Proximity proteomics reveals role of Abelson interactor 1 in the regulation of TAK1/RIPK1 signaling

Dysregulation of the adaptor protein Abelson interactor 1 (ABI1) is linked to malignant transformation. To interrogate the role of ABI1 in cancer development, we mapped the ABI1 interactome using proximity‐dependent labeling (PDL) with biotin followed by mass spectrometry. Using a novel PDL data filtering strategy, considering both peptide spectral matches and peak areas of detected peptides, we identified 212 ABI1 proximal interactors. These included WAVE2 complex components such as CYFIP1, NCKAP1, or WASF1, confirming the known role of ABI1 in the regulation of actin‐polymerization‐dependent processes. We also identified proteins associated with the TAK1‐IKK pathway, including TAK1, TAB2, and RIPK1, denoting a newly identified function of ABI1 in TAK1‐NF‐κB inflammatory signaling. Functional assays using TNFα‐stimulated, ABI1‐overexpressing or ABI1‐deficient cells showed effects on the TAK1‐NF‐kB pathway‐dependent signaling to RIPK1, with ABI1‐knockout cells being less susceptible to TNFα‐induced, RIPK1‐mediated, TAK1‐dependent apoptosis. In sum, our PDL‐based strategy enabled mapping of the ABI1 proximal interactome, thus revealing a previously unknown role of this adaptor protein in TAK1/RIPK1‐based regulation of cell death and survival.


Introduction
Signal transduction is central to the maintenance of cellular homeostasis, control of cell metabolism, response to external environmental cues, cell growth, proliferation, and death.Signaling dysregulation drives all malignancy.The functional nodes of signal transduction are structured complexes comprising multiple proteins.Resolving the composition, organization, and regulation of these complexes enables new therapeutic approaches that are better targeted to a protein's specific dysregulated context.Adaptor proteins lack enzymatic activity but contain structural motifs that facilitate both protein-protein interactions and context-specific organization of protein signaling complexes.Adaptor proteins are frequently dysregulated in cancer.Therefore, detailed mapping of adaptor protein interactomes has the potential to offer new insights into cancer pathophysiology and therapy [1].
Proximity-dependent labeling followed by mass spectrometry (PDL/MS) is a powerful approach to identify both directly and indirectly interacting protein partners involved in steady-state and transient cell signaling.PDL uses cellular expression of a fusion protein comprising a bait and inducible enzyme that catalyzes production of reactive substrates, which tag proximal proteins.Several methods for PDL in living cells are available, including a biotinylation/streptavidin-affinity capture-based method enabled by TurboID.TurboID is an Escherichia coli-derived biotin ligase, catalytically enhanced through directed evolution, which can be genetically fused to a protein of interest.Cellular Tur-boID is induced by exogenous biotin addition to produce reactive biotinoyl-5'-AMP, which rapidly binds exposed lysine residues on proximal proteins.Biotinylated proteins are pulled down from cell lysates using streptavidin affinity, then identified by MS to map the bait protein's proximal interactome [41,42].
Here we report the identification of the ABI1 proximal interactome using TurboID-based proximity labeling coupled with label-free, quantitative MS.PDL data are complemented with functional validations, using both ABI1 overexpressing (ABI OE) and ABI1 deficient (ABI KO) cells.Our obtained PDL results, while confirming the known role of ABI1 in the regulation of actin polymerization-dependent processes, reveal a new function of this adaptor protein in regulation of RIPK1 in TNFR-mediated cell death signaling.

Immunofluorescence imaging
Preparation and staining for immunofluorescence imaging were performed as described previously [2].Information about antibodies and stains is provided in Table S1.Confocal images were taken on a Nikon C1si (Tokyo, Japan) confocal microscope.

SDS/PAGE was performed in 19 NuPAGE MES SDS
Running Buffer (Thermo Fisher, NP0002).Gel electrophoresis was run on gradient gels (Thermo Fisher, WG1402BOX).Protein transfer to nitrocellulose membranes was performed in transfer buffer containing 20% methanol and 19 NuPAGE Transfer Buffer (Thermo Fisher, NP00061).Blocking was conducted using 5% milk in TBS-T, except for streptavidin, BirA (TurboID), and phosphorylated protein blots which used 5% BSA in TBS-T.Primary and secondary antibody incubations were conducted in 2% milk in TBS-T (except for streptavidin, BirA, and phosphorylated protein blots, which used 2% BSA in TBS-T).Signal was detected using SuperSignal West Pico PLUS Chemiluminescent Substrate (Thermo Fisher, 34579).Information about antibodies is provided in Table S1.

Wound healing assay
Stable NIH/3T3 cell lines expressing TurboControl or TurboABI1 (n = 6) were seeded in 6-well culture plates at 1e6 cells per well and incubated overnight at 37 °C, 5% CO 2 .100% confluent cells were washed with PBS, scratched with a 10 lL pipette tip, washed again, then imaged at timepoints indicated in Fig. S2, until wound closure.Images were captured using ECHO Revolve microscope system (San Diego, CA, USA) at 109 magnification.Wound healing was quantified using IMAGEJ (Bethesda, MD, USA) [43], and statistical comparisons were made between TurboControl and Tur-boABI1 groups using Student's two-tailed t-test.

EdU incorporation analysis
Cell lines from TurboControl and TurboABI1 (n = 6) were plated in 6-well plates at 2e5 cells per well then incubated overnight at 37 °C, 5% CO 2 .EdU incorporation and analysis were conducted using Click-iT Plus EdU Flow Cytometry Assay Kit with Alexa Fluor 647 picolyl azide (Thermo Fisher, C10634), according to the manufacturer's protocol.EdU incorporation was measured using a BD II LSR flow cytometer.Statistical comparisons were made between groups using Student's two-tailed t-test.

