Silencing of the Metabolic Gene HKDC1 Is Associated With Aging and Neurodegeneration in Mice and Humans
Zeenat Farooq, Vladimir Ilievski, James Boyett, Julianne Jorgensen, Yang Pan, Tanika Kelly, David Bennett, Orly Lazarov, Brian T. Layden
TL;DR
This study shows that the metabolic gene HKDC1 declines with age, leading to brain issues like memory loss and neuroinflammation in both mice and humans.
Contribution
The study identifies HKDC1 as a novel metabolic gene linked to aging and neurodegeneration through chromatin and transcriptional changes.
Findings
HKDC1 expression declines in humans and mouse models with cognitive decline and aging.
Reduced HKDC1 leads to memory loss, anxiety, and mitochondrial dysfunction in mice.
Chromatin changes block TFEB binding, causing HKDC1 downregulation and promoting neurodegeneration.
Abstract
Increased life expectancy brought about by improved healthcare and lifestyle has heightened the challenge of neurodegenerative disorders like Alzheimer's disease (AD) and other age‐related disorders. Neurodegeneration is known to be accompanied by loss of memory, changes in brain morphology, and neuroinflammation, and multiple factors contribute to the progression and pathogenesis of the condition. Of these factors, metabolic dysregulation is known to influence the process, but the precise mechanisms remain unexplored. In this study, we investigated the brain‐specific role of the metabolic enzyme hexokinase domain‐containing 1 (HKDC1) in neurodegeneration and observed that HKDC1 expression declines in humans with cognitive decline, which matches similar findings in mouse models of AD and aging. We observed age‐dependent anxiety, compromised memory and learning, senescence,…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
FIGURE 1
FIGURE 2
FIGURE 3
FIGURE 4
FIGURE 5
FIGURE 6- —National Institutes of Health10.13039/100000002
- —VA Office of Research and Development (VA‐ORD)10.13039/100006379
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsCancer, Hypoxia, and Metabolism · Biochemical Acid Research Studies · Mitochondrial Function and Pathology
Introduction
1
Due to improvements in lifestyle, healthcare, and a decline in infectious diseases, life expectancy has increased significantly globally over the last century (Stoyanova 2014; Vaupel 2010). One of the biggest challenges we face is an increase in the number of elderly individuals living with aging‐related disorders, including Alzheimer's, Parkinson's, and Huntington's diseases, which constitute a huge socioeconomic burden (Tay et al. 2024; Wehling and Groth 2011; Yang et al. 2020). Of these, Alzheimer's disease (AD) is the most common form of neurodegeneration and affects 40 million people worldwide, and is the seventh leading cause of death in the United States (“2024 Alzheimer's disease facts and figures,” 2024). Physiologically, all chronic brain aging or neurodegenerative diseases (NDD) have similar characteristic features: (i) progressive loss of neurons accompanied by impairment in memory, motor skills, and ultimately death (Martin 1999), (ii) neuroinflammation due to activation of inflammatory pathways in the brain (Giri et al. 2024) (iii) impairment of synaptic plasticity, which affects learning (Bae and Kim 2017; Galts et al. 2019; Gribkoff and Kaczmarek 2017; Keifer 2022). Studies in the field have also shown that all forms of neurodegeneration are associated with changes in metabolic pathways (Cai et al. 2012). NDDs also often present clinically with a coexisting metabolic dysfunction, which may exacerbate neurological symptoms. For instance, one of the major metabolic outcomes associated with AD is insulin resistance (Arrieta‐Cruz and Gutierrez‐Juarez 2016). In fact, AD has been referred to as Type III Diabetes due to the association of AD pathology with altered insulin action and dysregulation of glucose metabolism (Arrieta‐Cruz and Gutierrez‐Juarez 2016; Bellenguez et al. 2022). Clearly, the development of novel therapies requires a deeper understanding of molecular mechanisms underlying NDD, and research into mechanisms of metabolic control could offer new therapies to benefit the rapidly expanding cohort of patients with NDD. Uncovering the genes that play a role in the metabolic regulation of NDD is, therefore, highly significant.
Hexokinases (HK) are a family of enzymes that catalyze the first step of glucose metabolism, specifically by phosphorylating glucose to glucose‐6‐phosphate across all organisms, from bacteria to mammals (Farooq et al. 2023). HK enzymes exist in multiple isoforms across organisms. In mammals, there are four canonical hexokinases (HKI to HKIV), which differ in tissue distribution, enzyme kinetics, and the physiological state of the cell, among other factors (Bennett et al. 2003). HK1 is the most highly expressed hexokinase, with the highest levels of expression in the brain. More than 90% of the glucose in our body is utilized by the brain, and HK1 is responsible for its utilization to meet the energy demands (De Jesus et al. 2022). Previously, our lab helped identify the fifth novel hexokinase HKDC1 (Guo et al. 2015; Hayes et al. 2013). Our group and others have reported that HKDC1 is genetically associated with impaired oral glucose tolerance in gestational diabetes (Guo et al. 2015; Hayes et al. 2013; Zapater et al. 2022) and is involved in mice whole‐body glucose homeostasis in aging and pregnancy (Khan et al. 2019; Ludvik et al. 2016; Pusec et al. 2019; Zapater et al. 2022). Similar to HK1, HKDC1 possesses a mitochondrial binding domain, and we have recently reported that elevated hepatic HKDC1 plays a role in liver cancer progression through its interaction with mitochondria (Khan et al. 2022; Pusec et al. 2023). Due to the association of metabolic disease with NDDs, we assessed whether HKDC1 was expressed in the brain, observing that it was highly expressed in mouse brains. Considering this and its known association with mitochondrial homeostasis and senescence, we explored AD human sample databases and observed that HKDC1 expression declines with AD severity in humans and AD mouse models. Next, brain deletion of HKDC1 also had indications of memory deficit and anxiety, where this model was accompanied by an indication of dysregulation of mitophagy and an increase in senescence and neuroinflammation in mice. These findings suggest a novel role for HKDC1 as a key metabolic gene in the pathogenesis of NDD. Further studies to dissect the molecular mechanism of HKDC1 involvement in NDD could lead to a novel strategy for timely intervention in these debilitating chronic diseases.
Methods
2
Animals
2.1
C57BL/6J mice expressing Cre recombinase under the direction of the mouse Nestin 1 promoter (Nestin‐cre mice) were obtained from the Jackson Laboratory (Strain #003771), where Cre expression in these mice is observed in the central and peripheral nervous system, including neuronal and glial cell precursors. Samp8 and SamR1 mice were purchased from Inotiv (formerly Envigo). HKDC1^fl/fl^ mice (on a C57BL/6J background) were generated by the Wellcome Trust Sanger Institute and obtained from the Knockout Mouse Project Repository (www.komp.org). Mice were housed in the Biological Resources Laboratory at the University of Illinois at Chicago (UIC) under a 12‐h light/dark cycle. Mice were maintained on a chow diet throughout the study (Envigo #7012, Madison, WI, USA). All mouse studies were approved by the UIC Animal Care and Use Committee and performed in accordance with the Guide for the Care and Use of Laboratory Animals at UIC.
Creation of Brain‐Specific HKDC1 Knockout Transgenic Mice
2.2
Nestin‐cre transgenic male mice were mated with HKDC1^fl/fl^ female mice to create mice with the HKDC1 gene knocked out in Nes‐expressing brain tissue (HKDC1BKO mice). The presence of floxed HKDC1 and Nestin*‐Cre* was assessed through genotyping using GreenTaq ReadyMix PCR reaction mix (Sigma, St. Louis, MO, USA) with the primers listed in Table S1.
