Science - USA (2020-10-02)

been found to modulate lipid metabolism,

lysosomal pathways, and microglial metabolic

fitness ( 3 , 17 – 23 ).

How microglia may mediate synaptic

loss in AD

In AD, synaptic loss and dysfunction are region

specific, early, and strongly correlated with cog-

nitive impairment ( 24 ). Prefibrillar oligomeric

b-amyloid (Ab) and/or tau accumulate on syn-

apses and induce pathological synaptic dysfunc-

tion and loss ( 25 – 30 ). More than half of the

identified genetic risk factors in AD are expressed

by myeloid cells ( 31 ). Together, these data high-

light the need to understand how mutations

in risk genes and alleles impair the cross-

talk between microglia, the major myeloid

cell population in the brain, and neurons at

the synapse.

Multiple studies in animal models of AD

have suggested that there is dysregulation

of neuroimmune signaling pathways on syn-

apses involving classical complement cas-

cade, TREM2, phosphatidylserine (PtdSer),

and ApoE (Fig. 1). These studies raise the

intriguing question of whether the accumu-

lation of local pathological proteins on syn-

apses dysregulates neuron-glia interactions

that are critical for synaptic health. For ex-

ample, pathological Abor tau accumulated

on synapses up-regulate C1q in surrounding

microglia and promote complement activa-

tion on synapses and subsequent microglial

engulfment ( 26 , 29 , 32 , 33 ). Blocking the

activation of the classical complement cas-

cade in AD mouse models with genetic or

antibody-based means has been shown to

protect synapses from loss and dysfunction

and memory loss ( 26 , 29 , 32 – 34 ), which

suggests that the microglia-synapse pruning

pathway may be a potential therapeutic target.

What remains unclear is whether this pruning

mechanism—at least in the beginning—is a

beneficial process that then becomes dysregu-

lated in a chronic manner and impairs the

very neurons it was trying to save. Microglial

engulfment of synapses likely involves a fine

balance of“eat me”and“don’t-eat me”signals

( 35 ). Because many of the microglial functions

including synaptic pruning appear to be

activity-dependent ( 7 , 8 , 12 , 36 ), it will be

important to determine whether neuronal

hyperactivity observed in early-AD mouse

models ( 30 , 37 , 38 ) instructs microglia to

aberrantly engulf synapses ( 36 ). Insight into

the pathways that regulate pruning, as well

as the specific signals that guide microglia to

engulf synapses, will be crucial for iden-

tifying potential therapeutic targets against

cognitive decline and for developing biomarker

candidates to quantify microglial dysfunc-

tion in relation to synaptic loss.

Another biologically and therapeutically

important question is whether particular

synapses are targeted for elimination by microg-

lia. Proteomic studies in synaptosomes from

human and mouse AD brains have highlighted

synaptic mitochondrial dysfunction ( 14 , 39 ).

However, whether complement factors, includ-

ing C1q and C3b, target specific—i.e., dysfunc-

tional and/or damaged—synapses is not known.

Lipid signaling in neuron-glia interplay may

be a pivotal determinant. For example, TREM2,

a key damage sensor on microglia ( 40 ), has been

shown to mediate synaptic refinement in the

developing mouse hippocampus ( 41 ). A proposed

ligand for TREM2 is exposed PtdSer on the

outer leaflet of neuronal membranes ( 19 ). Thus,

exposed PtdSer on synapses may be an eat-me

signal for microglial TREM2 ( 42 , 43 ). Further-

more, recent work in a model of tauopathy has

suggested a potential link between TREM2

and microglia-mediated synaptic elimination

in AD ( 44 ). The AD risk variant of TREM2 was

associated with less synaptic localization of

C1q and fewer engulfed synaptic elements by

microglia compared with the common variant

of TREM2. Together, these data suggest a po-

tential role for microglial TREM2 in sensing

damaged synaptic membranes in AD, perhaps

through PtdSer signaling. Gangliosides, a family

of sialic acid–containing lipids enriched in the

brain, have also been postulated to be crucial for

Ab-induced synaptic dysfunction in mice ( 45 ).

GM1 ganglioside–bound Abis enriched on

membranes in early-AD brains ( 46 ). More-

over, anti–GM1 ganglioside antibodies have

been shown to fix complement on neuronal

membranes, and the same antibody targeting

C1q that was used in AD models ( 26 , 29 ) has

been shown to ameliorate antiganglioside

antibody-mediated neuronal injury in a mouse

model of acute motor axonal neuropathy ( 47 ).

These studies raise the question of whether

brain gangliosides contribute to synaptic loss

in AD and complement-mediated synaptic

engulfment by microglia.

An additional neuroimmune and lipid-

related pathway to consider is ApoE. Previous

research has suggested a possible link between

astrocytic ApoE and microglial synaptic prun-

ing: The ApoE allele-dependent rate of syn-

aptic engulfment by astrocytes appears to be

important for normal synapse plasticity ( 48 ).

This rate appears to slow down during aging,

thus potentially increasing vulnerability of

synapses to complement-mediated pruning

by microglia. Notably, ApoEe4 has been

associated with enhanced synaptic localiza-

tion of pathological Abin human AD brains

( 25 ). Furthermore, ApoE has recently been

shown to bind C1q and regulate the activa-

tion of the classical complement cascade ( 49 ).

Together, these data suggest a role for ApoE

at the synapse in astrocyte-neuron-microglia

cross-talk, which is of great interest, especially

in light of cell type–specific dysregulation of

ApoE in AD and critically linked cholesterol

and other lipid metabolic pathways.

Studies involving TREM2 and ApoE, two

of the major risk factors in late-onset AD,

suggest that lipid metabolism in microglia

may be a determinant of how well the brain’s

immune system can respond to the chronic

buildup of amyloids. For example, TREM2-

deficient microglia fail to properly metabolize

lipids in a chronic demyelination paradigm

( 21 ). Furthermore, TREM2 appears to be a

key regulator of ApoE, a major lipid trans-

porter ( 18 ). ApoE has been shown to trans-

port excess lipids from hyperactive neurons

to lipid droplets in astrocytes where they are

metabolized, which suggests a role for ApoE

SCIENCEsciencemag.org 2 OCTOBER 2020•VOL 370 ISSUE 6512 67

A

Hyperactivity

Mitochondria

Aβ oligomers

CR3/C3

C1q

Microglia

Tau

C

Neuron

TREM2

PtdSer

C1q

Astrocyte

Microglia

Lipid

droplet

ApoE

B Targeted

elimination

Elimination of

dysfunctional synapses

C1q

C3b

TREM2

Gangliosides

?

Fig. 1. Complement-mediated synapse loss by microglia
in AD.Potential mechanisms leading to complement-mediated
synapse elimination by microglia. (A) Whether microglia
target specific synapses is not known. Neuronal hyperactivity
and/or mitochondrial dysfunction observed in AD mouse
models and patients may lead to up-regulation of complement
factors (C1q and CR3) in microglia to target the dysfunctional
synapses. (B) In AD mouse models, pathological Abor tau
accumulated on gangliosides may up-regulate complement
signaling pathways through membrane damage sensors like
TREM2, which results in synaptic elimination by microglia.
(C) Recognition of exposed PtdSer on synapses by myeloid
TREM2 may lead to synaptic engulfment by microglia. Alternatively,
lipid transporter ApoE potentially ameliorates hyperactivity-
or membrane damage–induced lipid toxicity by transporting
ILLUSTRATION: MELISSA THOMAS BAUM/ excess lipids to lipid droplets in astrocytes and microglia.

SCIENCE