Proximity labeling and biotinylated protein enrichment
1.8e7 total cells per cell line were plated onto 10-cm tissue culture dishes, at 7.5e5 cells per dish containing complete DMEM culture medium and 19 P/S.Cells were incubated at 37 °C, 5% CO 2 overnight, and labeling was induced by adding biotin (Invitrogen, Waltham, MA, USA, B1595) from fresh 0.2 lm filtered 100 mM stock in DMEM to a final biotin concentration of 50 lM, then mixed by tilting.Plates were incubated at 37 °C, 5% CO 2 for 3 h before labeling was stopped by aspirating medium and washing twice with ice-cold PBS.Cells were scraped and pelleted at 3009g for 10 min while maintaining cold conditions.Cell pellets were lysed in 1 mL RIPA (Thermo Fisher, 89900) containing inhibitors (1 mM sodium orthovandate, 10 mM sodium pyrophosphate, 10 mM sodium fluoride, 19 protease inhibitor cocktail set III (EMD Millipore, 539134)).Lysis was conducted on ice for 15 min with vortexing every 5 min, then lysates cleared by centrifugation at 4 °C for 15 min at 20 0009g.Magnetic streptavidin beads (Pierce, Waltham, MA, USA, 88817) were washed in lysis buffer with inhibitors three times before adding cleared lysates and incubating overnight at 4 °C with rotation.Bound beads were then washed three times with 1 mL RIPA plus inhibitors and resuspended in 0.3 mL RIPA plus inhibitors before transport to the RIH CCCRD proteomics core facility.Small-scale biotinylation experiments were conducted as above but using 1e6 cells per 10-cm dish, and 500 lM biotin for 10 min, for analysis by SDS/PAGE.

Label-free proteomics
Label-free proteomics was performed as previously described [44].Samples were exposed to overnight onbead tryptic digestion at 37 °C on a rotator.Tryptic peptides were desalted using C18 Sep-Pak plus cartridges (Waters, Milford, MA, USA) and were lyophilized for 48 h to dryness.The dried peptides were reconstituted in 30 lL of buffer A (0.1 M acetic acid) and 5 lL was injected for each analysis.
The LC-MS/MS was performed on a fully automated proteomic technology platform that includes an Agilent 1200 Series Quaternary HPLC system (Agilent Technologies, Santa Clara, CA, USA) connected to a Q Exactive Plus mass spectrometer (Thermo Fisher Scientific, Waltham, MA, USA).The LC-MS/MS set up was used as described [44].Briefly, the peptides were separated through a linear reversed-phase 90 min gradient from 0% to 40% buffer B (0.1 M acetic acid in acetonitrile) at a flow rate of 3 lLÁmin À1 through a 3-lm 9 20 cm C18 column (OD/ID 360/75, Tip 8 lm, New Objectives, Woburn, MA, USA) for a total of 90 min run time.The electrospray voltage of 2.0 kV was applied in a splitflow configuration, and spectra were collected using a top-9 data-dependent method.Survey full-scan MS spectra (m/z 400-1800) were acquired at a resolution of 70 000 with an AGC target value of 3 9 10 6 ions or a maximum ion injection time of 200 ms.The peptide fragmentation was performed via higher-energy collision dissociation with the energy set at 28 normalized collision energy (NCE).The MS/MS spectra were acquired at a resolution of 17 500, with a targeted value of 2 9 10 4 ions or maximum integration time of 200 ms.The ion selection abundance threshold was set at 8e2 with charge state exclusion of unassigned and z = 1, or 6-8 ions and dynamic exclusion time of 30 s.

Database search and label-free quantitative analysis
Database search and label-free quantitative analysis was performed as previously described [45].Peptide spectrum matching of MS/MS spectra of each file was searched against the Uniprot Mus musculus database (TaxonID: 10090, downloaded on 02/09/2015) using the Sequest algorithm within PROTEOME DISCOVERER v 2.3 software (Thermo Fisher Scientific, San Jose, CA, USA).The Sequest database search was performed with the following parameters: trypsin enzyme cleavage specificity, two possible missed cleavages, 10 ppm mass tolerance for precursor ions, and 0.02 Da mass tolerance for fragment ions.Search parameters permitted dynamic modification of methionine oxidation (+15.9949Da) and static modification of carbamidomethylation (+57.0215Da) on cysteine.Peptide assignments from the database search were filtered down to a 1% FDR (false discovery rate).The relative labelfree quantitative and comparative among the samples were performed using the Minora algorithm and the adjoining bioinformatics tools of the PROTEOME DISCO- VERER 2.3 software.

TMT proteomics and data analysis
TMT proteomics and data analyses were performed as previously described [46,47].Proteins from flash-frozen cell pellets (1e6 cells per pellet, n = 3 per cell line) were reduced, alkylated, and purified by chloroform/methanol extraction prior to digestion with sequencing grade modified porcine trypsin (Promega, Madison, WI, USA).Tryptic peptides were labeled using tandem mass tag isobaric labeling reagents (Thermo Fisher) following the manufacturer's instructions and combined into one 10-plex sample group.The labeled peptide multiplex was separated into 46 fractions on a 100 9 1.0 mm Acquity BEH C18 column (Waters) using an UltiMate 3000 UHPLC system (Thermo Fisher) with a 50 min gradient from 99:1 to 60:40 buffer A:B ratio under basic pH conditions, and then consolidated into 18 super-fractions.Each super-fraction was then further separated by reverse phase XSelect CSH C18 2.5 lm resin (Waters) on an in-line 150 9 0.075 mm column using an UltiMate 3000 RSLCnano system (Thermo Fisher).Peptides were eluted using a 60 min gradient from 98:2 to 60:40 buffer A:B ratio.Buffer A-0.1% formic acid, 0.5% acetonitrile, buffer B-0.1% formic acid, 99.9% acetonitrile, both buffers were adjusted to pH 10 with ammonium hydroxide for offline separation.
Eluted peptides were ionized by electrospray (2.2 kV) followed by mass spectrometric analysis on an Orbitrap Eclipse Tribrid mass spectrometer (Thermo Fisher) using multi-notch MS3 parameters with real-time search enabled.MS data were acquired using the FTMS analyzer in top-speed profile mode at a resolution of 120 000 over a range of 375 to 1500 m/ z.Following CID activation with a normalized collision energy of 35.0, MS/MS data were acquired using the ion trap analyzer in centroid mode and normal mass range.Using synchronous precursor selection, up to 10 MS/MS precursors were selected for HCD activation with a normalized collision energy of 65.0, followed by acquisition of MS3 reporter ion data using the FTMS analyzer in profile mode at a resolution of 50 000 over a range of 100-500 m/z.Proteins were identified and MS3 reporter ions quantified using MAX- QUANT (Max Planck Institute, Munich, Germany) against the UniprotKB database with a parent ion tolerance of 3 ppm, a fragment ion tolerance of 0.5 Da, and a reporter ion tolerance of 0.003 Da.Scaffold Q + S (PROTEOME Software) was used to verify MS/ MS-based peptide and protein identifications [48].
Protein TMT MS3 reporter ion intensity values were assessed for quality using in-house the PROTEINORM app, a user-friendly tool for a systematic evaluation of normalization methods, imputation of missing values, and comparisons of different differential abundance methods [47].Popular normalization methods were evaluated including log2 normalization (Log2), median normalization (Median), mean normalization (Mean), variance stabilizing normalization (VSN) [49], quantile normalization (Quantile) [50], cyclic loess normalization (Cyclic Loess) [51], global robust linear regression normalization (RLR) [52], and global intensity normalization (Global Intensity) [52].The normalized data were used to perform statistical analysis using Linear Models for Microarray Data (LIMMA) with empirical Bayes (eBayes) smoothing to the standard errors [51].Proteins with an FDR-adjusted P-value < 0.05 and a fold change > 2 were considered significant.

