TDP-43 impairs glycolysis by sequestering hexokinase 1 in amyotrophic lateral sclerosis

TL;DR

This paper shows that TDP-43 in ALS disrupts glycolysis by trapping HK1, causing energy problems in neurons and suggesting glycolytic restoration as a treatment.

Contribution

The study reveals a novel mechanism where TDP-43 impairs glycolysis by sequestering HK1, linking metabolic dysfunction to ALS pathology.

Findings

Cytoplasmic TDP-43 reduces glycolytic capacity in patient-derived neurons by binding and sequestering HK1.

Compensating for HK1 loss improves motor performance and survival in TDP-43-associated ALS models.

TDP-43 mislocalization decreases HK1 mitochondrial association and enzymatic activity in multiple ALS models.

Abstract

Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disorder characterized by progressive motor neuron degeneration and cytoplasmic mislocalization of TDP-43. While metabolic dysfunction is increasingly recognized in ALS, the mechanistic link between impaired energy metabolism and TDP-43 pathology remains unknown. Here, we show that cytoplasmic TDP-43 directly disrupts glycolysis by targeting hexokinase 1 (HK1), the first rate-limiting enzyme of the pathway. In cells expressing a TDP-43 variant lacking its nuclear localization signal and in patient-derived iPSC motor neurons, TDP-43 accumulation in the cytoplasm reduces glycolytic capacity, indicating a neuron-intrinsic metabolic defect. Across cellular models including patient-derived neurons, TDP-43 mutant mice, and postmortem spinal cord tissue from ALS patients, we observe consistent decreases in HK1 protein level, mitochondrial association, and enzymatic activity, despite unchanged transcript levels. Mechanistically, cytoplasmic TDP-43 directly binds to HK1, disassociating it from mitochondria and promoting its sequestration into insoluble aggregates. This mislocalization impairs glycolysis and increases neuronal vulnerability. Notably, compensation for HK1 loss reduces cytoplasmic TDP-43 and ubiquitin accumulation, improves motor performance, and prolongs survival in TDP-43–associated ALS models. Together, these findings identify a previously unrecognized mechanism by which TDP-43 impairs glycolysis through HK1 misregulation and highlight glycolytic restoration as a potential therapeutic strategy in ALS.

The online version contains supplementary material available at 10.1007/s00401-026-02996-6.

Figures

a). In addition, consistent with prior reports implicating mitochondrial dysfunction in ALS [[39](#CR39), [44](#CR44)], mitochondrial respiration was also impaired in TDP-43^ΔNLS^ cells, reinforcing a global deficit in energy metabolism (Supplemental Fig. 1b).Fig. 1Pathological TDP-43 elicits glycolytic impairment in cells. a, b Agilent seahorse glycolytic stress test was performed on n = 3 independent biological repeats. Glycolysis and Glycolytic Capacity were measured according to manufacturer’s protocol. ECAR values were normalized to total protein then plotted against time. Individual glyc

a). Given the observed impairment in glycolytic function in TDP-43 mutant–expressing cell models (Fig. [1](#Fig1)), we next investigated whether key glycolytic enzymes are transcriptionally or post-transcriptionally affected. Because TDP-43 is a nuclear RNA-binding protein implicated in RNA processing [[55](#CR55)], we first examined transcript levels of major glycolytic genes in both TDP-43^WT^ and TDP-43^ΔNLS^-expressing stable cells. RT-qPCR analysis revealed no significant changes in glycolytic gene expression, suggesting that TDP-43–induced metabolic impairment may occur at the post-trans

b-c) was confirmed by the expression of the motor neuron marker Islet1/2 as previously described [[27](#CR27)]. Following a 24-h treatment with the proteasome inhibitor MG-132, a well-established method to exacerbate TDP-43 pathology in iPSC-derived neurons [[25](#CR25)], we observed a significant reduction in HK1 immunofluorescence intensity in Islet1/2+ motor neurons derived from both TDP-43^G298S^ and TDP-43^M337V^ iPSCs (Fig. [3](#Fig3)e, g and Supplemental Fig. 3b, d). Consistent with this finding, HK enzymatic activity was significantly reduced in motor neurons derived from both mutant l

a). These findings indicate the selective interaction of TDP-43 with HK1.Fig. 4TDP-43 directly interacts with HK1 and promotes HK1 dissociation from mitochondria. a Stable HEK293 cells overexpressing YFP (control) or TDP-43^ΔNLS^ were analyzed by PLA using HK1 and TDP-43 antibodies. Total puncta number were quantified in YFP+ cells (n = 3; 50 cells per biological repeat). Total puncta number in TDP-43^ΔNLS^ normalized to biological repeat control. All data were analyzed by Student’s t-test and are presented as mean ± SE. Scale bar: 20 µm. b Cortex from 2-month-old WT and TDP-43^A315T^ mice wer

a, b). To assess the impact of HK1 restoration on TDP-43 pathology, we examined cytoplasmic TDP-43 accumulation using a C-terminal–specific antibody. iPSC-derived motor neurons from both TDP-43^G298S^ and TDP-43^M337V^ lines exhibited a significantly increased ratio of cytoplasmic TDP-43 immunofluorescence intensity to total TDP-43 intensity in ChAT+ neurons compared with their respective isogenic controls (Fig. [5](#Fig5)d, e and Supplementary Fig. 5c, d). Notably, HK1 overexpression significantly reduced this cytoplasmic TDP-43 ratio in both mutant lines, while having no detectable effect in

a, b). Behavioral testing was performed at 8 and 10 weeks, coinciding with early symptomatic progression in this model (Fig. [6](#Fig6)a). Immunofluorescence confirmed successful motor cortical expression of HK1 in both WT and TDP-43^A315T^ mice (Supplemental Fig. 6c, d). HK1 overexpression significantly extended survival in TDP-43^A315T^ mice compared to eGFP-injected controls (Fig. [6](#Fig6)b), while no survival differences were observed in WT mice with or without HK1 overexpression (Supplemental Fig. 6e), indicating a favorable safety profile. Motor coordination, assessed by Rotarod perfor

Full text

Introduction

Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disease characterized by the loss of both upper and lower motor neurons (MNs), found in the cortex and spinal cord, respectively [9]. This degeneration leads to progressive muscle weakness, paralysis and ultimately death, primarily due to respiratory failure from decreased innervation to the respiratory system [28, 33, 36]. While multiple mechanisms have been proposed to explain MN vulnerability in ALS [59], mounting evidence suggests that metabolic dysfunction plays a key role in MN degeneration and disease progression [3, 8, 34, 49, 56]. Among the various metabolic alterations, mitochondrial dysfunction has emerged as a key contributor. ALS patient tissues and experimental models consistently show reduction in mitochondrial DNA content [41, 56], impaired electron transport chain activity [18, 23, 34], and decreased mitochondria density in regions with high energy demand [11]. These findings raise the possibility that alternative energy pathways, such as glycolysis, may be critically involved in maintaining neuronal function when mitochondrial respiration is impaired.

Glycolysis is a conserved cytosolic pathway that generates ATP through the breakdown of glucose [46]. Positron emission tomography (PET) using the radioactive glucose analogue ^18^fluoro-2-deoxyglucose (^18^F-FDG) has revealed reduced glucose uptake in the cortex of ALS patients [15]. Reductions in glucose uptake have also been reported in the motor cortex of TDP-43^A315T^ transgenic mice [53], and decreased glycolytic intermediates are observed in the cortex and spinal cord of SOD1 mutant ALS mouse model at symptom onset [47]. These observations suggest glycolytic impairment may contribute to ALS pathogenesis, although whether it is a downstream consequence of disease or a primary driver remains unresolved.

Genetic studies have linked numerous mutations to ALS, including in TARDBP, which encodes the RNA-binding protein TDP-43 [35, 50]. Mutations in TDP-43 lead to its nuclear depletion and cytoplasmic accumulation, pathological features observed in nearly 90% of sporadic ALS cases and many familial forms [29, 35, 45, 48]. Emerging evidence suggests TDP-43 pathology disrupts cellular metabolism, including glycolysis [53]. For instance, increasing glycolytic flux in a Drosophila model of TDP-43–related ALS delayed disease progression [32], while pharmacological activation of phosphoglycerate kinase 1 (PGK1) using terazosin improved phenotypes in TDP-43 zebrafish, mouse and cell culture models [12]. These findings implicate glycolysis in TDP-43–mediated neurodegeneration, but the molecular mechanisms remain poorly defined.

Hexokinase 1 (HK1) is the first rate-limiting enzyme of glycolysis, which catalyzes the phosphorylation of glucose to glucose-6-phosphate [24]. HK1 is typically anchored to the outer mitochondrial membrane (OMM) via interaction with the voltage-dependent anion channel (VDAC), coupling glycolysis to mitochondrial metabolism [17]. In Alzheimer’s disease (AD), reduced mitochondrial-bound HK1 in oligodendrocytes was associated with decreased glycolytic capacity, while HK1 overexpression restored metabolic function [5, 58]. In SOD1 mutant ALS model, mutant SOD1 competes with HK1 for VDAC binding [30], and a synthetic peptide mimicking HK1’s VDAC-binding domain restored mitochondrial metabolism by blocking this interaction [31]. However, whether TDP-43 disrupts HK1 function and subsequent glycolysis has not been explored.

