Characterization of huntingtin interactomes and their dynamic responses in living cells by proximity proteomics

Abstract

Huntingtin (Htt) is a large protein without clearly defined molecular functions. Mutation in this protein causes Huntington’s disease (HD), a fatal inherited neurodegenerative disorder. Identification of Htt-interacting proteins by the traditional approaches including yeast two-hybrid systems and affinity purifications has greatly facilitated the understanding of Htt function. However, these methods eliminated the intracellular spatial information of Htt interactome during sample preparations. Moreover, the temporal changes of Htt interactome in response to acute cellular stresses cannot be easily resolved with these approaches. Ascorbate peroxidase (APEX2)-based proximity labeling has been used to spatiotemporally investigate protein-protein interactions in living cells. In this study, we generated stable human SH-SY5Y cell lines expressing full length Htt23Q and Htt145Q with N-terminus tagged Flag-APEX2 to quantitatively map the spatiotemporal changes of Htt interactome to a mild acute proteotoxic stress. Our data revealed that normal and mutant Htt are associated with distinct intracellular microenvironments. Specifically, mutant Htt is preferentially associated with intermediate filaments and myosin complexes. Furthermore, the dynamic changes of Htt interactomes in response to stress are different between normal and mutant Htt. Vimentin is identified as one of the most significant proteins that preferentially interacts with muHtt in situ. Further functional studies demonstrated that mutant Htt affects the vimentin’s function of regulating proteostasis in healthy and HD human neural stem cells. Taken together, our data offer important insights into the molecular functions of normal and mutant Htt by providing a list of Htt-interacting proteins in their natural cellular context for further studies in different HD models.

Introduction

Protein-protein interaction (PPI) networks, or protein interactomes, are inherently dynamic in nature, adjusting in response to different stimuli and environmental conditions. This allows cells to adapt in a measured way to changing circumstances (Snider et al. 2015). Even a subtle change of protein interactomes can have major systemic consequences, leading to disease phenotypes (Gonzalez & Kann 2012; Ryan & Matthews 2005). Particularly, neurons have a low stress threshold, making them especially vulnerable to stresses as demonstrated in various neurodegenerative diseases (Saxena & Caroni 2011). Delineating in-depth, dynamic PPI networks in response to stresses is therefore critically important in understanding the molecular basis of neurodegenerative diseases and identifying new targets suitable for therapeutic interventions.

Huntington’s disease (HD) is an inherited fatal neurodegenerative disorder caused by an expanded polyglutamine repeat in the N-terminus of the huntingtin (Htt) protein. Accumulating evidence suggests that Htt is a scaffold protein involved in diverse cellular functions through interacting with many effector proteins. It is conceivable that mutant Htt (muHtt) can alter these interactions, resulting in pathological consequences. Furthermore, cellular stresses play important roles in HD pathogenesis (Atwal & Truant 2008; Jiang et al. 2016; Jin et al. 2012; Kumar & Ratan 2016; Sanchez et al. 2021). We (Erie et al. 2015; Huang et al. 2018) and others (Nath et al. 2015; Munsie et al. 2011) have shown that responses to cellular stresses are defective in HD, causing the accumulation of cellular damages overtime and eventually neurodegeneration. It is thus highly possible that normal and mutant Htt react differently to cellular stresses by altering their interactions with proteins, leading to increased vulnerabilities of HD cells to stresses.

Traditional approaches such as yeast two-hybrid systems and affinity purifications have greatly facilitated the identification of Htt-interacting proteins (Tourette et al. 2014; Goehler et al. 2004; Ratovitski et al. 2012; Shirasaki et al. 2012). However, due to their methodological limitations, they lack intracellular spatial information and often miss those interactions that are either weak or dynamic in nature. Proximity labeling-based methods including biotin identification (BioID) and engineered ascorbate peroxidase (APEX) provide an unbiased screening of protein proximity for a protein of interest in living cells (Han et al. 2018; Kim & Roux 2016; Roux 2013). APEX2 is an improved version of APEX with increased catalytic efficiency and can thus be expressed at lower levels (Lam et al. 2015). APEX2 has been successfully used in various studies to spatiotemporally identify protein networks in living cells, such as ligand-stimulated remodeling of G-protein coupled receptor (GPCR) networks (Lobingier et al. 2017; Paek et al. 2017), mapping proteomics associated with α-synuclein (Chung et al. 2017), stress granules (Markmiller et al. 2018) and synaptic clefts (Loh et al. 2016).

In this study, we generated cell lines stably expressing Flag-APEX2-Htt to label Htt-interacting proteins spatiotemporally in living cells at basal and under proteotoxic stress. In combination with unbiased quantitative proteomics, we identified an initial pool of 636 proteins. After applying stringent filtering steps and using a APEX2-labeled cytosolic probe as the control, 94 high confident proteins were remained for further quantitative analysis. Our data indicated that normal and mutant Htt exhibit distinct subcellular spatial distributions. Specifically, mutant Htt is more associated with intermediate filament and myosin complexes. Moreover, normal and muHtt interactomes respond differently to a mild acute proteotoxic stress. We further experimentally validated vimentin as preferential interactors for muHtt in situ and performed functional studies to determine the effect of muHtt on vimentin function in regulating proteostasis under proteotoxic stress. Taken together, our data should offer important reference points for gaining further insights into the spatial distribution and molecular functions of Htt and muHtt.

Methods

Plasmids and cloning

pcDNA3 Flag-APEX2-NES (plasmid #49386, Addgene) (Lam et al. 2015) was directly used in this study. Flag-APEX2-full length Htt23Q (Flag-APEX2-Htt23Q) and 145Q (Flag-APEX2-Htt145Q) were subcloned using the Polymerase Incomplete Primer Extension (PIPE) method (Klock et al. 2008; Klock & Lesley 2009). Specifically, full length Htt23Q and Htt145Q were PCR-amplified from pcDNA3.1-Htt23Q/Htt145Q with V-PIPE primers (Forward primer: 5’- ggactcagatctcgaATGGCGACCCTGGAAAAGCTGATGAAG-3’; Reverse primer, 5’- tccggtggatccctaGCAGGTGGTGACCTTGTGGACATTTCG-3’). The insert, which encodes Flag-APEX2, was PCR-amplified from APEX2-actin in pEGFP (plasmid # 66172, Addgene) (Lam et al. 2015) using the I-PIPE primers (Forward primer: 5’- tcgagatctgagtccGGAGCCCGAGCCCGAGGTCGAGCCC-3’; Reverse primer, 5’- tagggatccaccggaTCTAGATAACTGATCATAATCAGCC-3’). V-PIPE and I-PIPE PCR products were mixed to anneal the amplified DNA fragments together and the mixture was transformed into the XL10-Gold ultracompetent cells (Cat. No. 200314, Stratagene) for colony selection on Kanamycin-resistant LB-agar plates. The resulting plasmids were purified, and the DNA sequences were confirmed by Sanger sequencing (Genewiz).

