Irisin reduces Amyloid-β by inducing the release of neprilysin from astrocytes following downregulation of ERK-STAT3 signaling

Summary

A pathological hallmark of Alzheimer’s disease (AD) is the deposition of amyloid-β protein (Aβ) in the brain. Physical exercise has been shown to reduce Aβ burden in various AD mouse models, but the underlying mechanisms have not been elucidated. Irisin, an exercise-induced hormone, is the secreted form of fibronectin-domain III containing 5 (FNDC5). Here, using a three-dimensional (3D) cell culture model of AD, we show that irisin significantly reduces Aβ pathology by increasing astrocytic release of the Aβ-degrading enzyme neprilysin (NEP). This is mediated by downregulation of ERK-STAT3 signaling. Finally, we show that integrin αV/β5 acts as the irisin receptor on astrocytes required for irisin-induced release of astrocytic NEP, leading to clearance of Aβ. Our findings reveal for the first time a cellular and molecular mechanism by which exercise-induced irisin attenuates Aβ pathology, suggesting a new target pathway for therapies aimed at the prevention and treatment of AD.

In brief (eTOCBlurb):

Kim et al. demonstrate that exercise-induced myokine irisin reduces Aβ pathology using three-dimensional (3D) cell culture model of Alzheimer’s disease (AD) and delineate the key pathways through which irisin exerts neuroprotective effects. These findings offer strong support for the development of irisin as a potential therapeutic target for AD.

Introduction

Alzheimer’s disease (AD) is the most common form of age-related dementia, characterized by progressive memory loss and severe cognitive impairment. The amyloid cascade hypothesis postulates that excessive cerebral accumulation of the amyloid-β (Aβ) protein, which is liberated from the amyloid-β protein precursor (APP) via serial cleavage by β- and γ-secretase is the initiating pathological event leading to AD-related dementia¹. Therefore, an effective therapeutic approach for prevention of AD could entail reducing Aβ burden in the brain early in life.

Physical exercise has been shown to diminish various aspects of AD pathology in animal models, including cerebral Aβ levels and amyloid deposition^2,3 and neuroinflammation^4,5, leading to amelioration of cognitive dysfunction^3,6–8. The mechanism(s) by which exercise leads to reduced Aβ burden remain unclear. Enzymatic activity and/or levels of the Aβ-degrading endopeptidase, neprilysin (NEP) have previously been shown to be elevated in the brains of AD mice exposed to exercise and/or environmental enrichment, leading to reduced amyloid burden^2,9,10. How exercise leads to increased NEP activity/levels remains unknown.

Irisin, a myokine that is cleaved from its precursor protein, fibronectin type III domain containing protein 5 (FNDC5), regulates glucose and lipid metabolism in adipose tissue^11–14 and increases energy expenditure by accelerating the browning of white adipose tissue¹¹. FNDC5/irisin has been shown to be present in human and mouse brains, particularly in the hippocampus^15,16. Recently, it has been reported that irisin levels are reduced in the hippocampus and cerebrospinal fluid (CSF) of AD patients¹⁵ as well as in the brains of AD mouse models^15,17. Furthermore, CSF irisin has been reported to correlate positively with CSF Aβ42 and Mini-Mental State Exam (MMSE) scores of AD patients¹⁸. Additionally, exercise increases circulating irisin in humans¹⁹, Fndc5 gene expression in the hippocampus of wild-type mice¹⁶, and FNDC5 protein expression in the hippocampus of AD transgenic mice^7,16.

We previously developed a three-dimensional (3D) human neural cell culture model of AD which displays robust Aβ generation followed by tau pathology (referred as “3D-AD” cultures)^20–22. Here, we investigated whether irisin affects Aβ pathology in the 3D-AD culture system. We report that irisin significantly reduces Aβ levels by elevating the levels of soluble NEP secreted from astrocytes. We also show that irisin-induced enhancement of NEP activity/levels is mediated by downregulating extracellular signal-regulated kinase (ERK)-signal transducer and activator of transcription 3 (STAT3) signaling pathways, the latter of which is a critical regulator of astrogliosis. Finally, we show that integrin αV/β5 receptors on astrocytes are required to mediate the irisin effects on NEP level and consequent reduction in Aβ levels.

Results

Irisin reduces Aβ pathology in 3D-AD cultures.

ReN cells expressing FAD-associated APP_{Swedish/London} mutations (referred as ReN-GA cells) or expressing both APP_{Swedish/London} and PS1_ΔE9 mutations (referred as ReN-mGAP cells) were differentiated in the 3D culture systems as previously described^20,22 (Figures 1A and 1B). To test the impact of irisin on Aβ pathology, 0.5- or 3.5-week differentiated ReN-GA and Re-NmGAP cultures were treated with recombinant irisin protein for 1.5 weeks at 5 and 500 ng/ml (Figure 1B). We chose 5 ng/ml of irisin, which is within the normal physiological level in human plasma¹⁹, and 500 ng/ml, at which dose has been shown to have maximal effects on blocking osteocyte cell death²³. Conditioned media and 3D gel (cell pellets) samples were collected for further analysis from 2- or 5-week differentiated 3D-AD cultures. The endogenous irisin levels in the conditioned media were not different among 2- and 5-week ReN-GA and ReN-mGAP cultures (Figure S1A). Irisin was reduced in the 3D gels of 5-week ReN-GA cultures compared to 2-week ReN-GA, while no change was observed between 2- and 5-week ReN-mGAP cultures (Figure S1A).

Figure 1. Irisin reduces Aβ, dystrophic neurites, and hyperexcitability.

(A) Diagrams of lentiviral internal ribosome entry sites (IRES) constructs. CMV, cytomegalovirus.

(B) Schematic of irisin treatment schedule.

(C) Aβ levels in the media and in the 3D gels of 5-week 3D-AD cultures treated with irisin or PBS (0, vehicle, veh).

(D) Representative image of dystrophic neurites (arrows) in 5-week ReN-mGAP cultures. Scale bar: 50 μm.

(E) Quantification of dystrophic neurite area and dystrophic neurite number, normalized by total neurite area, in the 5-week 3D-AD cultures treated with irisin or PBS.

(F) Representative images from captured video of 5-week ReN-GA (10X objective; scale bar: 100 μm) and ReN-mGAP cultures (20X objective; scale bar: 50 μm) incubated with Cal-520 AM fluorescent dyes and the corresponding activity maps. Spontaneous Ca²⁺ transients (ΔF/F0) recorded in cells A and B. mins: minutes.

(G) Frequency distribution of Ca²⁺ transients of active cells (>0 transient per 4.5 minutes) in 3D-AD cultures treated with irisin or PBS.

(H) Fraction of hyperactive cells out of total GFP⁺ cells (sum of two videos per well) and average frequency of active cells (average of two videos per well) in 3D-AD cultures treated with irisin or PBS.

For (C), (E), and (H), *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001 between vehicle- and irisin-treated groups, one-way ANOVA with Dunnett’s test; ^#p < 0.05 and ^†p < 0.01 between ReN-GA (veh) vs ReN-mGAP (veh) groups, unpaired t-test. Data are represented as mean ± SEM. See also Figure S1.

Irisin significantly decreased Aβ40 and Aβ42 levels in the conditioned media of 5-week ReN-GA cultures at both 5 and 500 ng/ml doses and in the 3D gel matrix at 500 ng/ml dose (Figure 1C). In the ReN-mGAP cultures, which present more aggressive AD pathology than ReN-GA, 500 ng/ml of irisin reduced both Aβ40 and Aβ42 levels in the conditioned media and only Aβ42 levels in the 3D gels at 5 weeks of differentiation (Figure 1C). At 2 weeks of differentiation, irisin treatment significantly decreased Aβ42 levels in the conditioned media of ReN-GA and Aβ40 levels in the media of ReN-mGAP cultures (Figure S1B).

Irisin did not increase lactate dehydrogenase (LDH) release, an indicator of cell death/cytotoxicity, in 5-week 3D-AD cultures (Figure S1C), suggesting that irisin did not cause cell death at 5 and 500 ng/ml doses and that the Aβ reduction by irisin was not due to cell death. In fact, irisin indeed reduced LDH release at the dose of 500 ng/ml in 5-week 3D-AD cultures, when these cultures showed endogenous cell loss²², suggesting neuroprotective effects. It has been shown that FNDC5 overexpression increases expression of brain-derived neurotrophic factor (Bdnf) gene in the mouse hippocampus¹⁶ and that BDNF reduces Aβ generation by enhancing α-secretase processing of APP²⁴. BDNF levels were mostly below detectable ranges, and irisin did not change BDNF levels in the media of 5-week 3D-AD cultures (Figure S1D), suggesting that Aβ reduction by irisin treatment in 3D-AD cultures was not mediated by BDNF.

Analysis of western blots using G12A (APP C-terminus) or 6E10 antibodies revealed irisin did not significantly change the levels of full-length APP (fl-APP), APP-CTFα, and APP-CTFβ in the 5-week 3D-AD cultures (Figures S2A and S2B), suggesting that irisin does not affect APP processing.

Irisin reduces dystrophic neurites and hyperexcitability.

Dystrophic neurites are pathologically associated with both Aβ and tau pathologies^25–27. Although irisin did not change the levels of amyloid plaques, as detected by 3D6 antibody, potentially due to the low levels of plaques in 5-week 3D-AD cultures (Figure S2C), irisin treatment significantly reduced the area of dystrophic neurites in both 5-week ReN-GA and ReN-mGAP cultures and the number of dystrophic neurites in ReN-mGAP cultures (Figures 1D and 1E).

Aβ has been shown to be associated with neuronal hyperactivity in AD mouse models, in vitro, and in human^28–31 as well as astrocytic hyperactivity in APP/PS1 mice³². Thus, as previously described³³, we analyzed the spontaneous activity of cells in 3D-AD cultures at 5 weeks of differentiation by combining calcium activation and fast time-lapse imaging of all the active cells (>0 transient per 50 frames/4.5 minutes) in captured videos and defined hyperactive cells as cells with >6 transients per 50 frames/4.5 minutes (higher than the average frequency of active cells) (Figure 1F). We observed that the percentiles of hyperactive cells were slightly reduced by irisin in 3D-AD cultures (Figure 1G). The fraction of hyperactive cells among total GFP⁺ cells was significantly higher in 5-week ReN-mGAP cultures compared to ReN-GA cultures (Figure 1H). Importantly, we found that irisin markedly reduced the fraction of hyperactive cells among total GFP⁺ cells, but not the average frequency of active cells in ReN-mGAP cultures, whereas we did not observe any change in ReN-GA cultures by irisin treatment (Figure 1H). These results suggest that irisin successfully reduced Aβ and Aβ-associated dystrophic neurites and hyperexcitability of cells in 3D-AD cultures.

