Main

The global burden of Alzheimer disease (AD), already the most common type of dementia, is expected to increase still further owing to population ageing. AD not only causes severe distress for patients and caregivers, but also results in a large economic burden on society. Current major challenges in AD include the lack of reliable biomarkers for its early diagnosis, as well as the lack of effective preventive strategies and treatments1,2. Thus, increased understanding of the molecular pathogenesis of AD could lead to the development of improved diagnostic and therapeutic strategies.

AD is conventionally regarded as a CNS disorder. However, increasing experimental, epidemiological and clinical evidence has suggested that manifestations of AD extend beyond the brain. These systemic alterations might not be simply secondary effects of the cerebral degeneration seen in AD, but could reflect underlying processes linked to progression of the disease. AD pathogenesis is complex, involving abnormal amyloid-β (Aβ) metabolism, tau hyperphosphorylation, oxidative stress, reactive glial and microglial changes, and other pathological events. Given that Aβ is a major hallmark of AD and a fertile area of research in this disease, this Review focuses on the systemic role of Aβ in AD. We discuss the communication between peripheral and central pools of Aβ, and describe interactions between systemic abnormalities and AD pathogenesis in the brain. We review emerging findings of associations between systemic abnormalities and Aβ metabolism, and describe how these associations might interact with or reflect on the central pathways of Aβ production and clearance. On the basis of these findings, we suggest that interactions between the brain and the periphery might have a crucial role in the development and progression of AD.

Aβ biogenesis and catabolism

A steady accrual of data from laboratories and clinics is providing increasing support for the concept that an imbalance between the production and clearance of Aβ is a very early (and often initiating) factor in AD3. Normal metabolism of Aβ and maintenance of the homeostatic balance between Aβ production and clearance is, therefore, essential to maintain brain health. In fact, physiological metabolism of Aβ occurs not only in the brain but also in the periphery, and communication between these regions is possible (Fig. 1).

Figure 1: Physiological metabolism of Aβ in the brain and periphery.
figure 1

Amyloid-β (Aβ) is generated by neurons, microglia and astrocytes in the brain, and by platelets, skin fibroblasts, osteoblasts, and skeletal muscle cells in the periphery. The CNS and peripheral pools of Aβ can interact; some Aβ peptides in the CNS are cleared via phagocytosis or proteolytic degradation, whereas others are released into the blood via the blood–brain barrier (BBB), interstitial fluid (ISF) bulk flow or cerebrospinal fluid (CSF) egress pathways. Some Aβ peptides in blood are phagocytosed, including by monocytes or neutrophils, some are degraded by Aβ-degrading enzymes, and some are transported by carriers (such as erythrocytes, albumin and lipoproteins) to peripheral organs or tissues, where they are degraded by macrophages or hepatocytes, or excreted via the liver or kidney. BCSFB, blood–CSF barrier; RAGE, receptor for advanced glycation end products; RBC, red blood cell.

PowerPoint slide

Full size image

Central and peripheral production of Aβ

Aβ is derived from the proteolytic cleavage of amyloid precursor protein (APP), which is expressed not only in brain cells, including neurons, astrocytes and microglia, but also in peripheral organs and tissues, such as the adrenal gland, kidney, heart, liver, spleen, pancreas, muscles, and various blood and endothelial cells4,5. Aβ levels (in both peripheral tissues and the brain) are known to be lower in cognitively normal elderly individuals than in patients with AD (Table 1). The accumulation of Aβ aggregates in elderly patients with and without AD is an important factor that could influence the ratio of Aβ42 to Aβ40 in both brain and periphery. Given that skeletal muscle represents about one-quarter of body weight in humans and is just one of many peripheral sources of Aβ, peripherally derived Aβ is likely to represent a substantial proportion of the total. However, levels and profiles of the dominant Aβ species in brain and periphery still need to be measured in young people without Aβ deposition in future studies, for comparison purposes.

