Abstract

   Huntington’s disease (HD) is a debilitating neurodegenerative disorder
   affecting an individual’s cognitive and motor abilities. HD is caused
   by a mutation in the huntingtin gene producing a toxic
   polyglutamine-expanded protein (mHTT) and leading to degeneration in
   the striatum and cortex. Yet, the molecular signatures that underlie
   tissue-specific vulnerabilities remain unclear. Here, we investigate
   this aspect by leveraging multi-epitope protein interaction assays,
   subcellular fractionation, thermal proteome profiling, and genetic
   modifier assays. The use of human cell, mouse, and fly models afforded
   capture of distinct subcellular pools of epitope-enriched and
   tissue-dependent interactions linked to dysregulated cellular pathways
   and disease relevance. We established an HTT association with nearly
   all subunits of the transcriptional regulatory Mediator complex
   (20/26), with preferential enrichment of MED15 in the tail domain.
   Using HD and KO models, we find HTT modulates the subcellular
   localization and assembly of the Mediator. We demonstrated striatal
   enriched and functional interactions with regulators of calcium
   homeostasis and chromatin remodeling, whose disease relevance was
   supported by HD fly genetic modifiers assays. Altogether, we offer
   insights into tissue- and localization-dependent (m)HTT functions and
   pathobiology.

   Keywords: IP-MS, Mediator, Calcium, Thermal Profiling, Mass
   Spectrometry

   Subject terms: Neuroscience, Proteomics

Synopsis

   graphic file with name 44320_2025_96_Figa_HTML.jpg

   Multi-epitope immunopurifications of huntingtin (HTT) from vulnerable
   brain regions of Huntington’s disease (HD) models identify
   disease-linked protein interactions networks, capturing HD cell biology
   across pathways and subcellular compartments.
     * Epitope- and tissue-dependent HTT interactomes identify
       membrane-enriched and nuclear-associated protein networks that may
       explain the relative vulnerability of cells to mutant huntingtin
       toxicity.
     * Mutant HTT alters wild-type interaction with the Mediator complex,
       likely via MED15 or MED27, which shows progressive re-distribution
       to the cytoplasm.
     * The effect of polyQ expansion on the HTT interaction with TCERG1, a
       genetic modifier of HD, is brain region-specific.
     * HD fly genetic modifier assays highlight disease-associated
       candidates within striatum protein networks that regulate calcium
       homeostasis and chromatin remodeling.
     __________________________________________________________________

   Multi-epitope immunopurifications of huntingtin (HTT) from vulnerable
   brain regions of Huntington’s disease (HD) models identify
   disease-linked protein interaction networks, capturing HD cell biology
   across pathways and subcellular compartments.

   graphic file with name 44320_2025_96_Figb_HTML.jpg

Introduction

   Although discovered as the etiological agent of Huntington’s disease
   (HD) in the early 1990’s (MacDonald et al, [37]1993), the precise
   mechanisms by which polyglutamine (polyQ) expansion within the
   N-terminal region of HTT causes toxicity in vulnerable brain cells
   remains an active area of study (MacDonald et al, [38]1993).
   Significant progress has been made in understanding HD pathobiology by
   defining the molecular landscapes of HD models using quantitative omics
   approaches. For example, transcriptomic disturbances are key features
   of HD-induced pathobiology (Kumar et al, [39]2014), with recent efforts
   aimed at teasing apart the molecular contributions of polyQ versus
   uninterrupted CAG repeat lengths in striatum-enriched pathogenesis (Gu
   et al, [40]2022). Moreover, the granularity afforded by single
   cell/nuclear RNAseq is starting to reveal the contribution of different
   cell types and neuronal pathways to disease pathobiology (Matsushima et
   al, [41]2023; Pressl et al, [42]2024; Mätlik et al, [43]2024).
   Multiomics interrogation of an allelic series of knock-in HD mice
   (Zheng et al, [44]2012) has identified sequential polyQ-dependent
   disruptions of RNA, microRNA, and protein networks (Langfelder et al,
   [45]2016, [46]2018). Further integrative analysis of the allelic series
   datasets revealed consistent striatal-specific HD-associated disease
   signatures that are linked to disease progression (Obenauer et al,
   [47]2023). Evaluating these molecular signatures in conjunction with in
   vivo neuronal survival data helped to define cell type-specific
   temporal dynamics of homeostatic versus pathogenic responses in the
   striatum (Megret et al, [48]2021). HD models with perturbed omics
   networks have been complemented by the development and application of
   HD modifier screens in mice and flies, identifying genetic loci that
   affect disease progression (Wertz et al, [49]2020; Romero et al,
   [50]2008; Al-Ramahi et al, [51]2018; Miller et al, [52]2012).
   Therefore, an ongoing area of interest is focused on identifying
   striatal molecular signatures that are primary versus compensatory
   drivers of increased vulnerability to HD-induced pathology.

