Insights into T‐cell dysfunction in Alzheimer's disease

Abstract T cells, the critical immune cells of the adaptive immune system, are often dysfunctional in Alzheimer's disease (AD) and are involved in AD pathology. Reports highlight neuroinflammation as a crucial modulator of AD pathogenesis, and aberrant T cells indirectly contribute to neuroinflammation by secreting proinflammatory mediators via direct crosstalk with glial cells infiltrating the brain. However, the mechanisms underlying T‐cell abnormalities in AD appear multifactorial. Risk factors for AD and pathological hallmarks of AD have been tightly linked with immune responses, implying the potential regulatory effects of these factors on T cells. In this review, we discuss how the risk factors for AD, particularly Apolipoprotein E (ApoE), Aβ, α‐secretase, β‐secretase, γ‐secretase, Tau, and neuroinflammation, modulate T‐cell activation and the association between T cells and pathological AD hallmarks. Understanding these associations is critical to provide a comprehensive view of appropriate therapeutic strategies for AD.


| INTRODUC TI ON
Alzheimer's disease (AD), which usually progresses to dementia, is the most prevalent neurodegenerative disease in older adults (Hardy & Higgins, 1992). Symptoms of the disease manifest mainly as deficits in cognitive function, including memory loss, impairment of language, misidentifications, and behavioral disturbances (Burns et al., 2002). Neuropathologically, AD is remarkably characterized by two proteinaceous aggregate hallmarks, including tauhyperphosphorylation-induced intracellular neurofibrillary tangles (NFTs) and extracellular depositions of amyloid plaques induced by beta-amyloid peptide (Aβ) (Ittner & Gotz, 2011), which both contribute to synaptic damage and neuronal loss.
In the past few decades, various groups have devoted enormous efforts to explore AD pathogenesis and find key risk factors for the prevention and treatment of AD. However, it is disappointing that almost all AD-related clinical trials to date have failed to reverse cognitive decline and/or brain pathology. Undoubtedly, it is gratifying that there are different reasonable evidence-based hypotheses relating to underlying causes, such as the amyloid cascade hypothesis, the tau hypothesis, the Apolipoprotein E (ApoE) hypothesis, and the neuroinflammation hypothesis (Jiang et al., 2014; C. C. Lue et al., 2001;Morales et al., 2014;L. B. Yang et al., 2000).
Although it is widely accepted that neuroinflammation in AD is mediated by microglia and astrocytes, mounting evidence shows that T cells are involved in regulating the inflammatory response in AD through, but not limited to, the following two aspects (Figure 1).
First, during AD progression, the permeability of the blood-brain barrier (BBB) gradually increases due to decreased expression of the tight junction molecules ZO1 and occludin in vascular endothelial cells (Carrano et al., 2011;Cheng et al., 2014;Marco & Skaper, 2006). In addition, there is elevated peripheral T-cell expression of chemokine receptors, such as C-C motif chemokine receptor type 2 (CCR2), C-C motif chemokine receptor type 5 (CCR5), and C-X-C motif chemokine receptor 2 (CXCR2) (Goldeck et al., 2013;Liu et al., 2010;Town et al., 2005a). Both of these abnormal changes promote T-cell penetration into brain parenchyma. More important, activated CD8+ T and CD4+ T cells, the two major T-cell subsets, are neurotoxic and can induce substantial neuronal death through mechanisms dependent on cell-cell contact involving Fas ligand (FasL), lymphocyte function-associated antigen-1 (LFA-1), and CD40 (Giuliani et al., 2003). Increased T-cell infiltration also promotes crosstalk between T cells and microglia in a process dependent on major histocompatibility complex (MHC) class II, leading to further acceleration of neuroinflammation (Schetters et al., 2017). T-cell-derived cytokines can also impact local astrocyte-expressed chemokine function in inflammatory and neurodegenerative diseases (Williams et al., 2020).
