Bone marrow mesenchymal stem cells derived cytokines associated with AKT/IAPs signaling ameliorate Alzheimer’s disease development

Abstract

Background

Alzheimer’s disease (AD) is a progressive neurodegenerative condition affecting around 50 million people worldwide. Bone marrow-derived mesenchymal stem cells (BMMSCs) have emerged as a promising source for cellular therapy due to their ability to differentiate into multiple cell types and their paracrine effects. However, the direct injection of BMMSCs can lead to potential unpredictable impairments, prompting a renewed interest in their paracrine effects for AD treatment. The specific mechanism and central role of cytokines in this process have not been fully elucidated.

Methods

Mouse BMMSCs were isolated, validated, and then transplanted intracerebrally into APP/PS1 female mice. The behavioral tests, including open-field test, novel object recognition test, and Morris water maze were performed, followed by β-amyloidosis plaque and neuron apoptosis analyses. Then the tissue RNA sequencing and mBMMSC cytokine analysis were performed. A cytokine antibody array for BMMSCs and the brain slice models were performed with AD model tissues were used to elucidate the molecular mechanisms. Finally, APP/PS1 mice were administrated with cytokine mixture for cognitive recovery.

Results

Our results demonstrated that BMMSCs significantly improved cognitive function, reduced beta-amyloid plaque deposition, and decreased apoptotic neurons through the activation of the AKT signaling pathway. Using a cytokine antibody array, we identified three highly expressed AKT pathway regulated neuroprotective factors in BMMSCs: IGF1, VEGF, and Periostin2. These cytokines were found to upregulate inhibitors of apoptosis family proteins (IAPs) and suppress Caspase-3 activity in brain slices induced with beta amyloidosis (Aβ), okadaic acid (OA), and lipopolysaccharide (LPS). When injection of this cytokine mixture to APP/PS1 mice also resulted in a mitigation of cognitive impairment.

Conclusions

These findings suggest that the secretory factors IGF1, VEGF, and Periostin2 derived from BMMSCs play a crucial role in neuroprotection by modulating the AKT/IAPs pathway to restore neuronal function. These cytokine sets could be a potential therapeutic strategy for AD and lay the groundwork for promising clinical applications.

Background

Alzheimer’s disease (AD), also referred to as primary senile dementia is the most prevalent neurodegenerative disorder globally, affecting approximately 50 million individuals worldwide [1]. With the ongoing increase in global life expectancy, it is anticipated that the number of AD cases will triple by 2050 [2]. Clinically, AD manifests as the progressive degeneration of various behavioral functions, prominently marked by memory deficits and cognitive impairment [3]. The hallmark pathological features of AD encompass the aggregation of extracellular beta-amyloid (Aβ) plaques, intracellular hyperphosphorylated tau-laden neurofibrillary tangles, chronic neuroinflammation, and neuronal loss [4]. Despite notable advancements in comprehending the pathogenesis of AD, a paucity of effective treatment options persists.

Currently, the approved standard therapeutic approaches encompass cholinesterase inhibition and the utilization of the N-methyl-D-aspartate receptor antagonist, memantine. Regrettably, these interventions exhibit limited efficacy, merely offering temporary respite from the cognitive decline inherent in AD [2, 3]. Since the seminal findings presented in 2016, no novel agents addressing the cognitive symptoms of AD have received global approval [2]. Critically, there is an absence of treatments capable of either curing or preventing neuronal cell death in affected individuals, culminating in an inexorable decline in AD patients.

Mesenchymal stem cells (MSCs) constitute a category of multipotent progenitor cells characterized by their ability for localized self-renewal, multi-lineage differentiation, and low immunogenicity [5]. They are widely recognized as a promising cellular source for regenerative medicine in treating diseases such as AD [4]. MSCs are found distributed in various tissues, including bone marrow, adipose tissue, placenta, and umbilical cord [6, 7]. Among these, bone marrow-derived MSCs (BMMSCs) were the first to be identified and have been designated as the primary source for clinical applications [8, 9]. The use of BMMSCs offers a potentially effective therapeutic approach for AD due to their high accessibility and relative ease of implementation.

Numerous research studies have demonstrated a significant improvement in cognitive decline and neuropathological symptoms in AD-like animal models following treatment with BMMSCs [10]. However, the ability of BMMSCs to cross the blood-brain barrier (BBB) to reach the site of injury is limited, and surgery is invasive and may pose trauma and surgical complications to the patient [11]. As a result, the paracrine effect has emerged as a promising alternative approach for AD repair in future clinical applications. BMMSC-derived cytokines, including neurotrophic factors, growth-promoting factors, inflammatory cytokines, chemokines, fibrogenic cytokines, and pro- and anti-angiogenesis factors, play a crucial role in tissue remodeling through the microenvironment [12]. A comprehensive understanding of how BMMSCs improve AD-related behavioral mechanisms and identify the central sets of factors is essential for developing effective and convenient treatment strategies for AD.

This study aimed to investigate the potential of BMMSC transplantation to protect neurons from damage and improve cognitive impairment using an APP/PS1 mouse model. Additionally, we employed an antibody assay to detect the secretome from BMMSCs and utilized brain slices cultured in vitro to elucidate the underlying molecular mechanisms and assess their potential neuroprotective effects. Finally, we validated whether the neuroprotective cytokine sets could alleviate cognitive impairment in APP/PS1 mice through behavioral tests.