Cell culture

Cell lines:

The human neuroblastoma SH-SY5Y cell line (Cat. No. 94030304, RRID:CVCL_0019, Sigma-Aldrich) was cultured in a humidified incubator with 5% CO₂ at 37°C in Dulbecco’s Modified Eagle Medium/F-12K medium (DMEM/F12K) containing 10% fetal bovine serum (FBS) and 1% antibiotic-antimycotic solution. STHdhQ7 (CH00097, RRID:CVCL_M590) and STHdhQ111 (CH00095, RRID:CVCL_M591, Coriell Institute for Medical Research) cells were cultured in DMEM supplemented with 10% FBS, 1% glutamine and 1% antibiotic-antimycotic solution at 33°C in a humidified incubator with 5% CO₂. These two striatal cell lines were originally derived from wild type HdhQ7 and HD HdhQ111 knock-in mice, respectively (Trettel et al. 2000). All these cell lines were first expanded in the lab, stored in liquid nitrogen designated as passage 1 and used within 20 passages.

Generating stable cell lines expressing Flag-APEX2-NES, -Htt23Q and -145Q:

SH-SY5Y cells were transfected with Flag-APEX2-NES, -Htt23Q, and -Htt145Q plasmids by electroporation using the Neon electroporation system (pulse voltage = 1100V, pulse width = 50ms, pulse number = 1). Three days later, cells were selected with 500 μg/ml G418 until cell colonies were visible. Individual colonies were re-seeded into 24-well plates. After reaching confluency, cells were screened for APEX2 activity by the Amplex^™ Red assay (Cat. No. A12222, Thermofisher) as described (Martell et al. 2017). Positive clones were selected, serial-diluted into single cells in a 96-well plate and expanded for further characterization by Western blot and biotinylation assay as described below. Stable cells were expanded and maintained in a humidified incubator with 5% CO₂ at 37°C in DMEM/F12K medium supplemented with 10% FBS and 1% antibiotic-antimycotic solution. All stable cell lines were used within 10 passages after generation.

iPSCs maintenance and NSCs differentiation:

Age and sex-matched iPSCs from an apparently healthy individual (GM23476, female, 20 years old at sampling, RRID:CVCL_T841) and a HD patient (GM23225, female, 20 years old at sampling, RRID:CVCL_F169, CAG repeat length 71) were obtained from MIGMS cell repository through Coriell Institute for Medical Research. iPSCs were cultured in Matrigel-coated plates with mTeSR^™1 complete medium (STEMCELL Technologies). The differentiation of iPSCs to NSCs was performed using the STEMdiff^™ SMADi Neural Induction Kit based on the rosette formation and isolation method according to the manufacture’s instruction (STEMCELL Technologies). After differentiation, NSCs were further cultured and expanded in the complete STEMdiff^™ Neural Progenitor Basal Medium (STEMCELL technologies). NSC expansions were limited to 4 passages. The use of human iPSC lines was approved by Institutional Biosafety Committee of Florida Atlantic University (Approval number B20-21). All iPSCs utilized in this study were used within 10 passages from cryopreserved stocks previously determined to be karyotypically normal. None of the cell lines used in this study (SH-SY5Y, STHdhQ7 and Q111, iPSCs) are listed as commonly misidentified cell lines by the International Cell Line Authentication Committee (ICLAC). No further authentication was performed in the laboratory.

Chymotrypsin-like proteasomal activity assay

Cells were treated with 5 μM MG132 (Cat. No. AGCP30011M005, AdipoGen) for 30 min. Equal volume of dimethyl sulfoxide (DMSO, 0.1%) was added to the control group. For the stress recovery groups, cells were first incubated with 5 μM MG132 for 30 min, washed with normal complete medium twice and further incubated with complete medium for 4 hours. Chymotrypsin-like proteasomal activity was measured using the cell-based Proteasome-Glo^™ assay kit (Cat. No. G8621, Promega) with the luminogenic specific substrate, Suc-LLVY-aminoluciferin, according to the manufacturer’s instruction. The luminescent signals were read with a plate reader (CLARIOstar, BMG LABTECH).

APEX2-mediated proximity labeling in living cells and enrichment of biotinylated proteins

APEX2-mediated labeling was performed as described (Hung et al. 2016). Briefly, cells were first exposed to 500 μM biotin-phenol (Cat No. A8011, APExBIO) in complete culture medium for 1 hour and then challenged with 1 mM H₂O₂ for exactly 1 min. After the challenge, cells were quickly washed three times in the stop solution [10 mM NaN₃, 10 mM sodium ascorbate, 5 mM Trolox in phosphate-buffered saline (PBS)]. Cells were then directly lysed in 1x SDS sample buffer for Western blot. To enrich biotinylated proteins, cells were harvested in radioimmunoprecipitation assay (RIPA) lysis buffer containing 50 mM Tris, pH=7.4, 150 mM NaCl, 0.1% (wt/vol) SDS, 0.5% (wt/vol) sodium deoxycholate and 1% (vol/vol) Triton X-100, 1x protease inhibitor cocktail, 1 mM phenylmethylsulfonyl fluoride (PMSF), 10 mM NaN₃, 10 mM sodium ascorbate and 5 mM Trolox. Protein concentration was measured by Qubit^™ protein assay (Cat. No. Q33211, Thermofisher). The cleared supernatant (~360 μg) was incubated with 30μl Pierce^™ magnetic streptavidin beads (Cat. No. 88817, Thermofisher) overnight at 4°C. After incubation, the beads were washed sequentially with the following buffers to remove nonspecific proteins: twice with RIPA buffer, once with 1 M KCl, once with 0.1 M Na₂CO₃, once with 2 M urea in 10 mM Tris-HCl, pH 8.0, and twice with RIPA buffer. Finally, biotinylated proteins were eluted from the beads by boiling each sample in 30 μl of 3x protein loading buffer supplemented with 2 mM biotin and 20 mM dithiolthreitol (DTT) for 10 min and directly used for Western blot.

On-bead trypsin digestion for quantitative mass spectrometry (MS) analysis

About 4 mg biotinylated cell lysates in different groups were incubated with 300 μl Pierce^™ magnetic streptavidin beads overnight at 4°C. After incubation, the beads were serial-washed as described above and resuspended in 2 M urea/50 mM Tris (pH=7.5) containing 0.4 μg MS-grade trypsin (Cat. No.90057, Thermofisher). The beads suspension was digested at room temperature for 1 hour. At the end of digestion, the supernatant was transferred to a low-binding high-profile microcentrifuge tube. The remaining beads were further washed twice with 2 M urea/50 mM Tris (pH=7.5). The washes were combined with the original supernatant for reduction with 4 mM DTT at room temperature for 30 min followed by alkylation with 10 mM iodoacetamide (Cat. No. I1149, SigmaAldrich) for 45 min at room temperature in the dark. At the end of incubation, 0.5 μg trypsin was added for further digestion overnight at room temperature. Following digestion, samples were ZipTip concentrated according to the manufacture’s instruction (Thermofisher, Pierce^™ C18 Tips, 100 μl bed, Cat. No. 87784) and vacuum-dried at room temperature. Quadruplet samples from each group were submitted to Scripps Florida Proteomic Core for further Tandem Mass Tag labeling and MS analysis.

Tandem mass tag (TMT) quantitative proteomics.