   One hypothesis is that high-value disease-relevant candidates are
   proteins that interact with HTT in the striatum and are linked to HD
   pathogenesis and/or disease progression (Goehler et al, [53]2004;
   Kaltenbach et al, [54]2007; Neri, [55]2011; Shirasaki et al, [56]2012;
   Greco et al, [57]2022). While most HTT protein interactions that are
   disease modifiers have been defined in HD model organisms, there are
   several interactions identified as human genetic HD modifiers from
   genome-wide association studies, including TCERG1 (Holbert et al,
   [58]2001; Kaltenbach et al, [59]2007; Andresen et al, [60]2007; Lobanov
   et al, [61]2022), MSH3 (Shirasaki et al, [62]2012; Flower et al,
   [63]2019), and more recently MLH1-PMS2 (Lee et al, [64]2017; Sun et al,
   [65]2024) and POLD1 (Ripaud et al, [66]2014; Consortium et al,
   [67]2024). Over the last 20 years, protein interaction studies of HTT
   have evolved in concert with core technological advances and the
   development of increasingly powerful HD model systems (Silva Ramos et
   al, [68]2024). Computational platforms have been developed and made
   accessible to the community as tools for data integration and
   visualization, such as the HDinHD portal for HD multi-omic data and
   model systems and the HTT-OMNI for protein interaction analysis and
   visualization (Aaronson et al, [69]2021; Kennedy et al, [70]2022; Meem
   et al, [71]2023). Together, curation of HTT interacting proteins (HIPs)
   has produced a compilation of >3000 unique candidates (Aaronson et al,
   [72]2021). Ironically, the breadth of these interactions has created
   challenges in identifying HIPs that are disease-relevant and promising
   candidates for therapeutic targeting. Thus, progress towards
   identifying proximal, disease-relevant HIPs will require tighter
   integration of multifaceted experimental and computational approaches.
   These approaches would leverage the latest HD animal model systems and
   human specimens to prioritize HIPs through the integration of
   disease-dependent parameters, such as polyQ dependence, interaction
   stability and interface, tissue/cell type and subcellular localization,
   bulk and single-cell multiomics measurements, and modifiers of disease
   phenotypes. Towards these goals, our group has previously used
   quantitative proteomics to characterize the impact of expanded polyQ on
   the formation and relative stability of HIPs in the striatum (Greco et
   al, [73]2022) and cortex (Kennedy et al, [74]2022) in an age-depended
   manner in mice expressing FLAG-tagged HTT with expanded (Q140) or
   normal (Q20) polyQ. We found extensive polyQ-dependent HIP
   dysregulation occurring at an early age, prior to the onset of
   Q140-dependent HD phenotypes, i.e., as early as 8 weeks old in the
   striatum. Using bioluminescence-based two-hybrid assays, we showed that
   the dysregulated HIPs that we found in mouse HD models can also have
   direct interactions with HTT in human cells. To assess these
   dysregulated HIPs in the context of disease pathogenesis, a fly HD
   model showed, at the genetic level, a link between these interacting
   proteins and altered motor performance. Furthermore, comparing the
   molecular signatures of HD disease progression observed in the allelic
   series transcriptome and proteome networks, our study suggested that
   polyQ-dependent modulation of the interactome precedes the disruption
   of transcriptome and proteome (Greco et al, [75]2022).

