Ablation of cell-specific cholesterol synthesis affects cerebral β-amyloidosis

..................................................................................................................................... 6


The central nervous system and its cells
The central nervous system (CNS) consists of two parts, the brain and spinal cord. Despite the fact that the brain is only 2% of the body weight, the brain uses 20% of the overall oxygen consumption to fulfil functions ranging from transmission of neuronal stimuli to immune defense mechanisms (Gallagher et al., 1998;Hyder et al., 2013). To fulfill this variety of functions, the brain is equipped with highly specialized cell types that can be roughly classified as neurons and glia cells. Via synapses, neurons form chemo-electrical circuits in which impulses can be conveyed between cells. Oligodendrocytes are derived from oligodendrocyte precursor cells (OPC); these comprise myelinating oligodendrocytes, which generate myelin to ensheath axons for faster signal propagation, and satellite cells that regulate the perineuronal microenvironment (Simons & Nave, 2015). Astrocytes, the most abundant cell type in the CNS provide structural support, are involved in formation of the blood brain barrier (BBB), and maintain synaptic transmission. Regulation of CNS homeostasis, repair following injury, and modulation of inflammatory events are other tasks performed by astrocytes (Sofroniew & Vinters, 2010). The main immunoregulating and immunosurveilling cells in the brain are microglia cells. Microglia can acquire phagocytotic activity to remove cellular debris and abnormal deposits (Li & Barres, 2018).
Ependymal cells are located between nervous tissue and cerebrospinal fluid (CSF) with cilia protruding into the CSF to circulate the fluid. The exchange of molecules between the periphery and brain interstitial fluid is regulated by endothelial cells that form the BBB together with astrocyte end-feet and pericytes (Del Bigio, 2010). Pericytes partially cover endothelial cells and regulate tight junction (TJ) proteins, which regulates BBB permeability. Due to the BBB, lipid metabolism is largely separated from the peripheral circulation (Armulik et al., 2010).

CNS lipids and cholesterol
The brain is a highly lipid-rich organ and comprises roughly one quarter of the body cholesterol.
Lipids are essential for forming cell membranes in general and for myelin in particular as a membrane protrusion by maintaining membrane fluidity, permeability and electrical characteristics and thereby facilitating transmembrane signal transduction and vesicular trafficking (Orth & Bellosta, 2012). Due to the BBB, lipoproteins and thereby cholesterol uptake into the brain is prevented therefore the synthesis of cholesterol, the key structural component of lipid membranes takes place endogenously (Bjorkhem & Meaney, 2004). Each cell type has its individual need for sterols and therefore the capability to synthesize cholesterol. In vitro studies showed a 2-3 times higher production rate of cholesterol in astrocytes compared to neurons and an even higher synthesis activity in oligodendrocytes that could meet the needs for developmental myelination. Further studies suggest the capability of embryonic neurons to generate cholesterol and a redistribution of cholesterol production in the adult brain (Fünfschilling et al., 2007;Koper et al., 1984;Pfrieger, 2003;Saher et al., 2005;Saito et al., 1987). In the adult brain, when myelination is predominantly completed, stable but rather low levels of cholesterol synthesis ensure homeostasis. Cholesterol can be transferred between CNS cells depending on the cellular demands. Excess cholesterol can be exported from the CNS via two ways, in the shedding pathway surplus cholesterol is released into the CSF associated with apolipoprotein E (ApoE). In a second pathway cholesterol is exported via conversion to oxysterols that can pass the BBB due to their physical properties (Bjorkhem, 2006).

Cholesterol biosynthesis and metabolism
All mammalian nucleated cells have the ability to synthesize cholesterol. Cholesterol biosynthesis ( Figure 1) is an energy consuming, complex multistep process that involves many enzymes (Bloch, 1965). Under physiological conditions cholesterol levels are quite stable only a small proportion is replenished on regular basis. The daily production rate of sterols in the brain is in the magnitude of 15-20mg/g bodyweight in several species, including mice (Dietschy & Turley, 2004). The counterparts of biosynthesis and excretion are tightly regulated to maintain homeostasis. De novo synthesis occurs predominantly in the endoplasmic reticulum (ER). The first part of cholesterol synthesis involves the isoprenoid biosynthesis pathway. Acetyl-CoA and acetoacetyl-CoA are converted to 3-hydroxy-3-methylglutaryl-CoA (HMG-CoA) by the enzyme HMG-CoA synthase (HMGCS) followed by the reduction to mevalonate by HMG-CoA reductase (HMGCR). HMGCR, one of the rate-limiting enzymes of the sterol synthesis is the target of statins, a widespread group of cholesterol lowering drugs. Mevalonate is activated by two subsequent phosphorylation steps to form mevalonate 5-phosphate and mevalonate 5-diphosphate. The activated isoprenoid molecule isopentenyl pyrophosphate (IPP) results from the decarboxylation of mevalonate 5-diphosphate. In equilibrium with IPP is its isomer dimethylallyl pyrophosphate (DMAPP), IPP and DMAPP combine to form geranyl pyrophosphate (GPP). By adding another IPP molecule farnesyl pyrophosphate (FPP) is generated, the condensation of FPP results in squalene. Squalene formation is the first committed step of the cholesterol synthesis and is catalyzed by farnesyl-diphosphate farnesyltransferase 1 (FDFT1). Squalene epoxidase (also called squalene monooxygenase, SQLE) catalyzes the second rate-limiting step to lanosterol. Lanosterol is fed into two different pathways, the Bloch pathway and the Kandutsch-Russell pathway. These two pathways differ in the order of enzymatic steps. The conversion of lanosterol into desmosterol via the Bloch pathway starts with the reduction of the double bond catalyzed by the enzyme 14-alpha-demethylase (CYP51) and is followed by several reactions forming zymosterol, dehydrolathosterol, dehydrodesmosterol, and finally desmosterol. The intermediate desmosterol can be converted to cholesterol, catalyzed by the enzyme 24-dehydrocholesterol reductase (DHCR24) (Bloch, 1965). The post-lanosterol steps in the Kandutsch-Russell pathway  are dehydrolanosterol, zymostenol, lathosterol, and as terminal intermediate 7dehydrocholesterol, which can be converted to cholesterol by 7-dehydrocholesterol reductase (DHCR7) (Kandutsch & Russell, 1960). Flux analysis studies suggested a third pathway that combines both pathways in a "modified Kandutsch-Russel" pathway. In this modified pathway, cholesterol is produced within the Bloch pathway and at later steps changing to Kandutsch-Russel pathway using DHCR24 (Mitsche et al., 2015).

Regulation of brain cholesterol
Cholesterol metabolism is differentially regulated in diverse brain areas. This includes differentially expressed cholesterol synthesis genes, cholesterol shuttle proteins, intracellular transporters and lipoprotein receptors are region specific, even the abundance of the sterol sensing protein SREBP cleavage-activating protein (SCAP) varies. Transcription of the ratelimiting enzyme Hmgcr and Fdft1 is partially regulated by the transcription factor sterol regulatory element-binding protein 2 (Srebp-2) (Brown & Goldstein, 1997;Luo et al., 2020) (Figure 2). In high cholesterol circumstances the inactive SREBP-SCAP complex in anchored in the ER, the Insulin-induced gene 1 protein (INSIG) is bound to SCAP, which senses cholesterol in the ER membrane.
Low cholesterol conditions lead to unbound INSIG and thereby conformational changes of SCAP to transport the SREBP-SCAP complex to the Golgi apparatus via COPII vesicles. In the Golgi, SREBP undergoes proteolytic cleavage. The mature SREBP2 transcription factor binds sterol regulatory elements (SRE) in the nucleus to induce transcriptional regulation (Luo et al., 2020).
Another major regulator of cholesterol homeostasis are liver X receptors (LXRs) that serve to reduce cellular cholesterol levels. As SCAP, these nuclear receptors are cholesterol sensors and form heterodimers with retinoid X receptors (RXRs).
Endogenous ligands like oxysterols change the conformation of the heterodimer causing the dissociation of the corepressor complex and recruitment of the coactivator complex to initiate transcription of target genes like the apolipoprotein (ApoE) and the ABC-transporter Abca1 and Abcg1 (Bielska et al., 2012;Courtney & Landreth, 2016). The lipid transporters ABCA1/ABCG1 export cholesterol from the cells to lipidated APOE. APOE is the main component of the high density lipoproteins (HDL) like particles that are secrete into the intestinal fluid (Orth & Bellosta, 2012 (Kiss & Nagy, 2016;Muse et al., 2018;Wang et al., 2008;Zelcer et al., 2009).
To prevent accumulation of excess free cholesterol in the brain, cholesterol is hydroxylated to oxysterols to pass the BBB. Another way to reduce excess cholesterol is the esterification by acyl coenzyme-A cholesterol acyltransferase (ACAT1, also called SOAT1). SOAT1 can transform the free cholesterol into cholesterol esters (CE) for storage and transport via lipid droplets (Rogers et al., 2015). 1.1.4 Cholesterol turnover and excretion from the CNS CNS cholesterol is considerably stable, the sterol has a long half-life of four to six months and has a slow turnover with approximately 0,4% per day (Dietschy, 2009;Vance et al., 2005). About 99.5% of CNS sterols are unesterified cholesterol with a low quantity of desmosterol and cholesteryl esters. Around 70% of the brain cholesterol is stored in myelin sheaths and another considerable amount is located in plasma membranes. In plasma membranes, cholesterol is extensively enriched in membrane microdomains, these "lipid rafts" play a central role in A B membrane organization and trafficking (Bjorkhem & Meaney, 2004;Silvius, 2003;Simons & Ehehalt, 2002).
The synthesis of cholesterol is an energy consuming process and therefore an efficient apolipoprotein-dependent recycling mechanism in the brain results in a low excretion rate into the periphery. In the case of excess cholesterol, shedding of cholesterol in association with ApoE into the CSF is limited to 1-2 mg per day; about 6-7 mg per day can be released into the circulation in the form of oxysterols. Oxysterols can pass the BBB due to their lipophilic properties faster than non-oxidized cholesterol (Bjorkhem & Meaney, 2004;Orth & Bellosta, 2012). The excretion through the BBB is sustained by cholesterol 24 hydroxylase (CYP46) a highly brain specific enzyme that is mainly expressed in neurons. CYP46 converts cholesterol to 24 hydroxycholesterol (24-OHC) to pass the BBB. The hydroxylated cholesterol 27hydroxychoelsterol is mainly found in plasma but is imported into the CNS t of about 5mg daily (Vance et al., 2005). In the periphery the relation between cholesterol and oxysterols is more than 10.000-fold, in the brain the proportion of oxysterols is greater and varies between 1 to 500 and 1 to 1000. Oxysterols can be incorporated in lipoproteins and can be shuttled cells (Bjorkhem & Meaney, 2004;Orth & Bellosta, 2012).

