Monocytes release cystatin F dimer to associate with Aβ and aggravate amyloid pathology and cognitive deficits in Alzheimer’s disease

Age-matched controls and AD patients were recruited from Xuanwu Hospital of Capital Medical University and First Hospital of China Medical University (sTable 1). The National Institute of Neurological and Communication Disorders and the Stroke and Alzheimer’s Disease and Related Disorders Association (NINCDS-ADRDA) were used as the basis for the patient selection criteria, and patients with other types of dementia were excluded from this study [38]. All of the healthy control individuals had no substantial comorbidities that could impair brain function and no family history of dementia, and the patients with AD did not have immunological illnesses or vascular risk factors [39]. The Mini-Mental State Examination (MMSE), the Montreal Cognitive Assessment (MoCA) [40], and the Clinical Dementia Rating (CDR) [41, 42] were all used to measure cognitive function and the degree of impairment in all subjects. To evaluate object recognition memory, the Rey Auditory Verbal Learning Test (RAVLT) was used as described previously [43]. All participants (or their legal guardians) provided written informed consent.

The RNA of the primary monocytes from the AD patients and age-matched controls was extracted by using TRIzol reagent (Invitrogen, Carlsbad, CA, Cat No. 15596026CN) according to the instructions. RNA sequencing analysis was carried out by Biomarker Technologies Company (Beijing, China). For data analysis, the fold change (FC), an absolute fold change between the two groups for each transcript, was calculated when comparing the two groups with distinct profiles. A Student's t test was used to determine the statistical significance of the differences between groups. The significantly differentially expressed transcripts had a FC of at least 4 and p values less than 0.05. By dividing the total number of genes in the Gene Ontology (GO) category by the total number of differentially expressed genes, enrichment factors were computed. Gene Set Enrichment Analysis (GSEA) v2.0.14 software was used to perform GSEA analysis.

APP/PS1 transgenic mice were obtained from China Medical University. 5 × FAD transgenic mice were generous gifts from Professor Chao Wang of Chongqing Medical University. Monocyte-specific human cystatin F-overexpressing transgenic mice were generated as described in previous reports [44, 45]. Briefly, the open reading frame (ORF) of human cystatin F was obtained and subcloned and inserted into pBluescript SK (-) under the control of the human CD68 promoter. The fragment was removed and injected into fertilized C57BL/6 mice to construct transgenic mice named Hmo-cys F⁺ mice. Hmo-cys F⁺ and APP/PS1 AD model mice were bred to produce APP/PS1/Hmo-cys F⁺ double transgenic mice. All transgenic lines were backcrossed to the C57BL/6J strain. Mice were housed under specific pathogen-free conditions on a 12 h light/12 h dark cycle with free access to water and food and no more than five mice per cage. The temperature and humidity were 18–29 °C and 45–55%, respectively. In this study, only male 9-month-old APP/PS1 mice and 3-month-old 5XFAD mice were included. All the animal experimental procedures were approved by the Animal Experimentation Ethics Committee of China Medical University (CMU20240091).

The spatial memory and learning abilities of the mice were assessed by the MWM test. One day before the experiment, the mice were allowed to become accustomed to the water maze (100 cm in diameter). The pool was filled with water that was rendered opaque, which was drained every day, and the temperature of the water was maintained at 19 °C to 22 °C. During the training period, the mice were permitted to swim freely for 60 s to locate the platform (9 cm in diameter), which was fixed at a depth of 1 cm below the surface of the water. Mice that did not find the platform were directed to it and allowed to rest for 30 s. The mice were trained four times per day for six days. On day seven, the platform was removed, and the swimming activity of each mouse, including the latency, number of platform crossings and velocity within 1 min, was recorded. All the processes were monitored by using a video camera mounted overhead and automatically recorded via ANY-maze behavioural tracking software (Stoelting, Wood Dale, IL, USA).

The NOR test was conducted as described previously [46]. Initially, the mice were acclimated to the testing room for five days before testing. Then, the mice were allowed to explore freely in the empty arena (40 × 40 × 40 cm) for 10 min. For the training session, two identical cylinders (6 cm in diameter) were placed in the arena 5 cm away from the wall. The mice were individually placed in the arena and allowed to freely explore for 10 min. Twenty-four hours after the training session, the NOR test was carried out. One of the familiar cylinders was replaced with a cube (with a side length of 6 cm). The mice were allowed to explore for 10 min, and a camera tracking system (TopScan Suite, CleverSys Inc., Reston, VA, USA) was used to record the amount of time the mice spent exploring the objects. The recognition index = time spent exploring the new object/(time spent exploring the new object + time spent exploring the familiar object) × 100%.