In this study, we demonstrate that glycolytic capacity is reduced in various TDP-43-related ALS models and uncover a novel mechanism by which cytoplasmic TDP-43 directly impairs HK1 function. We show that TDP-43 mislocalization promotes HK1 disassociation from mitochondria, reduces its enzymatic activity, and drives its accumulation into insoluble aggregates. Notably, overexpression of HK1 mitigates TDP-43 pathology, improves motor performance, and prolongs survival in both patient induced pluripotent stem cell–derived motor neurons (iPSC-MNs) and transgenic mouse models. Together, these findings identify HK1 as a critical metabolic target of TDP-43 and suggest glycolytic restoration as a potential therapeutic approach in ALS.

Methods

Antibodies

The following antibodies against HK1 (C35C4) (#2024; IF-tissue 1:50, IF-cells 1:500) and PKM1/2 (C103A3) (#3190; IF-cells 1:500, WB 1:2000) were from Cell Signaling (Danvers, MA, USA). The antibodies against Actin (#A1978; WB 1:10,000), ChAT (#AB144P; IF-tissues 1:50, IF-cells 1:500), Flag (#F3165; Co-IP 1:1000, IF-cells 1:500, WB 1:2000) were from Sigma-Aldrich (St. Louis, MO, USA). The antibody against HK1 (#NBP1-51644; immunofluorescence-tissue 1:50) was purchased from Novus Biologicals (Centennial, CO, USA). The antibodies against ATPB (#17247-1-AP; WB 1:2000), cterm-TDP-43 (#12892-1-AP; Co-IP 1:300, IF-tissue 1:100, western blot 1:5000), HK1 (#19662-1-AP; Co-IP 1:300, WB 1:1000), hTDP-43 (#60019-2-Ig; IF-tissue 1:500, IF-cells 1:500), NeuN (#66836-1-Ig; IF-tissue 1:500), PFK (#55028-1-AP; IF-tissue 1:50, IF-cells 1:500, WB 1:2000), TDP-43 (#10782-2-AP; IF-tissue 1:100, western blot 1:2000) were from ProteinTech (Rosemont, IL, USA). The antibodies against Enolase (#sc-271384; WB 1:2000) and GFP (#sc-9996; WB 1:5000) were from Santa Cruz Biotechnology (Dallas, TX, USA). The antibody against ISLET1/2 (IF-Cells 1:100) (#39.4D5) was from DSHB (Iowa City, IA, USA). The following antibodies against Alexa 568, goat anti-rabbit, IgG (H + L) (#A11036; IF-Tissue 1:500, IF-Cells 1:1000), Alexa 488, goat anti-rabbit IgG IgG (H + L) (#A11034; IF-Tissue 1:500), Alexa 568, goat anti-mouse IgG (H + L) (#A11031; IF-Tissue 1:500, IF-Cells 1:1000), Alexa 568, goat anti-guinea pig IgG (H + L) (#A11075; IF-Tissue 1:500), Alexa 488, donkey anti-goat IgG (H + L) (#A11055; IF-Tissue 1:500), Alexa 568, donkey anti-Rabbit IgG (H + L) (#A10042; IF-Tissue 1:500) were purchased from Invitrogen (Waltham, MA, USA). The secondary HRP-conjugated antibodies Goat anti-mouse HRP (#31430; WB 1:5000) and goat anti-rabbit HRP (#31460; WB 1:5000) were purchased from Thermo Fisher Scientific.

Plasmids

The following constructs pcDNA3.2-YFP (addgene #84910), pcDNA3.2-TDP-43^WT^-YFP (human) (addgene #84911), pcDNA3.2-TDP-43^ΔNLS^-YFP (human) (addgene #84912), pLD-puro-Cc-TARDBP-A315T_VA Plasmid (human TDP-43 A315T) (addgene #141329), pLD-puro-Cc-TARDBP-WT_VA Plasmid (human TDP-43 WT) (addgene #141327), and shRNA control (addgene: #8453) were purchased from addgene. The following shRNA vectors HK1 (TRCN0000037656), GPI (TRCN0000049149), PFKM (TRCN0000037772), ALDOA (TRCN0000052504), TPI1 (TRCN0000289672), PGK1 (TRCN0000006180), PGAM1 (TRCN0000353685), ENO1 (TRCN0000029326), PKM (TRCN0000196588), LDHA (TRCN0000164922), and GAPDH (TRCN0000025828) were purchased from Sigma-Aldrich. Lentiviral and Adenoassociated virus vectors were designed and purchased from vector builder pAAV[Exp]SYN1 > ORF_Stuffer:EGFP (VB130822-1415euv), pAAV[Exp]-SYN1 > mHk1_GenBank L16949.1(ns):EGFP (VB240709-1621pyc), pLV[Exp]-Puro-SYN1 > ORF_Stuffer:EGFP (VB230822-1413dun), and pLV[Exp]-Puro-SYN1 > [HK1(ns)}:EGFP (VB240513-1763xeg).

Recombinant proteins

The recombinant protein for human HK1 protein (#ab85918) was purchased from Abcam (Cambridge, UK). The recombinant protein for Human TDP-43/TARDBP Protein, CF (#TD3-H5145) was purchased from ACRObiosystems (Newark, DE, USA). The TDP-43-LCD recombinant protein was previously described [26].

Hexokinase activity assay

Post-mortem patient spinal cord and cell culture total lysates were homogenized in HK activity assay buffer and HK activity was measured and recorded according to manufacturer’s protocol (abcam #ab136957). Patient spinal cord HK activity was normalized to total protein concentration. Total cell lysate concentration was tested before the assay and the same amount of protein was added for each samples for each repeat.

RNA extraction

RNA was extracted from frozen patient spinal cords, HEK293 cells and NSC-34 cells with TRIzol™ Reagent (Invitrogen: #15596026) according to manufacturer’s protocol. Reverse transcription of RNA to cDNA utilizing 1000 ng of RNA was completed according to manufacturer’s reaction setup via the ZymoScript™ RT PreMix Kit (Zymo Research: #R3012). Quantitative PCR (qPCR) was completed with PowerUp™ SYBR™ Green Master Mix (Applied Biosystems™: #A25742). Reactions completed on ALS patient spinal cord samples (Patient Demographic information: Supplemental Table 1) and human derived cell lines were completed with human specific primers (Supplementary Table 2). Reactions with mouse derived cell line (NSC-34) were completed with mouse specific primers (Supplemental Table 3). All samples completed in duplicate and normalized to RPL13 housekeeping gene [6]. All results were shown as standardized to experimental control.

Western blot

Mouse tissue and cell pellets were lysed with triton buffer (10 mM HEPES–NaOH pH 7.9, 150 mM NaCl, 1 mM EGTA, 1% Triton-X 100) and patient spinal cords were lysed with RIPA buffer (50 mM Tris pH 7.4, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% SDS, 0.5% sodium deoxycholate) including protease inhibitor cocktail (Sigma-Aldrich #P8340) and phosphatase inhibitor (Sigma-Aldrich #P5726). Samples were lysed for 30 min at 4 °C then centrifuged at 12,000 rpm for 10 min at 4 °C, separating soluble and insoluble fractions. Insoluble fractions were further lysed with urea buffer (7 M urea, 2 M thiourea, 4% CHAPS, 30 mM Tris–HCl pH 8.5) allowing aggregated proteins to dissolve in solution. Protein concentrations were measured using Bio-Rad Protein Assay Dye Reagent Concentrate (#5000006EDU). Samples were prepared using the same concentration for each experiment with a 1X concentration of laemmli sample buffer (Bio-Rad catalog #161-0747). Prepared samples were loaded onto an SDS-PAGE gel then transferred to a nitrocellulose membrane (Bio-Rad #1620112) at 100 V for 100 min. Membranes were blocked in tris-buffered saline with 0.1% Tween-20 (TBST) with 5% nonfat powdered milk for 1 h at room temperature. Membranes were incubated with primary antibodies in TBST with 5% milk overnight at 4 °C at previously specified dilutions. The next day, the membranes were washed 3 times for 10 min each in 0.1% TBST. Secondary HRP-conjugated antibodies were diluted according to previously specified dilutions in 0.1% TBST with 5% milk powder and incubated on membranes for 2 h at room temperature. Membranes were washed 3 times for 10 min in 0.1% TBST. Membranes were imaged using the Chemiluminescent Western Blot Imager (Azure 300; #AZI300-01) and signals were amplified with ECL solution (1 M Tris–HCl pH 8.5, 0.3% H_2_O_2_, 250 mM Luminol, 90 mM Coumaric acid). All experiments were completed in triplicate. Each sample was normalized to intra-well control then normalized relative to control.