Novel Object Location (NOL) Test
2.3
The test was adapted from the novel object recognition test with a modification that, instead of changing the object itself, the position of the object was changed to test spatial memory (Bellenguez et al. 2022; Lazarov and Hollands 2016). The apparatus, having a measurement of 30 1/2″ × 20 1/4″ × 16 5/8″ inches, was used to perform the NOL experiment. It consisted of a cubic chamber, open at the upper part, where the mice were placed, and a digital camera was positioned overhead to record behavior and facilitate subsequent analysis. Briefly, two identical objects were placed in distinct corners of the chamber, which the mice were allowed to explore for two 10‐min sessions during the training or until the animals explored the two objects for a total of 20 s. “Exploration of the object” was defined as the head direction pointing towards the object while the animal touches it. A habituation and familiarization trial was also performed for two consecutive days, 5 min each day, prior to training to familiarize the mice with the chamber environment. Testing was performed 24 h after training, during which one of the two objects (randomly chosen) was moved to a different corner. The time spent exploring both objects over 10 min or until the two objects were explored for 20 s was recorded. The box was cleaned with 70% alcohol between trials to eliminate olfactory cues (Mishra et al. 2022). Mice were introduced right in the middle of the chamber, facing the same side, to further eliminate exploration bias. Increased time spent exploring the object in the novel location was interpreted as successful retention of spatial memory for the object that had not been moved. The % exploration time and Discrimination ratio were calculated using the following equations (Abdullah et al. 2025):
Open Field (OF) Test
2.4
The open‐field test, which measures baseline locomotor activity and anxiety, was performed as described previously (Bellenguez et al. 2022). Similar to NOL, mice were trained and tested on two consecutive days in the same arena, and the % time spent in the open arena was calculated using the following equation:
For NOL and OF studies, block randomization was performed to ensure balanced group numbers and maintain consistency between groups when animals were added in cohorts (Karp and Fry 2021). Blinding was implemented by coding both groups (fl/fl and BKO) by a third party, and cages were labeled accordingly. For power calculations, we initially used “resource equation” method based on the law of diminishing return (Charan and Kantharia 2013) where we measured “E” (total number of animals−total number of groups), which is the degree of freedom of analysis of variance (ANOVA) and used at least 6 mice per group. However, due to anxiety associated with thigmotaxis and reduced exploration in 9‐month‐old brain knockout male mice, formal power calculation was not possible. However, we have used a sufficient number of animals (n ≥ 5) for statistical analysis (Figure 2) based on previous studies using similar mouse models (Shen et al. 2025).
Quantitative Polymerase Chain Reaction
2.5
Total RNA was extracted from 50‐mg minced whole brain tissue using TRIzol reagent (ThermoFisher Scientific, #15596026) and chloroform (Invitrogen, Carlsbad, CA, USA) for phase separation. RNA was purified using RNeasy Mini Kit (Qiagen, Germantown, MD, USA) and resuspended in nuclease‐free water (Qiagen). 1 μg of purified RNA was used for reverse transcription using RevertAid First Strand cDNA Synthesis Kit (Thermo Fisher Scientific, K1621) and quantified via qPCR using iTaq Universal SYBR Green Supermix (Bio‐Rad Laboratories Inc., 1,725,124). Relative expression levels were calculated for different targets using the ΔΔCT method and normalized to 18S RNA levels. Each reaction was performed in triplicate with a final volume of 10 μL, where the final primer concentrations were 0.625 μM for each reaction. Data were collected and analyzed using the CFX Connect Real‐Time PCR Detection System (Bio‐Rad) and CFX Maestro 2.2 software (Bio‐Rad), respectively. The sequences of primers utilized for qPCR are in Table S2.
Blood Glucose Measurements
2.6
Blood glucose was measured from tail vein blood using a OneTouch UltraMini glucose monitor (Lifescan Inc., Milpitas, CA, USA).
Measurement of Serum Aβ‐42 Levels
2.7
Blood was collected from the tail vein in heparinized capillary tubes, transferred to nonheparinized Eppendorf tubes, centrifuged at 2000 rpm at 4°C for 20 min, and the supernatant containing serum was collected. Serum Aβ‐42 levels were determined using an enzyme‐linked immunosorbent assay according to the manufacturer's protocol (Mouse Aβ‐42 ELISA Kit, Invitrogen, KMB3441, Waltham, MA, USA).
Intraperitoneal Glucose Tolerance Tests
2.8
After a 16‐h overnight fast, mice were administered 2 g/kg body weight of glucose via intraperitoneal injection, and blood glucose levels were assessed as described above for up to 120 min.
Insulin Tolerance Test
2.9
Mice were fasted for 4 h, then administered 0.75 units/kg body weight of Humalog insulin (Eli Lilly & Co., Indianapolis, IN, USA). Blood was collected and assessed for glucose as described earlier (Pusec et al. 2019).
Immunoblotting
2.10
Brain sections were homogenized in RIPA lysis buffer of components 10 mM Tris–HCl (pH 8.0), 1 mM EDTA, 0.5 mM EGTA, 1% Triton X‐100, 0.1% sodium deoxycholate, 0.1% SDS, 140 mM NaCl, 1 mM phenylmethylsulfonyl fluoride (PMSF) (R0278, Sigma Aldrich, Millipore, Burlington, MA, USA), supplemented with a protease inhibitor cocktail (ThermoFisher Scientific) and phosphatase inhibitor cocktail (PhosSTOP, #490684500, Roche). Protein concentrations were determined using Pierce Dilution‐Free Rapid Gold BCA Protein Assay (ThermoFisher, A55860). 30 μg protein was used per sample and added to Laemmli sample buffer (Bio‐Rad) at a ratio of 1:3, separated by sodium dodecyl sulfate‐polyacrylamide gel electrophoresis using Mini‐PROTEAN TGX 4% to 15% Gels (Bio‐Rad), and transferred to Trans‐Blot Turbo Mini‐size nitrocellulose membranes (Bio‐Rad). Membranes were blocked with 5% nonfat dried milk in tris‐buffered saline supplemented with 0.1% Tween‐20 for 1 h at room temperature, followed by incubation with desired primary antibodies overnight at 4°C [β‐actin (Cell Signaling technologies, 13E5), GAPDH (Abcam, ab8245), HKDC1 (Abcam, ab228729), HK1 (Abcam, ab154839), COX‐2 (Proteintech, #66351–1‐Ig), Cox‐IV (Cell Signaling technologies, #4850 T), Phospho‐tau (Invitrogen, #MN10120), TFEB (Cell Signaling technologies, #37785S), LC3B (Cell Signaling technologies, #2775), Pink1 (Proteintech, #23274–1‐AP), Parkin1 (Cell Signaling technologies, #4211 T), p‐Parkin1 (Abcam, #ab315376), and P62 (Cell Signaling technologies, #5114 T)]. All nitrocellulose membranes were then incubated with HRP‐conjugated secondary antibodies (Promega Express, #W4011 and W4021, Madison, WI) at 25°C for 1 h. Membranes were washed and incubated with SuperSignal Chemiluminescent Substrate (ThermoFisher Scientific, #34580). The emitted light signal was detected and analyzed using a ChemiDoc MP Imager (Bio‐Rad) and Image Lab software, version 6.0 (Bio‐Rad), respectively.
Brain Tissue Fractionation
2.11
Cytosolic, mitochondrial, and nuclear fractions from brain tissue samples were separated using the method of Ivan Dimauro et al. with slight modifications (Dimauro et al. 2012). Briefly, tissue samples were washed two times with ice‐cold 1× PBS and resuspended in 400 μL of ice‐cold STM buffer (250 mM sucrose, 50 mM Tris–HCl pH 7.4, 5 mM MgCl_2_, protease and phosphatase inhibitor cocktails) and homogenized for 1 min on ice using a tight‐fitting Teflon pestle attached to a homogenizer. The suspension was incubated on ice for 30 min, vortexed at maximum speed for 15 s, and then centrifuged at 800 g for 15 min. The pellet (P) was used for the isolation of nuclear proteins, and the supernatant (S) was used for the subsequent isolation of mitochondrial and cytosolic fractions.
