key: cord-0291726-wxj44fxt
authors: Freitag, Kiara; Sterczyk, Nele; Obermayer, Benedikt; Schulz, Julia; Houtman, Judith; Fleck, Lara; Braeuning, Caroline; Sansevrino, Roberto; Hoffmann, Christian; Milovanovic, Dragomir; Sigrist, Stephan J.; Conrad, Thomas; Beule, Dieter; Heppner, Frank L.; Jendrach, Marina
title: The autophagy activator Spermidine reduces neuroinflammation and soluble amyloid beta in an Alzheimer’s disease mouse model
date: 2021-10-28
journal: bioRxiv
DOI: 10.1101/2021.10.28.466219
sha: 758f0b0e9832b40449add6348ec1dfab783ad522
doc_id: 291726
cord_uid: wxj44fxt

Deposition of amyloid beta (Aβ) along with glia cell-mediated neuroinflammation are prominent pathogenic hallmarks of Alzheimer’s disease (AD). In recent years, impairment of autophagy has been found to be another important feature, contributing to AD progression and aging. Therefore, we assessed the effect of the autophagy activator Spermidine, a small body-endogenous polyamine often used as dietary supplement and known to promote longevity, on glia cell-mediated neuroinflammation. Spermidine reduced TLR3- and TLR4- mediated inflammatory processes in microglia and astrocytes by decreasing cytotoxicity, inflammasome activity and NF-κB signaling. In line with these anti-inflammatory effects, oral treatment of the amyloid prone AD-like APPPS1 mice with Spermidine reduced neuroinflammation and neurotoxic soluble Aβ. Mechanistically, single nuclei sequencing revealed microglia as one of the main targets of Spermidine treatment, with increased expression of genes implicated in cell motility and phagocytosis. Thus, Spermidine provides a promising therapeutic potential to target glia cells in AD progression.

Neuroinflammation plays an essential role in the development and progression of various neurodegenerative diseases. It belongs to the main hallmarks of Alzheimer´s disease (AD), alongside extracellular plaques containing amyloid-beta (Aβ) peptides and neurofibrillary tangles consisting of hyperphosphorylated microtubule-associated protein (MAP) tau [1] . The link between neuroinflammation and neurodegenerative diseases was further strengthened by the profound effects of maternal immune activation on the development of neurodegenerative diseases [2, 3] . Injection of the viral mimetic PolyI:C, a synthetic analog of double-stranded RNA, into wild type (WT) mice was sufficient to induce an AD-related pathology [4] , further demonstrating a crucial role of inflammatory events in the initiation of a vicious cycle of neuropathological alterations. Microglia, the brain´s intrinsic myeloid cells, and astrocytes are the predominant cytokine-producing cells of the CNS. Both cell types are essential for maintaining brain homeostasis and respond to danger signals by transforming into an activated state, characterized by increased proliferation and cytokine release [1] .

A growing set of data, including those derived from genome-wide association studies of various human diseases by the Wellcome Trust Case Control Consortium [5] , indicates that autophagy, one of the crucial degradation and quality control pathways of the cell, can regulate inflammatory processes. Mice deficient in the autophagic protein ATG16L1 exhibited a specific increase of IL-1β and IL-18 in macrophages and severe colitis, which could be ameliorated by anti-IL-1β and IL-18 antibody administration [6] . Similar results were achieved with mice lacking the key autophagic factor MAP1LC3B after exposure to LPS or caecal ligation and puncture-induced sepsis [7] . Recently, we could show that a reduction of the key autophagic protein Beclin1 (BECN1), which is also decreased in AD patients [8, 9] , resulted in an increased release of IL-1β and IL-18 by microglia [10] .

The multimeric NLRP3 inflammasome complex, responsible for processing Pro-IL-1β and Pro-IL-18 into its mature forms by activated Caspase-1 (CASP1) [11] , can be degraded by autophagy [10, 12] . Lack of the NLRP3-inflammasome axis induced amelioration of neuroinflammation and disease pathology in several neurodegenerative mouse models [13] [14] [15] , thus emphasizing that activation of autophagy presents an intriguing therapeutic target to counteract neuroinflammation.

The small endogenous polyamine and nutritional supplement Spermidine is known to induce autophagy by inhibiting different acetyltransferases [16, 17] and to extend the life span of flies, worms and yeast [17] [18] [19] [20] . In addition, Spermidine supplementation improved clinical scores and neuroinflammation in mice with experimental autoimmune encephalomyelitis (EAE) [20, 21] , protected dopaminergic neurons in a Parkinson's disease rat model [22] , and exhibited neuroprotective effects and anti-inflammatory properties in a murine model of accelerated aging [23] . Consistent with these observations, Spermidine decreased the inflammatory response of macrophages and the microglial cell line BV2 upon LPS stimulation in vitro [24] [25] [26] . Recent data showing that polyamines improved age-impaired cognitive function and tau-mediated memory impairment in mice [27, 28] and impaired COVID-19 virus particle production [29] , highlighting the need to investigate the therapeutic potential of Spermidine and its molecular mechanisms also in chronic inflammation and AD pathology, which has been missing so far.

Here, we can show that Spermidine interfered with key inflammatory signaling pathways in vitro and AD-associated neuroinflammation in vivo. Applying single nuclei sequencing, we identified microglial changes in the expression of genes associated with cell motility and phagocytosis as one target mechanism of Spermidine´s action in the brains of APPPS1 mice, correlating with reduced soluble Aβ levels. We therefore propose considering Spermidine as a therapeutic option for neuroinflammation and AD pathology.

We began by assessing the effects of Spermidine on the acute neuroinflammatory response of glial cells in vitro. First, neonatal microglia were pre-treated with Spermidine and subsequently stimulated with an established protocol for activation of the TLR4 pathway by application of LPS followed by ATP (Fig. 1a) . Indeed, the presence of Spermidine reduced the LPS/ATP-induced release of IL-1β, IL-18, IL-6 and TNF-α into the cell supernatant dosedependently as measured by ELISA, whereas IL-1β reacted most sensitively (Fig. 1b-e) .

Expanding this analysis by using electrochemiluminescence (MesoScale Discovery panel), we detected reduced levels of eight out of ten cytokines after Spermidine treatment ( Supplementary Fig. 1a ).

