Innate immune cells
Inside the brain, ongoing AD pathology leads to the differentiation of microglia into a novel type associated with neurodegenerative diseases with altered molecular expression profile and limited phagocytic capacity [
77,
78]. The precise origin of plaque-surrounding amoeboid myeloid cells has long been debated, owing to the technical challenge of distinguishing infiltrating myeloid cells from locally-activated microglia. Nevertheless, recent evidence suggests that peripherally-derived macrophages can engraft the brain and maintain a unique functional and transcriptional identity in CNS [
79]. Infiltrating peripheral myeloid cells have been demonstrated to participate in Aβ clearance, leading to the notion that monocytes are superior phagocytes when traversed inside the AD brain [
80,
81]. However, such a conclusion is challenged by studies showing that infiltrating peripheral myeloid cells, in replacing ablated microglia, adopt a microglia-like phenotype in the brain,
i.e. with limited phagocytic capacity, likely influenced by local molecular cues [
82,
83]. Whether additional signals are needed to boost their phagocytic function
in situ requires further investigation.
Rare variants in Triggering Receptor Expressed on Myeloid cells 2 (TREM2) increase the risk of developing AD by 2~3 fold. Significant TREM2-dependent phenotypes in mouse models of AD shed light on the role of TREM2 in regulating neuroinflammation during AD pathogenesis [
84]. Besides CNS upregulation, expression of TREM2 mRNA and protein are increased in the peripheral leukocytes of AD patients and correlated with cognitive deficits and hippocampal atrophy [
85‐
88]. It is unclear if elevated peripheral TREM2 expression has functional impact on disease pathogenesis or merely reflects the ongoing systemic inflammation of AD.
Neutrophils are the most abundant myeloid cells in human peripheral blood and participate critically in protective innate immunity. Neutrophils have been detected in the brain parenchyma of 5xFAD and 3xTg-AD mice and were shown to promote amyloid plaques and tau tangles as well as cognitive decline [
89]. Yet, a separate study showed that long term treatment of 3xTg-AD mice with a TNF-α modulator led to increased neutrophil infiltration in the brain, which coincided with improved learning and memory, and reduced tau and amyloid pathology [
90]. Thus, the functional features of infiltrating neutrophils remain to be fully elucidated.
Adaptive immune cells
Distinct from innate immunity, adaptive immune responses protect the host in a manner that is long-lasting and antigen-specific. Although innate immune responses at the site of brain inflammation are well-established, the role played by adaptive immunity in AD remains vague, primarily due to the scarcity of T cells or B cells inside the parenchyma. Nevertheless, a recent large-scale genome-wide association study strongly implicated the involvement of both adaptive and innate immunity in AD [
91]. In addition, the association of a single-nucleotide polymorphism (SNP) of MHC class II HLA-DRB5 with AD implies a potential T cell-mediated process [
92]. To assess the overall function of the adaptive immunity, two groups analyzed severely immunocompromised transgenic mice expressing human mutant APP and obtained conflicting results. PSAPP:Rag2
-/- mice lacking B and T cells showed reduced Aβ pathology accompanied by highly phagocytic microglia [
93], whereas 5xFAD:Rag2
-/-:Il2rg
-/- mice lacking B, T and NK cells displayed exaggerated Aβ plaque deposition and increased neuroinflammation [
94]. While the two mouse models do differ in several aspects, the primary reason for the discrepant observations is unclear. Given the memory and behavior impairment in Rag2
-/- mice [
68,
95,
96], how this may confound Aβ-associated functional decline was not addressed.
B cells in adaptive immunity
Humoral immune response (
i.e. antibody production) by B cells to amyloid β has been studied for over 20 years. It was first discovered that immortalized B cells from the peripheral blood of an AD patient secreted antibodies that specifically recognize Aβ peptide [
97]. Circulating anti-Aβ antibody at various levels has since been detected in the blood of human subjects with or without AD [
98‐
101]. The groundbreaking study by Schenk et al, in which Aβ immunization prevented the development of amyloid plaques, neuritic dystrophy and gliosis in PDAPP mice, demonstrated the therapeutic potential of B cell-mediated immune response [
102]. However, human vaccination trial AN1792 based on the same principle had to be halted due to life-threatening encephalitis, which was later attributed largely to the pathogenic autoimmune T
H1 response stimulated by active immunization (see next section). Subsequent studies in animal models demonstrated the ability of individual Aβ-specific antibody clones to attenuate AD pathogenesis without the involvement of T cells, lending support for the clinical trials to passively immunize AD patients with anti-Aβ antibodies. However, such a strategy has yielded mixed outcomes with several trials still ongoing (reviewed by [
103‐
105]; also see section
Passive Immunotherapy)).