Lipid transport between cells
Given the low rate of cholesterol synthesis in some CNS cells in the adult brain, sterols need to be transported between brain cells via CNS lipoproteins to maintain lipid homeostasis. Lipoprotein particles consist of a hydrophobic core and a hydrophilic surrounding. Within the lipophilic core polar lipids can be transported through aqueous environment like the blood circulation. A similar but less complex transport system is also present in the brain. Some components of lipoproteinmediated transport exist in the periphery and in the brain whereas some apolipoproteins alternate.
All brain cells express the mRNA to synthesize LDLR for the uptake of apolipoproteins. A second uptake route could be the LDL receptor-related protein (LRP) (Boyles et al., 1985;Bu et al., 1994;Danik et al., 1999). In addition to this, cholesterol efflux via ATP-binding cassette transporters (ABCA1) has a significant role in sterol flux between cells (Karasinska et al., 2009;Wellington et al., 2002). ABCA1 facilitates secretion and lipidation of ApoE and consequently the transport of cholesterol (Hirsch-Reinshagen et al., 2004;Wahrle et al., 2004) Apolipoprotein E is the most abundant apolipoprotein in the brain, other lipoproteins found in the brain are apolipoprotein A1 (ApoA1), apolipoprotein J (ApoJ) and apolipoprotein D (ApoD). ApoE is believed to be mainly produced by astrocytes and can be taken up by other brain cells. In humans, three different alleles of ApoE are distinguished, ApoE2, ApoE3, ApoE4. Carriers of the ApoE2 allele contain lower blood cholesterol and low-density lipoprotein levels but higher high-density lipoprotein and triglyceride levels compared to Apoe4 carriers. In the human brain, ApoE4 leads to an altered lipidomic profile and the reduction in important lipid classes (Lefterov et al., 2019;Wolters et al., 2019). ApoE isoforms differ in the lipidation status, ApoE2 is highly lipidated whereas Apoe4 is poorly lipidated which has an effect on their functions like cholesterol and lipid transport (Lanfranco et al., 2020). Beside the central role of ApoE for cholesterol metabolism it is the leading risk factor for AD (Corder et al., 1993;Strittmatter et al., 1993 (Galvagnion et al., 2016;O'Leary et al., 2018). The lysosomal storage disorder Niemann-Pick disease is caused by mutation of either of two genes involved in lysosomal storage of cholesterol and sphingolipids, Npc1 and Npc2. (Porter et al., 2010;Winkler et al., 2019). Reduced expression of genes related to sterol synthesis were detected in post mortem cortex and striatum tissue of Huntington disease patients and animal models. Studies indicated positive effects on behavioral outcomes of HD mouse models when infusing cholesterol into the striatum (Birolini et al., 2020;Leoni et al., 2011). A further imbalance of cholesterol and lipoprotein levels in cerebrospinal fluid (CSF) and serum was detected in patients with Amyotrophic lateral sclerosis (ALS), a neurodegenerative disease that mainly affects motor neurons.
The role of circulating cholesterol levels is not well understood, elevated serum cholesterol levels showed a protective prognosis but contradicting to that a higher risk for developing ALS (Abdel-Khalik et al., 2017;Dodge et al., 2021;Ingre et al., 2020). Unprovoked recurrent seizures are the hallmark of epilepsy. Several studies investigate the impact of cholesterol on neuronal excitation and hereby the likelihood for seizures using an inhibitor of cholesterol 24-hydroxylase, to reduce the conversion of cholesterol to oxysterols (Aird & Gurchot, 1939;Hawkins et al., 2021;Warren et al., 2018). Experimental autoimmune encephalomyelitis (EAE) is a model for Multiple sclerosis, which represents the inflammatory and demyelinating aspects of the disease. In the human disease as well as in EAE phagocytes are recruited to the lesion area to remove myelin debris.
In the mouse model, repair fails when microglia lack intact sterol synthesis. Moreover, disrupted cholesterol export from phagocytes leads to reduced remyelination Cantuti-Castelvetri et al., 2018;Spiteri et al., 2022). In Alzheimer´s disease the connection between lipid metabolism and different components of the disease are very complex and will be discussed in a separate section. report about a 50-year-old woman called Auguste Deter. In the paper "Über eine eigenartige Erkrankung der Hirnrinde"/"An Unusual Illness of the Cerebral Cortex" he described symptoms ranging from paranoia and disorientation to progressive memory loss. In autopsy material he found neurofibrillary changes, small accumulations caused by a substance deposition, reactive glia cells, and brain atrophy. In 1908 Emil Kraepelin, the mentor of A. Alzheimer published the 8 th edition of the textbook Psychiatrie and proposed the name Alzheimer´s disease to describe the characterized "unusual illness of the cerebral cortex" (Alzheimer, 1907;Hippius & Neundörfer, 2003;Kraepelin, 1913;Stelzmann et al., 1995).
Alzheimer´s disease, also called Morbus Alzheimer is the most common cause of dementia characterized by a progressive disease course, accelerating pathological accumulations of extracellular amyloid and intracellular tau, and decline of cognitive functions. Nowadays, approximately 50 million people suffer from dementia worldwide and AD accounts for 60-80% of these cases ("2020 Alzheimer's disease facts and figures," 2020; "World Alzheimer Report 2019,"). AD causes the decline of memory functions, language disabilities, and alterations of behavioral and social skills. Disease progression leads to a broad range of neuropsychiatric disorders such as sleep disturbances, mood change, depression, anxiety or aggressiveness.
Together with irritability and delusions this can result in personality changes. These common symptoms are often summarized as behavioral and psychological symptoms of dementia. The first histopathological brain abnormalities establish themselves years before the diagnosis of dementia. Cognitive tests and neurological evaluation in addition to neuroimaging methods such as magnet resonance imaging (MRI) and computerized tomography (CT). In the CSF, a decrease in Aβ42 or increases in total tau and phosphorylated tau are reflecting the disease process. And are nowadays used as AD biomarkers. To a certain extent this is even reflected in the blood but the quality of AD blood biomarkers is currently controversially discussed (Janelidze et al., 2020;Khan et al., 2020). A confirmed diagnosis of AD can only be achieved post-mortem, by histopathological evidence of the pathognomonic signs of AD. The histopathological hallmarks of AD are intraneuronal excess of hyperphosphorylated tau resulting in neurofibrillary tangles and extracellular aggregations of beta amyloid forming senile plaques (Hansson, 2021;Porsteinsson et al., 2021). Heiko and Eva Braak assessed these pathological characteristics in different stages of disease visualized by Bielschowsky silver staining and classified the disease progression in Braak stages I to VI (Braak & Braak, 1991). In the surrounding of senile plaques, activated microglia induce an inflammatory microenvironment. At a later disease stage astrocytes become reactive and also arrange around the plaques however more distant and fewer in numbers (DeTure & Dickson, 2019). In addition to the formation of amyloid plaques and neurofibrillary tangles, patients with Alzheimer´s dementia suffer from a drastic loss of synapses which correlates with the degree of cognitive impairment (Terry et al., 1991).
There are two subtypes of AD, the hereditary autosomal dominant familial AD (FAD) and sporadic AD (SAD). FAD generally has an early onset. Only about 1% of the AD cases develop before the age of 65 years. These cases are classified as Early-onset (EOAD) and are predominantly caused by surplus amyloid production (Blennow et al., 2006). FAD/EOAD is caused by mutations in the genes encoding APP, Presenilin-1 (PSEN1) or Presenilin-2 (PSEN2).
The majority (90-95%) of AD cases include a late-onset (LOAD) and unknown etiology (Harman, 2006). Aging is the primary risk factor for AD. Further risk factors include female sex and in respect to lipid metabolism the major genetic risk factor Apolipoprotein E ε4 allele (apoE4).
Relations to cardiovascular disease, obesity, diabetes, hypercholesterolemia and inflammation are discussed. Furthermore, environmental factors (reduced brain capacity and mental ability, low physical activity, sleep disturbances, late-life depression, and head injuries) may serve as risk factors for SAD (A. Armstrong, 2019). Beside ApoE4 as a metabolic risk factor, a genome wide study revealed clusterin (CLU), phosphatidylinositol-binding clathrin assembly protein (PICALM), complement C3b/C4b receptor 1 (CR1), bridging integrator 1 (BIN1), ATP-binding cassette subfamily A member 7 (ABCA7) and others (Kamboh et al., 2012). Some variants of the triggering receptor expressed on myeloid cells 2 (TREM2) gene have been associated with AD by altering the microglial interaction with the amyloid plaques and phagocytic activity. Due to the complexity of the disease, the underlying cause is not resolved yet.