Second, peripheral T cells can modulate glial cell neuroinflammation mediated by the release of proinflammatory factors into the central nervous system (CNS), such as interferon gamma (IFNγ) (Town et al., 2005b). Activation of peripheral T cells in AD patients is elevated compared to healthy controls, and these T cells further promote peripheral blood mononuclear cells (PBMCs) to release proinflammatory factors including IL-6, tumor necrosis factor α (TNFα), and IL-1, all of which are essential immune modulators in AD pathology (Mietelska-Porowska & Wojda, 2017;Tan et al., 2002).
Despite accumulating evidence suggesting that T cells participate in immunological and pathological stages of AD concomitant with changes in T-cell phenotype, the mechanism of T-cell abnormality in AD remains unknown. In this review, we summarize T-cell dysfunction in AD and demonstrate the association between T cells and the critical pathological features of AD. Furthermore, we propose that aberrant activation of T cells contributes to AD pathogenesis, and that the critical elements of AD can also mediate biological processes involving T cells.

| T-CELL AB NORMALITIE S IN AD
Although the mechanisms by which T cells contribute to AD pathophysiology are unclear, considerable work has shown that normal T-cell function and T-cell markers in both AD mouse models (Table 1) and AD patients are different when compared to measures from respective control groups (Table 2). As early as 1981, for example, Miller and colleagues first demonstrated that concanavalin A (ConA)-induced T-cell suppression in AD patients was significantly elevated and lymphocyte proliferation was lower when compared with an elderly control group (Miller et al., 1981). However, another group found no difference in phytohaemagglutinin (PHA)-induced T-cell proliferation between T cells derived from AD patients and those from age-matched controls (Leffell et al., 1985). This discrepancy may be accounted for by different types or intensities of mitogen stimulation used. Although interpretation of mitogen-induced T-cell activation is debated due to such discrepancies, variation in T-cell populations in peripheral blood from both AD mouse models and AD patients appears convincing, as described below. A higher ratio of CD4+/CD8+ T cells was found in the in peripheral blood of AD patients and AD mouse models compared to relevant controls, concomitant with lower CD8+ T-cell counts and lower total CD3 expression (Hu et al., 1995;Pirttila et al., 1992;Schindowski et al., 2007;Shalit et al., 1995;St-Amour et al., 2014, 2019. Furthermore, analyses of peripheral blood from AD patients have shown that the number of IL-2R+, HLA-DR+, CD25+, and CD28+T cells was significantly higher than controls (Ikeda et al., 1991;Lombardi et al., 1999;Speciale et al., 2007), indicating an immune response in the peripheral system. Similarly, a consistent finding is that the numbers of CD4+ and CD8+ T cells in the brain parenchyma and cerebrospinal fluid (CSF) of AD patients are significantly higher than normal, with CD8+ T cells having the advantage over CD4 + T cells in absolute numbers, and that both subtypes exhibit the CD45RA-CD45RO+ F I G U R E 1 Summary of T-cell roles in AD. With the development of AD, T-cell infiltration into the brain increases, and a large number of inflammatory cytokines derived from T cells in the peripheral blood also enter the brain, which eventually exacerbate neuroinflammation and accelerate neuronal death phenotype (Ferretti et al., 2016;Laurent et al., 2017;Merlini et al., 2018;Rogers et al., 1988;Togo et al., 2002), indicating that infiltrating T cells were activated and may be cytotoxic.
In addition to apparent changes in the number and proportion of AD-derived T cells, alterations have also been observed in the intracellular signaling pathway in T cells from AD patients, such as changes in calcium response. For example, there are in vitro data which show that both baseline cytosolic calcium and PHA-induced calcium responses in T cells from AD patients were higher than measures from control groups (Adunsky et al., 1995;Sulger et al., 1999).
However, beta-amyloid fragment Aβ25-35 led to a substantial reduction in mitogen-induced calcium signaling rise in the PBMC of AD patients compared with that of age-matched controls (Eckert et al., 1995). Therefore, the precise mechanism of calcium homeostasis in T cells may depend on the specific type of stimulus and the local microenvironment but may not rely only on AD status.