Besides the humoral response specific to proteins associated with AD pathology, the peripheral immunoglobulin repertoire is dysregulated in AD. Spontaneously secreted by B cells without exogenous stimulation, natural antibodies are abundant in normal sera, typically poly- or self-reactive, and fulfill important functions of targeting pathogens and removing cellular and molecular waste in the body [
106]. The levels of natural IgG recognizing self-antigens was shown to be influenced by age, sex and disease [
107]. Significant reduction of the number of autoantibodies was detected in the sera from patients with AD, Parkinson’s disease (PD), and multiple sclerosis, a neuroinflammatory condition [
107]. The cause and functional significance of such a decrease is unclear. Unexpectedly, a recent time-course global proteomic analysis of APP
NL-F mice reported an inverse correlation between Aβ burden in the brain and IgM levels in the blood [
108]. Again, the implication of this finding remains to be seen. Nevertheless using a panel of autoantibodies as biomarkers, DeMarshall et al. were able to differentiate mild-moderate AD patients from age-matched controls, MCI, and other neurological disorders with high confidence [
109].
Another insight on the role of immunoglobulin in AD came from the study by Marsh et al. [
94]. Increased mouse IgGs were detected in association with microglia in the brains of 5xFAD mice. Lacking specificity for Aβ, these proteins were shown to engage with microglia Fc receptor, induce an activating signaling pathway, and stimulate phagocytosis of Aβ, which lead to a decreased plaque load. Although the antigen-specificity of the infiltrating IgG was not studied in this study, the protective effect of mouse IgG is reminiscent of the beneficial influence previously reported with intravenous Immunoglobulin (IVIg), a pooled human IgG product [
110‐
112]. Whether any specific subset of the natural IgGs confers enhanced neuroprotection remains to be investigated.
T cells in adaptive immunity
Inside the post-mortem brains of AD patients, both CD4
+ and CD8
+ T cells were detected, occasionally next to the neuritic plaques or microglia [
113‐
115]. Similarly, increased numbers of T cells have been found infiltrating the brain parenchyma of multiple transgenic mice expressing mutant human APP [
116,
117]. In one study, significant fractions of infiltrating T cells were found to produce IFNγ and IL-17 in APP/PS1 brain [
117]. However, Ferretti et al. reported that infiltrating T cells consistently display an inactivated phenotype with reduced IFNγ production and lack of local proliferation and do not co-localize with amyloid plaques in three AD models (ArcAβ, APP/PS1 and Tg2576) [
116]. In line with this, peripheral CD4
+ T cells were found hyporesponsive to Aβ peptides in Tg2576 mice, which resulted in a defective Aβ-specific antibody response [
118]. In humans, an early study reported unresponsiveness to Aβ40 peptide by peripheral lymphocytes from AD patients [
119]. However, peripheral T cells reactive to Aβ42 peptide, which is more immunogenic than Aβ40, was later found significantly increased in AD patients and older adults over middle-aged healthy controls [
120]. T regulatory cells (Tregs) are a crucial T cell subset that suppresses effector immune responses and maintains immune tolerance. AD and Down syndrome patients both display Aβ plaque accumulation and similarly have increased Aβ-specific IL-10-producing Treg cells in their blood [
121]. Elevated Treg levels and suppressive activities in the periphery of AD patients were reported by other groups [
122‐
124], which is affirmed by the increased FoxP3
+ Tregs in the spleens of 5xFAD mice [
125].
Given the distinct and powerful roles played by T helper subsets in numerous diseases, it would be vital to dissect how different T
H cell populations specifically modulate AD. In the APP/PS1 model, adoptive transfer of Aβ-specific T
H1 cells, but not T
H2 or T
H17 cells, led to their homing to the brain and worsened AD pathology along with impaired cognitive function and exaggerated microglia activation [
117]. The pathogenic and plaque clearing roles of T
H1 cells were demonstrated in an earlier study, in which, post-Aβ vaccination, Aβ-specific IFNγ-producing T cells infiltrated the brain of J20 mice to clear amyloid plaques but induced meningoencephalitis, mirroring the pathogenic events of the failed human vaccination trial AN1792 [
126]. However, direct cerebrospinal injection of Aβ-specific T
H1 cells led to amyloid plaque clearance and increased neurogenesis in the absence of autoimmunity, implying an added impact of peripheral T
H1 cells in APP/PS1 model [
127]. On the other hand, adoptively transferred Aβ-specific T
H2 cells, while with no evidence of brain filtration, improved working memory of APP/PS1 mice in conjunction with reduced systemic inflammation and vascular amyloidosis [
128]. Therefore, Aβ-specific T
H1 and T
H2 effector subsets seemingly alter the Aβ pathology in distinct ways (Table
2).