Irisin increases secreted neprilysin (secNEP) levels.

We next determined whether irisin reduces Aβ by enhancing enzymatic degradation or phagocytosis of Aβ. Neprilysin (NEP), insulin degrading enzyme (IDE), and their secreted forms, secNEP and secIDE, are proteases produced by astrocytes that have been shown to degrade Aβ^34–38. We first measured the activities of secNEP and secIDE in the conditioned media of 3D-AD cultures treated with irisin and found that irisin significantly increased secNEP activity, but not secIDE activity (Figures 2A and S2D). To test whether increased secNEP activity was the consequence of increased secNEP expression or enhanced enzymatic activity, we measured the levels of secNEP in the conditioned media of the 3D-AD cultures treated with irisin. The molecular weight of NEP is known to range from 80 to 110 kDa depending on the glycosylation^39–43. In the wild-type mouse brain lysates, NEP was detected at slightly lower than 98 kDa; whereas in the mouse plasma samples, bands were detected at slightly higher than 98 kDa, which could potentially be glycosylated secNEP forms (Figure S2E). In the conditioned media of 3D-AD cultures, secNEP was detected at >98 kDa, the size of which corresponds to NEP detected in the mouse plasma (Figure 2B and S2E). Importantly, irisin significantly increased the secNEP protein levels in the conditioned media of both 5-week ReN-GA and ReN-mGAP cultures, while it did not change secIDE protein levels (Figure 2B). 5 ng/ml of irisin, which did not increase secNEP levels in the media of the ReN-mGAP cultures, did not increase secNEP activity (Figures 2A and 2B). These data indicate that the irisin-induced increase in secNEP activity is the result of upregulation of secNEP protein levels by irisin. In the 3D gels, NEP was detected at <98 kDa (Figure S2F), the size of which is similar with NEP detected in the mouse brain lysates (Figure S2E). Meanwhile, irisin changed neither the protein nor mRNA levels of Nep in the 3D gels of 5-week ReN-GA cultures (Figures 2C and S2F). Sacubitril, a NEP/secNEP inhibitor, completely blocked secNEP activity in the media of 3D-AD cultures (Figure S2G). Co-treatment of sacubitril and irisin significantly increased Aβ42 levels in the 3D-AD cultures compared to irisin only treated groups (Figure 2D), suggesting that increased secNEP activity is required for irisin to reduce Aβ42 levels in both ReN-GA and ReN-mGAP cultures. Co-treatment of sacubitril and irisin still reduced Aβ42 in the ReN-GA cultures compared to vehicle-treated group, indicating that there might be additional mechanisms by which irisin reduces Aβ42 in ReN-GA cultures. Although sacubitril treatment alone significantly reduced secNEP activity in the 3D-AD cultures (Figure S2G), it did not affect Aβ levels (Figure 2D). Low levels of endogenous secNEP in the 3D-AD cultures might not be affected by sacubitril treatment to change Aβ levels.

Figure 2. Irisin increases secNEP levels from astrocytes.

(A-B) secNEP activity levels (A) and western blots of secNEP and secIDE (B) in the media of 5-week 3D-AD cultures treated with irisin or PBS. Graphs represent densitometric quantifications.

(C) Nep mRNA levels, normalized by Gapdh, in 5-week ReN-GA cultures treated with irisin or PBS.

(D) Aβ levels in the 3D gels of 5-week 3D-AD cultures co-treated with irisin (500 ng/ml or PBS) and sacubitril (40 μM or DMSO).

(E) secNEP activity levels in the media of hiPSC-Astro cultures treated with irisin or PBS. For (A-E), *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, ns=not significant, one-way ANOVA with Dunnett’s test (A-C, E) and Fisher’s LSD test (D). Data are represented as mean ± SEM. See also Figure S2.

Our 3D-AD cultures consist of neurons, astrocytes, and oligodendrocytes^7,22. To test whether astrocytes are the source of the irisin-induced increased secNEP, we treated human induced pluripotent stem cell-derived astrocyte (hiPSC-Astro) cultures with irisin. Irisin treatment increased secNEP activity in the conditioned media of hiPSC-Astro cultures (Figure 2E), similar to our observation in 3D-AD cultures. Irisin did not affect Aβ uptake in hiPSC-Astro cultures (data not shown). These results suggest that irisin reduces Aβ pathology in the 3D-AD cultures by increasing secNEP release from astrocytes.

The integrin αV/β5 receptor in astrocytes is required for irisin-induced reductions in Aβ levels.

We previously identified the integrin complex αV/β5 as an irisin receptor in osteocytes and adipose tissues^23,44. In the central nervous system (CNS), the integrin β5 subunit is highly expressed in astrocytes and used as an astrocyte-specific gene suitable for their isolation; microglia also express integrin β5^45,46. The integrin αV subunit is an exclusive partner for integrin β5⁴⁷. We confirmed the expression of integrin αV and β5 subunits in 5-week ReN-mGAP cultures by mass spectrometry and in ReN-GA cultures by western blot (Figure 3A). Ablating astrocytes by treatment with L-α-aminoadipate (L-AAA), an astrocyte-selective gliotoxin⁴⁸, in 5-week ReN-mGAP cultures dramatically reduced protein levels of integrin β5 and S100β (an astrocytic marker) (Figure 3B), indicating that integrin β5 is primarily expressed in astrocytes in the 3D-AD cultures.

Figure 3. Integrin αV/β5 receptor in astrocytes mediates irisin effects on reducing Aβ.

(A) Identified integrin subtypes expressed in 5-week ReN-mGAP cultures by mass spectrometry. Integrin αV and β5 highlighted in yellow. Paired t-test. Western blots of integrin αV and β5 in the 3D gels of 5-week ReN-GA cultures.

(B) Western blots of integrin β5 and S100β in the 3D gels of 5-week ReN-mGAP cultures treated with irisin and/or L-AAA.

(C-D) Western blots of p-FAK, total FAK, p-CREB, and total CREB in the 3D gels of ReN-mGAP cultures and in protein lysates of hiPSC-Astro cultures (D), treated with irisin (500 ng/ml) at indicated time points. min: minutes. Graphs represent densitometric quantifications.

(E) Aβ42 levels in the 3D gels of 5-week ReN-mGAP cultures co-treated with irisin (500 ng/ml) and either an integrin αV/β5 antibody (0.9 μg/ml) or αV/β5 inhibitor (SB273005, 10 μM). Co-treatment of PBS and IgG (0.9 μg/ml) used as control.

(F) secNEP activity in the media of ReN-mGAP cultures co-treated with irisin (or PBS) and SB273005 (or DMSO).

(G) Representative confocal images and western blots of integrin β5 from 5-week ITGB5 KD-ReN-GA cultures treated with doxycycline (Dox, 2 μM). Green, GFP-APP_Swe/Lon; Red, RFP-tagged inducible lentiviral ITGB5 shRNA. Scale bar, 100 μm.

(H) Aβ levels in the media and 3D gels of 5-week ITGB5 KD-ReN-GA cultures treated with irisin (500 ng/ml).

For (D-F) and (H), *p < 0.05, **p < 0.01, ***p < 0.001, ns=not significant, one-way ANOVA with Dunnett’s test (D) and Fisher’s LSD test (E-F), and Unpaired t-test (H). Data are represented as mean ± SEM. See also Figure S3A–C.

Upon ligand binding, heterodimeric integrin receptors trigger canonical signaling by phosphorylation of focal adhesion kinase (FAK) and cyclic AMP response element-binding protein (CREB)^49,50. Irisin treatment induced a rapid activation of integrin receptor as indicated by increased phosphorylation of FAK (p-FAK) and CREB (p-CREB) in the ReN-mGAP and hiPSC-Astro cultures (Figures 3C and 3D), suggesting that integrin signaling is present and activated by irisin in the 3D-AD and hiPSC-Astro cultures.

To determine whether integrin αV/β5 receptors are required for irisin to reduce Aβ levels, integrin β5 activity was inhibited pharmacologically in ReN-mGAP cultures or genetically in ReN-GA cultures. To inhibit integrin αV/β5 activity pharmacologically, 3.5-week ReN-mGAP cultures were treated with an antagonistic integrin αV/β5 antibody or an integrin αV/β5 inhibitor (SB273005) for 1.5 weeks. The treatment of integrin αV/β5 antibody or SB273005 completely abrogated the effects of irisin on reducing Aβ42 levels in the 3D gels of 5-week ReN-mGAP cultures (Figure 3E). SB273005 treatment also abolished the ability of irisin to increase secNEP activity in the media of ReN-mGAP cultures (Figure 3F). To block integrin β5 activity genetically, we generated integrin β5 (ITGB5) knockdown (KD)-ReN-GA lines by overexpressing inducible lentiviral ITGB5 shRNA, for which expression is controlled by doxycycline (Figure 3G). We found that irisin lost its ability to lower Aβ40 and Aβ42 levels both in the media and 3D gels following ITGB5 deletion in 5-week ReN-GA cultures (Figures 3H). Doxycycline treatment alone did not reduce Aβ levels (Figure S3A). These results show that integrin αV/β5 functions as the receptor of irisin in astrocytes and mediates the ability of irisin to reduce Aβ levels in the 3D-AD cultures. Furthermore, we observed that both pharmacological and genetic inhibition of integrin αV/β5 alone, in the absence of irisin treatment, led to decreased levels of Aβ40 and Aβ42 in the 3D-AD cultures (Figures S3B and S3C).

Irisin inhibits STAT3 signaling via integrin αV/β5.