To assess whether Spermidine exerts microglia-specific anti-inflammatory effects, the TLR4 pathway of neonatal astrocytes was activated according to the scheme depicted in Fig. 1a .

Similar to microglia, astrocytes showed a dose-dependent reduction of IL-6 in the supernatant in response to Spermidine pre-incubation (Fig. 1f) , while IL-1β was released at low levels ( Supplementary Fig. 1b) .

To examine whether the anti-inflammatory effects of Spermidine were limited to TLR4mediated inflammation, we studied its effects on the TLR3 pathway by using the viral dsRNA PolyI:C (Fig. 1g) . Confirming published observations [30, 31] , PolyI:C treatment induced the release of IL-6 and TNF-α by neonatal microglia and IL-6 by neonatal astrocytes, as measured in the cell supernatant by ELISA ( Fig. 1h-j) . Similar to the observed effects on the TLR4 pathway, Spermidine treatment dose-dependently reduced the PolyI:C-induced release of IL-6 and TNF-α in neonatal microglia and neonatal astrocytes ( Fig. 1h-j) , with microglia showing a higher sensitivity to Spermidine. Again, we gained a broader overview of the cytokine specificity of Spermidine by using electrochemiluminescence (MesoScale Discovery panel).

Apart from IL-6 and TNF-α, the release of IL-12, IL-10, and KC/GRO was also significantly induced by PolyI:C treatment, all of which were significantly blocked by Spermidine ( Supplementary Fig. 1c, d) , emphasizing its broad interference spectrum beyond TLR4mediated inflammation.

To exclude that a putative Spermidine-induced cytotoxicity interfered with the cytokine response, lactate dehydrogenase (LDH) release was measured in the cell supernatant for all stimulation schemes, showing no cytotoxic effect of the Spermidine concentrations used here, instead revealing protective effects after LPS/ATP treatment ( Supplementary Fig. 1e-f ). (f) The IL-6 concentration in the cell supernatant of neonatal astrocytes (neoAC) was determined by ELISA; n = 4.

(g-j) Cells were treated with PolyI:C (50 µg/ml) and Spermidine at various concentrations for 6 h as depicted in the scheme (g).

(h) The IL-6 concentration in the cell supernatant of neonatal microglia was determined by ELISA; n = 3-7.

(i) The IL-6 concentration in the cell supernatant of neonatal astrocytes was determined by ELISA; n = 5-12.

(j) The TNF-α concentration in the cell supernatant of neonatal microglia was determined by ELISA; n = 3-7.

Mean ± SEM, one-way ANOVA, Dunnett´s post hoc test (reference= LPS/ATP-treated or PolyI:C-treated cells), * p < 0.05, ** p < 0.01, *** p < 0.001.

Next, we investigated the underlying molecular mechanism behind cytokine reduction after pre-treatment with Spermidine. Transcriptional analysis by RT-qPCR revealed that Spermidine reduced the amount of LPS/ATP-induced Il-1β, Il-6 and Tnf-α mRNA dose-dependently in neonatal microglia (Fig. 2a) and PolyI:C-induced gene expression of II-6 and Tnf-α in both neonatal microglia and astrocytes (Fig. 2b-c) .

In the BV2 cell line, Spermidine has been shown to target the transcription factor NF-κB, responsible for the induction of IL-1β, IL-6 and TNF-α expression (Choi and Park 2012).

Therefore, we assessed NF-κB p65 phosphorylation by Western blot. In neonatal microglia, Spermidine reduced the LPS/ATP-mediated increase in total NF-κB p65 and phosphorylated NF-κB p65 significantly (Fig. 2d) . In neonatal astrocytes, Spermidine reduced PolyI:C-induced phosphorylation of NF-κB p65, whilst not modulating total NF-κB levels (Fig. 2e ).

Taken together, Spermidine decreased TLR4-and TLR3-mediated cytokine release in neonatal microglia and neonatal astrocytes by reducing the induction of the respective mRNA expression via diminished NF-κB activity.

As we have shown previously that BECN1-mediated autophagy controls IL-1β and IL-18 release by the degradation of NLRP3 and inflammasomes [10] , we assessed whether pre-treatment with Spermidine also affects the inflammasome pathway. The inflammasome is a cytosolic oligomeric signaling platform, controlling IL-1β and IL-18 levels post-translationally. Their precursors, Pro-IL-1β and Pro-IL-18, are processed by activated CASP1 at the inflammasome, which is formed after LPS/ATP stimulation and consists of NLRP3 and ASC [32] .

While Nlrp3 and Pro-Casp1 mRNA expression as well as NLRP3 protein expression were not significantly altered upon Spermidine treatment ( Supplementary Fig. 1g,h) , protein expression of Pro-IL-1β was reduced by administering 30 µM Spermidine ( Supplementary Fig. 1i) , correlating with the Il-1β mRNA levels (Fig. 2a) . In addition, protein levels of cleaved CASP1 p20 in the supernatant were decreased, while levels of Pro-CASP1 were strongly increased after treatment with Spermidine, (Supplementary Fig. 1j, k) , implying that Spermidine regulates inflammasome activity in addition to NF-κB-controlled transcription. Neonatal microglia (neoMG) and neonatal astrocytes (neoAC) were either treated with LPS (1 µg/ml) and ATP (2 mM) or PolyI:C (50 µg/ml) and the indicated Spermidine concentrations as depicted in Fig. 1a or Fig. 1g . 1g. Its expression was normalized to Actin and displayed as fold change compared to non-treated control microglia; n = 3.

(c) The gene expression of Il-6 and Tnf-α was assessed by RT-qPCR after treatment of neonatal astrocytes as depicted in Fig.   1g . Its expression was normalized to Actin and displayed as fold change compared to non-treated control astrocytes; n = 6-7.

(d) Levels of phosphorylated NF-κB (pNF-κB) and NF-κB were determined by Western blot in neonatal microglia treated as depicted in Fig. 1a . Representative images are shown and protein levels are displayed as fold changes compared to nontreated controls normalized to ACTIN; n = 7.

(e) Levels of phosphorylated NF-κB (pNF-κB) and NF-κB were determined by Western blot in neonatal astrocytes treated as depicted in Fig. 1g . Representative images are shown and protein levels are displayed as fold changes compared to nontreated controls normalized to ACTIN; n = 3-4.