Mitochondrial fractionation

Cells were cultured in a 10 cm plate then lysed with mitochondria lysis buffer (250 mM Sucrose, 20 mM Hepes–NaOH pH 7.9, 10 mM KCl, 1.5 mM MgCl_2_ hydrate, 1 mM EDTA, 1 mM EGTA) including protease inhibitor cocktail (Sigma-Aldrich #P8340) and phosphatase inhibitor (Sigma-Aldrich #P5726) for 30 min at 4 °C. Spatula used to scrape cells from surface of the plate and pipetted into Eppendorf tubes. Samples were homogenized with a 25-gauge syringe 20 times on ice. Cells were centrifuged for 10 min at 800g, maintaining a constant temperature at 4 °C. Supernatant placed into new tube and centrifuged for 20 min at 10,000g. The supernatant (cytosolic fraction) was placed in a new tube and labeled appropriately. The pellet was resuspended in 500 µL of mitochondrial lysis buffer then centrifuged again for 20 min at 10,000g. Remove supernatant and resuspend pellet in 50µL containing 1% TritonX-100. Samples sat on ice for 30 min then prepared for western blot analysis.

Co-Immunoprecipitation (Co-IP)

HEK293 cells were transiently co-transfected with YFP (control) /TDP-43^WT^/TDP-43^ΔNLS^ and Flag/HK1 plasmids for 48 h as previously described. Cell pellets were collected as previously described. 1000ug of each sample was added to each respective tube then filled to 1 mL with total lysis buffer. Specified antibodies were added to samples and incubated overnight at 4 °C while under continuous rotation. The next day, 30 µL of protein A/G PLUS-Agarose beads (Santa Cruz #2003) was added to samples and placed back at 4 °C under continuous rotation for 2 h. After incubation, samples were centrifuged at 4000 rpm for 3 min. Remove supernatant then wash pellet with lysis buffer and centrifuge again, repeat 4 times. After last wash, add 15 µL loading buffer and boil samples for 10 min at 100 °C. Samples were cooled for 10 min then centrifuged at 4000 rpm for 3 min. Each repeat contained an IgG group to validate successful interaction. Inputs for each sample were created according to western blot protocol.

500 ng of each recombinant protein were incubated together in Eppendorf tube then filled to 1 mL with interaction buffer (20 mM Tris–HCl pH 7.5, 100 mM KCl, 2 mM MgCl_2_, 0,1% TritonX-100). Specified antibodies were added to samples and incubated overnight at 4 °C while under continuous rotation. The next day, 20 µL of protein A/G PLUS-Agarose beads (Santa Cruz #2003) was added to samples and placed back at 4 °C under continuous rotation for 2 h. Samples were then processed as the cell culture model, utilizing the interaction buffer.

Cell culture

Stable cell culture

HEK293 and NSC-34 cells were maintained in Dulbecco’s Modified Eagle Medium (DMEM) (HEK293 Cytiva: #SH30243.FS, NSC-34 Sigma Aldrich: #D5796). DMEM was supplemented with 10% Fetal Bovine Serum (FBS) (Corning: #35-011-CV) and 1% antibiotic solution (100 U/mL penicillin and 100 µg/mL streptomycin) (Cytiva: #SV30010). Transfection was completed using TransIT^®^-2020 transfection reagent according to manufactures protocol (Mirus Bio: #MIR5406). HEK293 were stably transfected with YFP (control) (addgene #84910), YFP tagged TDP-43^WT^ (addgene: #84911), or YFP tagged TDP-43^ΔNLS^ (addgene: #84912) plasmids and selected for at least 2 weeks using 200 µg/mL of G418 selection reagent (Enzo: #ALX-380-013). NSC-34 cells were transfected with flag (control) (addgene: #52535), flag tagged TDP-43^WT^ (addgene: #141327), or flag tagged TDP-43^A315T^ (addgene: #141329) plasmids and were selected for at least 2 weeks using (2 µg/mL) of puromycin.

Transient transfection

HEK293 cells were transiently co-overexpressed with either YFP (control)/TDP-43^WT^/TDP-43^ΔNLS^ and Flag/Flag-tagged HK1 plasmids with TransIT®-2020 according to manufacturer’s protocol. Media was changed 24 h after transfection. Total lysates were lysed and collected as previously depicted.

Cell viability (MTT assay)

Stable YFP (control) /TDP-43^WT^/TDP-43^ΔNLS^ were transiently transfected with previously mentioned control and various glycolysis shRNA and successful silencing of targets were confirmed via qPCR analysis. Using Cell Proliferation Kit 1 (MTT) (Roche: 11465007001), cell viability was determine according to manufacturer’s protocol. Output values were collected from Molecular Devices SpectraMax 340. Results were normalized YFP transfected control cells with control shRNA. Successful knockdown confirmed with RT-qPCR.

Immunocytochemistry

Cell culture

Coverslips were coated with Poly-D-Lysine (MiliporeSigma: #p6407) overnight (O/N) at 4 °C in a 24-well plate. Next day, coverslips were coated with laminin (Invitrogen: #23017-015) diluted in PBS and incubated in the 37 °C CO_2_ incubator for 1 h then washed with PBS. Cells were counted and plated according to experimental needs. After cellular experimental procedures were completed, cells were fixed in 4% paraformaldehyde (4% PFA) for 20 min at room temperature (RT). Cellular permeabilization was completed for 5 min with 0.1% PBST (phosphate buffered saline with 0.1% Triton-X [Fisher Scientific: #BP151-500]). Cells were blocked for 1 h at RT in blocking solution (0.05%PBST with 2% normal goat serum [NGS]). Primary antibodies were diluted in blocking solution then placed on cells O/N at 4 °C. The next day, cells were incubated with secondary antibodies diluted in blocking serum for 2 h at RT. Coverslips were mounted with mounting media (Dako: #S3025) onto slides (Fisher Scientific: #125442).

Frozen mouse tissue

Frozen mouse tissue was fixed with ice cold acetone for 2 min then dried at room temperature for 1 h. Slides were washed three times with 1X PBS. After wash, antigen retrieval was performed with 0.01 M sodium citrate buffer with 0.05% Tween20 at pH 6.0. Samples were microwaved in sodium citrate buffer for 15 min then cooled at room temperature for 1 h. After cooling, slides were blocked for 1 h at room temperature in humid chamber (10% NGS in 0.3% PBST). Slides were incubated overnight with primary antibodies (3% NGS in 0.3% PBST) at 4 °C. The following day, slides were incubated with secondary antibodies diluted in blocking serum for 2 h at RT. Coverslip was mounted in slides with mounting media.

Seahorse

HEK293 cells were counted and 20,000 cells were plated 2 days before experiment. iPSC-MNs were plated in seahorse plates on day 15 of differentiation and differentiated until day 27, then, if applicable, treated with MG-132 for 24 h. All cells were plated in triplicate and normalized to total protein concentration. The average of the three wells was normalized to the control of each individual experiment.

Using the Agilent XF HS Mini Analyzer (Agilent #S7852A), Seahorse XFp Glycolysis Stress Test (Agilent #103017) and Seahorse XFp Cell Mito Stress Test (Agilent #103010) kits were completed. The glycolysis stress test measures the extracellular acidification rate (ECAR) after glucose (10 mM/well), oligomycin (2 μM/well), and 2-deoxyglucose (2-DG) (50 mM/well) injections. After injections, results can be interpreted as basal glycolysis (maximum reading before oligomycin minus last measurement before glucose injection) and glycolytic capacity (maximum reading after oligomycin minus last measurement before glucose injection. The mito stress test measures the oxygen consumption rate (OCR) after oligomycin (2uM/well), Carbonyl cyanide-4 (trifluoromethoxy) phenylhydrazone (FCCP) (2uM/well), and rotenone/antimycin A (rot/AA) (0.5uM/well) injections. After injections, results can be interpreted as basal respiration (Last reading before oligomycin minus last reading after rot/AA), ATP-linked (maximum reading before FCCP minus last reading after rot/AA), and maximum respiration (maximum after FCCP minus last after rot/AA).

Proximity ligation assay (PLA)

Stable HEK293 cells were plated on 35 mm Dish with a precoated with Poly-d-Lysine No. 1.5 Coverslip (Mattek #P35GC). Frozen mouse cortex samples were prepared for immunostaining as previously described. Experiment was completed according to manufacturer’s protocol utilizing mouse and rabbit primary antibodies and Duolink^®^ PLA Duolink Reagnets Red (Sigma-Aldrich: #DUO92002, #DUO92004, #DUO92008, #DUO82049).