   Encouraged by the findings described above supporting the idea that
   HIPs are key molecular players in mediating mutant HTT (mHTT)
   pathobiology, attention should be turned to several fundamental aspects
   of the HIP landscape that remain unclear. First, we lack a complete
   understanding of the subset of HIPs that preferentially occur in
   regions of the brain most impacted by disease (striatum and cortex)
   compared to regions that are relatively spared of pathology (e.g., the
   cerebellum). An early pioneering study by Shirasaki and colleagues used
   immunopurification-based proteomics to identify mHTT-dependent
   interactions and reveal functional modules associated with the
   striatum, cortex, and cerebellum interactions (Shirasaki et al,
   [76]2012). Yet, the available proteomic technologies did not afford
   direct quantitative comparisons of HIP abundances between tissues.
   Second, protein interactions identified with endogenous, non-tagged HTT
   are vastly underrepresented in the HIP knowledgebase (Kennedy et al,
   [77]2022; Aaronson et al, [78]2021), providing an unknown level of bias
   in current candidate HIPs. Third, the interaction interfaces of
   proteins with full-length HTT have not been well characterized. Our
   recent efforts to define direct HTT interactions using two-hybrid dual
   bioluminescence assays highlighted the potential for spatial
   segregation of HIPs between the N- versus C-terminal domains of HTT.
   The development of distinct HD pathologies between different brain
   regions and the availability of HTT antibodies that cover unique
   epitopes provides an unexplored path to localize HIPs to specific HTT
   domains within different tissues.

   Towards addressing these gaps in knowledge, we integrated multi-epitope
   protein interaction studies, subcellular fractionation, thermal
   proteome profiling, and genetic modifier assays to characterize
   tissue-specific and epitope-enriched protein interactions of endogenous
   wild-type and mutant HTT. Leveraging human cell, mouse, and fly models
   allowed us to profile distinct subcellular pools of HTT complexes,
   which were linked to dysregulated cellular pathways involved in calcium
   homeostasis, chromatin remodeling, and transcription. Tissue-enriched
   HIPs were annotated to non-overlapping cellular processes. We also
   found an HTT association with nearly all subunits of the canonical
   Mediator complex (20/26), a major transcriptional regulator, with
   preferential enrichment of the MED15 subunit in the tail domain. The
   subcellular localization of this interaction was differentially
   modulated by mHTT in the striatum of HD mice at an early versus later
   stage of HD-like pathology, while loss of HTT in human neuroblastoma
   cells altered the subunit assembly state of Mediator. Lastly, we
   demonstrated that the striatal-enriched HTT interaction with
   voltage-dependent calcium channels (CACNB and CACNA1) and chromatin
   remodeling factors, SMARCA5 and CHD2, were modifiers of mHTT phenotypes
   in HD fly genetic assays. Taken together, this study highlights tissue-
   and localization-dependent (m)HTT functions that identify candidate
   sensitizing or protective molecular factors during disease progression.