Materials and methods

Animals

The APP/PS1 double-transgenic mouse lines, APPswe/PS1dE9, were obtained from the Institute of Laboratory Animal Science, Chinese Academy of Medical Science (CAMS). This study utilized seven-month-old female APP/PS1 mice (n = 32) and their wild-type (WT) littermate C57BL/6J controls (n = 14). All animals were accommodated in IVC cage at a specific, pathogen-free (SPF) room with controlled temperature and light/dark cycles. Mice designated for BMMSC isolation were 5–7 days old (n = 3) and procured from Beijing HFK Bioscience Company (Beijing, China). All animal experiments were approved by the Institutional Animal Care and Use Committee of the Institute of Laboratory Animal Science (Approval date: 1th March 2019; Approval number: WKW19001; Project title: Molecular mechanism study of bone marrow mesenchymal stem cell therapy for Alzheimer’s disease), which is carried out according to ARRIVE guidelines (https://www.nc3rs.org.uk/arrive-guidelines).

mBMMSC preparation

Briefly, mBMMSCs were isolated from both femurs and tibias using the collagenase II (Invitrogen, C6885, USA) bone digestion method under sterile conditions and as previously described [13]. The resultant BMMSCs underwent cultivation in α-MEM (Gibco, 12571-063) supplemented with 10% fetal bovine serum (FBS, Gibco, 10099141 C) and 1% penicillin-streptomycin (PS, Gibco, 15070063). Cultivation transpired in a 5% CO₂-humidified atmosphere at 37 °C, with passages occurring every 2–3 days using 0.25% trypsin containing EDTA (Gibco, 25200-056) when cells reached 70% confluence. The mBMMSCs utilized in our experimentation spanned passages 2 to 4. Osteogenic, adipogenic, and chondrogenic differentiations were achieved through induction with respective media (Cyagen: MUBMX-90021, MUBMX-90031, and MUBMX-90042), followed by staining with Alizarin Red S, Oil Red O, and Alcian blue, respectively.

For the characterization of mBMMSCs, the expression of surface-specific markers, namely CD29 (Invitrogen 12-0291-83, MA, USA), Sca I (Invitrogen 25-5981-82), CD90 (Invitrogen 12-0909-42, MA, USA), CD73 (Invitrogen 17-0739-42, MA, USA), CD34 (Invitrogen 11-0341-81), and CD45 (Invitrogen 45-0451-82), was assessed through flow cytometry analysis. Briefly, mBMMSCs (5 × 10⁵) were incubated with specific antibodies in the dark for approximately 20 min. The cells were then washed with PBS, filtered using a nylon membrane, and subjected to analysis using a flow cytometer.

mBMMSC transplantation

The hippocampal DG district of both APP/PS1 and WT control mice received the transplantation of either mBMMSCs suspension or α-MEM complete medium in the third passage (1.3 cm, 2.1 cm, and 1.7 cm). The littermates were randomly assigned as WT + MEM, WT + BMMSCs, APP/PS1 + MEM, and APP/PS1 + BMMSCs, n = 7 per group. Mice were immobilized with a stereotaxic apparatus and subjected to anesthesia with inspiratory isoflurane. The injection, utilizing a 10-µl syringe and an automated syringe pump, delivered the BMMSC suspension. Bilateral injections were administered, and approximately 2 × 10⁵ cells (2 µl) were injected at a rate of 0.3 µl/min, adhering to the manufacturer’s instructions. To facilitate absorption, the syringe remained in its injecting position for approximately 5 min after each injection. Following transplantation, a three-month interval preceded the commencement of all behavioral testing and pathological analyses for the mice. This transplantation was performed as a non-blinded approach.

Cytokine mixture injection in APP/PS1 mice

APP/PS1 mice were randomly divided into two groups: the control and cytokine groups: APP/PS1 + PBS, and APP/PS1 + cytokines, n = 9 per group. A cytokine mixture comprising IGF1 (20 µg), VEGF (2 µg), and Periostin2 (2 µg) was suspended in 100 µl of PBS. Subsequently, animals were administered weekly injections of 100 µl PBS or cytokines via the tail vein. The animals were assigned as control and cytokines. All the mouse weights were recorded weekly. Following six injections, behavioral tests were conducted. The injection was performed as a non-blinded approach.

Behavioral tests

After a three-month period, cognitive and learning abilities of the animals were assessed sequentially through the Open-field test, novel object recognition test, and Morris water maze. Prior to testing, a 1-h habituation period in the behavioral unit was implemented for all mice. Ethovision XT’s real-time video tracking system (Noldus Information Technology, Wageningen, The Netherlands) was utilized for the recording and analysis of all behavioral tests. The behavioral tests were conducted between 9 a.m and 5 p.m. Animals which spent a remarkably long time immobile, can not swim, or showed very different exploratory performances than other animals were excluded for analysis. Finally, there are 5 animals per group for BMMSCs therapy AD and 7 mice per group for cytokines therapy AD for statistical analysis. I am aware of the group allocation at all the different stages of the experiments, whereas the technician did not know the group assignment at the stage of “the conduct of the experiment”, and output “the outcome assessment”.

Open-field test

The assessment of motor and anxiety-like behaviors was conducted utilizing the open field paradigm. The experiment was commenced at 10:00 a.m. The square arena, measuring 50 × 50 cm with walls 30 cm high, was partitioned into distinct areas: center, transition, and border areas. Mice, placed in the center of the field, were allowed 5 min of unrestricted movement, recorded by a CCD camera linked to a computer. The field was cleaned with 75% alcohol between tests, and each mouse underwent testing only once. Motor behavior was quantified using both velocity and distance as indices. Anxiety-like behavior was assessed based on time spent and crossing entries into the center area, where increased time in the center area indicated reduced anxiety-like behavior [14, 15]. The cumulative duration in center zone (s) and frequency in center zone (times) were recorded for further analysis.