Mean ± SEM, one-way ANOVA, Dunnett´s post hoc test (reference= LPS/ATP-treated or PolyI:C-treated cells), * p < 0.05, ** p < 0.01, *** p < 0.001.

To further investigate the effect of Spermidine on inflammasome activity, we treated microglia with Spermidine after priming them with LPS and then added ATP afterwards (Fig.   3a ). In this setting, we investigated the effects of Spermidine on previously activated microglia, thereby mimicking a therapeutic scenario.

Cytokine analysis showed that Spermidine reduced IL-1β and IL-18 release into the cell supernatant dose-dependently (Fig. 3b, c) , while not altering the release of IL-6 and TNF-α ( Fig. 3d, e) . We expanded our analysis by using electrochemiluminescence assays and found that in fact only IL-1β and IL-18 out of all measured cytokines were regulated by Spermidine ( Supplementary Fig. 3a) . Again, Spermidine-mediated cytoprotective effects were shown by reduced LDH release in activated Spermidine-treated microglia ( Supplementary Fig. 3b ).

The transcriptional regulation of Il-1β, Il-6 and Tnf-α by Spermidine revealed no alterations ( Supplementary Fig. 2c ). However, increased protein levels of Pro-IL-1β were found (Fig. 3f) , indicating reduced processing. Also increased levels of uncleaved Pro-CASP1 in microglia treated with 100 µM Spermidine were detected, correlating with a reduction of cleaved and activated CASP1 in the supernatant (Fig. 3g ). In agreement with this observation, cleavage of Gasdermin D (GSDMD), another substrate of CASP1, was also strongly reduced in Spermidinetreated microglia (Fig. 3h ).

Upon observing that the precursors of the IL-1β processing pathway accumulated while the active forms were reduced, we investigated the effects of Spermidine on the inflammasome.

NLRP3 expression was not altered on the mRNA or protein level ( Supplementary Fig. 2d ).

However, staining and quantification of ASC specks/inflammasomes revealed that Spermidine addition reduced the number of ASC specks significantly. A similar reduction was also detected in Casp1 -/microglia (Fig. 3i) , indicating that Spermidine did not directly interfere with Pro-CASP1 cleavage but rather with inflammasome formation. To test this hypothesis, the ASColigomerization inhibitor MCC950 [33] was added before addition of ATP. As no additive effects of MCC950 to the Spermidine-mediated effects with regard to IL-1β release or number of ASC specks could be detected, we deduce that Spermidine is indeed interfering with ASColigomerization and with the inflammasome formation in activated microglia (Fig. 3j) .

Consistent with this hypothesis, Western blot analyses for ASC after chemical crosslinking showed reduced appearance of ASC oligomers in Spermidine-treated cells ( Supplementary   Fig. 2e ), while the amount of ASC monomers was not altered (Supplementary Fig. 2f ). Thus, Spermidine treatment of activated microglia reduced IL-1β processing by interfering with the oligomerization of ASC-positive inflammasomes, elucidating a novel regulatory mechanism of Spermidine in addition to targeting NF-κB-mediated transcription of pro-inflammatory genes.

Taken together, while Spermidine-mediated protective effects were shown in neurons, dendritic cells, macrophages and BV2 cells [25, [34] [35] [36] [37] , our results indicate that Spermidine has the potential to target various glial populations in the brain as well.

Neonatal microglia (neoMG) were treated with LPS (1 µg/ml) and Spermidine at indicated concentrations for 1.45 h and ATP (2 mM) as depicted in the scheme (a).

(b) The IL-1β concentration in the cell supernatant was determined by ELISA; n = 4-8.

(c) The IL-18 concentration in the cell supernatant was determined by ELISA; n = 3.

(d) The IL-6 concentration in the cell supernatant was determined by ELISA; n = 4-8.

(e) The TNF-α concentration in the cell supernatant was determined by ELISA; n = 4-8.

(f) Pro-IL-1β protein levels were determined by Western blot and normalized to ACTIN. Representative images are shown and values are displayed as fold changes compared to LPS/ATP-treated cells; n = 8-9.

(g) Cellular Pro-CASP1 and cleaved CASP1 levels in the supernatant were determined by Western blot (* non-specific band).

Pro-CASP1 was normalized to ACTIN (n = 4-8) and CASP1 was normalized on whole protein content determined by Ponceau S staining (n = 3). Values are displayed as fold changes compared to LPS/ATP-treated cells.

(h) Cellular and cleaved GSDMD (C-terminal fragment) levels in the supernatant were determined by Western blot (* nonspecific band). GSDMD was normalized to ACTIN (n = 4) and the C-terminal fragment was normalized to whole protein content determined by Ponceau S staining (n = 9). Values are displayed as fold changes compared to LPS/ATP-treated cells. 

Several beneficial effects of Spermidine are attributed to Spermidine-mediated induction of autophagy [16, 17] . Thus, we assessed whether the Spermidine-mediated anti-inflammatory effects in glial cells can be linked to autophagy. Spermidine treatment significantly upregulated the key autophagic proteins LC3-II and BECN1 in LPS/ATP-and PolyI:C-treated microglia and PolyI:C-treated astrocytes ( Supplementary Fig. 3a, b) . In agreement with a previous study, LC3-I was only weakly expressed by primary microglia [15] . Since the key transcription factor TFEB, that controls autophagosomal and lysosomal biogenesis [38, 39] , has been found to be regulated by Spermidine in B cells [40] , Tfeb mRNA expression was assessed. Indeed, while LPS/ATP and PolyI:C significantly reduced Tfeb mRNA in neonatal microglia and astrocytes, Spermidine preserved Tfeb levels comparable to non-treated cells ( Supplementary Fig. 3c ).

Next, we determined whether autophagy induction is crucial for the anti-inflammatory action of Spermidine. Here, either full medium or amino acid-free starvation medium HBSS, known to achieve a robust induction of autophagy, was used. As shown previously [10] , cell starvation impaired the release of cytokines upon LPS/ATP or PolyI:C treatment significantly, underlining that autophagy induction reduced the inflammatory response of glial cells. In addition, no further reduction of cytokine release after Spermidine treatment became discernable ( Supplementary Fig. 3d ,e).

Previous data showed that impairment of autophagy reversed several beneficial effects of Spermidine [16] [17] [18] 41] . Therefore, cells were treated according to the scheme in Fig. 3h ). Taken together, these data confirm that Spermidine induces autophagy in glial cells, thus mediating the reduction of cytokine release.