Results

Mapping brain region-dependent HIPs using multi-epitope immunocapture

   We sought to characterize the protein interactions of endogenous HTT by
   considering several layers of regulation in the context of HD—tissue
   type, polyQ- and age-variation, as well as epitope selectivity
   (Fig. [79]1A). Striatum, cortex, and cerebellum were dissected from
   8-week and 40-week old HD mice containing a humanized HTT exon 1 with
   either a normal (Q20) or expanded (Q140) polyQ region. This expanded
   polyQ length causes progressive behavioral and neuromotor dysfunction
   in mice, similar to that observed in human disease (Menalled et al,
   [80]2003; Hickey et al, [81]2008). To assess epitope selectivity,
   endogenous HTT was isolated with antibodies that target different
   regions of the protein. Individually and in combination, we tested nine
   anti-HTT antibodies that target the N-terminus, proline-rich region,
   the helical HEAT repeats, the bridge domain (located between the third
   and fourth HEAT domains), or the C-terminus (Fig. [82]EV1A). We
   observed several high-performing antibodies for HTT capture, such as
   EPR5526, whose epitope lies in the first 17 amino acids (N17)
   (Fig. [83]EV1A). In contrast, the C-terminal antibodies, 138 and 139,
   showed lower capture of HTT (Fig. [84]EV1A). For several antibodies
   with neighboring epitopes, their IP capture performance was additive
   when used in combination (Fig. [85]EV1B). Specifically, we optimized IP
   conditions for antibody combinations that targeted the N-terminus and
   proline-rich regions (2B7 and 4C9 antibodies, respectively), as well as
   the second HEAT repeat and bridge regions (3E10 and 4E10 antibodies,
   respectively) using western blotting analysis of Q20 whole brain
   lysates. Next, we evaluated antibody specificity using immunoaffinity
   purification paired with tandem mass spectrometry (IP-MS) (Dataset
   EV[86]1). Consistent with the western blot analysis, C-terminal
   antibodies 138–139 showed ~50–100× lower capture of HTT. The captured
   proteins with the highest abundance were aquaporin-4 (Aqp4) and Hsp10
   (Hspe1), which were uniquely identified in the 138–139 IPs
   (Fig. [87]EV1C; Dataset EV[88]1), suggesting potential off-target
   capture. For the isolations with EPR5526, 2B7-4C9, or 4E10-3E10
   antibodies, HTT and its known interacting partner, HAP40 (F8a1) were
   consistently isolated and among the most abundant captured proteins
   (Fig. [89]EV1C). However, we observed that EPR5526 captured a unique
   population of interacting proteins compared to 2B7-4C9 and 4E10-3E10
   (Fig. [90]EV1C,D). We investigated this further in the striatum and
   found that, despite the efficient capture of HTT/HAP40 in an ~1:1
   ratio, EPR5526 also captured synaptojanin-1 (Synj1) with ~10× higher
   abundance than HTT/HAP40 (Fig. [91]EV1E). Interestingly, the capture
   may be based on a shared conformational epitope. The primary sequence
   alignment of HTT-Synj1 and the Synj1 crystal structure (AlphaFold
   prediction) supported an alpha-helix conformation (Fig. [92]EV1F), but
   we did not observe an anti-EPR5526 immunoreactive band in the expected
   region (~173 kDa) by western blotting. Overall, based on our IP
   efficiency and specificity analyses, we proceeded with 2B7-4C9 and
   4E10-3E10 antibodies for the complete workflow (Fig. [93]1A). Both of
   these antibody combinations showed effective capture of HTT relative to
   the isotype-matched controls for Q20 and Q140 genotypes from the whole
   brain (Fig.  [94]EV1C) and striatum (Fig. [95]EV1D,G), as well as the
   enrichment for validated HTT interacting proteins (Fig. [96]EV1G),
   including HAP40.

Figure 1. Characterizing protein interactions of endogenous WT and mutant HTT
in HD mice.

   [97]Figure 1
   [98]Open in a new tab

   (A) IP-MS strategy and computational analysis workflow. Mice aged 8
   weeks or 40 weeks from two genotypes (Q20 and Q140 HTT) were chosen as
   a model for HD. The striatum, cortex, and cerebellum were dissected
   from sacrificed animals and lysed. Antibodies that target different
   regions of HTT (2B7 & 4C9: N-terminus; 3E10 & 4E10: Central) and
   isotype-matched control antibodies (IgG) were utilized for IP followed
   by LC-MS/MS. Interaction specificity was determined by SAINT analysis.
   Precursor-based label-free quantification and bioinformatic analyses
   were performed to characterize age, genotype, tissue, and
   epitope-driven HTT interactions. (B) Domain map of HTT with antibody
   epitope regions denoted. (C) IP-MS protein abundance levels of HTT,
   isolated from the striatum, cortex, and cerebellum (mean ± SD, n = 3–5
   biological replicates). HTT abundance was also scaled to the whole-cell
   lysate amounts of HTT quantified by DIA-MS (line plots, top). (D) The
   number of HTT interactions in each mouse genotype and age that passed
   SAINT specificity filtering for striatum, cortex, and cerebellum IP
   conditions. (E) PCA analysis of HTT interacting protein abundances
   quantified by IP-MS in striatum, cortex, and cerebellum. Interaction
   abundances were normalized to respective HTT bait abundances.

Figure EV1. Evaluating IP efficiency and specificity of HTT antibodies for
isolation of endogenous WT and mutant HTT from whole brain lysates of HD
mouse models.