Table 1 Amounts of Aβ in the CNS and periphery
Full size table

Important differences have been found between the central and peripheral pools of Aβ. First, Aβ42, which is the most aggregation-prone and most neurotoxic form of Aβ, is the dominant molecular species in the brain, whereas Aβ40 is dominant in the periphery, although the mechanism underlying this difference is still unclear. Differential expression of APP isoforms in the brain and periphery is one possible explanation. APP695 is the dominant species produced by neurons, whereas APP751 and APP770 are the dominant species produced by peripheral cells, including platelets and leukocytes6. Another possible explanation is that the tissue microenvironment differs between the CNS and periphery, which could result in differential processing of APP by γ-secretase and generation of different Aβ species4,5. Second, levels of Aβ in the central pool are higher than those in the peripheral pool. Concentrations of Aβ in cerebrospinal fluid (CSF) are at least 5–15 times higher than those in plasma7,8. One possible interpretation is that APP processing in peripheral cells probably occurs via α-cleavage (rather than the β-cleavage used in neurons)9,10, resulting in decreased peripheral production of Aβ. Another explanation for the lower Aβ levels in the periphery is that the periphery contains abundant Aβ-binding proteins (lipoproteins and albumin)11 and Aβ-binding cells, such as erythrocytes12, which all contribute to Aβ transportation and clearance. Additionally, the high volume of the circulatory system and the blood-dilution effect efficiently reduce systemic Aβ concentrations.

These factors might also help to explain why Aβ aggregates are mainly deposited in the brain and cerebral vessel walls, and only rarely in peripheral organs (although detection of Aβ aggregates has been claimed in skin, subcutaneous tissue, intestinal tissues, and heart)13,14,15. The aggregation (oligomerization and fibrillogenesis) of Aβ peptides is determined by the relative proportions of Aβ species, their concentrations, the pH, temperature and ionic strength of solution, and incubation time16. In the periphery, therefore, an increased proportion of Aβ40 might lead to sequestration of Aβ42 in a stable mixed formation, thereby preventing its oligomerization and aggregation17, whereas an increased proportion of Aβ42 in the brain might render Aβ42 susceptible to aggregation.

Central and peripheral clearance of Aβ

The failure to clear Aβ (especially Aβ42) is an important cause of sporadic AD, which accounts for 99% of AD cases. Understanding how Aβ is physiologically cleared from the brain is, therefore, essential. Several potential pathways could clear Aβ from the brain: phagocytosis, endocytosis and macropinocytosis by various cells (such as microglia, perivascular macrophages, astrocytes, oligodendroglia and neurons); proteolytic degradation by various enzymes (including neprilysin, insulin-degrading enzyme (IDE) and matrix metalloproteinases); and efflux of Aβ to the peripheral circulation, via transportation across the blood–brain barrier (BBB) and blood–CSF barrier, interstitial fluid bulk flow and CSF egress pathways, including arachnoid villi and glymphatic–lymphatic pathways18 (Fig. 1). Some endogenous inhibitors of Aβ aggregation, such as the secreted ectodomain of tumour necrosis factor receptor superfamily member 16 (also known as low affinity neurotrophin receptor p75NTR)19 and the N-terminal domain of myelin basic protein20, prevent Aβ deposition in the brain and facilitate its efflux into the circulation.

How Aβ is cleared in the periphery is poorly understood. Previous studies have suggested that ∼60% of brain Aβ is cleared via transportation to the periphery21,22. Our group has demonstrated, in a mouse model of AD, that brain-derived Aβ can be physiologically cleared in the periphery, and that a singular peripheral system can remove ∼40% of the Aβ produced in the brain23. These findings indicate that peripheral clearance has a crucial role in removing brain-derived Aβ, and suggest that effective peripheral Aβ clearance can improve the efficacy of Aβ efflux from the brain. In fact, several peripheral tissues or organs participate in Aβ catabolism and constitute potential Aβ clearance pathways. These include uptake and phagocytosis or endocytosis by monocytes, macrophages, neutrophils, lymphocytes, and hepatocytes24,25; excretion via bile or urine26,27; proteolytic degradation by Aβ-degrading enzymes28; and clearance from blood mediated by Aβ-binding proteins and cells, such as erythrocytes, albumin, antithrombin III and lipoproteins, including apolipoprotein E (ApoE) and apolipoprotein J (ApoJ)11,12. These central and peripheral pathways might interact with each other and work synergistically to clear Aβ from the brain.