Immunofluorescence Microscopy and Immunohistochemistry
2.12
Immunofluorescence (IF) microscopy was performed to detect HKDC1, NeuN (a marker for intact neurons), GFAP (astrocytes), IL‐1β (a proinflammatory cytokine), and nuclei (DAPI). Briefly, tissue sections were first deparaffinized in xylene and rehydrated through a graded ethanol series. Antigen retrieval was conducted by microwaving the sections in either 1 mM EDTA (pH 8.0) or 10 mM citrate buffer (pH 6.0) for 5 min, repeated four times. Cell and nuclear membrane permeabilization were achieved by incubating the sections in 0.25% Tween‐20 in PBS for 30 min. Tissue sections were independently incubated with primary antibodies to detect specific cell markers and cytokines: (1) mouse monoclonal anti‐NeuN antibody (ab104224, Abcam, Cambridge, MA; 1:100) for intact neurons, (2) mouse monoclonal anti‐GFAP antibody (3670, Cell Signaling Technology, Danvers, MA; 1:100) for astrocytes, (3) rabbit polyclonal anti‐IL1β antibody (ab9722, Abcam; 1:100), and (4) rabbit polyclonal anti‐HKDC1 antibody (ab228729, Abcam; 1:100). All primary antibodies were incubated overnight at 4°C. Following primary antibody incubation, appropriate secondary antibodies were applied: donkey anti‐rabbit Alexa Fluor 594‐conjugated antibody (A21207, Thermo Fisher Scientific, MA) or donkey anti‐mouse Alexa Fluor 488‐conjugated antibody (R37114, Thermo Fisher Scientific, MA). Isotype‐matched negative controls were included to assess non‐specific antibody binding. Nuclei were counterstained using ProLong Gold Antifade Mountant with DAPI (P36931, Thermo Fisher Scientific, Waltham, MA).
Senescent cells were identified using a histochemical assay for β‐galactosidase activity, performed on frozen tissue sections according to the manufacturer's protocol (Senescence Histochemical Staining Kit, #CS0030‐1KT). Stained sections were then imaged using immunofluorescence (IF) microscopy.
Chromatin Immunoprecipitation (ChIP)
2.13
ChIP assays were performed as described previously (Banday et al. 2016; Braveman et al. 2004; Farooq et al. 2019). Briefly, brain tissue was extracted and minced, followed by cross‐linking with 1% formaldehyde in PBS for 10 min with gentle shaking at room temperature. Next, quenching was done using 125 mM glycine in PBS for 5 min. Tissue lysis was performed using a homogenization method. Cross‐linked chromatin was sonicated to yield DNA fragments of an average size of 200–600 bp. Immunoprecipitations were carried out using different antibodies [anti‐RNA pol II (Abcam, #AB5131), anti‐TFEB (Abcam, #AB2636), anti‐H3K27ac (Abcam, #AB4729), and anti‐H3 (Abcam, #AB1791)]. Primers used in the PCR were first analyzed for both efficiency and linearity range by real‐time PCR with a LightCycler (BioRad). All ChIP experiments were performed three times, yielding similar results. qPCR was performed using the SYBR Green reagent (Thermo Fisher Scientific, #4364346) according to the manufacturer's protocol using a Real‐Time Cycler (BioRad). The results shown are based on three independent experiments with standard errors. Data were corrected for the nucleosome occupancy using the total H3 signal. Primers used to amplify different TFEB binding sites on the HKDC1 promoter are given below in Table S3. Multiple pairs of pairs were designed to achieve better coverage of the binding sites of TFEB on the HKDC1 promoter. All other ChIP qPCR primer sequences are also included in Table S3.
Cell Culture and Reagents
2.14
CCF‐STTG1 cells (ATCC, #CRL‐1718) were grown in RPMI‐1640 medium (Gibco; #11875093). All growth media were supplemented with 10% fetal bovine serum (FBS; Sigma; F0926) and 1% penicillin–streptomycin (Gibco; #15070063). Cells were plated in T‐25 flasks (Thermo Fisher Scientific; #156430) and incubated in a 37°C humidified incubator with 5% CO_2_ until they formed a monolayer and achieved the desired confluence.
Transfection
2.15
Lipofectamine‐based transfection using RNAi MAX lipofectamine (Thermo Fisher Scientific, #13778030) was performed as per the manufacturer's protocol for the knockdown of HKDC1. CCF‐STTG1 cells were transfected with HKDC1 siRNA (Millipore Sigma; # SASI_Hs01_0021–3759/HKDC1) against human HKDC1. Briefly, cells were grown to a confluency of 90% in 6‐well plates. For each well of a 6‐well plate, 150 μL of opti‐MEM (Gibco; 31‐985‐070) was mixed with 2 μL of siRNA. This was added to a mixture containing 150 μL of opti‐MEM (Gibco; #31985062) and 9 μL RNAi MAX lipofectamine (#13778030), and incubated for 5 min. The mixture was then added to the wells in a dropwise manner following the addition of 1.5 mL of fresh opti‐MEM, and cells were incubated for 6 h. Next, 1.5 mL of fresh DMEM:F12 media was added to each well and incubated overnight. After 24 h, cells were harvested for protein and RNA extraction.
Induction of Autophagy and Reactive Oxygen Species Production (ROS)
2.16
CCF‐STTG1 cells were transfected as described above with either siScr or siHKDC1 siRNA followed by Accutase (Gibco; A1110501) for detachment after 6 h and plated in 6‐well clear and 96‐well black‐wall plates at 600,000 and 30,000 cells per well, respectively, overnight. At 20 h post‐transfection, cells were treated with FCCP (MedChemExpress; HY‐100410) according to their experimental group for protein isolation or quantification of ROS production, respectively. The 6‐well plate for protein was treated with a 20 μM dose for 2 h prior to three washes with ice‐cold PBS and lysis with supplemented RIPA buffer (Sigma; R0278) supplemented with Halt protease and phosphatase inhibitor (ThermoFisher; 1,861,281). Quantification and immunoblotting were performed with 20 μg of protein per sample, as described above.
For the 96‐well black‐wall plate, cells were treated with 20uM 2′7′‐Dichlorodihydrofluorescein Diacetate (Sigma, 35,845) diluted in 2X‐PBS for 40 min prior to aspiration and treatment with DMSO vehicle (Sigma, D2650), 100uM TBHP (Sigma, 168,521) for maximum ROS production, or 5 μM, 10 μM, and 20 μM FCCP for 4 h. Following treatment, fluorescence was measured using a BioTek Synergy H1 plate reader at an excitation wavelength of 500 nm and an emission wavelength of 540 nm. Maximal fluorescence was quantified for each transfection condition from the TBHP treatment group and used to calculate % of maximum ROS production per sample.
Gene Expression Profiling (RNA‐Seq) and Bioinformatics Analysis
2.17
HKDC1‐BKO and HKDC1^fl/fl^ mice were sacrificed at 9 months. Brains were harvested, minced, and RNA was isolated using TRIzol reagent. RNA quality control (QC) was performed using a bioanalyzer. RNA‐Seq analysis was performed in the Genomic Core Facility at the University of Chicago on samples with a 7–10 RIN (RNA Integrity Number). Bioinformatics analysis was performed by the University of Illinois Research Informatics Core. Raw reads were aligned to the reference genome in a splice‐aware manner using the STAR aligner (Dobin et al. 2013). Gene expression was quantified using FeatureCounts (Liao et al. 2014) against Ensembl gene annotations. Differential expression statistics (fold change and p value) were computed using edgeR (McCarthy et al. 2012), and p values were adjusted for multiple testing using the false discovery rate (FDR) correction of Benjamini and Hochberg (Reiner et al. 2003). Raw data are deposited in the GEO database (accession ID awaited). Results from RNA‐seq analysis were uploaded to Metascape.org for pathway analysis, as described previously (Zhou et al. 2019). The top 100 genes that showed significant changes in HKDC1‐BKO compared to control samples were subjected to IPA core analysis based on IPA annotation databases.