Dried cleaned-up peptides were resolubilized in 25 μl of 100 mM triethyl ammonium bicarbonate pH 8.5, labelled with TMT labels (10-plex) according to the manufacturer’s instructions (Cat. No. 90111, Thermo Fisher Scientific) and pooled. TMT labeling was randomized and the strategy is shown in the Supplementary table S1 of Additional File 1. The pooled plexed samples were then dried under vacuum, resolubilized in 1% trifluoroacetic acid and finally desalted using 2μg capacity ZipTips (Millipore, Billeric, MA) according to the manufacturer’s instructions.

Peptides were on-line eluted into a Fusion Tribrid mass spectrometer (Thermo Scientific) from an EASY-Spray PepMap^™ RSLC C18 column (Cat No. ES803, 2μm, 100Å, 75 μm x 50cm, Thermo Scientific), using a gradient of 5-25% solvent B (80/20 acetonitrile/water, 0.1% formic acid) in 180 min, followed by 25-44% solvent B in 60 min, 44-80% solvent B in 0.1 min, a 5 min hold of 80% solvent B, a return to 5% solvent B in 0.1 min, and finally a 20 min hold of solvent B. All flow rates were 300 nL/min delivered using a nEasy-LC1000 nano liquid chromatography system (Thermo Scientific). Solvent A consisted of water and 0.1% formic acid. Ions were created at 1.7kV using an EASY Spray source (Thermo Scientific) held at 50ºC. A synchronous precursor selection (SPS)-MS³ mass spectrometry method was used based on the work of Ting et al to overcome reporter ion ratio distortion phenomena resulting from co-isolation and co-fragmentation of interfering ions (Ting et al. 2011), scanning between 380-2000 m/z at a resolution of 120,000 for MS1 in the Orbitrap mass analyzer at an AGC target of 4E5 and a maximum injection of 50 msec, and performing CID in the linear ion trap of peptide monoisotopic ions with charge 2-8 above an intensity threshold of 5E3, using a quadrupole isolation of 0.7 m/z and a CID energy of 35%. The ion trap AGC target was set to 1.0E4 with a maximum injection time of 50 msec. Dynamic exclusion duration was set at 60 sec and ions were excluded after 1 time within +/− 10 ppm mass tolerance window. The top 10 MS² ions in the ion trap between 400-1200 m/z were then chosen for HCD at 65% energy and detection occurred in the Orbitrap at a resolution of 60,000, an AGC target of 1E5 and an injection time of 120 msec (MS³). All scan events occurred within a 3 sec specified cycle time.

Proteomic data processing and statistical analysis

Quantitative analysis of the TMT experiments was performed simultaneously with protein identification using Proteome Discoverer 2.5 software. The precursor and fragment ion mass tolerances were set to 10 ppm and 0.6 Da, respectively. Trypsin enzyme was used with a maximum of 2 missed cleavages. The databases used were: Uniprot human proteome, common contaminant.fasta, and Streptavidin only database FASTA files. SEQUEST search was performed and Percolator feature of Proteome Discoverer 2.5 was used to set a false discovery rate (FDR) of 0.01. The impurity correction factors obtained from Thermo Fisher Scientific for each kit were included in the search and quantification. The following settings were used to search the data; dynamic modifications of Oxidation / +15.995Da (M), Deamidated / +0.984 Da (N, Q), and static modifications of TMT6plex / +229.163 Da (N-Terminus, K), Carbamidomethyl / +57.021 Da (C). Only unique+ Razor peptides were considered for quantification purposes. Co-isolation threshold and SPS Mass Matches threshold were set to 50 and 65, respectively. The abundance values for master proteins were imported into JMP PRO 15.2.0 (SAS). Western blot values of APEX2 protein were used to normalize the raw protein abundance. Protein Abundance Based method was used to calculate the protein level ratios. All contaminant proteins, and proteins that were not found in at least 3 of the 4 plexes were excluded. The resulting proteins were analyzed first with One way ANOVA per treatment (Control, Stress and Stress recovery) after ln(abundance+1) transformation, Benjamini -Hochberg method was used to control FDR at 0.05 , and finally Tukey’s Honestly Significant Difference test was used to perform pairwise comparisons.

Immunostaining

Stable cell lines were seeded on poly-L-lysine coated coverslips (No. 1.5). After biotinylation, cells were fixed in 4% paraformaldehyde (PFA) in PBS for 15 min at 37°C. NSCs were seeded on poly-L-ornithine/laminin coated coverslips and fixed in 4% PFA for 15 min at 37°C. Cells were then permeabilized with 0.25% Triton X-100 in PBS for 10 min. Cells were then blocked and immunostained as we previously described (Huang et al. 2018). Immunofluorescent signals were detected using a laser scanning confocal microscope (Nikon A1R). DAPI (4′,6-diamidino-2-phenylindole) was used as a nuclear counterstain. Images were taken with a 60X oil objective (CFI Plan Apochromat Lambda 60X Oil, numerical aperture =1.4). The following antibodies were used: rabbit monoclonal anti-huntingtin (1:500, D7F7, Cat. No. 5656S, RRID:AB_10827977, Cell Signaling), mouse monoclonal anti-Flag M2 (1:500, Cat. No. F1804, RRID:AB_262044, SigmaAldrich), mouse monoclonal vimentin (1:1000, Cat. No. 60330-1-lg, RRID:AB_2881439, Proteintech), Rabbit monoclonal anti-K48-linked ubiquitin (clone Apu2, 1:1000, Cat. No. 05-1307, RRID:AB_1587578, Millipore). To label biotinylated proteins, Alex Fluor^™ 488-conjugated streptavidin (1:5000, Cat. No. S11223, Invitrogen) was used.

Co-immunoprecipitation

Co-immunoprecipitation was performed as we previously described with modifications (Rush et al. 2012). Briefly, cells were lysed in co-immunoprecipitation (co-IP) lysis buffer (20 mM Tris pH 7.5, 100 mM NaCl, 0.5% NP-40, 0.5 mM EDTA) for 15 min at 4°C. Total lysates were cleared by centrifugation at 500 x g for 10 min at 4°C to remove undissolved debris and retain vimentin filaments in the supernatant. After determining protein concentration, ~500 μg of cleared cell lysates were incubated with Flag M2 (2 μg), Htt (D7F7, 1:50), rabbit polyclonal vimentin (1.6 μg) or rabbit control IgG (1.6 μg) for 2 hours at 4°C with end-to-end mixing. At the end of the incubation, 20 μl of PureProteome^™ protein G magnetic beads (Cat. No. LSKMAGG10, Millipore) were added and further incubated with end-to-end mixing overnight at 4°C.The immobilized protein G-bound complexes were washed three times with cell lysis buffer the next day and eluted with 2x SDS sample buffer for western blot analysis.