Signal transducer and activator of transcription 3 (STAT3) knockout specifically in astrocytes in APP/PS1 mice have been shown to decrease Aβ burden accompanied by increased NEP protein levels, although the cellular source of NEP was not specified⁵¹. We assessed gene expression levels of Stat3 along with selective astrocyte reactivity markers, including Gfap, Cp, Cd44, C3, Serping1, Amigo2, and Emp1⁵² in 5-week ReN-GA cultures treated with irisin (Figure 4A). One-way ANOVA with Dunnett’s test showed a trend of reduced C3 (p=0.086) and Stat3 (p=0.094) gene expression following 500 ng/ml irisin treatment (Figure 4A). But these reductions attained statistical significance for the vehicle versus irisin 500 ng/ml groups using a t-test (Figure S3D). The protein levels of NF-κB p65, which is upstream of C3 and C3aR^53–55, C3, C3aR, and STAT3 were decreased by irisin treatment in the 3D gels of 5-week ReN-GA cultures (Figure 4B). Co-treatment with an integrin αV/β5 antibody abrogated the ability of irisin to reduce STAT3, C3aR, and NF-κB p65, while treatment of αV/β5 antibody alone did not change their levels (Figures 4B, S3E, and S3F), suggesting that irisin-induced reductions in these proteins are mediated by astrocytes via integrin αV/β5 receptor. The ability of irisin to reduce protein levels of NF-κB p65, C3, C3aR, and STAT3 in the hiPSC-Astro cultures lend further support to these findings (Figure S3G). Next, mass spectrometry and western blot analyses showed that the levels of apolipoprotein E (APOE), which is downstream of STAT3 in astrocytes and exclusively expressed in glial cells^51,56, were reduced by irisin in the 5-week 3D-AD cultures and in the hiPSC-Astro cultures, further demonstrating the effects of irisin in reducing STAT3 (Figures 4C, S3H, and S3I). Irisin-induced attenuation of APOE levels was again abolished by co-treatment of the ReN-mGAP cultures with the integrin αV/β5 antibody or antagonist, SB273005 (Figures 4C and S3F).

Figure 4. Irisin inhibits STAT3 signaling and attenuates reactive astrocyte gene and protein expression via integrin αV/β5.

(A) mRNA levels of astrocyte reactivity genes (normalized by Gapdh) in 5-week ReN-GA cultures treated with irisin or PBS.

(B-C) Western blots of NF-kB p65, C3, C3aR, and STAT3 in the 3D gels of 5-week ReN-GA cultures (B) and APOE in ReN-mGAP cultures (C), co-treated with irisin and either an integrin αV/β5 antibody (0.9 μg/ml) or integrin inhibitor SB273005 (SB, 10 μM). Co-treatment of PBS and IgG (0.9 μg/ml) used as control. Graphs represent densitometric quantification. (D) Western blots of GFAP and S100β in the 3D gels of 5-week ReN-mGAP cultures treated with irisin (500 ng/ml) and densitometric quantification.

(E) Transcriptional profile of neuron and astrocyte clusters in 5-week ReN-mGAP cultures.

(F) Identification of cell types based on the gene expression of cell-specific markers.

(G) Number of gene changes in neurons and astrocytes from 5-week ReN-mGAP cultures treated with irisin (500 ng/ml). Red, upregulated genes; blue, downregulated genes.

(H-I) Volcano plots for differential gene expression (DEG) and gene ontology analysis.

Additional 44 and 46 genes are not shown (I).

(J-K) Violin plots of gene expression levels of Vim, S100b, and Gfap (J) as well as Egfr, Mapk1/Erk, IL6st, and Apoe (K) in astrocytes from 5-week ReN-mGAP cultures co-treated with irisin and integrin αV/β5 antibody (0.9 μg/ml). Co-treatment of PBS and IgG (0.9 μg/ml) used as control.

For (A-D), (J), and (K), *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, one-way ANOVA with Dunnett’s test (A) and Fisher’s LSD test (B-C), unpaired t-test (D), and Wilcoxon rank sum test (J-K). Data are represented as mean ± SEM. See also Figures S3D–K and S4.

STAT3 has been shown to regulate the reactivity of astrocytes^57,58. To test whether irisin-mediated reduction of STAT3 changes reactivity of astrocytes, we measured protein levels of the astrocyte reactivity markers, GFAP and S100β, in 2- and 5-week ReN-mGAP cultures treated with irisin. Western blot analysis showed that irisin treatment significantly reduced GFAP and S100β in the 3D gels of 5-week ReN-mGAP cultures (Figures 4D and S3J). These reductions were observed even at 5 ng/ml of irisin, which failed to reduce Aβ pathology in 2-week ReN-mGAP cultures (Figures S1B and S3K). These data suggest the reduction of GFAP engendered by irisin was not due to secondary effects of Aβ reduction owing to irisin-induced increase in the secretion of NEP.

Irisin attenuates reactive astrocyte gene and protein expression.

To characterize cell-type specific transcriptomic changes induced by irisin treatment, we performed single cell RNA sequencing (scRNAseq) analysis from 5-week 3D-AD cultures, following co-treatment with irisin and an anti-integrin αV/β5 antibody. In ReN-GA cultures, we extracted 2671 astrocytes and 1207 neurons from the full dataset on the basis of marker expression (astrocytes: Gfap, Aqp4, Id4; neurons: Wsb1, Kcnq1ot1, Syt1, Kcnj6) (Figures S4A and S4B). Irisin treatment caused significant changes in gene expression of both neurons and astrocytes (Figure S4C). In astrocytes, gene ontology analyses revealed that irisin upregulated the genes involved in gene expression, RNA export from nucleus, cytoplasmic translation, RNA metabolic process, protein targeting to ER, and regulation of interleukin-1 production (FDR < 0.1) (Figure S4D). In neurons, irisin downregulated the genes involved in stress-induced premature senescence, amyloid fibril formation, positive regulation of ERK1 and ERK2 cascade, negative regulation of cellular protein metabolic processes, and regulation of Aβ formation, while it upregulated the genes involved in cellular protein metabolic process, positive regulation of telomere maintenance via telomere lengthening, generation of neurons, RNA metabolic process, and neuron development (FDR < 0.1) (Figure S4E).

In ReN-mGAP cultures which exhibit more severe neuronal loss compared to ReN-GA cultures²², we extracted 1,821 astrocytes and 695 neurons (Figures 4E and 4F). Irisin treatment caused significant changes in gene expression of both neurons and, more particularly, astrocytes (Figure 4G). In astrocytes, gene ontology analyses showed that irisin downregulated the genes involved in negative regulation of cellular process, regulation of apoptotic process, cellular metal ion homeostasis, import into cell, and negative regulation of cellular metabolic process (FDR < 0.1) (Figure 4H). In neurons, irisin downregulated the genes involved in aggrephagy, positive regulation of signal transduction by p53 class mediator, negative regulation of ubiquitin protein ligase activity, translation, and SRP-dependent cotranslational protein targeting to membrane (FDR < 0.05) (Figure 4I).

Importantly, irisin significantly reduced gene expression levels of the astrocyte reactivity markers, Vimentin (Vim), S100b, Hspb1, Camk2d, Gpx1, Lgals1, Lgals3, Flna, Vcan, and Nes, which was abrogated by inhibiting integrin αV/β5 (Figures 4J and S4F). Additionally, in astrocytes, irisin reduced expression of epidermal growth factor receptor (Egfr) gene, its downstream effector genes including interleukin (IL)-6 pathway (Mapk1/Erk, Il6st), and Apoe gene (consistent with our mass spectrometry and western blot results), which are upregulated in reactive astrocytes^57,59–62, via integrin αV/β5 receptor (Figure 4K).

Our mass spectrometry results further support that irisin suppressed protein expression of EGFR and its downstream pathways that are critical for astrocyte reactivity (consistent with our scRNAseq analysis, EGFR/MEK-ERK/IL-6; consistent with our western blot results, NF-kB p65) in 5-week ReN-mGAP cultures (Figure S4G). Irisin altered the protein expression of reactive astrocytic genes in 5-week ReN-mGAP cultures (Figure S4H). We observed a trend of decreased protein expression of CAMK2D and GPX1 (consistent with our scRNAseq analysis) (Figure S4H). There was a trend toward downregulation of ‘pan’-astrocyte marker genes following irisin treatment, although we did not observe any clear transition between ‘A1’ and ‘A2’ activation states⁶³ (Figure S4H). When these genes were categorized by gene ontology, protein expression of those associated with extracellular matrix (ECM) remodeling (CD44), IL-6 signaling (RRAS2), and GAP43 (neuromodulin) were found to be significantly decreased by irisin in the ReN-mGAP cultures (Figure S4H).

Blocking STAT3 signaling increases secNEP levels.

Next, we tested whether the ability of irisin to increase secNEP levels is mediated by STAT3 downregulation. Thus, we generated ReN-GA cells carrying a doxycycline-regulatable shRNA transgene, which silences STAT3 gene expression (Figure 5A). Knockdown of STAT3 significantly increased secNEP levels and its activity in the conditioned media of the 5-week STAT3 KD-ReN-GA cultures, whereas it reduced secIDE levels (Figures 5B and 5C). STAT3 knockdown also led to significant reduction of Aβ40 and Aβ42 levels in the media and 3D gels of 5-week ReN-GA cultures without irisin treatment (Figure 5D), whereas doxycycline treatment alone did not reduce Aβ levels (Figure S3A). As the deletion of STAT3 was not astrocyte-specific in the 3D-AD cultures, further confirmation for the effects of astrocyte-specific STAT3 deletion is warranted. Overall, these results suggest that reducing STAT3 expression in astrocytes is a key event in the molecular mechanism by which irisin induces increased NEP release leading to Aβ reduction in 3D-AD cultures.

Figure 5. Blocking STAT3 signaling increases secNEP levels.

(A) Representative images and western blots of STAT3 from 5-week STAT3 KD-ReN-GA cultures treated with Dox (2 μM). Green, GFP-APP_Swe/Lon; Red, RFP-tagged inducible lentiviral STAT3 shRNA. Scale bar: 100 μm.

(B-C) secNEP activity (B) and western blots of secNEP and secIDE (C) in the media of 5-week STAT3 KD-ReN-GA cultures with or without Dox (2 μM) treatment. Graphs represent densitometric quantification. Unpaired t-test.

(D) Aβ levels in the media and 3D gels of 5-week STAT3 KD-ReN-GA cultures treated with irisin (500 ng/ml or PBS). One-way ANOVA with Fisher’s LSD test.

For (B-D), *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, ns=not significant. Data are represented as mean ± SEM.

Irisin reduces STAT3 levels by inhibiting IL-6/ERK signaling leading to increased release of secNEP.

The mitogen-activated protein kinase (MAPK)/extracellular signal-regulated kinase (ERK) signaling pathway is chronically activated in human reactive astrocytes and AD brains^64,65, and IL-6 is an upstream activator of both ERK and STAT3 signaling pathways⁶⁶. FNDC5/irisin regulates MAPK/ERK signaling pathways⁶⁷. Since our scRNAseq and mass spectrometry results indicated that irisin suppressed the IL-6/MEK-ERK/STAT3 pathway (Figures 4K, S4E, and S4G) and attenuated astrocyte reactivity (Figures 4 and S4), we tested whether irisin changes the levels of pro-inflammatory cytokines, including IL-6, phosphorylated ERK1/2 (p-ERK at Thr202/Tyr204), and total ERK in the 5-week 3D-AD cultures.