In order to further assess Spermidine´s anti-inflammatory potential, microglia were isolated from adult WT mice and treated with the established stimulation schemes shown in Fig. 1 and Fig. 4c ).

Expanding our analysis to an in vivo-like chronic inflammatory setting, APPPS1 mice harboring transgenes for the human amyloid precursor protein (APP) bearing the Swedish mutation as well as presenilin 1 (PSEN1) were used, which develop a strong Aβ pathology including neuroinflammation [42] . Acute whole hemisphere slice cultures derived from WT mice or 200 day old APPPS1 mice were pre-treated with Spermidine and subsequently stimulated with LPS and ATP (Fig. 4a) . LPS/ATP treatment of APPPS1 slice cultures revealed a massive release of IL-1β and IL-6 compared to slices from WT mice. Spermidine treatment reduced the IL-1β and IL-6 release of slice cultures of both genotypes (Fig. 4a) , suggesting that Spermidine might influence neuroinflammation in APPPS1 mice in vivo.

Thus, APPPS1 mice were treated with 3 mM Spermidine [27] via their drinking water starting prior to disease onset (namely substantial Aβ deposition), at the age of 30 days (Fig. 4b) .

Compared to control APPPS1 mice that received water (H 2 O), Spermidine-supplemented animals showed no differences in fluid uptake per day ( Supplementary Fig. 4d ). To examine the neuroinflammatory status of 120 day and 290 day old male Spermidine-treated APPPS1 mice, ten cytokines were quantified by electrochemiluminescence in brain homogenates containing soluble proteins. Spermidine supplementation significantly reduced the AD- 

To gain insights into the underlying mechanisms behind these changes and the cell populations affected by Spermidine treatment in vivo, we performed comparative single nuclei sequencing (snRNA-seq) using hemispheres of male Spermidine-treated APPPS1 mice, H 2 O APPPS1 controls as well as WT mice in triplicates at 180 days, representing a midpoint in pathology (Fig. 5a) . Using fluorescence-activated cell sorted single nuclei and a 10x Genomics platform ( Supplementary Fig. 5a ), we detected between 6500 and 10000 cells per mouse at a median depth of 1400-1700 genes. Automated clustering revealed 34 clusters which we grouped into 7 major cell types, including neurons, oligodendrocytes, microglia, oligodendrocyte precursors (OPC), astrocytes, macrophages and fibroblasts/ vascular cells, using label transfer from a previously published mouse brain reference dataset In agreement with previous single cell transcriptomic analyses of APPPS1 mice [43, 44] , we detected two microglia subpopulations. The microglia 2 cluster appeared only in APPPS1 but not in WT mice, thus presenting an AD-associated activated microglia phenotype largely equivalent to the classical disease-associated microglia published by Keren-Shaul et al. [43] (Supplementary Fig. 5b-e) . To characterize the main characteristics of both microglia clusters, differential gene expression followed by gene set enrichment analysis between these populations was performed. Compared to cluster 1, the AD-associated cluster 2 revealed a downregulation of genes involved in phagocytosis, endocytosis, cell adhesion and cell polarity, while upregulating neuroinflammatory responses, cell-cycle transition and autophagy ( Supplementary Fig. 5f ).

By focusing on the changes induced by Spermidine treatment of APPPS1 mice, we found the strongest transcriptional changes in microglia. Fewer genes were altered in oligodendrocytes, neurons, and astrocytes, while OPC and macrophages remained almost unaffected (Fig. 5d,   Supplementary Fig. 5g, h) . Interestingly, we observed that Spermidine significantly increased the abundance of microglia cluster 2 (Fig. 5e) . Thus, genes differentially expressed in Spermidine-treated compared to H 2 O APPPS1 mice were specifically assessed in microglia clusters 1 and 2 (Fig. 5f) . We found the following anti-inflammatory-associated genes to be regulated: Pfn1 [45] , Glp2r [46] , Per1 [47] and Sirt3. The upregulation of Sirt3 by Spermidine in the AD-associated microglia cluster 2 was especially striking (Fig. 5g) , as the NAD-dependent deacetlyase SIRT3 is known to exhibit anti-inflammatory effects, targeting several cytokines including the NLRP3 inflammasome in the IL-1β processing pathway [48, 49] . By RT-qPCR analysis, we could confirm that activation of neonatal microglia induced a reduction in Sirt3 levels, which was rescued by Spermidine treatment, thus making SIRT3 a potential driver of the observed Spermidine-mediated effects (Fig. 5g) .

Notably, among the top differentially expressed genes in microglia were genes associated with cell motility, cell migration (Arpc3, Capns1, Pfn1, Plxna2, Aamp, Erbb4, Ywhae, Hpgd) , Capns1, Pfn1, Dctn6, Rin2), autophagy (Arpc3, Capns1, Ets2, Per1, Ghr) , Ets2, Erbb4, Hpgd, Glp2r, Ghr, Jtb, Sra1, Ywhae) , transcription and alternative splicing (Celf2, Eif4a1, Rsrp1, Khdrbs3) ( Fig. 5f-i) . The increased abundance of microglia cluster 2 could be connected to the induction of proliferation-associated genes by Spermidine. Accordingly, gene set enrichment analysis revealed the following Gene Ontology terms to be significantly regulated by Spermidine: glial cell migration, microtubule organization center localization, cell matrix adhesion and the semaphorin plexin signaling pathway. Indeed, Spermidine treatment of neonatal microglia increased expression of the actin nucleation gene Arpc3 (Fig. 5h) and reduced the levels of the anti-proliferatory gene Hpgd ( Supplementary Fig. 5i) , matching the snRNA-seq findings. Notably, Spermidine might exert some of its function not solely by affecting the transcriptome but also the proteome, as indicated by our in vitro findings. scale indicates normalized expression, grey dots represent no data. right panels: Neonatal microglia were treated with the indicated concentrations of Spermidine in combination with LPS (1 µg/ml) and ATP (2 mM) or with PolyI:C (50µg/ml) and the gene expression was assessed by RT-qPCR. Their expression was normalized to Actin and displayed as fold change compared to non-treated control cells; n = 4-6, one-way ANOVA, Dunnett´s post hoc test. Mean ± SEM, * p < 0.05, ** p < 0.01, *** p < 0.01.

(i) As in (g-h) for additional genes.