Coimmunoprecipitation
WT NIH/3T3 grown to 70% confluency were washed twice with prewarmed PBS, then lysed on ice for 5 min with ice-cold lysis buffer containing 25 mM Tris-HCl (pH 7.4), 150 mM NaCl, 5% glycerol, 1% NP-40, and protease/phosphatase inhibitors.Cells were scraped, then collected lysates were passed 20 times through a 25.5 G needle, then five times through a 21.5 G needle.Lysates were cleared by centrifuging at 20 0009g for 10 min at 4 °C.10% cleared lysate volume was collected for lysate analysis.Five microgram mouse IgG (Jackson ImmunoResearch, West Grove, PA, USA, 015-000-003) and 50 lL lysis-bufferwashed Protein G Dynabeads (Invitrogen, 10004D) were added to the remaining cleared lysate and incubated with rotation overnight at 4 °C.Supernatant was separated from beads using a magnet, then split evenly between two tubes for pulldowns.Five microgram mouse IgG1 isotype control (Cell Signaling Technology, Danvers, MA, USA, 5415S) was added to one tube, and 5 lg anti-ABI1 (clone 1B9, MBL) to the other.Tubes were incubated with rotation overnight at 4 °C, then added to 50 lL washed Protein G beads and incubated with rotation overnight at 4 °C.Beads were then washed three times in wash buffer of the same composition as the lysis buffer above, except with 0.1% NP-40.Sixty microliter sample reducing buffer was added to the beads, then samples were heated for 10 min at 98 °C.Supernatants were separated from beads using a magnet, then analyzed by western blot.Information about antibodies is available in Table S1.

TNFa stimulation
ABI1 deficient [Abi1(À/À) KO MEF cells: clone#3-6 (KO1), clone#3-11 (KO2) and Abi1(fl/fl) WT MEF cells, clone#3-PP (KOC)].Cells were a kind gift from Dr. Leszek Kotula [2].These cells were derived from mouse embryos as described [2].These cells were not authenticated by Short Tandem Repeat (STR) profiling.Cells were tested for mycoplasma upon receipt, when cell stocks are generated, and every 3 months on all actively growing cells using the MycoAlert mycoplasma Detection Kit (Lonza, LT07-318) and DAPI staining.ABI1 overexpressing (OE) cells and ABI1 OE control (OEC) were TurboABI1 #1 and TurboControl #1 cell lines.7.5e5 cells per condition were seeded in 10 cm plates and grown overnight in complete DMEM + P/S.Cells were washed once with prewarmed PBS, then 8 mL prewarmed complete DMEM or complete DMEM supplemented with 30 ngÁmL À1 TNFa (Stemcell Technologies, Vancouver, BC, Canada, 78069.1)was added to each plate.Cells were incubated at 37 °C for 30 min or 12 h, then washed once with ice-cold PBS.Two hundred and fifty microliter ice-cold RIPA plus inhibitors was added to each plate and plates were incubated on ice for 5 min.Cells were scraped, then collected lysates were centrifuged at 20 0009g for 15 min at 4 °C.Cleared supernatants were prepared for SDS/PAGE.The experiment was repeated three times.Densitometry comparing ABI1 KO and ABI1 OE proteins was reported as a ratio to unstimulated control cells, normalized to GAPDH for RIPK1 and c-CASP3, and further normalized to RIPK1 for RIPK1 phosphosites (n = 3 per protein).
For stimulation studies aimed at detecting phosphorylated TAK1, preparation conditions were like those described above, but 10 ngÁmL À1 TNFa and 10 nM Calyculin A (Cell Signaling Technology, 9902S) were used for 15 min incubation with ABI1 KO and ABI1 KOC cells.Densitometry presented is from three replicate experiments, measured using IMAGEJ [43].Densitometry from ABI1 KO MEF clones was averaged and compared to ABI1 KOC MEF clone per experiment, and statistical tests were performed on ABI1 KO experimental average vs.ABI1 KOC densitometry measurements using two-tailed Student's t-test.Information about antibodies is available in Table S1.
For stimulation assays in the presence of TNFR inhibitors, 100 ngÁmL À1 TNFa was used.Takinib (Med-Chem Express, Monmouth Junction, NJ, USA, HY-103490) was used at a final concentration of 100 nM, necrostatin (MedChem Express, HY-15760) was used at a final concentration of 500 nM, and Z-VAD (R&D Systems, Minneapolis, MN, USA, FMK001) was used at a final concentration of 10 lM.Annexin V orange dye (Sartorius, Goettingen, Germany, 4641) was used at a final dilution of 1:3200.The assay was performed on ABI1 KO and ABI1 KOC cells plated on 96-well plates in quadruplicate.INCUCYTE AUTO-ANALYZER software was used to quantify confluency by phase and apoptosis by annexin V image-integrated intensity.Annexin V was normalized to confluency per timepoint, and all timepoints were reported as a ratio of time = 0.
To determine the phenotypic differences between Tur-boABI1 and TurboControl-expressing cell lines, we assessed their wound-healing capability by scratch assay over 40 h.We did not observe significant differences in wound closure between the TurboABI1 and TurboControl cell lines (Fig. 1C and Fig. S2).We also assessed cell cycle status by EdU incorporation assay and did not observe significant differences in cell cycle status between the TurboABI1 and TurboControl cell lines (Fig. 1D).Based on these results, we concluded that the single cellderived cell lines stably expressing biotin ligase or biotin ligase tagged ABI1 had similar growth phenotypes.
The TurboABI1 #1 and TurboControl #1 clonal cell lines were selected for PDL/MS experiments.TurboID localization and expression in the two cell lines were similar as measured by immunofluorescence and immunoblotting (Fig. 2A,B).We observed increased biotinylation in TurboControl #1 and TurboABI1 #1 upon incubation with biotin, confirming the catalytic activity of TurboID.Wild type (WT) NIH/3T3 cells, as expected, did not show increased biotinylation upon biotin induction (Fig. 2C).
To determine the global effect of ABI1 overexpression in the TurboABI1 clone, we measured protein expression differences between TurboABI #1, Turbo-Control #1, and WT NIH/3T3 cell lines using Tandem Mass Tag (TMT) proteomics.We identified 106 proteins that showed a statistically significant (adjusted Pvalue ≤ 0.05) expression level difference of 2-fold or greater between the TurboABI1 #1 and TurboControl #1 cell lines.Of these 106 proteins, 82 were upregulated, and 24 were downregulated in TurboABI1 #1 cells compared to TurboControl #1 cells (Fig. 2D,E).Gene ontology (GO) biological process analysis of these 106 proteins identified significantly enriched processes of oxidation-reduction, cell-cell adhesion, and response to oxidative stress in TurboABI1 #1 when compared to TurboControl #1 cells (Fig. 2F).StringDB analysis of these proteins identified an interaction cluster associated with the oxidoreductase complex (Fig. 2D).The average log 2 FC values of ABI1 upregulation in the TurboABI1 #1 cell line were 4.8 and 4.6 in comparison to the TurboControl #1 and WT cell lines, respectively (Table S9).Based on these observations, we concluded that levels of ABI1 overexpression in the TurboABI1 clone, while modestly affecting the phenotype of the cells, would enable detection of proximal interactions that might otherwise remain below detection limits.