Novel object recognition test

The assessment of exploratory behavior employed a square testing arena measuring 70 × 70 cm with walls 35 cm high, akin to the open-field test dimensions. The arena contained two sets of wooden objects: one set comprising two identical green cubes, and the other consisting of a green cube and a violet triangle cone.

Two trials were conducted. In the initial trial, mice were placed at the arena’s center for 5 min of unrestricted exploration. The experiment was commenced at 10:00 a.m. Subsequently, in the second trial, conducted after a 1-h interval, mice were tested for 5 min following the same sequence. To eliminate olfactory influences, the apparatus and objects were cleaned with 75% alcohol before testing each new animal. The time a mouse spent with the new object was denoted as A, while the time spent with the familiar object was denoted as B. Enhanced time around the new object signified stronger memory retention³⁴. The cumulative duration during new object (s), frequency in new object (times) were recorded for further analysis.

Morris water maze

This method constitutes a classic spatial learning and memory test. The maze, measuring 1.2 m in diameter and 30 cm in height, was filled with 20 cm of opaque water achieved through the use of non-toxic, white paint. The maze was quadrant-divided, featuring a 10 cm diameter platform submerged 6–8 mm below the water surface in the first quadrant.

The mice underwent training for five consecutive days and were tested on the sixth day. During the training phase, mice experienced the water maze three times daily, being randomly placed in either the second, third, or fourth quadrants and carried out from 9:00 a.m. They were given 60 s to find the submerged platform. If successful, the mice remained on it for 15 s; if unsuccessful, they were guided to the platform and allowed to remain for 15 s. The time taken for the mice to climb onto the platform was recorded as latency. On test day (day 6), the platform was removed, and the mice, placed in the third quadrant, had 60 s to explore, with all movements recorded by a CCD camera connected to a video monitor and computer. The testing time was also scheduled for 9:00 a.m, following the same sequences.

A visual experiment tested their eyesight. The platform, fixed in the first quadrant, prompted mice in the second quadrant to explore for 5 min. If the mouse did not find the platform, its location changed, and the test was repeated. If the mouse still could not find the target, it was excluded from all subsequent statistical analyses. All qualified mice were included in the final statistical analysis regarding their latency and crossing times. Animals’ latency (s), frequency (times), cumulative duration (s), speed (cm/s) were recorded for further analysis.

Tissue preparation

Following the completion of the behavioral tests, the mice underwent rapid sacrifice through cervical dislocation. Brains were promptly dissected, and both cortex and hippocampi tissues were partitioned. All harvested tissue was immersed in ice-cold phosphate buffer saline (PBS, Gibco, 14190-144) and washed twice. Tissue designated for subsequent histopathological examination was fixed in 4% paraformaldehyde at 4 °C for 2–3 days. Tissue allocated for protein/RNA detection underwent washing and then rapid freezing in liquid nitrogen. The frozen tissue was then rapidly transferred to a -80 °C refrigerator for subsequent analysis.

Histology and immunohistochemistry

Fixed tissues underwent washing, alcohol dehydration, xylene clearing, and paraffin embedding at 56 °C for an additional 24 h. Sections of 4 μm thickness were then stained with hematoxylin and eosin (H&E). For all immunohistochemistry procedures, sections were dewaxed using xylene and gradient ethanol. Antigen repair was executed via citric acid in a microwave oven, followed by quenching of endogenous peroxidase enzyme activity and blocking with H₂O₂ and 5% goat serum. The following primary antibodies were incubated overnight at 4 °C: anti-beta amyloid (Biolegend, 800709, 1:150, USA), anti-NeuN (abcam, ab177487, 1:200, USA), anti-GFAP (abcam, ab7260, 1:100), and anti-Iba 1 (abcam, ab5076, 1:100) antibodies. Following PBS washes, tissues were subjected to enhanced second antibody reagents at room temperature for 10–30 min in accordance with the manufacturer’s instructions. After incubation, tissues underwent three PBS washes, and the DAB reagent was applied to the sections until a visible color change occurred. Two blinded observers conducted all histopathological processing and assessments to mitigate bias.

Thioflavin S staining

Tissue sections underwent thioflavin S labeling to assess the presence of amyloidogenic fibrils in plaques. Following sequential ethanol dewaxing and immersion in PBS, the dewaxed tissue sections were stained with Thioflavin S solution (0.4 M in 50% ethanol) for 1–2 min, followed by rapid washing (1–2 times). Subsequently, all tissue slices were sealed with 20% glycerin. Fluorescence microscopy was employed for plaque imaging. Two blinded observers conducted all histopathological processing and assessments to mitigate bias.

Transmission electron microscopy

Tissue sections underwent fixation overnight at 4 °C with 2.5% glutaraldehyde, followed by three rinses with PBS. The tissues were then fixed with 1% osmic acid at 4 °C for 2 h, followed by three washes with deionized water washes. The sections underwent dehydration in a gradient of 70%, 90%, and 100% ethanol, along with propylene oxide replacement treatment. Post-dehydration, they were immersed in resin, embedded, and precision-sliced into ultrathin sections (90 nm thick). Following staining with uranyl acetate and lead citrate, imaging was conducted using a JEM - 1400Plus electron microscope. Two blinded observers conducted all histopathological processing and assessments to mitigate bias.