Three-month-old 5xFAD mice were given tail intravenous injections of murine cystatin F dimer protein at a dose of 200 μg/kg every three days for one month. The experiments were performed 48 h after the last injection.

Briefly, mice were transcardially perfused with phosphate-buffered saline (PBS) followed by ice-cold 4% paraformaldehyde (PFA) in PBS (pH 7.4). After perfusion, the brain was removed and postfixed in 4% PFA in PBS at 4 °C overnight and then transferred to 30% sucrose for 2-3 days until it sank to the bottom of the container. Coronal sections of the brain (30 μm) were cut using a cryostat (Minux FS800A, RWD Inc., China), collected serially on coated glass slides, and stored at − 20 °C until use. Six sections obtained from each brain containing the cortex and CA1 area of the hippocampus were chosen for immunofluorescence staining. The brain sections were incubated with PBS containing 0.5% Triton X-100 for 10 min and 5% BSA for 1 h at room temperature. The brain sections were sequentially incubated overnight at 4 °C with primary antibodies recognizing Aβ (1:800; Cell Signaling Technology, MA, USA, Cat No. 8243S), His tag (1:500; ABclonal Technology, Wuhan, China, Cat No. AE003) and IBAI (1:1000; Abcam, Cambs, UK, Cat No. ab178846). Then, the sections were incubated with Alexa Fluor 488-conjugated donkey-anti-rabbit IgG (1:200; Invitrogen, CA, USA, Cat No. A-21206) for 3 h in the dark at room temperature. After counterstaining with 4′,6-diamidino-2-phenylindole (DAPI) (Sigma Aldrich, MO, USA, Cat No. 28718-90-3) for 5 min, the coverslips were mounted and the slides were analysed under a laser scanning confocal microscope (LSCM).

The levels of soluble and insoluble Aβ in the prefrontal cortex were measured by ELISA method. Briefly, fresh brain tissues from the mice were homogenized in RIPA buffer [50 mM Tris (pH 7.4), 150 mM NaCl, 1% NP-40, and 0.5% sodium deoxycholate] supplemented with protease inhibitor cocktail, followed by centrifugation at 12,000 × g for 30 min at 4 °C. The supernatant was collected as the RIPA fraction, and the pellet was solubilized in 2% SDS and 25 mM Tris-HCl (pH 7.4) followed by centrifugation at 12,000 × g for 30 min at 4 °C as the SDS fraction. The concentrations of Aβ40 and Aβ42 were measured using ELISA kits (Elabscience Biotechnology Company, Wuhan, China, Cat No. E-EL-M3009; E-EL-M3010) according to the manufacturer’s instructions. The Aβ40 and Aβ42 in human plasma were also determined by the ELISA kits (Elabscience Biotechnology Company, Wuhan, China, Cat No. E-EL-H0542; E-EL-H0543). In addition, the levels of cystatin F in the plasma of patients with AD or AD mice were quantitatively analysed by ELISA kits (CUSABIO Company, Wuhan, China, Cat No. CSB-E17504h; CSB-EL006095MO). The data were normalized to the total protein concentration.

M-MLV reverse transcriptase was used to reverse transcribe the total RNA from the cells. RT-qPCR was conducted using a SYBR Premix Ex Taq Kit (Vazyme,Nanjing,China, Cat No. Q321-02) and an ABI 7500 RT-PCR instrument. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was selected as the reference gene. The conditions for PCR were 95 °C for 10 s, followed by 40 cycles of 95 °C for 30 s and 60 °C for 30 s. Standard curves were established to calculate target gene transcript quantities in each sample.