Based on current literature, microglia in early AD pathology are characterized by acute activation including enhanced proliferative, phagocytic and migratory behavior, which ceases and reverts during the progression of the disease and the chronic activation of microglia [50, 51] . The characterization of the AD-associated cluster 2 indicates that these APPPS1 microglia seem to be in the progressed state during which endocytosis and phagocytosis are downregulated ( Supplementary Fig. 5f ). However, Spermidine appears to rescue the impairment of phagocytosis and endocytosis. We therefore hypothesize that Spermidine prolongs the early activated state of microglia characterized by increased phagocytosis, cell motility, migration and proliferation, thus maintaining the surveillance mode of microglia and potentially reducing Aβ. In line with this hypothesis, Spermidine treatment reduced the levels of the transcriptional regulator Celf2 (Fig. 5f ), which negatively regulates the phagocytic receptor TREM2 [52] and increased the expression of the phagocytic receptors Trem2 and Cd36 in activated neonatal microglia in vitro (Fig. 6a, b) .

To test whether Spermidine directly modifies the phagocytic behavior, neonatal microglia were pre-treated with Spermidine before addition of oligomeric and fibrillary fluorescently labelled Aβ for 24 h. Spermidine treatment significantly decreased the mean Aβ signal per phagocytic cell (Fig. 6c) , while the percentage of phagocytic cells was not altered ( Supplementary Fig. 6a) , indicating that Spermidine enhances the degradation of Aβ. A histogram, showing the intensity signal of each phagocytic cell, revealed that the amount of cells with low Aβ signal was increased upon Spermidine treatment, while a lower percentage with high fluorescent signal was found, supporting the idea of increased Aβ degradation in Spermidine-treated cells (Supplementary Fig. 6b ).

Finally, Aβ pathology was analyzed in Spermidine-treated APPPS1 mice over time at 120 and 290 days, when pathology is known to have reached a plateau [42] (Fig. 6d) . After consecutive protein extractions, soluble and insoluble Aβ40 and Aβ42 levels were measured by electrochemiluminescence (MesoScale Discovery panel). In line with the snRNA-seq findings and the in vitro data, Spermidine supplementation significantly reduced soluble Aβ40 levels in both 120 day old and 290 day old APPPS1 mice by 40 % and soluble Aβ42 in 290 day old mice by 49 % (Fig. 6d) , whilst not affecting insoluble Aβ (Supplementary Fig. 6c) . These findings were further substantiated as no differences in insoluble Aβ plaque covered area or plaque size after staining tissue sections with the fluorescent dye pFTAA were observed ( Supplementary Fig. 6d ). As APP processing was not affected by Spermidine treatment ( Supplementary Fig. 6e ), we conclude that Spermidine treatment targeted soluble Aβ by maintaining microglial phagocytic and migratory behavior. (d) APPPS1 mice were treated with 3 mM Spermidine via their drinking water starting at 30 days (d) until mice reached an age of 120 days or 290 days according to the depicted treatment scheme. Spermidine-treated APPPS1 mice were compared to non-treated controls (H 2 O). The Aβ40 and Aβ42 content was measured in the TBS (soluble) fraction of brain homogenates of (b) 120d old or (c) 290d old Spermidine-treated mice and water controls using electrochemiluminescence (MesoScale Discovery panel). Values were normalized to water controls. 120d APPPS1 H 2 O (n = 14), 120d APPPS1 Spermidine (n = 14), 290d APPPS1 H 2 O (n = 14), 290d APPPS1 Spermidine (n = 12).

Mean ± SEM, two-tailed t-test, * p < 0.05, ** p < 0.01, *** p < 0.01.

Preventing AD or at least delaying disease progression still presents an urgent unmet clinical need. Based on recent advances in our understanding of AD pathogenesis that resulted in the appreciation of the impact of neuroinflammation and autophagy, we assessed the therapeutic effects of the autophagic activator Spermidine on the main cytokine-producing glial cells, namely microglia and astrocytes. By modeling neuroinflammation in vitro through activation of the TLR3 and TLR4 pathway, we found that Spermidine treatment abolished the release of pro-inflammatory cytokines by microglia and astrocytes by targeting various pathways: NF-κB-mediated transcription of cytokines, cytotoxicity and the assembly of the NLRP3 inflammasome. In agreement with previous studies [16, 17] , Spermidine-induced autophagy in glial cells accounts for the observed reduction of neuroinflammation. In line with our in vitro findings, we observed a substantial reduction of inflammation in brain slice cultures as well as in 290 day old male APPPS1 mice treated with Spermidine in vivo. Consistent with these findings, pro-inflammatory cytokine production was decreased and disease pathology was ameliorated in Spermidine-treated EAE mice, a murine model for multiple sclerosis [20, 21] .

Interestingly, single nuclei sequencing of a hemisphere from 180 day old Spermidine-treated APPPS1 mice revealed microglia as the main cytokine-producing cell type targeted by Spermidine. On the contrary, astrocytes were barely affected, despite the in vitro data showing anti-inflammatory responses after Spermidine treatment in both cell types, which could be explained by their decreased sensitivity to Spermidine shown in vitro. Among the significantly regulated candidate genes in microglia was the NAD-dependent deacetylase Sirt3.

Similarly to Spermidine, SIRT3 was shown to exhibit anti-inflammatory effects, specifically targeting the NLRP3 inflammasome and IL-1β processing pathway as well as IL-6 and TNF-α [48, 49] , indicating that Spermidine might exert some of its effects by modulating SIRT3. In addition, SIRT3 was shown to be decreased in AD patients and in microglial BV2 cells upon Aβ treatment [53, 54] and to promote the migratory behavior of microglia [55] , thus targeting all pathways shown to be affected by Spermidine treatment. Of note, clear anti-inflammatory effects of Spermidine became only apparent at 290 days, when APPPS1 mice show neuroinflammation changes. Thus, only a few differentially expressed genes related to inflammation could be found at 180 days, correlating with the lack of anti-inflammatory effects of Spermidine at 120 days.

While changes in Sirt3 and other genes with anti-inflammatory properties (Pfn1, Glp2r, Per1) that were up-regulated in 180 day old Spermidine-treated APPPS1 mice might pave the path for the reduction in neuroinflammation observed at 290 days, our in vitro analyses on the NLRP3 inflammasome pathway as well as recent publications on a post-translational modification called hypusination [27, 28] , indicate that Spermidine might exert some of its function also on the post-translational level. Among those might be autophagic proteins, which are mainly regulated post-translationally.