We measured interferon-γ (IFN-γ), IL-1β, IL-2, IL-4, IL-6, IL-8, IL-10, IL-12p70, IL-13, and tumor necrosis factor-α (TNF-α) in the conditioned media of 5-week 3D-AD cultures, and found that irisin at 5 and 500 ng/ml dramatically reduced IL-6 levels (Figure 6A). Additionally, 500 ng/ml of irisin increased IL-2 levels (Figure S5A). Western blot analyses revealed that irisin treatment for 24 hours significantly reduced the levels of p-ERK (normalized to total ERK) in ReN-GA and ReN-mGAP cultures (Figures 6B and S5B).

Figure 6. Irisin inhibits IL-6/ERK signaling to reduce STAT3 signaling for increasing secNEP levels.

(A) IL-6 levels in the media of 5-week 3D-AD cultures treated with irisin or PBS. N.D.=not detected.

(B) Western blots of p-ERK and total ERK in the 3D gels of 5-week ReN-mGAP cultures treated with irisin (500 ng/ml) at indicated time points. Graphs represent densitometric quantification. hrs: hours.

(C-D) Western blots of p-ERK, total ERK, p-STAT3, and total STAT3 in the 3D gels (C) as well as secNEP and secIDE in the media and NEP in the 3D gels (D) in 5-week ReN-GA cultures treated with U0126 (25 μM) or DMSO (veh). Graphs represent densitometric quantification.

(E) Aβ levels in the media and 3D gels of 5-week ReN-GA cultures co-treated with sacubitril (40 μM) and U0126 (25 μM).

For (A-E), *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, ns=not significant, one-way ANOVA with Dunnett’s test (A-B) and Fisher’s LSD test (E), and unpaired t-test (C-D). Data are represented as mean ± SEM. See also Figure S5.

Next, we asked whether ERK inhibition by irisin contributes to the increased NEP activity/levels, by reducing STAT3 levels. Thus, we treated 3.5-week ReN-GA cultures with U0126, which is a highly selective inhibitor of MEK1/2, a type of MAPK/ERK kinase, for 1.5 weeks. U0126 (25 μM) significantly decreased protein expression levels of both total STAT3 and its phosphorylated form (p-STAT3 at Tyr705) in 5-week ReN-GA cultures (Figure 6C), demonstrating that ERK activity regulates STAT3 protein expression. Importantly, U0126 treatment significantly increased secNEP activity and secNEP protein levels in the conditioned media of ReN-GA cultures, leading to reduced levels of Aβ40 and Aβ42 in their media and 3D gels (Figures 6D, 6E, and S5C). Blocking NEP activity with co-treatment of sacubitril abolished the effects of U0126 on reducing Aβ42 levels in both the media and 3D gels (Figure 6E). These findings indicate that increased secNEP activity is required for U0126 to reduce Aβ levels. Similar to irisin, U0126 treatment did not change mRNA levels of Nep in the 3D gels of ReN-GA cultures (Figure S5D). We further found that U0126 treatment increased secNEP protein levels in the media while it decreased NEP levels in cell lysates of hiPSC-Astro cultures, suggesting that ERK inhibition might facilitate NEP secretion from astrocytes (Figures S5E and S5F).

Of note, IL-6 is one of the most highly induced NF-kB-dependent cytokines⁶⁸. Additionally, we found that ERK suppression led to downregulation of NF-kB p65 protein expression in the 3D gels and IL-6 levels in the media of 5-week 3D-AD cultures (Figures S5G and S5H). These results indicate that irisin inhibits the ERK-STAT3/NF-kB pathway to increase NEP activity/levels, leading to reduced Aβ pathology in the 3D-AD cultures (Figure 7).

Figure 7. Schematic diagram of the mechanism of action of irisin.

Depiction of IL-6/ERK pathway leading to STAT3 activation. Irisin inhibits IL-6/ERK-STAT3 pathway, promoting secNEP-mediated Aβ clearance via integrin αV/β5 receptor. SRE, STAT response element; RGD, Arg-Gly-Asp; ECM, extracellular matrix proteins.

Treatment of A1 inducers reduces NEP levels in the cell lysates of hiPSC-Astro cultures.

NEP has previously been shown to be strongly expressed in GFAP⁺ reactive astrocytes surrounding plaques^69–71. However, our findings show that NEP secretion is enhanced in astrocytes with reduced GFAP expression following irisin treatment (Figures 4–7). It should be also noted that exercise, which increases NEP activity, attenuates the reactivity of astrocytes^{2,5,10,72–75}. To further understand the relationship between astrocytic NEP secretion and astrocyte reactivity, hiPSC-Astro cultures were treated with A1 inducers, consisting of IL-1α (3 ng/ml), TNF-α (30 ng/ml), and C1q (400 ng/ml) recombinant proteins as previously reported⁷⁶. The treatment of IL-1α, TNF-α, and C1q significantly increased the levels of pro-inflammatory cytokines but not C3, which resulted in no change in NEP and IDE protein expression levels in both media and cell lysates (Figures S6A–D). Interestingly, when we co-treated with IL-1β (10 ng/ml) and IFN-γ (10 ng/ml) in addition to IL-1α, TNF-α, and C1q to induce C3⁺ A1 astrocytes (Figure S6E), we observed reduced protein expression of NEP in the cell lysates, but not in the media of hiPSC-Astro cultures (Figures S6E–G). These findings suggest that NEP secretion is not affected despite depletion of intracellular NEP in neurotoxic A1 astrocytes.

Irisin reduces phosphorylated tau levels.

To examine the effects of irisin on tau phosphorylation, we treated 3.5-week ReN-mGAP line, which develops robust tau pathology²², with irisin for 1.5 weeks. Irisin (500 ng/ml) significantly reduced sarkosyl-soluble and -insoluble PHF-1⁺ pTau (Ser396/Ser404) levels in 5-week ReN-mGAP cultures (Figures S7A and S7B), while pTau/total tau ratio was not changed due to the trend of reduced total tau levels by irisin treatment (Figures S7C and S7D). Irisin significantly reduced PHF-1⁺ pTau levels when normalized by β-actin (PHF-1/β-actin) (Figure S7E). These results suggest that irisin has potential to reduce tau pathology.

Discussion

We examined whether and how irisin affects AD-related Aβ pathology. First, we found that irisin treatment led to a remarkable reduction of Aβ pathology. Second, we showed this effect of irisin was attributable to increased NEP activity owing to increased levels of NEP secreted from astrocytes. Third, we found irisin’s ability to reduce Aβ levels via increased NEP activity is mediated by ERK-STAT3 signaling. Fourth, we found integrin αV/β5 functions as the irisin receptor on astrocytes mediating irisin’s ability to increase NEP activity and lower Aβ levels (Figure 7).

Our findings indicate that irisin is a major mediator of exercise-induced increases in NEP activity/levels leading to reduced Aβ burden (Figures 1 and 2). Irisin increases NEP secretion from astrocytes by reducing IL-6/ERK-STAT3 signaling (Figures 5 and 6). Previous studies documented that genetic loss of STAT3 in astrocytes increases NEP levels in AD transgenic mice and that the ERK pathway is involved in NEP secretion from astrocytes, though mixed findings report ERK activation or inactivation causing increased NEP secretion^{37,51,77–79}. We found that ERK inhibition significantly reduced the levels of total and phosphorylated STAT3 in the 3D-AD cultures, accompanied by increased secNEP activity/levels in the media (Figures S5C and 6C–6E). These results suggest that STAT3 signals downstream of ERK and that ERK inactivation results in increased NEP secretion from astrocytes by inhibiting STAT3. In considering the detailed molecular mechanism underlying enhanced NEP secretion, it is important to recognize that production of soluble NEP can be regulated by ectodomain shedding via sheddases, such as a disintegrin and metalloprotease 17 (ADAM17), or by release of extracellular vesicles^80–82. Further studies should examine whether irisin changes ADAM17 activity and/or the exosome secretion process in AD.

Neuroinflammation including reactive gliosis strongly correlates with the severity of AD symptoms and neuronal loss. Both IL-6 upregulation and astroglial ERK activation occur early in AD pathogenesis^83–86. STAT3 is also activated in AD patients and mouse models^51,58. We found that irisin attenuates astrocyte reactivity by targeting the major pathways across all levels from the main extracellular stimuli (IL-6, C3, and S100β) and cellular membrane (C3aR) to the nucleus (STAT3 and NF-kB p65) where it initiates the extensive and coordinated transcription of various target genes including IL-6, C3, Stat3, Apoe, and Gfap⁵⁷ (Figures 4–6). Consistent with our findings, irisin’s anti-inflammatory effects have been observed in cultured astrocytes, diabetic mice, adipose tissues, and in AD mice^17,87–90.

Interestingly, 5 ng/ml of irisin, of which dose did not reduce Aβ levels in the 5-week ReN-mGAP cultures (Figure 1C), reduced the levels of dystrophic neurites, hyperactive cells (Figures 1D–1H), and IL-6 (Figure 6A). It has been shown that astrocytic S100β expression correlates with the degree of dystrophic neurites⁹¹ and that reactive astrogliosis leads to hyperexcitability of the hippocampal network via deficits in neuronal inhibition⁹². Therefore, irisin might possibly reduce reactive astrogliosis, thereby decreasing neurite damage and hyperexcitability from reactive astrocytes. Reactive astrogliosis includes altered levels of integrins and ECM components^63,93. Integrins link the ECM to the intracellular actin cytoskeleton, a process required for cell movement during astrogliosis^63,94. Indeed, irisin reduced several key ECM components controlling glial scar formation, such as CD44, laminin, brevican, N-cadherin, and dystroglycan (Figures S4H and S4I). Future studies should monitor changes in astrocyte adhesion and movement close to plaques in response to irisin treatment⁹⁵.

We identified integrin αV/β5 as an irisin receptor on astrocytes, which aligns with our earlier report of irisin binding to αV/β5 receptors on astrocytes in adult hippocampal derived neural stem cell cultures¹⁷. We showed that blocking integrin αV/β5 pharmacologically or genetically abolished the effect of irisin on reducing Aβ levels (Figure 3). Notably, integrin αV/β5 inhibition itself decreased Aβ levels (Figures S3B and S3C) in 3D-AD cultures. In contrast to irisin’s effects on reducing Aβ levels by reducing STAT3 and increasing secNEP, inhibiting αV/β5 receptor using αV/β5 antagonistic antibody did not reduce STAT3 (Figure S3F). These results suggest that different mechanism(s) might reduce Aβ by blocking integrin αV/β5 receptor from other ligands compared to activating αV/β5 receptor by irisin. Interestingly, it has been shown that, unlike typical large ligands with RGD (Arg-Gly-Asp) motif, such as ECM, irisin is a small polypeptide composed of a single fibronectin type III domain without an RGD motif. Irisin binds to integrin αV/β5 receptor at a site distinct from the well-studied RGD-binding pocket⁹⁶, suggesting potential differences in downstream pathways between irisin- and other RGDmediated integrin signaling.