   [99]Figure EV1
   [100]Open in a new tab

   (A) Domain map of HTT with IP antibodies marking their approximate
   epitope regions (see Materials and Methods). Blue cylinders represent
   the 7 HEAT repeat domains of HTT. Antibody IP efficiency was evaluated
   in Q20-8 weeks whole brain by western blotting using Odyssey Infrared
   detection (bottom). Antibodies in blue text had sufficient IP
   efficiency for further evaluation. FT = flow-through fraction, E =
   eluted fractions. (B) Western blot densitometry quantification of
   percent of HTT depleted from the flow-through (FT) vs. mIgG control
   lanes, and amount of HTT recovered from the elution (E) vs. mIgG
   control lanes. Antibodies are ordered (top to bottom) by epitope
   location (N-terminal to C-terminal). (C) Heatmap of the top 5% most
   abundant HTT interacting proteins (n = 79) isolated and quantified by
   IP-MS from whole brain using different HTT antibodies (EPR5526,
   2B7-4C9, 4E10-3E10, 138-129). Raw MS protein abundances were normalized
   by their respective replicate medians and theoretical number of tryptic
   peptides and expressed as the log[2] average of two biological
   replicates in Q20 and Q140-8 weeks conditions. Normalized protein
   abundances were analyzed by hierarchical clustering (distance metric:
   Euclidean and method: Ward) with no additional normalization or
   scaling. Missing values were imputed with the minimum quantified value.
   (D) Violin plot of log[2] enrichment ratio (HTT IP vs mIgG IP)
   distribution for proteins co-isolated with HTT Q20 or Q140 from the
   striatum (8 weeks) using antibodies EPR5526, 2B7-4C9 pair, 4E10-3E10
   pair, and D7F7. Enrichment ratio of HTT bait is indicated by orange
   circle. (E) Comparison of Htt, Hap40 (F8a1), and Synj1 levels in IPs of
   HTT Q20 from the striatum of 8 weeks old mice using EPR5526, 2B7-4C9,
   or 4E10-3E10 antibodies (n = 4, mean ± SD). Protein abundances were
   calculated as in (C). (F) Sequence alignment of Htt N17, corresponding
   to the EPR5526 epitope region, and Synj1 (aa. 399–408). The predicted
   crystal structure (AlphaFold 2.0) is shown, labeled with selected
   aligned amino acids of Synj1/HTT in an alpha-helix (dark blue), which
   forms hydrogen bond contacts with a proximal alpha-helix (light blue).
   (G) Scatterplot comparisons of co-isolated (prey) protein log[2]
   enrichment ratios (HTT vs mIgG IP) in Q20 vs Q140 for N-terminal (left)
   and Central (right) antibody sets. Bait HTT and selected positive
   control interactions are denoted in red and orange circles,
   respectively. HTT interactions that were annotated in at least 2/3 of
   the previously reported HTT IP-MS studies from mouse brain are
   indicated in light blue circles. (H) Scatterplot of log[10] abundance
   (MS summed intensity) for HTT vs. HAP40 showing high correlation in all
   individual IP-MS experiments by tissue. (I) Principal component
   analysis of all co-isolated log[2] protein abundances stratified by
   antibody set (N-terminal and Central) and tissue. Abundances were
   median normalized. (J) Same as (I), except stratified by antibody set
   and polyQ length, for striatum (left), cortex (middle), and cerebellum
   (right).

   Following antibody optimizations and specificity assessment, we
   proceeded to characterize HTT interactions from HD and control mice
   striatal, cortical, and cerebellar tissues collected at 8 and 40 weeks
   of age. Tissue lysates were divided between IPs performed with optimal
   N-Terminal (N-Term) and Central HTT antibodies (Fig. [101]1B) and
   isotype-matched IgG controls. In parallel, a portion of the lysate was
   reserved for label-free proteomics to control for proteome abundance
   changes of HTT and its interactions, which could influence the
   interpretation of interaction levels. For example, polyQ expansion
   resulted in diminished Triton-soluble HTT at the whole proteome level,
   which was exacerbated by mouse age. This phenotype was stronger in the
   striatum and cortex (~60% decrease relative to Q20) relative to the
   cerebellum tissue (~30% decrease). Subsequently, immunoisolates were
   digested and analyzed by tandem mass spectrometry as above. We first
   assessed HTT capture across all conditions. In the N-term IPs, we saw a
   polyQ- and age-dependent reduction of captured HTT. While most of this
   effect likely resulted from the lower HTT at the proteome level, it is
   also possible that the expanded polyQ contributes to the reduced
   capture performance of the N-term targeting antibodies (Fig. [102]1C,
   left panels). Also noteworthy, targeting Central epitopes upon polyQ
   expansion (Q140) resulted in greater capture of HTT relative to Q20,
   which cannot be explained by changes in HTT protein abundance. Rather,
   this may be contributed by changes in epitope availability due to
   conformational or PPI-driven effects of polyQ expansion. Next, as a
   positive control, we evaluated the co-isolation of HAP40, which
   directly interacts with HTT and is the only structurally characterized
   PPI to date (Guo et al, [103]2018; Harding et al, [104]2021).
   HTT-associated HAP40 levels were reproducibly quantified across all
   conditions (age, genotype, tissue, and antibody). HAP40 abundance
   correlated linearly with HTT abundance and was independent of antibody,
   tissue, or genotype (Fig. [105]EV1H).