Proximity labeling experiments and quantitative analysis of ABI1 PDL/MS data
To establish the ABI1 interactome, we performed three independent PDL/MS experiments using TurboABI1 #1, TurboControl #1, and WT NIH/3T3 cell lines.We performed three technical replicate MS injections for each independent experiment, resulting in a total of 27 MS runs (Fig. 3A).To assure stringent data filtering, we used both protein peak area (PA; also known as protein ion intensity or abundance) and peptide spectral match (PSM) data for quantitative comparisons between TurboABI1 and controls.PA and PSM data both showed log-normal distribution (Fig. S3); therefore, significance testing was performed on logtransformed data.TurboABI1 measurements were compared to TurboControl and WT controls based on one-tailed, TurboABI1-sided significance tests to determine probable ABI1 interactors in the dataset (Fig. 3B).Principal component analyses indicated that PA and PSM data showed clustering by cell line and experiment (Fig. 3C).In three independent experiments, a total of 4082 unique proteins were identified, with 2997 proteins identified as enriched in the Tur-boABI1 #1 group (Fig. 3D).Taking into consideration only proteins identified with an average of 1 or more PSM per injection resulted in a higher proportion of targets with PA P-value ≤ 0.05 for the TurboABI1 #1 vs. TurboControl #1 comparison (Fig. 3E).Interestingly, more known ABI1 interactors were identified within the group of targets with PA FDR ≤ 0.05 for TurboABI1 #1 vs. TurboControl #1 (Fig. 3F).Together, these data indicated an expected directionality of TurboABI1 #1 labeling while informing optimal FDR and average TurboABI #1 PSM thresholds.To determine optimal TurboABI1 #1 vs. TurboControl #1 PA and PSM ratios defining the most probable ABI1 proximal interactors, we plotted PA and PSM TurboABI1 #1:TurboControl #1 ratios vs. the number of known ABI1 interactors identified.For PA and PSM, a TurboABI1 #1 vs. TurboControl #1 ratio of 1.5 was close to the best-fit line (Fig. 3G).This agreed with the TurboID/BioID field standard reporting cutoff of a 1.5-fold enrichment.Labeling intensity in WT cell lines was negligible compared to TurboABI1 #1 and TurboControl #1 (Fig. S3).Considering these statistical interpretations, we defined the most probable ABI1 proximally interacting hits as follows: Tur-boABI1 #1 vs. controls PA ratio ≥ 1.5 (FDR ≤ 0.05), PSM ratio ≥ 1.5 (FDR ≤ 0.05), and minimum average 1 PSM per TurboABI1 #1 MS injection.A webapp that functions as a user interface to interpret this dataset under different statistical parameters was created and is available at https://maxpetersen.shinyapps.io/turboabi_data_ui_v2/ (GitHub: https://github.com/maxjohan/turboabi_data_ui, Table S6).We conclude from our analyses that utilization of both PA and PSM for quantitative MS data interpretation, in combination with analysis of MS data based on enrichment of known interactors, represents a stringent approach to establish PDL/MS-based interactomes with high statistical confidence.

Known and new ABI1 interactors identified by PDL/MS
Of the 4082 proteins identified by PDL/MS, 212 proteins met the aforementioned filtering criteria.As expected, ABI1 itself was the most labeled protein by average TurboABI1 PSM (Fig. 4B).Of the 212 proteins, 32 were known ABI1 interactors based on curated data in the StringDB, MINT, IntAct, and Bio-GRID interaction evidence databases (Table S8).Gene Ontology (GO) Biological Process (BP) enrichment analysis of the 212 ABI1 proximal interactors yielded both expected and unexpected significant (FDR ≤ 0.05) biological processes (BP).As anticipated, processes linked to actin cytoskeleton organization, including actin cytoskeleton organization, endocytosis, cell migration, cell-cell adhesion, actin filament organization or lamellipodium assembly were the most abundant within the top 15 hits.Surprisingly, positive regulation of IKK/NF-jB signaling was detected as one of the top 15 hits (Fig. 4C).StringDB interaction mapping of ABI1 proximally interacting proteins showed, as expected, a cluster affecting cytoskeletal organization (Fig. 4D), as well as an unexpected cluster associated with IKK/NF-jB signaling (Fig. 4E).The latter included TAK1 and TAK1-interacting proteins.These analyses, considering the enrichment of known roles of ABI1, provided reasonable support that the unexpected GO BP enrichments were genuine.Because of our previous findings describing a role of ABI1 in regulating NF-jB signaling in hematopoietic stem/ progenitor cells from MPN patients, 24 we sought to further characterize the role of ABI1 in TAK1mediated regulation of NF-jB.