Communication between Aβ pools

Brain-derived Aβ can be transported into the peripheral pool via the BBB, blood–CSF barrier, arachnoid villi or glymphatic–lymphatic pathway. Several transporters mediate Aβ flow out of the brain across the BBB, including LDL-related protein 1 (LRP1) and ATP-dependent efflux transporter P-glycoprotein29. The arachnoid villi absorb Aβ in the CSF and mediate its release into the circulation30. The glymphatic–lymphatic pathway, which consists of the glymphatic pathway in the brain and the CNS lymphatic vessels (discovered in 2015)31,32, might also transport Aβ from the brain to the periphery for clearance18,33. However, the glymphatic–lymphatic pathway and arachnoid villi are unidirectional; they only mediate Aβ efflux from the CNS to the periphery18.

Whether peripherally generated Aβ can enter the brain and exert neurotoxic effects there remains poorly understood. In the absence of a relevant transport mechanism, systemic amyloidosis might not necessarily lead to AD. However, peripheral inoculation of Aβ-containing brain extracts induces cerebral Aβ deposition in both mice and humans, suggesting that peripherally generated Aβ is able to enter the brain and participate in the pathogenesis of AD34,35,36,37. Receptor for advanced glycation end products (RAGE) has been suggested to transport Aβ across the BBB, from the blood into the brain38. Expression of the AGER gene (encoding RAGE) is upregulated in the AD brain vasculature39,40, indicating that influx of peripheral Aβ into the brain is increased in AD. The contribution of peripherally derived Aβ to amyloidosis in the AD brain needs to be determined in future studies.

A decline in peripheral Aβ clearance might also impede efflux of Aβ from the brain to the periphery, and thereby attenuate central clearance of Aβ. Moreover, the influx and efflux of Aβ might result in equilibrium between the central and peripheral pools of Aβ. Mechanisms that might regulate this equilibrium need to be understood.

Systemic abnormalities in AD

An increasing number of studies indicate that a series of systemic abnormalities can exacerbate the progression of AD (Fig. 2). In turn, the downstream effects of processes in the AD brain can also drive these systemic disorders, forming feedback loops. Mechanisms that might underlie the effects of systemic abnormalities or alterations on Aβ metabolism are outlined in Table 2. Here, we discuss the interactions between Aβ metabolism in the brain and periphery, and place them in a systemic context.

Figure 2: Systemic abnormalities in AD.
figure 2

Various systemic abnormalities have been found in patients with Alzheimer disease (AD). Red boxes highlight AD risk variants in genes related to innate immunity, phagocytosis of amyoid-β (Aβ) by immune cells, and lipid metabolism in periphery and brain, respectively, which were identified in genome-wide association studies and candidate-gene studies of sporadic AD. APP, amyloid precursor protein; BA, basilar artery; COPD, chronic obstructive pulmonary disease; CRP, C-reactive protein; HDL-C, HDL-cholesterol; IGF, insulin-like growth factor; LDL-C, LDL-cholesterol; LPS, lipopolysaccharide; LTICA, left terminal internal carotid artery; OSA, obstructive sleep apnoea; PI, pulsatility index; RBC, red blood cells; RI, resistance index; RTICA, right terminal internal carotid artery; TNF, tumour necrosis factor.

PowerPoint slide

Full size image
Table 2 Potential mechanisms of systemic abnormalities that affect Aβ metabolism
Full size table

Disorders of systemic immunity

One of the primary pathways of Aβ clearance in the brain is phagocytosis or endocytosis by professional phagocytes and microglia, as well as by astrocytes, oligodendrocytes and neurons. Accumulations of Aβ in the periphery can similarly be phagocytosed by monocytes and neutrophils in the blood, and by macrophages in tissues41. Of note, in transgenic mice with AD, expression of Aβ scavenger receptors and Aβ-degrading enzymes in circulating mononuclear phagocytes decreases substantially as these mice age42, and the phagocytic functions of these cells are impaired in both mice and humans with AD43,44,45. Infusion of monocytes derived from peripheral human umbilical cord blood reduces the Aβ burden and improves cognitive deficits in a mouse model of AD46, implying that peripheral mononuclear phagocytes have an important role in Aβ clearance. Promoting the phagocytic function of peripheral blood monocytes or promoting the recruitment of peripheral macrophages into the brain might, therefore, improve Aβ clearance in the brain47, although the existence of conflicting data48 renders this approach controversial.