The plasma protein concentrations were determined by using a nondenaturing gel. Briefly, the plasma was diluted with 1X PBS, and nondenaturing loading buffer (50% glycerol, 0.3 mol/L pH = 6.8 Tris-HCl, 10% SDS, 0.01% bromophenol blue) was added to the samples. To detect the monomeric form of cystatin F, loading buffer with 20% dithiothreitol (DTT) (Sangon biotech, shanghai, China, Cat No. A620058) was used, and the samples were boiled at 95 °C for 10 min. Coomassie Brilliant Blue (CBB) staining of the plasma proteins served as an internal control. Cells were lysed with radioimmunoprecipitation assay lysis buffer (Beyotime Biotech, Beijing, China, Cat No. P0013B) containing 1 mM PMSF (Beyotime Biotech, Beijing, China, Cat No.ST506). Then, the protein samples were separated on 12% SDS polyacrylamide gels. For western blot detection, the primary antibody against cystatin F (ABclonal Technology, Wuhan, China, Cat No. A8164) was diluted 1:1000, and Horseradish Peroxidase (HRP) goat anti-rabbit IgG (ABclonal Technology, Wuhan, China, Cat No. AS014) was diluted 1:10,000. Immunoreactive bands were visualized with the SuperSignal West Pico chemiluminescent substrate (Pierce, IL, USA) using an LAS-3000 mini (Fujifilm, Tokyo, Japan). For quantitative analysis, the mean density of each band was measured by ImageJ software.

The ORF of cystatin F was subcloned and inserted into the pcDNA3.1 (+) myc-His A vector, and the recombinant plasmids were transfected into 293T cells with polyethylenimine (PEI) (Polysciences, IL, USA, Cat No. 02371-500) at a ratio of 1 µg of DNA to 2.5 µg of PEI [47, 48]. The cell culture supernatant was collected after 48 and 72 h of transfection, and the secreted cystatin F protein was purified by using His-tag Purification Resin (Beyotime, Shanghai, China, Cat No. P2210).

The Aβ internalization by monocytes were assessed by flow cytometry analysis (FCM), ELISA and LSCM methods. For FCM and ELSIA assay, 3 × 10⁵ cells per well were cultured in 24-well plates, and the cystatin F dimer protein was added at final concentrations of 0 ng/mL, 50 ng/mL, 250 ng/mL, and 1000 ng/mL for 30 min. Then a final concentration of 1 μg/mL Aβ42-Alexa Fluor 647 (Kaneka Eurogentec, Liege Province, Belgium, Cat No. AS-64161) or soluble Aβ42 was added for a further incubation for 30 min at 37 °C, after which the cells were subjected to FCM and ELISA analysis, respectively. For LSCM analysis, cells were pretreated with the cystatin F dimer at a final concentration of 250 ng/mL for 30 min and then cells were incubated with Aβ42-Alexa Fluor 555 (Kaneka Eurogentec, Liege Province, Belgium, Cat No. AS-60480-01) for an additional 30 min at 37 °C. To analysis the binding ability of cystatin F dimer to Aβ at the cell membrane by FCM and ELISA methods, cells were pretreated with the cystatin F dimer at a final concentration of 250 ng/mL for 30 min at 37 °C, and then cells were incubated with Aβ42-Alexa Fluor 647 or soluble Aβ42 for an additional 30 min at 0 °C, after which the cells were subjected to FCM and ELISA analysis, respectively. For LSCM analysis, cells were pretreated with the cystatin F dimer at a final concentration of 250 ng/mL for 30 min at 37 °C, and then cells were incubated with Aβ42-Alexa Fluor 555 for an additional 30 min at 0 °C. The coverslips were labelled with 15 μg/mL 3,3′-dioctadecyloxacarbocyanine perchlorate (DiOC18(3), Beyotime, Shanghai, China, Cat No. C1038) in 4% PFA solution for 15 min, and stained with DAPI at a final concentration of 1 μg/mL for 4 min.

For the uptake inhibition assay, cytochalasin D (GLPBIO, CA, USA, Cat No. D22914GC13440) or EIPA (MedChemExpress, NJ, USA, Cat No. HY-101840) at final concentrations of 5 μg/mL or 40 μM were added to the cells to pretreat for 30 min and 1 h, respectively. The cells were incubated with the cystatin F dimer at a final concentration of 250 ng/mL for 30 min. Aβ42-Alexa Fluor 647 at a final concentration of 1 μg/mL was added to the cells, which were incubated for 30 min and analysed via FCM method. The data were analysed with InCyte software (Millipore, Darmstadt, Germany) and visualized using FlowJo software (Tree Star, Inc., CA, USA).

