Abstract
The insulin/IGF-1 signaling (IIS) regulates a wide range of biological processes, including aging and lifespan, and has also been implicated in the pathogenesis of Alzheimer's disease (AD). We and others have reported that reduced signaling by genetic ablation of the molecules involved in IIS (e.g., insulin receptor substrate 2 [IRS-2]) markedly mitigates amyloid plaque formation in the brains of mouse models of AD, although the molecular underpinnings of the amelioration remain unsolved. Here, we revealed, by a transcriptomic analysis of the male murine cerebral cortices, that the expression of genes encoding extracellular matrix (ECM) was significantly upregulated by the loss of IRS-2. Insulin signaling activity negatively regulated the phosphorylation of Smad2 and Smad3 in the brain, and suppressed TGF-β/Smad-dependent expression of a subset of ECM genes in brain-derived cells. The ECM proteins inhibited Aβ fibril formation in vitro, and IRS-2 deficiency suppressed the aggregation process of Aβ in the brains of male APP transgenic mice as revealed by injection of aggregation seeds in vivo. Our results propose a novel mechanism in AD pathophysiology whereby IIS modifies Aβ aggregation and amyloid pathology by altering the expression of ECM genes in the brain.
SIGNIFICANCE STATEMENT The insulin/IGF-1 signaling (IIS) has been recognized as a regulator of aging, a leading risk factor for the onset of Alzheimer's disease (AD). In AD mouse models, genetic deletion of key IIS molecules markedly reduces the amyloid plaque formation in the brain, although the molecular underpinnings of this amelioration remain elusive. We found that the deficiency of insulin receptor substrate 2 leads to an increase in the expression of various extracellular matrices (ECMs) in the brain, potentially through TGF-β/Smad signaling. Furthermore, some of those ECMs exhibited the potential to inhibit amyloid plaque accumulation by disrupting the formation of Aβ fibrils. This study presents a novel mechanism by which IIS regulates Aβ accumulation, which may involve altered brain ECM expression.
Introduction
Alzheimer's disease (AD) is the most frequent cause of dementia in the elderly. The brains of patients with AD show progressive neurodegeneration with neuropathological changes, including accumulation of senile plaques and neurofibrillary tangles, composed of amyloid-β (Aβ) peptides and tau protein, respectively. The findings derived from rare autosomal dominant forms of AD, along with neuropathological observations, biomarker evidence, and experimental data from cell culture and animal studies, together support the amyloid hypothesis (i.e., the idea that Aβ is the causal factor of the pathogenic cascade of AD) (Selkoe and Hardy, 2016).
Aging is the greatest risk factor of AD (Hou et al., 2019). One of the major molecular pathways implicated in the regulation of aging is the insulin/insulin-like growth factor-1 (IGF-1) signaling (IIS) (Kenyon, 2010). In mammalian IIS, signals arising from ligand-activated insulin and IGF-1 receptors are transduced via insulin receptor substrate (IRS) proteins to downstream intracellular effectors, such as PI3K/Akt. Inhibition of IIS by gene disruption or pharmacological intervention extends lifespan across multiple animal species (e.g., nematodes, flies, and mice) and affects the expression of transcription factors that regulate genes related to anti-aging effects (e.g., stress responses and chaperones) (Kenyon, 2010). Furthermore, in mammals, IIS in the brain has also been implicated in the regulation of lifespan and aging (Taguchi et al., 2007; Kappeler et al., 2008). Thus, IIS may be a key molecular pathway linking aging and AD pathogenesis. Accordingly, reducing IIS has been shown to diminish Aβ42 toxicity in a Caenorhabditis elegans model (Cohen et al., 2006). Studies in AD model mice carrying mutant APP have also demonstrated that genetic inactivation of one of the IIS components, IRS-2 (Freude et al., 2009; Killick et al., 2009; Wakabayashi et al., 2019), IGF-1 receptor (IGF-1R) (Freude et al., 2009; Gontier et al., 2015), or insulin receptor (Murakami et al., 2011; Stöhr et al., 2013), reduces Aβ accumulation in the brain. Deletion of IRS-2 (Killick et al., 2009) and IGF-1R (Cohen et al., 2009; Gontier et al., 2015) has been shown to ameliorate behavioral impairments, too. We have shown that, unlike IRS-2 deficiency, IRS-1 deficiency does not alter either the insulin signaling activity or Aβ deposition in the brains of AD model mice, providing further support for a link between the brain IIS activity and Aβ pathology (Ochiai et al., 2021).
These observations together led to the consensus that reduced IIS inhibits the formation of amyloid pathology. However, it has not been clarified what are the key changes in the amyloidogenic pathomechanisms of Aβ caused by the reduction in IIS. Some reports have attributed the decrease in Aβ levels to reduced Aβ production, while others suggest altered Aβ clearance (Freude et al., 2009; Killick et al., 2009; Stöhr et al., 2013; Gontier et al., 2015). However, no reports have clearly demonstrated the mechanism of changes in brain Aβ dynamics caused by reduced IIS in vivo. Also, it remains unsolved which molecular pathways downstream of IIS are directly involved in the suppression of Aβ accumulation. Importantly, comparative studies on amyloid pathology of the impact of deficiency of the neuronal insulin receptor and IGF-1R, as well as of IRS-2 and IRS-1, have suggested that the protective effects of reduced IIS on AD pathology may be regulated by unknown molecular pathway(s) distinct from the canonical IIS downstream involved in lifespan extension (Stöhr et al., 2013; Ochiai et al., 2021).
In this study, we aimed to elucidate the molecular mechanisms underlying the inhibitory effect of reduced IIS on Aβ accumulation. RNA sequencing (RNA-Seq) analysis of the brains of APP transgenic mice identified a group of extracellular matrix (ECM) genes that were upregulated by IRS-2 deletion, possibly through activation of the TGF-β/Smad signaling. These ECM molecules exhibited an inhibitory effect on Aβ aggregation in vitro, and seed-induced Aβ accumulation was suppressed in the absence of IRS-2 in vivo. Together, our findings propose a novel link between the IIS and Aβ pathology via alteration in the brain ECM.
Materials and Methods
Mice
A7 mice are transgenic mice overexpressing human APP695 harboring familial AD mutations (KM670/671NL and T714I) in neurons under the control of Thy1.2 promoter (Yamada et al., 2009). A7 mice were backcrossed and maintained on a C57BL/6J background. Irs2−/− and Irs1−/− mice were generated as previously described (Tamemoto et al., 1994; Kubota et al., 2000). Irs2+/− and Irs1+/− mice were crossed with A7 mice to obtain Irs2−/−;A7 and Irs1−/−;A7 mice, and littermate controls were used. Animals were maintained on a 12 h light/dark cycle and provided ad libitum access to food and water. The animal care and experimental procedures were approved by the animal experiment committee of the University of Tokyo Graduate School of Medicine.
RNA-Seq and analysis
Nine-month-old male A7 and Irs2−/−;A7 mice were sacrificed after fasting for 4-6 h. The cerebral cortices were excised, snap frozen in liquid nitrogen, and stored at −80°C. Total RNA was extracted from the cerebral cortices with TRIzol Plus RNA Purification Kit (Thermo Fisher Scientific) according to the manufacturer's instructions. DNase (2270A, Takara Bio) treatment was performed for digestion of genomic DNA, followed by purification of total RNA with PureLink RNA Mini Kit (Thermo Fisher Scientific). Total RNA quality was assessed by Agilent Bioanalyzer 2100 (Agilent Technologies) using Agilent RNA 6000 Pico Kit (Agilent Technologies). The concentration of total RNA was measured by Qubit 2.0 Fluorometer (Thermo Fisher Scientific) using Qubit RNA BR Assay Kit (Thermo Fisher Scientific). Strand-specific mRNA library preparation, sequencing with HiSeq 2500 (2 × 100 bp) (Illumina), read cleaning with trimmomatic (version 0.36), mapping to the mouse reference genome (mm10) with BWA (version 0.7.17), and differential expression analysis of TMM normalized counts with edgeR (version 3.16.1) were performed (Eurofins Genomics). Likelihood ratio test was used for differential expression analysis. A false discovery rate (FDR) was set to <0.05. A volcano plot was generated using Galaxy (version 22.01). Enrichment analysis was performed using Gene Set Enrichment Analysis (GSEA) software (version 3.0; https://www.gsea-msigdb.org/gsea/) (Mootha et al., 2003; Subramanian et al., 2005). Gene sets database and chip platform used for this study were c5.all.v6.1.symbols.gmt and ENSEMBL_mouse_gene.chip. The number of permutations was 1000.