Western blot

About 20 μg of sample lysates were separated on a 4-20% Mini-PROTEAN TGX (Tris-Glycine eXtended) precast protein gel (Bio-Rad) and transferred to nitrocellulose membranes. The membrane was first blocked with blocking buffer for fluorescent western blotting (Cat. No. MB-070, Rockland Inc.) for 2 hours at room temperature, then incubated with primary antibodies diluted in Tris-buffered saline with 0.1% Tween 20 (TBS-T) overnight at 4°C. The next day, the membrane was washed in TBS-T 3 x 15 min and then incubated with goat anti-rabbit Alexa Fluor Plus 800 (Cat. No. A32735, 1:8000, Invitrogen) and goat anti-mouse Alexa Fluor 680 (Cat. No. A21058, 1:5000, Invitrogen). For 1 hour at room temperature followed by extensive washes in TBS-T. Fluorescent signals were detected with a LI-COR Odyssey Fc system and the images were quantified with the provided Image Studio software. The following primary antibodies were used: rabbit monoclonal anti-huntingtin (1:500, D7F7), mouse monoclonal anti-huntingtin (1:500, mEM48, Cat. No. MAB5374, RRID:AB_10055116, Millipore), mouse monoclonal ubiquitin (clone P4D1, 1:1000, Cat. No. 14-6078-82, RRID:AB_837154, eBiosciences), mouse monoclonal anti-Flag M2 (1:500), rabbit polyclonal vimentin (1:500, 10366-1-AP, RRID:AB_2273020, Proteintech), mouse monoclonal vimentin (1:500, Cat. No. 60330-1-lg, Proteintech), rabbit polyclonal LaminA/C (1:500, Cat. No. 10298-1-AP, RRID:AB_2296961, Proteintech), rabbit polyclonal MYH9 (1:500, Cat. No. 11128-1-AP, RRID:AB_2147294, Proteintech), rabbit polyclonal MYH10 (1:500, Cat. No. 19673-1-AP, RRID:AB_10858231, Proteintech), rabbit polyclonal pyruvate carboxylase (1:500, Cat. No. 16588-1-AP, RRID:AB_1851513, Proteintech), mouse monoclonal anti-actin (1:1000, Cat. No. sc-47778, RRID:AB_626632, Santa Cruz Biotechnology), and rabbit polyclonal GAPDH (1:1000, Cat. No. 10494-1-AP, RRID:AB_2263076, Proteintech).

Proximity ligation assay (PLA)

In situ interaction of Htt with vimentin was detected using the Duolink^® In Situ Detection Reagent Red (Cat. No. DUO92008, Sigma-Aldrich) with Duolink^® In Situ PLA^® Probe Anti-Rabbit Plus (Cat. No. DUO82002, RRID:AB_2810940, Sigma-Aldrich) and Anti-Mouse Minus (Cat. No. DUO82004, RRID:AB_2713942, Sigma-Aldrich) as we previously described (Davis et al. 2018). Rabbit anti-Htt (1:500), and mouse anti-vimentin antibodies (1:500) were used. After ligation, cells were mounted using ProLong^™ Diamond Antifade Mountant (Cat. No. P36961, Invitrogen) with DAPI and observed with an A1R Nikon confocal microscope at 20x objective (Nikon Plan Apo 20x/0.75). PLA signals were recognized as red fluorescent puncti. PLA-positive puncti and cell numbers as indicated by DAPI staining in one captured image (1024 x 1024 pixels) were counted using Fiji/Image J (RRID:SCR_0022285). PLA-positive puncti per cell was calculated as the division of the total number of PLA-positive puncti per field by the total number of cells per field.

Data analysis

No sample size calculation was performed in the study. However, sample sizes, such as the number of samples for image or western blot quantification, were similar to, or exceeded those reported in our recent publications (Xu et al. 2021; Bensalel et al. 2021) and others (Ratovitski et al. 2012; Shirasaki et al. 2012). No blinding except for proteomic analysis was performed in the study. All data were expressed as means ± S.E.M. To establish significance, data were subjected to unpaired student’s t-tests, one-way or two-way ANOVA followed by the Tukey’s multiple comparison test using the GraphPad Prism software statistical package 9.0 (RRID:SCR_002798, GraphPad Software) as specified. Normality of the data was not assessed due to the limited sample size. No exclusion criteria were pre-determined. The ROUT test revealed no outliers. The criterion for significance was set at p ≤ 0.05.

Results

Characterization of the stable SH-SY5Y cell lines expressing Flag-APEX2-Htt for labeling Htt-interacting proteins in living cells.

We first generated stable clones expressing Flag-APEX2-full length Htt23Q (hereafter referred to as Htt23Q), Flag-APEX2-full length Htt145Q (hereafter referred to as Htt145Q) and Flag-APEX2-NES (hereafter referred to as NES, nuclear export signal) in the human neuroblastoma SH-SY5Y cell line. NES is used here as a cytosolic control for nonspecific proteins (Lee et al. 2016). Structural studies revealed that the polyglutamine tract resides at the external surface (Vijayvargia et al. 2016) and the flexible N-terminus has limited influence on the overall architecture of Htt complexes (Guo et al. 2018). We therefore attached Flag-APEX2 to the N-terminus of Htt (Fig. 1A). APEX2 is a 27-KDa monomeric protein which is the same size as that of green fluorescent protein (GFP). N-terminal GFP-tagged huntingtin has been used in various studies for localization and functional studies (Xia et al. 2003; Jung et al. 2020; Wang et al. 2009). A flexible glycine-rich 16 amino acid spacer is present between APEX2 and Htt to minimize the interference of tags with the structure and function of Htt (Fig. 1A). We screened about 27 clones for each plasmid using Amplex Red assay and in vivo biotinylation assay. We chose the single clone that had strong signals in both assays for the rest of experiments in this study (Additional File 1: Supplementary Fig. S1). Western blot analysis confirmed the overexpression of Htt protein in Htt23Q and Htt145Q cell lines (Fig. 1B, top blot) and the expression of Flag epitope in all stable cell lines (Fig. 1B, bottom blot). The expression level of Flag-APEX2-NES is higher than those of the much larger constructs of Flag-APEX2-Htt, probably due to the higher plasmid number integration in NES samples. The purity of the clone was checked by co-immunostaining with antibodies against Htt and Flag (Additional File 1: Supplementary Fig. S2). We next performed biotinylation assay to investigate whether APEX2-Htt could biotinylate its nearest proteins in proximity in living cells. As shown in the middle panel of Fig. 1B, biotinylated proteins were clearly increased in all stable cell lines after H₂O₂ exposure. The two heavy bands (~130KDa and ~75KDa ) in the absence of H₂O₂ could indicate endogenously biotinylated proteins as reported (Hung et al. 2016). We also performed imaging studies to examine protein biotinylation by APEX2-Htt in situ. Consistent with the Western blot analysis, biotinylated proteins were significantly increased after H₂O₂ exposure as detected by fluorescently labeled streptavidin (Fig. 1C).

Fig. 1. Characterization of Flag-APEX2-Htt stable SH-SY5Y cell lines.

A. A linearized diagram illustrating the construct of Flag-APEX2-Htt. The sequence of the molecular spacer between APEX2 and Htt is shown. B. Representative images of Western blot analysis of Htt expression and APEX2-Htt-mediated biotinylation of endogenous proteins in NES, Htt23Q and Htt145Q stable cell lines. Cells were pre-incubated with biotin-phenol for 1 hour and then challenged in the presence or absence of H₂O₂ for 1 min. Cell lysates were analyzed by Western blot for Htt expression using two Htt antibodies against different epitopes. Biotinylated proteins were visualized by streptavidin. Flag tag expression was detected by M2 antibody. Without stripping, the same membrane was re-probed with Actin antibodies as a loading control. C. Representative confocal images showing in situ visualization of biotinylated proteins by immunostaining in NES, Htt23Q and Htt145Q stable cell lines. Cells were pre-incubated with biotin-phenol for 1 hour and then challenged in the presence or absence of H₂O₂ for 1 min. After biotinylation, cells were fixed with 4% PFA. Biotinylated proteins were labeled with Alexa Fluor 488-conjugated streptavidin. DAPI was used as a nuclear counterstain.