Monocytes were cultured on coverslips, and after treatment with 250 ng/mL cystatin F protein for 30 min, the cells were incubated with 1 μg/mL soluble Aβ42-Alexa Fluor 555 for 30 min at 0 °C and washed three times with PBS. The monocytes on the coverslips were fixed with 4% PFA solution for 15 min and were analysed under a TIRFM.

The 96-well ELISA plates were coated with 4 ng/μL of dissolved Aβ1-40, Aβ1-42, Aβ40-1, or Aβ42-1 for 16 h at 4 °C. Bull serum albumin at a 5% concentration was used as a blocking agent and incubated for 2 h at 37 °C. Recombinant cystatin F dimer protein was added to the plates for 3 h at various concentrations (18.75 ng/mL, 37.5 ng/mL, 75 ng/mL, 150 ng/mL, and 300 ng/mL). Cystatin F antibody (R&D Systems, MN, USA, Cat No. MAB1889) was used to identify the binding of the two proteins. A secondary antibody coupled to HRP and 3,3′,5,5′-Tetramethylbenzidine reagent was used to induce a colour reaction, and the absorbance was read at 450 nm.

Protein Aβ42 (PDB ID: 6SZF) and cystatin F dimer (PDB ID: 2CH9) crystal structures were retrieved from the RCSB Protein Data Bank (http://​www.​rcsb.​org/​). The protein cystatin F was set as the receptor before docking, while Aβ42 was set as the ligand. Cluspro1, HDOCK2, and MOE3 were utilized to determine the binding mechanism of Aβ42 and the cystatin F dimer using three molecular docking tools.

The BiFC was carried out as previously reported with modifications [49]. Briefly, full-length ORF of cystatin F conjugated with the N-terminal fragment Venus and of Aβ conjugated with the C-terminal fragment Venus were constructed. The recombinant plasmids were transfected into 293T cells, and the protein supernatants were purified. The two proteins were incubated at 200 ng/mL for 2 h at 37 °C, and the fluorescence intensity was detected at 529 nm.

Data were presented as mean ± standard deviation (SD). Student’s t test was used to compare two groups while one-way ANOVA was used to compare more than two groups. A receiver operating characteristic (ROC) curve was generated to evaluate the diagnostic value of the cystatin F in AD monocytes. Correlations were assessed by using Pearson correlation coefficient. All data were analysed by GraphPad Prism 8.0 statistical software. Statistical significance was indicated by a probability (p) value of less than or equal to 0.05.

First, we collected samples from sporadic AD patients and isolated monocytes for RNA-seq analysis (sTable 1 and sFig. 1A). We performed a  GO functional enrichment analysis of the differentially expressed genes with four fold changes in expression (Fig. 1A and sFig. 1B and C). Since monocytes are the main peripheral innate immune cells involved in the immune response and clearance of pathogenic agents, we selected the GO terms “immune response” (GO:0006955), “immune system process” (GO:0002376), and “cell activation” (GO:0001775) from the top 10 GO pathways to perform a GSEA analysis of these differentially expressed genes (Fig. 1B). There were 19 core enriched terms in the three pathways (Fig. 1C and D). Among them, only four enrichments were specific to immune cells (sTable 2). Ultimately, we chose the cystatin F protein as a target because it is expressed in myeloid leukaemia cells involved in the innate immune response (Fig. 1D and sTable 2). Cystatin F also has been reported to be expressed in microglia, which are central innate immune cells [50, 51]. Next, we examined the mRNA level of cystatin F in monocytes isolated from AD patients by RT-qPCR with an expanded sample size (n = 40; sTable 1). The results showed that the mRNA expression of cystatin F was obviously greater in the AD group than in the elderly control group (Fig. 1E). Neutrophils and lymphocytes are the main types of immune cells that participate in innate and adaptive immune responses, respectively. We also examined the mRNA expression of cystatin F in the two types of immune cells. The results indicated that the transcript level was not significantly different between the AD group and age-matched controls (Fig. 1E). Furthermore, the specificity and sensitivity of the mRNA expression of cystatin F in AD monocytes were analysed, and ROC analysis indicated that the area under the curve (AUC) reached 0.8513, which was statistically significant (Fig. 1F). These results demonstrated that aberrant expression of cystatin F principally occurs in AD monocytes.