In this regard, a cluster of genes associated with the risk of sporadic AD (including CD33, CR1, MS4A6A, MS4A4E, ABCA7 and TREM2)49,50 encode proteins that are involved in innate immunity. Variants in these genes, especially in TREM2 and CD33, are associated with compromised phagocytic function of peripheral monocytes or macrophages and altered Aβ accumulation in AD brains24,51. Interestingly, CR1 (encoding complement receptor-1, also known as CD35) is expressed primarily in peripheral leukocytes and erythrocytes, but not in any brain cells.

In regard to adaptive immunity, much attention has been focused on autoimmunity and autoreactive antibodies related to the pathogenesis of AD, including naturally occurring antibodies and autoantibodies. These autoreactive antibodies are ubiquitous in human blood and CSF, and profiles of these antibodies are altered in patients with AD52,53,54,55,56. Identification of the most antigenic epitopes targeted by human antibodies against Aβ aggregates could lead to development of an effective immunotherapy for AD. Aducanumab, derived from a naturally occurring human autoantibody against Cu2+-modified Aβ aggregates (which are the most neurotoxic Aβ species in the AD brain), showed promise in clearing brain Aβ deposits and improving cognition in a 2016 phase Ib trial57. Lymphocytes (including B cells, T cells and natural killer cells) also participate in Aβ clearance via immunoglobulin-mediated adaptive phagocytosis58,59. Future studies will help to elucidate the crosstalk between innate immunity and adaptive immunity, and to discover how the interaction of these two immune systems might synergistically affect AD pathogenesis.

Blood abnormalities

Besides monocytes and leukocytes, other blood components are also involved in Aβ metabolism. Increased expression of APP, altered APP isoform ratios and processing patterns, and enhanced β-secretase activity are observed in platelets from patients with AD60,61,62, and these changes could result in overproduction of Aβ in the periphery. Blood levels of albumin, an Aβ carrier, are reduced in patients with AD63. The quantity and function of erythrocytes (another Aβ carrier) are also altered in such patients, and their erythrocytes show compromised binding of Aβ64,65. In addition, the activity of Aβ-degrading enzymes in serum is thought to be decreased in AD28. These changes impede Aβ transportation and clearance in the periphery.

Some anti-ageing molecules, such as growth and differentiation factor 11, granulocyte–macrophage colony stimulating factor, and metalloproteinase inhibitor 2, have been identified in blood from young mice and in human umbilical cord plasma66,67. Whether levels of these anti-ageing molecules are reduced in patients with AD, and whether they are pathophysiologically relevant to this disease, remain unknown. However, identification of these protective components could be of importance in understanding the pathogenesis of AD and in developing systemic rejuvenation therapies68,69,70,71.

Metabolic disorders

Diabetes mellitus. How diabetes mellitus affects Aβ catabolism and AD risk is not yet well understood. Patients with diabetes mellitus are estimated to be 1.4–2.0-fold more likely than healthy individuals to develop AD72,73, although these claims need to be verified in patients with biomarker-confirmed AD. In patients with diabetes mellitus, insulin resistance substantially compromises the positive effects of insulin on both cognition and hepatic clearance of circulating Aβ74,75, resulting in AD-like alterations in the brain. Moreover, excess insulin can competitively inhibit IDE-mediated Aβ degradation76. Some other pathological features of diabetes mellitus — including oxidative stress, BBB disruption and reduced cell energy supply — can also affect Aβ generation and clearance77,78. In addition, amylin (a misfolded protein deposited in the pancreas in patients with type 2 diabetes mellitus) can enter the brain, where it accelerates and exacerbates the misfolding and aggregation of Aβ79. However, atherosclerosis and small vessel disease (discussed in more detail below) can be important causes of cognitive dysfunction in patients with diabetes mellitus80, and should be considered in the differential diagnosis of AD in this setting.