Identification of TAK1 and TAK1 associated proteins within the network of ABI1 proximal interactors
To further explore ABI1 in the context of TAK1 regulation, we performed bioinformatic analysis to characterize the intersection between the measured ABI1 proximal interactome and the established TAK1 interactome curated from the StringDB, MINT, IntAct, and BioGrid interaction evidence databases.Of 227 curated TAK1 interactors (Table S8), ABI1 was found to proximally interact with six proteins: DAB2, TAK1, TAB2, RIPK1, GSK3B, and TRAF3IP2.Using StringDB, we identified 13 secondary TAK1 interactors and 28 tertiary TAK1 interactors that were also identified as significant ABI1 proximal proteins (Fig. 5A,B).This bioinformatic analysis suggested the role of ABI1 as a regulator of TAK1 activity and identified specific proteins that may be involved in ABI1associated regulation of TAK1.

ABI1 directly interacts with proteins involved in TAK1 signaling
Based on bioinformatic analysis of ABI1 proximal interactors that are also direct or indirect interactors of TAK1, we used coimmunoprecipitation to confirm the interaction between ABI1 and selected TAK1associated proteins in unstimulated WT NIH/3T3 cells.In selecting targets for the analysis, we prioritized TAK1-associated proteins by strength of proximity labeling by TurboID-ABI1 and by the fold-enrichment relative to TurboID control cells (Fig. 4B).The results of the coimmunoprecipitation experiments supported a direct interaction between ABI1 and RIPK1, SQSTM1, and ERC1 (Fig. 6A).We confirmed the proximal labeling of ERC1 by TurboID-ABI1 in a small-scale biotinylation experiment followed by western blot of the whole cell and streptavidin-enriched lysates (Fig. S4).We also analyzed for coimmunoprecipitation between ABI1 and TAK1, TNIP1, TAB2, and TBK1.The results supporting these interactions were not obtained.Based on ABI1 coimmunoprecipitations of TAK1-interacting proteins, and because of the prominent roles TAK1 and RIPK1 play in regulating TNFR-mediated balance of cell survival and death, we next focused on characterizing how ABI1 regulates TAK1 and RIPK1 in functional assays.C) Western blot probed with streptavidin-horseradish peroxidase (HRP) to detect biotinylated proteins in WT, Tur-boControl #1, and TurboABI1 #1 cell lines that were either uninduced or induced with 500 lM biotin for 10 min (n = 1).(D) List of proteins measured by global tandem mass tag (TMT) proteomics that are 2-fold changed (adjusted P-value ≤ 0.05) between uninduced TurboABI1 #1 and TurboControl #1 cell lysates (n = 3).(E) Volcano plot showing differentially expressed proteins in TurboControl #1 and TurboABI1 #1 cell lines.The vertical dashed lines, from left to right, correspond to Log 2 Fold-Change (FC) of À1 and 1, respectively.The horizontal dashed line corresponds to adjusted P-value 0.05.The points in color correspond to proteins listed in (A).Blue points are significantly downregulated in TurboABI1 #1 compared to TurboControl #1, and red points are significantly upregulated.(F) Gene Ontology (GO) biological process analysis of differentially expressed proteins measured by global proteomics, listed in (A), of uninduced TurboControl #1 and TurboABI1 #1 cell lines, showing processes identified with false discovery rate (FDR) ≤ 0.05.(G) Oxidoreductase interaction cluster derived from physical StringDB analysis, with interaction score set to 0.9 (highest confidence), of differentially expressed proteins measured from global TMT proteomics of uninduced TurboControl #1 and TurboABI1 #1, presented in (A).