To explore the role of cystatin F in monocytes in the context of AD, we generated transgenic mice on the C57BL/6 background with monocyte-specific expression of human cystatin F under the human CD68 promoter (Hmo-cys F⁺) (Fig. 2A) [44, 45]. Hmo-cys F⁺ mice were hybridized with APP/PS1 mice to obtain APP/PS1/Hmo-cys F⁺ mice via double transgenics to mimic the high cystatin F levels observed in the monocytes of AD patients (Fig. 2B-D). First, we detected Aβ deposition and measured the Aβ concentration in the brain. The Aβ plaque coverage within the brains of APP/PS1/Hmo-cys F⁺ mice were markedly greater than that within the brains of APP/PS1 mice according to immunofluorescence staining (Fig. 2E). Additionally, there was a significant increase in soluble Aβ40 and Aβ42 levels (Fig. 2F, G) and insoluble Aβ40 and Aβ42 levels (Fig. 2H, I) in the brains of APP/PS1/Hmo-cys F⁺ mice. Furthermore, we assessed the spatial learning and memory of the mice through the MWM test. Compared with APP/PS1 mice, APP/PS1/Hmo-cys F⁺ mice showed significant spatial learning and memory deficits, as reflected by longer escape latency time during the training trials and a lower number of platform crossings (Fig. 2J-L, N). However, there was no significant alterations in swimming speed (Fig. 2M). In addition, the levels of soluble Aβ40 and Aβ42 were significantly increased in the plasma of APP/PS1/Hmo-cys F⁺ transgenic mice compared with APP/PS1 mice (Fig. 2O, P). These results suggested that human cystatin F specifically aggravated Aβ deposition in the brain and cognitive impairment in AD model mice.

Mature cystatin F, which contains a signal peptide, is synthesized, and secreted into the extracellular space of cultured cells [50, 51]. Plasma is a major extracellular fluid makes up 55% of the peripheral blood, and the presence of cystatin F in plasma has not yet been investigated. Therefore, we used a nondenaturing gel to analyse the expression of cystatin F in the plasma of patients with AD. As shown in Fig. 3A, the level of cystatin F, which is approximately 50 kDa in molecular weight (MW), in the plasma from patients with AD was significantly greater than that in the plasma from controls. Cystatin F dimerizes through two disulfide bridges involving Cys26 on one subunit and Cys63 on the other [35]. We selected DTT, a reducing agent that can disrupt disulfide bonds to reduce dimers to monomers, to determine the structure of cystatin F in plasma. We found that the MW of cystatin F changed to approximately 25 kDa when DTT was added, suggesting that cystatin F in human plasma is a dimer in the physiological state (Fig. 3B). Furthermore, we examined the level of native cystatin F in cultured monocytes. Using a nondenaturing gel, the results showed that the intracellular cystatin F protein remains predominantly dimeric with few monomers and the extracellular proteins remained absolutely dimeric (sFig. 2). This finding suggested that the level of secreted cystatin F significantly increased in AD plasma as dimeric structure.

Next, we used ELISA to analyse the level of secreted cystatin F in the AD plasma with expanded cases, showing that cystatin F is significantly higher than that in age-matched controls (Fig. 3C), in accordance with the changes in its transcript levels. As the disease progressed, plasma cystatin F levels progressively increased depending on the AD stages (Fig. 3D). Furthermore, we examined the correlation between the cystatin F dimer concentration and the plasma Aβ40 or Aβ42 concentration. The results showed that the cystatin F dimer concentration was significantly positively correlated with the Aβ40 or Aβ42 concentration (Fig. 3E, F). These findings indicated that cystatin F dimer levels were correlated with Aβ levels. We also attempted to analyse the relationship between the cystatin F dimer concentration and the clinical manifestations of AD. The MMSE, MoCA, RAVLT-I, RAVLT-D, and CDR are important tools that provide complete and accurate information about mental health and cognitive status [40, 52]. Correlation analysis revealed that the plasma cystatin F concentration in patients with AD was significantly negatively correlated with the MMSE, MoCA, RAVLT-I, and RAVLT-D scores and positively correlated with the CDR (Fig. 3G-K). These results implied that high levels of cystatin F dimers in the plasma of patients with AD are associated with Aβ levels, the severity of AD, and the clinical features of AD.

