field M.C. Kramer S. Comparative proteomics of the two T. brucei PABPs suggests that PABP2 controls bulk mRNA.PLoS Negl. Trop. Dis. 2018; 12e0006679Crossref PubMed Scopus (20) Google Scholar, S. H. et a in and the of in of Cell Sci. PubMed Scopus Google Scholar). The nuclear export receptor MEX67 between the and the with M. and nuclear export of Biol. PubMed Scopus Google Scholar, S. M. in for proteins of and PubMed Scopus Google Scholar). The NPC protein to in and in is to the outer rings of the NPC (3Obado S.O. Field M.C. Chait B.T. Rout M.P. High-efficiency isolation of nuclear envelope protein complexes from trypanosomes.Met. Mol. Biol. (Clifton, NJ). 2016; 1411: 67-80Crossref PubMed Scopus (21) Google Scholar). these bait proteins for the we of biotinylation on a protein is and including a range of protein and this to be and we have cell and in some mass data (10Zoltner M. Krienitz N. Field M.C. Kramer S. Comparative proteomics of the two T. brucei PABPs suggests that PABP2 controls bulk mRNA.PLoS Negl. Trop. Dis. 2018; 12e0006679Crossref PubMed Scopus (20) Google Scholar, 11Obado S.O. M. Chait B.T. et the of the nuclear pore Biol. 2016; PubMed Scopus Google Scholar). four proteins were from fused to TurboID and the used for affinity capture, that in that fusion was to used two cell cells and cells fused to TurboID and an experiments were in form the brucei in the were to to proteins by a proteins were detected in whereas the cell including the proteins. were major differences in the of proteins for each bait and also in the of the detected in a of proteins than in both consistent with that PABP2 with a cohort of proteins with PABP1 (10Zoltner M. Krienitz N. Field M.C. Kramer S. Comparative proteomics of the two T. brucei PABPs suggests that PABP2 controls bulk mRNA.PLoS Negl. Trop. Dis. 2018; 12e0006679Crossref PubMed Scopus (20) Google Scholar). In we of the TurboID bait proteins by in biotinylation a For the we of the two and for distinct of the nuclear MEX67 TurboID a labeling as with some from the but were to and The absence of from the nuclear envelope can be by the by the which antibody in this of the pore but not binding. the labeling indicates high spatial labeling of TurboID we purified proteins by affinity for each cell in and by in the BioID were into and by in and on in of S. T. A. T. et for of 2016; 13: PubMed Scopus Google Scholar) the BioID were with cells to identify proteins as as proteins that to the affinity A cells a to identify proteins in a manner. For the proteins PABP1 and PABP2, we parameters to for both the bait proteins and the false proteins as in and in A and and which were from the of PABP1 and PABP2 candidate The of less parameters for the as used for or MEX67 would have in the of as for protein and the bait PABP1 the proteins were in or with or H. Z. protein in brucei controls by the of the Biol. PubMed Scopus Google Scholar, S. a trypanosome protein PubMed Scopus Google Scholar, H. K.J. et interactions in brucei identify of and PubMed Google Scholar, M. J. N. S. C. et into by employing the as a molecular PubMed Scopus Google Scholar, C. M. Kramer S. The nuclear of PubMed Scopus Google Scholar, B. S. M. et in brucei identified using 12: PubMed Scopus Google Scholar), as as proteins two of this cohort were to the BioID likely interactions, whereas the proteins were also identified in other BioID experiments in more than in also and to the less proteins with a confined the NUP158, and we less parameters to the A and and With these proteins were as of these in of these proteins were in the The were from the of or for proteins that were in for the bait proteins and in for in The cohort an in proteins with nuclear and two proteins with to the including the T. brucei NPC S.O. M. Chait B.T. et the of the nuclear pore Biol. 2016; PubMed Scopus Google Scholar, J.A. A. Field M.C. et for a nuclear pore complex that is from the common Cell PubMed Scopus Google Scholar). In few proteins were proteins In we determined the parameters to the data for each individual BioID bait to false positives, while including interaction not the protein from a true interacting was identified as here but was identified by BioID with protein and as to the indicates that this is likely a true H. K.J. et interactions in brucei identify of and PubMed Google Scholar). some of the and the within the cohort of may be true the for of the dataset. T. brucei has two PABP1 and PABP2 have under but PABP2 to cells are S. B. S. Field M.C. et of the two T. proteins to the and suggests to distinct mRNA Scopus Google Scholar, A. of to in PubMed Scopus Google Scholar) and can also to the under conditions S. B. S. Field M.C. et of the two T. proteins to the and suggests to distinct mRNA Scopus Google Scholar, C. M. et characterization of protein with distinct to and protein 9: PubMed Scopus Google Scholar). have previously