We previously showed that HD cells are more vulnerable to proteotoxic and ER stresses (Xu et al. 2021; Huang et al. 2018), implying that the composition of the PPI networks between normal and mutant Htt may be distinctive in response to cellular stresses. In current study, we aimed to address how normal and mutant Htt interactomes change in response to an acute proteotoxic stress induced by MG132, a potent reversible proteasome inhibitor. Using changes in chymotrypsin-like proteasomal activities (Huang et al. 2018) as a readout for MG132 inhibition, we showed that a 30-min treatment with 5 μM MG132 significantly blocked over 90% of the chymotrypsin-like proteasomal activities, which was half-recovered 4 hours after removing MG132 (Fig. 2A). Consistently, polyubiquitinated proteins were increased in MG132-treated samples and returned to baseline in the recovery group (Fig. 2B). This condition was therefore adopted in the rest of the study to induce the acute mild proteotoxic stress. Taken together, we established stable cell lines expressing APEX2-Htt and a control cell line expressing APEX2-NES for quantitatively analyzing the early dynamic changes of Htt interacting proteins in response to mild proteotoxic stress.

Fig. 2. Proteotoxic stress induction in APEX2-Htt cells.

A. Measurements of chymotrypsin-like proteasomal activities in NES, Htt23Q and Htt145Q cells under basal, stress and stress recovery conditions. Data are representative of four independently treated cultures. ****p < 0.0001. Two-way ANOVA with Tukey’s post-hoc test. The full statistical reports of the analysis can be found in Additional File 3. B. Representative images of Western blot analysis of protein ubiquitination in NES, Htt23Q and Htt145Q cells under basal, stress and stress recovery conditions. GAPDH was used as the loading control. Semi-quantification of polyubiquitinated proteins by densitometry from three independently treated cultures shown at the bottom. **p < 0.01, ***p < 0.001, ****p < 0.0001, One-way ANOVA with Šidák’s post-hoc test for selected pair comparisons. The full statistical reports of the analysis can be found in Additional File 3.

Identification of time-resolved normal and mutant Htt interactomes in response to proteotoxic stress

We next proceeded with biotinylation experiments to label Htt-interacting proteins under control, proteotoxic stress and post-stress recovery conditions in NES, Htt23Q and Htt145Q cells. The experimental timeline corresponding to sample treatments and labeling is summarized in Fig. 3A. NES was used as a spatial cytosolic probe in this study since Htt is mainly present as a soluble cytosolic protein. After labeling, we designed a multi-step proteomic workflow to quantitatively identify biotinylated proteins in control, MG132-treated (stress) and post-stress recovery (stress recovery) samples (Fig. 3B). Four independent replicates were included in each experimental group. Therefore, the final samples contained four independent 10-plex tandem mass tag (TMT) batches (F1-F4, referred to as four plexes).

Fig. 3. Experimental flow for quantitative proteomics and data quantification.

A. The Experimental design for biotinylation labeling in living cells under basal, stress and stress recovery conditions. B-P, biotin-phenol. B. The experimental workflow for quantitative proteomic analysis. Four independent cultures per group were included. C. A diagram illustrating the data filtering process. D. Venn diagram comparison of Htt-interacting proteins identified in this study with two published studies from Culver et al. and Shirasaki et al. E. Normalized protein abundance in each individual experimental group.

We identified a total of 636 proteins from the four plexes and quantified their abundance (Additional File 2). The correlation matrix analysis showed high correlation efficiency among replicates within each cell type. Moreover, Htt23Q and NES samples showed closer relationships than Htt145Q samples (Additional File 1: Supplementary Fig. S3). We then applied a filter to exclude 51 contaminants including different keratin isoforms and non-Homo sapiens proteins that possibly came from the culture medium. This resulted in 585 proteins in the list (Fig. 3C, Additional File 2). Out of the 585 proteins, 123 and 132 proteins are present in the high-confidence list of Htt-interacting proteins identified from postnatal day 15 wildtype and Q140 mouse brains (Culver et al. 2012) and wildtype and BACHD mouse brains (Shirasaki et al. 2012), respectively (Fig. 3D). Proteins that were not found in three or four sets of the four plexes were further excluded in the present study due to the inability to apply statistical calculations. After applying these filtering steps, our final proteomic list consisted of 94 proteins (Fig. 3C and Additional File 2), in which 39 and 23 proteins were present in the studies from Culver, et al. (Culver et al. 2012) and Shirasaki, et al. (Shirasaki et al. 2012) respectively.

We noticed that raw protein abundances in NES samples were generally higher than those in the Htt samples. This is likely due to the higher APEX2 expression levels in NES samples as reflected by the Flag immunostaining (Fig. 1B). To balance this expressional difference, we normalized the raw abundance with Flag expression values quantified from three independent Western blots (Fig. 1B, NES/Htt23Q: 3.449; NES/Htt145Q: 4.485). The normalized protein abundance was similar among all groups except the groups with post-stress recovery which is unexpected and not readily explainable (Fig. 3E, Additional File 2). We therefore focused our analysis on the control and stress group in the present study.

Htt23Q and Htt145Q are associated with distinct microenvironments in the cytosol

By pairwise comparing the protein abundance of the 94 proteins across different experimental groups against its significance (Additional File 2), our goals in this study were to (1) map the spatial distribution of Htt23Q and Htt145Q in the cytosol; (2) identify differential interacting proteins for Htt23Q and Htt145Q under normal and stress conditions; and (3) characterize dynamic changes in the Htt23Q and Htt145Q interactomes in response to the stress.

As a soluble cytosolic protein, APEX2-NES would biotinylate cytosolic proteins in a random manner. In agreement with this notion, we observed a less variation in the protein abundance distribution for all detected proteins in the NES group compared to the Htt groups (Additional File 1: Supplementary Fig. S4). We reasoned that if a protein is spatially closer to and/or has a stronger interaction with Htt, its abundance in the Htt group should be higher than that of the NES group, i.e., the ratio of protein abundance (FoldChange) between the Htt group and the NES group should be above 1. By comparing the FoldChange of each protein in the Htt group to the NES group, we identified two groups of proteins that have either increased (Group 1, red symbols, FoldChange > 1.5, p < 0.05, Fig. 4A) or decreased (Group 2, blue symbols, FoldChange < 0.67, p < 0.05, Fig. 4A) affinities for Htt. A full list of proteins in these two groups is shown in Fig. 4B. Proteins in Group 1 have little overlap between the Htt23Q and Htt145Q group (Fig. 4B), suggesting that Htt145Q and Htt23Q are associated with distinct intracellular microenvironments. Cellular Component (CC) enrichment analysis suggest that Htt145Q is more likely to be associated with proteins overrepresented in myosin II filament (GO: 0097513), stress fiber (GO: 0001725) and intermediate filament (IF) (GO: 0005882) (Fig. 4C). Under stress, additional CC terms appeared for Htt145Q (Fig. 4C). Proteins in Group 2 interact with Htt23Q and Htt145Q at a much lower frequency than the soluble NES-APEX2 (blue symbols, Fig. 4A). Interestingly, there was a large overlap in these proteins between the Htt23Q and Htt145Q group under both control and stress conditions (Fig. 4A-B). Taken together, these data suggest that normal and mutant Htt are associated with different intracellular microenvironments.