The most profound effects of Spermidine on the transcriptome were seen in the ADassociated microglia cluster 2, which was characterized by increased migration, cell motility, phagocytosis, autophagy and cell proliferation. While acute activation of microglia in early disease pathology induces microglial phagocytosis and migration towards plaques, later stages of AD pathology and chronic priming of microglia with Aβ have adverse effects [50, 51] .

In accordance, microglial motility in the presence of Aβ plaques was found to be decreased in APPPS1 mice compared to control mice when a focal lesion was performed by a laser [56] . By promoting genes known to be involved in cell motility, migration, phagocytosis and autophagy, Spermidine seems to delay the late stage AD-associated microglial phenotype. In addition, Spermidine also increased the abundance of microglia cluster 2. Although it is still under discussion whether proliferation of microglia in AD is beneficial or detrimental [57] , we conclude that Spermidine mediates an enlargement of a microglial population characterized by increased phagocytosis and cell motility. Several regulated genes as Arpc3 [58] , Glp2r [59] , Sirt3 [53] and Per1 [60, 61] have been shown to exert protective effects in neurodegenerative diseases or reverse memory deficits in various models, underlining the observed protective effects of Spermidine.

In line with the snRNA-seq findings, Spermidine treatment also significantly reduced Aβ in vitro as well as soluble Aβ levels at 120 days and at 290 days, while Aβ plaque burden and size were not altered. These results correlate with recent data by De Risi et al. [62] , who observed that Spermidine decreased soluble Aβ and α-synuclein in a mouse model with mild cognitive impairment. Currently, it is thought that soluble Aβ causes more synaptotoxicity than plaquebound insoluble Aβ by altering synaptic transmission and mediating synaptic loss and neuronal death, thus stressing the importance of targeting soluble Aβ in AD [63] [64] [65] . In 120 d old mice only Aβ40, was reduced, while at the later time point both Aβ species were less present in the soluble fraction. Interestingly, studies have shown that soluble Aβ40 has a substantial pathogenetic effect on neuronal survival underlining its qualitative impact in AD [66] [67] [68] . In addition, very recent data imply that microglia create core plaques as a protective measure in order to shield the brain from soluble Aβ [69] , thus presenting an additional mechanism of Spermidine-mediated mitigation of AD pathologic progression. As the APPPS1 mouse model exhibits a fast disease progression with a strong genetically-driven amyloid pathology appearing at 60 days [42] , the substantial effect of Spermidine on soluble Aβ supports the potential of Spermidine to counteract or at least slow AD progression. Additionally, neuroinflammation is a known driver for plaque formation [2, 3] , thus the anti-inflammatory effects of Spermidine might potentially affect plaque size in APPPS1 mice older than the analyzed 290 day old animals.

In agreement with our findings, several studies suggest an involvement of autophagic mechanisms in regulating AD. Similarly to Spermidine, strong activators of autophagy such as fasting or caloric restriction, Rapamycin, an inhibitor of the mechanistic Target of Rapamycin (mTOR), and Metformin were found to prolong the life span of several species and to reduce Aβ deposition in different mouse models [70] [71] [72] [73] , emphasizing that the herein described effects of Spermidine in APPPS1 mice on Aβ burden are genuine. Spermidine's multifarious intracellular interference points are advantageous, particularly in light of its good tolerability and its uncomplicated oral administration method. Therefore, we consider the bodyendogenous substance Spermidine as a novel, attractive therapeutic dietary supplement in AD as it attenuated AD-relevant neuroinflammation and reduced synaptotoxic soluble Aβ.

Since Spermidine supplementation is already tested in humans, the extension of Spermidine supplementation from individuals with subjective cognitive decline [28, [74] [75] [76] to clinical trials aimed at testing Spermidine efficacy in AD patients appears justified.

Mice and Spermidine treatment APPPS1 +/− mice [42] were used as an Alzheimer's disease-like mouse model. conducted in accordance with animal welfare acts and were approved by the regional office for health and social service in Berlin (LaGeSo).

Mice were anesthetized with isoflurane, euthanized by CO 2 exposure and transcardially perfused with PBS. Brains were removed from the skull and sagitally divided. The left hemisphere was fixed with 4 % paraformaldehyde for 24 h at 4°C and subsequently immersed in 30% sucrose until sectioning for immunohistochemistry was performed. The right hemisphere was snap-frozen in liquid nitrogen and stored at -80°C for a 3-step protein extraction using buffers with increasing stringency as described previously [77] . In brief, the hemisphere was homogenized in Tris-buffered saline (TBS) buffer (20 mM Tris, 137 mM NaCl, pH = 7.6) to extract soluble proteins, in Triton-X buffer (TBS buffer containing 1% Triton X-100) for membrane-bound proteins and in SDS buffer (2% SDS in ddH 2 O) for insoluble proteins. The protein fractions were extracted by ultracentrifugation at 100,000 g for 45 min after initial homogenization with a tissue homogenizer and a 1 ml syringe with G26 cannulas. The respective supernatants were collected and frozen at -80°C for downstream analysis. Protein The brains of C57Bl/6J and APPPS1 mice were harvested, the cerebellum removed and the hemispheres mounted on a cutting disk using a thin layer of superglue. Hemispheres were cut using the Vibratome platform submerged in chilled medium consisting of DMEM medium (Invitrogen, 41966-029) supplemented with 1% penicillin/streptomycin (Sigma, P0781-20ML).

Coronal slicing was performed from anterior to posterior after discarding the first 1 mm of tissue generating 10 x 300 µm sequential slices per brain with vibrating frequency set to 10 and speed to 3. Brain slices were cultured in pairs in 1 ml culture medium at 35 °C, 5 % CO 2 in 6-well plates. Pre-treatment with the indicated Spermidine concentrations was started immediately for 2 h. Subsequently, LPS (10 µg/ml) was added to the medium for 3 h followed by the addition of ATP (5 mM) for an additional 3 h. Afterwards, the culture medium was frozen for subsequent analyses. Cell culture of neonatal microglia and astrocytes Newborn mice (1-4 days old) were sacrificed by decapitation. Mixed glial cultures were prepared as described previously [10] . In brief, brains were dissected, meninges removed and brains mechanically and enzymatically homogenized with 0.005% trypsin/EDTA. Cells were 488-conjugated anti-rabbit IgG, Invitrogen A21206) for 3 h at room temperature. Cell nuclei were counterstained with DAPI (Roche, 10236276001) and coverslips embedded in fluorescent mounting medium (Dako, S3023). Images were acquired using Leica TCS SP5 confocal laser scanning microscope controlled by LAS AF scan software (Leica Microsystems, Wetzlar, Germany). Z-stacks were taken and images presented as the maximum projection of the z-stack. The number and size of ASC specks was assessed using ImageJ software as described before [10] .