Further, integrins have been proposed as targets for Aβ-mediated neurotoxicity despite mixed results being reported. For example, significant expression of Itgb5 was observed in astrocytes in AD patients but not in control subjects, and Itgb5 was identified as an Aβ plaque-induced gene⁹⁷. Higher expression of integrin αV/β5 was also observed in microvessels of AD patients⁹⁸. Plasma integrin αV/β5 level was associated with lower amyloid burden in human brain⁹⁹. A positive association between Itgb5 expression and inflammatory response-related genes including NF-kB pathway has been observed in glioblastoma cells¹⁰⁰. Furthermore, upregulated Itgb5 has been reported to be reversed by exercise in human AD brain¹⁰¹. Interestingly, we observed that irisin treatment led to significant reduction in Itgb5 protein expression in the 5-week ReN-mGAP cultures by mass spectrometry (Figure 3A). Blockage of integrin α1 activation was reported to prevent Aβ-induced MAPK activation and subsequent cell death in hippocampal neurons¹⁰². Future studies should determine integrin αV/β5’s roles on AD pathogenesis and the mechanisms by which blocking integrin αV/β5 changed Aβ levels.

It should be also determined whether other integrin αV subtypes, such as integrin αV/β1, which is expressed in astrocytes as well as in neurons, could function as another irisin receptor in AD. Integrin β5 is also expressed in microglia^46,55,103. Thus, we cannot exclude the possibility that irisin also modulates microglial functions. IL-2 increase and APOE decrease have been independently suggested to be protective in the AD brain by ameliorating Aβ plaque burden^104–106. Furthermore, our scRNAseq results indicate that irisin changes genes related to neuronal aggrephagy¹⁰⁷, amyloid fibril formation, and regulation of Aβ formation in the 3D-AD cultures (Figures 4I and S4E). Further studies should investigate the precise potential roles of these changes in the beneficial effects of irisin (Figures 4C, 4K, S3H, S3I, and S5A).

FNDC5/irisin overexpression in the brain has been shown to result in significant decrease in soluble Aβ42 and a trend of decreased insoluble Aβ42 in the hippocampus, but not in the cortex, of APP/PS1M146L mice¹⁵. We recently found that irisin is the active moiety that confers the cognitive benefits of exercise and that elevation of peripheral irisin using AAV8-irisin did not change detergent-soluble Aβ levels in 11-month-old APP/PS1ΔE9 mice¹⁷. Irisin treatment in late stages of AD in the 3D-AD cultures (8-week differentiation) did not change Aβ levels (data not shown). Therefore, irisin would most likely be optimally efficacious during the early stages of AD pathogenesis and may be brain region-specific, which will be investigated in future studies.

In summary, we present the experimental data showing that irisin reduces Aβ pathology by increasing NEP activity/level secreted from astrocytes. Furthermore, we have delineated the involved molecular mechanism from identifying the astrocyte receptor, integrin αV/β5, to demonstrating the inhibition of IL-6/ERK and NF-κB-STAT3 signaling, resulting in increased secretion of NEP and reduction of Aβ levels. Thus, our findings offer strong support for developing irisin as a therapeutic target to reduce Aβ burden for AD treatment and prevention.

STAR METHODS

RESOURCE AVAILABILITY

LEAD CONTACT

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Rudolph E. Tanzi (Tanzi@helix.mgh.harvard.edu).

MATERIALS AVAILABILITY

This study did not generate new unique reagents.

DATA AND CODE AVAILABILITY

• scRNAseq data and mass spectrometry proteomics data have been deposited in NCBI-GEO and iProX, respectively. They are publicly available as of the date of publication, and accession numbers are listed in the key resources table.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the Lead Contact upon request.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-CD10 (NEP)	Santa Cruz Biotechnology	Cat# SC-46656; RRID: AB_626828
Anti-IDE	Abcam	Cat# AB32216; RRID: AB_775686
Anti-FAK	Cell Signaling Technology	Cat# 3285S; RRID: AB_2269034
Anti-p-FAK (Tyr397)	Cell Signaling Technology	Cat# 3283S; RRID: AB_2173659
Anti-CREB (86B10)	Cell Signaling Technology	Cat# 9104S; RRID: AB_490881
Anti-P-CREB (Ser133) (87G3)	Cell Signaling Technology	Cat# 9198S; RRID: AB_2561044
Anti-ERK	Cell Signaling Technology	Cat# 4695S; RRID: AB_390779
Anti-p-ERK (Thr202/Tyr204)	Cell Signaling Technology	Cat# 4370S; RRID: AB_2315112
Anti-β-Actin-Peroxidase	Sigma-Aldrich	Cat# A3854; RRID: AB_262011
Anti-GFAP	DAKO	Cat# Z0334; RRID: AB_10013382
Anti-integrin αV/β5 (Clone P1F6)	EMD Millipore	Cat# MAB1961Z; RRID: AB_94466
Anti-integrin β5 (D24A5)	Cell Signaling Technology	Cat# 3629S; RRID: AB_2249358
Anti-C3	Santa Cruz Biotechnology	Cat# SC28294; RRID: AB_627277
Anti-C3aR	Abcam	Cat# AB126250; RRID: AB_11143014
Anti-STAT3	Cell Signaling Technology	Cat# 30835S; RRID: AB_2798995
Anti-pSTAT3 (Y705)	Cell Signaling Technology	Cat# 9145S; RRID: AB_2491009
Anti-APOE	Abcam	Cat# AB1906; RRID: AB_302668
Anti-NF-kB p65	Abcam	Cat# AB16502; RRID: AB_443394
Anti-S100β	Abcam	Cat# AB52642; RRID: AB_882426
Anti-PHF-1	Dr. Peter Davis118	N/A
Anti-β-Amyloid, 1–16 Antibody (6E10)	Biolegend	Cat# 803015; RRID: AB_2565328
Anti-APP (G12A)	Pierce Custom Antibodies	N/A
Anti-3D6	Eli Lilly	N/A
Anti-total tau	DAKO	Cat# A0024; RRID: AB_10013724
Anti-MAP2	Abcam	Cat# AB5392; RRID: AB_2138153
DyLight™ 405 AffiniPure Donkey Anti-Chicken IgY (IgG)	Jackson immunoresearch	Cat# 703–475-155; RRID: AB_2340373
Alexa Fluor® 647 AffiniPure Goat Anti-Mouse IgG (H+L)	Jackson immunoresearch	Cat# 115–605-003; RRID: AB_2338902
Mouse IgG1 Isotype Control (Clone 11711)	R&D systems	Cat# MAB002; RRID: AB_357344
Anti-mouse IgG, HRP-linked Antibody	Cell Signaling Technology	Cat# 7076; RRID: AB_330924
Anti-rabbit IgG, HRP-linked Antibody	Cell Signaling Technology	Cat# 7074; RRID: AB_2099233
ChromPure Donkey IgG, whole molecule	Jackson immunoresearch	Cat# 017–000-003; RRID: AB_2337256
Chemicals, Peptides, and Recombinant Proteins
Paraformaldehyde	Thermo Fisher Scientific	Cat# AA433689L
OneMinute®Plus WB Stripping Buffer	GM bioscience	Cat# GM6015
SuperSignal™ West Femto Maximum Sensitivity Substrate	Thermo Fisher Scientific	Cat# 34096
RGDS peptide	R&D systems	Cat# 3498/10
SB273005	Selleck	Cat# CS7540
10 His-tag irisin	Lake Pharma	Kim et al. 201823
Lysis buffer 6	R&D systems	Cat# 895561
RIPA buffer	Boston bioproducts	Cat# BP-115
L-α-Aminoadipic acid	Santa Cruz	Cat# SC-202200
β-Amyloid (1–42), HiLyte Fluor 555	Ana Spec	Cat# AS-60480–01
Halt™ Protease/ Phosphatase Inhibitor Cocktail	Thermo Scientific	Cat# 78440
U0126	Tocris Bioscience	Cat# 1144
Corning® Cell Recovery Solution	Corning	Cat# 354253
Corning® Matrigel® Basement Membrane Matrix	Corning	Cat# 354234
N-Lauroylsarcosine sodium salt	Sigma-Aldrich	Cat# L5777
Sacubitril calcium salt	Sigma-Aldrich	Cat# SML1380
Cal-520, AM	AAT Bioquest	Cat# 21130
IL-1α recombinant protein	Sigma-Aldrich	Cat# I3901
TNFα recombinant protein	Cell signaling Technology	Cat# 8902SF
C1q recombinant protein	Mybiosource	Cat# mbs143105
IL-1β recombinant protein	R&D systems	Cat# 201-LB
IFN-γ recombinant protein	R&D systems	Cat# 285-IF
Doxycycline	Stemcell technologies	Cat# 72742
Critical Commercial Assays
V-PLEX Aβ Peptide Panel 1 (4G8) Kit	Meso Scale Discovery	Cat# K15199E
V-PLEX Aβ Peptide Panel 1 (6E10) Kit	Meso Scale Discovery	Cat# K15200E
V-PLEX Proinflammatory Panel 1 Human Kit	Meso Scale Discovery	Cat# K15049D
Neprilysin Activity Assay Kit (Fluorometric)	Biovision	Cat# K487
Insulin Degrading Enzyme Activity Assay Kit	Ana Spec	Cat# AS-72231
Pierce™ BCA Protein Assay Kit	Thermo Fisher Scientific	Cat# 23225
miRNeasy Mini Kit	Qiagen	Cat# 217004
iScript™ gDNA Clear cDNA Synthesis Kit	BioRad	Cat# 1725035
SsoAdvanced™ Universal Probes Supermix	BioRad	Cat# 1725282
Promega CytoTox-ONE™ Homogeneous Membrane Integrity Assay	Promega	Cat# G7891
Irisin ELISA	Phoenix Pharmaceuticals	Cat# 067–29
BDNF ELISA	R&D systems	Cat# SBNT00
Deposited Data
Raw and processed data files for scRNAseq	This paper	NCBI-GEO: GSE240161
Raw and processed data files for mass spectrometry	This paper	iProX: IPX0006887001
Experimental Models: Cell Lines
ReNcell VM human neural precursor (ReN) cells	EMD Millipore	Cat# SCC008
Human iPSC-Derived Astrocytes	Axol Bioscience	Cat# AX0665
Oligonucleotides
Nep (Hs00153510_m1)	Life Technologies	Cat# 4331182
Gfap (Hs00909233_m1)	Life Technologies	Cat# 4331182
Cp (Hs00236810_m1)	Life Technologies	Cat# 4331182
Cd44 (Hs01075864_m1)	Life Technologies	Cat# 4331182
C3 (Hs00163811_m1)	Life Technologies	Cat# 4331182
Serping1 (Hs00163781_m1)	Life Technologies	Cat# 4331182
Amigo2 (Hs05001325_s1)	Life Technologies	Cat# 4331182
Emp1 (Hs00608055_m1)	Life Technologies	Cat# 4331182
Stat3 (Hs00374280_m1)	Life Technologies	Cat# 4331182
Gapdh (Hs02786624_g1)	Life Technologies	Cat# 4331182
Recombinant DNA
Inducible ITGB5 shRNA	Dharmacon	Cat# V3SH7669-229525403
Inducible STAT3 shRNA	Dharmacon	Cat# V3SH7669
Software and Algorithms
ImageJ software	NIH	https://imagej.nih.gov/ij
LICOR Image Studio (Version 5.2.5)	LI-COR	https://www.licor.com
GraphPad Prism 9	GraphPad Software	https://www.graphpad.com; RRID:SCR_002798
NIS-Elements Advanced Research Software	Nikon	https://www.microscope.healthcare.nikon.com
Gen5 Microplate Reader Software	Biotek	https://www.biotek.com
Discovery Workbench 4.0	Meso Scale Discovery	https://www.mesoscale.com