Given that monocytes/macrophages are crucial phagocytic cells in the innate immune system, we first investigated whether high levels of the cystatin F dimer impact the internalization of Aβ by monocytes. By the addition of purified cystatin F dimer protein to cultured monocytes to mimic the high concentration of cystatin F in the plasma of AD patients, we found that the uptake of Aβ42-AlexaFluor 647 by the monocytic cell line THP-1 gradually decreased with increasing concentrations of the cystatin F dimer via FCM analysis (Fig. 4A). The ELISA data showed that the intracellular Aβ42 levels decreased in a dose-dependent manner as the cystatin F dimer concentration increased (Fig. 4B). Additionally, we observed less Aβ42-Alexa Fluor 555 in THP-1 cells pretreated with 250 ng/mL cystatin F dimer protein by LSCM (Fig. 4C). Notably, when the substrates were changed to fluorescence-labelled transferrin or dextran, monocytes exhibited similar levels of phagocytosis in the presence and absence of cystatin F (sFig. 3A and B), indicating that cystatin F inhibits Aβ phagocytosis by monocytes with a certain specificity. The process of internalization by phagocytic cells is known to be actin dependent [53]. We further selected cytochalasin D and EIPA, which are specific inhibitors of phagocytosis and macropinocytosis, respectively, to investigate the effect of the cystatin F dimer on intracellular phagocytic signalling [54, 55]. As shown in Fig. 4D, the two inhibitors effectively hindered the internalization of Aβ by monocytes; moreover, the suppressive effect was apparently reversed specifically by the cystatin F dimer protein, whereas cystatin F failed to reverse the inhibition of Aβ internalization as determined by the amount of fluorescently labelled dextran (sFig. 3C).

The recruitment of exogenous substances by monocytes is a prerequisite for phagocytosis. LSCM showed a significant reduction in Aβ42-Alexa Fluor 555 adhesion to the monocyte membrane when the cells were incubated in advance with the cystatin F dimer at 250 ng/mL (Fig. 4E). In addition, we used total TIRFM to observe the enrichment of Aβ42-Alexa Fluor 555 on the cell membrane of monocytes. The images showed that less Aβ42-Alexa Fluor 555 adhered to the surface of monocytes which were pretreated with purified cystatin F dimer protein (Fig. 4F, G). Moreover, co-incubating the substrate with cells at a lower temperature effectively prevented its internalization, showing that the cystatin F dimer was able to markedly inhibit the adhesion of Aβ to monocytes at 0 °C, as determined by FCM and ELISA analysis (Fig. 4H, I). These results demonstrated that the cystatin F dimer blocks monocyte phagocytosis by hindering the binding of Aβ to the monocyte membrane.

High concentrations of the cystatin F dimer present with Aβ in the circulation and inhibit the internalization of Aβ by monocytes; therefore, we first investigated whether there is a potential relationship between them. Cystatin F shares a high homology with cystatin C, which belongs to the same family and can interact with Aβ [56]. We explored the potential associations, assuming that dimeric cystatin F binds to Aβ. A binding assay that can estimate the binding capacity of the two proteins showed that the cystatin F dimer protein directly binds to Aβ40 and Aβ42 in a dose-dependent manner (Fig. 5A, B). Both GST and His pull-down assays further demonstrated a strong interaction between the cystatin F dimer and Aβ (Fig. 5C, D). Furthermore, we used BiFC to investigate how the cystatin F dimer interacts with Aβ in vitro. After tagging the two proteins at their N-terminal and C-terminal fragments of fluorescent Venus protein (Fig. 5E), the cystatin F dimer-VN protein and Aβ-VC protein were expressed in 293T cells and purified with a nickel agarose column. When the purified cystatin F dimer-VN protein was incubated with the Aβ-VC protein, the fluorescence intensity was measured at 529 nm, and the results indicated that the cystatin F dimer is capable of binding to Aβ (Fig. 5F). We further investigated whether the interaction between the cystatin F dimer and Aβ led to the inhibition of the recruitment and internalization of Aβ. We predicted the amino acids that interact with the cystatin F dimer and Aβ42, and three potential binding mechanisms were identified by the Cluspro1, HDOCK2, and MOE3 methods (Fig. 5G and sTable 3). The predicted amino acids on the cystatin F dimer that are potentially capable of interacting with Aβ were mutated to alanine, namely, cystatin F (KSRKKKWWRK), cystatin F (RTSTK) and cystatin F (YNR). GST pull-down assays revealed that cystatin F (RTSTK) had a weaker ability to bind to Aβ42 than did cystatin F (KSRKKKWWRK) and cystatin F (YNR) (Fig. 5H–J), indicating that the Thr41, Ser73, Arg77, Lys119, and Thr121 residues in cystatin F form hydrogen bonds and that salt bridges with the Asp7, Tyr10, His14, Gly33 and Leu34 residues in Aβ42 are involved in these interactions (sFig. 3). Moreover, compared with the cystatin F dimer protein, the cystatin F (RTSTK) protein or supernatant of 293T cells transfected with the cystatin F (RTSTK) plasmid restored Aβ uptake according to FCM and ELISA (Fig. 5K, L and sFig. 4A). In addition, we observed that monocytes pretreated with the cystatin F (RTSTK) mutant protein had an enhanced ability to recruit Aβ compared to monocytes pretreated with the wild type (WT) cystatin F dimer protein through TIRFM and LSCM (Fig. 5M and sFig. 4B). In parallel, the FCM analysis results showed that the adhesion of Aβ to monocytes that were pretreated with the cystatin F (RTSTK) protein was significantly improved (Fig. 5N). These results implied that the complex formed by the interaction of the cystatin F dimer and Aβ extracellularly is the major contributor to the failure of monocytes to effectively recruit and internalize Aβ.

