Abstract

   This study investigates NADPH oxidase 4 (NOX4) involvement in
   iron-mediated astrocyte cell death in Alzheimer’s Disease (AD) using
   single-cell sequencing data and transcriptomes. We analyzed AD
   single-cell RNA sequencing data, identified astrocyte marker genes, and
   explored biological processes in astrocytes. We integrated AD-related
   chip data with ferroptosis-related genes, highlighting NOX4. We
   validated NOX4’s role in ferroptosis and AD in vitro and in vivo.
   Astrocyte marker genes were enriched in AD, emphasizing their role.
   NOX4 emerged as a crucial player in astrocytic ferroptosis in AD.
   Silencing NOX4 mitigated ferroptosis, improved cognition, reduced Aβ
   and p-Tau levels, and alleviated mitochondrial abnormalities. NOX4
   promotes astrocytic ferroptosis, underscoring its significance in AD
   progression.

Supplementary Information

   The online version contains supplementary material available at
   10.1186/s13578-024-01266-w.

   Keywords: Single-cell sequencing, Ferroptosis, NADPH oxidase 4,
   Astrocytes, Alzheimer’s disease, Differential gene analysis,
   Immunofluorescence staining, Mouse model validation

Introduction

   Alzheimer’s disease (AD) is a prevalent neurodegenerative condition
   characterized by memory loss, cognitive decline, and impaired
   behavioral abilities [[35]1–[36]3]. Despite some advances in
   understanding AD’s pathological mechanisms, the specific molecular
   underpinnings remain elusive, and effective treatments targeting its
   root causes are lacking [[37]1, [38]4–[39]6]. A comprehensive
   investigation into AD’s pathogenesis holds paramount importance for its
   prevention and treatment. In recent years, ferroptosis, an emerging
   form of cell death reliant on iron, has garnered significant attention
   in the realm of neurodegenerative diseases [[40]7–[41]10]. Astrocytes
   are one of the most abundant subtypes of glial cells in the central
   nervous system [[42]11, [43]12]. They are associated with brain
   development and function, such as regulating synaptic formation and
   function, controlling neurotransmitter release and uptake, producing
   trophic factors, and maintaining neuronal survival [[44]13–[45]16]. As
   major glial cells in the central nervous system, astrocytes also play a
   role in various physiological and pathological processes in the brain
   and may have significant implications for the pathogenesis of AD
   [[46]17–[47]19]. Previous studies have indicated higher levels of
   astrocyte damage in Alzheimer’s disease patients [[48]20].

   Ferroptosis, characterized by heightened lipid peroxidation and
   iron-dependent reactive oxygen species generation, constitutes a novel
   iron-dependent cell demise pathway [[49]21–[50]23]. There is a close
   relationship between oxidative stress and iron death. Both iron excess
   and deficiency can induce oxidative stress, leading to cell death and
   other related diseases [[51]24]. Studies suggest that iron death may
   play a significant role in neurodegenerative diseases, including
   Alzheimer’s disease [[52]25]. Furthermore, literature indicates the
   presence of high concentrations of iron in the brains of AD patients
   and transgenic mouse models, where excess iron can exacerbate oxidative
   damage and cause cognitive impairment. Disruption of iron homeostasis
   is considered to be associated with Alzheimer’s disease [[53]26,
   [54]27]. Increasing evidence indicates that iron death can lead to
   AD-mediated neuronal cell death [[55]28]. Nevertheless, iron’s precise
   roles and regulatory mechanisms in Alzheimer’s, particularly its
   interplay with astrocytes, remain enigmatic [[56]29]. Star-shaped glial
   cells play pivotal roles in the central nervous system’s physiological
   and pathological processes, with certain key regulatory genes or
   pathways potentially offering valuable insights into Alzheimer’s
   pathogenesis [[57]30–[58]32].

   Single-cell sequencing technology, noted for its remarkable sensitivity
   and high resolution, has gained widespread adoption in biomedical
   research [[59]33–[60]35]. This innovative approach facilitates the
   analysis of gene expression and regulatory networks at the single-cell
   level, revealing cellular heterogeneity and dynamic changes under
   physiological and pathological conditions [[61]36]. Single-cell
   sequencing affords the capacity to dissect distinct neural cell types,
   including neurons and glial cells, and their roles in disease
   progression, offering invaluable insights into neurodegenerative
   conditions such as Alzheimer’s [[62]37].