EXPERIMENTAL MODEL AND SUBJECT DETAILS

3D-Familial Alzheimer’s Disease (FAD) hNPCs (ReN-GA and ReN-mGAP)

3D cultures were prepared as described previously with minor modifications^20,22. ReNcell VM human neural progenitor line (hNPCs) was purchased from Millipore (CA, USA) and has been validated for the expression of Nestin and Sox2 and for its self-renewal and multi-lineage differentiation capacities. The company also provided certificates of analysis, including sterility, mycoplasma, and karyotyping. Genetically modified ReN cells used in this study were stably transfected to express either a green fluorescent protein (GFP) or both GFP and mCherry, and they have been routinely monitored for purity and morphologies by optical and fluorescence microscopy. Microscopy and immunostaining were also used to check the morphology of the differentiated neurons and astrocytes from ReN cells. The ReNcell VM hNPCs which express amyloid precursor protein (APP)_{Swedish/London} (ReN-GA) or APP_{Swedish/London} and presenilin-1 (PS1)_ΔE9 (ReN-mGAP) were originated from a single clone to avoid mixed cell heterogeneity²². Cell lines have been routinely monitored for their shape, differentiation capacity, and expression of AD marker proteins (e.g., Aβ). The cell passage number was limited to 15.

Both ReN-GA and ReN-mGAP lines were grown and expanded in BD Matrigel (BD Biosciences)-coated T25 or T75 cell culture flasks (BD Biosciences). The media used for expansion is composed of DMEM/F-12 media (Life Technologies) supplemented with 2 μg/ml heparin (StemCell Technologies), 2% (v/v) B27 neural supplement (Life Technologies), 20 ng/ml EGF (Sigma-Aldrich), 20 ng/ml bFGF (Stemgen), 1% (v/v) penicillin/streptomycin, and 1% amphotericin-B solution (Lonza). EGF and bFGF were excluded in the differentiation media. For 3D cultures, BD Matrigel was added to ice-cold cell suspensions (1:10 dilution ratio). ReN-GA and ReN-mGAP cells were plated into 96 well plates and 3D differentiated according to the treatment schedules. Cells were incubated at 37°C in a humidified atmosphere containing 5% CO₂. Media was changed every 3–4 days.

Human induced pluripotent stem cell (hiPSC)-derived astrocytes

hiPSC-derived astrocytes were purchased from Axol Bioscience (Cambridge, UK) and were cultured according to the manufacturer’s instruction. hiPSC-derived astrocytes have been validated by immunostaining for expressing astrocyte markers, such as GFAP, S100β, AQP4, ALDHL1L, and EAAT1, as well as other relevant markers, such as Vim, Kir4.1 (potassium channel), and CD44, with very low or negative expression of a neuronal marker, microtubule associated protein 2 (MAP2). The cell passage number was limited to 3. Cells were incubated at 37°C in a humidified atmosphere containing 5% CO₂. Media was changed every 3 days.

METHOD DETAILS

Cell culture experiments

10-His-tag recombinant irisin protein was produced in HEK 293 cells via previously established protocols [(Lake Pharma, Hopkinton, MA;²³] The protein was diluted in sterilized phosphate-buffered saline (PBS) for experiments. Irisin was prepared in fresh differentiation media. In 3D-AD culture experiments, cells were subjected to three times of irisin treatments every 3–4 days for approximately 10 days (1.5 week). To test the effects of irisin on integrin signaling by phosphorylation of FAK and CREB, differentiation media was replaced 4 hours before irisin treatment. For treatments of irisin and an integrin inhibitor, SB273005 (Selleck Chemicals) was administered as pre-treatment for 1 hour before co-treatment with irisin. Similarly, cells were pre-treated with integrin αV/β5 antagonistic antibody (Millipore) or monoclonal mouse IgG (R&D systems) as control for 1 hour before every irisin treatment. For treatments of irisin and NEP inhibitor, cells were pre-treated with sacubitril (Sigma-Aldrich) for 1 hour before co-treatment with irisin. L-α-aminoadipate (L-AAA, Santa Cruz) was only co-treated with irisin. In order to inhibit ERK activity, U0126 was treated three times, every 3–4 days, for 10 days total (1.5 week). The conditioned media was collected after every treatment. hiPSC-derived astrocytes were treated with irisin, U0126 (Tocris), or A1 inducers [IL-1α (Sigma-Aldrich), TNF-α (Cell Signaling Technology), C1q (MyBioSource), IL-1β (R&D systems), and IFN-γ (R&D systems)] twice during a period of 7 days.

Generation of inducible STAT3 and ITGB5 shRNA cell lines

ReN-GA cells were transduced in one well of a 6 well plate at low multiplicity of infection (MOI, 0.3 TU/cell) to generate a population of cells with one integration event per cell. After infection of inducible ITGB5 or STAT3 shRNA (Dharmacon), the media was replaced with fresh media on the next day. Cells were grown, sub-cultured in a T25 cell culture flask, and treated with doxycycline (2 μM) for 3 days. Top 10% of red fluorescent protein (RFP) positive cells were selected after cell suspensions were subjected to fluorescence-activated cell sorting (FACS)-assisted single cell sorting using FACSAria Fusion (MGH Pathology, flow, image, and mass cytometry cores). The knockdown of genes in the FACS sorted RFP-positive cells was confirmed by western blot analysis. ITGB5 knockdown (KD)-ReN-GA and STAT3 KD-ReN-GA cultures were cultured in the differentiation media with or without doxycycline (2 μM) until the treatment of irisin was finished at 5 weeks.

Analysis of Aβ levels

3D gels were lysed using lysis buffer 6 (R&D systems) containing Halt^™ protease and phosphatase inhibitor cocktail (Thermo Fisher Scientific). Samples of conditioned media and 3D gel lysates were subjected to analysis of Aβ levels using the Meso Scale Discovery (MSD) 96-well Aβ V-PLEX assay as outlined in the manufacturer’s protocol. Briefly, 25 μl each of sample and calibrator were added to the plate coated with an array of Aβ capture antibodies. The plate was then incubated for 2 hours with vigorous shaking at room temperature, followed by washing with wash buffer (provided in the kit). A volume of 25 μl of the detection antibody solution was added, and plate was incubated for 2 hours with vigorous shaking at room temperature. The plate was washed with wash buffer before adding 150 μl 2X MSD read buffer to it, and immediately read on a Meso QuickPlex SQ 120.

Enzyme-linked immunosorbent assays (ELISAs)

Irisin ELISA (EK-067–29, Phoenix Pharmaceuticals) was performed according to the manufacturer’s instruction. Briefly, samples were incubated with primary antibody and biotinylated peptide for 2 hours, and after four times of washing with assay buffer, samples were incubated with streptavidin-horseradish peroxidase (SA-HRP) solution for an additional hour followed by washes. This was followed by incubation with tetramethylbenzidine (TMB) substrate solution for an hour, and the reaction was terminated with 2N hydrochloric acid (HCl) before reading absorbance at 450 nm using a microplate reader.

Total BDNF Quantikine ELISA (SBNT00, R&D systems) was performed according to the manufacturer’s instruction. Briefly, assay diluent RD1–123 and samples were incubated for 2 hours at room temperature. After three times of washing, Total BDNF conjugate was added and incubated for 1 hour, followed by washes. After incubation of substrate solution for 30 minutes at room temperature, stop solution was added, and the optical density of each well was determined using a microplate reader by subtracting readings at 540 nm from the readings at 450 nm.

3D6 and microtubule associated protein 2 (MAP2) double immunostaining

5-week differentiated ReN-mGAP cells were fixed with 4% paraformaldehyde solution (Thermo Fisher Scientific) at room temperature for 24 hours. Fixed cells were blocked with 4% skim milk in Tris-buffered saline (TBS) with 0.1% (v/v) Tween-20 (TBS-T) supplemented with 4% Donkey IgG for additional 24 hours at 4 °C. After washes with TBS-T for three times, the cells were permeabilized with a buffer containing 0.5% Triton X-100 in TBS-T for 1 hour at room temperature. After brief washing with TBS-T, primary antibodies were incubated in TBS-T buffer containing 4% milk and 1% donkey IgG for 24 hours at 4°C (Mouse 3D6 antibody, 1:200, a gift from Lilly; chicken MAP2 antibody, 1:200, Abcam). Cells were washed five times with TBS-T and incubated with secondary antibodies for 24 hours at 4°C (DyLight^™405-conjugated AffiniPure donkey anti-chicken IgY††(IgG), 1: 200, Jackson Immunoresearch; Alexa Fluor^® 647 AffiniPure goat anti-mouse IgG, 1:200, Jackson Immunoresearch). Three individual images were captured in the same region of each well unbiasedly by Nikon A1 confocal laser scanning microscope (Nikon Instruments Inc.). 3D6-positive Aβ aggregates of each image (% area) were calculated by ImageJ, and the average of three images per well was graphed.