Human Brain Proteomic Data Analysis
2.18
Proteomic data from the dorsolateral prefrontal cortex (dPFC) of postmortem brain samples donated by participants in the Religious Orders Study and Memory and Aging Project (ROSMAP) were profiled using isobaric tandem mass tag mass spectrometry (TMT‐MS) (Bennett et al. 2018; Ping et al. 2020; Robins et al. 2021). All participants in both cohorts were enrolled without known dementia and agreed to undergo annual clinical evaluations and to donate their brains upon death. Both studies were approved by the Institutional Review Board of Rush University Medical Center. All participants signed the Anatomical Gift Act and both informed and repository consents. The data processing and protein abundance quantification steps, including batch correction and normalization, have been described previously (Ping et al. 2020; Robins et al. 2021). In brief, 10,030 proteins were retained after filtering for missingness (≤ 70%), followed by log2 transformation and principal component analysis (PCA)‐based outlier removal, from a total of 971 samples, with batch and post‐mortem interval regressed out for analysis. After excluding samples with missing covariates, up to 78 participants were included in the final analysis. AD dementia diagnosis was made as reported (Bennett et al. 2006). Postmortem pathological evaluation of brain samples was conducted and described elsewhere (Bennett et al. 2003). AD‐related pathologies, including global AD pathology, amyloid beta, and neurofibrillary tangles, were considered as dichotomized as high versus low based on the median (Wilson et al. 2007). Logistic regression models were applied to assess the association between HKDC1 protein abundance and AD‐related pathologies, adjusting for age at death, sex, race, education, and clinical measures from the final visit, including depression, and prior history of stroke. Results were reported as odds ratios (OR) per 1 standard deviation (SD) greater HKDC1 protein levels, with 95% confidence intervals (CI) and corresponding p values.
For two‐sample Mendelian randomization (MR) analysis, we leveraged data from a recent large‐scale brain protein quantitative trait loci (pQTL) study (Robins et al. 2021) to identify genetic instruments associated with HKDC1 protein levels in the DLPFC. Summary statistics from two large and independent AD genome‐wide association studies (GWAS), namely EDBA stage I (Bellenguez et al. 2022) and FinnGen R7 (Kurki et al. 2023), were used for the outcome. Two‐sample MR analyses were conducted using the R package TwoSampleMR tool (Hemani et al. 2018), with sensitivity analyses performed to account for horizontal pleiotropy and weak instrument bias. Results from the two GWAS datasets were combined in a meta‐analysis using a fixed/random‐effects model, depending on heterogeneity.
Human Brain Single‐Nuclei RNAseq Data Analysis
2.19
The frozen postmortem brain tissues were processed to isolate nuclei on ice or at 4°C with a previously reported protocol (Kurki et al. 2023; Mathys et al. 2023). Chromium Single Cell 3 Reagent Kits v3 were used to prepare libraries following the protocol from 10x Genomics. The snRNA‐seq libraries were then sequenced with either NextSeq 500/550 High Output v2 kits or NovaSeq 6000 S2 Reagent Kits (Kurki et al. 2023; Mathys et al. 2023). The snRNA‐seq libraries were then sequenced with either NextSeq 500/550 High Output v2 kits or NovaSeq 6000 S2 Reagent Kits (Kurki et al. 2023; Mathys et al. 2023). Reads were aligned to the GRCh38 genome and mapped to pre‐mRNA to account for unspliced nuclear transcripts. All libraries were then aggregated to generate a gene count matrix using Cell Ranger software (version 3.0.2) (10× Genomics). (Velmeshev et al. 2019). For quality control, doublets and poor‐quality reads were identified as extreme outliers (3× outside the lower and upper quantiles) for percentage of mitochondrial genes, number of genes, or unique molecular identifiers and excluded, resulting in 2,359,994 total cells. Cell clusters were identified with FindNeighbors and FindClusters (Bennett et al. 2018; Mathys et al. 2023). Cell clusters were annotated with published marker genes, and marker genes were identified separately for each high‐resolution major cell class (Mathys et al. 2023). Astrocytes were then compared between groups with different cognitive impairment.
Results
3
Loss of HKDC1 Expression in the Human and Mouse Brain Is Associated With Aspects of Neurodegeneration
3.1
Considering the association of diabetes with dementia and previous links of HKDC1 to gestational and type II diabetes, we profiled HKDC1 expression in different organs in one of our mouse models and observed that it is highly expressed in the brain (Figure S1A,B). Next, leveraging the mouse brain atlas, we further observed that it is most highly expressed in the cortex and hippocampus in mice (Figure S1C,D). We also looked at publicly available datasets on AD to look at the expression of HKDC1 using the gene expression omnibus (GEO) profiling tool and observed a reduction in the expression of HKDC1 in AD; while it is expressed in all the major regions of the normal human brain (Figure S1E,F). Because of these data, we sought to test if there is a relationship between HKDC1 expression in the brain and dementia. We then explored the association of HKDC1 protein abundance with AD pathologic traits in the ROSMAP. We leveraged 78 brain proteomes profiled from the dorsolateral prefrontal cortex (DLPFC) using isobaric tandem mass tag mass spectrometry (TMT‐MS) with paired AD dementia and AD pathologic indices. While no association was found between HKDC1 protein abundance and AD dementia (data not shown), one standard deviation higher abundance was associated with 24% lower odds of global AD pathology burden (95% confidence interval [CI]: 0.65–0.89; p < 0.001) and 24% lower odds of Amyloid beta protein level (95% CI: 0.64–0.88; p < 0.001), after adjusting for demographic and study covariables. We then performed a two‐sample Mendelian randomization (MR) analysis, integrating AD GWAS summary statistics with HKDC1 brain protein quantitative trait loci (pQTL) data. Genetically predicted brain HKDC1 protein abundance conferred a potentially protective effect against AD, with marginal significance in a meta‐analysis of two independent AD GWAS datasets (OR: 0.51, 95% CI: 0.23–1.12; p = 0.07) (Figure 1A,B). And finally, considering these relationships in human databases, we assessed if HKDC1 expression is impacted in mouse models of AD by evaluating its expression in 5X FAD mice, and we observed a drastic decrease in RNA and protein expression levels (Figure S2A,B).
Associations of HKDC1 protein abundance in the human brain with Alzheimer's disease (AD)‐related brain pathologies and AD. Forest plots showing: (A) associations between HKDC1 protein abundance and AD‐related brain pathologies. Odds ratios (ORs) are per 1 standard deviation (SD) increase in HKDC1 levels, adjusted for age at death, sex, race, education, and clinical measures at the final visit (depression and stroke history). Results that remain statistically significant after Bonferroni correction (α = 0.016) are shown in bold font; and (B) associations of genetically predicted higher HKDC1 protein abundance with AD, based on two large AD GWAS summary statistics. CI, confidence interval; N, sample size used for logistic regression or Mendelian randomization analyses.