   NADPH oxidase 4 (NOX4), an enzyme with protein catalytic activity that
   generates reactive oxygen species (ROS), exerts pivotal roles in
   various physiological and pathological processes encompassing cell
   proliferation, migration, and cell death [[63]38–[64]40]. Current
   research highlights NOX4’s potential significance in neurodegenerative
   diseases, including Parkinson’s and Alzheimer’s diseases. Nonetheless,
   the precise mechanisms by which NOX4 influences Alzheimer’s,
   particularly its association with astrocytic iron-mediated cell death,
   remain to be fully elucidated [[65]41, [66]42].

   This study’s primary objective is to elucidate the critical role of
   NADPH oxidase 4 (NOX4) in iron-triggered astrocytic cell death and its
   implications for Alzheimer’s disease (AD) pathogenesis. Employing
   single-cell sequencing technology, the GEO database, and transcriptome
   sequencing data, this study delves deeply into the cellular populations
   and associated genes of AD patients. Researchers meticulously label and
   analyze various cell types, identifying distinctive marker genes for
   astrocytes. Furthermore, this study uncovers NOX4’s pivotal role in
   iron-induced astrocyte demise. These findings substantially augment our
   comprehension of Alzheimer’s disease etiology, particularly elucidating
   the nexus between NOX4, astrocytic iron-mediated death, and AD.
   Importantly, these insights hold clinical relevance for the diagnosis
   and treatment of Alzheimer’s disease.

Materials and methods

Transcriptome sequencing data acquisition

   The single-cell transcriptome sequencing data of AD-related samples in
   the [67]GSE164089 dataset was analyzed using the Seurat package in R
   software. To ensure data quality, quality control criteria were
   applied, including nFeature_RNA > 500, 1000 < nCount_RNA < 20,000, and
   percent.mt < 10%. Additionally, the top 1000 highly variable genes were
   selected based on their variance. Furthermore, AD-related microarray
   dataset [68]GSE48350 was obtained from the GEO database
   ([69]https://www.ncbi.nlm.nih.gov/geo/). This dataset consists of 173
   normal brain tissue samples and 80 AD brain tissue samples [[70]43].

TSNE clustering analysis

   To reduce the dimensionality of scRNA-Seq datasets, we employ principal
   component analysis (PCA) based on the top 1000 genes with the highest
   variance in expression. We used the Elbowplot function of the Seurat
   package and selected the top 15 principal components for downstream
   analysis. Using the FindClusters function provided by Seurat, we
   identified different subpopulations of cells with the default
   resolution (res = 0.5). Next, we use the t-SNE algorithm to reduce
   nonlinear dimensionality on scRNA-seq sequencing data. We also used the
   Seurat package to identify marker genes for individual cell
   subpopulations and combined the single package with the online website
   CellMarker ([71]http://xteam.xbio.top/CellMarker) for cell type
   annotation analysis [[72]44, [73]45].

GO and KEGG enrichment analysis

   The differential expression genes (DEGs) were subjected to Gene
   Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG)
   enrichment analysis using the “clusterProfiler”, “org.Hs.eg.db”,
   “enrichplot”, and “ggplot2” packages in the R language. Bubble plots
   and circular plots were generated to visualize the enrichment results
   of the three categories, namely Biological Processes (BP), Cellular
   Components (CC), and Molecular Functions (MF), in the Gene Ontology
   (GO). Additionally, a bubble plot was generated to display the
   enrichment results of the KEGG pathway analysis [[74]46].

Differential gene expression screening

   The “limma” package in R software was utilized for the selection of
   differentially expressed genes. Differentially expressed genes between
   normal samples and AD samples were filtered based on the criteria
   |logFC| > 0 and P.adjust < 0.05 [[75]47].

Lentivirus infection

   To construct a lentivirus-mediated NOX4 silencing vector, the
   pSIH1-H1-copGFP (sh-) interference vector (catalog number SI501A-1,
   System Biosciences, USA) was purchased. The silencing sequence can be
   found in Table [76]S1. The lentiviral particles carrying the vector
   were packaged into HEK-293T cells (CRL-3216, ATCC, USA) using the
   lentivirus packaging reagent kit (catalog number A35684CN, Invitrogen,
   USA). After 48 h, the supernatant was collected to obtain lentivirus
   with a titer of 1 × 10^8 TU/ml. Researchers interested in rapidly and
   efficiently constructing a lentiviral vector that mediates NOX4
   silencing may consider adopting the methodology utilized in our
   laboratory [[77]48, [78]49].