Immunohistochemistry (IHC)

Frozen mouse cortex tissue were fixed with ice cold acetone for 2 min then dried at room temperature for 1 h. Slides were washed 3 times with 1X PBS. After wash, antigen retrieval was performed with 0.01 M sodium citrate buffer with 0.05% Tween20 at pH 6.0. Samples were microwaved in sodium citrate buffer for 15 min then cooled at room temperature for 1 h. After cooling, slides were blocked for 1 h at room temperature in humid chamber (10% NGS in 0.3% PBST). Slides were incubated overnight with primary antibodies (3% NGS in 0.3% PBST) at 4 °C. The following day, the samples were washed with 0.1%PBST 3 times then incubated in a 3% H_2_O_2_ solution diluted in methanol. Utilizing the Vectastain^®^ Elite^®^ ABS Universal Kit (Vector Laboratories: #PK-7200), slides were incubated for 60 min at room temperature with the Biotinylated Universal Antibody then washed with 0.1%PBST. Subsequently, the samples were incubated with 30 min at room temperature with the R.T.U vectastain Elite ABC reagent then washed with 0.1%PBST. 3,3′-Diaminobenzidine (DAB) staining was completed using impact^®^ DAB Substrate Kit (Vector Laboratories: #SK-4105). One drop (~ 30 µL) of DAB regent was diluted in 10 mL impact DAB diluent. Slides were incubated then immediately placed in running water for 5 min. Slides were then treated with nuclear Hematoxylin QS Counterstain (Vector Laboratories: #H-3404). Slides were incubated then immediately placed in running water for 5 min. Lastly, slides were dehydrated in increasing increments of ethanol for 1 min each (50%, 70%, 95%, 100%), then in xylene 3 times for 3 min each. Slides were mounted with a xylene based mounting medium (Electron Microscopy Sciences: #15322) and let dry overnight at room temperature.

iPSC differentiation into motor neurons

iPSCs harboring the TDP-43^G298S^ mutation and matching isogenic control were differentiated according to a modified version of a previously published protocol [13]. The TDP-43^M337V^ and isogenic control line are a genetically modified iPSC line purchased from Jackson Laboratory (Isogenic Control: JIPSC001110; TDP43^M337V^:JIPSC001106) [37]. iPSCs were plated in a vitronectin coated (ThermoFisher: #A14700) 6-well plate and cultured in mTeSR medium (StemCell #100-0276) and changed every other day for up to one week. From days 1–5 neuronal induction was performed via Dual SMAD inhibition. Cells were incubated in Basal Media, consisting of DMEM/F12 50/50 Media (Corning: #10-092-CV), B27 (ThermoFisher: #12587010), N2 (ThermoFisher: #17502048), 100U/mL penicillin and 100 µg/mL streptomycin (Cytiva: #SV30010), and ascorbic acid (MedChemExpress: #HY-B0166) supplemented with dosomorphin (1 µM) (Tocris: #3093), SB 431542 (10 µM) (Tocris: #1614), CHIR 99021 (3 µM) (Tocris: #4423), and Rock Inhibitor (RI) (5 µM) (Tocris: #1254). Day 6 of the protocol began SHH activation with the addition of Retinoic Acid (RA) (1.5 µM) (Millipore Sigma: #R2625), SAG (200 nM) (EMD Millipore: #566660), and an increase in RI (10 µM). Day 7–14, cells were cultured in the same media as day 6, except RI was returned to its original concentration, 5 µM. A new 6-well plate was coated with Poly-l-Ornithine (1:400) (Millipore Sigma: #P3655) O/N at 4 °C, then treated with laminin (1:240) (Millipore Sigma: #L2020) for 4 h in the 37 °C incubator. Cells were split on day 15 with a 1:1 solution of accutase (Innovative Cell Technologies: #AT104) and accumax (Innovative Cell Technologies: #AM105), cells were placed in the incubator and checked every 5 min until they were no longer adhesive. Cells were centrifuged at 200 g for 2 min then resuspended in day 15 media; basal media supplemented with BDNF (2 ng/mL) (PeproTech: #450-02), CNTF (2 ng/mL) (PeproTech: #450-13), GDNF (2 ng/mL) (PeproTech: #450-10), RA (1.5 µM), SAG (200 nM), RI (10 µM). Days 16–21 media was the same as day 15 only with 5 µM RI. Days 22–24 media consisted of BDNF, CNTF, GDNF (at the same previous concentrations) along with DAPT (2 µM) (Tocris: #2634), RI (2 µM), and Culture 1 (1:100) (ThermoFisher: #A3320201). From day 25+ DAPT was removed from media and maintained in day 25+ media. Cell media was changed every other day. MG-132 (1 µM) (Cayman Chemical: #10012628) treatment was completed 24 h before cell collection (day 26). All iPSC-MNs were collected by day 27.

Lentiviral transduction in ipsc induced motor neurons

Lentiviral vectors were overexpressed in iPSCs on day 20 of differentiation protocol. Positive fluorescence signal was exhibited beginning on day 25, however, lentiviral transduction was completed until day 27.

Imaging systems

All immunofluorescence imaging was completed using The Olympus Fluoview 3000 (#FV3000). Immunohistochemical staining was imaged using All in one imager (Keyence #BZ-700). Imaging analyses were completed using ImageJ software.

Animal model of ALS

TDP-43^A315T^ mice (JAX stock #010700) were purchased from Jackson laboratory. These mice exhibit the ALS phenotype as early as 2 months of age as previously described [54]. Mice were bred on C57BL/6J (stock No. 000664) background.

All animal experiments were conducted in accordance with protocols approved by the Institutional Animal Care and Use Committee of Case Western Reserve University and were performed based on the National Institutes of Health Guide for the Care and Use of Laboratory Animals. Sufficient procedures were employed for reduction of pain or discomfort of subjects during the experiments.

AAV injection

One-month-old TDP-43^A315T^ male mice were anesthetized with inhaled isoflurane and mounted on a stereotactic device. Stereotaxic surgery was performed using a model 1900 stereotax (Kopf), as previously described [43]. Briefly, a small craniotomy was made using a 33-gauge drill bit above the desired coordinate. A small pulled glass pipette containing AAV was attached to a Nanoject II (Drummond) and inserted to the appropriate depth. Injections were performed at a rate of 90 nL/min. The coordinates used for bilateral injections were anteroposterior, 1.5 mm from bregma; mediolateral, ± 1.5 mm; dorsoventral, − 0.8 mm. AAV5-EV and AAV5-HK1 were injected into the primary motor cortex (MOp) of TDP-43^A315T^ mice and wild-type littermates at the age of one month for HK1 overexpression. The concentration (ddTiter) of the AAVs was 1.0e12 GC/mL. One microliter was injected into each hemisphere of the mice. Consequently, the number of viral particles of AAVs was 1.0e9. Mice were sacrificed at 2.5 months of age and cortex were dissected for immunohistochemistry and biochemical analyses.

Behavioral test

All behavioral analyses were conducted by an experimenter who was blinded to the genotypes and treatment groups. All mice were subjected to a series of behavioral measurements to monitor balance and motor coordination (Rotarod test) and muscular strength (Grip test). Body weights were recorded throughout the study period.

Rotarod

Locomotor activity was measured on mice 4 and 6 weeks after injection. Latency to fall (seconds) was measured according to when mice fell off the constantly accelerating rod over a 300 s timeframe. Mice were trained three times per day for three days. On test day, mice were placed on rod and the trial ended when mice either fell off the rod or remained stable on the rod for 300 s. Test day procedure was repeated for three trials. Final value was calculated as the average of the three repeats.

Grip strength

Mouse grip strength was measured utilizing a grip strength meter (Bioseb #Bio-GT3). Mice were placed on a metal grid allowing for all limbs to grasp metal grid and pulled by the tail. Once the mice were fully removed from the metal grid, the device gave a total strength reading as generated by force. This procedure was completed three times per mouse with 30 s rest in between each trial.

Tissue collection

For Immunofluorescence analysis, mice were transcardially perfused with 4% PFA. Whole mouse brain samples were collected and placed in 4% PFA for 24 h at RT. 4% PFA was then replaced with 1% PFA for 72 h at 4 °C. The samples were then placed in in a 30% sucrose solution (w/v) in 1XPBS for 72 h at 4 °C. OCT solution was used to embed tissue for sectioning.

For total tissue collection, mice were transcardially perfused with 1XPBS. Cortex samples were flash frozen in liquid nitrogen then placed in -80 °C freezer for long term storage.

Quantification and statistical analysis

A previously published study depicts a power analysis for animal number. Each group consisted of at least 14 animals per group for behavioral tests and n = 3–6 mice for biochemical analyses. For animal studies, randomization and blinded evaluations were performed. Each cell culture study was completed independently at least three individual times. All quantifications for imaging studies was completed by a blinded observer. All image analysis and quantification was performed using ImageJ software. All statistical analyses were completed using GraphPad Prism 8.0. Unpaired student’s t test was used to compare two groups. One way Anova was used to compare three or more groups, followed by Tukey’s post hoc test. All values are reported as mean ± SE). We considered p < 0.05 as statistically significant.