qRT-PCR
Mice were sacrificed after fasting for 4-6 h. The cerebral cortices and the hippocampi were excised, snap frozen in liquid nitrogen, and stored at −80°C. Total RNA was extracted from the tissues or isolated brain cells with TRIzol Plus RNA Purification Kit and from primary cultured cells with PureLink RNA Mini kit according to the manufacturer's instructions. RNA purity and concentration were measured with NanoDrop (Thermo Fisher Scientific). Total RNA was reverse-transcribed into cDNA using ReverTra Ace qPCR RT Master Mix with gDNA Remover (Toyobo). qRT-PCR was performed on the LightCycler 480 System II (Roche Diagnostics) using THUNDERBIRD SYBR qPCR Mix (Toyobo). Threshold cycle values were normalized to Hprt and Sdha for tissues and primary cultured cells and to Actb for isolated brain cells. The primer pairs used in this study are as follows: 5′-CAGTCCCAGCGTCGTGATTA-3′ and 5′-GGCCTCCCATCTCCTTCATG-3′ for Hprt; 5′-TCGACAGGGGAATGGTTTGG-3′ and 5′-TAATCTTCCCTGGCATGGGC-3′ for Sdha; 5′-ACTGAGCTGCGTTTTACACC-3′ and 5′-AGCCATGCCAATGTTGTCTC-3′ for Actb; 5′-CTTCCAGTTCTTCCTCCAAAGC-3′ and 5′-GCACAGATTCCAGGTCGTTATG-3′ for Ogn; 5′-AAGGACGTGGTCACTCTGAAC-3′ and 5′-TGCTGGTTTCGGAGGTATTGG-3′ for Fmod; 5′-TGGAATTTGGACACGTTGGC-3′ and 5′-TAGCGCTCCACATCCTTTGTC-3′ for Myoc; 5′-AGACGCCTGCCCAATTAATG-3′ and 5′-ATCGCCAACACAGTCTTTGC-3′ for Thbs1; 5′-GACCCTGCCCAGATGGTTAC-3′ and 5′-TCACATCTGAAACCCGGAGC-3′ for Thbs4; 5′-TTGGCTGGCTCTTTGCTTTG-3′ and 5′-TCGGCAGTTTGAGTGATACG-3′ for Nid2; 5′-CAGATGGCGAATACTGGGTC-3′ and 5′-TTTGGGCCAAGAAGTGATTCTG-3′ for Col5a1; 5′-AGCGGGTACACCAGGAAAAG-3′ and 5′-ACCATCATTACCAGCCGGAC-3′ for Col5a2; 5′-ACATACCGGCGCAATTTCAC-3′ and 5′-AGCCTGGCACTCAAAAGAAC-3′ for Col6a1; 5′-ACTCCATTGCCTGTGACAAG-3′ and 5′-TGTTTGGCAGGGAAGTTCTG-3′ for Col6a2; 5′-GCAACGTTTTCAAGCGGATG-3′ and 5′-ATAAGCAAAGACGGCAAGGC-3′ for Col6a5; 5′-CTCCCAGATCTGACGTGCTC-3′ and 5′-GCCATCACATTTAGGCTTGGC-3′ for Col8a1; 5′-GGTAAAGTATGTGCAGCCCATG-3′ and 5′-CATCGGTAGAGGCATTTCCAAG-3′ for Col8a2; 5′-AAATTGGTCCACAGGGCATC-3′ and 5′-ATTGGCCCTTCTCTCCTTTCTC-3′ for Col9a2; 5′-AAAGGCGAGAAGGGAGAACC-3′ and 5′-TCCTGTGCTGCCTTTTTGTC-3′ for Col9a3; 5′-GACATTGGTGTTGGCATTGC-3′ and 5′-TGCACCCATCTTGCCATAAC-3′ for Col16a1; 5′-ACAAAGTGTGCAGGTGGATG-3′ and 5′-TCAAGTTCAATGCCCGTAGC-3′ for Tnxb.
Isolation of brain cell populations
The detailed experimental procedures have been described previously (Ochiai et al., 2021). Briefly, after transcardial perfusion with Hanks balanced salt solution without calcium or magnesium (HBSS, Thermo Fisher Scientific), the cerebral cortices and hippocampi of 11-month-old C57BL/6J mice and Irs2−/− mice were harvested and kept in ice-cold HBSS. Small pieces of the tissue were dissociated with papain (Worthington) at a final concentration of 34 U/ml in Hibernate A minus calcium medium (BrainBits) with 0.5 mm Glutamax (Thermo Fisher Scientific) for 30 min at 37°C with shaking at 170 rpm. After the centrifugation of the mix, the pellet was resuspended in Dulbecco's PBS (Thermo Fisher Scientific) to obtain a single-cell suspension. Neurons, astrocytes, and microglia were magnetically isolated using Neuron Isolation Kit (neuron), Anti-ACSA-2 MicroBead Kit (astrocyte), CD11b (microglia) MicroBeads, and MACS Separator (Miltenyi Biotec) after the removal of cell debris and erythrocytes.
Protein extraction
Mice were sacrificed after fasting for 4-6 h. The cerebral cortices and the hippocampi were excised, snap frozen in liquid nitrogen, and stored at −80°C. For Aβ ELISA and immunoblotting, brain tissues were homogenized in a 1:10 (w/v) volume of TBS and centrifuged at 347,600 × g for 20 min at 4°C. The supernatants were stored as TBS-soluble fractions. The resulting pellets were homogenized in a 1:10 (w/v) volume of 2% Triton X-100 in TBS and centrifuged at 347,600 × g for 20 min at 4°C, and the supernatants were stored as TX-soluble fractions. For immunoblotting, brain tissues were homogenized in a 1:10 (w/v) volume of RIPA buffer (1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS in TBS) and centrifuged at 17,800 × g for 20 min at 4°C. The supernatants were stored as RIPA-soluble fractions. The resulting pellets were homogenized in a 1:5 (w/v) volume of 6 m urea in RIPA buffer and centrifuged at 347,600 × g for 20 min at 4°C. The supernatants were stored as urea-soluble fractions. Protein concentration was measured with BCA protein assay kit (Takara Bio). All the buffers contained cOmplete protease inhibitor (Merck) and PhosSTOP phosphatase inhibitor cocktails (Merck).
Immunoblot analysis
Brain homogenates were mixed with Laemmli sample buffer and 2-mercaptoethanol whose final concentration was 1% and boiled for 5 min at 95°C. Proteins were separated by SDS-PAGE under a reducing condition using a Tris-glycine gel system, transferred to PVDF membranes (Merck), and reacted with antibodies. The signals were visualized with ImmunoStar Reagents (Fujifilm Wako) or SuperSignal West Femto (Thermo Fisher Scientific), and detected with ImageQuant LAS 4000 mini (Fujifilm). Relative levels of signal intensity were analyzed by Image Quant TL (Fujifilm) or ImageJ software (National Institutes of Health). The antibodies used in this study were as follows: anti-Thrombospondin1 (ab85762, Abcam), anti-Thrombospondin4 (sc-390734, Santa Cruz Biotechnology), anti-COL6A1 (sc-377143, Santa Cruz Biotechnology), anti-α-tubulin (DM1A, Merck), anti-phospho-Smad2 (Ser465/467) (138D4, 3108, CST), anti-Smad2 (D43B4, 5339, CST), anti-phospho-Smad3 (Ser423/425) (ab52903, Abcam), anti-Smad3 (C67H9, 9523, CST), anti-phospho-FoxO1(Ser256) (9461, CST), anti-FoxO1 (C29H4R, 2880, CST), anti-phospho-FoxO3a (Ser253) (9466, CST), anti-FoxO3a (75D8, 2497, CST), anti-phospho-4E-BP1 (Thr37/46) (9459, CST), anti-4E-BP1 (9452, CST), anti-phospho-p70 S6 kinase (Thr389) (9205, CST), anti-p70 S6 kinase (9202, CST), anti-Nrf2 (ab62352, Abcam), anti-heme oxygenase-1 (ADI-SPA-895, Enzo Life Sciences), anti-KEAP1 (D6B12, 8047, CST), anti-heat shock transcription factor 1 (HSF-1) (ADI-SPA-901, Enzo Life Sciences), anti-HSP70/HSP72 (C92F3A-5, ADI-SPA-810, Enzo Life Sciences), anti-Hsc70 (10654-1-AP, Proteintech), anti-APP (18 961, IBL), anti-BACE1 (D10E5, CST), anti-IDE (sc-393887, Santa Cruz Biotechnology), anti-Neprilysin (AF1126, R&D Systems), anti-MMP-9 (sc-393859, Santa Cruz Biotechnology), and species-specific HRP-conjugated secondary antibodies (Jackson ImmunoResearch Laboratories).