RNA-based next-generation transcriptome sequencing

Hippocampal tissue underwent dissection and extraction utilizing TRIzol reagent (Invitrogen, 15596018). Following the assessment of RNA integrity, concentration, and purity, RNA sequencing libraries were constructed using an RNA Library Prep Kit (NEB, USA). Subsequent to library quantification and size determination via a Fragment Analyzer, Illumina performed the sequencing as facilitated by Majorbio. DEGs, with a fold change exceeding 2 folds and a P value < 0.05, were identified for further analysis, presented in Venn diagrams generated by an available tool (http://bioinformatics.psb.ugent.be/webtools/Venn/). KEGG and GO analyses were executed using KOBAS software and the GOseq R package, respectively. GSEA employed the GSEA algorithm. All RNA sequencing data accession numbers were deposited in the NCBI database and are available in GEO under accession number PRJNA38746.

Antibody array analysis

The antibody array analysis utilized the Mouse Cytokine Array Q400 (RayBiotech Inc., Norcross, GA, USA) following the manufacturer’s instructions. Upon reaching 80% confluence, mBMMSCs were cultured, washed with PBS, and then incubated with serum-free α-MEM. After 24 h, the conditioned medium was collected and applied to array membranes in a blocking buffer for a 30-min incubation. Subsequently, biotin-conjugated antibodies were applied, followed by a 1-h incubation with fluorescent dye-conjugated streptavidin after three washes. Signal intensities were quantified using Array-Pro Analyzer^® Ver. 4.5 (Mediacy Bernetics, Inc., Bethesda, MD, USA).

Brain slices culture and induction by Aβ, OA, and LPS

Mice aged 6–8 weeks (C57/BL) were euthanized, and their brains were promptly dissected. The cerebral cortex and hippocampus, desired brain regions, were then sliced into 300 μm sections using a vibratome (Leica VT1200/S, Germany). These slices were kept in ice-cold, oxygenated medium throughout the process. Subsequently, the tissues were immersed in DMEM/F12 complete medium (DMEM/F12 + 10% FBS + 1% PS) for a rotating culture. Aβ protein (1 µg/ml), okadaic acid (OA) (20 nM), and lipopolysaccharide (LPS) (1 µg/ml) (Sigma-Aldrich, MO, USA) at specified concentrations were introduced to the brain slices, which were then cultured at 37 °C in a 5% CO₂ environment for 24 h. Following this, IGF1 (0.5 µg/ml), VEGF (2 µg/ml), and Periostin 2 (2 µg/ml) (Pepro Tech, New Jersey, USA) were added to the culture system, allowing for a 24-h rotating incubation. The tissue mass was washed with PBS, and then collected and stored at -80 °C for subsequent qPCR and immunoblotting analyses.

Caspase 3 activity

Caspase3 activity of brain tissue sections were performed using the Caspase3 Activity Assay Kit (Beyotime Biotechnology, C1115, Shanghai, China). Tissue sections (5 mg) were lysed using lysis buffer for 5 min on ice bath. The homogenate supernatants were harvested by centrifugation at 12 000 rpm for 10 min at 4 °C. The protein concentration was measured using a BCA protein assay kit (Thermo Scientific, 23225, MA, USA). Subsequently, tissue lysates were added in a 96 well microplate, and incubated with Ac-DEVD-pNA (2 mmol/L) at 37 °C for 2 h. After incubation, absorbance was read at a wavelength of 405 nm using Multiskan FC (Thermo Scientific, MA, USA).

Immunoblotting analysis

Brain tissues or sections were obtained and lysed on ice using RIPA lysate (Beyotime Biotechnology, P0013C, Shanghai, China) supplemented with a protease inhibitor cocktail (Roche, 11836170001, Basel, Germany) for 20 min. The cell lysate was then centrifuged at 12,000 rpm for 10 min at 4 °C, resulting in the collection of the supernatant. The protein concentration was determined using a commercially available BCA reagent kit (Beyotime Biotechnology, P0012). Next, 20 µg of total protein was loaded and separated through 10–15% SDS-PAGE, followed by transfer to PVDF membranes (GE Healthcare Life Sciences, RPN303F, IL, USA) at 300 mA for 1–2 h based on the molecular weight of the target proteins. Following transfer, the membranes were blocked with 5% skim milk or bovine serum albumin (for phosphorylated protein) for 1 h at room temperature. Subsequent to blocking, the membranes were incubated overnight at 4 °C with primary antibodies targeting specific proteins. The blots were washed three times with TBST buffer and then incubated with the appropriate secondary antibodies for 1–2 h at room temperature. Signal bands were visualized using a commercially available ECL reagents kit (Millipore, WBULS0500, MA, USA).

Reverse transcription and quantitative PCR

Cell pellets and brain tissue mRNA were extracted utilizing TRIzol reagent with the inclusion of chloroform and isopropanol, followed by dissolution in DEPC water. Then, 2 µg of RNA was reverse transcribed to cDNA using the Revert Aid First Strand cDNA Synthesis Kit (ThermoFisher Scientific, K1622, MA, USA) and oligo (dT) primers. Quantitative PCR was performed using TB Green Premix Ex Taq TM II (TliRNaseH Plus) (TaKaRa, RR82WR, Japan) on the Step One Realtime Cycler. The standard two-step procedure was applied, and the relative expression was determined using the ΔΔCt relative quantification method.