Amyloid cascade hypothesis
The amyloid cascade hypothesis was first described in 1991 and has become the most widely accepted hypothesis ever since (Beyreuther & Masters, 1991;Hardy & Allsop, 1991;Selkoe, 1991). The amyloid cascade hypothesis states that the β-amyloid protein is central for developing Alzheimer´s disease and triggers a detrimental cascade. The strongest indication for β-amyloid as the driving force of AD is the fact that the FAD genes APP, PSEN1 and PSEN2 are all involved in Aβ generation (Karran et al., 2011). Another genetic correlation referring to amyloid as starting point of AD is the elevated risk for developing the disease with the chromosomal anomaly trisomy 21. This genetic condition is caused by an extra copy of the chromosome 21, the chromosome where APP is localized. Most patients have significant increased levels of amyloid and tau by the age of 40 years (Masters et al., 1985). However, the amyloid cascade hypothesis does not consider the direct interaction between amyloid and tau as the two hallmarks of the disease.
Mutations in the tau gene can lead to autosomal dominant dementia of the frontotemporal lobe but not to AD with additional amyloid pathology. Considering this fact, tau is more likely downstream of amyloid (Hutton et al., 1998). Further studies using APP transgenic mice show that reduced tau levels are beneficial for behavioral changes and amyloid-mediated deficits, which is in line with the amyloid hypothesis (Roberson et al., 2011;Roberson et al., 2007).
At present, there is no cure for AD and current therapies focus on symptomatic treatment. Several clinical trials that tested a broad range of drugs to target the processing and production of amyloid production, preventing amyloid aggregation or enhance its clearance, failed. Last year the antiamyloid antibody, Aducanumab was approved by U.S. Food and Drug Administration (FDA) to treat AD. In clinical trials Aducanumab treatment reduced amyloid burden but was inefficient in alleviating the cognitive dysfunction and its approval has sparked controversy in the AD research field (Fillit & Green, 2021).
The lack of therapeutic achievements points in a direction of an intervention before clinical symptoms become apparent. Other approaches focus on targeting tau, inflammatory responses, and pathogenic infections among others (Liu et al., 2019). After many years of unsuccessful clinical trials focusing on amyloid reduction, the direction of AD therapy has expanded in numerous trajectories. One trajectory is "untangling" the tau protein, which is directly connected to cognitive decline. Another focus is on the role of neuroinflammation and the consequence of alternating the inflammatory surrounding. One possibility would be to rejuvenate exhausted immune cells for enhanced clearing and thereby more effective plaque reduction (Chee & Solito, 2021). As a common future perspective, the intervention at an early prodromal disease stage is the prevailing opinion for a successful treatment and therefore an early diagnosis involving valid biomarkers is an essential prerequisite (Porsteinsson et al., 2021).

Amyloid precursor protein processing
The amyloid precursor protein (APP) is the origin of the pathogenic amyloid-β protein. APP is located on the long arm of chromosome 21 and is ubiquitously expressed but highly abundant in neurons. Different isoforms of APP are generated by alternative splicing, ranging from 365 to 770 amino acid residues. APP695 is the most abundant in the brain followed by APP751 and APP770. (Sandbrink et al., 1996;Zhang et al., 2011). When newly synthesized APP passes through the secretory pathway, this single-pass transmembrane protein undergoes several posttranslational modifications including proteolytic processing, glycosylation, phosphorylation and sulfation, which is called APP maturation. In the maturation process APP relocates from the ER to the Golgi apparatus to the plasma membrane. Most of the APP protein remains associated to the Golgi apparatus. Some APP is localized in the plasma membrane, but a substantial part is internalized again and sorted into early endosomes and either recycled to the plasma membrane or degraded in lysosomes (Caporaso et al., 1994;Haass & Selkoe, 1993;Koo & Squazzo, 1994). cleavage. Aβ can be removed by Aβ-degrading enzymes or accumulates to form extracellular deposits as senile plaques (adjusted from David Goodsell, (Berman et al., 2000)).
Proteolytic processing of APP involves consecutive cleavage, starting with αor β-cleavage, followed by intramembrane γ-cleavage ( Figure 3). The processing is distinguished in the amyloidogenic pathway and the non-pathogenic pathway. In the non-amyloidogenic pathway, the α-secretase ADAM10, a disintegrin and metalloproteinase domain-containing protein 10 cleaves at the c-terminal side generating an 83 residue c-terminal fragment (C83) and the soluble APP α (sAPPα). C83 is then cleaved by the γ-secretases Presenilin-1 or Presenilin-2 to form P3, a 3 kDa non-amyloidogenic peptide. The non-amyloidogenic pathway takes place mainly in the plasma membrane where α-secretase is located. The amyloidogenic pathway includes beta-secretase 1 (BACE1) and subsequent cleavage by PSEN1/2 resulting in a 99 residue c-terminal fragment (C99) and as an end product the Aβ peptide. The β-secretase BACE1 is mainly located in the trans-Golgi network (TGN) and endosomes where APP is processed in the amyloidogenic pathway. This pathway was shown to be pH dependent, therefore, an acidic compartment is an important prerequisite for β-cleavage (Haass & Selkoe, 1993;Zhang & Song, 2013). γ-cleavage can result in different Aβ peptides depending on the cleavage site in the transmembrane domain.
The primary amino acid sequence of amyloid-β was initially sequenced by Glenner and Wong in 1984 from purified fresh frozen autopsy material of CAA patients. The amino acid sequence comprises peptides from 37 to 49 residues, Aβ40, a 40 amino acid long peptide is the major form but Aβ42 plays a pivotal role in AD. (Glenner & Wong, 1984) Mutations in APP can be classified into mutations in proximity to the Aβ locus, immediately before or directly after the Aβ sequence or within. The Swedish mutation (APPswe) is a double mutation at amino acid 670 and 671 just prior to the sequence of Aβ1-x. This mutation causes enhanced BACE1 cleavage and accelerated Aβ generation. Mutations carboxyterminally of the Aβ sequence affect the γ-secretase activity. The mutation at position 717 (APP717) increases the Aβ42/ Aβ40 ratio (Savonenko et al., 2015). APP mutations within the Aβ sequence can lead to conformational changes of the peptide that change aggregation kinetics. Most of the mutations within APP modify biochemical properties and thereby often creating highly pathological variants (Hatami et al., 2017). Most of the mutations either increase the Aβ production, the Aβ42/ Aβ40 ratio, or aggregation properties of Aβ. Beyond these mutations, there are some protective mutations, the best known is A673T also called Icelandic mutation. This mutation is located near the Nterminal end of Aβ and reduces amyloidogenic cleavage by interfering with BACE1 cleavage (Jonsson et al., 2012;Tambini et al., 2020).

Aβ aggregation and plaque seeding
In contrast to pathological levels, physiological levels of Aβ and other APP processing products support synaptic plasticity and memory (Lazarevic et al., 2017;Puzzo et al., 2011;Puzzo et al., 2008).
Pathological levels of Aβ42 play a critical role in AD pathology. Aβ monomer to oligomer transition is the first step of the formation of insoluble Aβ fibrils and aggregation of plaques Nag et al., 2011). Aβ oligomer levels are linked to neurotoxicity and the development of cognitive deficits. However, these correlations are weak because of the instable nature of Aß oligomers and their transition to form fibrils (Jongbloed et al., 2015;McLean et al., 1999;Ono et al., 2009). Primary murine neurons treated with stabilized Aβ42 oligomers revealed the highly neurotoxic nature in comparison to Aβ fibrils. Amyloid fibrils harbor a secondary β-sheet structure which is a common biochemical property to form fibril and protein aggregations (Ahmed et al., 2010).
Senile plaques are composed of Aβ fibrils. Fibrillary aggregates originated from Aβ42 are different compared to fibrils from Aβ40. These peptides differ only in two amino acid residues, but Aβ42 tend to form β-sheet like structures (Kim et al., 2007). Furthermore, Aβ42 aggregates faster under artifact-free conditions when compared to Aβ40 (Nirmalraj et al., 2020). Interestingly, the concentration of Aβ in CSF is lower than the concentration which is needed for spontaneous formation of fibrils (Seubert et al., 1992). Further conversion from fibrils to amyloid plaques is not well understood but some studies described intracellular Aβ accumulations in various cells after repetitive synthetic Aβ induction (Friedrich et al., 2010;Gellermann et al., 2006). Another study showed, neuroglia co-cultures treated with Aβ42 protofibrils generates neurotoxic extracellular vesicles, inducing axonal swellings, neuronal cell body vacuolization, and pathological lysosomal cholesterol deposits (Beretta et al., 2020). Previously, intraneuronal amyloid accumulations has not been the focus of AD research because extracellular aggregates within plaques outnumber the intracellular accumulations. A number of reports, however showed aggregates with in neurons and a connection to synaptic pathology (Bayer & Wirths, 2010;Oddo et al., 2006;Takahashi et al., 2017). Present data suggest that amyloid plaques originate from intracellular Aβ accumulations that get released during degeneration of synapses or neurites and form the starting point for expansion into the extracellular parenchyma (Gouras et al., 2010) . Toxic oligomeric Aβ is deposited into developing amyloid plaques which could be a possible mechanism for reducing the circulating Aβ levels in brain tissue. With increased amyloid pathology inflammation exacerbates, leading to the formation of dystrophic neurons and the corralling of activated glia around senile plaques in later stages of the disease.