In vivo microdialysis
For the analyses of interstitial fluid (ISF) Aβ, a guide cannula (BASi) was implanted in the hippocampus (bregma: −2.8 mm, ML: 0.5 mm, DV: −1.3 mm (A7); bregma: −2.5 mm, ML: 0.5 mm, DV: −1.3 mm (Irs2−/−;A7), at 37.5° angle). For in vivo insulin stimulation, the guide cannula was implanted in the hippocampus (bregma: −3.1 mm, ML: −2.5 mm, DV: −1.2 mm at 12° angle). A microdialysis probe with 30 kDa molecular weight cutoff membrane (BASi) was inserted into the cannula and aCSF (1.3 mm CaCl2, 1.2 mm MgSO4-7H2O, 3.0 mm KCl, 0.4 mm KH2PO4, 25 mm NaHCO3, and 122 mm NaCl) containing 0.15% BSA was perfused at 1 μl/min. To determine the half-life of ISF Aβ, compound E (Merck) was infused via reverse microdialysis for 6 h at 1 μm after overnight collection of the baseline ISF. ISF samples were collected every 1 h, and Aβ half-life was determined as described previously (Cirrito et al., 2003). For in vivo insulin stimulation, insulin (Humulin-R, Lilly) was infused directly into the hippocampus at 400 nm in aCSF containing 0.15% BSA at 1 μl/min for 1 h using microdialysis probes with 30 kDa molecular weight cutoff membranes (BASi). Controls received regular aCSF containing 0.15% BSA. Mice were immediately sacrificed after 1 h of treatment, and the hippocampus around the probe was harvested for analysis.
Primary mixed neuronal/glial cultures
Primary mixed neuronal/glial cultures were prepared from fetuses of C57BL/6J mice on gestational day 17. The cortex and hippocampus of the excised fetal brain were isolated, cut into small pieces, and digested with enzyme digestion solution (137 mm NaCl, 5 mm KCl, 7 mm Na2HPO4·12H2O, 25 mm HEPES, pH 7.2) containing 10 mg/ml trypsin (Sigma) and 0.5 mg/ml DNase (Sigma) for 20 min at 37°C. Trypsin was inactivated by 20% FBS in HBSS, and the tissues were washed 3 times with HBSS. The cells were dispersed by pipetting in cell separation solution (12 mm MgSO4·7H2O in HBSS) containing 0.5 mg/ml DNase. The cells collected by centrifugation were added to Neurobasal medium (Thermo Fisher Scientific) containing 10% FBS and 1% GlutaMAX (Thermo Fisher Scientific) and cultured at 37°C in a 5% CO2 humidified incubator. On the following day, the medium was replaced with Neurobasal medium containing 2% B-27 Supplement (Thermo Fisher Scientific) and 1% GlutaMAX. Experiments on stimulation with TGF-β and insulin were performed on day 14 of culture using Neurobasal medium containing 2% B-27 (minus insulin) (A1895601, Thermo Fisher Scientific) and 1% GlutaMAX; 2 nm TGF-β (10021, PeproTech) was added to the medium and incubated for 24 h. For simultaneous insulin treatment, the cells were preincubated in the medium containing 1 μm insulin for 6 h and then incubated in the medium containing 1 μm insulin and 2 nm TGF-β for 24 h.
Osmotic pump implantation
ALZET Osmotic Pumps (model 2004, Durect) with a pumping rate of 0.25 µl/h were filled with 22.5 mm RepSox (S7223, Selleck), a selective inhibitor of TGF-β receptor I/ALK5, or 15% DMSO in sterile 0.9% saline (vehicle). After priming at 37°C for >40 h, the prefilled pump was implanted subcutaneously on the back of a 14-month-old A7 mouse under anesthesia. Three weeks after the first implantation surgery, the pump was removed and another prefilled pump was implanted. Three weeks after the second surgery, the animals were sacrificed.
In situ thioflavin T (ThT) fluorescence assay for Aβ fibril detection
Synthetic human Aβ (1-42) (4349-v, Peptide Institute) and Aβ (1-40) (4307-v, Peptide Institute) peptides were solubilized in 1,1,1,3,3,3-hexafluoro-2-propanol (Fujifilm Wako) at a concentration of 1 mg/ml. 1,1,1,3,3,3-Hexafluoro-2-propanol was allowed to evaporate under vacuum using SpeedVac, resolubilized in PBS, pH 7.4, including 2% DMSO at a concentration of 22 μm, and filtered through a 0.22 µm pore size membrane immediately before use. Recombinant mouse thrombospondin-1 (7859-TH, R&D Systems), recombinant mouse thrombospondin-4 (7860-TH, R&D Systems), native human collagen VI α1 (COL6A1) (NBP1-97 270, Novus Biologicals), human α1-microglobulin (PRO-407, ProSpec-Tany TechnoGene), or vehicle (PBS) was added to the Aβ solution, and the solutions were mixed with ThT. The final concentrations of ECM protein, Aβ, and ThT were 250 nm, 5 μm, and 3 μm for the Aβ42 assay, 197 nm, 15 μm, and 3 μm for the Aβ40 assay with thrombospondin-1, and 750 nm, 15 μm, and 3 μm for the Aβ40 assay with the rest of the proteins, respectively. The concentration of collagen VI was calculated and adjusted assuming a trimer of collagen VI α1 because monomers of collagen VI consist of trimers of subunit chains. Dialysis was performed to remove acetic acid contained in COL6A1 solution before use of COL6A1. Four 100 µl aliquots of the mixed solutions were transferred into wells of 96-well microplates (#655900, Greiner Bio-One), and incubated at 37°C in an Infinite M200 PRO microplate reader (Tecan). Fluorescence (excitation at 443 nm, emission at 484 nm) was measured at 5 min intervals with top reading. ThT-monitored fibrillation kinetic curve can be fitted to a sigmoidal curve, as previously described (Uversky et al., 2001; Gade Malmos et al., 2017). The lag time and the apparent rate constant for the growth of fibrils were determined based on fitted curves. Curve fitting analysis was performed with GraphPad Prism 6.
ELISA quantification of Aβ levels
TBS-soluble fractions of hippocampal homogenates and ISF samples were mixed with guanidine hydrochloride, whose final concentration was 0.5 m. These samples were incubated for 30 min at room temperature. The levels of Aβ were measured with Human/Rat β Amyloid(40) ELISA Kit (Fujifilm Wako), Human/Rat β Amyloid(42) ELISA Kit (Fujifilm Wako), or High Sensitive (Fujifilm Wako) according to the manufacturer's instructions.
Intrahippocampal injection of mouse brain extract
A cerebral hemisphere of a 29-month-old A7 mouse was homogenized in a 1:10 (w/v) volume of PBS, pH 7.4, and centrifuged at 347,600 × g for 20 min at 4°C. The supernatant was aliquoted, frozen, and stored at −80°C until use. A glass capillary pipette was inserted into the hippocampus (bregma: −2.5 mm, ML: 2.0 mm, DV: −1.8 mm (A7); bregma: −2.25 mm, ML: 1.8 mm, DV: −1.62 mm (Irs2−/−;A7)) of 3-month-old A7 and Irs2−/−;A7 mice under anesthesia, and 2.5 µl of the brain extract was injected at a rate of 0.35 µl/min. The inner diameter of the point of the pipette was ∼100 µm. Ten minutes after injection, the pipette was slowly withdrawn and the incision was sutured.
Immunohistochemical analysis
Mouse brains were fixed with 4% PFA in PBS, pH 7.4, for 24 h, dehydrated, and embedded in paraffin. Sections were cut serially at 4 μm thickness. Deparaffinized sections were treated with microwave (700 W) in citrate buffer, pH 6.0, for 20 min, followed by digestion with 100 μg/ml proteinase K (Worthington) in TBS for 6 min at 37°C. After blocking by incubation with 10% calf serum in TBS, the sections were incubated with an anti-Aβ antibody 82E1 (IBL) and then a biotinylated anti-mouse IgG antibody (Vector Laboratories), followed by visualization by avidin-biotin complex method (ABC elite, Vector Laboratories) using 3,3′-diaminobenzidine as chromogen. The area percentage occupied by Aβ-positive staining was measured using ImageJ software (National Institutes of Health) as previously described (Yamamoto et al., 2015). Averages of five sections 100 μm distance from each other proximal to the injection site were analyzed.