Lactate dehydrogenase (LDH) assay

CytoTox-ONE^™ homogeneous membrane integrity assay (Promega) measures lactate dehydrogenase released into the conditioned media through compromised cell membranes. Equal volumes of CytoTox-ONE reagent and cell culture media samples were added into 96 well assay plates. The plates were incubated at 37°C for 30 minutes, and the fluorescent signal was measured using a microplate reader (Ex/Em=560/590 nm).

Western blot analysis

Protein concentration was calculated using the Pierce BCA protein assay kit (Thermo Fisher Scientific). Protein samples were resolved by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and then transferred onto PVDF membranes. iBlot 2 transfer stacks were used to transfer proteins using the iBlot 2 gel transfer device. Membranes were incubated overnight with primary antibodies at 4°C. Horseradish peroxidase (HRP)-conjugated secondary antibodies were applied for 1 hour at room temperature. Secondary antibodies were detected via the enhanced chemiluminescence system (ECL, Thermo Fisher Scientific). β-actin was used as a loading control, and quantitative analysis was performed with ImageJ and LICOR image studio (Version 5.2.5). OneMinute Plus western blot stripping buffer (GM Bioscience) was used to strip the membranes.

Reverse transcription quantitative real-time polymerase chain reaction (RT-qPCR)

Total RNA was extracted using QIAzol lysis reagent combined with a column-based RNA extraction (Qiagen), according to the manufacturer’s protocol. Equal amounts (0.5 μg) of RNA were reverse transcribed into cDNA using iScript^™ gDNA clear cDNA synthesis kit (Bio-Rad). The cDNA was amplified for Gfap, Cp, Cd44, C3, Serping1, Amigo2, Emp1, Stat3, Mme (Nep) with corresponding primers and TaqMan probes with a dye label (FAM) (TaqMan gene expression assays) together with SsoAdvanced^™ universal probes supermix (Bio-Rad). Gapdh expression was used as an internal control in each PCR reaction to calculate deltaCT.

Aβ uptake assay

The experiments were performed as described previously with minor modifications^108,109. hiPSC-derived astrocytes were incubated with fluorescence (FAM)-labelled Aβ42 (500 nM, AnaSpec) in astrocyte culture medium for 2 hours at 37°C; afterwards, cells were lysed with RIPA buffer (1% CHAPS, 1% deoxycholate, 0.2% SDS, 140 mM NaCl, 10 mM Tris-HCl, pH 7.4) containing protease and phosphatase inhibitor cocktail. The internalized Aβ42 levels were analyzed using the MSD 96-well Aβ V-PLEX assay as described above.

Neprilysin (NEP) activity assay

NEP activity levels in the conditioned media were determined by NEP activity assay kit (fluorometric, BioVision). The kit utilizes the ability of an active NEP to cleave a synthetic substrate (Abz-based peptide provided in the kit) to release a free fluorophore. Conditioned media samples from the 3D-AD cultures were incubated with NEP substrates at 37°C. The released Abz was quantified using a fluorescence microplate reader (fluorescence excitation/emission, Ex/Em=330/430 nm).

Insulin degrading enzyme (IDE) activity assay

IDE activity levels in conditioned media were determined by IDE activity assay kit (fluorometric, SensoLyte). When active IDE cleaves the fluorescence resonance energy transfer (FRET) substrate, it results in an increase of 5-FAM (5-carboxyfluorescein) fluorescence. Conditioned media samples from the 3D-AD cultures were incubated with IDE substrates for 1 hour at 37°C. The fluorescence intensity of the samples was measured using a fluorescence microplate reader (Ex/Em=490/520 nm).

Cytokine measurement

Conditioned media was used to measure 10 cytokines: interferon-γ (IFN-γ), IL-1β, IL-2, IL-4, IL-6, IL-8, IL-10, IL-12p70, IL-13, and tumor necrosis factor-α (TNF-α). The measurement of cytokines was performed using the MSD 96-well Human Pro-Inflammatory V-PLEX assay as outlined in the manufacturer’s protocol. Briefly, 25 μl each of sample and calibrator were added to the plate coated with an array of cytokine capture antibodies. The plate was then incubated for 2 hours with vigorous shaking at room temperature, followed by washing with wash buffer (provided in the kit). A volume of 25 μl of the detection antibody solution was added, and plate was incubated for 2 hours with vigorous shaking at room temperature. The plate was washed with wash buffer before adding 150 μl 2X MSD read buffer to it, and immediately read on a Meso QuickPlex SQ 120.

Sarkosyl fractionation

Sarkosyl-insoluble tau proteins were prepared from 3D gels of 5-week differentiated ReN-mGAP cells. 3D gels were lysed with 1% sarkosyl buffer added with protease/phosphatase inhibitors. After centrifugation at 85,000 rpm for 1 hour, sarkosyl-insoluble pellets were added with 2X sodium dodecyl sulfate (SDS) sample loading buffer. Both sarkosyl-soluble and -insoluble samples were subjected to western blot using PHF1 anti-phospho tau antibody (1:1,000, courtesy of Peter Davies) and total tau (Dako).

Quantification of dystrophic neurites

5-week differentiated 3D-AD cultures were fixed with 4% paraformaldehyde solution (Thermo Fisher Scientific) at room temperature for 24 hours. For the artificial intelligence (AI) analysis, NIS-Elements AR 5.21.03 64-bit software was used. Dystrophic neurites (the swollen bulbous processes) were manually identified among a collection of images of the ReN-GA cells and a collection of images of the ReN-mGAP cells (20X objective; Nikon A1 confocal laser scanning microscope), respectively, based on a definition of dystrophic neurites in the previous literature¹¹⁰. These images were then used to train the Segment.AI to recognize the dystrophic neurites in each cell line. The resulting algorithms were then used on the images taken of the ReN-GA and ReN-mGAP cultures to identify dystrophic neurites. Three individual images were captured in each well for total 8 wells per group in ReN-GA cultures and for total 10 wells per group in ReN-mGAP cultures. The dystrophic neurite area (average of three images per well) and dystrophic neurite number (sum of three images per well) were normalized by total neurite area (average of three images per well) in the graphs. Images with ‘0’ total neurites were removed for the analyses.

Calcium imaging

3D-AD cultures at 3.5 weeks of differentiation were subjected to three times of irisin treatments every 3–4 days for approximately 10 days (1.5 week). On the last day of the treatments at 5 weeks of differentiation, to assess neural activity, cellular calcium dynamics were monitored using Cal-520 AM (AAT Bioquest), a Ca²⁺ indicator. 3D-AD cultures were incubated for 30 minutes at 37°C in the differentiation media supplemented with 20 μM of Cal-520 AM, washed twice with fresh media, and incubated for 30 minutes at 37°C before imaging. Fluorescence intensity dynamics were measured using time-lapse imaging by Nikon A1 confocal laser scanning microscope (Green; ReN-GA, 10X objective; ReN-mGAP, 20X objective) at the frame time rate (1 frame/5.4 seconds, total 50 frames/4.5 minutes). Two individual areas per well were captured in the same region of each well and used for analyses. Raw intensity values of all the active cells (>0 frequency per 4.5 minutes) in the captured videos were extracted using ImageJ time series analyzer V3, and relative changes in fluorescence intensity to baseline [(F – F0)/F0] were regarded as calcium signals. For each region of interest (soma), a transient was accepted as a signal when its amplitude was greater than 10 times the standard deviation of the noise in the baseline. Hyperactive cells were defined as cells with >6 transients per 50 frames/4.5 minutes (higher than the average frequency of active cells). GFP-positive cells were counted using ImageJ cell counter in the whole region of every video recorded for ReN-mGAP cultures (area of 1.54 mm² per well). For ReN-GA cultures (area of 6.22 mm² per well), GFP-positive cells were counted in the same selected region of each video (area of 0.25 mm² per well), and the fraction of hyperactive cells (the number of hyperactive cells/mm² divided by the number of GFP-positive cells/mm²) was calculated for both ReN-GA and ReN-mGAP cultures.

Mass spectrometry (MS) analysis

Proteomics analysis was performed as follows:

Protein digestion and isobaric labelling

After three times of irisin treatment, on the harvest day, cells were washed with PBS 3–4 times to remove heparin that might interrupt mass spectrometry and replaced with fresh differentiation media without heparin 4 hours before harvest. After two times of PBS wash, cell recovery solution was added into cells. After incubation in the shaker at 4°C for 4 hours, the cell pellet was acquired by centrifugation followed by 4 times washing with PBS.

Cell pellets were lysed with 500 μl SDS lysis buffer (2.0% SDS (w/v), 250 mM NaCl, 5 mM TCEP, EDTA-free protease inhibitor cocktail (Promega), and 100 mM HEPES, pH 8.5) using a 22-gauge syringe. Lysates were reduced at 57°C for 30 minutes, and cysteine residues were alkylated with iodoacetamide (14 mM) in the dark for 45 minutes at room temperature. Proteins were purified by TCA precipitations (20%), and the protein pellets were washed three times with ice-cold methanol. Pellets were resuspended in 100 μl of 200 mM HEPES, pH 8.0, and digested overnight with 2 μg LysC (Wako). Proteins were further digested with 2 μg of trypsin (Promega) for 6 hours at 37°C. 10 μl of tandem mass tag (TMT) reagents (Thermo Fisher Scientific) were added to each solution for 1 hour at room temperature (25°C). After incubating, the reaction was quenched by adding 1 μl of 5% (w/v) hydroxylamine. Labelled peptides were combined and subsequently desalted via StageTips before liquid chromatograph–tandem mass spectrometry (LC–MS/MS) analysis.

Liquid chromatography (LC)–MS/MS analysis

Data were collected using an Orbitrap Fusion Lumos mass spectrometer (Thermo Fisher Scientific) coupled with a Proxeon EASY-nLC 1200 LC pump (Thermo Fisher Scientific). Peptides were separated on a 75 μm inner diameter micro-capillary column packed with 45 cm of Accucore C18 resin (2.6 μm, 100 Å, Thermo Fisher Scientific). Peptides were separated using a 3-hour gradient of 6–25% acetonitrile in 0.125% formic acid with a flow rate of ~400 nl/minute. Each analysis used an MS3-based TMT method as previously described¹¹¹. The data were acquired using a mass range of m/z 400 – 1400, resolution at 120,000, automatic gain control (AGC) target of 1 × 10⁶, a maximum injection time 150 milliseconds, dynamic exclusion of 180 seconds for the peptide measurements in the Orbitrap.