Aβ degradation and clearance
Increased Aβ levels enhance the likelihood for accumulation of the peptide, making nonenzymatic clearance and enzymatic degradation of Aβ important mechanism to maintain homeostasis. Emerging evidence suggest impaired clearance as a crucial factor for the development of AD, especially in sporadic AD (Cheng et al., 2020;Weller et al., 2000). There are several ways to reduce the amount of Aβ in the CNS. The non-enzymatic clearance can be achieved by Aβ drainage via lymphatic pathways. The excretion can be accomplished via CSF drainage from the meningeal lymphatics or along the cranial nerves into the deep cervical lymph nodes (Ahn et al., 2019;Cheng et al., 2020). A further possibility is the perivascular pathway, Aβ can be diminished by interstitial fluid drainage through capillary walls into the internal carotid artery (Weller et al., 2009). Clearance based on the glymphatic pathway uses the paravascular space between vessels, glia endfeet and the leptomeninges to reduce CSF Aβ (Feng et al., 2020;Iliff et al., 2012). Additionally, to lymphatic efflux, receptor mediated transport into blood vessels via the BBB and subsequent transfer into the circulation exists. Excretion is orchestrated mainly by the endothelial cell receptors LDL Receptor Related Protein 1 (LRP1) in association with Phosphatidylinositol Binding Clathrin Assembly Protein (PICALM) (Hartz et al., 2018;Shibata et al., 2000). An abnormal clearance via the blood vessels can result in cerebral amyloid angiopathy (CAA) which is part of the pathological pattern of many AD patients (Love, 2004). Furthermore, via large vacuoles CSF can drain directly through the BBB into the circulation, known as the arachnoid granule-venous sinus pathway (Tripathi & Tripathi, 1974). Another clearance mechanism is the uptake of soluble Aβ and the fibrillary form by astrocytes and to a greater extent by microglia cells (Rogers et al., 2002). However, microglial Aβ clearance rate declines with age and disease progression (Flanary et al., 2007;Floden & Combs, 2011;Hickman et al., 2008;Wyss-Coray, 2006). Microglia degrade the incorporated Aβ by common degradation pathways involving lysosomes, endosomes and autophagosomes (Majumdar et al., 2008). Another possibility is the enzymatic degradation of Aβ by neurons, including various Aβ-degrading enzymes (ADE): insulysin (also called insulin-degrading enzyme, IDE), neprilysin (NEP), matrix metalloproteinase (MMP9) and glutamate carboxypeptidase II (GCPII) and others. Despite effective in animal studies, the beneficial effect of the Aβ cleavage from senile plaques is under debate, considering that smaller Aβ-derived peptides have toxic properties as well (Eckman et al., 2006;Farris et al., 2007;Iwata et al., 2004).

Mouse models of AD
Animal models are a widely used tool in AD research to model aspects of the complex disease pathogenesis in vivo. Murine models are the most common model due to the rather low husbandry costs, the convenience of a fast generation of descendants and the ease of genetic manipulation (Götz et al., 2018). An optimal model would harbor the histopathological hallmarks of AD, Aβ plaques and neurofibrillary tangles of hyperphosphorylated tau accompanied by gliosis and synaptic and neuronal aggravations. Furthermore, the pathological changes should be comparable to the human disease, both in area and progression. To evaluate clinical relevance, cognitive decline should be a feature of such an animal model as well. For mimicking AD several mouse models are available, some of them present with amyloid plaques and others with tau tangles. The disease onset, the severity of gliosis, synaptic changes and cognitive impairment is very different. For example, the APP NL-G-F knock-in mouse builds senile plaques, the APP23 mouse model develops plaques and hyperphosphorylated tau but no tau tangles, the PS19 mouse displays tangles but lacks the amyloid burden.
A widely used mouse model of AD is the 5xFAD strain harboring 5 mutations linked to familial Alzheimer´s disease. This transgenic mouse line contains the Swedish (K670N/M671L), Florida (I176V), and London (V717I) mutation of the APP gene and two mutations in PSEN1, M146L and L286V driven by the neuronal Thy1 promotor (Oakley et al., 2006). These mutant mice present with an aggressive and early-onset AD-related pathology including severe amyloid pathology, gliosis, and synaptic degeneration and progressive cognitive deficits. The major disadvantage of this model is the lack of tau tangles, the second major hallmark of the human Alzheimer´s disease.

The role of microglia under physiological conditions and in neurodegenerative disease
Microglia were first visualized and described by Pio del Hortega in 1919 and originate from the yolk sac. During embryogenesis erythromyeloid progenitors migrate into the brain parenchyma and develop into immature microglia that start ramifying and populating the brain (Kierdorf et al., 2013). During CNS homeostasis, microglia are quite stable and nearly half of the microglia in a mouse brain can survive throughout the entire life of the animal (Fuger et al., 2017). Microglia fulfill an exceptionally broad variety of tasks in the healthy brain, ranging from innate immune system functions, trophic support, and angiogenesis to cell-cell communication to maintain the healthy brain microenvironment. However, as soon as the homeostasis is disturbed by infections, protein aggregations, cell death or another inflammatory stimulus, microglia start proliferating and change their homeostatic expression profile to a reactive signature.
Microglia cells take responsibilities for neuroprotection and produce growth factors for supporting neuronal survival, induce programmed cell death of surplus and defective neurons, eliminate malfunctioning synapses and clear the cellular debris (Badimon et al., 2020;Frade & Barde, 1998;Marıń-Teva et al., 2004;Paolicelli et al., 2011;Wakselman et al., 2008). In the developing white matter, microglia contribute to physiological myelinogenesis by maintaining oligodendrocyte progenitor maturation and maintenance (Hagemeyer et al., 2017). The interactions between astrocytes and microglia are bidirectional, as both cell types accomplish immune functions, has homeostatic responsibilities and micro-environmental balancing obligations. Astrocytes supply cholesterol loaded apoliporotein E and trophic support to microglia (Baxter et al., 2021;Bohlen et al., 2017;Saher & Stumpf, 2015).
In neurodegenerative diseases astrocytes and microglia transfer into reactive glia cells. Microglia participate in Aβ plaque clearing, contribute to the compaction of Aβ, and could thereby counteract the spreading (Clayton et al., 2021;Dionisio-Santos et al., 2019). Another remarkable study showed that microglia depletion before plaques arise, eliminates plaque deposition, suggesting a need of microglia for plaque deposition (Spangenberg et al., 2019). Further studies suggest an initial restriction of the disease progression by microglia and an age dependent transition closely associated with the exacerbation of amyloidosis. The gradual progression of the β-amyloid pathology leads to microglial production of proinflammatory cytokines and chemokines consequently reducing phagocytic activity (Heneka et al., 2010;vom Berg et al., 2012;Wang et al., 2015). Other studies showed phagocytes becoming more senescent with age, thereby limiting phagocytosing capacity and could lead to dysfunctional β-amyloid clearance (Flanary et al., 2007;Hickman et al., 2008). According to their expression profile, activated microglia have been classified as pro-inflammatory "M1" and an anti-inflammatory "M2" state (Tang & Le, 2016). Another microglia signature is associated with several neurodegenerative diseases, the so-called disease-associated microglia

Lipids in Alzheimer´s disease
Multiple connections have been made between cholesterol and Alzheimer´s disease (AD).
Several of the high-risk genes of AD like ApoE4 are associated with cholesterol metabolism. The ApoE4 allele has a population frequency of 13.7% and increases the risk of developing AD by 3fold in heterozygous carriers and more than 10-fold in homozygous carriers (Liu et al., 2013). A protective effect has been shown for homozygous carriers of the APOE2 allele. In physiological conditions, ApoE is mainly produced by astrocytes whereas in neurodegenerative diseases neurons as well as activated microglia also produce ApoE. It has been proposed, that microglia- Additionally, cholesterol content affects βand γ-secretase activity and low cholesterol levels shift the APP processing to the nonamyloidogenic processing in cell culture (Grimm et al., 2008;Kojro et al., 2001;Simons et al., 1998;Xiong et al., 2008). Decreased cholesterol levels and the adaption of the APP processing leads to reduced Aβ production (Ehehalt et al., 2003). APP can directly bind cholesterol and can localize to cholesterol-rich membrane rich lipid rafts. APP processing can be regulated by different lipid compositions which could explain the altered cholesterol and cholesterol ester levels of post mortem AD brains (Beel et al., 2008;Chan et al., 2012;Fabelo et al., 2014;Tajima et al., 2013). Furthermore, multiple genes that are highly upregulated in microglia during disease are connected to the regulation of lipid metabolism and microglia activation e.g. Trem2. In the DAM profile of activated microglia Lpl and Cst7 are upregulated. These genes are involved in the process of lipid uptake and phagocytosis and lipidrich debris (Hammond et al., 2019;Keren-Shaul et al., 2017). A study reported a correlation between increased expression of cholesterol 25 hydroxylase (Ch25h) with reduced phagocytic activity in microglia (Ofengeim et al., 2017). The multitude of links between AD and lipid metabolism suggest a pivotal role of lipids in the amyloidosis and the etiology of AD. However, despite decades of research, the detrimental function of lipids in this devastating disease is still not fully elucidated. We hypothesize that the analysis of cell type-specific responsibilities will help understanding disease processes.