Experimental design and statistical analysis
All data are presented as mean ± SEM. Significance was determined by two-tailed unpaired t test for two-group data, or one-way ANOVA followed by Tukey's post hoc tests for multiple group comparisons using GraphPad Prism 6. Sample sizes for the experiments are indicated in the figure legends.
Results
Deletion of IRS-2 increased the expression of a subset of ECM genes in the brain
We first aimed to elucidate the molecular underpinnings that contribute to the suppression of amyloid deposition by IRS-2 deficiency using a transcriptomic approach. Our previous analyses in A7 mice have shown that the loss of IRS-2 reduced Aβ levels in the cerebral cortex at as early as 9 months of age, while amyloid plaques were not apparent until after ∼12 months of age (Wakabayashi et al., 2019). Therefore, RNA-Seq of the cerebral cortices of Irs2−/−;A7 and control A7 mice was performed at 9 months of age, when the influence of pathologic changes because of the appearance of Aβ plaques was expected to be minimal. RNA-Seq followed by differential gene expression analysis revealed a total of 268 differentially expressed genes (DEGs) with an FDR of <0.05, of which 183 were upregulated and 85 were downregulated in the cerebral cortices of Irs2−/−;A7 mice compared with A7 mice (Fig. 1A; Extended Data Table 1-1). GSEA showed that gene sets associated with the ECM, including “extracellular matrix component,” “extracellular matrix structural constituent,” “extracellular matrix,” and “proteinaceous extracellular matrix,” were enriched in the cerebral cortices of Irs2−/−;A7 mice (Fig. 1B,C). qRT-PCR analysis subsequently verified that IRS-2 deficiency elevated the expression of genes that were both identified as DEGs by RNA-Seq and included in the gene sets of “extracellular matrix,” in the cerebral cortices of A7 mice (Fig. 1D). Many of the genes identified in the series of analyses have been implicated in the ECM in diverse ways: Col5a1, Col6a1, Col6a2, Col6a5, Col8a2, and Col16a1, for example, belong to the collagen superfamily. Proteins encoded by Col5a1 (Wenstrup et al., 2004), Col6a1 (Minamitani et al., 2004), Fmod (Chen et al., 2010), Ogn (Ge et al., 2004), and Tnxb (Minamitani et al., 2004) regulate collagen fibrillogenesis. Thrombospondin-1 and thrombospondin-4, encoded by Thbs1 and Thbs4, respectively, bind to ECM molecules (e.g., collagen proteins) and affect the organization of ECM (Mumby et al., 1984; Narouz-Ott et al., 2000; Tan and Lawler, 2009). Together, these results suggest that the expression of molecules that constitute the brain ECM was upregulated by the loss of IRS-2.
Comparison of cortical gene expression profiles between A7 and Irs2−/−;A7 mice. A, Volcano plot represents the result of differential gene expression analysis of RNA-Seq data of the cerebral cortices of 9-month-old male A7 (n = 3) and Irs2−/−;A7 (n = 3) mice (FDR < 0.05). The list of DEGs can be found in Extended Data Table 1-1. B, Normalized enrichment scores of gene sets that were significantly upregulated in Irs2−/−;A7 mice in GSEA (FDR < 0.05). C, GSEA enrichment plots of ECM-related gene sets that were significantly enriched in Irs2−/−;A7 mice. D, Validation of RNA-seq data by qRT-PCR in the cerebral cortices of 9-month-old male A7 (n = 10) and Irs2−/−;A7 (n = 10) mice. qRT-PCR was performed on select genes that were identified as DEGs by RNA-Seq and included in the gene sets of “extracellular matrix.” Data are mean ± SEM. *p < 0.05; **p < 0.01; ***p < 0.001; unpaired t test. See Extended Data Fig. 1-1.
Figure 1-1
Molecules involved in the pathways that regulate aging were not altered by IRS-2 deficiency in mouse brains. A-D, Immunoblot analysis of FoxO1 and FoxO3 (A), 4E-BP1 and S6K1 (B), HSF-1 and its target gene products HSP70 and HSC70 (C), and NRF2, heme oxygenase-1, and its regulator KEAP1 (D) in RIPA-soluble cortical fractions of 9-month-old male and female A7 (n = 7) and Irs2−/−;A7 (n = 7) mice were analyzed (left panels). The graphs on the right show the levels of each protein. The levels of phospho-FoxOs, phospho-4E-BP1, and phospho-S6K1 were normalized against the total amount of each protein. The levels of other proteins were normalized against α-tubulin (indicated in D). Data are mean ± SEM and analyzed by unpaired t test. E, qRT-PCR analysis of the cerebral cortices of male C57BL/6J mice at 2 months (n = 10) and 20 months (n = 10) of age. qRT-PCR was performed on the same set of ECM genes analyzed in Figure 1D. Data are mean ± SEM and analyzed by unpaired t test. *p < 0.05; **p < 0.01; ***p < 0.001. Download Figure 1-1, TIF file.
Table 1-1
A list of DEGs between A7 and Irs2−/−;A7 mice identified by RNA-Seq analysis. In Irs2−/−;A7 mice, 183 genes were upregulated and 85 were downregulated. FDR < 0.05. Download Table 1-1, XLSX file.
IRS-2 deficiency increased the expression of ECM genes in neurons and astrocytes
To further explore the altered ECM gene expression by IRS-2 deficiency, we focused on Thbs1, Thbs4, and Col6a1, whose protein products are reliably detectable in brain tissue and have potential relevance to AD pathology. Thrombospondins have been localized to senile plaques in the brains of AD patients, while they are reduced in a subset of pyramidal neurons that are vulnerable in AD, implicating thrombospondins in neuronal degeneration and amyloid accumulation (Buée et al., 1992; Cáceres et al., 2007; Son et al., 2015). Collagen VI, encoded by Col6a1 gene, was shown to competitively inhibit the binding of Aβ oligomers to neurons, suggesting the involvement of collagen VI in Aβ dynamics in the brain (Cheng et al., 2009). The expression of these three genes was upregulated also in the hippocampi of Irs2−/−;A7 mice (Fig. 2A). Moreover, immunoblot analysis of the cerebral cortices revealed that the protein levels of thrombospondin-1, thrombospondin-4, and collagen VI α1 chain, encoded by Col6a1, in Irs2−/−;A7 mice were significantly higher than those in control A7 mice (Fig. 2B,C).
IRS-2 deficiency increased the expression of ECM genes in neurons and astrocytes. A, qRT-PCR analysis of Thbs1, Thbs4, and Col6a1 mRNA expression in the hippocampus of 9-month-old male A7 (n = 10) and Irs2−/−;A7 (n = 10) mice. B, C, Immunoblot analysis of thrombospondin-1 (TSP1), thrombospondin-4 (TSP4), collagen VI α1 chain (COL6A1), and α-tubulin in urea-soluble cortical fractions of 9-month-old A7 (n = 10) and Irs2−/−;A7 (n = 10) mice (B). Graphs represent the levels of each polypeptide normalized to α-tubulin (C). Data are mean ± SEM and analyzed by unpaired t test. D, qRT-PCR analysis of Thbs1, Thbs4, and Col6a1 mRNA expression in neurons, astrocytes, and microglia purified from the brains of 11-month-old male C57BL/6J mice (neurons: n = 4, astrocytes: n = 6, microglia: n = 6). Data are mean ± SEM and analyzed by one-way ANOVA with Tukey's post hoc test. E, qRT-PCR analysis of Thbs1, Thbs4, and Col6a1 mRNA expression in neurons, astrocytes, and microglia purified from the brains of 11-month-old Irs2+/+ (n = 4) and Irs2−/− (n = 8) mice. Data are mean ± SEM. *p < 0.05; **p < 0.01; ***p < 0.001; unpaired t test.