Fig. 4. Htt23Q and Htt145Q have different spatial distributions in the cytosol.

A. Volcano plots of FoldChanges of protein abundance for each protein in Htt145Q and Htt23Q groups compared to the NES group against their significances under control and MG132 treatment. The vertical lines indicate a threshold of FoldChange >1.5 or < 0.67, i.e., −0.4 < ln(FoldChange) < 0.4. The horizontal line indicates a threshold of p < 0.05. B. A heatmap showing the ln(FoldChange) of significant proteins identified in A under different conditions. C. GO enrichment analysis based on Cellular Component. D. Volcano plots of FoldChanges of protein abundance for each protein in the Htt145Q group compared to the Htt23Q group against their significances under control and MG132 treatment. The vertical lines indicate a threshold of FoldChange >1.5 or <0.67, i.e., −0.4 < ln(FoldChange) < 0.4. The horizontal line indicates a threshold of p < 0.05. E. Representative Western blot images showing the presence of MYH10, MYH9, PC, LMNA, and VIM proteins in enriched biotinylated samples that were pulled down by streptavidin beads. Actin was used as the loading control. F. Volcano plot of FoldChanges of protein abundance for each protein in the stress group compared to the control group against their significances in Htt145Q, Htt23Q and NES cells. The vertical lines indicate a threshold of FoldChange >1.5 or <0.67, i.e., −0.4 < ln(FoldChange) < 0.4. The horizontal line indicates a threshold of p < 0.2.

We next compared protein abundance between the Htt145Q and the Htt23Q group. Under both control and MG132 treatment, five proteins were identified as preferential Htt145Q interactors (FoldChange > 1.5 and p < 0.05). These five proteins are vimentin (VIM), lamin A/C (LMNA), myosin IIa heavy chain (MYH9), myosin IIb heavy chain (MYH10) and nestin (NES) (Fig. 4D). On the other hand, proteomic analysis indicates that pyruvate carboxylase (PC), a mitochondrial protein, has a stronger association with Htt23Q under control, which appeared to be weakened after MG132 treatment (Fig. 4D). The presence of these proteins in APEX2-Htt biotinylated samples was further confirmed by Western blot (Fig. 4E).

We lastly investigated the dynamic changes of Htt interactome in response to the proteotoxic stress. This was done by comparing the protein abundance between the control and MG132-treated samples within each cell type. If there are no dynamic changes in response to MG132 treatment, we would expect that most proteins remain at the base of the volcano plot. This is what we observed in the NES group (right panel, Fig. 4F), indicating that stress did not affect the random labeling of APEX2-NES. Notably, Htt145Q and Htt23Q interactomes exhibited different shifting patterns in response to MG132 treatment (left and middle panels, Fig. 4F). More proteins were left-shifted in the Htt145Q group (left panel, Fig. 4F), indicating that Htt145Q weakened its interactions with some proteins in response to MG132 treatment. In comparison, the Htt23Q group had more proteins right-shifted (middle panel, Fig. 4F), indicating that Htt23Q strengthened some PPIs in response to MG132 treatment. Taken together, the data strongly suggest that Htt145Q and Htt23Q respond distinctively to MG132 treatment. Since no proteins achieved a significance of p < 0.05 possibly due to the short incubation time with MG132, we lowered the p-value to p < 0.2 with an absolute value of FoldChange > 1.5. The proteins with decreased (blue symbols) or increased (red symbols) associations with Htt in response to stress are shown in Fig. 4F.

Vimentin is a preferential interacting protein with mutant Htt

Vimentin is shown to be one of the most prominent proteins that are preferentially associated with muHtt as quantified from the TMT labeling experiment, indicating that muHtt has a stronger interaction with vimentin in situ (Fig. 5A). To further characterize this interaction, we performed co-immunoprecipitation (co-IP) experiments. Since MG132 treatment had little effects on the interaction (Fig. 4D), we focused on untreated samples. It is known that vimentin is assembled into filaments in cells and becomes highly insoluble (Perez-Sala et al. 2015). We therefore lowered the centrifugal force during sample clearance to maintain vimentin levels in the co-IP input fractions. It is still notable that vimentin fraction in the NES group is appreciably lower than in the Htt overexpression groups (Fig. 5B, supernatant input). But importantly, the vimentin levels in the Htt23Q and Htt145Q groups remain comparable. We first used Flag M2 antibodies to pull down Flag-APEX2-Htt proteins. We could detect vimentin in the co-immunoprecipitated complexes (Fig. 5B). The presence of Htt was confirmed by Htt antibodies. When the Htt complex was pulled down by Htt antibodies, we could detect the presence of vimentin (Fig. 5C middle panels) and vice versa (Fig. 5C, right panels). A minor band of vimentin was detected in samples immunoprecipitated with rabbit control IgG, which could be nonspecific background signals. Alternatively, vimentin may have a low affinity for the Fc fragment of IgG as reported (Hansson et al. 1984). Nonetheless, it is obvious that Htt was not present in control IgG-pulled samples and the vimentin signal from vimentin antibody-pulled samples was much stronger than control IgG-pulled samples. Collectively, the co-IP experiment indicates that Htt interacts with vimentin although it did not suggest the preferential association of vimentin with muHtt, possibly due to the fact that the nature of vimentin and Htt interactions in vivo is not defined and could be affected during the co-IP sample preparations.

Fig. 5. Validation of the interactions between Htt and top candidate proteins.

A. Quantitative analysis of the abundance of biotinylated vimentin in different samples from four independent biotinylation experiments that were quantified by TMT labeling (Additional File 2). *p < 0.05, ***p < 0.001, ****p < 0.001 One-way ANOVA with Tukey’s post-hoc test. The full statistical reports of the analysis can be found in Additional File 3. B. Representative Western blot images from three biologically independent experiments showing that vimentin is present in Flag M2- immunoprecipitated Htt protein complexes. Htt and Flag antibodies were used to confirm the presence of Flag-APEX2-Htt and Flag-APEX2-NES in the immunoprecipitated samples. GAPDH was used as the input loading control. C. Representative Western blot images from three biologically independent experiments showing that vimentin is present in Htt protein complexes immunoprecipitated by Htt antibodies (middle panels) and vice versa (right panels). Rabbit control IgG was used as the negative control. D. Representative confocal images showing the interaction of Htt and VIM by PLA assay in STHdhQ7 and Q111 cells. E. Quantification of PLA-positive puncti per cell in STHdhQ7 and Q111 cells. Unpaired two-tailed Student’s t-test. t = 1.126, df = 9, p = 0.29. N=5-6 confocal fields from two independent culture preparations. F. Representative confocal images showing the interaction of Htt and VIM by PLA assay in healthy and HD NSCs. G. Quantification of PLA-positive puncti per cell in healthy and HD NSCs. *p < 0.05, Unpaired two-tailed Student’s t-test. t = 2.545, df = 6, p = 0.04. N=4 confocal fields from two independent culture preparations.