Aim of the study
In addition to aging as most prominent risk factor, there is growing evidence that CNS cholesterol and lipid metabolism is a fundamental player in AD. Already in 1907, Alois Alzheimer described lipid accumulations as a hallmark of AD, besides the well-known appearance of senile plaques and tau tangles. In AD, dysfunctional plaque associated microglia with lipid accumulations have been described that show reduced phagocytic capacity (Marschallinger et al., 2020). These lipid accumulations have been linked to lipid droplets that are observed in post-mortem human AD tissue (Farmer et al., 2020).
Aβ generation within the amyloidogenic pathway of APP processing includes β-and γ-secretase, both intensively expressed in neurons and located within the cholesterol-rich membrane. More than two decades ago studies showed the association between high cholesterol levels and Aβ accumulations as well as increased β-secretase activity resulting in more amyloidogenic processing (Marquer et al., 2011;Wolozin, 2004).
Despite this strong correlative data, it still remains enigmatic how cell-specific cholesterol synthesis contributes to the amyloid pathology.
In this study, I investigated the relationship between cell type-specific cholesterol metabolism and amyloidosis. Therefore, I generated conditional mutants of sterol synthesis using the Cre To get insight into the relevance of cell type-specific cholesterol on Aβ generation, I investigated APP processing and amyloid generation in the complex mutants. In addition to amyloid generation, Aβ clearance could be influenced by the CNS cholesterol metabolism. To address the amyloid turnover independent of plaque formation, I performed ex vivo experiments by seeding isolated microglia on 5xFAD brain slices. It has been hypothesized that progressive plaque load is linked to increasingly inefficient microglial phagocytosis and turnover of Aβ. The explore the role of microglial cholesterol in this process, I determined phagocytosis rates in primary BMDM isolated from conditional mutants that were treated with Aβ. AD microglia show a specific signature of activated DAM genes. At present it is unclear whether these activation markers are required for amyloid deposition and clearance or are a marker of impaired turnover of amyloid.             dark and 12h light cycle. All mice used for this study (Table 12 List of generated mice) were facilitated at a C57BL/6 background and compared to sex and age matched controls, either as littermate controls or closely related controls. Genotyping of the mice used for experiments were performed from ear punches after weaning and confirmed with a second biopsy after sacrificing.

General laboratory devices
DNA isolation of ear biopsies was achieved by lysis with 50mM NaOH under boiling alkaline conditions and subsequent neutralization with 1M Tris/HCl pH8.0. Performing mutation specific polymerase chain reactions (Table 13 List of genotyping) and following gel electrophoresis for visualizing the amplified DNA fragments completes the determination of the diverse genotypes.
This method was performed mainly by Annika Schmidke and Vanessa Schlotzig (both student assistants).

Light sheet microscopy of whole tissue
Whole brain imaging was performed on the perfused tissue following the iDisco protocol (Liebmann et al., 2016;Renier et al., 2014) displayed in (Table 16 Dehydration and staining for light sheet microscopy). Protocol was applied by Dr. Constanze Depp and Andrew Sasmita . Following the clearing and staining protocol samples were imaged with the UltraMicroscope II (LaVision Biotec) in an Eci filled sample holder. The red fluorescence of congo red as acquired 80% laser power and a 585/40 emission filter in the mosaic acquisition mode (settings: 5µm thickness, 20% sheet width, 0.154 sheet numerical aperture, 4µm z-step size, 1000x1600 px, 4x4 tiling, dual light sheet illumination, 100ms camera exposure). Acquired and stitched whole brain images were annotated based on Allen brain atlas to analyze regions of interest separately .

Immunolabeling and epifluorescent microscopy
For fluorescent labeling of tissue sections, fixed brains were embedded in paraffin (Table 17 Paraffin embedding) using the STP 120 tissue processing machine (Leica) and HistoStar embedding station (Epredia) and cut into 5µm sections to mount and store on glass slides.
Paraffin sections were stained according to a well-established staining protocol (Table 18 Staining protocol for fluorescent labeling) and the listed antibodies (Table 7 List of Primary antibodies, Table 8 List of Secondary Antibodies, Table 9 List of Dyes) and finally mounted with Aqua Poly/Mount.  were sacrificed and brain was removed and meninges were removed from brain surface through rolling on whatman paper. Brain was sliced using a 1mm brain matrix to cut coronar sections and separate specific brain areas and immediately frozen on dry ice.
RNA was extracting as stated in following (Table 19 Tissue RNA extraction with Qiazol). After RNA isolation, concentration was measured and the calculated amount used for cDNA synthesis (Table 20 cDNA synthesis). Quantitative Real-time PCR was performed in 384 well plates using the SybrGreen mix (Promega), therefore 10ng cDNA was used for each of the triplicates. Gene specific intron spanning primes (Table 6 List of qRT-PCR primer sequences) were provided from the in-house AGCT lab and polymerase chain reaction was executed using a 45-cycle heating protocol (Table 21 qRT-PCR procedure) in the LightCycler 480 II. The corresponfing LightCycler 480 software was used for background substraction and thresholding and afterwards CT values (cycle thresholds) were exported to excel for further normalization to housekeeping genes (Table   5 List of housekeeping gene primer sequences). Results were analyzed using the ∆∆CT method and displayed significance was calculated using the Student´s t-test.

Cell isolation
For cell specific analysis cells were isolated according to the manufacturer's protocol of the Adult brain dissociation kit (Miltenyi Biotec) displayed in short (Table 22 Cell isolation via magneticactivated cell sorting). Magnetic-activated cell sorting (MACS) separates cell fractions using antibody covered beads for surface protein detection. The cell fraction was used for analysis of gene expression changes and therefore RNA was extracted using the RNeasy micro Kit (Table   23 RNA isolation of cells in RLT buffer), established for small amounts of mRNA and either send for transcriptomic analysis or precipitated and amplified using the Ovation PicoSL WTA System V2 (Tacan, Switzerland) Table 24 RNA precipitation with glycogen, Table 25 Single primer isothermal amplification).

Table 22 Cell isolation via magnetic-activated cell sorting (MACS)
Step

Solution Process
Tissue preparation DPBS Extract brain   (Table 24 RNA precipitation with glycogen) and SPIA (Table 25 Single primer isothermal amplification).  and Laemmli loading dye and loaded into separate lanes of gradient Tris-Tricine SDS PAGE gels (10-20%, Thermo Fisher). Gels were run at 120V for approximately 1 hour in Tricine running buffer and stopped when low molecular weight proteins are at the end of the gel for better resolution of c-terminal fragments. Proteins were transferred onto low fluorescent membrane (Immobilon-FL membrane, 0.45µm pore size, Merck) in a wet blot system for 1 hour at 500mA in blotting buffer.
Determination of the protein content was given by FastGreen total protein staining solution, the protein concentration was used for normalizing the concentration of the proteins of interest. After the FastGreen staining the dye was removed with FastGreen washing buffer, ethanol and water washes and blocked afterwards with 5% BSA for 1 hour. The blocked membrane was incubated overnight in the APP antibody in 5% BSA at 4°C (Table 7 List of primary antibodies). After incubation membrane was washed 3 times with TBS-T and incubated in the fluorescent secondary antibody in 5% BSA for 1 hour at room temperature (Table 8 List (Table 10 List of cell culture materials). Isolated microglia cells were used for two approaches, for the ex vivo phagocytosis assay and for treating cells with Aβ. For evaluating the plaque clearance efficiency microglia (from McKO and Cre) were seeded on 10µm thick fresh frozen 5xFAD sections that were pre-treated with anti-amyloid antibody to prime microglia. Microglia were kept in culture for 5 days for clearing plaques from amyloid burdened brain and efficiency was evaluated by fluorescent staining and microscopy of microglia and plaques. The reduction of plaques comparing to the consecutive slide that was incubated without microglia displays the clearance efficiency (Protocol was kindly provided by Dr.

