Is Alzheimer’s Disease a Liver Disease of the Brain?

Production

Three loci that modify Aβ accumulation and deposition in the brains of a mouse model of AD have been previously described: amyloid-β protein precursor (AβPP), presenilin 1 (PSEN1), and presenilin 2 (PSEN2). Mutations in these loci result in abnormal processing of AβPP and increased generation of Aβ₄₂, which aggregates as insoluble β-pleated sheets [5]. One of these, the PSEN2 gene encoding presenilin 2, a component of the γ-secretase activity is responsible for generating Aβ by proteolysis. Activity of mouse PSEN2, as measured by levels of mRNA accumulation, has unexpectedly been shown to be heritable in the liver, but not the brain, suggesting that the liver is the origin of brain Aβ deposits [6]. Sutcliffe and colleagues showed that peripheral administration of the anticancer drug, Imatinib, commonly known as Gleevac (a specific inhibitor of a number of tyrosine kinase enzymes), resulted in a 50% reduction in plasma and brain Aβ levels. As Imatinib does not cross the blood-brain barrier (BBB), this provided evidence that Aβ produced peripherally was contributing to brain Aβ. Imatinib lowers Aβ levels through indirect inhibition of γ-secretase activity [7]. Imatinib also renders AβPP less susceptible to proteolysis by β-secretase (BACE) without inhibiting BACE enzymatic activity or the processing of other BACE substrates [8]. However, plasma levels of Aβ₄₂ did not change in patients with chronic myeloid leukemia treated with Imatinib [9], bringing into question whether it may have a role in individuals with Aβ dyshomeostasis.

The important study by Sutcliffe et al. [6] built on earlier findings that peripherally derived Aβ₄₀ and Aβ₄₂ can cross the BBB and that circulating Aβ could thus contribute to neurotoxicity [10, 11]. The implication that Aβ homeostasis was an interconnected system involving the liver and BBB to regulate brain Aβ was discussed at the time [12] but needs re-emphasizing now, with the addition of new data from the last decade.

The AβPP gene is located at chromosome 21q21 and individuals with Down’s syndrome, which results from trisomy of chromosome 21, develop AD neuropathology (reviewed in [13]). Individuals with partial trisomy of chromosome 21, which does not include the AβPP gene, fail to develop AD neuropathology, demonstrating that excess Aβ production is sufficient to cause AD (Fig. 1). In addition, recent tissue-specific metabolomic analysis revealed that the liver was the earliest affected organ in AβPP/PS1 mice during amyloid pathology progression [14]. Genetic variants affecting AβPP and Aβ processing are associated not only with early-onset autosomal dominant AD but also with LOAD [15].

OPEN IN VIEWER

Fig.1

Some factors altering the balance between amyloid Aβ production and clearance leading to dyshomeostasis. Peripheral clearance can remove 40–50% of Aβ burden in the brain [25, 26]. Aβ, amyloid-β; AβPP, amyloid-β protein precursor, LRP1, low-density lipoprotein receptor-related peptide 1; LDLR, low-density-lipoprotein receptor; P-gp, P-glycoprotein.

Generation of the AβPP by β- and γ-secretases occurs in early endosomes, followed by routing of Aβ to multivesicular bodies in HeLa and N2a cells and subsequently, a minute fraction of Aβ peptides can be secreted from the cells in association with exosomes [16]. Exosomes are extracellular membrane vesicles actively secreted by cells into the circulation that are involved in cell-to-cell communication in normal homeostasis. They can carry cargo across the BBB [17] and are carriers of Aβ in AD [18]. The BBB keeps neurotoxic plasma derived components out of the central nervous system, but recent studies suggest an early BBB breakdown in AD [19]. The receptor for advanced glycation end products (RAGE) has been implicated in the transport of Aβ across the BBB [11, 20]. Inhibiting RAGE has been shown to have significant therapeutic benefit in AD models [21]. High dietary advanced glycation end products have also been found to accelerate Aβ deposition in an AD murine model mediated by overexpression of RAGE [22].

Clearance

LOAD, the common form of AD, is characterized by an overall impairment in Aβ clearance [23]. Aβ clearance is a complex event that involves more than neurons and microglia [24]. Peripheral clearance of brain-derived Aβ exists physiologically. Efflux of Aβ to peripheral blood accounts for 50% of total brain Aβ clearance in humans [25], suggesting that the physiological Aβ clearance capacity of the peripheral system provides an important mechanism against Aβ accumulation in the brain. Hence, dysfunction of peripheral Aβ clearance may contribute to the development of AD.