Labeling. A 1-42 peptides (Cayman Chemicals) were resuspended in hexafluoroisopropanol to obtain 1 mM solution, evaporated and stored as aliquots. For each preparation, 125 µg of amyloid- was dissolved in 2 µL DMSO, sonicated for 10 min in the waterbath and supplemented with 3x molar excess of NHS-ester ATTO647N dye (Sigma) in 1x PBS (phosphate buffer saline, Gibco) and pH was adjusted to 9 with sodium bicarbonate. After 1 h of labeling reaction in the dark at room temperature, the labeled peptides were separated using spin columns (Mobicol, Mobitec) and loaded with 0.7 mL of Sephadex G25 beads (Cytiva). Clean In short, to obtain oligomeric forms, A was resuspended in the final concentration of 1x PBS and incubated at 4°C overnight. For obtaining A fibrils, the labelled peptides were resuspended in the final concentration of 1 mM HCl and incubated at 37°C overnight.

Neonatal microglia seeded on coverslips were pre-treated for 2 h with Spermidine. 0.5 µM fibrillary and oligomeric 647-labelled Aβ was added and after 24 h cells were fixed and counterstained with Iba-1. Quantification of Z-stacks taken at the confocal microscope was performed with Image J: a mask was created for each Iba-1 stained cell body and the intensity of the Aβ signal in every cell was determined. The mean intensity/phagocytic cell was calculated as well as the number of Aβ-containing/phagocytic cells.

Mouse hemispheres were harvested from male mice at the age of 180 days and immediately snap frozen in liquid nitrogen and stored at -80°C until further processing for nuclei isolation.

Nuclei were isolated from a single mouse hemisphere in 2 ml of pre-chilled EZ PREP lysis buffer (NUC-101, Sigma) using a glass Dounce tissue grinder (D8938, Sigma) (25 strokes with pastel A and 25 strokes with pastel B) followed by incubation for 5 minutes on ice with additional 6 ml of EZ PREP buffer. During incubation, 1 µM DAPI was added to the homogenate and subsequently filtered through a 35 µm strainer. Intact nuclei were sorted with a BD FACSAriaIII with a 70 µm configuration into 1.5 ml-Eppendorf tubes with 40 µl of 4 % BSA in PBS and RiboLock RNase Inhibitor (25 U/µl, EO0381, ThermoFisher). A FSC/SSC based gate was used to exclude debris followed by exclusion of damaged nuclei in a DAPI-A/DAPI-H (see Supplementary Fig. 5a) . 150,000 events were sorted for each sample. The concentration of sorted nuclei was determined based on brightfield images and DAPI fluorescence using a Neubauer counting chamber and a Leica DMi8 microscope.

Single nuclei libraries were generated according to the Chromium Next GEM Single Cell 3′ Reagent Kits v3.1 User Guide (CG000204) by 10x Genomics. Briefly, a droplet emulsion was generated in a microfluidic chip followed by barcoded cDNA generation inside these droplets.

Purified and amplified cDNA was then subjected to library preparation and sequenced on a NovaSeq 6000 instrument (Illumina) to a median depth of 45-90k reads per cell.

Sequencing libraries were processed with CellRanger (v3.1.0) against the mouse genome (mm10) augmented by intronic sequences and the APP and PS1 transgenes, followed by background removal with CellBender (v0.2.0; REF doi: 10.1101/791699). snRNA-seq data analysis was performed in R (v3.6.3) with Seurat (v3.2.1) [79] . Cells with at least 250 but less than 6000 genes and less than 10% mitochondrial RNA content were combined from each library, and clustering and UMAP embedding was computed based on log normalized gene counts, with total RNA content regressed out during scaling and using 20 PCA components.

Cluster annotation was aided by marker expression and label transfer with Seurat's TransferData workflow using a previously published mouse brain dataset as reference "pseudo-bulk" counts for each cluster from each sample. Pathway analysis was performed using tmod [80] (v0.44) and Gene Ontology terms from the msigdbr package (v7.2.1).

All values are presented as mean ± SEM (standard error of the mean). For pairwise comparison between two experimental groups, the student´s t-test was used. Statistical differences between more than two groups were assesses with One-way ANOVA using the indicated post hoc test. Statistically significant values were determined using the GraphPad Prism software and are indicated as follows: *P < 0.05, **P < 0.01 and ***P < 0.001 

Immune attack: the role of inflammation in Alzheimer disease

Prenatal Immune Activation Interacts with Genetic <em>Nurr1</em> Deficiency in the Development of Attentional Impairments

Deciphering the mechanism underlying late-onset Alzheimer disease

Systemic immune challenges trigger and drive Alzheimer-like neuropathology in mice

Genome-wide association study of 14,000 cases of seven common diseases and 3,000 shared controls

Loss of the autophagy protein Atg16L1 enhances endotoxin-induced IL-1beta production

Autophagy proteins regulate innate immune responses by inhibiting the release of mitochondrial DNA mediated by the NALP3 inflammasome

The autophagy-related protein beclin 1 shows reduced expression in early Alzheimer disease and regulates amyloid β accumulation in mice

Microglial beclin 1 regulates retromer trafficking and phagocytosis and is impaired in Alzheimer's disease. Neuron

Beclin1-driven autophagy modulates the inflammatory response of microglia via NLRP3

The NLRP3 inflammasome: molecular activation and regulation to therapeutics

Activation of autophagy by inflammatory signals limits IL-1beta production by targeting ubiquitinated inflammasomes for destruction

NLRP3 is activated in Alzheimer's disease and contributes to pathology in APP/PS1 mice

NLRP3 inflammasome activation drives tau pathology

TLR4 (toll-like receptor 4) activation suppresses autophagy through inhibition of FOXO3 and impairs phagocytic capacity of microglia