Lipid and lipoprotein risk factors. Some evidence suggests that abnormal lipid metabolism is associated with an increased risk of AD81. Several potential AD risk genes (including APOE, BIN1, CLU, SORL1, PICALM and PLD3) encode proteins linked to lipid metabolism82. Among these, ApoE is known to participate in Aβ production, aggregation, and clearance in an isoform-dependent manner83,84.

Cholesterol levels in the brain can affect Aβ synthesis, clearance and neurotoxicity. High serum cholesterol levels are associated with an increased cerebral burden of Aβ85,86. Abnormal cholesterol levels could reflect unmeasured genetic factors or dietary patterns that might affect the pathogenesis of AD85,86. Cell membrane fluidity is strongly affected by its lipid composition, and increased membrane fluidity of platelets and leukocytes has been reported in patients with AD, as well as in individuals with Down syndrome (who have a greatly increased risk of developing dementia and AD)87,88,89,90. Cell membrane fluidity also influences the processing of APP and cellular phagocytosis, both of which might affect Aβ generation and clearance. A high dietary intake of polyunsaturated fatty acids such as docosahexaenoic acid, which maintain membrane fluidity, have a beneficial effect on cognition in patients with AD91,92. The study of membrane fluidity in AD might provide insights into the alteration of membrane-dependent biological functions related to Aβ, such as phagocytosis, endocytosis, macropinocytosis and autophagy.

Cardiovascular disease

Emerging evidence indicates that cardiovascular disease (CVD) is a major comorbidity in patients with sporadic AD. A low cardiac index and heart failure are both associated with dementia, and perhaps also with AD93,94,95. However, as CVD and AD are both complex and multifactorial age-related diseases, the association between them might be attributed partly to shared risk factors, such as diabetes mellitus, hypertension, hypercholesterolemia and stroke96,97.

The presence of compromised myocardial function and intramyocardial deposits of Aβ in patients with AD suggests that peripheral Aβ accumulation could affect heart function in patients with AD15,98. In addition, cardiac systolic dysfunction could affect Aβ generation and clearance in the brain as a result of reduced cerebral blood flow99,100,101,102. Regional cerebral blood flow and glucose uptake or metabolism are consistently decreased in Aβ-positive patients with AD103, and correlate inversely with AD severity104. Indeed, some degree of cerebral small vessel disease almost always accompanies AD. Increased stiffness of small vessel walls might attenuate Aβ clearance via the BBB, interstitial fluid bulk flow and glymphatic pathways, thereby accelerating AD105,106. However, a clear understanding of the interaction between CVD and AD is lacking. Most of the evidence points to CVD being an independent risk factor for cognitive impairment, and having an additive rather than synergistic effect on the AD neurodegenerative process.

Hepatic dysfunction

The liver is the major organ responsible for system-wide metabolic regulation, protein synthesis and metabolic detoxification. Circulating Aβ is predominantly cleared by either degradation in hepatocytes or direct excretion in bile; several peptide clearance experiments have suggested that soluble Aβ has a short half-life of 2.5 min to 2.5 h in the circulation26,107. LRP1 is thought to mediate the uptake of Aβ by hepatocytes25. The liver might also indirectly influence Aβ clearance by regulating albumin levels and Aβ-related lipid metabolism. Plasma Aβ levels inversely correlate with liver function, suggesting that hepatic dysfunction attenuates peripheral Aβ clearance108. Liver tissue from patients with AD contains less Aβ than that from healthy individuals, which implies that the Aβ-clearance function of liver is compromised in patients with AD5.

Whether liver dysfunction also increases the Aβ load in the brain remains unknown; however, treatments that enhance LRP1-mediated Aβ uptake by the liver alleviate both the burden of Aβ in the brain and cognitive impairment74,109. These observations suggest that improving the Aβ clearance capacity of the liver is a potential systemic therapeutic approach for AD.