In a previous study that analyzed the contribution of different IRS proteins to brain insulin signaling activity, we observed that IRSs are expressed primarily in astrocytes and neurons (Ochiai et al., 2021). We therefore sought to determine which cell types in the brain contribute to the increased expression of ECM genes because of IRS-2 deficiency. Neurons, astrocytes, and microglial cells were isolated from the cerebral hemispheres of adult C57BL/6J mice by magnetic cell sorting, and the distribution of each ECM gene expression was evaluated by qRT-PCR analysis. Thbs4 exhibited an astrocyte-specific expression pattern, whereas Thbs1 and Col6a1 showed the highest expression in neurons (Fig. 2D). Col6a1 was also expressed in astrocytes. We then investigated in which cell types the loss of IRS-2 affected the expression of the ECM genes (Fig. 2E). IRS-2-deficient neurons showed significantly higher expression of Thbs1 and Col6a1 than control neurons. IRS-2-deficient astrocytes showed significantly higher expression of Thbs4 and Col6a1 than control astrocytes. Microglial cells exhibited no differences in the mRNA levels of these three genes between the two genotypes. Thus, the present results that IRS-2 deficiency increased the expression of ECM genes in neurons and astrocytes were consistent with the expression pattern of IRS-2 in both cell types (Ochiai et al., 2021). Notably, these results also demonstrated that IRS-2 deficiency increased the expression of ECM genes not only in A7 mice but also in nontransgenic mice. Together, our data demonstrate that deletion of IRS-2 increased the mRNA and protein levels of a subset of ECM genes in the brain.
Molecules involved in the pathways that regulate aging were not altered by IRS-2 deficiency in mouse brains
Aging leads to a decline in protein homeostasis (i.e., proteostasis), which in turn results in the development of protein-misfolding diseases (Taylor and Dillin, 2011). Previous analyses using C. elegans and mouse models have suggested that reduced IIS may act on the proteostasis of Aβ, reducing its aggregation and toxicity (Cohen et al., 2006, 2009). We therefore asked whether the effect of IRS-2 deficiency on ECM gene expression is regulated by any of the canonical anti-aging-related molecules downstream of the IIS pathway (Taylor and Dillin, 2011). However, the group of genes identified as DEGs by RNA-seq or enriched by GSEA did not include genes implicated in anti-aging effects downstream of IIS. We further examined the expression of proteins involved in the regulation of aging, for example, FoxO transcription factors, mTOR substrates, HSF-1, and an antioxidant transcription factor NRF2, in the cerebral cortices of Irs2−/−;A7 mice. Phosphorylation levels of FoxO1 and FoxO3a, as well as those of 4E-BP1 and S6K1, substrates of mTOR, were not affected by IRS-2 deficiency (Extended Data Fig. 1-1A,B). Also, the levels of HSF-1 and its target gene products, HSP70 and HSC70, were comparable between Irs2−/−;A7 and A7 mice (Extended Data Fig. 1-1C). Likewise, the levels of NRF2, its target gene product heme oxygenase-1, and KEAP1, a regulator of NRF2 activity, were comparable between the two groups (Extended Data Fig. 1-1D).
We also examined whether the expression of ECM genes identified by RNA-seq changes with aging in mice. Expression of a series of ECM genes examined in Figure 1D was compared in the brains of 2-month-old (young) and 20-month-old (aged) C57BL/6J mice, which showed no clear changes in expression with aging (Extended Data Fig. 1-1E). These results collectively suggest that changes in ECM expression may not be downstream of the canonical aging-regulating signals downstream of IIS.
IRS-2 deficiency increased the phosphorylation level of Smad2 and Smad3 in the cerebral cortices
We then sought to elucidate the mechanism by which the loss of IRS-2 upregulated the expression of ECM genes in the brain. One of the most important signaling pathways regulating the production of ECM is the TGF-β signaling. TGF-β/Smad signaling controls various biological processes, such as cell proliferation/differentiation, immune responses, as well as the synthesis of ECM proteins (Verrecchia and Mauviel, 2002). Previous studies have reported that TGF-β stimulation upregulated the expression of Thbs1 (Song et al., 2003; Nakagawa et al., 2004; Pal et al., 2016), Thbs4 (Qian et al., 2018), and Col6a1 (Verrecchia et al., 2001) in a variety of non-neuronal cells, although knowledge on the regulation of these genes by TGF-β signaling in the CNS is limited. TGF-β induced expression of Thbs1 (Nakagawa et al., 2004), Thbs4 (Qian et al., 2018), and Col6a1 (Verrecchia et al., 2001) via activation of the transcription factors Smad2 and Smad3. Interestingly, several studies have shown that insulin and IGF-1 inhibit the transcriptional responses to TGF-β through suppression of Smad3 activation in NRP-152 cells and Hep3B cells (Song et al., 2003; Conery et al., 2004; Remy et al., 2004). Hence, we hypothesized that the increased expression of ECM genes in IRS-2-deficient brains is mediated by enhanced activation of the Smad proteins. To test this possibility, we evaluated the phosphorylation of Smad proteins in the cerebral cortices at 9 months of age. Immunoblot analysis showed that the phosphorylation levels of both Smad2 and Smad3 in the cerebral cortices of Irs2−/−;A7 mice were significantly higher compared with those of control A7 mice (Fig. 3A,B).
Insulin signaling regulated the activation of Smad2/3 and TGF-β-dependent brain ECM expression. A, B, Immunoblot analysis of phosphorylation and total amount of Smad2 and Smad3 in RIPA-soluble cortical fractions of 9-month-old male A7 (n = 10) and Irs2−/−;A7 (n = 10) mice (A). Graphs represent the levels of phospho-Smad normalized against total Smad (B). Data are mean ± SEM and analyzed by unpaired t test. C, D, Immunoblot analysis of phosphorylation and total amount of Smad2 and Smad3 in RIPA-soluble hippocampal fractions of female A7 mice after 1 h of aCSF (vehicle, n = 7) or insulin (400 nm, n = 9) perfusion (C). Graphs represent the levels of phospho-Smad normalized against total Smad (D). Data are mean ± SEM and analyzed by unpaired t test. E, qRT-PCR analysis of Thbs1, Thbs4, and Col6a1 mRNA expression in primary mixed neuronal/glial cultures treated with vehicle (n = 10), TGF-β (n = 12), and both TGF-β and insulin (n = 11). Data are mean ± SEM. *p < 0.05; **p < 0.01; ***p < 0.001; one-way ANOVA with Tukey's post hoc test.
To ascertain the effect of insulin signaling on phosphorylation of Smad2/3 in the brain, we examined the changes in the levels of Smad phosphorylation in the hippocampus after insulin administration. The hippocampi of A7 mice were perfused with 400 nm insulin for 1 h via reverse microdialysis. We have previously shown that, under this experimental condition, insulin perfusion activates the insulin receptor and triggers activation of the PI3K/Akt pathway in vivo (Ochiai et al., 2021). Immunoblot analysis showed that insulin perfusion significantly decreased phosphorylation of Smad3 in the hippocampus in vivo; phosphorylation of Smad2 showed a decreasing trend but was not statistically significant (Fig. 3C,D). These results indicate that the increased activation of Smad2/3 because of IRS-2 deficiency may be the result of an abrogation of the inhibitory effect of IIS on Smad2/3 activity.
Next, we examined the regulation of ECM expression by TGF-β and insulin signaling in brain-derived cells. Primary mixed neuronal/glial cultures from C57BL/6J mice were treated with TGF-β or both TGF-β and insulin for 24 h to examine the effects on the expression of thrombospondins and collagen VI. qRT-PCR analysis showed that TGF-β stimulation significantly increased the mRNA level of Thbs4 (Fig. 3E). We observed a trend of increase in the mRNA levels of Thbs1 and Col6a1, which was not significant (Fig. 3E). Moreover, to examine the effect of insulin on the TGF-β-induced upregulation of these ECM genes, primary cultures were treated with both TGF-β and insulin. Addition of insulin significantly reduced the TGF-β-induced upregulation of Thbs1 and Thbs4 (Fig. 3E). We found a nonsignificant decrease in the mRNA levels of Col6a1 in insulin-treated cultures (Fig. 3E). The results suggest that TGF-β upregulates the expression of the ECM genes, which is inhibited by insulin in the CNS.
Deletion of IRS-1 did not alter the phosphorylation level of Smad or the expression of ECM genes
IRS-1, encoded by Irs1, also is a major substrate of the insulin receptor. However, we have previously shown that, unlike IRS-2 deficiency, deletion of IRS-1 does not attenuate the brain insulin signaling activity; accordingly, amyloid plaque formation in the brains of A7 mice is markedly suppressed at 15 months of age by loss of IRS-2, while it is not altered by IRS-1 deficiency (Extended Data Fig. 4-1) (Ochiai et al., 2021). To examine whether the brain insulin signaling activity is correlated with Smad2/3 phosphorylation and the expression of ECM genes, we analyzed the brains of IRS-1-deficient A7 mice at 9 months of age. The phosphorylation levels of Smad2 or Smad3 in cortical homogenates of Irs1−/−;A7 mice were at similar levels to those of control A7 mice (Fig. 4A,B). The hippocampal mRNA levels of Thbs1, Thbs4, and Col6a1 were also comparable between the two genotypes (Fig. 4C). The levels of these ECM proteins in the cerebral cortices of Irs1−/−;A7 mice were also similar to those in control A7 mice (Fig. 4D,E). These results support the idea that attenuated insulin signaling increased the activity of Smad2/3 as well as the expression of a group of ECM genes. The observation that IRS-1 did not affect Aβ deposition further suggests that the change in the activity of Smad proteins and the expression of those ECM genes is associated with the change in Aβ accumulation.