   When considering the abundance profiles of HTT co-isolated proteins,
   these were primarily separated by tissue (Fig. [106]EV1I) and antibody
   (Fig. [107]EV1J), as shown by Principal Component analyses. Further
   filtering by interaction specificity using the Significant Analysis of
   INTeractome (SAINTexpress) (Teo et al, [108]2014) tool identified ~1100
   high-confidence HTT interacting proteins as a union across all
   conditions. This computational tool models the MS signal distribution
   of proteins from the control (IgG) IPs and the experimental (HTT) IPs
   to define experiment-optimized thresholds for removal of non-specific
   interactions. The overall number of SAINT-filtered proteins showed a
   polyQ-dependent effect that correlated with the known relative tissue
   vulnerabilities (Fig. [109]1D). Specifically, a polyQ-dependent
   increase in the number of SAINT-specific interactions was observed in
   the striatum, which was much less pronounced in the cortex, while the
   number of interactions was relatively constant in the cerebellum
   (Fig. [110]1D). Moreover when considering the abundance profiles of
   SAINT-filtered HTT interacting partners isolated from the striatum and
   cortex, the Q140 40-week-old mice were more distinct when using HTT
   antibodies targeting the N-terminal versus the Central epitope, as
   shown by this sample group’s variance in PC1 (Fig. [111]1E, blue vs.
   gray). In contrast, HTT interaction profiles in the cerebellum showed
   the most separation by the antibody variable, but less so by the mHTT
   or age variables (Fig. [112]1E, diamonds vs circles). Overall, these
   results suggest that different subsets of proteins interacting with
   wild-type or mutant HTT may cause preferential immunocapture depending
   on the targeted epitopes. Taken together with our observation that mHTT
   appears to alter antibody affinity differentially between N-term and
   Central antibody capture (Fig. [113]1C), our approach targeting
   multiple HTT epitopes can provide a more complete perspective of
   tissue-specific interactions and dysregulated PPIs.

Epitope-driven HTT PPI signatures reveal tissue and polyQ-dependent cellular
functions

   Given our finding of epitope-driven PPI signatures, we asked which PPIs
   were the primary drivers of these unique profiles, whether specific
   biological processes are enriched in these signatures, and if they were
   influenced by tissue and polyQ expansion. PPI abundance can vary as a
   factor of the amount of HTT captured and contributions from biological
   influences at the proteome level (Fig. [114]1C). Therefore, we scaled
   PPI abundances within each tissue based on the HTT level in each IP.
   Next, we calculated fold changes and the significance of PPI relative
   abundance changes between the N-terminal and Central region IPs within
   each mouse genotype and age for each tissue. This comparison allowed us
   to identify a core set of ~700 consistent HIPs, as well as distinct
   sets of HIPs that exhibit a preference for specific epitopes. A range
   of 209–295 HIPs (between tissues) demonstrated preference (>log[2] 2.5)
   for the N-terminal IP condition, while between 108–130 PPIs
   preferentially enriched with the Central region IP (Fig. [115]2A–C). As
   a next step, we characterized enriched features of the
   epitope-influenced HIPs across tissues. We first performed gene
   ontology (GO) enrichment for subcellular localization of the HTT
   interacting proteins. In the striatum, the N-terminal antibodies
   enriched for proteins associated with the cytoplasm, synapse, and
   cytoskeleton (Fig. [116]2D), which are among the well-represented
   annotations of previously reported HIPs (Kennedy et al, [117]2022).
   Surprisingly, the PPIs enriched from the cortex and cerebellum by the
   N-terminal IP showed a different localization annotation pattern. The
   enrichment in localization to the synapse, cell projection, and cell
   junction were all diminished in the cortex (Fig. [118]2E) and
   cerebellum (Fig. [119]2F)—with the largest decrease in these terms
   associated with the cerebellum. Instead, in these tissues, the
   N-terminal IPs enriched for HIPs associated with the nucleus, which was
   more prominent in the cortex than the cerebellum (Fig. [120]2E,F,
   N-Term). The nuclear annotation was missing from the N-terminal IP in
   the striatum but curiously was recovered in the Central IPs
   (Fig. [121]2D). These results suggest that the complement of HIPs may
   be highly tissue-specific and would require access to different
   epitopes for their capture by antibody-based approaches.