Renal dysfunction

Soluble Aβ is a normal component of human urine27. In addition, animal experiments have shown that, after intracranial or intravenous infusion of 125I-labelled Aβ, radioactivity is subsequently detected in the kidney and urine23,110. These findings indicate that the kidney might participate in physiological clearance of Aβ by filtering Aβ from blood to urine. Conversely, renal dysfunction probably leads to impaired peripheral Aβ clearance. In support of this notion, serum Aβ levels inversely correlate with measures of renal function (estimated glomerular filtration rate and creatinine levels) in patients with chronic kidney disease111,112. Moreover, human kidney donors have decreased estimated glomerular filtration rates and increased circulating levels of Aβ113, suggesting that the reduction in renal function reserve associated with having a single kidney also attenuates peripheral Aβ clearance.

Whether renal dysfunction increases Aβ burden in the brain or facilitates AD processes remains unknown. Renal dysfunction increases the risk of both cognitive impairment and dementia114, and this association could involve AD pathogenetic pathways. However, kidney transplantation can reduce plasma Aβ levels113, and haemodialysis alleviates Aβ deposition in the brain of patients with chronic kidney disease115. These observations suggest that improvement of renal function is a promising approach to AD prevention and treatment.

Respiratory and sleep disorders

Patients with AD have an increased incidence of respiratory disorders, such as bronchopneumonia, obstructive sleep apnoea (OSA) and sleep-disordered breathing116,117. In addition, sleep-disordered breathing is associated with an increased risk of mild cognitive impairment or dementia and with earlier onset of AD118,119,120.

Compared with healthy control individuals, patients with OSA or chronic obstructive pulmonary disease exhibit higher blood levels of Aβ, which negatively correlate with pulmonary function121,122. OSA is also associated with altered levels of AD biomarkers in CSF, including decreased levels of Aβ42 and elevated levels of phosphorylated tau123. OSA and chronic obstructive pulmonary disease could contribute to AD processes via hypoxia, inflammation, or sleep disruption124. Sleep disruption has been suggested to increase Aβ production and aggregation, suppress glymphatic clearance of AD pathogenic proteins (tau as well as Aβ) and aggravate oxidative stress, inflammation and synaptic damage125,126.

Gut microbiota disturbance and infection

The establishment of the gut–brain axis revealed a clear association between the gastrointestinal microbiota and cognition127. Gram-negative bacterial species (such as Escherichia coli K99) are the predominant sources of bacteria-derived factors in normal human brains, and levels of these molecules are increased in AD brains, along with levels of the bacterial cell wall component lipopolysaccharide128. Lipopolysaccharide colocalizes with Aβ in plaques in AD brains128, suggesting that Gram-negative bacteria are associated with AD pathogenesis. Probiotic supplementation is associated with improved cognition in patients with AD129, which further supports a role for the gut microbiota in AD development.

In addition to its relationship with normal microbial flora, emerging evidence indicates that AD is associated with exposure to an ever-increasing number of pathogens130,131. Moreover, a high infectious burden is associated with increased serum levels of Aβ and proinflammatory cytokines in patients with AD and in healthy controls130. However, the underlying mechanisms through which the microbiota or pathogens influence AD remain to be determined. Whether the microbiota or pathogens contribute to AD development, or whether an increased infectious burden is a consequence of AD, also remains unknown.

Systemic inflammation

Chronic reactive gliosis and microgliosis are neuroinflammatory responses that are important contributors to AD pathology. These processes might participate in a positive feedback loop of Aβ deposition, neurofibrillary tangle formation, and damage to synapses and neurons. Several studies have shown that other conditions involving chronic systemic inflammation, such as rheumatoid arthritis and periodontitis, are associated with an increased risk of AD132,133. These conditions are also associated with elevated levels of C-reactive protein and proinflammatory cytokines, such as tumour necrosis factor, IL-6 and IL-1β. These proinflammatory molecules could participate in AD pathogenesis either directly, by affecting brain Aβ metabolism (via entry to the CNS through the BBB or neural afferent pathways such as the vagus nerve)134,135 or indirectly, by affecting Aβ metabolism in the periphery. In this regard, the results of observational studies show that NSAID use is associated with a reduced risk of AD136. These findings suggest that chronic systemic inflammation promotes the AD process.