HKDC1 Expression in the Mouse Brain Is Important for Memory, Learning, and Age‐Associated Anxiety
3.2
To determine the physiological consequences of loss of HKDC1 in the brain, we generated a brain specific HKDC1 knock‐out mouse model (BKO) by crossing HKDC1^fl/fl^ female mice with C57BL/6 Nestin‐cre male mice (Jackson Laboratories) (Figure S2C). This model was validated for knockout of HKDC1, as seen in Figure S2D–F where we found that HKDC1 mRNA expression levels within the brain were significantly reduced (p < 0.0001) in the BKO mice compared to expression levels in the HKDC1^fl/fl^ littermate control group, and the expression remained unchanged in other tissues (Figure S2D), including at the protein level (Figure S2E). To determine chronic, age‐dependent differences of HKDC1 loss on memory and learning, we performed a novel object location (NOL) test on our mice at 4 and 9 months of age (Figure 2A–C). We observed that 9‐month‐old male and female HKDC1 brain knockout mice had memory and learning deficits compared to the HKDC1^fl/fl^ littermate control group (Figure 2D,E), which was not observed at 4 months (Figure S3). We next tested anxiety levels in our 4‐ and 9‐month‐old mice, employing an open maze test (Figure 2F). We observed that 9‐month‐old HKDC1 brain knockout male mice had significant anxiety compared to the HKDC1^fl/fl^ littermate control group (Figure 2G), which was not observed at 4 months (Figure S3). Interestingly, female HKDC1 brain knockout mice were protected from anxiety even at 9 months of age (Figure 2G). To begin to understand the brain‐specific role of HKDC1, we explored the impact of chronic HKDC1 loss on the brain transcriptome via bulk RNA sequencing of the mouse brain (Figure S4). The RNA sequencing showed distinct transcriptomic profiles with transcripts significantly different between the groups (Figure S4A–D). Examining the ingenuity pathway analysis, we observed that many pathways were downregulated in HKDC1 BKO animals, with the most significantly downregulated pathways included those involved in synaptic signaling, behavior, cognition, neuron projection development, memory, and learning, regulation of membrane potential, and so on (Figure 2H). Overall, chronic loss of HKDC1 in the brain shifts the transcriptomic profile towards loss of expression of genes associated with memory, learning, and neuronal development.
*HKDC1 expression in the brain of mice is important for memory and learning. (A) Schematic of HKDC1fl/fl and HKDC1‐BKO mice, timelines of behavioral tests conducted. (B) Schematic of the novel object location (NOL) test conducted on male and female HKDC1fl/fl and HKDC1‐BKO mice. (C) Image of the chambers used to conduct novel object location (NOL) test. (D, E) Quantification of percent time spent by (D) male mice and (E) female mice around the objects at the old (O) and new (N) location (n = 8). (F) Schematic of the open maze test conducted to test anxiety in HKDC1fl/fl and BKO male and female mice. (G) Quantification of percent time spent in the open arena by male and female mice (n = 8). Data is Mean ± SD; *p < 0.05; **p < 0.01; ***p < 0.001; ***p < 0.0001; by student t‐test with Welchs correction. Effect size (r) = −0.98 (d) = −11.37 (D), (r) = −0.990 (d) −14.21 (E), (r) = 0.96 (d) = 7.05 (G). (H) Ingenuity Pathway Analysis identified major pathways associated with synaptic signaling and neuronal development. Gene Ontology biological process terms show significantly enriched pathway with downregulated genes in HKDC1‐BKO versus HKDC1‐fl/fl mouse brain. The graph shows downregulated genes in each functional category.
Loss of HKDC1 Expression in the Brain Does Not Lead to an Overt Metabolic Phenotype
3.3
To investigate if changes in metabolic homeostasis due to the absence of HKDC1 in the brain occurred and impacted our outcomes, both male and female BKO and HKDC1^fl/fl^ mice underwent a metabolic characterization on a standard chow diet from week 4 to week 44 (9 months) of age (Figure S5). No differences in either body weight or ad libitum glucose (Figure S5A–D), percent lean and fat mass (Figure S5E,F) were observed between the two groups throughout the study. We also compared the brain tissues of both groups following sacrifice at 9 months and observed that HKDC1^fl/fl^ and BKO mice also did not differ in brain morphology, brain weight, or brain/body weight ratios (Figure S5G,H). Furthermore, no differences in fasting glucose levels, glucose tolerance by intraperitoneal injection (Figure S5I,J), or insulin tolerance (Figure S5K,L) were observed. Collectively, these data indicate that no major congenital or time‐dependent differences in metabolic homeostasis are observed in HKDC1‐BKO mice due to their transgenic composition.
To understand if the absence of brain HKDC1 impacts other hexokinases that could potentially compensate for HKDC1 loss, we evaluated their expression in the brain of male and female BKO and HKDC1^fl/fl^ mice at 4 and 9‐months of age and observed that HK1 RNA and protein expression remain consistent, regardless of age, sex and HKDC1 ablation (Figure S6A–E), where HK1 is the other predominant HK in the brain. It is well known that HK1 is primarily associated with mitochondria through its mitochondrial binding domain and is responsible for channeling glucose towards glycolysis (respiration). Consistent with previous reports, subcellular fractionation of HKDC1^fl/fl^ and BKO mouse brains showed no difference in HK1 localization or total protein amount (Figure S6F,G). As expected, HK2 and HK3 expression were not observed (Figure S6H).
Loss of HKDC1 Expression in the Brain Correlates With Aging in Mice
3.4
Aging is one of the most important hallmarks of AD and related neurodegenerative diseases (NDD). We next wanted to evaluate the expression landscape of HKDC1 in the brain during aging. We observed that HKDC1 protein and RNA expression in the brain of HKDC1^fl/fl^ mice declines with age (Figure 3A–C). Subcellular fractionation of the mouse brain samples showed that, similar to HK1, HKDC1 in the brain exclusively localizes to mitochondria, with no expression in the cytosol (Figure 3D). To further confirm an age‐associated decline in HKDC1 expression in mice, we assessed its expression in the aging mouse model SamP8 and its control SamR1 (Figure 3E). We observed that expression of HKDC1 in SamP8 was considerably lower compared to the control mice SamR1, for both males and females (Figure 3F) (Cui et al. 2024). Additionally, we observed that the expression of HKDC1 in SamP8 mice decreases from 4 to 9 months of age in males and females, strongly suggesting that loss of HKDC1 expression is associated with aging in mice (Figure 3G).
*Loss of HKDC1 in the brain positively correlates with aging in mice. (A) Immunoblot of HKDC1 expression in 4‐ and 9‐Month‐old HKDC1 fl/fl male and female mice (n = 3). Actin was used as a loading control. (B) Densitometric analysis of A. (C) RNA expression of HKDC1 in 9‐Month‐old HKDC1 fl/fl male and female mice (n = 10) using 18S as an internal control. (D) Western blotting of HKDC1 expression in mitochondrial and cytosolic fractions in HKDC1 fl/fl mice. GAPDH was used as a loading control. (E) Schematic showing the aging mouse model Samp8 and the timeline of the analysis of expression of HKDC1. (F) RNA expression of HKDC1 in 9‐Month‐old Samp8 and SamR1 mice (n = 4) using 18S as an internal control. (G) RNA expression of HKDC1 in 4‐ and 9‐month‐old Samp8 mice (n = 4) using 18S as an internal control. Values are mean ± SD; *p < 0.05; **p < 0.01; ***p < 0.001; ***p < 0.0001 by student t‐test with Welchs correction (B, C) or ordinary two‐way ANOVA (F, G). Effect size (r) = 0.658 (d) = 1.75 (C, male), (r) = 0.665 (d) = 1.78 (C, female), (r) = −0.98 (d) = −10.74 (F), (r) = 0.87 (d) = 3.32 (G).