IRS-1 deficiency did not alter the phosphorylation of Smad or brain ECM expression. A, B, Immunoblot analysis of phosphorylation and total amount of Smad2 and Smad3 in RIPA-soluble cortical fractions of 9-month-old male A7 (n = 10) and Irs1−/−;A7 (n = 8) mice (A). Graphs represent the levels of phospho-Smad normalized against total Smad (B). Data are mean ± SEM and analyzed by unpaired t test. C, qRT-PCR analysis of Thbs1, Thbs4, and Col6a1 mRNA expression in the hippocampus of 9-month-old male A7 (n = 10) and Irs1−/−;A7 (n = 8) mice. Data are mean ± SEM and analyzed by unpaired t test. D, E, Immunoblot analysis of TSP1, TSP4, COL6A1, and α-tubulin in RIPA-soluble cortical fractions of 9-month-old male A7 (n = 10) and Irs1−/−;A7 (n = 8) mice (D). Graphs represent the levels of each polypeptide normalized to α-tubulin (E). Data are mean ± SEM and analyzed by unpaired t test. See Extended Data Fig. 4-1.
Figure 4-1
Differential effects of IRS-1 and IRS-2 deficiency on amyloid plaque formation in A7 mouse brain. A, B, Representative images of Aβ-immunostained brain sections from 15-month-old A7 and Irs2−/−;A7 mice (A), and 15-month-old A7 and Irs1−/−;A7 mice (B). Download Figure 4-1, TIF file.
Inhibition of TGF-β signaling increased Aβ deposition
We next examined whether Aβ deposition is affected downstream of TGF-β signaling by means of pharmacological interventions. In the canonical TGF-β signaling, TGF-β receptor I(TGF-βRI)/ALK5 is a receptor for TGF-β, which plays an important role in the downstream signaling by phosphorylating Smad2 and Smad3 through its kinase activity. We thus used RepSox, a small-molecule inhibitor of TGF-βRI/ALK5, to examine whether suppression of the signaling exerts the opposite effect of IRS-2 deficiency (i.e., increased Aβ accumulation in the brain).
In a preliminary study of a single bolus intraperitoneal injection of 20 mg/kg RepSox into C57BL/6J mice, Smad3 phosphorylation in the brain was significantly reduced by ∼25% at 30 min after administration (pSmad3 ratio: 1.000 ± 0.079 (vehicle), 0.742 ± 0.021 (RepSox), p = 0.020, n = 4 per group), but this effect was abolished by 5 h. These observations indicated that Smad3 is basally activated in the brain in a TGF-β receptor-dependent manner, which can be inhibited by peripheral administration of RepSox. We then performed continuous RepSox administration to evaluate the long-term effects on Aβ pathology. RepSox was administered at 22.5 mM to 14-month-old A7 mice at a flow rate of 0.25 μl/h for 6 consecutive weeks using subcutaneously implanted osmotic pumps designed for long-term infusion. Immunoblot analysis of extracted fractions of the cerebral cortex showed a strong trend of decrease in Smad3 phosphorylation levels in the RepSox-treated group (p = 0.058), suggesting suppression of brain TGF-β/Smad signaling (Fig. 5A,B). As expected, we observed no changes in the levels of phosphorylation at Ser473 and Thr308 of Akt (pAkt Ser473/Akt ratio: 1.005 ± 0.091 (vehicle), 1.078 ± 0.117 (RepSox), p = 0.628; pAkt Thr308/Akt ratio: 1.001 ± 0.025 (vehicle), 1.024 ± 0.033 (RepSox), p = 0.600, n = 9 per group). Immunohistochemical analyses showed an increase in amyloid plaques in the brains of RepSox-treated A7 mice compared with the vehicle-treated A7 mice (Fig. 5C). Quantification of the percentage of Aβ-positive area revealed that RepSox significantly increased Aβ deposition (Fig. 5D). These in vivo experiments supported the notion that Aβ deposition is regulated by the downstream pathway of the TGF-β/Smad signaling in the brain.
Administration of a TGF-β signaling inhibitor increased Aβ deposition in A7 mouse brain. A, B, Immunoblot analysis of phosphorylation and total amount of Smad3 in RIPA-soluble cortical fractions of 15-month-old male A7 mice after 6 consecutive weeks of vehicle or RepSox administration with subcutaneously implanted ALZET osmotic pumps (n = 9 per group) (A). Graph represents the levels of phospho-Smad3 normalized against total Smad3 (B). Data are mean ± SEM and analyzed by unpaired t test. C, D, Immunohistochemical analysis using an anti-Aβ (82E1) antibody (C) and morphometry of percentage Aβ-positive areas (D) in the cerebral cortex. Data are mean ± SEM. *p < 0.05; unpaired t test. Scale bar, 500 μm.
Thrombospondins and collagen VI inhibited the aggregation of Aβ in vitro
Next, we determined how thrombospondins and collagen VI contribute to the mitigation of Aβ accumulation. Several ECM proteins expressed in the brain, including laminin (Bronfman et al., 1996; Monji et al., 2000), collagen IV (Kiuchi et al., 2002), CLAC/collagen XXV (Kakuyama et al., 2005; Osada et al., 2005), and Reelin (Pujadas et al., 2014), have been shown to inhibit fibril formation of Aβ in vitro. We therefore speculated that thrombospondins and collagen VI also inhibit the fibril formation of Aβ. To test this hypothesis, we examined the effect of these ECM proteins on fibril formation of Aβ by an in vitro assay using an amyloid-specific fluorescent dye, ThT. Synthetic Aβ42 peptides were mixed with ThT and incubated at 37°C, and the fluorescence intensity was measured every 5 min. Based on the fitted curves, the lag time and elongation rate of fibril formation of Aβ were estimated (Uversky et al., 2001; Gade Malmos et al., 2017). The lag time reflects a nucleation phase, in which Aβ monomers undergo conformational change and associate to form oligomeric nuclei, whereas the elongation rate reflects the process at which the nuclei grow by further addition of monomers and form amyloid fibrils (Harper and Lansbury, 1997). Thrombospondin-1, thrombospondin-4, or collagen VI proteins did not alter the fluorescence intensities at the starting point. The addition of these three proteins to the reactions significantly increased the lag time and decreased the elongation rate (Fig. 6A–I). By contrast, the addition of α1-microglobulin as a control protein altered neither the lag time nor the elongation rate (Fig. 6J–L). Similarly, the effects of ECM proteins on fibril formation of Aβ40 peptide also were examined (n = 4 for each test). Upon coincubation of thrombospondin-4 with Aβ40, the lag time was significantly prolonged to 643.4 ± 30.3 min, compared with 353.0 ± 18.5 min on incubation of Aβ40 alone (p = 0.0002). Thrombospondin-4 significantly reduced the elongation rate to 0.0137 ± 0.0033 min−1 compared with 0.0335 ± 0.0038 min−1 on incubation of Aβ40 alone (p = 0.0078). Collagen VI significantly prolonged the lag time to 766.0 ± 16.8 min compared with 603.2 ± 12.0 min on incubation of Aβ40 alone (p = 0.002). Collagen VI significantly reduced the elongation rate to 0.0148 ± 0.0006 min−1 compared with 0.0274 ± 0.0043 min−1 on incubation of Aβ40 alone (p = 0.0276). Of note, thrombospondin-1 showed the most potent inhibition, with no increase in ThT fluorescence intensity. Similarly to the assays with Aβ42, α1-microglobulin had no effect on Aβ40 fibril formation (lag time: 534.9 ± 42.6 min (Aβ40 alone) and 515.4 ± 43.2 min in the presence of α1-microglobulin; elongation rate: 0.0353 ± 0.0051 min−1 (Aβ40 alone) and 0.0373 ± 0.0027 min−1 in the presence of α1-microglobulin). These results indicate that thrombospondin-1, thrombospondin-4, and collagen VI inhibit aggregation of Aβ in vitro by affecting both nucleation and fibril elongation processes.