Cell culture and screening

   Human normal astrocytes were purchased from ATCC (ATCC, USA) and
   cultured in human astrocyte medium (catalog number 1801, ScienCell,
   USA). The medium consisted of a basal medium (catalog number 1801), 2%
   (v/v) fetal bovine serum (FBS, catalog number 0010), 1% (v/v) astrocyte
   growth supplement (AGS, catalog number 1852), and 1% (v/v)
   penicillin/streptomycin solution (P/S, catalog number 0503). The cells
   were cultured at 37 °C in a 5% CO2 incubator. To simulate Aβ-induced
   neuronal injury, the cells were treated with Aβ25–35 peptide (catalog
   number A107853-25 mg, Aladdin, Shanghai, China) at a concentration of
   20 μM for 24 h. The cells were divided into the following groups:
   control, AD, AD + sh-NC (infected with negative control lentivirus
   expressing sh-NC), and AD + sh-NOX4 (infected with lentivirus
   expressing sh-NOX4). After adding 1 × 10^5 TU lentivirus to the
   astrocytes, the cells were incubated for 48 h, except for the control
   group, which was incubated for an additional 24 h in the medium
   containing 20 μM Aβ25–35 peptide [[79]50, [80]51].

Alzheimer’s disease APP/PS1 mouse model

   The male transgenic mice with overexpression of human amyloid precursor
   protein (APP) and mutant forms of presenilin 1 (PS1) were purchased
   from Jackson Laboratory (Bar Harbor, ME, USA, Stock #034829). The
   wild-type C57 male mice were purchased from Weitonlihua Experimental
   Animal Technology Co., Ltd. in Beijing, China for the Alzheimer’s
   disease model experiments. They were maintained under non-pathogenic
   conditions at a temperature of 26–28 ℃ and humidity of 50–65%, with
   free access to food and water. All mice were acclimated for one week
   prior to the experiments. The experimental procedures were conducted in
   accordance with ethical standards and were approved by our
   institution’s Animal Ethics Committee.

   For the experiments, the mice were randomly divided into five groups:
   WT, APP/PS1, APP/PS1 + sh-NC, APP/PS1 + sh-NOX4, and
   APP/PS1 + sh-NOX4 + erastin, with six mice in each group. To silence
   NOX4 in the neurons in vivo, sh-NOX4 (4 × 10^5 TU) or sh-NC (4 × 10^5
   TU) was slowly injected into the bilateral hippocampi of APP/PS1 mice.
   In the APP/PS1 + sh-NOX4 + erastin group, erastin (10 μM; HY-15,763,
   MedChem Express, New Jersey, USA) was dissolved in a water bath at 37℃
   with gentle shaking, and then 5% dimethyl sulfoxide with corn oil
   (C8267, Sigma-Aldrich, USA) was added. The mice were treated for 20
   days. At the end of the experiment, all mice were euthanized with an
   overdose of anesthetic (pentobarbital sodium). The tissues were
   processed by perfusing the ascending aorta with a 0.9% sodium chloride
   solution, followed by fixation of the brain tissues in 4%
   paraformaldehyde solution and embedding in paraffin [[81]38,
   [82]52–[83]54]. The behavior test experiment was finished, and the
   above processing methods were executed.

Immunofluorescent staining

   The brain tissue embedded in paraffin was sectioned into 4 μm thick
   slices and permeabilized using 0.5% Triton-X (T8787, Sigma-Aldrich,
   USA). The slices were then blocked in CAS-Block™ tissue blocking
   reagent (008120, Thermo Fisher Scientific, Waltham, MA, USA).
   Immunostaining was performed using the following antibodies: rabbit
   anti-GFAP antibody (ab207165, Abcam, Cambridge, UK), rabbit anti-NOX4
   antibody (MA5-32090, 1:50, ThermoFisher, USA), rabbit anti-4-HNE
   antibody (MA5-45789, 1:50, ThermoFisher, USA), and rabbit
   anti-malondialdehyde (MDA) antibody (MA5-45803, 1:50, ThermoFisher,
   USA), as well as rabbit monoclonal anti-GFAP antibody (ab207165, Abcam,
   Cambridge, UK). The slices were then incubated with secondary
   antibodies goat anti-rabbit IgG (H + L) Alexa Fluor 488 (A11008, 1:100,
   Thermo Fisher Scientific, USA) and goat anti-mouse IgG (H&L) Texas Red
   (ab6787, 1:100, Abcam, Cambridge, UK) at 25 °C for 2 h. Nuclear
   staining was performed using Fluoroshield™ with DAPI (F6057,
   Sigma-Aldrich, USA). The stained brain tissue sections were analyzed
   using THUNDER Imager Tissue (Leica Microsystems Ltd., Wetzlar, Germany)
   and quantified using LAS X imaging software (Leica Microsystems Ltd,
   Wetzlar, Germany) and ImageJ software v1.52a (Bethesda, MD, USA)
   [[84]38].