Statistical analysis

Data are expressed as mean ± SEM, and statistical comparisons between two groups were conducted using Student’s t-test. One-way ANOVA was applied to the multiple group comparison followed by a Tukey post-hoc test. P value < 0.05 was considered statistically significant. Data were derived from at least three independent experiments. Statistical analyses were performed using GraphPad 8.0 and SPSS 19.0.

Results

Isolation, validation, and characterization of mBMMSCs

Primary mouse BMMSCs were isolated from the femur and tibia through bone splice digestion, as depicted in Fig. 1A. Microscopic examination confirmed the typical morphology of mesenchymal stem cells, characterized by a homogeneous spindle shape (Fig. 1B). The CCK8 assay demonstrated an exponential growth curve for these cells (Supplementary Fig. 1A).

Fig. 1

Isolation, validation, and characterization of mBMMSCs. (A) Flow chart illustrating the isolation of mBMMSCs through bone slice digestion. (B) Morphological characterization using an inverted microscope. Scale bar, 100 μm. (C) Representative images of Oil Red O (left), Alizarin Red S (middle), and Alcian blue (right) staining following adipogenic, osteogenetic, and chondrogenic induction, respectively. (D) Flow cytometry analysis of surface antigen markers CD34, CD45, CD29, and Sca-I

To assess BMMSCs’ differentiation pluripotency, the third generation of BMMSCs underwent examination. Positive staining with Oil red O, Alizarin Red S, and Alcian blue following adipogenic, osteogenic, and chondrogenic induction, respectively, was observed (Fig. 1C). FACS analysis of cell surface marker expression revealed positivity for CD29, Sca I, CD90, and CD73, while hematopoietic markers CD34 and CD45 were present in less than 5% of cells (Fig. 1D, Supplementary Fig. 1B). Collectively, these results affirmed the MSC characteristics of these BMMSCs.

Treatment of APP/PS1 mice with mBMMSCs improves cognitive and memory function

To assess the potential alleviation of cognitive and memory dysfunction in APP/PS1 mice through in vivo induction of mBMMSCs, bilateral injections were administered into the hippocampus, followed by behavioral testing after a three-month interval (Fig. 2A).

Fig. 2

Treatment of APP/PS1 mice with mBMMSCs improves cognitive and memory function. (A) Schematic representation of mBMMSCs injection and behavioral study regime. (B, C) Representative paths (B) and quantification of cumulative duration in center zones (C) (n = 5). A, center area, B, transition area, C, border area. Values are mean ± SEM. (D, E) Representative paths (D) and quantification of cumulative duration during new object exploration (E) (n = 5). A, new object, B, familiar object. Values are mean ± SEM. (F) Water maze escape latencies recorded during 5-day training (n = 5). (G) Representative swimming paths in the probe test. (H) Water maze escape latencies in the 60-second probe trial (n = 5). (I) Number of times each mouse crossed the target zone during the probe trial (n = 5). I, WT + MEM, II, WT + BMMSCs, III, APP/PS1 + MEM, IV, APP/PS1 + BMMSCs. ns, not significant; *, P < 0.05

In the open-field test, mice subjected to mBMMSCs treatment exhibited a discernible inclination towards an increased cumulative duration and frequency in the center zone, suggesting reduced anxiety levels in comparison to the control groups (Fig. 2B, C, Supplementary Fig. 2A). Subsequently, the novel object recognition test revealed a noteworthy cumulative duration spent with the new object in the mBMMSCs-implanted AD group (Fig. 2D, E, Supplementary Fig. 2B). Notably, during the Morris water maze test training trials, the APP/PS1 mice treated with mBMMSCs demonstrated a significantly shorter escape latency than the control group (Fig. 2F). In the probe test, the APP/PS1 – mBMMSCs group exhibited a reduced escape latency and increased frequency in platform crossings (Fig. 2G-I). The increased cumulative duration in platform and mean speed trend were also observed in APP/PS1-mBMMSCs mice compared with APP/PS1-MEM, although without reaching statistical significance (Supplementary Fig. 2C, D). In summary, these results collectively suggest that the mBMMSCs-treated groups displayed enhanced search strategies and improved cognitive and memory function.

mBMMSCs injection decreases β-amyloidosis plaque deposition and neuron apoptosis in APP/PS1 mice

Our initial investigation focused on assessing the impact of mBMMSCs on Aβ deposition, employing Aβ and Thioflavin S staining techniques. The results revealed a notable reduction in Aβ plaques in both the cortex and hippocampi of mice treated with mBMMSCs, with approximately half the plaques observed in the cortex and one-third in the hippocampi compared to control mice (Fig. 3A, B). Similar results were observed in the Thioflavin S staining assay, demonstrating a 1.5- to 3-fold decrease in plaque deposition in mBMMSCs-treated tissue relative to the α-MEM group (Fig. 3C, D). We then analyzed cell-type specificity using Neu N, Iba-1, and GFAP, revealing a significant increase in neuron numbers and a notable decrease in microglia and astrocytes. This phenomenon indicates that mBMMSCs transplantation engages various cell types, promoting neurogenesis and reducing the inflammatory response (Supplementary Fig. 3). Since Aβ peptide deposition induced neurotoxicity, we next investigate whether mBMMSCs can protect neurons from death. In AD-like mice treated with mBMMSCs, we observed a significant downregulation of cleaved Caspase3 via immunoblotting analysis. There was only a slight reduction in p-MLKL/ MLKL, and minimal changes in GPX4 and FACL4 (Fig. 3E, Supplementary Fig. 4). These results indicated that the reduction of apoptosis is the major mechanism of cell death in response to mBMMSCs injection. Given the pivotal role of mitochondria in cellular homeostasis and their implications for cognitive deficits in AD, we conducted an assessment of cellular ultrastructure through transmission electron microscopy. Mitochondria in APP/PS1 mice exhibited swollen and broken cristae, along with membrane rupture. In contrast, BMMSCs treatment led to an improvement in mitochondrial morphology in AD mice (Fig. 3F).