Results

Pathological TDP-43 impairs glycolytic function and reduces cell viability

To assess the impact of pathological TDP-43 on glycolysis, we performed Seahorse glycolytic stress tests in stable HEK293 cell lines expressing either YFP (control), YFP-tagged TDP-43 wild-type (TDP-43^WT^), or YFP-tagged TDP-43 harboring mutations in the nuclear localization signal (TDP-43^ΔNLS^), promoting accumulation in the cytoplasm and mimics disease-relevant mislocalization [20]. Compared to YFP-expressing controls, TDP-43^ΔNLS^-expressing cells exhibited a significant reduction in both basal glycolysis and glycolytic capacity (Fig. 1a). Note that YFP control expression did not affect glycolytic function (Supplemental Fig. 1a). In addition, consistent with prior reports implicating mitochondrial dysfunction in ALS [39, 44], mitochondrial respiration was also impaired in TDP-43^ΔNLS^ cells, reinforcing a global deficit in energy metabolism (Supplemental Fig. 1b).Fig. 1. Pathological TDP-43 elicits glycolytic impairment in cells. a, b Agilent seahorse glycolytic stress test was performed on n = 3 independent biological repeats. Glycolysis and Glycolytic Capacity were measured according to manufacturer’s protocol. ECAR values were normalized to total protein then plotted against time. Individual glycolysis and glycolytic capacity values were normalized to total protein level then to individual control from each biological repeat. a Stable HEK293 cells were overexpressed with either YFP (control), TDP-43^WT^ or TDP-43^ΔNLS^. Data were analyzed by one-way ANOVA followed by post hoc Tukey’s test. b TDP-43^G298S^ patient iPSC and control iPSC were differentiated into motor neurons with our established protocol. On day 26 following differentiation, cells were treated with MG-132 (1 μM) for 24 h. Data were analyzed by Student’s t-test. c YFP (control), TDP-43^WT^ or TDP-43^ΔNLS^ stable cells were transiently transfected with various glycolysis shRNAs. Cell viability was measured via MTT assay. Decreased cell viability in all TDP-43^WT^ and TDP-43^ΔNLS^ with various glycolysis shRNA (n = 3). Data were analyzed by one-way ANOVA followed by post hoc Tukey’s test. All data are mean ± SE

To examine the disease relevance of these findings, we assessed glycolytic function in induced iPSC-MNs derived from an ALS patient harboring the TDP-43^G298S^, and the human KOLF2.1 J iPSCs with TDP-43^M337V^ [37]. Following treatment with MG-132, a proteasome inhibitor that enhances TDP-43 aggregation and cytoplasmic accumulation [25], both TDP-43^G298S^ and TDP-43^M337V^ iPSC-MNs exhibited marked reductions in glycolysis and glycolytic capacity compared to isogenic controls (Fig. 1b and Supplementary Fig. 1c). No differences were observed under vehicle-treated conditions (Supplemental Fig. 1d), indicating that TDP-43 pathology, rather than genotype alone, drives glycolytic dysfunction.

To determine whether glycolytic impairment contributes to cellular vulnerability, we performed shRNA-mediated knockdown of individual glycolytic enzymes in TDP-43^WT^- and TDP-43^ΔNLS^-expressing HEK293 cells and assessed viability via MTT assay. Both TDP-43^WT^ and TDP-43^ΔNLS^ expression reduced viability compared to YFP (control). Notably, further suppression of glycolysis by knocking down individual glycolytic enzymes exacerbated cell death across all conditions (Fig. 1c). Knockdown efficiency for each enzyme was validated by RT-qPCR (Supplemental Fig. 1e). Together, these results suggest that pathological TDP-43 impairs glycolytic metabolism and that reduced glycolytic capacity increases susceptibility to TDP-43–induced cellular toxicity.

Cytoplasmic TDP-43 reduces HK1 protein levels and promotes its mitochondrial disassociation

We quantified nuclear and cytoplasmic distribution of YFP (control), YFP-TDP-43^WT^, and YFP-TDP-43^ΔNLS^ constructs, and validated predominantly nuclear localization of TDP-43^WT^ and robust cytoplasmic localization of TDP-43^ΔNLS^ (Supplementary Fig. 2a). Given the observed impairment in glycolytic function in TDP-43 mutant–expressing cell models (Fig. 1), we next investigated whether key glycolytic enzymes are transcriptionally or post-transcriptionally affected. Because TDP-43 is a nuclear RNA-binding protein implicated in RNA processing [55], we first examined transcript levels of major glycolytic genes in both TDP-43^WT^ and TDP-43^ΔNLS^-expressing stable cells. RT-qPCR analysis revealed no significant changes in glycolytic gene expression, suggesting that TDP-43–induced metabolic impairment may occur at the post-transcriptional level (Supplemental Fig. 2b). To assess this, we evaluated protein levels of three key rate-limiting glycolytic enzymes: HK1, phosphofructokinase (PFK), and pyruvate kinase (PKM). Immunoblotting revealed a slight reduction in HK1 protein in TDP-43^WT^ and more prominently in TDP-43^ΔNLS^-expressing cells, whereas PFK and PKM levels remained unchanged (Fig. 2a–c). Further, immunofluorescence analyses were performed to assess protein changes specifically within transfected cells. Quantification restricted to YFP-positive cells confirmed reduced HK1 intensity in TDP-43^ΔNLS^–expressing cells, with no significant change observed in TDP-43^WT^–expressing cells (Fig. 2d). There were no observable differences in PFK or PKM (Supplemental Fig. 2c-d), indicating a selective downregulation of HK1 at the protein level.Fig. 2HK1 was selectively decreased in TDP-43-related ALS models. Total lysates of stable HEK293 cells expressing YFP (control), TDP-43^WT^ or TDP-43^ΔNLS^ were harvested and subjected to western blot analysis. a–c Total protein levels of the specified proteins were measured and normalized to the intra-well control, Actin. Biological repeats (n = 3) were normalized to individual controls. Western blot analysis was utilized to determine changes in; a HK1 total protein b PFK total protein c PKM total protein. d Stable HEK293 cells expressing YFP (control), TDP43^WT^, TDP-43^ΔNLS^ were stained with anti-HK1 antibody and HK1 immunodensity in YFP+ cells was measured. HK1 immunodensity was quantified by drawing regions of interest (ROIs) around YFP-positive cells. A fixed threshold was applied across all groups, and integrated fluorescence intensity was measured for each individual cell. HK1 immunodensity in TDP-43^ΔNLS^ was normalized to YFP control cells. Total n = 3 biological repeats with at least 50 cells/repeat). Data were analyzed by Student’s t-test. Scale bar: 10 µm. e Cellular fractionation of TDP-43 stable cells was conducted. HK1 in cytosolic and mitochondrial fractions were analyzed by western blot and probed for total HK1 protein level. Mitochondrial HK1 protein level was normalized to ATPB loading control. Biological repeats (n = 3) were normalized to individual control samples. Data were analyzed by Student’s t-test. f HK enzymatic activity was measured in YFP (control), TDP-43^WT^ or TDP-43^ΔNLS^ stable HEK293 cells. Enzymatic activity was normalized to total protein concentration. Biological repeats (n = 3) were normalized to individual control samples. Data were analyzed by one-way ANOVA followed by post hoc Tukey’s test. g Total cortical brain lysates of TDP-43^A315T^ and wildtype mice (2 months) collected and HK1 total protein level analyzed via western blot analysis. HK1 total protein was normalized to loading control Actin. Biological repeats (n = 3) were normalized to individual control samples. Data were analyzed by Student’s t-test. h Brain sections of TDP-43^A315T^ and wildtype mice (2 months) were stained with HK1 antibody. HK1 immunodensity in NeuN+ cells was quantitated. HK1 immunodensity was quantified by drawing ROIs around NeuN-positive cells. A fixed threshold was applied across all groups, and integrated fluorescence intensity was measured for each individual cell. Biological repeats (n = 3–4) were normalized to individual control samples. Data were analyzed by Student’s t-test. All data are mean ± SE. Scale bar: 20 µm

HK1 predominantly localizes to OMM, where its association with VDAC facilitates coupling between glycolysis and mitochondrial metabolism [38]. Previous studies reported that HK1 disassociation from the mitochondria suppresses its enzymatic activity and initiates the inhibition of glycolysis [58]. We therefore investigated whether pathological TDP-43 alters HK1 subcellular distribution. Mitochondrial fractionation showed a significant reduction in mitochondrial-bound HK1 in TDP-43^ΔNLS^-expressing cells compared to controls (Fig. 2e). Immunofluorescence co-staining for HK1 and the OMM protein, TOM20, revealed reduced colocalization in TDP-43^ΔNLS^–expressing cells, further supporting decreased mitochondrial association of HK1 in the presence of cytoplasmic TDP-43 (Supplementary Fig. 2e). Consistent with these findings, HK enzymatic activity was significantly reduced in YFP-TDP-43^ΔNLS^–expressing cells (Fig. 2f). Together, these data indicate that cytoplasmic TDP-43 is associated with reduced HK1 abundance and mitochondrial association, accompanied by impaired HK enzymatic activity. However, the present data do not distinguish whether this reflects a consequence of mitochondrial dysfunction or a more direct effect of cytoplasmic TDP-43 on HK1.