Flow cytometric cell sorting

   The mouse brain tissue samples were cut into small pieces and digested
   at 37 °C in PBS solution containing 0.8 mg/mL Collagenase IV (Merck,
   C4-BIOC, USA). After a wash with PBS buffer, the cell suspension was
   filtered through a 50 μm sieve to remove residual solid components such
   as organic matter and neurons. The cell suspension was then centrifuged
   at 1000 rpm for 5 min. The supernatant was discarded, and the cell
   pellet was retained. Washing with cell culture medium was performed,
   followed by grinding the remaining tissue cell clusters using a
   homogenizer. After repeated washes, the cell suspension was added to
   approximately 1 mL of washing buffer for selection. Subsequently, the
   cell suspension was placed in a column containing magnetic beads
   labeled with S100β antibody.

   In the spatial sliding column, cells were bound to the S100β antibody
   (# 9550 S, Cell Signaling Technology, Danvers, MA, USA). Through
   negative selection, non-stellar glial cells were removed while stellar
   glial cells were retained. Subsequent washing steps were performed to
   eliminate cells and impurities not bound to the magnetic beads. The
   cell suspension was then transferred to a sterile culture dish for
   inspecting the integrity and activity of stellar glial cells under a
   microscope. After identifying and collecting the target cell samples,
   the cells were fixed using PBS solution containing 3.7% formaldehyde.
   Permeabilization of the fixed cells was done using 0.1% Triton X-100,
   followed by labeling of the stellar glial cells using GFAP antibody
   (ab207165, Abcam, Cambridge, UK). The labeled cell samples were
   injected into a flow cytometer, and cell fluorescence intensity was
   measured by laser excitation to obtain a purity of 90% for stellar
   glial cells. Finally, the collected stellar glial cells were subjected
   to Western blot, lipid peroxidation detection, and measurement of iron
   ion content using the same culture conditions [[85]55].

Western blot

   Tissue total protein was extracted using RIPA lysis buffer (P0013C,
   Beyotime, Shanghai, China) containing PMSF. The extraction process
   involved incubation on ice for 30 min, followed by centrifugation at
   4 °C and 8000 g for 10 min to collect the supernatant. The total
   protein concentration was measured using a BCA assay kit (Catalog
   number: 23,227, ThermoFisher, USA). 50 μg of protein was dissolved in
   2x SDS loading buffer and boiled for 5 min at 100 °C prior to SDS-PAGE
   gel electrophoresis. The proteins were then transferred to a PVDF
   membrane.

   The PVDF membrane was blocked with 5% non-fat milk at room temperature
   for 1 h, followed by incubation overnight at 4 °C with the diluted
   primary antibodies: rabbit anti-GPX4 (ab125066, 1:1000, Abcam,
   Cambridge, UK), rabbit anti-NOX4 (MA5-32090, 1:1000, ThermoFisher,
   USA), rabbit anti-4-HNE antibody (MA5-45789, 1:1000, ThermoFisher,
   USA), rabbit anti-malondialdehyde (MDA) antibody (MA5-45803, 1:1000,
   ThermoFisher, USA), rabbit anti-SAT1 antibody (ab105220, 1:500, Abcam,
   Cambridge, UK), rabbit anti-FTH1 antibody (PA5-27500, 1:500,
   ThermoFisher, USA), rabbit anti-SLC7A11 (711,589, 1:1000, ThermoFisher,
   USA), rabbit anti-ACSL4 (PA5-27137, 1:1000, ThermoFisher, USA), and
   rabbit anti-GAPDH (ab181602, 1:10000, Abcam, Cambridge, UK) as internal
   references. Subsequently, the membrane was washed three times with TBST