Fig. 3

mBMMSCs injection decreases β-amyloidosis plaques and neuron apoptosis in APP/PS1 mice. (A) Coronal brain sections immunostained with anti-Aβ antibody. Scale bar, 200 μm. (B) Quantification of aggregated Aβ42 per field in the cortex (top) and hippocampus (bottom) (n = 5). Values are mean ± SEM. (C) Brain sections from the cortex (top) and hippocampus (bottom) were stained with thioflavin S in WT and APP/PS1 mice injected with MEM or mBMMSCs. Scale bar, 100 μm. (D) Quantification of thioflavin S positive plaque burden per field in the cortex (top) and hippocampus (bottom) (n = 5). Values are mean ± SEM. (E) Immunoblotting analysis of Caspase 3 in AD mouse hippocampus following mBMMSC transplantation. The full-length blots and the original blots were presented in Supplementary Fig. 7. (F) Representative images of mitochondria in hippocampal sections. Scale bar, 500 nm. Values are mean ± SEM. **, P < 0.01; ****, P < 0.0001

mBMMSCs mediated cognitive deficits recovery via activation of AKT signaling pathway

The mediation of cognitive deficits recovery by mBMMSCs was further explored through the activation of the AKT signaling pathway. RNA sequencing analysis of hippocampal tissues from WT-MEM, WT-BMMSCs, APP/PS1-MEM, and APP/PS1-BMMSCs mice identified 302 differentially expressed genes (DEGs). Among these, 122 were regulated in the WT-MEM vs. WT-BMMSCs group, and 199 were regulated in the AD-MEM vs. AD-BMMSCs group (Fig. 4A-C). Gene ontology (GO) term enrichment elucidated the functional categories and biological pathways of the 180 specifically regulated DEGs in the AD group. Notably, cell growth, wound healing functions, and the AKT signaling pathway dominated the enriched pathways compared to the MEM control group (Fig. 4D). Gene set enrichment analysis (GSEA) further highlighted the significant activation of downstream target genes in the AKT signaling pathway in the mBMMSCs-treated group (Fig. 4E).

Fig. 4

mBMMSCs-mediated cognitive deficits recovery via activation of AKT signaling pathway. (A) Heatmap illustrating the DEGs in mBMMSCs-treated versus control hippocampi in WT (top) and APP/PS1 (bottom) mice. Red indicates upregulated genes; blue indicates downregulated genes. (B) Volcano plots depicting gene expression in hippocampi from mBMMSCs-treated versus control (left) and APP/PS1 (right). Red indicates upregulated genes, blue indicates downregulated genes, and grey indicates non-significant genes. (C) Venn diagram showing DEGs in WT and APP/PS1 brains treated with either MEM or mBMMSCs. (D) Biological pathway based on gene ontology (GO) term analysis. (E) Gene set enrichment analysis (GSEA) of the AKT signaling pathway in mBMMSCs-treated versus control hippocampus in APP/PS1 mice

Growth factors secreted from BMMSCs rescues AD brain slice apoptosis by AKT signaling

Having elucidated the impact of BMMSCs secreted factors, our focus shifted to understanding the mechanisms through which BMMSCs modulate cognitive function recovery and mitigate neurological damage. Employing a cytokine antibody array, we identified key secreted factors (Supplementary Table 1), including IGF1, VEGF, and Periostin2, with robust signals exceeding 100,000, known for their involvement in tissue repair (Supplementary Fig. 5). To simulate amyloid deposition, Tau protein phosphorylation, and neuroinflammation, mouse brain slices were induced with Aβ, OA, and LPS, respectively, followed by treatment with IGF1, VEGF, and Periostin2 (Fig. 5A). In groups subjected to Aβ, OA, and LPS damage, elevated Caspase 3 activity and cleaved/total Caspase 3 were identified compared to control cells, whereas incubation with IGF1, VEGF, and Periostin2 effectively reversed these changes to baseline levels (Fig. 5B, C).

Fig. 5

Growth factors secreted from BMMSCs rescue AD brain slice apoptosis via AKT signaling. (A) Working model of brain slice culture and the time course for Aβ, OA, LPS, as well as VEGF, Periostin 2, and IGF1 treatment. (B) Caspase 3 activity assessed in brain slices treated with VEGF, Periostin 2, and IGF1 in the presence or absence of Aβ, OA, or LPS (n = 3). (C) Expression of Caspase 3 in brain slices analyzed in the indicated groups by immunoblotting. Numbers above the panels represent normalized cleaved/total Caspase3 expression levels, with β-actin as the loading control. The full-length blots and the original blots were presented in Supplementary Fig. 7. (D) Heatmap illustrating qRT-PCR analysis of IAPs (Apollon, cIAP1, cIAP2, Livin, NAIP, Survivin, XIAP) mRNA expression levels in the indicated brain slice groups. The full-length blots and the original blots were presented in Supplementary Fig. 7. (E) Livin and NAIP assessed in the indicated groups by immunoblotting. Numbers above the panels represent normalized protein expression levels, with β-actin as the loading control. The full-length blots and the original blots were presented in Supplementary Fig. 7. (F) Immunoblot analysis of brain slices for p-AKT and AKT in the indicated brain slice groups. Numbers above the panels represent normalized p-AKT expression levels, with β-actin as the loading control

Inhibitor of apoptosis family proteins (IAPs), recognized endogenous apoptosis inhibitors [16, 17], were evaluated through qPCR analysis. NAIP and Livin demonstrated consistent upregulation at the mRNA level (Fig. 5D, Supplementary Table 2), a finding corroborated by immunoblotting analysis at the protein level (Fig. 5E).