Table 2
Aβ-specific T cell subsets regulate AD pathogenesis in experimental models
TH1 | Parenchyma | APP/PS1 | Aβ | Adoptively transferred cells increased microglia activation and Aβ deposition | |
TH1 | Parenchyma | J20 with Aβ peptide vaccination | Aβ | Migrated to Aβ plaques with increased clearance, while inducing transient meningoencephalitis | |
TH1 | Parenchyma | APP/PS1 | Aβ | Cells injected to cerebrospinal ventricle migrated to Aβ plaques, increased Aβ clearance and promote neurogenesis | |
TH2 | Outside the brain | APP/PS1 | Aβ | Adoptively transferred cells improved working memory, decreased microgliosis and reduced plasma cytokines. No effect on plaque load inside the parenchyma but reduced vascular amyloidosis | |
Substantial yet contradictory impacts of Tregs on AD pathogenesis have been demonstrated by recent studies in several experimental models (Table
3). By way of antibody-mediated depletion, adoptive transfer of purified cells, and low-dose IL-2-induced peripheral expansion, multiple groups independently provided experimental evidence that collectively suggested an overall protective role of Tregs in restoring memory deficits, reducing plaque load and decreasing microglia activation as well as inflammation in several APP transgenic models [
129‐
132]. These results are consistent with the impaired learning and memory in IL-2
-/- mice, which lack functional Tregs [
133]. Whether Tregs convey neuroprotection in an antigen-specific manner is not clear at this time.
Table 3.
Treg cells regulate AD pathogenesis in experimental models
Treg | Systemic | APP/PS1 | bulk | Transient depletion of Treg accelerated cognitive decline; increased Treg with low-dose IL-2 treatment restored cognitive functions | |
Treg | Systemic | 3xTg | bulk | Adoptively transferred cells improved cognitive functions and reduced Aβ deposition; long-term Treg depletion resulted in exacerbated spatial learning deficits, Aβ plaque load and microgliosis | |
Treg | Systemic and Parenchyma | App/PS1 ΔE9 | bulk | AAV-IL-2 expression within the brain induced Treg expansion and astrocyte activation, reduced Aβ plaque and improved synaptic plasticity and spine density | |
Treg | Systemic | AβPPswe/PS1 Δ E9 | bulk | Adoptively transferred cells improved cognitive function, while reducing Aβ deposition, microgliosis and systematic inflammation | |
Treg | Systemic | 5xFAD APP/PS1 | bulk | Transient depletion or pharmacological inhibition of Treg lead to Aβ plaque clearance, reduced neuroinflammation and reversal of cognitive decline. It affected CP with increased recruitment of peripheral monocytes and Tregs to Aβ plaque | |
Treg | Systemic | 5xFAD | bulk | Anti-PD1 treatment stimulated IFNƔ-dependent systematic immune response, which resulted in the recruitment of peripheral monocytes and Tregs to Aβ plaque, clearance of plaque, and improvement of cognitive performance. Repeated treatments maintained a long-lasting beneficial effects | |
Treg | | ThyAPP/PS1m146L ThyAPP/PS1A246E PD-APP | bulk | Anti-PD1 treatments had no effect on amyloid pathology nor induced infiltration of peripheral monocyte into the brain | |
On the other hand, two consecutive reports from Michal Schwartz’s group strongly argued for detrimental effects of Tregs in AD pathogenesis [
125,
134]. They first demonstrated that 5xFAD mice have reduced expression in choroid plexus of molecules critical for transepithelial migration, which correlated with the loss of IFNγ signaling, a condition similar to brain aging [
125]. This led to a hypothesis that a dysregulated CP exacerbates AD pathogenesis by blocking protective immune cell infiltration. Since Tregs are the primary suppressor of IFNγ-producing T
H1 cells, depleting or disabling Treg function was shown to restore IFNγ signaling in CP, increase the number of peripheral immune cells in the parenchyma, clear amyloid plaques, and rescue memory and behavior defects in 5xFAD mice [
125,
134]. In more detail, Baruch et al. perturbed the Treg population by various experimental schemes, which included selective depletion in 5xFAD:FoxP3-DTR mice with diphtheria toxin, chemical treatments to boost or diminish the functionality of Tregs or stimulate their induction, and administrating antibody against PD1 to break Treg suppression in 5xFAD and/or APP/PS1 models (Table
3).
At this time, it is unclear whether these directly opposing results reported on Tregs are affected by the genetic background of the AD models or other factors. Nevertheless, a T cell-based intervention with anti-PD1 antibody, which is largely analogous to the cancer checkpoint immunotherapy, was proposed as a translatable approach of AD therapeutics based on the findings by Baruch et al. [
135]. However most recently, a joint study by three pharmaceutical companies questioned the effectiveness of systemic PD-1 blockade in modifying amyloid burden in several AD mouse models, where the authors also failed to detect monocyte infiltration into the brain ([
136]; Table
3). Hence, the mechanism of how Tregs modulate Aβ pathology remains to be fully elucidated and the pre-clinical support for PD-1 blockade approach in AD patients is weak at this time.