By contrast, acute systemic inflammatory responses seem to protect against AD — at least in animal models — by recruiting monocyte-derived macrophages into the brain, where they clear cerebral Aβ47. However, most studies in this area have not been repeated, and their results have not yet been validated in patients with biomarker-confirmed AD.

A systemic approach to understanding AD

The close interaction between the brain and the periphery, in terms of Aβ metabolism, provides novel insights into the pathogenesis of AD, and could lead to new approaches to the diagnosis and treatment of AD, based on systems biology and systems neurophysiology paradigms.

Pathogenesis

Our current understanding of the role of Aβ in AD focuses on its contribution to brain pathology and symptoms. However, as already discussed, this view might not be the whole story. First, although Aβ peptides are generated in the brain, a considerable amount of Aβ is also generated in peripheral systems. Second, Aβ can be cleared — from peripheral organs or tissues as well as from the brain — by professional phagocytes, which can transmigrate through the BBB. Third, Aβ deposits have been detected in the periphery — although this claim has not yet been replicated and its pathophysiological relevance remains unknown. Last, a series of systemic abnormalities are both driven by and contribute to AD progression. On the basis of these findings, we propose that AD might not be solely a brain disorder, in the sense that systemic factors might interact with the brain to modify the AD process.

As discussed, the central and peripheral Aβ pools interact with and influence each other. For example, the rate of peripheral catabolism of Aβ seems to affect the rate of Aβ efflux from the brain, and peripherally derived Aβ can enter the brain and accelerate the progression of cerebral AD pathology34,35. Therefore, we hypothesize that the peripheral pool of Aβ is not simply associated with AD, but is causally linked to this disease. Indeed, interactions between the brain and the periphery might have a crucial role in the natural history of AD, and elucidation of the effects of peripheral processes on AD development could lead to improved understanding of its pathogenesis. The crucial questions to answer would be precisely how the brain and periphery interact with each other to affect AD progression, and whether interventions that target systemic factors can modulate the pathogenesis or development of AD.

Diagnosis

Several PET radiotracers can be used to detect Aβ in the brain, and a few biomarkers for AD have been validated for diagnostic use, including CSF levels of Aβ42, total tau and phosphorylated tau. However, these approaches are either invasive or expensive, and are impractical for the early diagnosis of patients without obvious cognitive complaints. The search for peripheral blood or plasma biomarkers for AD that reflect AD-related processes in the brain has, therefore, received considerable attention. However, owing to the complexity of blood components, the accurate measurement of plasma levels of Aβ or tau is very challenging. A study published in 2017 did not find a statistically significant difference in plasma levels of free Aβ between patients with AD and age-matched controls137, and similarly negative results have been published for plasma tau levels138. However, these results do not indicate the end of the road for AD biomarker studies139. With the development of advanced and highly sensitive techniques that are able to promote efflux of Aβ and tau from the brain, and accurately measure their levels even when bound to other serum proteins, researchers might eventually find a method to monitor cerebral Aβ accumulation. In addition, misfolded oligomeric species of Aβ and tau proteins have been detected in CSF and suggested as potential biomarkers for AD140,141. If these misfolded protein species also prove to be present in blood, they might be useful for AD diagnosis, although no published reports yet exist.