Sabina Tahirovic).
A second experiment using the primary microglia isolated from McKO and Cre is the in vitro treatment with synthetic Aβ. The Aβ was diluted to 100µM and incubated at 37°C for 2 days to build pre-fibrils and treat microglia with 10pM. After 3 incubation cells were lysed for mRNA isolation and expression analysis (Table 23 RNA isolation of cells in RLT buffer,  3 Results

Neuronal cholesterol synthesis does not alter Aβ generation and deposition
In vitro reduced neuronal cholesterol levels inhibited amyloidogenic APP processing (Ehehalt et al., 2003) and decreased generation of Aβ (Simons et al., 1998). For evaluating the role of neuronal cholesterol synthesis on amyloidosis in vivo we generated mice lacking Fdft1 in neurons driven by the CamKIIα-Cre line (Funfschilling et al., 2012;Minichiello et al., 1999). Recombination of projections neurons in the cortex and hippocampus starts postnatally (Dragatsis & Zeitlin, 2000). Mutants used for this study harbored the genotype CamKIIα-Cre Fdft1 fl/fl 5xFAD tg (called NcKO-AD) and were compared to age and sex matched controls with the genotype Fdft1 fl/fl 5xFAD tg (NCtrl-AD). Surprisingly, the loss of cholesterol synthesis in neurons did not change plaque load at 3.5 months of age as quantified in 2D slices and fluorescent labeling of plaques in the subiculum and in 3D light-sheet microscopic analysis of the cortex and hippocampus (Figure   6a-c). Due to the Thy1 promotor of the 5xFAD model hAPP is restricted to neurons. In vivo neuronal cholesterol synthesis did not affect neuronal APP processing. Immunoblot analysis using the anti-APP antibody A8717 binding to the carboxyterminal fragment of APP uncovered no alterations of full-length APP and C-terminal fragments ( Figure 6 d-e). In addition to the steadystate APP processing and amyloid deposition, the glial response to plaques was similar in mutants and control. Summarizing the observations in NcKO-AD mice, ablation of neuronal cholesterol synthesis had no effect on amyloid plaque density with the applied methods.

Amyloid burden is reduced when astrocytes lack sterol synthesis
In response to pathological conditions astrocytes undergo morphological changes and become reactive. The astrogliosis is detectable in post-mortem brains as well as in mouse models of AD and can be correlated to dementia (Rodríguez et al., 2009). Astrocytes are a central player of brain cholesterol homeostasis, because astrocytes are the main producers of cholesterol in the brain and supply other brain cells via ApoE. To evaluate the impact of the astrocyte-specific cholesterol synthesis on APP processing, Aβ generation and glial response, mice lacking FDFT1 in astrocytes using Aldh1l1-CreERT2 (Winchenbach et al., 2016) were created and combined with the 5xFAD mouse model, here referred to AcKO-AD. The generated AcKO-AD (Aldh1l1-CreERT2 Fdft1 fl/fl 5xFAD tg ) were compared to age-and sex-matched controls further called ACtrl-AD (Fdft1 fl/fl 5xFAD tg ). The recombination of astrocytes was induced by applying tamoxifen at the age of 5 weeks. The plaque burden was investigated by in toto brain imaging of congo red+ stained plaques (Figure 7a). Reduced plaque numbers were found in the hippocampus and a tendency of reduction as well in the cortex at the age of 3.5 month (Figure 7b). Next the effect of the astrocyte specific loss of sterol production on tissue level was analyzed by quantitative RT-PCR (Figure 7c  Quantification of 3D plaque number in the cortical and hippocampal area of one hemisphere at 3.5 months. Circles represent single animals (n=3). Asterisks represent significant differences with *p < 0.05, **p < 0.01, ****p < 0.0001 (unpaired Student´s t-test). (C-E) Expression analysis of cortical tissue of 3.5-month-old mice (n=5) depicting glial changes, adaptions of the cholesterol synthesis and lipid transport, alterations of the amyloid regulation to the lack of sterol synthesis in astrocytes. Heatmap represents fold changes normalized the mean of to ACtrl-AD expression levels.
Asterisks represent significant differences with *p < 0.05, **p < 0.01, ****p < 0.0001 (unpaired Student´s t-test). (F) Immunoblots of full-length APP and the carboxyterminal fragments in cortical tissue from 3.5-month-old mice is presented (Immunoblot was performed together with Hoang Duy Nguyen, BA student). The protein levels are was downregulated in the mutant cortex. Additional to that, Trem2 was shown to modulate lipid metabolism in microglia (Nugent et al., 2020) (Figure 8f). This alteration can affect the microglia amyloid response and thereby the plaque burden. One cause of the elevated plaque burden can be higher Aβ production. The full-length APP and the processed c-terminal APP fragments can be depicted by the protein abundance in the membrane-bound fractions. The fAPP and CTFs were visualized using the c-terminal APP antibody A8717 and were normalized to the total protein content stained with FastGreen solution (Figure 8g). CTFs and fAPP of the microglia specific cholesterol synthesis depleted mice were comparable to the controls suggesting no changes in the processing of the amyloid precursor protein (Figure 8h). To conclude, highly elevated plaque density in 5xFAD mice lacking sterol synthesis in microglia were independent of APP processing changes but are likely the result of altered glial activity, plaque clearance or plaque building. (n=5-6) Asterisks represent significant differences with *p < 0.05, **p < 0.01, ****p < 0.0001 (unpaired Student´s t-test).

Morphological alterations of plaque associated microglia
Microglia morphology changes with activation of the phagocytosing cells. A ramified morphology is linked to a homeostatic state. Reactive microglia are part of the healthy brain during the whole life to control debris removal, pathogen clearance, maintenance and immune surveillance functions, while an imbalance towards activated microglia is described in pathological conditions. Switches in microglia morphology, especially in the microglia cells surrounding plaques have been observed in human Alzheimer´s patients and mouse models of AD and have been referred to "corralling" microglia (Hammond et al., 2019;Keren-Shaul et al., 2017;Krasemann et al., 2017).
The microglial activation state of the plaque-corralling microglia was shown to facilitate plaque modulation and Aβ removal (Huang et al., 2021;Kulkarni et al., 2022). Analysis of the microglia number and corralling phenotype around plaques in the cortex was implemented using two separate approaches. Using concentric circles around plaques to analyze the area covered by surrounding microglia pointed towards a higher coverage of plaques by sterol synthesis-deficient microglia (Figure 9a). The IBA1-positive area around plaques was enhanced in all sizes of plaques therefore suggesting a size independent but genotype-based alteration (Figure 9b).
Using a microglia and plaque co-staining microglia numbers with cellular contact to a plaque were analyzed (Figure 9c). Sterol-synthesis deficient microglia clustered more extensively around the plaques (Figure 9d). The enhanced plaque load and elevated number of microglia surrounding the plaques suggests a higher ratio of plaque-associated microglia to parenchymal microglia. For uncovering the transcriptional signal of the "hypercorralling" microglia in comparison to microglia of control mice and if the activation profile is altered, isolated microglia were analyzed by bulk seq transcriptional analysis.  (Figure 10b). In the volcano plots the distribution of up-and downregulated genes was visualized. In the ten most upregulated genes of Cre-AD and McKO-AD, shown in the first volcano plot, five genes were similar in both conditions, lipoprotein lipase (Lpl), brain-specific angiogenesis inhibitor 1-associated protein 2-like protein 1 (Baiap2l2), cholesterol 25-hydroxylase (Ch25h), cystatin F (Cst7) and glycoprotein Nmb (Gnmb). Lpl, Cst7 and Ch25h have been associated with the DAM signature that is common in AD (Figure 10c).
The second volcano plot is depicting the differentially regulated genes between McKO-AD vs Cre.
Surprisingly, the overall significance of upregulation was way higher in Cre-AD than McKO-AD and the previously mentioned DAM genes were stronger regulated in Cre-AD microglia. In the right volcano plot the significantly regulated genes comparing McKO-AD with Cre-AD were clearly reduced. In summary, microglia lacking cholesterol synthesis detect senile plaques and react to the amyloid burden through an upregulation of activation markers known for their disease association. On the other hand, the upregulation is considerably lower than in control microglia, suggesting only a partially activated status of all microglia or might propose that a lower number of microglia initiate the DAM signature. 3.6 DAM signature is reduced in sterol synthesis deficient microglia The RNA-seq analysis gave profound insights into the importance of microglial cholesterol synthesis in the presence of amyloid. Selected pathways of the transcriptomic analysis were displayed and examined in more detail. Effects on cholesterol synthesis and lipid handling were analyzed as well as extensive effect on microglia activation. The basal knockout of Fdft1 in microglia only slightly affected the cholesterol homeostasis and the lipid transport. This was expected due to the fact that in the healthy brain microglia barely contribute to the cholesterol production ( Figure 11 a-b). Furthermore, the microglia signature was neither changed in homeostatic genes nor in activation genes. In the isolated microglia bulk sequencing data set from pure AD mice, cholesterol synthesis genes were upregulated and the lipid transport was enhanced as well. This suggested enhanced cholesterol turnover in microglia. These gene sets were not upregulated any longer in McKO-AD mice. Furthermore, in ordinary 5xFAD microglia homeostatic markers were downregulated and the widely studied activation genes of microglia in amyloid pathology were significantly upregulated. This transition from homeostatic microglia to the activated DAM signature was less prominent in microglia lacking sterol synthesis (Figure 11c). This finding indicated that lack of microglial sterol synthesis affected the inflammatory response to amyloidosis. Microglia with altered activation and inflammatory response could suggest a change in phagocytosis and thereby plaque clearance. Lipid transport, lipid storage and excretion of cholesterol is depicted. (C) Gene expression changes of microglia activation status was categorized in activated and homeostatic marker genes. (n=3-5) Asterisks represent significant differences with *p < 0.05, **p < 0.01, ****p < 0.0001 (unpaired Student´s t-test).