To assess whether similar HK1 dysregulation occurs in vivo, we analyzed the TDP-43^A315T^ transgenic mouse model, which expresses a familial ALS-linked mutant form of TDP-43 and recapitulates hallmark features of the disease [21, 54]. Immunoblotting of motor cortex tissue from symptomatic TDP-43^A315T^ mice showed a significant reduction in total HK1 protein levels relative to wild-type littermates (Fig. 2g). Furthermore, TDP-43^A315T^ mice exhibited a pronounced decrease in HK1 immunodensity within NeuN+ neurons compared to wild-type controls (Fig. 2h), indicating that HK1 loss occurs within neurons in vivo. Motor neurons in the spinal cord are also directly affected by TDP-43 pathology; accordingly, immunofluorescence staining of spinal cord sections from TDP-43^A315T^ mice revealed a similar reduction in HK1 immunodensity in choline acetyltransferase (ChAT)+ motor neurons compared with wild-type controls (Supplementary Fig. 2f).

To further validate these findings, we transfected the motor neuron–like NSC-34 cell line with Flag, Flag-TDP-43^WT^, or Flag-TDP-43^A315T^ constructs. TDP-43^A315T^ did not affect mRNA, however, overexpression significantly reduced HK1 protein levels (Supplemental Fig. 2g-h). In contrast, PFK and PKM protein levels remained unchanged (Supplemental Fig. 2i-j). Mitochondrial fractionation further confirmed reduced mitochondrial HK1 in cells expressing TDP-43^A315T^, while no change observed in TDP-43^WT^-expressing cells (Supplementary Fig. 2k).

Collectively, these results demonstrate that pathological or cytoplasmic TDP-43 reduces HK1 protein abundance and disrupts its mitochondrial localization, contributing to impaired glycolytic function in both cellular and in vivo models of ALS.

Loss of HK1 in ALS patient spinal cord and iPSC-derived motor neurons

To assess the translational relevance of our findings, we examined postmortem spinal cord tissue from individuals with ALS and non-ALS controls (see Supplemental Table 1). In line with our observations in cellular and mouse models, RT-qPCR analysis showed no significant differences in glycolytic gene transcripts between ALS patients and control subjects’ spinal cords (Supplemental Fig. 3a), consistent with our above data that HK1 dysregulation occurs at the post-transcriptional level. Immunoblotting of spinal cord lysates revealed a marked reduction in HK1 protein abundance in ALS patients compared to controls (Fig. 3a). Immunofluorescence staining further confirmed a pronounced decrease in HK1 immunodensity within ChAT+ motor neurons in ALS spinal cords (Fig. 3b, c). Notably, total HK enzymatic activity was significantly reduced in ALS spinal cord lysates (Fig. 3d), supporting the notion that HK1 protein loss translates into a functional glycolytic deficit in patient tissue.Fig. 3. Decreased HK1 in motor neurons of ALS patients. Demographic information for ALS patients is provided in Supplemental Table 1. a Western blot analysis of total lysates from postmortem ALS patient spinal cords performed for total HK1 protein levels. HK1 protein levels were normalized to intra-well control Actin. All results were normalized to the average of the control patients (n = 4–5). b, c HK1 immunodensity from postmortem ALS patient spinal cord samples. Quantification of HK1 immunodensity in ChAT+ spinal motor neurons from ALS patients normalized to individual control patients (n = 4–5). Immunodensity was collected utilizing ROI that were drawn around ChAT+ cells, threshold value was set and maintained across all groups, and integrated density was recorded for each individual cell. Scale bar: 20 µm. d HK enzymatic activity measured in total spinal cord lysates from ALS patients Enzymatic activity was normalized to total protein concentration. Relative enzyme activity was normalized to the average of the control patients (n = 5). e TDP-43^G298S^ patient iPSCs and control iPSCs differentiated into motor neurons. On day 26 following differentiation, cells were treated with MG-132 (1 μM) for 24 h and collected on day 27. Scale bar: 20 µm. f Differentation efficiency calculated by dividing total Islet1/2+ cells by the total number of nuclei (DAPI). g HK1 immunodensity in Islet1/2+ cells was imaged and quantified (n = 3, ≥ 100 cells per group). Immunodensity was collected utilizing ROI that were drawn around Islet1/2+ cells, threshold value was set and maintained across all groups, and integrated density was recorded for each individual cell. Histograms for iPSC immunostaining experiments are shown beneath immunofluorescence images. All data were analyzed by Student’s t-test and are presented as mean ± SE. h HK enzymatic activity measured in total lysates of TDP-43^G298S^ and control cells. Enzymatic activity was normalized to total protein concentration. Relative enzyme activity was normalized to the isogenic control (n = 3)

To determine whether HK1 dysregulation also occurs in ALS patient–derived iPSC-MNs, we differentiated TDP-43^G298S^ and TDP-43^M337V^ iPSCs into motor neurons. Successful differentiation (Fig. 3e, f and Supplementary Fig. 3b-c) was confirmed by the expression of the motor neuron marker Islet1/2 as previously described [27]. Following a 24-h treatment with the proteasome inhibitor MG-132, a well-established method to exacerbate TDP-43 pathology in iPSC-derived neurons [25], we observed a significant reduction in HK1 immunofluorescence intensity in Islet1/2+ motor neurons derived from both TDP-43^G298S^ and TDP-43^M337V^ iPSCs (Fig. 3e, g and Supplemental Fig. 3b, d). Consistent with this finding, HK enzymatic activity was significantly reduced in motor neurons derived from both mutant lines (Fig. 3h and Supplementary Fig. 3e). Together, these results demonstrate that HK1 protein levels and enzymatic activity are diminished in both ALS patient spinal cord tissue and patient iPSC-MNs. These findings validate and extend our earlier observations in experimental models, indicating HK1 loss as a conserved and disease-relevant feature of TDP-43–associated motor neuron pathology.

Cytoplasmic TDP-43 binds to HK1 and promotes HK1 mitochondrial dissociation

Cytoplasmic mislocalization of TDP-43 is a defining pathological hallmark of ALS [45], yet its impact on downstream metabolic enzymes remains unclear. To investigate whether cytoplasmic TDP-43 alters the localization or solubility of HK1, we first assessed their spatial proximity using a PLA. In TDP-43^ΔNLS^-expressing cells, which accumulate TDP-43 in the cytoplasm [20], we observed a significant increase in PLA puncta between TDP-43 and HK1 compared to control cells, indicating a close spatial association (< 40 nm) (Fig. 4a). To confirm these findings in vivo, we performed PLA in cortex sections from TDP-43^A315T^ transgenic mice and detected elevated TDP-43–HK1 proximity compared to wild-type controls (Fig. 4b), consistent with the cytoplasmic accumulation of TDP-43 in this model [22]. In contrast, no PLA-positive signal was detected between TDP-43 and PFK, another rate-limiting glycolytic enzyme, in the brains of TDP-43^A315T^ mice (Supplemental Fig. 4a). These findings indicate the selective interaction of TDP-43 with HK1.Fig. 4TDP-43 directly interacts with HK1 and promotes HK1 dissociation from mitochondria. a Stable HEK293 cells overexpressing YFP (control) or TDP-43^ΔNLS^ were analyzed by PLA using HK1 and TDP-43 antibodies. Total puncta number were quantified in YFP+ cells (n = 3; 50 cells per biological repeat). Total puncta number in TDP-43^ΔNLS^ normalized to biological repeat control. All data were analyzed by Student’s t-test and are presented as mean ± SE. Scale bar: 20 µm. b Cortex from 2-month-old WT and TDP-43^A315T^ mice were analyzed by PLA using HK1 and TDP-43 antibodies. Total puncta number were quantified and normalized to biological control (n = 3 animals). All data were analyzed by Student’s t-test and are presented as mean ± SE. Scale bar: 20 µm. c HEK293 cells were co-overexpressed with Flag or Flag-HK1 and YFP (control) or TDP-43^ΔNLS^. Co-IP with Flag antibody performed then followed by western blot of total lysates with TDP-43 antibody. TDP-43^ΔNLS^ and TDP-43^WT^ are seen at 70KD due to the addition of the YFP tag. Representative blot from 3 independent experiments are shown. d HEK293 cells were co-overexpressed with Flag or Flag-HK1 and YFP (control) or TDP-43^ΔNLS^. Subcellular compartmentalization performed and Co-IP with Flag antibody performed then followed by western blot of total lysates with TDP-43 antibody. TDP-43^ΔNLS^ and TDP-43^WT^ are seen at 70KD due to the addition of the YFP tag. Representative blots from 3 independent experiments are shown. e Co-IP of recombinant TDP-43 LCD and HK1 proteins. The isolated LCD of TDP-43 is known to present around 15KD. Representative blots from 3 independent experiments are shown. and Co-IP with C-terminal TDP43 antibody performed then followed by western blot with HK1 antibody. f Insoluble fractions from HEK293 cells overexpressing YFP (control), TDP-43^WT^, or TDP-43^ΔNLS^ were isolated with urea buffer. Insoluble fractions were analyzed via western blot analysis for HK1, pTDP-43 and cleaved TDP43. pTDP-43 presents at 70KD (TDP-43 with the addition of the YFP tag), cleaved TDP-43 is known to be at 25 and 35 KD, with the addition of the YFP tag, cleaved TDP-43 presents at 50KD and 65KD, respectfully. Total protein levels were quantified as relative to biological control (n = 3). All data represent mean ± SE