determined an interactome for PABP1 and PABP2 by affinity capture using from cells of the PABPs to from (10Zoltner M. Krienitz N. Field M.C. Kramer S. Comparative proteomics of the two T. brucei PABPs suggests that PABP2 controls bulk mRNA.PLoS Negl. Trop. Dis. 2018; 12e0006679Crossref PubMed Scopus (20) Google Scholar). from four experiments with high and two with low were to a high of PPIs that proteins that were in four more than one in the bait and were in a cohort of proteins with of which two proteins and with extremely high of PABP2 proteins proteins with PABP1 or PABP2 have a function in T. brucei mRNA are mRNA and S. C. B. in into the of the 2016; PubMed Scopus Google or have a S. a trypanosome protein PubMed Scopus Google a cohort of In the of proteins identified with BioID was larger and and proteins in and and proteins in which of the proteins were also identified with BioID proteins identified in the PABP1 four were in of the BioID including the bait and the two proteins and A proteins were in proteins were not in the BioID of these are low molecular weight proteins and may be because of of which we as a of BioID in The two and were the proteins in the PABP1 pulldown and hence may suggest a less or more and with the proteins identified in the PABP2 are also in BioID including the bait and proteins with in the pulldown. A proteins were in the the are low molecular weight and may not be detected as The is a nuclear protein that in the to with the The protein proteins identified by BioID were also in the pulldown the proteins identified in the PABP1 BioID were in two of the were in one were not but and were not the proteins from the PABP2 BioID we detected in two of the of the a in one not but and proteins proteins are to be in mRNA these data we evidence for in mRNA metabolism for each on and published and of PPIs identified by BioID in for PABP1 and PABP2, evidence for a in mRNA the of proteins with a in mRNA metabolism was the BioID proteins that were also identified by affinity capture and the BioID proteins by affinity While there is between BioID and pulldown BioID identifies a larger of potential by the of interactions and the dynamic of complexes in which the function. with a S. J. J. M. of and two proteins in Biol. PubMed Scopus Google Scholar) BioID to the PPIs of two T. brucei and a of BioID to have a the cohort of is likely because of the of PABP1 and PABP2 and in addition by PABPs are and with complexes that may have proteins not in mRNA PABPs are in and in proximity to proteins that they the absence of direct that proteins were not found biotinylation of the entire MEX67 a with and this complex is the major mRNA export complex and is most A. A. J. et a for nuclear mRNA to both and nuclear J. 16: PubMed Scopus Google Scholar, J. A. M. E. The nuclear mRNA export is from yeast to J. 1999; PubMed Scopus Google Scholar). MEX67 its mRNA both or proteins and mRNA export to the by interacting with While mRNA export in in from the in S. mRNA and mRNA export from to 2021; PubMed Scopus Google Scholar), has been in the and to have a function in mRNA export S. M. in for proteins of and PubMed Scopus Google Scholar, S. mRNA and mRNA export from to 2021; PubMed Scopus Google Scholar, A. T. C. N. M. et of in the of in PubMed Scopus Google Scholar, A. S. M. B. The nuclear mRNA export receptor of brucei a and PubMed Scopus Google Scholar). Two from are for T. brucei A with a tag identified and a receptor for nuclear A. S. M. B. The nuclear mRNA export receptor of brucei a and PubMed Scopus Google Scholar), whereas a affinity capture using a MEX67 as bait identified S.O. M. Chait B.T. et the of the nuclear pore Biol. 2016; PubMed Scopus Google Scholar). the affinity capture data S.O. M. Chait B.T. et the of the nuclear pore Biol. 2016; PubMed Scopus Google Scholar) for different extraction conditions to a of proteins have then BioID with MEX67 to compare the PPIs identified by these different In contrast to the PABP1 and PABP2 the of MEX67 PPIs identified with BioID was to the of PPIs identified by pulldown suggest that this is likely a of the more confined of the MEX67 which results in less MEX67 BioID identified proteins with a localization, both and and the of proteins were to the or the NPC whereas of proteins to the with nuclear identified by BioID be proteins that with nuclear proteins in the pulldown PPIs from the BioID PPIs in a larger of and a of in for a of proteins. specific to the pulldown PPIs were proteins to the and nuclear these likely with the but are not in with MEX67 and hence not with BioID it be that or proteins also be to the BioID labeling but not to direct affinity capture. While MEX67 the entire MEX67 BioID identified of proteins exposed to channel The identified by BioID is which is also with the and may have to for MEX67, BioID proteins interacting with MEX67, whereas the pulldown