In a murine model of AD using parabiosis (the anatomical joining of two animals for physiological research), it has been shown that parabiosis reduces brain Aβ burden through clearance by peripheral tissues and organs, including the liver [26]. In this model, it was calculated that the periphery can remove 40% of Aβ burden in the brain, similar to other estimates of the importance of peripheral physiological Aβ clearance [25]. Studies in another model system found a short half-life of both Aβ₄₀ and Aβ₄₂ after injection of only 2.5–3.0 min with the liver being the major organ responsible for plasma clearance, accounting for > 60% of the peptide uptake. Indeed, it was suggested that the capability of the liver to take-up, catabolize, and excrete large doses of Aβ, several orders of magnitude above its physiologic concentration may explain not only the femtomolar plasma levels of Aβ, but also the minor fluctuation observed with age and disease stages [27].

There are several potential pathways for the efflux of brain Aβ into the periphery. These include clearance via the glial-lymphatic (glymphatic) system [28] and transport across the BBB, mediated by low-density lipoprotein receptor-related peptide 1 (LRP1) [29, 30] and the low-density-lipoprotein receptor (LDLR). In animal models of AD, lack of LDLR enhances amyloid deposition in the brain [31], while LDLR overexpression increases the rate that Aβ enters the blood from the brain [32].

In plasma, a soluble form of LRP1 (sLRP1) is the major transport protein for peripheral Aβ [33]. Improving the binding of Aβ to a sLRP1 variant has been shown to increase the efficiency of Aβ clearance [34], suggesting that this binding prevents re-entry of Aβ to the brain. Impaired sLRP1 binding of plasma Aβ has also been reported to be an early biomarker for mild cognitive impairment preceding AD [35]. Brain-derived Aβ in the arterial blood is cleared physiologically when it goes through the capillary bed of the peripheral organs and tissues, including the liver [26]. LRP1 is the major receptor responsible for the saturable uptake of plasma free Aβ₄₀ by the liver [36]. The remarkable therapeutic effect of the ayurvedic agent, Withania somnifera (also known as poison gooseberry or winter cherry, from the nightshade family), mediated through upregulation of liver LRP indicates that targeting the periphery offers a unique mechanism for Aβ clearance as this therapy reverses the behavioral deficits and pathology seen in AD models [37]. Atorvastatin has also been shown to upregulate liver LRP1 and this effect is mediated by the sterol response element-binding protein-2 (SREBP-2) in vitro and in vivo [38]. Statins can reduce AD risk and the effect varies with statin molecule, sex, and race/ethnicity [39].

Transthyretin, a protein involved in the transport of thyroid hormones and retinol, has been proposed as a protective protein in AD [40]. Transthyretin acts as a carrier of Aβ at the BBB and liver using LRP1 [41].

Thus, LRP1 is involved in three stages of the homeostatic control of Aβ clearance including 1) cell-surface LRP1 at the BBB and cerebrovascular cells, mediating brain-to-blood Aβ clearance, 2) circulating LRP1 providing a key endogenous peripheral ‘sink’ activity for plasma Aβ which prevents free Aβ access to the brain [42], and 3) LRP1 in the liver mediating systemic Aβ clearance [43] (Fig. 2).

OPEN IN VIEWER

Fig.2

Schematic representation of Alzheimer’s disease homeostasis showing amyloid-β (Aβ) production from amyloid-β protein precursor (AβPP) in the liver, dysregulated influx/efflux across blood-brain barrier (BBB) (1), transport in serum via soluble LRP1 (2) and exosomes [61] and saturable uptake of Aβ by liver via low-density lipoprotein receptor-related peptide 1 (LRP1) and low-density-lipoprotein receptor (LDLR) (3) with subsequent biliary clearance. RAGE, receptor for advanced glycation end products.

In a human study using amyloid PET with [¹¹C]PiB, the C667T polymorphism of the LRP1 gene has been shown to be moderately, but significantly associated with global and regional amyloid deposition [44]. This finding is compatible with the Aβ hypothesis that impaired amyloid clearance contributes to amyloid deposition in LOAD.