The role of microglia on plaque clearance
Amyloid plaque removal is highly dependent on the functional detection of plaques by microglia, their motility, and clearance efficiency. It has been suggested that with disease progression microglia cells become less efficient in amyloid phagocytosis and thereby plaque clearance . To investigate whether microglia plaque clearance depend on cholesterol synthesis, an ex vivo phagocytosis assay was performed. Brain sections from a one-year old plaque loaded 5xFAD brain were incubated with an anti-Aβ-antibody to prime microglia for amyloid and transferred into medium. CD11b+ cells were acutely isolated from mutant (McKO) and control (Cre) mice, and seeded on the brain slices (Figure 12a). After 5 days in culture slices were fixed and stained for plaques and microglia and plaque load was analyzed compared to the untreated consecutive slice (Figure 12b). Microglia displayed a partially ramified morphology and enriched around the plaques in both conditions (Figure 12c). The density of plaques was reduced in both conditions, to a similar extent (Figure 12d). These data suggest that, both microglia genotypes can clear amyloid plaques under ex vivo conditions. It is possible that the clearance is comparable because the microglia cells were purified from healthy brains and were only exposed to amyloid for the time of the assay. How amyloid was ingested and how it is processed cannot be answered by this assay. The visualization within cells is difficult due to the background signal originating from the brain tissue. Each circle is a separate assay with microglia isolated from 2 different mice per group. Asterisks represent significant differences with *p < 0.05, **p < 0.01, ****p < 0.0001 (unpaired Student´s t-test).

Aβ accumulation is altered in cholesterol synthesis deficient macrophages
To evaluate the ingestion of amyloid and its processing the amount of intracellular fluorescently labeled Aβ (fluoro Aβ) after 24 hours was depicted. Therefore, primary bone marrow derived macrophages (BMDMs) from McKO and Cre mice taken in culture and the knockout was induced by in vitro tamoxifen for 10 days. Afterwards, fluorescently labeled Aβ (fluoro Aβ) was added for 24 hours to assess the ingestion and the intracellular handling. The cells were seeded on cover slips and treated with fluoro Aβ for 24 hours, fixed, stained and analyzed using fluorescent imaging (Figure 13a-b). Surprisingly, the treatment with fluoro Aβ after 24 hours resulted in different appearances of ingested amyloid. These appearances of amyloid were categorized by the fluorescent signals into 1) dense aggregated protein further referred to "accumulated" and 2) more evenly distributed in the cell soma and smaller in spot size further called "spotty" ( Figure   13c). For evaluation of the Aβ intake the total amount of internalized Aβ was normalized to Iba1+ macrophages, the intake was similar in both genotypes. However, the fraction with accumulated appearance was increased in the control group compared to McKO, whereas the spotty particles were more frequent in mutant macrophages. The lysosome-associated membrane protein 1 (LAMP1) intensity normalized to the Iba1+ cell area was increased in sterol synthesis deficient cells (Figure 13d). Lamp1 is a transmembrane protein mainly localized to lysosomal membranes.
Lysosomes serve as hubs for degrading components by autophagy or endocytosis. The change of LAMP1 signal after Aβ internalization could either suggest a processing in different cellular compartments or different stages of digestion.

Amyloid phagocytosis changes microglial expression profile
Investigating microglial in vivo phagocytosis of Aβ and the differentiation between phagocytosing and non-phagocytosing microglia could unravel a modification in amyloid handling. Altered proportions of phagocytosing and non-phagocytosing or different expressional regulations between these groups could expose an adaptation of microglia lacking sterol synthesis. A variety of studies showed different properties of microglia with amyloid inclusions and non-laden microglia. More specifically, morphological alterations, inflammatory and phagocytic properties as well as alterations in channel activity are some changes between these distinct microglia populations (Grubman et al., 2021;Hemonnot-Girard et al., 2021).  Asterisks represent significant differences with *p < 0.05, **p < 0.01, ****p < 0.0001 (unpaired Student´s t-test). Previously, cholesterol-rich membrane microdomains of neurons were associated with Aβ generation due to the subcellular location of the β-and γ-secretase and APP to these domains.
Several in vitro studies utilizing neuronal cultures established a link between cholesterol metabolism and APP processing. More specifically, a shift of APP processing towards the amyloidogenic direction and Aβ secretion was connected to enhanced cholesterol levels (Bodovitz & Klein, 1996;Marquer et al., 2011). Conversely, a recent study revealed reduced interactions between BACE1 and APP in human iPSC-derived neurons under low cholesterol levels, achieved by statin treatment, in vivo versus ex vivo studies (Langness et al., 2021).
However, they modified cholesterol by applying statins, these drugs inhibit sterol synthesis at the level of HMG CoA reductase. My approach of genetically targeting Fdft1, cholesterol synthesis is diminished at a later step and thereby no other pathways regulated by intermediates of the synthesis pathway are affected. Surprisingly, my data presented in this thesis shows that the lack of cholesterol synthesis in neurons did not affect the plaque burden in the 5xFAD mouse model.
In the disease establishing phase, amyloidosis is progressing in the hippocampus and first plaques establish in the cortex layer 5. Aβ generation is dependent on APP processing which occurs primarily in neurons in this AD mouse model. In mutant brain lysates neither full length APP nor c-terminal fragments were modified. This is in agreement with unaltered plaque burden in these animals. How can these findings be explained in comparison to earlier studies? First, previous studies were restricted to investigating the effect of changes in cholesterol metabolism in vitro. Second, neuronal cholesterol levels might not be depleted by ablation of cholesterol synthesis in neurons in our in vivo approach, as other brain cells could potentially supply sufficient amount of cholesterol to neurons via horizontal transfer. The latter possibility has to be further investigated in the future.
An alternative source of cholesterol for neurons could be astrocytes that have been previously shown to be the main producers of cholesterol and shuttle lipids via lipoproteins (Fünfschilling et al., 2007;Quan et al., 2003). Likewise, ApoE is the most frequent carrier of lipids and cholesterol in the cerebrospinal fluid and is almost exclusively produced by astrocytes. The strongest genetic risk factor for LOAD is the ApoE4 allele, indicating a connection between astrocytes, ApoE and the generation of Aβ by determining the cholesterol content of neurons. Corresponding to this possible interaction, plaque quantity was significantly reduced in 5xFAD mice with astrocytes unable to generate cholesterol (Mahan et al., 2022). However, APP processing in cortical tissue seemed unaffected, suggesting an effect on amyloidosis independent of Aβ generation. Tissue expression analysis of mice with deactivated cholesterol synthesis in astrocytes revealed enhanced levels of Abca1, a protein relevant for cholesterol export and lipidation of APOE. The robustly upregulated Ldlr is a receptor for APOE and mediates cholesterol uptake. Ldlr also mediates Aβ clearance in astrocytes (Kim et al., 2009) suggesting astrocytic cholesterol synthesis affects APOE mediated or direct effects on Aβ drainage. Additionally, Ece1, an Aβ-degrading enzyme, is highly enriched in the cortex of mice deficient in astrocyte sterol synthesis. In addition to the upregulation of microglia activation genes e.g. Trem2 and Tyrobp, these findings suggest an effect on Aβ clearance rather than Aβ generation in animals with astrocytes deficient in cholesterol synthesis. Taken together, cholesterol deficient astrocytes do not affect APP processing but rather could affect plaque clearance and could thereby alter disease progression.
The most prominent effect on the plaque burden was observed when microglia cells lacked the ability to synthesize cholesterol. Here, plaque burden was massively exacerbated. Many risk genes for LOAD, such as Trem2, CD33, and Ms4a are expressed by microglia and are associated with microgliosis. While microglia sustain brain homeostasis in health, in AD the homeostatic gene signature shifts to an activated condition and may reflect a highly phagocytic state. In the early disease stage, the activation of microglia could be part of a beneficial attempt to resolve the inflammation causing agent, with disease progression microglia might detrimentally fail in handling amyloid deposits. Considering the extremely elevated plaque density in microglia deficient in sterol synthesis, the expression pattern was only minimally changed. These mainly physiologic expression levels could be assigned to the low specificity of whole tissue analysis.
Intriguingly in animals with microglia deficient in cholesterol synthesis, Trem 2 expression, was slightly downregulated. TREM2 is known to enhance phagocytosis and microglia activation (Kleinberger et al., 2014). In addition to these functions the receptor affects inflammatory signaling (Jay et al., 2015). The reduced expression could contribute to less Aβ handling by microglia cells.
For further conclusions cell specific expression analysis of microglia is needed. As in the other cell-specific sterol synthesis mutants, APP processing was not dependent on cholesterol production in microglia.
Altogether, glial cholesterol production but not neuronal production seems to be highly involved in controlling amyloid plaque burden with astrocytic synthesis favoring and microglial synthesis diminishing plaque deposition. Importantly, both effects seem not be mediated by direct changes to APP metabolism in neurons. This suggests a mechanism of action that involves variations in plaque deposition or Aβ clearance. 4.2 "Hypercorralling"microglia lacking sterol synthesis organize around senile plaques Plaque clearance is dependent on microglial detection and phagocytosis of amyloid fibrils (Huang et al., 2021). Plaque-associated microglia present with morphological changes that resemble a classical activated phenotype. The cells surrounding senile amyloid plaques are thought to limit the plaque expansion by forming a barrier for brain protection (Condello et al., 2015). On the other hand, some studies suggest that microglia cells are involved in initial plaque seeding. Cholesterol synthesis affected the interaction between microglia and plaques: more amoeboid microglia surrounded the plaques. Supposing that microglia shield the brain parenchyma to protect unaffected brain areas from the senile plaques, why should more microglia be detrimental? The efficacy of plaque clearance could be diminished by lack of sterol synthesis, by possibly altering the phagocytosis efficiency or the handling of ingested amyloid. Eventually this could lead to an increased recruitment of microglia to plaques, explaining the hypercorraling.