Due to the increased expression of the cystatin F dimer in a collection of patients with sporadic AD, we further explored whether cystatin F is associated with mutations in AD-related genes. We detected the expression of cystatin F in 5XFAD transgenic mice, which express human APP and PSEN1 transgenes with a total of five AD-linked mutations, namely, the Swedish (K670N/M671L), Florida (I716V), and London (V717I) mutations in APP and the M146L and L286V mutations in PSEN1 [57]. As shown in sFig. 5A, the expression of the cystatin F protein in the plasma of 5XFAD transgenic mice did not significantly increase compared with that in the plasma of control mice. Then, the 5XFAD transgenic mice, which exhibited many AD-related phenotypes at a relatively early stage, received a high dose of purified mouse cystatin F dimer via the tail vein to disperse into the peripheral blood (Fig. 6A–C). We first assessed Aβ levels from the brain to the periphery. In the brain, Aβ plaque coverage within the cortex and hippocampus significantly increased (Fig. 6D), and the levels of Aβ40 and Aβ42 markedly increased in 5XFAD-cystatin F dimer mice (Fig. 6E, F). The ELISA results showed that the levels of soluble Aβ40 and Aβ42 was increasing in the plasma of 5XFAD-cystatin F dimer mice compared with the control mice (Fig. 6G, H). Next, we investigated whether blood-borne cystatin F dimers affect central Aβ clearance in the brain. Through the detection of His tag-fused recombinant cystatin F proteins in brain sections, we found that circulating high-level cystatin F failed to diffuse into the brain (sFig. 5B) and that microglia were not affected by increased peripheral levels of cystatin F in 5XFAD mice, either morphologically or in terms of numbers (sFig. 5C). Additionally, we used a human brain microvascular endothelial cell (HBMEC) monolayer cultured on a Transwell insert by adding Aβ to the lower chamber to mimic the AD blood brain barrier model in vitro [58]. We found that monocyte transendothelial migration was not altered by elevated levels of the cystatin F dimer (sFig. 5D and E). These results suggested that high levels of the cystatin F dimer influence Aβ metabolism mainly in the periphery.

Furthermore, various behavioural tests were carried out to assess the effect of the cystatin F dimer protein on learning and memory in 5XFAD mice. The MWM results showed that 5XFAD-cystatin F dimer mice exhibited longer escape latencies during the training trials and fewer platform crossings but no significant alterations in swimming speed (Fig. 6I–M). In the NOR test, 5XFAD-cystatin F dimer mice spent less time exploring new objects, as evidenced by a significant decrease in the recognition index (Fig. 6N, O). Compared with those of the controls, the time, ambulation, and number of visits in the novel arms were obviously decreased in the 5XFAD-cystatin F mice during the Y maze tests (sFig. 5F–I). These results indicated that a high plasma cystatin F dimer level could rapidly exacerbate Aβ deposition and a memory deficit in 5XFAD mice as a risk factor.