LRP1 is capable of recognizing a wide variety of structurally-distinct ligands; Apolipoprotein E (ApoE) is one. ApoE polymorphic alleles are major genetic determinants of AD. Individuals carrying the epsilon (ɛ) 4 allele (APOE ɛ4) are at increased risk of AD, compared to those carrying the more common ɛ3 allele, whereas the ɛ2 allele decreases risk. Thus, at age 85 years, the lifetime risk of AD without reference to APOE genotype is 11% in males and 14% in females, compared to 51% for male APOE4 homozygotes and 60% for female APOE4 homozygotes, consistent with autosomal co-dominant inheritance of a moderately penetrant gene variant [45]. In a murine model, Aβ was mainly sequestered in the liver and its peripheral clearance was by influenced by ApoE [46]. A number of subsequent studies suggest that ApoE4 inhibits Aβ clearance and/or is less efficient in mediating Aβ peripheral clearance compared with ApoE3 and ApoE2 [47] (reviewed in [48, 49]). A biologically inspired nanostructure, ApoE3-reconstituted high-density lipoprotein, with high binding affinity to Aβ, rescues memory loss of mice with AD by accelerating the clearance of Aβ [50].

ApoE expression is transcriptionally induced through the action of the nuclear receptors peroxisome proliferator-activated receptor gamma (PPARγ) and liver X receptor (LXR) in coordination with retinoid X receptors (RXRs).

In the liver, LRP1 functions in concert with LDLR in the clearance of ApoE-containing particles circulating in plasma [51]. Biliary clearance of Aβ is not only mediated by LRP1, but also by the drug efflux pump, P-glycoprotein encoded by ABCB1 gene [52]. P-glycoprotein dysfunction in BBB active efflux of xenobiotics has been shown by imaging studies in individuals with early AD [53]. This raises the possibility that common pharmacological inhibitors of P-glycoprotein, such as amiodarone, lansoprazole, omeprazole, and other proton-pump inhibitors, tamoxifen and verapamil [54], could impact on Aβ clearance. A recent in vitro study using synthetic fluorescein-labelled Aβ₄₀ and Aβ₄₂ spiked into human liver homogenates has shown that Aβ degradation rates are lower in AD-derived homogenates as compared with those from non-demented control subjects, even after accounting for the covariates of age, sex, and APOE genotype. The authors conclude that their results “support the possibility that impaired hepatic Aβ degradation could be a factor contributing to increased brain Aβ accumulation and AD” [55]. In addition, serum-based bile acid metabolites are associated with AD biomarkers, providing further evidence that bile acid pathways play a role in AD pathophysiology [56].

Plasma assays as screening tests for AD

Plasma Aβ levels tend to be nearer the lower limits of detection, but there is emerging consensus that “recent improvements in technologies to assess plasma levels of amyloid beta indicate that a single sample of blood could provide an accurate estimate of brain amyloid positivity” [57]. For example, measurement of plasma Aβ biomarkers by immunoprecipitation coupled with mass spectrometry has recently been shown to correlate with brain Aβ burden and levels of Aβ₄₂ in cerebrospinal fluid (CSF) [58]. The biomarkers measured were ratios of AβPP_669-711/Aβ₄₂ and Aβ₄₀/Aβ₄₂ and their composites, confirming that plasma Aβ might reflect brain amyloid deposition. Another group has shown that the secondary structure distribution of Aβ in blood plasma, measured by an immuno-infrared sensor, is an excellent biomarker for AD, reflecting the Aβ burden in the brain [59]. The performance of Elecsys (Roche Diagnostics, Basel, Switzerland) immunoassays to measure plasma Aβ₄₂ and Aβ₄₀ has also recently been shown to predict Aβ status in all stages of AD and their accuracy can be increased by analyzing APOE genotype [60]. A recently reported promising approach is measuring exosome-bound Aβ; this has shown to better reflect PET imaging of brain amyloid plaques than unbound or total circulating Aβ and hence enable early diagnosis and disease monitoring [61].

Circadian rhythm

A rapidly growing body of research suggests that disturbances in the circadian system precede the emergence of the characteristic cognitive and motor symptoms of AD [62]. Aggregation of Aβ into extracellular plaques in the brain likely begins 20 years before the onset of dementia. Aβ concentrations in both humans and mouse models show Aβ concentrations rise during wakefulness and fall during sleep, that is, an Aβ diurnal pattern. Studies on sleep raise the possibility that altering sleep quality might impact Aβ deposition and may also regulate the clearance of Aβ from the brain [63]. Indeed, sleep has been identified as a factor which alters the production and/or clearance of Aβ in stable isotope labelling kinetic (SILK) studies measuring Aβ turnover in blood and within the brain [64]. SILK studies have shown that the Aβ₄₂:Aβ₄₀ turnover rate positively correlates with amyloid plaque load and demonstrate that understanding Aβ dynamics in different compartments including CSF, blood, and brain tissue is crucial to improving therapy. Thus, studies showing significantly lower levels of CSF Aβ₄₂ in AD patients with more severe cognitive impairment [65] are unable to provide information on changes in Aβ turnover occurring either during the circadian cycle or during the progression of disease.