ABI1 dysregulation affects NF-jB activation, RIPK1 phosphorylation status, and caspase cleavage
To measure the effect of ABI1 on TAK1-RIPK1 signaling, we stimulated ABI1 overexpressing (OE) cells, ABI1 deficient (KO) cells, and the respective control MEF cell lines (OEC, KOC) with TNFa for 30 min and 12 h.We then assessed the TAK1-RIPK1-NF-jB pathway activation status by immunoblotting in three independent experiments (Fig. 6B,C and Fig. S5).We observed general, independent of TNFa stimulation length, elevation of TAK1, RIPK1, and NF-jB pathway components including IKKb, IjBa, and NF-jB p65 and their phosphorylated species in ABI1 KO cells compared to ABI1 OE and their control cells (Fig. 6B and Fig. S5).After 30 min of TNFa stimulation, we noted increased levels of RIPK1 in ABI1 KO cells compared to ABI1 KOC, OE, and OEC cells, regardless of stimulation status, although this measurement did not reach statistical significance (Fig. 6B,C).RIPK1-normalized RIPK1 S166 phosphorylation status upon 30 min TNFa stimulation, indicative of RIPK1 autoactivation prior to apoptosis [57], was largely unchanged in ABI1 OE cells compared to ABI1 OEC, but decreased in ABI1 KO cells compared to ABI1 KOC and to ABI1 OE.Without stimulation, ABI1 KO showed 1.6-fold significantly less RIPK1 S166 phosphorylation than ABI1 KOC, and 1.5-fold significantly less than ABI1 OE cells.With stimulation, ABI1 KO showed 1.8-fold significantly less RIPK1 S166 phosphorylation than ABI1 OE cells (Fig. 6B,C and Fig. S5).RIPK1 normalized RIPK1 S321 phosphorylation, indicative of inhibition of apoptosis [58,59], showed a 3.0-fold decrease in ABI1 OE compared to control after 30 min without stimulation and was decreased, but did not reached statistical significance after 30 min of stimulation.RIPK1 normalized RIPK1 S321 phosphorylation levels were not significantly affected in ABI1 KO compared to control after 30 min with or without TNFa stimulation (Fig. 6B,C and Fig. S5).Compared to unstimulated conditions, ABI1 OE showed 7.6-fold significantly increased RIPK1-normalized RIPK1 S321 phosphorylation (ABI1 OEC showed 3.7-fold, P = 0.11), and ABI1 KO showed a 1.7-fold increase, although this was not significant (ABI1 KOC showed a 2-fold increase, P = 0.28; Fig. 6C).ABI1 KO showed 2.7-fold significantly more RIPK1-normalized RIPK S321 phosphorylation than ABI1 OE without stimulation, and 1.7fold less after 30-min stimulation, although this did not meet statistical significance (P = 0.13; Fig. 6C).In congregation, these data indicated decreased levels of proapoptotic phospho-RIPK1 S166 in ABI1 KO cells compared to ABI1 OE and controls, and elevated levels of RIPK1 and phospho-RIPK S321 in ABI1 KO cells, linking loss of ABI1 to decreased susceptibility to TNFa-induced apoptosis.To further assess the role of ABI1 in modulating apoptosis, we examined the status of caspase 3 cleavage upon TNFa exposure.ABI1 OE cells showed elevated caspase 3 cleavage compared to control after 30 min or 12 h stimulation with or without TNFa stimulation, whereas ABI1 KO showed only modestly elevated caspase 3 cleavage compared to control after 12 h with TNFa stimulation, although no statistical significance was observed among ABI1 OE, KO, and control cell lines (Fig. 6B,C and Fig. S5).These data suggest resistance to TNFa-induced apoptosis in ABI1 KO cells, and support the interpretation that elevated activity of IKK/NF-jB survival signaling and increased RIPK1-dependent antiapoptotic signaling is present in ABI1 KO cells.Fig. 3. Quantitative analyses of proximity-dependent labeling with biotin followed by mass spectrometry (PDL/MS) data from wild type (WT), TurboID-expressing (TurboControl #1), and TurboID linked to Abelson interactor 1-expressing (TurboABI1 #1) NIH/3T3 cell lines.(A) Experimental workflow describing cell lines used, experimental conditions, and biological and technical replicates resulting in 27 mass spectrometry (MS) runs.(B) One-tailed TurboABI1 #1 vs. TurboControl #1 enrichment plots generated for peak area (PA; left) and peptide spectral match (PSM; right) measurements.Percentage enrichment, on the x-axis, is calculated as (100*((TurboABI1 #1 peak area (PA) or peptide spectral match (PSM))/(TurboABI1 #1 + TurboControl #1 PA or PSM))), as an alternative to pseudocounts to include proteins not labeled by TurboControl High #1.The dashed lines are drawn at ratio and false discovery rate (FDR) thresholds defining ABI1-proximally interacting hits (TurboABI1 #1 vs. TurboControl #1 ratio ≥ 1.5, FDR ≤ 0.05).Points in red represent proteins considered hits (n = 9).(C) Principal component analyses showing clustering by cell line and experiment of PA (left) and PSM (right) data.Points in red represent TurboABI1 #1, points in blue represent TurboControl #1, and points in green represent WT.Circles represent the first PDL/MS experiment, squares the second, and triangles the third.(D) Number of streptavidin-enriched proteins identified per cell line.Overlaps indicate identification in multiple cell lines.(E) P-value distribution of TurboABI1 #1 vs. TurboControl #1 PA ratio, subset by minimum average number of PSM per MS injection in TurboABI1 group (green: PSM ≥ 0, blue: > 0, red: ≥ 1).(F) Density plot of TurboABI1 #1 vs. TurboControl #1 PA FDR, subset by published evidence of ABI1 interaction curated from MINT, IntAct, StringDB, and Biogrid interaction evidence databases.(G) TurboABI1 #1 vs. TurboControl #1 PA (left) or PSM (right) log-ratio vs. number of known ABI1 interactors.Vertical segments are drawn between data points and the best-fit line was calculated by linear regression.

ABI1 deficiency protects cells from TNFamediated apoptosis, dependent on TAK1 activity
To further characterize the effect of ABI1 deficiency on TNFa-mediated apoptosis, we stimulated ABI1 KO and KOC cells with TNFa in the presence of TNFa pathway inhibitors and measured apoptosis using the apoptotic marker annexin V (Incucyte Annexin V Orange Dye) and an Incucyte live-cell imaging system.We used the TAK1 autophosphorylation inhibitor takinib (at 100 nM [60]), the RIPK1 S166 autophosphorylation inhibitor necrostatin (at 500 nM [61]), and the pancaspase inhibitor Z-VAD (at 10 lM [62]).Without stimulation, ABI1 KO and KOC cells showed similar levels of confluency-normalized apoptosis over a 24 h period (Fig. 6D).After 15 h of TNFa stimulation, ABI1 KO cells showed a statistically significant 3-to 4fold reduction in apoptosis compared to KOC cells (Fig. 6D).In the presence of takinib alone, ABI1 KO cells showed significantly less cell death than KOC cells over 24 h (Fig. 6D).However, in the presence of TNFa and takinib, both KO and KOC cells showed similarly increased levels of apoptosis compared to unstimulated conditions (Fig. 6D), indicating that ABI1 acts downstream of TAK1 upon TNFa stimulation.In the presence of necrostatin, Z-VAD, and necrostatin or Z-VAD with TNFa, ABI1 KO cells showed significantly less apoptosis than KOC cells (Fig. 6D), indicating that loss of ABI1 affects communication between TAK1 and RIPK1 and RIPK1 and caspases.Together, these data suggest that ABI1 deficiency protects cells from TNFa-mediated cell death, and this protection is decreased upon TAK1 inhibition.