T cells from AD patients and AD mouse models are hyperactive to Aβ stimulation, which increases expression of activation marker CD69 and enhances cytokine production (Ciccocioppo et al., 2008;Miscia et al., 2009;Pellicano et al., 2010). Cytokines secreted by activated T cells are significant modulators of microglia and astrocyte function in AD. For example, IL-10, TNFα, IL-6, IL-1β, and monocyte chemoattractant protein-1 (MCP-1) released from T cells are markedly elevated in AD patients (Lombardi et al., 1999;Man et al., 2007), and microglia and astrocytes are the main targets of these inflammatory factors (Hanisch, 2002;Ramesh et al., 2013). Thus, dysregulation of T cells which overexpress these cytokines is potentially harmful and likely contributes to chronic neuropathology in AD.
Furthermore, proinflammatory cytokines may be responsible for the prevalence of elevated Th1 and Th17 cells in AD Oberstein et al., 2018;Saresella et al., 2011;Zhang et al., 2013). Th1 and Th17 cells are two major proinflammatory T-cell subtypes which are typically elevated in neurodegenerative diseases, including AD. In addition, Th1 cells are significant sources of IFNγ secretion. At the same time, microglia and astrocytes can be activated by IFNγ and disturb cell homeostasis, thereby contributing to Aβ deposition, impaired synaptic plasticity, and acceleration of cognitive deficits in APP/PS1 mice (Browne et (McGeer et al., 1988) 10 AD Increased the presence of CD4+ and CD8+ T cells in capillaries of the brain (CD8+ T much more prevalent then CD4+ T) 5 age-matched controls (Rogers et al., 1988) 10 AD Increased the presence of CD4+ and CD8+ T cells in the brain parenchyma and blood vessels 6 age-matched controls (Leonardi et al., 1989) 26 AD Increased T-cell proliferative response in AMLR 2013). However, injection of Aβ-specific Th1 cells into 5xFAD mice enhances Aβ uptake due to T-cell-activation-induced expansion of brain-endogenous MHCII+microglia, which exhibit stronger phagocytic activity (Mittal et al., 2019). These findings suggest that the function of Th1 cells in AD may be linked to disease stage. Therefore, manipulation of Th1 cells at an appropriate period may be a helpful approach for AD therapy.
Alternatively, elevated Th17 cells in AD may also be detrimental and induce neuronal apoptosis (Zhang et al., 2013), through the release of proinflammatory factors such as FasL, IL-17, and IL-22.
In contrast to the destructive function of Th17 cells, the adoptive transplantation of Th2 cells (one of the immunosuppressive T-cell subtypes) into APP/PS1 AD mice benefits cognitive function and reduces pathological features (Cao et al., 2009). The function of regulatory T cells (Tregs, another immunosuppressive T-cell subtype) has also been described in AD. Transient depletion of Tregs in 5xFAD mice is beneficial for Aβ clearance and cognitive function by affecting the choroid plexus, which regulates the recruitment of immunoregulatory cells such as monocyte-derived macrophages and Tregs into cerebral pathological sites (Baruch et al., 2015). A consistent finding is that IL-10 (the main Tregs effector cytokine) signaling transduction is accelerated in the brains of AD patients, and IL-10deficient APP/PS1 mice show restricted cerebral amyloidosis and less cognitive decline than controls via a rebalancing of abnormal innate immunity (Guillot-Sestier et al., 2015). This result suggests that aberrant elevated IL-10 signaling in AD patients may be a therapeutic target for AD.
In summary, T cells display abnormal phenotypes and dysfunction in AD, and transplantation or deletion of different T-cell subtypes into AD mice has the potential to alter the progression of AD pathology. Accordingly, AD treatment strategies targeting T cells have been proposed (Table 3). However, it is still unclear whether

TA B L E 2 (Continued)
these changes in T cells are related to disease progression. Thus, the critical question is whether abnormal T-cell parameters are driving AD progression or do the abnormalities, including the lack of different phenotypes, occur only after onset of AD. Future work is needed to clarify the link between T-cell abnormalities and disease severity.
In particular, analyses of these abnormalities at different AD stages would be beneficial. These results should provide immunologybased guidance for treating AD.