We next asked whether TDP-43 physically interacts with HK1. Co-IP in HEK293 cells co-expressing Flag-tagged HK1 and YFP-tagged TDP-43 variants revealed a strong interaction between HK1 and TDP-43^ΔNLS^, but not with YFP (control) or TDP-43^WT^ (Fig. 4c). Subcellular fractionation localized this interaction to the cytoplasmic compartment (Fig. 4d), consistent with the loss of mitochondrial-bound HK1 in cells expressing mutant TDP-43. To further support a direct interaction, AlphaFold2 structural modeling predicted a plausible binding interface between the low-complexity domain (LCD) of TDP-43 and HK1 (Supplemental Fig. 4b, c). This prediction was experimentally confirmed in vitro: both full-length TDP-43 and its LCD fragment directly bound to HK1 in pull-down assays using purified recombinant proteins (Fig. 4e; Supplemental Fig. 4d), providing biochemical evidence for a direct TDP-43–HK1 interaction.

Given these findings, we investigated whether TDP-43 binding affects the solubility of HK1. Immunoblotting of the detergent-insoluble fraction revealed a pronounced accumulation of HK1 in the insoluble fraction of TDP-43^ΔNLS^-expressing cells but no significant change in TDP-43^WT^ expressing cells, compared to YFP (control) (Fig. 4f). This was accompanied by elevated levels of cleaved and phosphorylated TDP-43 (pTDP-43), hallmark features of pathological TDP-43 aggregates [40]. These results indicate that cytoplasmic TDP-43 promotes the sequestration of HK1 into insoluble aggregates.

Together, our results support a model that cytoplasmic TDP-43 binds to HK1 and disassociates it from the mitochondria and that TDP-43/HK1 cytoplasmic interaction further promotes HK1 insolubility. This aberrant redistribution of HK1 likely contributes to the observed glycolytic impairment and neuronal vulnerability in TDP-43–associated ALS.

Overexpression of HK1 mitigates TDP-43 pathology and reduces ALS-associated phenotypes

Given the observed reduction in HK1 protein across multiple TDP-43–associated ALS models, we next investigated whether restoring HK1 expression could ameliorate disease-relevant phenotypes. Using a lentiviral approach, we overexpressed HK1 in iPSC-MNs harboring either TDP-43^G298S^ or TDP-43^M337V^ mutations (Fig. 5a). Immunostaining confirmed HK1 overexpression in Islet 1/2+ motor neurons derived from both mutant iPSC lines (Fig. 5b, c and Supplementary Fig. 5a, b). To assess the impact of HK1 restoration on TDP-43 pathology, we examined cytoplasmic TDP-43 accumulation using a C-terminal–specific antibody. iPSC-derived motor neurons from both TDP-43^G298S^ and TDP-43^M337V^ lines exhibited a significantly increased ratio of cytoplasmic TDP-43 immunofluorescence intensity to total TDP-43 intensity in ChAT+ neurons compared with their respective isogenic controls (Fig. 5d, e and Supplementary Fig. 5c, d). Notably, HK1 overexpression significantly reduced this cytoplasmic TDP-43 ratio in both mutant lines, while having no detectable effect in control cells (Fig. 5d, e and Supplementary Fig. 5c, d). Consistently, co-expression of HK1 in TDP-43^ΔNLS^-expressing cells reduced the levels of cleaved TDP-43 (Supplemental Fig. 5g), supporting a role for HK1 in mitigating TDP-43 proteinopathy.Fig. 5HK1 overexpression alleviates TDP-43 iPSC-derived motor neuron pathology. a Schematic of iPSC-MN differentiation and lentiviral transduction. TDP-43^G298S^ and TDP-43^M337V^ patient iPSCs and control iPSCs were differentiated into motor neurons. On day 20 of differentiation, cells were transduced with either GFP control or GFP-tagged HK1 lentiviral vectors. On day 26 following differentiation, cells were treated with MG-132 (1 μM) for 24 h. b, c Confirmation of HK1 overexpression in TDP-43^G298S^ iPSC-MNs. b Representative images of TDP-43^G298S^- and control iPSC-MNs stained with HK1 and Islet1/2 antibodies. Scale bar: 20 µm. c HK1 immunodensity was quantified by drawing ROIs around Islet1/2+ cells. A fixed threshold was applied across all groups, and integrated fluorescence intensity was measured for each individual cell. Data was quantified from 3 independent experiments. Data were analyzed by one-way ANOVA followed by Tukey’s post hoc test. d–f Cytoplasmic TDP-43 levels were measured in iPSC-MNs transduced with GFP or HK1 in ChAT+ neurons. Scale bar: 20 µm (d). TDP-43 immunodensity was quantified by drawing ROIs around ChAT+ cells. A fixed threshold was applied across all groups, and integrated fluorescence intensity was measured for each individual cell. e The ratio of cytoplasmic TDP-43 intensity to total TDP-43 intensity. Data were obtained from 3 independent experiments and analyzed by one-way ANOVA with Tukey’s post hoc test. All values represent mean ± SE

Next, to determine whether these protective effects extend in vivo, we overexpressed HK1 in the motor cortex of TDP-43^A315T^ mice via stereotaxic delivery of AAV-HK1-eGFP at 4 weeks of age (Supplemental Fig. 6a, b). Behavioral testing was performed at 8 and 10 weeks, coinciding with early symptomatic progression in this model (Fig. 6a). Immunofluorescence confirmed successful motor cortical expression of HK1 in both WT and TDP-43^A315T^ mice (Supplemental Fig. 6c, d). HK1 overexpression significantly extended survival in TDP-43^A315T^ mice compared to eGFP-injected controls (Fig. 6b), while no survival differences were observed in WT mice with or without HK1 overexpression (Supplemental Fig. 6e), indicating a favorable safety profile. Motor coordination, assessed by Rotarod performance, showed significant impairment in TDP-43^A315T^ mice at 10 weeks (6 weeks post-injection), which was markedly improved by HK1 overexpression (Fig. 6c). No motor differences were detected at earlier time points (Supplemental Fig. 6f). Similarly, grip strength testing revealed reduced strength in all limbs of TDP-43^A315T^ mice, which was significantly rescued by HK1 expression (Fig. 6d). Lastly, no changes in body weight were found between any groups 6 weeks post injection (Supplemental Fig. 6g).Fig. 6. Compensation for HK1 loss restores motor neuron function and reduces neuropathology in TDP-43^A315T^ mice. a Experimental timeline showing stereotaxic injection of AAV-HK1 or AAV-GFP into the motor cortex of TDP-43^A315T^ mice. b Survival analysis demonstrating increased survival in TDP-43^A315T^ mice injected with AAV-HK1 compared with AAV-GFP controls (log-rank test: p = 0.0006; Gehan-Breslow-Wilcoxon test: p = 0.0006; n = 17–18 mice/group). Survival time measured as days post-injection (42 days = 6 weeks post-injection). c Rotarod performance assessed 6 weeks post-injection (n = 14–17 mice/group). d Grip strength evaluation at 6 weeks post-injection (n = 14–17 mice/group). Mice were sacrificed after testing, and cortical tissue was collected for biochemical analyses. e Immunohistochemistry for ubiquitin in cortical sections (n = 3 mice/group). f Immunohistochemistry for TDP-43. Cytoplasmic TDP-43 levels were quantified by subtracting nuclear TDP-43 from total TDP-43 (n = 3 mice/group). All data are presented as mean ± SE, and were analyzed by one-way ANOVA followed by Tukey’s post hoc test