Microglia DAM signature is attenuated through sterol deficiency
Corresponding to the different tasks of microglia during development and the entire life of a mouse, microglia are a heterogeneous population of cells with their transcriptomic profile depending on their function. Similarities between populations arising in the early development and microglia in neurodegenerative disease were reported (Hammond et al., 2019). Microglia in neurodegenerative diseases and development, present with an inflammatory signature which in the development could relate to the phagocytosis of synapses and cell debris and in disease the activation through pathological protein accumulations (Salter & Stevens, 2017;Zia et al., 2020).
Microglia from AD mice are characterized by downregulating of homeostatic markers and the upregulation of activation profiles. (Keren-Shaul et al., 2017;Wang et al., 2015). However, the role of the activated disease associated microglia is not completely understood. For instance, in Trem2 KO mice plaque recognition and corralling by microglia is reduced along with the expression of DAM markers is diminished (Ulland & Colonna, 2018). This raises the question whether some microglia activation patterns are beneficial for the disease, by actively clearing amyloid from senile plaques (Butovsky & Weiner, 2018).
The RNA sequencing data of isolated microglia from AD mice (Cre-AD) displayed the expected signature of microglia in amyloid disease. Lpl, Cst7 and the Integrin Subunit Alpha X (Itgax) were highly upregulated compared to microglia from healthy controls. The overall inflammatory response is elevated compared to microglia from mice without 5xFAD, this displayed the activated status of microglia. In contrast to the extensive transcriptional changes of pure AD microglia, microglia lacking the ability to synthesize sterol failed to upregulate these genes to the same extent. The deeper analysis uncovered several pathways that depend on cholesterol synthesis during amyloidosis. Expression levels of cholesterol synthesis depleted microglia in AD a rather low level more comparable to the profile of microglia in a healthy microenvironment. Genes implicated in lipid transport were highly enriched in Cre-AD microglia, ranging from lipid export genes and lipid receptors to enzymes converting cholesterol to oxysterols. This indicates not just an accelerated sterol synthesis but an enhanced lipid handling and transfer in 5xFAD mice. In the McKO-AD expression profile, lipid transport transcripts were at normal levels compared to the microglia with functional sterol synthesis. Genes involved in making cholesterol more accessible were regulated, oxysterol conversion (Cyp46a1, Cyp7b1, Cyp27a1) and ester formation (Lcat) genes were reduced and in addition to that neutral cholesterol ester hydrolase 1 (Nceh1) was upregulated. Not all lipid related genes were regulated in a contrary manner, as genes contributing to lipid droplet stabilization (Plin2, Plin3) were enhanced similar to Cre-AD thereby suggesting a certain degree of preservation of lipid droplet formation capability independent of sterol synthesis.
Lipid droplets are storage organelles enhanced in neurodegenerative disease and have been shown to be upregulated by DAM genes in AD (Hamilton et al., 2015;Moulton et al., 2021;Ralhan et al., 2021). DAM genes were only partially induced in McKO-AD mice. The classical marker genes were highly expressed in the control microglia as it has been shown in other studies of AD mouse models and human AD (Deczkowska et al., 2018;Sobue et al., 2021;Wang, 2021). While in AD microglia homeostatic markers were downregulated, in the sterol synthesis deficient expression levels were comparable with non-AD microglia. These changes in the classical induction of microglia activation in response to amyloid plaques could imply a lower number of microglia upregulating DAM genes. Furthermore, it could be a consequence of a blocked transition of microglia and thereby only a partially activated status. Importantly, the sterol synthesis mutants without amyloid pathology were similar compared to Cre control microglia. Only minor changes in healthy conditions could hint towards the importance of microglial cholesterol synthesis only in pathology. This was expected because microglia contribute to cholesterol metabolism only to a low level.
In summary, sterol synthesis deficiency in microglia altered the induction of lipid transport in AD and modified the activation status of microglia. In conjunction with the fact that microglia lacking sterol synthesis were able to recognize and migrate to the plaque, the question is if the mutant microglia are able to phagocytose and digest Aβ? 4.4 Sterol deficient microglia phagocytose plaques but the Aβ digestion is altered Activated microglia are linked to amyloid clearance (Bacskai et al., 2001;Wyss-Coray et al., 2001). With disease progression and increasing plaque load and thereby prolonged activation, microglia slow down their phagocytic activity (Hu et al., 2021). Exhausted or poorly functioning microglia give rise to amyloid plaques (Flanary et al., 2007;Flanary & Streit, 2004). As a consequence, the question arises whether microglia with reduced DAM signature show phagocytosis defects. I addressed this question in three different experimental paradigms.
In the functional assessment of plaque clearance in the ex vivo phagocytosis assay control microglia as well as mutant microglia reduced the plaque number to a similar degree. This result can be a confirmation of the ability of the sterol synthesis deficient microglia to phagocytose plaques or these microglia have never encountered amyloid-β in vivo. Although this finding shows the capability of McKO microglia to reduce plaques at a similar rate like Cre controls, this might not translate well to the in vivo situation. Within the AD brain parenchyma microglia could be fatigued by longer exposure to Aβ in vivo, overloaded with fibrils, or gathering problems regarding the processing of the internalized amyloid-β.
Considering the fact, that imaging approaches to unravel the internalization of amyloid into the cells is difficult when seeded on a brain slice, in vitro experiments were performed. The uptake of synthetic Aβ was investigated using bone marrow derived macrophages. Despite an overall similar uptake of amyloid-β in Cre and McKO cells, the internalization pattern was different. In Cre controls internalized Aβ accumulated into dense amyloid aggregations whereas in the sterol synthesis deficient macrophages a spotty signal distributed over the entire cell body was observed. In addition, the enhanced LAMP1 intensity in the mutant cells suggests an ongoing digestion or an internalization in different compartments. It was shown that LAMPs sustain essential function in the endocytic pathway, especially the successful phagosome-lysosome fusion (Huynh et al., 2007). In AD, LAMP1 positive microglia predominantly localized around senile plaques and may either indicate Aβ removal or deposition (Barrachina et al., 2006).  (Dani et al., 2018;Sobue et al., 2021). Further studies could correlate anti-inflammatory treatment and thereby more homeostatic microglia and elevated plaque levels (Jantzen et al., 2002;Wyss-Coray et al., 2001). Taking together all these findings indicates a prominent role of microglial activation for amyloid plaque control which in turn seems to be highly dependent on functional cholesterol synthesis. In addition, the disease associated microglia signature is regulated by sterol metabolism in microglia. Interestingly, this signature can be traced back to the substantial non-phagocytosing microglia population. Due to the fact that mutant microglia are capable of phagocytosing amyloid, the disturbance of sterol homeostasis might affect the intracellular handling of amyloid. Considering the evidence that depleting microglia before plaque seeding, eradicates plaque formation, microglia are essential for plaque deposition. Could cholesterol synthesis control the plaque deposition by modifying amyloid processing in the different compartments within microglia and thereby seeding of new plaques?

Translation into human AD
Uncovering the relevance of sterol metabolism in human Alzheimer´s disease is more complex.
In comparison to the 5xFAD model, including an assured disease onset due to the inserted mutations most of the human AD cases originate spontaneous.
Additionally, the complexity of the cholesterol metabolism is even higher considering that the horizontal transfer of cholesterol between cells is mainly achieved by ApoE, which is polymorphic in humans. The different alleles are structurally different and harbor their individual lipid binding affinities. Surprisingly, the residues responsible for lipid binding are not conserved between species suggesting distinguished lipid binding properties in various species (Frieden et al., 2017).
To perform normal functions, ApoE has to be adequately lipidated. The AD risk gene ApoE4 is less lipidated compared to ApoE2, whereas ApoE2 is more capable of supporting cholesterol efflux. (Lanfranco et al., 2020). Could the alteration of cholesterol synthesis affect the lipidation status of ApoE and thereby be a possible intervention to reduce plaque burden? In human AD a further complication can be expected by the appearance of hyperphosphorylated tau tangles. The translation of effects from mice to human is always uncertain but regarding the simplified pathology in mouse models compared to the complex pathology in humans the impact on phosphorylated tau as further disease aspect is even more difficult to assess.
Targeting cholesterol synthesis specific in a certain cell type in humans is another challenge. The common way of lowering cholesterol synthesis, is the application of statins that block sterol synthesis at the level of HMG CoA reductase. The overall intervention in cholesterol metabolism is not advisable given the fact that alterations in the cholesterol synthesis of each cell type has different consequences. 6 Appendix 6.1 List of tables  Table 2   Miso and Heiko for making the great Light Microscopy Facility available and for support with data generation and analysis.
Finally, I would like to thank my family and friends for their loving support and for enabling me to finish my PhD-studies.
Thanks to Sotirios Pavlidis and Enzymel for supporting me in the last weeks and for helping me with the graphics.
Thanks to Jana Vieten for making the Bachelor easier and more fun and for your friendship since.
An enormous thank you to Miriam Steimmig, my best friend and for always being there, in good and bad times.
The biggest thank you goes to my mother Andrea, my father Udo and my sister Janina for being my greatest supporter and making me the person I am today.