Loss of HKDC1 in the Brain Is Associated With Senescence and Inflammation
3.5
Senescence and inflammation are closely linked to aging and neurodegeneration in both mice and humans. As HKDC1 expression has been recently associated with protection against senescence (Cui et al. 2024), we tested whether age‐dependent loss of HKDC1 expression in mice impacts senescence and inflammation. By evaluating the expression of senescence markers (p21 and p53) and inflammation markers (TNFα, COX‐2, and IL‐1β), we observed that loss of HKDC1 results in an increase in the expression of markers of neuroinflammation (Figure 4A–C, Figure S7A,B) and senescence (Figure 4D,E). Mitochondrial dysfunction is one of the greatest factors that lead to senescence and neuroinflammation. To determine whether mitochondrial dysfunction contributes to senescence and inflammation in our mouse model, we assessed the expression of the mitochondrial health marker PINK1 in BKO and HKDC1^fl/fl^ male and female mice. Our data indicated that PINK1 expression undergoes a significant decline as a result of the loss of HKDC1, suggesting a potential relationship of HKDC1 to mitochondrial health (Figure 4F,G). We also looked at the levels of Aβ‐42 in the serum and observed that the levels between HKDC1^fl/fl^ and BKO were not significantly different, suggesting that our 9‐month‐old mice are tipping towards senescence and neuroinflammation, without measurable neurodegeneration, supporting the notion that the observed changes in inflammation and senescence are attributable to HKDC1 expression and not influenced by Aβ‐42 accumulation (Figure 4H). We also looked at the levels of p‐tau in our protein samples and observed slight expression of p‐tau in BKO samples (Figure 4I). To gain further insights into the effect of HKDC1 loss on autophagy, we performed immunoblotting for LC3‐I and II and observed that LC3‐II expression, which is a marker of autophagosomes, was trending higher in HKDC1‐BKO animals without reaching significance (Figure S7C,D), indicating that the absence of HKDC1 alone at 9 months is not sufficient to upregulate autophagy.
*Loss of HKDC1 in the brain is associated with inflammation and senescence. (A) Schematic representation of mouse genotypes (HKDC1 fl/fl and BKO) and time‐points of analysis of mitochondrial health and neuroinflammation markers. (B) RNA expression of inflammatory markers in 9‐month‐old HKDC1 fl/fl and BKO male mice (n = 6) using 18S as an internal control. Values are mean ± SD; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001 by student t‐test (C) Immunoblot of COX‐2 expression in 9‐month old HKDC1 fl/fl and BKO male and female mice (n = 2). Actin was used as a loading control. (D) Light microscopy images of HKDC1fl/fl and BKO brain sections at 40× magnification. Scale bars, 25 μm. n = 7–9 cells/group from experiments repeated three times. (E) RNA expression of senescence markers in 9‐month‐old HKDC1 fl/fl and BKO male mice (n = 6) using 18S as an internal control. (F, G) Protein and RNA expression of mitochondrial health marker PINK1 in 9‐month‐old HKDC1 fl/fl and BKO male mice (n = 6) using Actin (F) and 18S (G) as an internal control. (H) Bar graph showing relative concentration of Aβ‐42 in the serum of 9‐month‐old HKDC1 fl/fl and BKO mice. APP‐PS1 Alzheimer's disease mouse model (9‐month) was used as a positive control. (I) Western blot of p‐tau expression in 9‐month‐old HKDC1 fl/fl and BKO male mice (n = 4). 5X‐ FAD Alzheimer's disease mouse model (6‐month) was used as a positive control. Values are mean ± SD; *p < 0.05; **p < 0.01; ***p < 0.001; ***p < 0.0001 by student t‐test (A, C) with Welch's correction or ordinary one‐way ANOVA (H). Effect size (r) = −0.63 (d) = −1.63 (TNFα), (r) = −0.75 (d) = −2.27 (COX‐2), (r) = −0.957 (d) = −6.63 (IL‐1β) (B), (r) = −0.68 (d) = −1.85 (p21), (r) = −0.991 (d) = −15.23 (p53) (E), (r) = 0.87 (d) = 3.63 (G).
Loss of Astrocyte‐HKDC1 Leads to Increased ROS Production and Impaired Autophagy
3.6
To investigate the direct effects of HKDC1 loss on mitochondrial health and resulting autophagy, we looked to determine where HKDC1 was most highly expressed. For this, we reanalyzed single‐nuclei RNA sequencing data (data not shown) generated by the Rush Alzheimer's disease center (RADC) from the ROSMAP study (Bennett et al. 2018) and observed that the expression of HKDC1 is highest in astrocytes in the human brain and is significantly reduced in patients with AD compared to individuals with no or mild cognitive impairment (p adj < 0.01) (Figure S8A–C). To confirm that this was also the case in our mouse models, we performed immunofluorescence to examine HKDC1 expression in astrocytes and observed co‐localization between the astrocyte marker GFAP and HKDC1 (Figure S8D). Based on these findings, we employed a human astrocyte cell line CCF‐STTG1 and depleted HKDC1 by using a small interfering RNA for HKDC1 (siHKDC1) in comparison to control (siScr) (Figure 5A–C). As previously seen in the BKO mice, loss of HKDC1 corresponded with a decline in PINK1 expression (Figure 5D) as well as a significantly increased expression of markers of inflammation within astrocytes (Figure 5E–H).
*Loss of HKDC1 expression in CCF‐STTG1 human astrocyte cell line causes inflammation, increase in reactive oxygen species (ROS) production and autophagy. (A) Schematic representation of CCF‐STTG1 astrocyte cells indicating silencing of HKDC1 expression, RNA and protein extraction. (B) Immunoblot of HKDC1 expression in CCF‐STTG1 astrocytes to confirm siRNA mediated silencing of HKDC1. (C) RT‐PCR of HKDC1 in siScr and siHKDC1 CCF‐STTG1 astrocytes using 18S as an internal control (n = 4). (D–G) RNA expression of mitophagy marker PINK1 (D), inflammatory markers IL‐6 (E), S100B (F), IL‐1β (G), and TNF‐α (H) using 18S as an internal control (n = 4). (I) Schematic representation of CCF‐STTG1 silencing with FCCP treatment for measurement of ROS production and autophagy. (J) DCF fluorescence results from siScr and siHKDC1 astrocytes following 4 h treatment with vehicle (DMSO) and 5 μM, 10 μM, and 20 μM FCCP, as compared to 100 μM treatment with TBHP for maximal ROS production. Values are mean ± SD; *p < 0.05; **p < 0.01; ***p < 0.001; ***p < 0.0001 by student t‐test (C–G) with Welch's correction or two‐way ANOVA (J).
Next, to determine if the loss of HKDC1 impacted mitochondrial ROS production and corresponding autophagy, we treated both siScr and siHKDC1 cells with a dosage scale of the mitochondrial decoupler FCCP, as low doses have been shown to effectively decouple mitochondrial membrane potential, leading to increased mitochondrial ROS production without inducing mitophagy, while induction of mitophagy by high doses leads to decreased ROS production (Berezhnov et al. 2016; Brennan et al. 2006). Prior to FCCP treatment, these cells were treated with 2′,7′‐Dichlorodihydrofluorescein Diacetate (DCF), a lipophilic and non‐fluorescent compound that is oxidized to fluorescent DCF (2′,7′‐dichlorofluorescein) in response to ROS exposure, where we observed that not only did loss of HKDC1 increase ROS production alone, but that siHKDC1 astrocytes showed an inability to reduce ROS production in response to high doses of FCCP, suggesting an inability to properly conduct autophagy of decoupled mitochondria (Figure 5I,J).