Discussion
The use of proximity-dependent labeling with biotin by TurboID combined with stringent analysis of MS data allowed us to identify a previously unrecognized role of ABI1 in regulating TAK1-RIPK1 signaling.We attained high depth proximity labeling and detection by using large numbers of cells in replicate experiments with stable TurboID-ABI1 and TurboID control single-cell-derived cell lines that expressed similar levels of TurboID ligase.Additionally, since MSgenerated PA and PSM measurements are uniquely affected by protein properties [56,63,64], we considered both metrics to identify ABI1 proximal interactors with high confidence.We developed the 'TurboAbi data UI' program to assist interpretation of proximity labeling data, based on both quantitative analysis and cross-referencing to bioinformatic databases.
TAK1 is known to control cell viability and inflammation by activating downstream effectors such as NF-jB, but also through NF-jB-independent pathways including a RIPK1 signaling axis activated in response to TNFa or IL-1 [65].RIPK1 is found both bound to the activated TNFR complex and uncoupled within the cytoplasm [30,59,[66][67][68].In our dataset, we did not identify TNFR complex I-associated proteins such as TRADD, TRAFs, or cIAP1/2 as ABI1 proximal interactors.Instead, we identified TAK1, TAB2, ERC1, PPP6C, and TBK1, all of which are known to proximally interact with TAK1 bound to TNFR complex I-based RIPK1-linked polyubiquitin chains.Based on these observations, we hypothesize that the placement of the ABI1-RIPK1 interaction is not at the Fig. 4. Analyses of significant Abelson interactor 1 (ABI1) proximally interacting proteins identified by proximity-dependent labeling with biotin followed by mass spectrometry (PDL/MS) of wild type NIH/3T3 cells or NIH/3T3 cells expressing TurboID (TurboControl #1) or TurboID linked to ABI1 (TurboABI1 #1) identifies the role of ABI1 in tumor necrosis factor receptor (TNFR)-transforming growth factor b-activated kinase 1 (TAK1)-nuclear factor kappa-light-chain-enhancer of activated B cells (NF-jB) signaling.(A) Data filtering strategy defining hits, or significant ABI1 proximally interacting proteins, and the number of proteins at each filtering step.(B) List of identified hits, sorted by average peptide spectral matches (PSMs) in TurboABI1 #1 group.Cells shaded in blue are associated with cytoskeleton organization, and cells shaded in purple are associated with inhibitor of nuclear factor kappa-B kinase (IKK)/NF-jB regulation indicated by Gene Ontology (GO) biological process analysis of hits.Proteins in red text were selected for further ABI1 interaction validation by coimmunoprecipitation.* = false discovery rate (FDR) ≤ 0.05, ** = FDR ≤ 0.005, *** = FDR ≤ 0.0005 for TurboABI1 #1 vs. TurboControl #1 peak area (PA) or PSM.(C) GO biological process analysis of hits, sorted by number of identified proteins assigned to each process and colored by GO biological process enrichment FDR.Boxed biological processes were selected for further validation by physical and functional experiments.(D) Top: StringDB interaction analysis of hits involved in cytoskeletal regulation, indicated by GO biological process analysis of hits.Thicker edges indicate a higher interaction score (0.4-0.9).Proteins outlined in red were selected for further validation by ABI1 coimmunoprecipitation.TNFR under-membrane assembly site but downstream of TAK1 (Fig. 7).
In addition to regulation of membrane-proximal RIPK1 by ubiquitination and deubiquitination, phosphorylation of both membrane-proximal and cytosolic RIPK1 influences the balance between TNFRmediated cell survival and death.RIPK1 is both autophosphorylated and phosphorylated by other kinases, RIPK1 itself being the only known RIPK1 kinase substrate [57].While around 40 RIPK1 phosphosites have been recorded [69], RIPK1 S166 phosphorylation is accepted as the principal biomarker of RIPK1 cell death-promoting activity [57].RIPK1 S166 autophosphorylation opposes TAK1-mediated survival signaling by initiating a caspase cleavage cascade leading to apoptosis.However, RIPK1 S166 phosphorylation alone is not sufficient to induce cell death [70], indicating the importance of other RIPK1 phosphosites and protein interactions.RIPK S321 phosphorylation is associated with decreased RIPK1 S166 phosphorylation and decreased cell death, preventing complex IIb formation but not affecting NF-jB signaling [58,59].The mechanism of RIPK1 S321 phosphorylation is unclear, as RIPK1 S321 was shown by Geng et al. [58] to be phosphorylated by TAK1 in response to TNFa stimulation, and by Jaco et al. [59] to be dependent on a TNFR/LPS-induced phosphorylation cascade involving TAK1, p38, and MK2.RIPK1 plays a dual role in TNFR signaling, acting as a structural element upholding complex I formation and prosurvival signaling through NF-jB activation, independent of RIPK1 kinase activity, and as a promoter of cell death through autophosphorylation and caspase activation.A better understanding of RIPK1dependent regulation seemingly holds the key to understanding TNFR-regulated balance of survival and death.
Our results indicated decreased RIPK1 S321 phosphorylation in unstimulated ABI1 overexpressing cells, which is maintained upon TNFa stimulation, suggestive of sequestration and blockade of free RIPK1 by ABI1 and an overall negative regulatory effect of ABI1 on RIPK1 S321 phosphorylation and its associated antiapoptotic signal.In ABI1 KO cells, we detected elevated levels of TAK1-NF-jB pathway components in steady state, including RIPK1, and consistently elevated NF-jB pathway activation upon TNFa stimulation, in agreement with a prosurvival signal.Additionally, we observed decreased cell death in ABI1 KO compared to control cells upon TNFa stimulation, which was not observed in response to TAK1 inhibition.In our assay conditions we were not able to detect active TAK1, autophosphorylated on Thr184/187, in response to 30 min or 12 h TNFa stimulation.However, using phosphatase inhibitor calyculin A in combination with TNFa [71,72], we observed increased TAK1 activation in ABI1 KO compared to control cells after a 15 min stimulation (Fig. S6), which may be linked to elevated baseline levels of TAK1 and RIPK1, consistent with more active S321 phosphorylation of RIPK1 by TAK1.This interpretation is consistent with high caspase cleavage observed in stimulated ABI1 OE cells, while the ABI1 absence appeared to provide an apoptosis protection effect by comparison.Furthermore, downstream inhibition of phospho-RIPK1 S166 using necrostatin, or caspase cleavage using Z-VAD, maintained the apoptosis protection observed in ABI1 KO compared to control cells.Together, this suggests that the apoptosis protection observed in ABI1 KO cells is granted by increased total phospho-RIPK1 S321, through a TAK1-dependent mechanism that primes cells for protection against TNFa-induced death by depleting the RIPK1 pool available for S166 autophosphorylation and subsequent apoptosis, enabling prosurvival NF-jB activation.Overall, these data suggest that ABI1 represses antiapoptotic signaling by sequestering RIPK1 to attenuate S321 phosphorylation by TAK1 (proposed model shown in Fig. 7), supporting a role for ABI1 in TAK1-RIPK1-NF-jB-mediated balance of death and survival.
ABI1 loss in hematopoietic stem/progenitor cells was previously shown to be associated with development of a myeloproliferative neoplasm.This Fig. 5. Bioinformatic analysis of significant proximal Abelson interactor 1 (ABI1) interactors shows relevance to transforming growth factor b-activated kinase 1 (TAK1) interaction.(A) List of proximity-dependent labeling with biotin followed by mass spectrometry (PDL/MS) data for statistically significant ABI1 proximal interactors that were also identified as primary, secondary, or tertiary TAK1 interactors by StringDB physical interaction mapping (interaction score [IS] ≥ 0.4).Hits are stratified by TAK1 interaction degree and sorted by average peptide spectral matches (PSM) identified in biotin-stimulated cells expressing TurboID linked to ABI1 (TurboABI1 #1).(B) StringDB physical interaction map (IS ≥ 0.4) of statistically significant ABI1 proximal interactors that are also primary, secondary, or tertiary TAK1 interactors.TAK1 was manually input, and proteins were stratified by degree of TAK1 interaction.Proteins in green are primary TAK1 interactors, proteins in yellow are secondary TAK1 interactors, and proteins in orange are tertiary TAK1 interactors according to StringDB.Thicker edges indicate a higher interaction score.association was linked to elevated NF-jB signaling [24], prompting the question of how ABI1 mechanistically affects NF-jB signaling to promote a malignant cell phenotype.In the present study, using proximitydependent labeling, we identified proximal interactions between ABI1 and components of the TNFR-NF-jB signaling pathway, including TAK1 and RIPK1.Based on our findings, we conclude that increased antiapoptotic RIPK1 phosphorylation, mediated by TAK1, offers a mechanistic link to sustained NF-jB prosurvival signaling and resistance to RIPK1-dependent cell death in ABI1 deficient cells.