Induction of autophagy by spermidine promotes longevity

Spermidine induces autophagy by inhibiting the acetyltransferase EP300

Cardioprotection and lifespan extension by the natural polyamine spermidine

Spermidine in health and disease

Spermidine alleviates experimental autoimmune encephalomyelitis through inducing inhibitory macrophages

Spermidine alleviates severity of murine experimental autoimmune encephalomyelitis

Neuroprotective potential of spermidine against rotenone induced Parkinson's disease in rats

Spermidine and spermine delay brain aging by inducing autophagy in SAMP8 mice

Spermidine endows macrophages anti-inflammatory properties by inducing mitochondrial superoxide-dependent AMPK activation, Hif-1α upregulation and autophagy

Anti-inflammatory effects of spermidine in lipopolysaccharidestimulated BV2 microglial cells

Spermidine Protects against Oxidative Stress in Inflammation Models Using Macrophages and Zebrafish

eIF5A hypusination, boosted by dietary spermidine, protects from premature brain aging and mitochondrial dysfunction

Dietary spermidine improves cognitive function

SARS-CoV-2-mediated dysregulation of metabolism and autophagy uncovers host-targeting antivirals

Microglia recognize double-stranded RNA via TLR3

Astrocytes recognize intracellular polyinosinic-polycytidylic acid via MDA-5. Faseb j

Inflammasomes in the CNS

Pivotal Advance: Arginase-1-independent polyamine production stimulates the expression of IL-4-induced alternatively activated macrophage markers while inhibiting LPS-induced expression of inflammatory genes

Spermine Alleviates Acute Liver Injury by Inhibiting Liver-Resident Macrophage Pro-Inflammatory Response Through ATG5-Dependent Autophagy

Spermidine protects from age-related synaptic alterations at hippocampal mossy fiber-CA3 synapses

Spermidine Suppresses Inflammatory DC Function by Activating the FOXO3 Pathway and Counteracts Autoimmunity. iScience

A gene network regulating lysosomal biogenesis and function

TFEB links autophagy to lysosomal biogenesis

Polyamines Control eIF5A Hypusination, TFEB Translation, and Autophagy to Reverse B Cell Senescence

Restoring polyamines protects from age-induced memory impairment in an autophagy-dependent manner

Abeta42-driven cerebral amyloidosis in transgenic mice reveals early and robust pathology

A Unique Microglia Type Associated with Restricting Development of Alzheimer's Disease. Cell

The TREM2-APOE Pathway Drives the Transcriptional Phenotype of Dysfunctional Microglia in Neurodegenerative Diseases

Profilin-1 is dysregulated in endometroid (type I) endometrial cancer promoting cell proliferation and inhibiting pro-inflammatory cytokine production

Protective potential of glucagon like peptide 2 (GLP-2) against the neurodegeneration

PER1 prevents excessive innate immune response during endotoxin-induced liver injury through regulation of macrophage recruitment in mice

Overexpression of Sirt3 inhibits lipid accumulation in macrophages through mitochondrial IDH2 deacetylation

Sirtuin 3-induced macrophage autophagy in regulating NLRP3 inflammasome activation

Molecular Mechanisms of Microglial Motility: Changes in Ageing and Alzheimer's Disease. Cells

Alzheimer's disease: microglia targets and their modulation to promote amyloid phagocytosis and mitigate neuroinflammation

CELF2 regulates the species-specific alternative splicing of TREM2

Sirtuin 3 mRNA Expression is Downregulated in the Brain Tissues of Alzheimer's Disease Patients: A Bioinformatic and Data Mining Approach. Medical science monitor : international medical journal of experimental and clinical research

Alterations of Transcription of Genes Coding Anti-oxidative and Mitochondria-Related Proteins in Amyloid β Toxicity: Relevance to Alzheimer's Disease

Sirtuin 3 promotes microglia migration by upregulating CX3CR1

Functional Impairment of Microglia Coincides with Beta-Amyloid Deposition in Mice with Alzheimer-Like Pathology

Selective proliferative response of microglia to alternative polarization signals

Dendritic Spines in Alzheimer's Disease: How the Actin Cytoskeleton Contributes to Synaptic Failure. International journal of molecular sciences

Research on the role of GLP-2 in the central nervous system EPK signal transduction pathway of mice with vascular dementia

A clock gene, period, plays a key role in long-term memory formation in Drosophila

Temporal dynamics of mouse hippocampal clock gene expression support memory processing

Mechanisms by which autophagy regulates memory capacity in ageing

Neuronal hyperactivity--A key defect in Alzheimer's disease?

The Cellular Phase of Alzheimer's Disease. Cell

Increased soluble amyloid-beta causes early aberrant brain network hypersynchronisation in a mature-onset mouse model of amyloidosis

Soluble amyloid-b oligomers as synaptotoxins leading to cognitive impairment in Alzheimer's disease. Frontiers in Cellular Neuroscience

Neurotoxicity and Memory Deficits Induced by Soluble Low-Molecular-Weight Amyloid-β<sub>1-42</sub> Oligomers Are Revealed <em>In Vivo</em> by Using a Novel Animal Model

Naturally secreted oligomers of amyloid β protein potently inhibit hippocampal long-term potentiation in vivo

Microglia use TAM receptors to detect and engulf amyloid β plaques

Inflammatory cytokine levels correlate with amyloid load in transgenic mouse models of Alzheimer's disease

Caloric restriction attenuates amyloid deposition in middle-aged dtg APP/PS1 mice

Metformin Improves Learning and Memory in the SAMP8 Mouse Model of Alzheimer's Disease

Rapamycin and Alzheimer's disease: Time for a clinical trial

The effect of spermidine on memory performance in older adults at risk for dementia: A randomized controlled trial

Effects of spermidine supplementation on cognition and biomarkers in older adults with subjective cognitive decline (SmartAge)-study protocol for a randomized controlled trial

Spermidine intake is associated with cortical thickness and hippocampal volume in older adults

Age-dependent changes in brain, CSF, and plasma amyloid (beta) protein in the Tg2576 transgenic mouse model of Alzheimer's disease

Tricine-sodium dodecyl sulfate-polyacrylamide gel electrophoresis for the separation of proteins in the range from 1 to 100 kDa

Gene set enrichment for reproducible science: comparison of CERNO and eight other algorithms