HKDC1 Expression in the Brain Is Regulated by Transcription Factor EB
3.7
Transcription factor EB (TFEB) is a master transcription regulator that is known for regulating HKDC1 expression (Cui et al. 2024). Previous data from the literature demonstrate that HKDC1 interaction with the mitochondrial protein VDAC is important for maintaining mitochondrial biogenesis, mitochondrial and lysosomal contacts, and the ablation of HKDC1 results in mitochondrial dysfunction and senescence (Cui et al. 2024; Khan et al. 2022). Loss of HKDC1 expression in the brain correlates with increased expression of senescence and neuroinflammation markers. Importantly, loss of PINK1 expression in HKDC1 knock‐out mice suggests a possible involvement of HKDC1 in mitochondrial health, and the PINK1 pathway is known to be controlled by TFEB to maintain mitochondrial biogenesis (Cui et al. 2024; Wang et al. 2021). To understand the role of HKDC1 in this axis, we looked at the promoter of HKDC1 in mice to assess TFEB binding. We observed that the HKDC1 mouse promoter has multiple binding sites for TFEB, possibly controlling its expression levels to control mitochondrial function (Figure 6A). We performed chromatin immunoprecipitation of mouse brain from HKDC1^fl/fl^ female mice at 4 and 9 months of age and observed that TFEB binding to the HKDC1 promoter declines with age, resulting in a decline in HKDC1 expression (Figure 6B,C). We also compared the binding of RNA polymerase 2 (Figure 6G,H) and H3K27 acetylation (Figure 6I,J) at the HKDC1 locus as an indicator of active transcription and observed that RNA pol 2 binding and H3K27ac decline at the HKDC1 locus with aging (Figure 6G–J). Additionally, we observed that the expression of TFEB does not change with age or absence of HKDC1 in the brain, indicating that aging up to 9 months specifically reduces recruitment of TFEB to the HKDC1 promoter without changing its overall level of expression (Figure 6D–F, Figure S7E). Furthermore, TFEB is a known master transcriptional regulator of the autophagy pathway beyond mitophagy. To understand the impact of the age‐associated decline in HKDC1 on TFEB activity and sub‐cellular localization, we performed TFEB immunofluorescence but did not observe a significant change between 4‐ and 9‐month‐old mice. We also failed to observe a detectable amount of signal in the nucleus, indicating that TFEB‐controlled autophagy pathway genes are not upregulated as a result of loss of HKDC1 at 9‐months alone in the absence of an acute physiological insult (Figure S7F).
*Transcription factor EB regulates HKDC1 expression through binding to its promoter. (A) Schematic representation of HKDC1 mouse promoter (−5000 to TSS) highlighting binding sites of transcription factor EB. (B) Schematic representation of chromatin immunoprecipitation and quantitative PCR of 4 and 9‐month‐old HKDC1 fl/fl mice. (C) TFEB ChIP‐ qPCR analysis of HKDC1 mouse promoter in 4‐ and 9‐month‐old HKDC1 fl/fl mice. (D, E) RT‐PCR analysis of TFEB in 9‐month HKDC1 fl/fl and BKO mice (D) and between 4‐ and 9‐month‐old HKDC1 fl/fl mice (E). (F) Western blot of TFEB expression in 4‐ and 9‐month‐old HKDC1 fl/fl and BKO mice (n = 3). β‐actin was used as a loading control. (G) Schematic depicting change in chromatin conformation of HKDC1 promoter with age and exclusion of RNA polymerase binding from it. (H) RNA polymerase II ChIP‐ qPCR analysis of HKDC1 mouse promoter in 4‐ and 9‐month‐old HKDC1 fl/fl mice. (I) Schematic depicting change in chromatin conformation of HKDC1 promoter with age and exclusion of H3K27ac binding from it. (J) H3K27ac ChIP‐ qPCR analysis of HKDC1 mouse promoter in 4‐ and 9‐month‐old HKDC1 fl/fl mice. Values are mean ± SD; *p < 0.05; **p < 0.01; ***p < 0.001; ***p < 0.0001 by student t‐test with Welch's correction. p = 0.1042 (J). Effect size (r) = 0.83 (d) = 2.97 (C), (r) = 0.91 (d) = 4.47 (H).
Discussion
4
Aging and aging‐associated senescence are irreversible physiological processes that can lead to neurodegenerative disorders (NDDs) and cause a huge global socioeconomic burden. Therefore, it is important to understand how we can limit neuronal loss, as seen in aging and NDDs. Abnormal energy metabolism and mitochondrial dysfunction are well‐known hallmarks of aging and neurodegenerative diseases (Belanger et al. 2011; Reeve et al. 2008), which makes it reasonable to study the mechanisms of involvement of glucose metabolism and mitochondrial function in the brain to devise therapies targeting metabolism and mitochondrial function. Transcription factor EB, the master transcription regulator, has previously been reported to be responsible for age‐associated mitochondrial dysfunction in mice (Cui et al. 2024; Reeve et al. 2008; Wang et al. 2021). However, there is a significant gap in our understanding of the full array of proteins that coordinate with TFEB in regulating mitochondrial function and energy metabolism in the brain (Cui et al. 2024; Wang et al. 2021; Sollvander et al. 2016; Frost and Li 2017; Rodriguez‐Giraldo et al. 2022). In this study, we provide the first characterization of the fifth hexokinase enzyme, HKDC1, in relation to its role in brain aging and senescence. HKDC1 expression is significantly lost in AD and aging mouse models. Analysis of human proteomics data from RADC provides supporting evidence of a loss of HKDC1 protein expression correlating with AD‐related pathologies in human brains. Consistent with our finding, downregulation of HKDC1 in AD was also detected using large‐scale deep proteomic profiling and advanced network biology modeling of the parahippocampal gyrus (PHG) recently (Wang et al. 2025).
Additionally, our results show that genetic deletion of HKDC1 at 9 months leads to memory and learning deficits in both sexes and anxiety in male mice (Figure 2 and Figure S3). We also provide evidence that HKDC1 is highly expressed and exclusively localized to mitochondria in the mouse brain, and its expression is controlled by TFEB. Furthermore, we confirmed that aging changes the chromatin architecture of the HKDC1 gene locus, indicated by reduced RNA polymerase II binding and reduced H3K27ac, resulting in loss of HKDC1 expression (Figure 6). We also reported an increase in the expression of senescence markers like p53, p21, and neuroinflammation markers like COX‐2, TNFα, and IL‐1β in our HKDC1 ablated mice (Figure 4) but not in TFEB target genes PGC 1alpha and TFAM (data not shown). While HKDC1 has been shown to be involved in preventing senescence in a TFEB‐dependent manner (Cui et al. 2024), this is the first demonstration that HKDC1 is essential for preventing senescence, inflammation, and oxidative stress in the mouse brain as a direct transcriptional target of TFEB. We also observed an inverse relationship of HKDC1 expression with cognition and episodic memory (Table S4) upon analysis of single‐nucleus RNA Sequencing data from the ROSMAP study (data not shown). Taken together, the loss of HKDC1 expression in the brain is associated with cognitive decline or AD‐related pathology in humans and mice.
Previous studies also reported that TFEB‐mediated senescence and oxidative stress are caused by loss of TFEB expression (Bellenguez et al. 2022). While we did not observe changes in TFEB expression, our study included 4‐ and 9‐month‐old mice, and changes in TFEB expression in other studies with wild‐type mice were seen at 18 months (Bellenguez et al. 2022). This observation suggests that the loss of HKDC1 expression represents an early event in aging, tipping the cells towards senescence and oxidative damage before TFEB expression changes, potentially serving as an early marker for detecting aging and providing a plausible opportunity for arresting the progression of the condition. In the future, it would be interesting to capture the expression and activity (nuclear and cytosolic localization) profiles of TFEB in older control and BKO mice (12 and 18 months).