Conclusion
Decreased ABI1 expression was found in hematopoietic stem/progenitor cells in patients with myeloproliferative neoplasm (MPN), and murine bone marrowtargeted depletion of ABI1 was shown to be associated with an MPN-like phenotype mechanistically linked to activation of SFKs, STAT3, and NF-jB pathways [24].We used proximity-dependent labeling to detail mechanistic links between ABI1 and SFKs, STAT3, and NF-jB and uncover details of involvement of ABI1 in cancer-linked signaling pathways.We identified proximal interactions between ABI1 and components of the TNFR-NF-jB signaling pathway, and we uncovered that increased antiapoptotic RIPK1 phosphorylation, mediated by TAK1, constitutes a mechanistic link to sustained NF-jB prosurvival signaling and resistance to RIPK1-dependent cell death in ABI1deficient cells.Our proximity-dependent labeling-based strategy enabled mapping of the ABI1 proximal interactome, revealing a previously unknown role of this adaptor protein in TAK1/RIPK1-based regulation of cell death and survival.
Fig.4.Analyses of significant Abelson interactor 1 (ABI1) proximally interacting proteins identified by proximity-dependent labeling with biotin followed by mass spectrometry (PDL/MS) of wild type NIH/3T3 cells or NIH/3T3 cells expressing TurboID (TurboControl #1) or TurboID linked to ABI1 (TurboABI1 #1) identifies the role of ABI1 in tumor necrosis factor receptor (TNFR)-transforming growth factor b-activated kinase 1 (TAK1)-nuclear factor kappa-light-chain-enhancer of activated B cells (NF-jB) signaling.(A) Data filtering strategy defining hits, or significant ABI1 proximally interacting proteins, and the number of proteins at each filtering step.(B) List of identified hits, sorted by average peptide spectral matches (PSMs) in TurboABI1 #1 group.Cells shaded in blue are associated with cytoskeleton organization, and cells shaded in purple are associated with inhibitor of nuclear factor kappa-B kinase (IKK)/NF-jB regulation indicated by Gene Ontology (GO) biological process analysis of hits.Proteins in red text were selected for further ABI1 interaction validation by coimmunoprecipitation.* = false discovery rate (FDR) ≤ 0.05, ** = FDR ≤ 0.005, *** = FDR ≤ 0.0005 for TurboABI1 #1 vs. TurboControl #1 peak area (PA) or PSM.(C) GO biological process analysis of hits, sorted by number of identified proteins assigned to each process and colored by GO biological process enrichment FDR.Boxed biological processes were selected for further validation by physical and functional experiments.(D) Top: StringDB interaction analysis of hits involved in cytoskeletal regulation, indicated by GO biological process analysis of hits.Thicker edges indicate a higher interaction score (0.4-0.9).Proteins outlined in red were selected for further validation by ABI1 coimmunoprecipitation.Bottom: PA and PSM measurements for indicated proteins from TurboABI1 #1, TurboControl #1, and wild type NIH/3T3 PDL/MS experiments (n = 9).Error bars represent standard deviation.* = FDR ≤ 0.05, ** = FDR ≤ 0.005, *** = FDR ≤ 0.0005.(E) Top: StringDB interaction analysis of hits associated with IKK/NF-jB regulation, indicated by GO biological process analysis of hits.Thicker edges indicate a higher interaction score (0.4-0.9).Proteins outlined in red were selected for further validation by ABI1 coimmunoprecipitation.Bottom: PA and PSM measurements for indicated proteins from TurboABI1 #1, TurboControl #1, and wild type NIH/3T3 PDL/MS experiments (n = 9).Error bars represent standard deviation.* = FDR ≤ 0.05, ** = FDR ≤ 0.005, *** = FDR ≤ 0.0005.