The clearance of amyloid from the brain during sleep is primarily via the glymphatic pathway [66, 67]. These lymphatic vessels exit the cranium along veins and arteries associated with the middle meningeal arteries, transporting waste via the deep cervical lymph nodes to the systemic circulation (reviewed in [68]). Initial functional studies in the sleeping brain have shown the importance of the glymphatic pathway in animal models. More recent research has demonstrated glymphatic efflux in patients with AD [69].

Proper functioning of the circadian system is determined by the orchestration of the suprachiasmatic nucleus in the hypothalamus and synchronized peripheral clocks in local tissues, including the liver. Mass spectrometry analyses of the mouse liver proteome has shown that many secreted proteins accumulate with a diurnal rhythm [70]. Circadian post-transcriptional and post-translational mechanisms play a key role in the temporal orchestration of liver-specific metabolic pathways [71]. Among those enriched in the liver cycling proteome are primary bile acid biosynthesis, bile secretion, protein processing in the ER, PPAR signaling pathway, and metabolism of xenobiotics. Hence poor sleep may be impacting on the hepatic production and biliary clearance of Aβ, as well as directly on the brain. The unfolded protein response (UPR) and circadian rhythm are intimately linked in the liver [72] and the UPR is emerging as a pharmaceutical candidate to combat neurodegenerative diseases [73]. It would hence be of interest to measure the diurnal pattern of plasma biomarkers of AD in relation to therapy.

Other nuclear receptor agonists

PPARγ may act as a master regulator of the transcription of several genes involved in LOAD pathogenesis [84]. PPARγ agonists such as the glitazone, pioglitazone, prescribed for the treatment of type 2 diabetes, promote amyloid clearance in animal models of AD [85]. A phase II study of pioglitazone in AD showed that it is safe and well tolerated and two large phase III trials are ongoing [86]. In patients with diabetes, pioglitazone treatment is a time- and dose-dependent protective factor against dementia [87]. Cilostazol enhances LRP1 expression in liver by activating PPARγ through the peroxisome proliferator response element in the LRP1 promoter [88]. In mice, combined PPARγ /LXR agonist treatment also reduces soluble and deposited forms of Aβ [89].

The RXR agonist, bexarotene (brand name: Targretin), which is approved by both the FDA and European Medicines Agency (EMA) for use in cutaneous T cell lymphoma, stimulates physiological Aβ clearance mechanisms, resulting in the rapid reversal of a broad range of Aβ-induced deficits in mouse models [90].

VDR - 1,25-dihydroxyvitamin D3(1,25(OH)2D3) treatment has been shown in vitro to enhance both Aβ₄₀ efflux across the BBB and hepatic uptake by HepG2 cells, accompanied by increased LRP1 expression. It was suggested that the effect was exerted via the nuclear VDR and could explain how 1,25(OH)2D3 exerts neuroprotection against AD [91].

Flavonoids

Genistein is an isoflavone derived from the dyer’s broom plant, Genista tinctoria, which activates PPARs. Treatment of an AD mouse model with genistein results in improvement in various parameters of cognition, associated with a lowering of Aβ levels in brain and in the number and the area of amyloid plaques [92]. The authors conclude that “our results strongly suggest that controlled clinical trials should be performed to test the effect of genistein as treatment of human AD”.

Newly identified flavonoids which selectively target another nuclear receptor/pathway, LXRβ have also been found to reduce total brain Aβ and plaque burden in AβPP/PS1 double transgenic mice [93].

Silymarin, the main flavonoid extracted from milk thistle, has long been used as a medicinal herb for liver diseases. In a mouse model, it has been shown that Silymarin treatment was associated with a decline in Aβ oligomer production [94]. Silymarin can control the production of Aβ by inhibiting the precursor substance of Aβ, AβPP, and has potential for the treatment of AD [95].

Gene silencing and genome editing

The property of a therapeutic agent in AD to penetrate the BBB has generally been regarded as a prerequisite for new drugs [96]. Recognition of the importance of the liver in Aβ production and clearance means that targeting the liver might be a promising therapeutic approach.