We therefore further analyzed the interaction between vimentin and Htt in situ by a proximity ligation assay (PLA) using two distinct HD cellular models. The first one is the immortal striatal cell lines derived from STHdhQ7/7 and STHdhQ111/111 knock-in HD mouse models (Trettel et al. 2000). There were apparently more PLA-positive puncti per cell in STHdhQ111 cells than in STHdhQ7 cells (Fig. 5D), although the quantification did not reach statistical significance (Fig. 5E, p = 0.29). We next used neural stem cells (NSCs). It is known that vimentin is naturally enriched in NSCs and gradually replaced by neurofilaments when neurons become mature (Yabe et al. 2003). We therefore examined the Htt-vimentin association in NSCs differentiated from induced pluripotent stem cells (iPSCs) derived from an apparently healthy individual and a HD patient. The identities of the iPSCs and NSCs were confirmed using specific iPSC (Oct4A, SSEA4) and NSC markers (PAX6) (Additional File 1: Supplementary Fig. S5). There were significantly more PLA-positive puncti in HD NSCs than in healthy ones (Fig. 5F-G). These data suggest that muHtt, compared to normal Htt, is in closer proximity to vimentin in situ.

Mutant Htt affects vimentin’s function in regulating proteostasis

Recently, a novel role of vimentin in coordinating protein turnover by formation of vimentin cages surrounding the aggresomes in NSCs was reported (Morrow et al. 2020). We previously showed disrupted proteostasis in HD cells during proteotoxic stress (Huang et al. 2018). To investigate whether muHtt affects vimentin’s function in regulating proteostasis, we examined the presence of vimentin cages surrounding aggresomes in NSCs treated with MG132. In both healthy and HD NSCs, vimentin filaments formed a network that extended from the nucleus toward the plasma membrane (Fig. 6A and B left panels). Remarkably, upon proteotoxic stress, vimentin filaments retracted from the plasma membrane, coalesced around the nucleus in healthy NSCs, but with a lesser extent in HD NSCs (Fig. 6A-B right panel), suggesting potentially impaired vimentin dynamics. Furthermore, vimentin formed cage-like structures around aggresomes in MG132-treated healthy NSCs (Fig. 6C, 6E). Although aggresomes in HD NSCs still formed in response to MG132 treatment, most of them were not surrounded by vimentin cages (Fig. 6D, 6F). Based on the 3D-confocal images from three independent batches of iPSC-differentiated NSCs, we examined between 60 and 100 K48-staining aggresomes as demonstrated in Fig. 6E-F and counted the number of aggresomes that were clearly surrounded by vimentin cages using the 3D reconstructed images. We estimated that ~76% of aggresomes in healthy NSCs and only ~25% of aggresomes in HD NSCs were surrounded by vimentin cages. In summary, these data suggest that muHTT in HD NSCs impairs the ability of vimentin to form cages around aggresomes in response to proteotoxic stress.

Fig. 6. Mutant Htt affects vimentin cage formation in NSCs in response to proteotoxic stress.

To induce aggresomes formation, cells were treated with 5μM MG132 for 4 hours before fixation in 4% PFA. A-B. Representative confocal images of vimentin staining in healthy (A) and HD (B) NSCs under control conditions or after MG132 treatment. C-D. Representative maximal projections of 3D confocal images of healthy (C) and HD (D) NSCs co-immunostained with K48-ubiquitin and vimentin antibodies after MG132 treatment. Healthy NSCs had more cage-like vimentin structures around aggresomes compared to HD NSCs. Arrows indicate aggresomes surrounded by vimentin cages based on 3D reconstruction. Arrowheads indicate the aggresomes not surrounded by vimentin cages based on 3D reconstruction. E-F. Enlarged maximal projections of 3D confocal images showing that vimentin formed a cage-like structure around an aggresome in healthy NSCs (E), but not HD NSCs (F) after MG132 treatment. All experiments were repeated with three independent differentiations of NSC cultures.

Discussion

Htt plays an important role in a number of fundamental cellular functions as a scaffold protein via protein-protein interactions. In this study, we applied an APEX2-based proximity labeling approach coupled with quantitative proteomics to spatiotemporally quantify and differentiate Htt interactomes of normal and mutant Htt. Some proteins detected in our current study overlap with reported Htt interactomes identified from traditional affinity purification studies (Ratovitski et al. 2012; Shirasaki et al. 2012). On the other hand, many known Htt-interacting proteins, such as huntingtin associated protein 1 (HAP1) (Li et al. 1995), HAP40 (Guo et al. 2018) and huntingtin-interacting protein 14 (HIP14) (Sanders et al. 2014) were not present in our list. This is not unusual given the methodological differences between live-cell proximity labeling and affinity purification by co-IP. Since APEX2 labeling occurs within a short radius of 20 nm (Rhee et al. 2013), we believe that our identified Htt-interacting proteins comprise the first shell of the Htt interactome, including proteins either directly interacting with or spatially proximal to Htt.

One important finding of the present study is that normal and mutant Htt are spatially associated with distinctive subcellular microenvironment. Comparing the protein abundance in the Htt23Q and Htt145Q samples to the NES sample under basal conditions provides insights into the local milieu of Htt. Our data revealed two classes of proteins with either higher (Group 1) or lower (Group 2) affinities for Htt (Fig. 4A). The nature of the interaction between Group 2 proteins and Htt deserves further investigations. It is possible that the bulky Htt protein is more spatially confined and has fewer interactions with Group 2 proteins as compared to the small, freely diffusible APEX2-NES proteins. For example, biotinylation of cell cycle associated protein 1 (CAPRIN1), a protein involved in stress granule formation, was significantly lower in both Htt groups compared to the NES group (Fig. 4A). This suggests that the interaction between CAPRIN1 and Htt is not dynamically dominant under basal or mild proteotoxic condition, although it could be enhanced under other stress conditions, such as ER stress (Ratovitski et al. 2012). When comparing Htt145Q to Htt23Q, the CAPRIN1 ratio was increased (Htt145Q/Htt23Q=1.3, p = 0.49, Additional File 2), suggesting that muHtt has a higher affinity for CAPRIN1 which is consistent with the previous reports showing that CAPRIN 1 is a preferable interactor for muHtt (Culver et al. 2012; Ratovitski et al. 2012; Eshraghi et al. 2021).

Since there is little overlap in Group 1 between the Htt23Q and Htt145Q group, we suggest that there is a distinctive difference in spatial distribution between normal and mutant Htt. Direct comparison of the protein abundance between Htt23Q and Htt145Q further supports this notion. Mutant Htt is preferentially associated with IFs and myosin complexes while normal Htt has a higher affinity for PC, a mitochondrial protein. Microtubules, IFs and actin microfilaments are the three cytoskeleton systems that crosstalk and co-regulate the structural organization of the cytoplasm of a cell (Chang & Goldman 2004). Beyond the structural support functions, IFs are increasingly recognized as signaling platforms that regulate cell migration, growth, survival and the response to stress (Pallari & Eriksson 2006). Given the diverse biological functions of huntingtin (Saudou & Humbert 2016), it remains interesting to further investigate the functional impacts of muHtt on IF.