identified the entire likely by interactions between individual to there was between the PPIs identified by the two different methods, and proteins of were common and Additional to MEX67, there were proteins with to the NPC or nuclear to but a its from yeast is a of that to the of the NPC B. H. et is an mRNA export that to of nuclear pore Biol. PubMed Scopus Google Scholar), and two proteins of and a protein C. Field M.C. of Biol. PubMed Scopus Google is likely in the mRNA the S.O. M. Chait B.T. et the of the nuclear pore Biol. 2016; PubMed Scopus Google Scholar). and S.O. M. E. S. M. et of in trypanosome and Scopus Google Scholar), in the were also in the BioID but a Two proteins common to both BioID and pulldown are to the and protein of the and S. a trypanosome protein PubMed Scopus Google Scholar, H. Z. and of the in Cell Sci. PubMed Scopus Google Scholar), both likely Surprisingly, the major MEX67 was identified by likely because of the low molecular weight of proteins were to the of these in A. Most of these to the NPC or nuclear envelope and most proteins including protein and proteins. proteins to the nuclear envelope and nuclear including the and whereas or In of proteins to the BioID a nuclear localization, four with NPC and or proteins to the BioID were and protein is in nuclear export of mRNA of a to be by of or MEX67 M. E. A. A. et both mRNA and nuclear export in an PubMed Scopus Google Scholar). the function of and in export and is of and nuclear export for mRNA and for which there is also evidence in yeast M. E. A interaction on the receptor an between mRNA and J. PubMed Scopus Google Scholar). While the absence of and in the pulldown against a interaction with MEX67, high in BioID suggests interactions with the proteins were and proteins with in nuclear mRNA including nuclear an the the and proteins in the of proteins to be in nuclear mRNA metabolism S. mRNA and mRNA export from to 2021; PubMed Scopus Google Scholar) was not identified with method both BioID and pulldown detected the interactions of MEX67 with In these are likely the major as in MEX67 can be fused to a nucleoporin and still its S. E. The export as a Cell Biol. PubMed Google Scholar). from these there was between the PPIs identified by the two While the pulldown the entire BioID identified nuclear proteins in mRNA metabolism and interactions between MEX67 and its mRNA that are in the pulldown. BioID is a valuable for of MEX67 function as it of the NPC in direct as the exposed the inner pore the other the pulldown interactions because of high of NPC The as MEX67 is to some of the of a because of its significant the NPC S. E. The export as a Cell Biol. PubMed Google Scholar). Furthermore, BioID to identify interactions, as by the of proteins with a function in mRNA and nuclear localization, is an and of outer and affinity capture was as S.O. M. Chait B.T. et the of the nuclear pore Biol. 2016; PubMed Scopus Google Scholar) but the entire to detected into and into and these are and with NUP158, these proteins form the outer complex S.O. M. Chait B.T. et the of the nuclear pore Biol. 2016; PubMed Scopus Google Scholar). significant are the nuclear protein and inner proteins and proteins as the inner proteins and the and and the protein were also but the proteins were identified as significant of which to the NPC and to the S. a trypanosome protein PubMed Scopus Google Scholar). BioID identified proteins and proteins in a in that also in the and proteins in In comparison to the and MEX67 BioID this is the of likely the confined of NUP158. these to the to the and to the envelope S. a trypanosome protein PubMed Scopus Google Scholar). between BioID and pulldown PPIs was identified as interaction partners in the pulldown were not identified in the BioID and were to the BioID and or the pulldown and The in common were and and there was one between pulldown and BioID is a protein of function but to the NPC S. a trypanosome protein PubMed Scopus Google both methods identify but PPIs for NUP158. The pulldown identifies outer proteins that are in A potential for the absence be that NUP158, within the the outer rings of the has a labeling radius because of or the pulldown is identification of PPIs for a valuable data can be with we found surprisingly between proteins identified by BioID and by affinity capture. suggests that these methods be as a distinct for a given bait rather than or A to the one here has been for protein complexes in cell M. C. Gingras biotinylation and affinity are complementary for the interactome of protein PubMed Scopus Google Scholar, B. Gingras of interaction by BioID and affinity coupled to mass Mol. Biol. PubMed Scopus Google Scholar) with BioID larger and the between pulldown and BioID was that both identified a cohort of bait proteins with greater variation of location and positional to provide

Topics & Concepts