Technical advances by the AD Neuroimaging Initiative have enabled the use of microarrays to detect serum anti-Aβ autoantibodies with 100% accuracy55; however, this promising technique requires validation in large independent cohorts. This approach is particularly interesting in the light of the promising results obtained with aducanumab, the first human anti-Aβ autoantibody to be developed for clinical trials57. In addition to Aβ and tau, β-secretase 1 (encoded by BACE1, an enzyme involved in Aβ production) has been suggested as a potential biomarker for AD142. In general, the identification of novel peripheral biomarkers for AD diagnosis and prognosis represents a promising and rapidly expanding research direction. (Box 1)

Treatment

Currently, effective agents for AD prevention or treatment are lacking. The traditional concept 'one target, one treatment' inevitably ignores the complexity of AD pathogenetic mechanisms143. After the failure of over 100 clinical trials of monotherapies targeting Aβ, multitargeted therapies that address various aspects of AD pathogenesis at different disease stages are needed144. We argue that a comprehensive strategy targeting both brain and peripheral (systemic) abnormalities might be more effective than strategies that target CNS abnormalities alone. As discussed, many comorbidities of (and risk factors for) AD — such as diabetes mellitus, metabolic disorders, cardiovascular diseases, and hypertension — are systemic disorders.

Many attempts have been made to prevent AD via peripheral interventions, and some have been associated with beneficial outcomes. Improvements in overall population health have led to a decreased incidence and prevalence of dementia over the past 10–30 years145,146,147,148, perhaps through improved management of cardiovascular risk factors145,146,147,148. For example, administration of statins (which reduce peripheral blood cholesterol levels) to healthy middle-aged individuals was associated with a reduced dementia risk in one large-scale prospective cohort study149; statins have also decreased the brain burden of Aβ in experimental models of AD150, although the results of most studies of statin treatment in patients with AD have been disappointing151. Furthermore, treatment with continuous positive airway pressure for sleep-disordered breathing or OSA might delay the onset of mild cognitive impairment and slow or even improve cognitive decline in patients with AD120,152,153. These observations support the view that systemic management of an individual's known comorbidities or risk factors, with the aim of maintaining bodily homeostasis, might help to prevent or slow the progression of AD.

Active removal of excess peripheral Aβ seems to be a particularly promising therapeutic strategy for AD23. Plasma albumin exchange both improves cognition and decreases the Aβ burden in patients with AD154. Peritoneal dialysis reduces blood Aβ levels in humans and also attenuates AD pathology in an AD mouse model155; patients who have undergone haemodialysis exhibit a reduction in Aβ deposition in the brain115. Approaches to improve peripheral Aβ clearance via enhancing phagocytosis45, proteolytic degradation and excretion155,156, and identification of rejuvenation factors in blood68, are other promising systemic therapeutic strategies for AD157.

Conclusions

AD might be not only a brain disorder, but also a systemic disease with widespread abnormalities beyond the brain. Thus, systemic factors might interact with brain-related factors to modify the AD process. AD diagnosis and treatment should have a corresponding focus not only on pathological changes in the brain but also on peripheral abnormalities, which vary among individuals. Identifying these peripheral abnormalities might offer new opportunities for diagnosis of early AD and lead to the design of specific treatment strategies for individuals with preclinical, prodromal or frank AD. In conclusion, the systemic view of AD proposed in this Review provides a novel perspective for understanding the pathogenesis of this disease, and fosters new opportunities for its early diagnosis and successful management.

Review criteria

Articles for inclusion in this Review were identified by searches of the PubMed database using the following terms: “Alzheimer disease”, “amyloid beta”, “peripheral clearance”, “innate immunity”, “adaptive immunity”, “macrophage”, “monocyte”, “phagocytosis”, “blood”, “circulation”, “plasma”, “platelet”, “erythrocyte”, “exosome”, “diabetes”, “insulin”, “amylin”, “apolipoprotein E”, “cholesterol”, “lipid”, “lipoprotein”, “cerebrovascular diseases”, “heart failure”, “cerebral blood flow”, “hypoperfusion”, “hypoxia”, “liver”, “hepatic function”, “kidney”, “renal function”, “respiratory”, “sleep”, “microbiome”, “gastrointestinal microbiota”, “gut flora”, “infection”, “inflammation”, “immune responses”, “GWAS”, “risk factor”, “biomarker”, “diagnosis”, “therapy”, “treatment”, “prevent”. Only articles published in English were retrieved. Full-text papers were available for most of the articles that were chosen for review, and the reference lists of these articles were searched for further relevant material.

Publisher's note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.