Given the crucial role of the AKT signaling pathway in IAP-mediated anti-apoptosis [18, 19], we assessed this pathway through immunoblotting. Treatment with IGF1, VEGF, and Periostin2 significantly upregulated AKT phosphorylation levels (Fig. 5F). These findings collectively demonstrate that BMMSCs secreted factors prevent AD-associated neural cell death by upregulating the AKT/IAPs signaling pathway.

Cytokines alleviate cognitive deficits in APP/PS1 mice

To evaluate the impact of IGF1, VEGF, and Periostin2 on cognitive function, a mixture of these cytokines was intravenously administered to APP/PS1 mice, followed by behavioral analysis (Fig. 6A). In the open-field test, cytokine-treated mice exhibited a tendency and significantly increased frequency to remain in the center zone (Fig. 6B, C, Supplementary Fig. 6A). The novel object recognition test indicated that mice in the cytokine group not showed an obvious preference for interacting with the novel object (Fig. 6D, E, Supplementary Fig. 6B). During the training trials of the Morris water maze test, all animals exhibited a gradual learning curve, manifesting in shortened escape latencies to the hidden platform. On the 4th and 5th day, the APP/PS1 + cytokine group demonstrated significantly shorter escape latencies compared to the APP/PS1 control group (Fig. 6F). In the probe test, animals in the cytokine group spent a shorter time swimming to the platform zone, and an increased crossed frequency and speed trend than APP/PS1 mice (Fig. 6G-I, Supplementary Fig. 6C, D). We then analyzed β-amyloid plaques and neuronal apoptosis in the cytokine-treated group. The results showed demonstrated that cytokine administration reduced plaques burden approximately 25% in the cortex and 35% in the hippocampus compared to control mice (Fig. 6J-L). Moreover, cleaved-Caspase 3 levels, indicative of neuronal apoptosis, were significantly reduced relative to the control group (Fig. 6M). These results mirrored those observed with mBMMSCs, suggesting that this cytokine sets can ameliorate certain phenotypic aspects of AD mice.

Fig. 6

Cytokines alleviate cognitive deficits in APP/PS1 mice. (A) Schematic diagram illustrating the therapeutic potential of a cytokine mixture in APP/PS1 mice via six tail vein injections. (B, C) Open-field test results, depicting representative paths (B) and quantification of cumulative duration in center zones (C) (n = 7). A, center area, B, transition area, C, border area, Values are mean ± SEM. (D, E) Representative paths (D) and quantification of cumulative duration with new objects (E) (n = 7). A, new object, B, familiar object. Values are mean ± SEM. (F) Water maze escape latencies recorded during 5-day training (n = 7). (G) Representative swimming paths in the probe test, with the green circle indicating the platform. (H) Water maze escape latencies in the 60-second probe trial (n = 7). (I) Number of times each mouse crossed the target zone during the probe trial (n = 7). (J) Coronal brain sections immunostained with anti-Aβ antibody. Scale bar, 200 μm. (K, L) Quantification of aggregated Aβ42 per field in the cortex (K) and hippocampus (L) (n = 7). Values are mean ± SEM. (M) Immunoblotting analysis of Caspase 3 in AD mouse hippocampus following cytokines injection. The full-length blots and the original blots were presented in Supplementary Fig. 7. (N) Schematic diagram of the therapeutic potential of mBMMSCs in AD. ns, not significant; *, P < 0.05; **, P < 0.01; ****, P < 0.0001

Discussion

AD imposes a wide-ranging impact on physical, psychological, social, and economic dimensions for individuals diagnosed with AD, as well as their caregivers, families, and society at large [20]. Over the last decade, several clinical trials targeting AD have been initiated, yet their outcomes have uniformly resulted in substantial failures [21,22,23]. Presently, there exists a critical unmet need for novel AD treatments, as there are no existing therapies capable of either curing AD or modifying its progression.

MSCs, a heterogeneous cluster of multipotent adult stem cells, have demonstrated effectiveness in the treatment of neurodegenerative diseases [24]. Isolated from diverse tissues, including bone marrow, adipose, brain, placenta, and umbilical cord, MSCs derived from bone marrow remain an attractive therapeutic option, considering ethical and clinical considerations regarding alloimmunogenicity [10]. In this study, mBMMSCs were isolated through bone digestion, as previously described, and bilaterally transplanted into the hippocampi of an AD-like mouse model [13]. The observed effects of mBMMSCs included improvements in cognitive dysfunction related to learning and memory in the APP/PS1 mouse model. Additionally, transgenic mice showed a reduction in the deposition of Aβ plaques and mitigated mitochondrial damage, along with increased neurogenesis and a decrease in the inflammatory response. Subsequent transcriptional sequencing revealed an enrichment of the AKT signaling pathway, correlating with improvements in animal behavior and pathological phenotypes.