These mice also show a reduction in expression of the mitophagy marker PINK1 (Figure 4), which is consistent with the role of HKDC1 in stabilizing PINK1 protein on mitochondria to allow PINK1‐parkin mediated healthy mitophagy and prevention of senescence (Cui et al. 2024) and its loss promoting mitochondrial dysfunction and senescence (Figure 4). Furthermore, our mouse RNA sequencing data show that loss of HKDC1 in the brain downregulates pathways related to synaptic signaling, memory, and learning (Figure 2H). Based on our overall findings, we propose that aging results in the modification of local chromatin around the HKDC1 gene, leading to a loss of TFEB recruitment and HKDC1 expression (graphical abstract). Consequently, the loss of HKDC1 prevents the stabilization of PINK1 on mitochondria, causing the accumulation of dysfunctional mitochondria and senescence. In the future, it will be important to dissect the detailed mechanism of the TFEB‐HKDC1‐PINK1 axis in the regulation of aging and neurodegeneration. Interestingly, fasting promotes the upregulation of TFEB in the brain of ketogenic diet (KD)‐treated mice, preventing mitochondrial dysfunction and improving brain health (Bellenguez et al. 2022; Gomora‐Garcia et al. 2023). While age‐related brain dysfunction is well known to be associated with metabolic diseases that stem, at least in part, from mitochondrial dysfunction and a consequent increase in oxidative stress (Milan et al. 2024) and considering that HKDC1 is a direct target of TFEB during glycolysis, studying the role of HKDC1 in time‐restricted and ketogenic diet‐fed mice at different ages would help bolster how HKDC1 can affect mitochondrial function to modulate brain aging as a target of TFEB. It would also be important to investigate age‐associated changes in epigenetic marks on the HKDC1 promoter that trigger changes in chromatin accessibility. Investigating the epigenetic landscape of male and female mice would be equally interesting to understand how 9‐month‐old HKDC1 knockout female mice are selectively protected from the development of anxiety (Figure 2). Studies have previously reported that estrogen and estrogen receptors play a role in maintaining body weight and overall homeostasis in females (Meyer et al. 2011). Studying older and ovariectomized female mice to mimic menopause and gather evidence on the role of female hormones in aging and NDD is an attractive area for future investigation. Estrogen is also known to positively influence cerebral blood flow (CBF) by promoting vasodilation (Choi et al. 2025), a process directly linked to memory and cognition in mice (Bracko et al. 2020). It would be interesting to perform longitudinal assessments in aging control and HKDC1‐BKO male and female mice (Milan et al. 2024) to establish the in vivo role of sex‐specific differences in memory resulting from hormonal differences coupled with HKDC1 perturbation.
Noteworthy, among the different cell types in single cell RNA Sequencing data that we analyzed from RUSH RADC database (data not shown), HKDC1 expression was found to be highest in astrocytes (Figure S8A–C), whereby these are primarily glycolytic cells that supply metabolic substrates to neurons and other cells and play a well‐established role in the removal of neuronal debris (Sollvander et al. 2016). Accumulation of misfolded proteins, including Aβ‐42, also occurs concomitantly with defects in astrocyte function (Frost and Li 2017; Rodriguez‐Giraldo et al. 2022; Sollvander et al. 2016). As HKDC1's role in cancer has previously been associated with increasing glycolytic capacity and anabolic pathway output, a similar role in astrocytes may exist, where a study of the cell‐type specific role of HKDC1 in the brain will help us understand this relationship. To gain insights into this relationship, we ablated HKDC1 expression in the human astrocyte cell line CCF‐STTG1 and observed an increase in markers of inflammation, accompanied by a concomitant decrease in the mitophagy‐related gene PINK1. Additionally, ablation of HKDC1 also led to increased ROS production and a decrease in astrocyte ability to reduce ROS production and increase autophagy in response to mitochondrial decoupling (Figure 5). This suggests that HKDC1 plays a crucial role in increasing autophagy in response to mitochondrial stress; however, further testing would be needed to establish a direct HKDC1‐driven mechanism for autophagic capacity and potential mitophagy.
Overall, our data suggest that HKDC1 expression in the brain is associated with cognitive decline in both humans and mice. TFEB, a known regulator of HKDC1 expression, is diminished in its capacity to maintain HKDC1 expression with aging. From recent studies, the loss of HKDC1 can compromise mitochondrial function and impact senescence and neuroinflammation. Our mouse model data support a similar mechanism occurring here, where more detailed studies with older mice are needed to test this phenomenon.
Author Contributions
Z.F. and B.T.L. designed the research; Z.F. performed the experiments, analyzed, and interpreted the results, and wrote the manuscript with B.L. V.I. performed confocal imaging and senescence assay, Z.F. and J.B. performed in vitro assays, J.B. and J.J. helped analyze single‐nucleus RNA Sequencing data, Y.P. and T.K. performed human brain proteomic data analysis, Z.F. and O.L. performed behavioral experiments, D.B. provided feedback on RNA Sequencing and human proteomic data.
Funding
BTL is supported by a VA Merit Review Award (I01BX003382), NIH grants (R01 DK104927 and U01 DK127378). JJ is supported by Rubenstein Award for Innovative Research into neurodegenerative Disease and Dementia Fund.
Ethics Statement
The human datasets in this study have been acquired after obtaining the required approval by the institutional ethics committee (RUSH University), and AD Knowledgebase Platform. This study has been reviewed and approved by the relevant institutional ethics committee at the University of Illinois at Chicago, ensuring adherence to guidelines for responsible data handling.
Conflicts of Interest
The authors declare no conflicts of interest.
Supporting information
Figures S1–S8: acel70419‐sup‐0001‐FiguresS1‐S8.docx.
Tables S1–S4: acel70419‐sup‐0002‐TablesS1‐S4.docx.
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1Abdullah, A. , A. Kumar , A. Z. Beg , et al. 2025. “Peripherally‐Restricted Recurrent Infection by Engineered E. coli Strain Modulates Hippocampal Proteome Promoting Memory Impairments in a Rat Model.” Gene 933: 148969. 10.1016/j.gene.2024.148969.39341518 · doi ↗ · pubmed ↗
- 22024. “2024 Alzheimer's Disease Facts and Figures.” Alzheimer's & Dementia 20, no. 5: 3708–3821. 10.1002/alz.13809.PMC 1109549038689398 · doi ↗ · pubmed ↗
- 3Arrieta‐Cruz, I. , and R. Gutierrez‐Juarez . 2016. “The Role of Insulin Resistance and Glucose Metabolism Dysregulation in the Development of Alzheimer s Disease.” Revista de Investigación Clínica 68, no. 2: 53–58.27103040 · pubmed ↗
- 4Bae, J. R. , and S. H. Kim . 2017. “Synapses in Neurodegenerative Diseases.” BMB Reports 50, no. 5: 237–246. 10.5483/bmbrep.2017.50.5.038.28270301 PMC 5458673 · doi ↗ · pubmed ↗
- 5Banday, S. , Z. Farooq , R. Rashid , E. Abdullah , and M. Altaf . 2016. “Role of Inner Nuclear Membrane Protein Complex Lem 2‐Nur 1 in Heterochromatic Gene Silencing.” Journal of Biological Chemistry 291, no. 38: 20,021–20,029. 10.1074/jbc.M 116.743211.PMC 502568827451393 · doi ↗ · pubmed ↗
- 6Belanger, M. , I. Allaman , and P. J. Magistretti . 2011. “Brain Energy Metabolism: Focus on Astrocyte‐Neuron Metabolic Cooperation.” Cell Metabolism 14, no. 6: 724–738. 10.1016/j.cmet.2011.08.016.22152301 · doi ↗ · pubmed ↗
- 7Bellenguez, C. , F. Kucukali , I. E. Jansen , et al. 2022. “New Insights Into the Genetic Etiology of Alzheimer's Disease and Related Dementias.” Nature Genetics 54, no. 4: 412–436. 10.1038/s 41588-022-01024-z.35379992 PMC 9005347 · doi ↗ · pubmed ↗
- 8Bennett, D. A. , A. S. Buchman , P. A. Boyle , L. L. Barnes , R. S. Wilson , and J. A. Schneider . 2018. “Religious Orders Study and Rush Memory and Aging Project.” Journal of Alzheimer's Disease 64, no. s 1: S 161–S 189. 10.3233/JAD-179939.PMC 638052229865057 · doi ↗ · pubmed ↗