The ECM proteins inhibited amyloid fibril formation of Aβ in vitro. The effects of thrombospondin-1 (TSP1) (A–C), thrombospondin-4 (TSP4) (D–F), collagen VI (Col6) (G–I), or α1-microglobulin (A1M) (J–L) on fibril formation of Aβ42 by an in vitro assay using ThT. Representative kinetics of ThT fluorescence intensity during incubation of synthetic human Aβ (1-42) in the absence or presence of TSP1 (A), TSP4 (D), Col6 (G), or A1M (J). The lag time (B,E,H,K) and the elongation rate of fibril formation (C,F,I,L) were calculated based on fitted curves (n = 4). Data are mean ± SEM. ***p < 0.001; unpaired t test.
IRS-2 KO suppressed seed-induced Aβ deposition
We next aimed to dissect and identify the step(s) of Aβ accumulation affected by IRS-2 deletion. The brain levels of Aβ are regulated by a balance between production, elimination, and aggregation. First, Aβ levels in the hippocampus of Irs2−/−;A7 mice were measured by ELISA to determine the starting point of inhibition in Aβ accumulation. In the cerebral cortices, IRS-2 deficiency in A7 mice decreased the levels of both soluble and insoluble Aβ (Wakabayashi et al., 2019). As in the cerebral cortex, we found that the levels of soluble Aβ40 and Aβ42 in the hippocampus were lower in Irs2−/−;A7 mice than in control A7 mice at 9 months of age (Fig. 7A). However, the levels of soluble Aβ40 and Aβ42 in the hippocampus of 6-month-old Irs2−/−;A7 mice were approximately equal to those in control A7 mice (Fig. 7B). These results raise the possibility that deletion of IRS-2 might not affect the production or elimination of Aβ, but rather may alter the pathologic process after Aβ has started to insolubilize.
IRS-2 deficiency suppressed seed-induced Aβ deposition. A, The levels of Aβ40 and Aβ42 in TBS-soluble hippocampal fractions of male and female 9-month-old A7 (n = 12) and Irs2−/−;A7 (n = 14) mice. B, The levels of Aβ40 and Aβ42 in TBS-soluble hippocampal fractions of 6-month-old female A7 (n = 6) and Irs2−/−;A7 (n = 6) mice. C–E, The half-life of ISF Aβ42 was analyzed by microdialysis. C, Hourly measurements of ISF Aβ42 during microdialysis in 9-month-old male and female A7 (n = 8) and Irs2−/−;A7 (n = 12) mice. ISF Aβ42 levels are shown as % of the mean baseline in A7 mice. D, Rate of reduction in ISF Aβ42 after administration of compound E. A semi-log plot of percentage ISF Aβ42 levels (2.0 at time 0) is shown versus time. E, Half-life of ISF Aβ42 based on data in D. F, G, Immunohistochemical analysis using an anti-Aβ (82E1) antibody (F) and morphometry of percentage Aβ-positive areas (G) of the hippocampus of 6-month-old male A7 (n = 5) and Irs2−/−;A7 (n = 6) mice injected at 3 months of age with brain extract from aged A7 mouse. Data are mean ± SEM. *p < 0.05; **p < 0.01; ***p < 0.001; unpaired t test. Scale bar, 500 μm. See Extended Data Fig. 7-1.
Figure 7-1
IRS-2 deficiency did not affect the molecules involved in the production or degradation of Aβ. A, B, Immunoblot analysis of the C-terminal fragments (CTFs) and full-length polypeptides of APP (A), and BACE1 (B), in TX-soluble cortical fractions of 9-month-old male and female A7 (n = 12) and Irs2−/−;A7 (n = 14) mice. C, Immunoblot analysis of Aβ degrading enzymes IDE, neprilysin, and MMP-9 in RIPA-soluble cortical fractions of 9-month-old male A7 (n = 10) and Irs2−/−;A7 (n = 10) mice. The graphs on the right show the levels of each protein. The levels of APP CTFs were normalized against the full-length APP. The levels of other proteins were normalized against α-tubulin. Data are mean ± SEM and analyzed by unpaired t test. Download Figure 7-1, TIF file.
To examine the effects on Aβ production, we analyzed the protein levels of APP and its cleavage products. There has been some disagreement among previous studies on the changes in APP processing in the brains of IRS-2-deficient APP transgenic mice (Freude et al., 2009; Killick et al., 2009). In this present study, the levels of CTF-α, CTF-β, and BACE1 in the cortical homogenates of 9-month-old Irs2−/−;A7 mice were comparable with those of control A7 mice (Extended Data Fig. 7-1A,B), supporting the lack of effects on APP processing by IRS-2 deficiency.
Next, we assessed the effects of IRS-2 deficiency on the elimination of soluble Aβ by in vivo microdialysis. We collected the ISF from the hippocampus every hour and estimated the half-life of Aβ42 by measuring the time-dependent changes in ISF Aβ42 levels after administration of a γ-secretase inhibitor, compound E, to block de novo Aβ production (Cirrito et al., 2003). The results showed that the ISF Aβ42 levels in 9-month-old Irs2−/−;A7 mice and control A7 mice were reduced similarly after inhibition of Aβ production (Fig. 7C). A semi-log plot of concentration versus time was linear, suggesting first-order kinetics for ISF Aβ42 (Fig. 7D). The half-life of ISF Aβ42 calculated based on the slope of decline was not different between Irs2−/−;A7 and control A7 mice (Fig. 7E). Because the protein levels of IDE have been reported to increase in the brains of IRS-2-deficient Tg2576 mice (Killick et al., 2009), we further examined the levels of amyloid degrading enzymes. Immunoblot analysis of the cerebral cortices showed no differences in the protein levels of IDE, neprilysin, or matrix metalloproteinase-9 between the two genotypes at 9 months of age (Extended Data Fig. 7-1C). Overall, our results did not support the effect of IRS-2 deficiency on the elimination of Aβ.
We then investigated the effects of IRS-2 deletion on Aβ fibril formation in vivo. To this end, we performed a brain injection experiment to evaluate seed-induced Aβ deposition (Kane et al., 2000). Brain extracts from aged A7 mice containing Aβ seeds, which harbor the ability to induce amyloid deposition in vivo, were injected into the hippocampi of 3-month-old A7 or Irs2−/−;A7 mice, which were sacrificed 3 months later and analyzed by immunohistochemistry to evaluate Aβ deposition. The amount of seed-induced Aβ deposition in the hippocampi of Irs2−/−;A7 mice was significantly lower than that of control A7 mice at 6 months of age (Fig. 7F,G). Since the amount of Aβ in the hippocampus of Irs2−/−;A7 mice was comparable to that of control A7 mice at 6 months of age (Fig. 7B), we reasoned that the observed suppression of seed-induced Aβ deposition was because of the inhibition of Aβ fibril elongation. Collectively, these findings suggest that IRS-2 deficiency inhibits Aβ fibril formation, leading to the reduced deposition of amyloid plaques in the brains of Irs2−/−;A7 mice.
Discussion
The molecular mechanisms by which reduced insulin/IGF-1 signaling suppresses the accumulation of Aβ in the brains of animal models of AD have remained incompletely understood. Here, we showed that IRS-2 deficiency promoted the phosphorylation of Smad2/3 and upregulated the expression of a set of ECM proteins in the brain. We also found that these ECM molecules inhibited the fibril formation of Aβ in vitro and that Aβ fibril formation was suppressed in IRS-2-deficient A7 mice in vivo. These findings led us to hypothesize that reduced IIS inhibits the aggregation process of Aβ through increased TGF-β signaling and ECM expression in the brain (Fig. 8).
Proposed mechanism of the regulation of Aβ accumulation by brain IIS. In neurons and astrocytes, TGF-β/Smad signaling induces the expression of a subset of ECM, which may influence ECM composition in the brain extracellular space. Several ECM proteins, including TSP1, TSP4, and collagen VI, have inhibitory effects on Aβ amyloid fibril formation. In the situation of IRS-2 deficiency (indicated by the red cross), reduced IIS activity may enhance TGF-β/Smad signaling, resulting in inhibition of Aβ amyloid fibril formation by the ECM and less amyloid plaque deposition in AD model mouse brains.
Limited findings have suggested a relationship between TGF-β/Smad signaling and ECM expression in the brain: the levels of ECM proteins were increased in the brains of astrocyte-specific TGF-β1 transgenic mice (Wyss-Coray et al., 1995); and conversely, loss of TGF-β1 reduced the ECM expression in the brain (Brionne et al., 2003). However, these studies have focused only on fibronectin and laminin, without reference to comprehensive changes in the ECM, including collagens and other components. In the present study, we used primary mixed neuronal/glial cultures to show that the expression of ECM genes affected by IRS-2 deficiency, Thbs1, Thbs4, and Col6a1, was induced by TGF-β. Together with the elevated phosphorylation of Smad2/3 in the brains of Irs2−/−;A7 mice, these results support the hypothesis that reduced IIS affects TGF-β/Smad signaling pathway, thereby altering the ECM expression.