Previous reports show an increase of ubiquitination and cytoplasmic TDP-43 in TDP-43^A315T^ mice [2, 22]. To test whether HK1 overexpression also modulates these pathological features, we performed immunohistochemistry on cortex sections. Consistent with previous findings [54], TDP-43^A315T^ mice displayed increased cytoplasmic ubiquitin compared to WT controls, which was reduced by HK1 overexpression (Fig. 6e). Similarly, cytoplasmic TDP-43 accumulation, detected with the C-terminal-specific antibody, was reduced in TDP-43^A315T^ mice overexpressing HK1 (Fig. 6f). No changes were observed in WT animals receiving eGFP vector or eGFP-HK1, indicating that HK1 overexpression does not induce off-target proteinopathy (Fig. 6e, f). Finally, to determine whether the observed molecular changes translated into altered neuronal viability, we performed Nissl staining on cortical sections from symptomatic TDP-43^A315T^ mice. This analysis revealed a significant reduction in neuronal survival in TDP-43^A315T^ mice injected with control vector compared with wild-type mice. Notably, HK1 overexpression significantly rescued neuronal survival in TDP-43^A315T^ mice (Fig. 7a).Fig. 7a Nissl staining for neuronal rescue in TDP-43 mutant mice infected with HK1. Number of neurons were quantified per image. The total number of surviving neurons was normalized to the WT + AAV-Control for each biological repeat (n = 3 mice/group). All data are presented as mean ± SE, and were analyzed by one-way ANOVA followed by Tukey’s post hoc test. b Schematic of the current study. ALS-associated mutations in TDP-43 promote its cytosolic accumulation. Mislocalization of TDP-43 from the nucleus to the cytoplasm enhances its interaction with HK1, leading to recruitment of HK1 from the outer mitochondrial membrane and subsequent sequestration into insoluble TDP-43 fractions. This redistribution reduces mitochondrial HK1, suppresses HK1 enzymatic activity and thereby suppresses glycolysis. Overexpression of HK1 counteracts these effects, rescuing TDP-43–induced pathogenic outcomes both in vitro and in vivo

Together, these findings demonstrate that HK1 overexpression alleviates TDP-43 pathology, enhances motor function, and prolongs survival in a TDP-43–associated ALS mouse model. These results indicate HK1 as a key metabolic effector of TDP-43–induced neurodegeneration and support glycolytic restoration as a promising therapeutic strategy in ALS.

Discussion

Metabolic dysregulation is increasingly recognized as a critical contributor to neurodegenerative disease pathogenesis, but the molecular links between altered energy metabolism and hallmark pathogenic proteins remain incompletely defined. In this study, we identify HK1, the first rate-limiting enzyme of glycolysis, as a direct and disease-relevant target of cytoplasmic TDP-43. Through complementary approaches across cell lines, patient-derived motor neurons, ALS mouse models, and patient postmortem tissue, we demonstrate that cytoplasmic TDP-43 interacts with HK1, disrupts its mitochondrial localization, and promotes its sequestration into insoluble aggregates. This mislocalization impairs HK1 enzymatic activity and leads to reduced glycolytic capacity and neuronal viability, providing a mechanistic link between TDP-43 pathology and glycolytic failure in ALS (Fig. 7b).

Our data demonstrates that cytoplasmic mislocalization of TDP-43 is sufficient to impair glycolytic function. In cells expressing the TDP-43^ΔNLS^ mutant along with TDP-43^G298S^ and TDP-43^M337V^ iPSC-MNs treated with MG-132, a proteasome inhibitor used to induce TDP-43 proteinopathy in iPSC-based models [25], we observed a significant reduction in both basal glycolysis and glycolytic capacity. Notably, knockdown of individual glycolytic enzymes in TDP-43–expressing cells further exacerbated cell death, indicating that impaired glycolysis contributes directly to neuronal vulnerability rather than representing a compensatory metabolic adaptation. In iPSC-MN models, disease-relevant metabolic phenotypes are often context- and stage-dependent rather than constitutively evident at baseline. Consistent with this, we did not observe a significant difference in glycolytic activity between TDP-43^G298S^ and control motor neurons under basal conditions at day 27 of differentiation. Previous studies have shown that metabolic and survival deficits in TDP-43 mutant iPSC-MNs can remain subtle at early stages and become more apparent following cellular challenge [1, 7, 16]. In these systems, neuronal vulnerability frequently reflects reduced stress resilience rather than overt basal dysfunction. Although spontaneous cytoplasmic TDP-43 pathology has been reported in some iPSC-MN differentiation paradigms, the timing and extent of mislocalization vary considerably across protocols and culture duration, particularly in short-term cultures [10, 19]. Proteostasis stress, including proteasome inhibition, has therefore been widely used to reliably induce cytoplasmic TDP-43 accumulation and downstream functional consequences in neuronal systems [42, 51]. Accordingly, MG-132 was used here as a controlled stressor to reveal latent metabolic vulnerability associated with cytoplasmic TDP-43 dysfunction, rather than to induce pathology de novo.

Among rate-limiting glycolytic enzymes, HK1 emerged as uniquely sensitive to TDP-43 mislocalization. Despite unchanged transcript levels, HK1 protein abundance was selectively decreased in cellular, mouse, and human ALS models. This reduction was not observed with TDP-43^WT^, suggesting a cytoplasmic gain-of-function mechanism. Overexpression of TDP-43^WT^ has previously been shown to disrupt cellular homeostasis and induce ALS-like phenotypes [57]. Therefore, reduced cell viability in TDP-43^WT^–expressing cells likely reflects broader cellular stress caused by TDP-43 overexpression rather than pathogenic cytoplasmic mislocalization. In contrast, TDP-43^ΔNLS^ models disease-relevant cytoplasmic TDP-43 pathology, which is the focus of this study. Consistent with the canonical localization of HK1 to the OMM, we found a significant loss of mitochondrial-bound HK1 in cells and tissues expressing mutant TDP-43, accompanied by reduced HK activity. These results are consistent with findings in SOD1-linked ALS and AD, where HK1 displacement from mitochondria has been shown to impair glycolysis and promote neuronal stress [30, 58].

Mechanistically, we demonstrate that cytoplasmic TDP-43 binds directly to HK1, primarily through its LCD. AlphaFold-based structural modeling predicted a potential interaction interface between TDP-43-LCD and HK1. Although the LCD is intrinsically disordered, resulting in low-confidence predictions [14], we experimentally validated this interaction using immunoprecipitation assays with recombinant full-length TDP-43 and HK1 proteins. Moreover, immunoprecipitation using recombinant LCD-TDP-43 protein further confirmed a direct interaction with HK1, supporting a specific binding role for the LCD. In addition, we showed both a mislocalization of HK1 and a reduction in total protein in TDP-43-associated ALS models. Since our previous results demonstrating a decrease in HK1 total protein was performed in triton soluble fractions of samples, we also homogenized the insoluble fraction with urea buffer and found an increase in HK1 within the insoluble fraction of cells expressing TDP-43^ΔNLS^ compared to control cells. Thus, when TDP-43 mislocalizes to the cytoplasm, the protein interacts directly with HK1 and shuttles HK1 to insoluble fractions commonly seen in ALS patients [35]. Collectively, we show that the interaction of TDP-43 and HK1 occurs in the cytosolic compartment and correlates with the redistribution of HK1 into detergent-insoluble aggregates, a pathological feature commonly observed in ALS patient tissue [4].

Restoration of HK1 expression rescued multiple aspects of TDP-43–associated pathology. HK1 overexpression in patient-derived iPSC-MNs reduced cytoplasmic TDP-43 accumulation, and in TDP-43^A315T^ mice, it significantly improved motor performance and extended survival. These therapeutic effects were accompanied by reductions in cytoplasmic ubiquitin and pathological TDP-43 in MNs, suggesting that restoring glycolytic capacity via HK1 normalization may alleviate proteostasis stress and neurodegeneration. Although HK1 overexpression clearly mitigates TDP-43 pathology, the precise downstream mechanisms warrant further investigation. One possibility is that restoring mitochondrial-bound HK1 enhances glycolytic flux and ATP production, alleviating energy stress in vulnerable neurons. Alternatively, HK1 may influence non-glycolytic processes such as apoptosis, as its mitochondrial anchoring has been shown to suppress pro-apoptotic signaling [24]. It is also plausible that HK1 indirectly affects TDP-43 turnover or solubility via its role in maintaining metabolic and redox homeostasis. Previous studies have reported that inhibition of TDP-43 mitochondrial mislocalization can mitigate mitochondrial defects and ultimately rescuing motor neuron defects [52], proposing overexpression of HK1 could inhibit further TDP-43 mitochondrial localization. Importantly, our results demonstrate a slight decrease in nuclear TDP-43 upon HK1 overexpression in iPSC models, reinforcing the necessity to investigate the function of HK1 after restoration. Future studies will be needed to dissect these potential contributions and determine whether glycolysis-independent roles of HK1 contribute to its protective effects.

In summary, our findings identify HK1 as a critical metabolic effector disrupted by cytoplasmic TDP-43 and establish a mechanistic link between TDP-43 pathology and impaired glycolytic metabolism in ALS. These results highlight the broader concept that metabolic enzymes may serve as both functional effectors and targets of proteinopathy in neurodegenerative disease. Targeting glycolytic restoration through HK1 or its interactome may offer a promising therapeutic strategy to counteract TDP-43–driven neurodegeneration.

Supplementary Information

Below is the link to the electronic supplementary material.Supplementary file1 Supplementary material is available at ACTA Neuropathologica online (PDF 9065 KB)