Figure 2. Antibody epitope target drives distinct HTT interaction signatures
between tissues of the brain.

   [122]Figure 2
   [123]Open in a new tab

   (A–C) HTT PPI (HIP) abundance ratio of N-terminus versus Central
   antibody IPs from the striatum (A), cortex (B), and cerebellum (C). PPI
   ratios of N-terminal versus Central abundance for each genotype-age
   condition were calculated, and the ratios with the greatest absolute
   value within each tissue were plotted. HIP ratios with |log[2]| >2.5
   were considered preferentially captured. P values were calculated by
   unpaired t tests (n = 3–4 per group). (D–F) Subcellular localization
   ontology enrichment analysis was performed for HIPs captured in
   N-terminal (top) or Central (bottom) IPs in the striatum (D), cortex
   (E), and cerebellum (F). (G, H) Pathway enrichment analysis for HIPs
   captured in N-terminal (G) or Central (H) IPs in the striatum, cortex,
   and cerebellum was performed using DAVID
   ([124]https://davidbioinformatics.nih.gov/). The top ten pathways were
   collated for each tissue type and the number of annotated proteins and
   enrichment significance (P value, Fisher’s Exact test) are displayed
   for each pathway. (I) The subset of HIPs identified in all three
   tissues was assembled into a STRING functional network. Nodes are
   color-coded to represent the HIP abundance age ratio (40 weeks/8 weeks)
   for each genotype (Q20 and Q140) in the striatum, cortex, and
   cerebellum to visualize age-dependent interactions. Node border color
   indicates whether the HIP was captured with the N-term (gray) or
   Central (blue) antibodies. (J) Targeted MS/MS (parallel reaction
   monitoring) quantification of CREBBP abundance in HTT IPs, as
   indicated. For each condition, the median normalized intensities (left
   axis, bars, n = 2 biological replicates) and chromatogram peak DotP
   score (right axis, line graph) for the CREBBP peptide were plotted. A
   DotP score >0.75 indicates a high-confidence assignment (orange
   circles). (K) Immunofluorescence microscopy of HTT staining (cyan) with
   epitope-targeted antibodies in BE(2)-C neuroblastoma cells. Nuclei were
   visualized with DAPI (grayscale). Scale bar = 10 μm. [125]Source data
   are available online for this figure.

   We next asked if the epitope-driven HTT PPI signatures were associated
   with distinct molecular functions. Using GO analysis to analyze PPIs
   captured by the N-terminal IPs, we observed a strong enrichment for
   terms associated with cell division, cell cycle regulation, metabolism,
   and vesicle trafficking, which were largely shared across all tissues
   (Fig. [126]2G). The Central IP condition correlated with protein
   biosynthesis, metabolism, and DNA repair for all tissue types
   (Fig. [127]2H). Interestingly, the cortex and cerebellum displayed an
   N-terminal IP enrichment for RNA metabolism, transcription, and
   processing that was missing in the striatum for this IP condition
   (Fig. [128]2G) but was present in the striatum for the Central IP
   condition (Fig. [129]2H). Our observations of tissue-dependent PPI
   epitope preferences suggest that a unique assembly of interactions