Crosstalk between IIS and TGF-β/Smad signaling pathways has been suggested by studies using cell lines derived from peripheral tissues, which showed that insulin and IGF-1 inhibited TGF-β-induced phosphorylation of Smad3 (Song et al., 2003; Conery et al., 2004; Remy et al., 2004). Our in vivo observation that intrahippocampal administration of insulin decreased the levels of phospho-Smad3 supports the crosstalk. A comprehensive gene expression analysis has also suggested that conditional deletion of neuronal IGF-1R affected TGF-β signaling and ECM expression in the hippocampus of both APP/PS1 transgenic and nontransgenic mice (George et al., 2017). However, no study has yet analyzed the impact of reduced IIS on the brain proteome. In addition, ECM proteins are underrepresented in proteomics datasets because of the extraction methods, such as decellularization for efficient enrichment, as we used in this study (McCabe et al., 2021). Although we focused on the three representative ECM proteins here, future studies should address more comprehensive analysis of the ECMs at the protein level using conditions optimized for ECM detection.
One possible mechanism for the inhibitory action of IIS on TGF-β/Smad signaling is that phosphorylated Akt may form a complex with Smad3 and inhibit its activation by sequestering free Smad3 (Conery et al., 2004; Remy et al., 2004). Our findings that only deletion of IRS-2, but not IRS-1, reduces basal phosphorylation levels of Akt (Ochiai et al., 2021) and increases phospho-Smad2/3 in the brain support the hypothesis for a role of Akt in suppressing Smad3 activation. However, other mechanisms may also be possible. TGF-β is secreted as a latent form in complex with LAP and LTBP and sequestered into the ECM. Thrombospondin-1 competes with LAP and activates TGF-β (Schultz-Cherry and Murphy-Ullrich, 1993; Crawford et al., 1998). It has also been suggested that the properties of ECM per se influence the activity and availability of TGF-β1 (Hinz, 2015). Thus, it might be tempting to speculate the presence of a positive feedback loop, in which IRS-2 deficiency increases the levels of thrombospondin-1 and other ECM proteins in the brain, resulting in the activation of TGF-β and enhancement of the downstream signaling.
TGF-β receptors are distributed in a wide variety of regions and cell types in the nervous system (Böttner et al., 2000). The cell type-specific analysis in this study showed an increased expression of Thbs4 and Col6a1 in astrocytes and Thbs1 and Col6a1 in neurons by loss of IRS-2, in agreement with our previous observation that IRS-2 expression is higher in astrocytes and neurons, while weaker in microglia (Ochiai et al., 2021). It also dovetails with the previous studies which showed that neuron-specific deletion of the insulin receptor and IGF-1R decreased the amyloid plaque formation, underscoring the importance of neuronal IIS in regulating Aβ accumulation (Stöhr et al., 2013; Gontier et al., 2015). We thus speculate that both neurons and astrocytes may contribute to the IIS-dependent regulation of ECM expression and affect the process of Aβ accumulation.
TGF-β signaling is involved in multiple pathologic conditions, most notably cancer, and also in AD. While increased TGF-β expression has been reported in the brains of AD patients, decreased TGF-β Type II receptor levels and abnormal subcellular localization of activated Smads have also been demonstrated (Flanders et al., 1995; Wyss-Coray et al., 1997; Lee et al., 2006; Tesseur et al., 2006; Ueberham et al., 2006), making it unclear how TGF-β signaling activity is involved in the pathogenesis of AD. A previous study in APP transgenic mice showed that astrocytic overexpression of TGF-β1 suppressed brain Aβ accumulation (Wyss-Coray et al., 2001). In contrast, reduction of TGF-β signaling in the brain by pharmacological inhibition of TGF-βRI/ALK5 promoted Aβ accumulation in our study. Because TGF-β and its receptors transmit diverse downstream signals (Derynck and Zhang, 2003), future experimental evidence on the regulation of Aβ pathology by interventions into the brain TGF-β-ECM pathway may provide further clues to the causal relationship between ECM proteins and Aβ accumulation in vivo.
Our in vivo and in vitro analyses suggested that the IRS-2-deficient condition and associated upregulation of ECM may suppress Aβ fibril formation, while no effect of IRS-2 deficiency on Aβ production or clearance was observed. Several ECM proteins (e.g., laminin, collagen IV, CLAC/collagen XXV, and Reelin) (Bronfman et al., 1996; Monji et al., 2000; Kiuchi et al., 2002; Osada et al., 2005; Kocherhans et al., 2010; Pujadas et al., 2014; Hashimoto et al., 2020) have been reported to inhibit fibril formation of Aβ and affect its accumulation in the brain. However, the effects of the individual products of the differentially expressed genes identified in this study on amyloid plaque formation remain largely unresolved. Thrombospondins have been observed immunohistochemically in senile plaques of AD brains, but their effects on Aβ fibril formation are unknown (Buée et al., 1992; Cáceres et al., 2007). Collagen VI has been shown to reduce the Aβ oligomers while promoting aggregate formation (Cheng et al., 2009). Our data showing the inhibitory effect of collagen VI on amyloid fibril formation may not contradict the former results, which might have evaluated nonfibrillar, large amorphous Aβ aggregates observed by atomic force microscopy (Herzberg et al., 2020). These findings support the notion that IIS has a broad impact on Aβ deposition by modulating multiple components of the brain ECM.
IIS has also been implicated in AD pathophysiology through the regulation of aging and lifespan. We and others have demonstrated in mouse models that there is no association between lifespan extension and suppressed Aβ deposition caused by genetic deletion of IIS components (Selman et al., 2008; Stöhr et al., 2013; Ochiai et al., 2021). Furthermore, signaling molecules that contribute to the lifespan-extending effects downstream of IIS, such as FOXOs and mTOR, were not altered by IRS-2 deficiency in our study. Interventions that reduce IIS activity, including genetic inhibition of IIS and caloric restriction, also prevent abnormal protein aggregation and cytotoxicity by maintaining protein homeostasis (Cohen and Dillin, 2008). Altered ECM expression downstream of the IIS proposes a novel mechanism by which the IIS maintains proteostasis through changes in the extracellular environment of the brain. In C. elegans, repression of daf-2, the nematode homolog of insulin and IGF-1 receptors, increases the expression of various collagen genes, resulting in ECM remodeling that delays aging (Ewald et al., 2015). More detailed analysis on the ECM remodeling during brain aging may facilitate our understanding of the pathogenesis of dementing neurodegenerative disorders, including AD.
It may also be possible that the progression of AD leads to impaired brain insulin signaling and subsequent pathogenic processes (i.e., synaptic/cognitive dysfunction). Mechanisms, such as increased serine phosphorylation of IRS-1, inhibition of Akt, and activation of GSK-3β by Aβ, have been implicated in this process (Jo et al., 2011; Bomfim et al., 2012). Based on these findings, AD therapies that attempt to activate brain insulin signaling have also been explored (Kellar and Craft, 2020). Thus, a deeper understanding of the diverse effects of brain IIS on AD pathogenesis at different stages of disease progression is crucial.
In conclusion, our results suggest that reduced brain IIS by IRS-2 deficiency leads to attenuated Aβ deposition by increasing ECM molecules in the brain that inhibit Aβ fibril formation. The potential relevance of the ECM to the pathogenesis of AD has been highlighted by recent large-scale proteome analyses showing AD-related changes in the “matrisome,” a collection of ECM-associated proteins (Johnson et al., 2022). Closer understanding of the mechanisms and roles of altered brain IIS and ECMs in AD pathophysiology will lead to the identification of molecular targets for interventions that mimic the beneficial effects of reduced IIS in the brain.
Footnotes
This work was supported by Japan Agency for Medical Research and Development Grant JP20dm0107056, Japan Society for the Promotion of Science KAKENHI Grant JP20H00525 and JP23K18416, and Daiichi-Sankyo Foundation of Life Science. We thank the members of our laboratory for helpful suggestions and discussions.
The authors declare no competing financial interests.
- Correspondence should be addressed to Tomoko Wakabayashi at tomoko-wakabayashi{at}umin.ac.jp or Takeshi Iwatsubo at iwatsubo{at}m.u-tokyo.ac.jp