Vimentin is one of the top candidates as a muHtt interactor in situ and was further investigated in this study. Vimentin is predominantly expressed in cells of mesenchymal and ectodermal origin (Ridge et al. 2022). Additionally, vimentin is expressed in multiple cell types of the central nervous system, including astrocytes, neural progenitor cells, immature neurons and some neurons (Ridge et al. 2022). The function of vimentin in CNS remains elusive. Vimentin is mainly expressed in NSCs and is replaced by neurofilament (NEFL) and α-internexin (INA) in mature neurons of CNS (Bott & Winckler 2020). However, neuronal expression of vimentin has been detected in Alzheimer’s disease (Levin et al. 2009) and injured nerve (Perlson et al. 2005). It is reported that vimentin formed cage-like structures around the N-terminal truncated muHtt aggregates (Bauer et al. 2012; Morrow et al. 2020). Since we did not detect any microscopically visible full-length muHtt aggregates in our stable cell line, it is unlikely that increased biotinylation of vimentin is simply due to the physical proximity between vimentin cages and aggregated muHtt. However, we cannot rule out the possibility that some muHtt microaggregates may form around vimentin filament structures.

We applied three complementary approaches to investigate the Htt-vimentin interaction. Although the co-IP experiment only suggests the interaction of vimentin with both normal and mutant Htt, the APEX2-mediated labeling and PLA experiments clearly indicate that vimentin is preferentially associated with muHtt in situ. Vimentin was identified in Htt co-immunoprecipitated mouse brain samples (Shirasaki et al. 2012). In a separate study of using HTT506-145Q and HTT506-23Q as the bait for co-IP in HEK293 cells combined with quantitative proteomic analysis, it is shown that vimentin preferentially interacts with HTT506-145Q (Hosp et al. 2015). It is not clear whether Htt interacts with filamentous or soluble vimentin from the co-IP experiment. It is likely that the interaction of Htt and vimentin is restricted to the in vivo microenvironment. Mutant Htt may affect vimentin dynamics, diminishing its ability to respond to stress. This is of particular interest in NSCs since it is shown that NSCs recover from disrupted protein homeostasis by asymmetrical segregation of protein aggregates during cell division that is mediated by vimentin (Morrow et al. 2020; Pattabiraman et al. 2020). Since HD has also been increasingly recognized as a neurodevelopmental disorder (Barnat et al. 2020; Wiatr et al. 2018; Godin et al. 2010), the preferential interaction of muHtt with vimentin may mediate some neurodevelopmental deficits observed in HD. In the adult brain, adult NSCs demonstrate an age-associated dysfunction in maintaining protein homeostasis with increased protein aggregates and decreased proliferation (Navarro Negredo et al. 2020). Interestingly, striatal neurogenesis, a unique pattern of neurogenesis in the adult human brain, is preferentially decreased in HD patients (Inta et al. 2016; Ernst et al. 2014). Therefore, it is likely that muHtt impairs adult NSC protein homeostasis through aberrant interaction with vimentin.

On the other side, our study suggest that normal Htt has a higher affinity for PC. Interestingly, another carboxylase enzyme, acetyl-CoA carboxylase alpha (ACACA), was also found to be enriched in Htt23Q (Fig. 4A, Additional File 2, Htt23Q/Htt145Q = 1.56, p =0.18), suggesting that normal Htt may participate in biotin binding activities and regulate carboxylase activities. One potential concern is that since APEX2-mediated labeling is based on protein biotinylation, endogenously biotinylated protein or biotin binding proteins that can be enriched by streptavidin pull-down could be potential contaminants. Biotin is a co-factor for carboxylase enzymes such as acetyl-CoA carboxylase (ACAC), PC, propionyl-CoA carboxylase (PCC) and 3-methyl crotonyl-CoA carboxylases (MCCC1) (Tong 2013). These proteins could be the background sources of biotinylated signals in the absence of H₂O₂ (Fig. 1B). In our analysis, we made pairwise comparisons among NES, Htt23Q and Htt145Q samples. If they are background proteins, their abundance should not change significantly across groups. Since the abundance of PC is significantly higher in the Htt23Q group than in the NES group (Fig. 4A) and Htt145Q group (Fig. 4D), we reason that it is a true member in the normal Htt interactome.

In this study, we also examined changes of the Htt interactomes in response to proteotoxic stress and showed that normal and mutant Htt interactomes exhibit different shifting patterns to stress response. Due to the short incubation time with MG132 that we intentionally designed to detect early changes, our analysis did not identify any proteins with significant changes in abundance between stress and control conditions. Nonetheless, our results established a platform for future studies of Htt-interacting proteins in stress response. It is possible that after prolonged stress, significant changes will be detected. Despite this, our data do point to a trend that normal and muHtt respond differently to stress. Specifically, we found that muHtt weakened its interaction with some specific types of histone proteins and ribosomal proteins (Fig. 4F), suggesting that muHtt may alter gene transcription and translation as an immediate response to proteotoxic stress. To support this, our recent RNA-seq analysis revealed that STHdhQ7 and Q111 cells exhibited distinct transcriptional reprogramming in response to ER stress recovery (Xu et al. 2021). In contrast, normal Htt gained interactions with proteins related to neurofilaments. The functional significance of the increase remains to be investigated. Taken together, the general left-shift of the muHtt interactome in Fig. 4F suggests a loss or modification of normal Htt function in affecting its effector proteins in response to stress, which can be attributable to HD pathogenesis. The RNA helicase, Eukaryotic initiation factor 4A1 (eIF4A1), a key component of the translation initiation machinery, has increased affinity for both normal and mutant huntingtin in response to stress (Fig. 4F). It is known that eIF4A1 forms stress granule with RNA molecules (Padrón et al. 2019). It is possible that the interaction temporally shuts down protein translation in response to proteotoxic stress, a common mechanism in the integrated stress response.

Although the application of APEX2 to cytosolic proteins has its own limitation due to the high labeling background, recent development of new labeling probes with increased labeling selectivity should help with reducing the background noise (Ke et al. 2021). Therefore, our study provides a platform that can be optimized and applied to different cellular models to further study the spatiotemporal changes of Htt interactome in response to stresses. Additionally, it provides a list of differential interacting protein candidates for both normal and mutant Htt in their natural cellular contexts. Further validation and functional characterization of these differential interactions in different cell types, should open new avenues of investigation into the molecular pathogenesis of HD.

List of abbreviations:

APEX

Ascorbate Peroxidase

BioID

Biotin Identification

CAPRIN1

Cell Cycle Associated Protein 1

CC

Cellular Component

DAPI

4′,6-diamidino-2-phenylindole

co-IP

co-immunoprecipitation

DMEM

Dulbecco’s Modified Eagle Medium

DMSO

Dimethyl Sulfoxide

DTT

dithiothreitol

FBS

Fetal Bovine Serum

GO

Gene Ontology

HD

Huntington’s Disease

Htt

huntingtin

muHtt

mutant Htt

IF

Intermediate Filament

iPSC

induced Pluripotent Stem Cells

MS

Mass Spectrometry

NES

Nuclear Export Signal

NSC

Neural Stem Cell

PBS

Phosphate-Buffered Saline

PC

Pyruvate Carboxylase

PFA

paraformaldehyde

PIPE

Polymerase Incomplete Primer Extension

PLA

Proximity Ligation Assay

PPI

Protein-protein Interaction

RIPA

radioimmunoprecipitation assay

TMT

Tandem Mass Tag

VIM

vimentin