The activation of endogenous AKT/IAPs plays a crucial role in synaptic plasticity, intracranial brain volume, and is essential for cell growth and apoptosis [19, 25]. In our study, we observed that NAIP and Livin are commonly upregulated in the regulation of neuronal death in an AD brain slice model. NAIP and Livin are involved in disrupting the apoptosis signaling pathway by binding to and inhibiting caspase activity, including Caspase-3, -7, -8, and − 9 [26]. Consistent with these findings, in vitro studies have shown that up-regulation of NAIP and decreased interaction of NAIP with its endogenous inhibitor, Smac, in the presence of neurotrophin-3, protect neurons from Aβ-induced death [27]. Additionally, Fucoidan has been found to be neuroprotective against Aβ-induced neurotoxicity by upregulating Livin expression [28]. These findings highlight the pathophysiological significance of anti-apoptosis in the pathogenesis of AD following transplantation of mBMMSCs.

MSCs engraftment is usually limited by the rigorous tissue structure around the injured site, where allografted cells undergo cell death or be engulfed by resident cells. MSCs to the brain due to potential unpredictable impairments caused by cell injection. In recent decades, an increasing body of evidence has demonstrated that not only mesenchymal stem cells but also conditioned media or exosomes can alleviate pathological cognitive decline [29, 30]. The media and exosomes varied from MSCs status and they are difficult to standardization. The therapeutic and regenerative effects of MSCs may extend beyond local mechanisms to involve the secretion of paracrine cytokines. Paracrine cytokines are considered valuable contributors to our understanding of a diverse range of molecular mechanisms. They may represent an ideal potential substitute therapy, given their ability to traverse the BBB with potency similar to their parental cells [30].

In this study, we screened 200 cytokines from conditioned media of BMMSCs, identifying several highly expressed neurotrophic factors and growth-promoting factors, including IGF1, VEGF, and Periostin2, associated with the AKT signaling pathway. They enhanced angiogenesis, promoted neuronal survival, improved neuronal plasticity, and facilitated tissue repair [31,32,33]. Subsequent examination revealed that these cytokines reduced neuronal death in an in vitro mouse brain slice system, providing an ex vivo mimicry of AD. Furthermore, considering the presence of transport receptors such as IGFR in the BBB, we formulated a mixture of these cytokines for intravenously injection [34]. Through three animal behavior tests, these cytokines demonstrated the ability to alleviate spatial cognitive and memory impairment symptoms in AD. The results were consistent with the therapeutic benefits observed with BMMSCs transplantation in an AD model, providing additional evidence that these cytokines may represent a crucially regulated paracrine secretion. IGF1 treatment stabilizes the microvascular cytoskeleton, reduces BBB permeability, and concurrently, therapeutic VEGF increases vascular permeability, which may contribute to tissue edema formation [35,36,37]. Consequently, optimizing cytokine combinations and performing dose-response studies are critical for the future investigations.

We also found that mBMMSC transplantation yielded statistically superior outcomes (Fig. 2) compared to cytokine administration (Fig. 6). The differential outcomes may be attributed to several factors. First, mBMMSCs exhibit the ability to migrate throughout the brain and differentiate into neurons, glial cells, and vascular endothelial cells. Second, mBMMSCs secrete a complex mixture of cytokines and extracellular vesicles. Our cytokine antibody array revealed that dozens of cytokines were expressed at signals exceeding 1000, contributing to proliferation, differentiation, apoptosis inhibition, and immunomodulation. These factors enable mBMMSCs not only to clear Aβ plaques and inhibit apoptosis, but also to enhance neurogenesis, differentiation, integration, angiogenesis, and immunomodulatory functions [10]. mBMMSCs also activate several signaling pathways, including ERK, TGF-β1, STAT3 signaling [38,39,40]. Regarding AKT signaling, while IGF1, VEGF, and Periostin2 are key contributors, other cytokines such as HGF, PAI-1, G-GSF, TIMP-1, and glial cell-derived neurotrophic factors as well as extracellular vesicle components like miR-223 andMicroRNA-631 also unregulated AKT signaling [41,42,43]. These diverse and synergistic effects may explain the superior therapeutic outcomes observed with mBMMSC treatment compared to cytokine administration.

Despite the limitations in replicating the complete cytokine milieu of IGF1, VEGF, and Periostin2, direct BMMSC injection poses risks of unpredictable adverse effects in patients. Thus, the paracrine effects of BMMSCs have emerged as a promising alternative for AD repair in future clinical applications. In this study, we confirmed that a cytokine composition of IGF1, VEGF, and Periostin2 mitigates cognitive deficits. Future experiments will focus on identifying additional candidate cytokines, characterizing their individual roles and therapeutic efficacy, and optimizing cytokine combinations in terms of dose, timing, and administration routed for the treatment of Alzheimer’s disease.

Conclusions

In conclusion, mBMMSCs have shown promise in improving cognitive function and reducing neuropathological symptoms in animal models of AD. The therapeutic mechanism of mBMMSCs likely involves enhancing neuronal anti-apoptosis through the activation of the AKT pathway. A cytokine combination, including IGF1, VEGF, and Periostin2, appears to increase NAIP and Livin levels while reducing Caspase-3 activity. Furthermore, this cytokine mixture has demonstrated the ability to improve spatial cognitive and memory impairment in behavioral tests. Our findings suggest that targeting AKT/IAPs could be a promising therapeutic approach, and cytokines derived from BMMSCs may pave the way for future AD therapies (Fig. 6N).