Data dependent MS2 spectra were acquired in the ion trap with a normalized collision energy (NCE) set at 35%, AGC target set to 2.0 × 10⁵ and a maximum injection time of 120 milliseconds. MS3 scans were acquired in the Orbitrap with a higher-energy collisional dissociation (HCD) collision energy set to 55%, AGC target set to 1.5 × 10⁵, maximum injection time of 200 milliseconds, resolution at 50,000, and with maximum synchronous precursor selection (SPS) precursors set to 10.

Mass spectrometry data processing and spectra assignment

In-house developed software was used to convert acquired mass spectrometric data from the .RAW file to the .mzXML format. Erroneous assignments of peptide ion charge state and monoisotopic m/z were also corrected by the internal software. SEQUEST algorithm was used to assign MS2 spectra by searching the data against a protein sequence database, including Human UniProt Database (downloaded March 2019), and known contaminants, such as mouse albumin and human keratins. A forward (target) database component was followed by a decoy component including all listed protein sequences. Searches were performed using a 20 ppm precursor ion tolerance and requiring both peptide termini to be consistent with trypsin specificity. 6-plex TMT labels on lysine residues and peptide N termini (+ 229.163 Da) were set as static modifications and oxidation of methionine residues (+15.99492 Da) as a variable modification. An MS2 spectra assignment false discovery rate (FDR) of less than 1% was implemented by applying the target-decoy database search strategy. Filtering was performed using a linear discrimination analysis method to create one combined filter parameter from the following peptide ion and MS2 spectra properties: Xcorr and ΔCn, peptide ion mass accuracy, and peptide length. Linear discrimination scores were used to assign probabilities to each MS2 spectrum for being assigned correctly, and these probabilities were further used to filter the data set with an MS2 spectra assignment FDR to obtain a protein identification FDR of less than 1%.

Determination of TMT reporter ion intensities

For reporter ion quantification, a 0.003 m/z window centered on the theoretical m/z value of each reporter ion was monitored for ions, and the maximum intensity of the signal to the theoretical m/z value was recorded. Reporter ion intensities were normalized by multiplication with the ion accumulation time for each MS2 or MS3 spectrum and adjusted based on the overlap of isotopic envelopes of all reporter ions. Following extraction of the reporter ion signal, the isotopic impurities of the TMT reagent were corrected using the values specified by the manufacturer’s specification. Total signal-to-noise values for all peptides were summed for each TMT channel, and all values were adjusted to account for variance and a total minimum signal-to-noise value of 200.

Single-cell RNA sequencing (scRNA-seq) via InDrops

Cell dissociation and capture were performed by scRNA-seq in the Harvard Medical School (HMS) scRNA-seq core laboratory using inDrops technology, as follows. First, 5-week differentiated ReN-GA and ReN-mGAP cultures were treated with Accutase for 10 minutes at 37°C. The cells were dissociated and resuspended in 1,000 μL of Dulbecco’s phosphate buffered saline (DPBS) to allow homogeneous resuspension and reduced clumping. Cell capture and library preparation were performed using a modified version of inDrops protocols involving encapsulation of cells into 3 nL droplets with hydrogel beads carrying barcoding reverse transcription primers¹¹². Following the within-droplet reverse transcription step, emulsions were split into batches of approximately 3,000 cells, frozen at −80°C, and subsequently processed as individual RNA-seq libraries. Approximately 3,000 cells were encapsulated for scRNA-seq.

The standard transcriptome RNA-seq libraries were processed as previously reported¹¹³. In brief, the single cell libraries were demultiplexed following the recommended inDrops pipeline (https://github.com/indrops/indrops) in order to generate count matrices for each sample. We used the repeat-masked primary assembly of the human genome GRCh38 (ENSEMBL) as a reference. Reads were filtered according to the protocol to remove those that had low quality or low complexity. After counting and sorting abundant barcodes, histograms were used to identify thresholds that separate cells from empty gel beads. Finally, the reads of each barcode were aligned to the reference genome with Bowtie. Next, demultiplexed count matrices of the four libraries were aggregated into one combined analysis for downstream analysis. After further filtering to remove putative doublets as well as stressed or dying cells (removing cells more than > 20% mitochondrial genes and nfeature RNA less than 200 genes), we performed linear dimensionality reduction with principal component analysis (PCA), which was then used as input for Louvain clustering and non-linear dimensionality reduction with Uniform Manifold Approximation and Projection (UMAP). Cell cycle stage was scored and classified using the strategy as previously described¹¹⁴. Differential expression was tested using hurdle models for sparse single-cell expression data implemented in model-based analysis of single-cell transcriptomics (MAST)¹¹⁵. 0 minimum UMI total for filtering cells, minimum of 3 cells with ≥ 3 counts for filtering genes, 80% as threshold of gene variability for filtering genes, 50 PCA dimensions for building graph, and a k of 5 for the k-nearest neighbors algorithm were used to create the graph. Annotation tracks (clusters) were imported from upstream analysis with the Seurat package (Louvain algorithm at resolution 0.4). Harmony¹¹⁶ was used to correct batch effect in conjunction with Seurat workflow. The final figures were plotted using ggplot2 package (R-CRAN) and edges.

scRNA-seq data visualization and differential gene analysis

5-week differentiated ReN-GA and ReN-mGAP cultures were read into Seurat (v4.1.2) for preprocessing and clustering analysis¹¹⁷. First, cells were log normalized, centered, and scaled (default settings) followed by PCA using all genes in the dataset. PCs 1:30 were used for clustering with a resolution parameter of 0.4. Clusters identified as doublets, dividing, or gene poor (representing damaged cells) were then discarded before further analysis. Cells passing these quality control (QC) parameters were then merged using the MergeSeurat function. Secondary QC cutoffs were then applied to include only cells with less than 20% mitochondrial genes, greater than 200 genes but less than double the median gene count. Data for these cells were log normalized, centered, and scaled, using the ‘negbinom’ general linear model, while regressing out library size differences, percent mitochondrial genes, and sex. A second round of PCA was then performed on the cleaned data, and interrogation of the PCs revealed a set of 50 highly variable genes contained across all samples, regardless of samples. These genes were present across multiple PCs and had a large effect on clustering but, as a gene set, had no discernable biological relevance. Therefore, we treated these genes as technical artifact and removed the set from the variable gene list before repeating subsequent PCA analysis. PCA was then performed using a set of variable genes selected according to expression and dispersion cutoffs (low expression cutoff = 0.01, high expression cutoff = 3, low dispersion cutoff = 1.1). Subsequent UMAP clustering was performed using PCs 1:30 at a resolution of 0.4. UMAP plots were generated using the same PCs used for clustering and variable genes were determined between clusters using the Wilcoxon Rank Sum Test.

QUANTIFICATION AND STATISTICAL ANALYSIS

N indicates the number of samples which corresponds to the number of wells in 96-well plates in most experiments except for RT-PCR assays. For the RT-PCR assays, 10 well samples from the 96-well plates were combined to extract sufficient amounts of RNA, where N indicates the number of samples with each sample comprising 10 well samples.

Figure 1C, ReN-GA (media): N=20 (veh), 19 (5 ng/ml irisin), and 21 wells (500 ng/ml irisin), in 3 sets; ReN-GA (3D gels): N=21 (veh and 500 ng/ml irisin) and 18 wells (5 ng/ml irisin), in 3 sets; ReN-mGAP (media): N=27 (veh), 26 (5 ng/ml irisin), and 28 wells (500 ng/ml irisin), in 4 sets; ReN-mGAP (3D gels): N=21 (veh and 500 ng/ml irisin) and 18 wells (5 ng/ml irisin), in 3 sets. Figure 1E, ReN-GA: N=6 wells (veh) and 8 wells (5 and 500 ng/ml irisin); ReN-mGAP: N=10 wells per group. Figure 1G, ReN-GA: N=91 cells (veh), 112 cells (5 ng/ml irisin), and 94 cells (500 ng/ml irisin); ReN-mGAP: N=267 cells (veh), 136 cells (5 ng/ml irisin), and 124 cells (500 ng/ml irisin). All groups in 4 wells each. Figure 1H, N=4 wells per group. Figure 2A, ReN-GA: N=6 wells per group; ReN-mGAP: N=13 wells per group in 2 sets. Figure 2B, ReN-GA: N=13 (veh and 500 ng/ml irisin) and 12 wells (5 ng/ml irisin), in 2 sets; ReN-mGAP: N=7 (veh and 500 ng/ml irisin) and 6 wells (5 ng/ml irisin). Figure 2C, N=7 samples per group; 10 wells per sample. Figure 2D, N=5 wells per group. Figure 2E, N=4 wells per group. Figure 3D, N=3 wells per group. Figure 3E, N=8 wells per group. Figure 3F, N=12 wells per group (2 sets, N=6 wells per set). Figure 3H, N=16 wells per group (2 sets, N=8 wells per set). Figure 4A, N=4 samples per group; 10 wells per sample. Figure 4B, N=4 wells per group for NF-kB p65 and C3aR; N=5 wells per group for C3 and STAT3. Figure 4C, N=3 wells per group. Figure 4D, N=6 wells per group. Figure 5B, N=10 wells per group (2 sets, N=5 wells per set). Figure 5C, N=5 wells per group. Figure 5D, N=12 wells per group (2 sets, N=6 wells per set). Figure 6A, ReN-GA: N=21 (veh) and 16 wells (5 ng/ml and 500 ng/ml irisin), in 3 sets; ReN-mGAP: N=20 (veh), 15 (5 ng/ml irisin), and 18 wells (500 ng/ml irisin), in 3 sets. Figure 6B, N=4 wells per group. Figure 6C, N=6 wells per group. Figure 6D, N=5 wells per group. Figure 6E, N=5 wells per group.

For all graphs, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Data are represented as mean ± standard error of the mean (SEM). Statistical analysis was performed using Prism V8 software. All statistical analyses were performed using a student’s t-test, one-way ANOVA followed by a post-hoc Dunnett’s or Fisher’s LSD, or two-way ANOVA with Šídák’s multiple-comparisons. Wilcoxon rank sum test was performed for comparing groups in the violin plots from scRNA-seq. An outlier analysis (ROUT [Q = 1%]) was performed to remove outliers. For all graphs, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Highlights.

Irisin promotes neprilysin (NEP) secretion, leading to Aβ decrease in 3D-AD culture

Integrin αV/β5 serves as the irisin receptor on astrocytes to induce NEP secretion

Irisin-induced NEP secretion is mediated by downregulating ERK-STAT3 signaling