The last decade has witnessed renewed interest in novel therapeutic agents aiming to prevent the production of disease-causing proteins at the level of mRNA, reviewed in [97]. Broadly, two oligonucleotide-based technologies are being deployed in this way, antisense oligonucleotides (ASO) and small interfering RNAs (siRNA), which respectively prevent translation or trigger RNA induced silencing complex (RISC)-dependent cleavage of a specific RNA target. A large number of disease targets reside in the liver where they are susceptible to modulation by oligonucleotide therapies [98]. The ASO mipomersen, targeting apolipoprotein B in the liver to treat familial hypercholesterolemia, was among the first such agents to be marketed [99]. The recent entry of large numbers of gene silencing agents into clinical trials owes much to the development of conjugates that enable specific delivery of the oligonucleotide to the cytoplasm of the desired target cell, allowing lower doses with fewer side effects. In particular, addition of N-Acetylgalactosamine conjugates, which bind with high specificity and affinity to the asialoglycoprotein receptor on hepatocytes has elicited robust gene silencing in vivo [100].

Proof of concept for silencing the production of a form of aberrant amyloid by the liver in order to prevent or reverse damage to the nervous system is provided by the rare metabolic disease, hereditary transthyretin amyloidosis. In this condition, mutant transthyretin is produced in the liver resulting in amyloid fibril deposition in various organs and heterogeneous clinical symptoms including peripheral neuropathy and cardiomyopathy [101]. Patirisan, a gene silencing agent delivered to the hepatocyte in a lipid nanoparticle, reduces the production of abnormal transthyretin and can halt or even reverse the process. Both Patirisan, an siRNA and Inotersen, an untargeted ASO, are approved by the FDA and EC for this disorder [102].

Whereas gene silencing provides a temporary fix, all be it with single dose duration of action above six months now regularly achievable, the ability to permanently alter the human genome remains an attractive possibility for patients with deleterious genetic mutations. This has been made possible by a series of technologies collectively known as genome editing for which CRISPR-Cas9 was the archetype. In murine models, liver-directed somatic genome editing with CRISPR-Cas9 is a novel and versatile approach with therapeutic potential in metabolic disorders [103]. The proof of principle that a gene can be targeted in mammalian hepatocytes in vivo would suggest that sequence-specific gene editing might be viable in humans [104–107]. In the future, liver-specific gene editing may be used to alter hepatic gene transcription for therapeutic purposes in AD. It is tempting to speculate that mutations in AβPP, PSEN1 [108], and ApoE4 [109] could be targeted with these technologies. In addition, as lifelong overexpression of wild-type AβPP causes AD in individuals with trisomy 21 (Down’s syndrome), so the normal hepatic AβPP gene could be targeted in AD to decrease production and alter the balance in favor of Aβ₄₂ clearance.

Hepatitis C virus (HCV)

Chronic HCV infection has been found to be associated with dementia in a large population-based cohort (Hazard ratio 1.36 [95% CI 1.27–1.42]) [112]. More recently, a predictor of cirrhosis in chronic HCV infection, an elevated aspartate aminotransferase to alanine aminotransferase ratio, has been reported to be associated with AD diagnosis (Odds ratio: 7.932) [113]. ApoE, a critical player in Aβ homeostasis, is intimately involved in production of infectious HCV particles [114] and is important in HCV cell entry. HCV may cross the BBB leading to neuroinflammation and neuropsychiatric symptoms [115]. We have reported ApoE deficiency in HCV associated depression [116]. Interestingly low plasma levels of ApoE are associated with increased risk of future AD and all dementia in the general population, independent of ɛ2/ɛ3/ɛ4 APOE genotype [117].

NAFLD, metabolic syndrome, and type 2 diabetes

Non-alcoholic fatty liver disease (NAFLD) is a chronic liver disease which is increasing in prevalence, in tandem with the obesity epidemic. NAFLD defines a spectrum of conditions from simple steatosis to non-alcoholic steatohepatitis and cirrhosis and is regarded as the hepatic manifestation of the metabolic syndrome. Hepatic insulin resistance is associated with NAFLD and is a major factor in the pathogenesis of type 2 diabetes and the metabolic syndrome [118].

Various epidemiological studies have shown that metabolic syndrome and type 2 diabetes [119] [120] are correlated with AD. In addition, short sleep duration and poor sleep quality are associated with an increased risk of both NAFLD [121] and AD [62].

In murine models, correlations between a high fat diet and elevated brain and serum Aβ₄₂ have been observed [122], and NALFD induces signs of AD in wild-type mice and accelerates pathological signs of AD [123]. Advanced glycation end-products exacerbate progression of experimental NAFLD [124] and AD [22].