| ApoE modulates T-cell activation
ApoE is a polymorphic protein involved in lipoprotein conversion and metabolism, produced by organs such as the brain, liver, kidneys, and spleen (Huang & Mahley, 2014). ApoE4, one of the protein isoforms of ApoE, interacts with Aβ more efficiently than ApoE3, which results in increased Aβ deposition and amyloid plaques in AD (Sanan et al., 1994;Schmechel et al., 1993;Strittmatter et al., 1993).
Recent evidence revealed that ApoE4 can lead to BBB dysfunction in cognitively unimpaired individuals and yet more severe dysfunction in cognitively impaired individuals, independently of CSF Aβ and tau status (Montagne et al., 2020).
It has been shown that plasma lipoproteins containing ApoE have a role in inhibiting T-cell activation and proliferation induced by PHA in vitro in an ApoE-concentration-dependent manner by downregulating the secretion of bioactive IL-2 (Kelly et al., 1994;Macy et al., 1983). Consistent findings suggest that lack of ApoE exacerbates the production of proinflammatory factors including TNFα, IFNγ, IL-12, and IL-6 during LPS-induced responses, whereas treatment with exogenous ApoE can normalize these cytokines levels (Ali et al., 2005).
Similarly, in another neuroinflammation mouse model induced by ApoE deletion, an ApoE mimetic peptide reversed upregulated expression of proinflammatory factors (IL-17, IL-12, TNFα, IFNγ, IL-6, and IL-1β) (Wei et al., 2013). Moreover, two major proinflammatory T-cell subtypes, Th1 and Th17 cells, were elevated in this model, and IL-17 levels secreted by Th17 cells were elevated, promoting mononuclear cell infiltration and activation . Consistent with these results, treatment with IL-17 antibody led to a significant amelioration of atherosclerotic symptoms (Smith et al., 2010).
ApoE also disturbs the balance of Th17 and Treg in the spleen during arteriosclerosis; however, the mechanism remains unknown (Xie et al., 2010).

| Aβ regulates T-cell activation
Amyloid precursor protein (APP) is sequentially cleaved by βsecretase and γ-secretase in an amyloidogenic pathway to produce Aβ, which aggregates into amyloid plaques (Cole & Vassar, 2007;Selkoe, 2001;Vassar, 2004). It is clear that the level of Aβ in the brain is significantly elevated during the progression of AD, which ultimately results in neuronal death and inflammation.
Autoantibodies to Aβ have been detected and were found to be elevated in both AD patients and elderly AD mice, which suggests that Aβ can act as a self-antigen to initiate humoral immune response (Nath et al., 2003). Moreover, Aβ-reactive T cells were also detected in the peripheral blood of AD patients (Monsonego et al., 2013). The presentation of peripheral Aβ to T cells is typically detected in lymph glands; however, it has been reported that antigen-presenting cells (APCs) can present Aβ to T cells infiltrating the parenchyma, although it is unknown why T cells recognize selfantigen Aβ (Archambault et al., 2005).
The role of Aβ-reactive T cells is complicated and controversial.
On the one hand, it has been shown that Aβ-reactive T cells in certain AD mouse models are beneficial for Aβ clearance via enhancement of microglial activation in an IFNγ-dependent manner (Fisher et al., 2010;Monsonego et al., 2006). On the other hand, Aβ-reactive T cells may be detrimental because they promote pathogenic immune responses in AD. These abnormal T cells lead to strong secretion of proinflammatory factors, including TNFα, IL-1β, and IL-6, contributing to chronic neuroinflammation and neurotoxicity (Mietelska-Porowska & Wojda, 2017). Although the beneficial and detrimental effects of Aβ-reactive T cells in Aβ pathology are still not fully elucidated, it has well accepted that Aβ as a specific antigen can be captured by APCs and then be recognized by T cells to induce T-cell activation and proliferation. These processes are tightly related to AD pathology.
In addition to being presented as a specific antigen to T cells, Aβ may have a more direct effect on the regulation of T-cell function; evidence had revealed that T cells can synthesize and secret APP upon activation to initiate the immune response (Bullido et al., 1996;Monning et al., 1990Monning et al., , 1992. Consistently, the lymphoblastoid cell line established from familial Alzheimer's disease (FAD) expresses higher APP than control patients (Matsumoto & Fujiwara, 1991).
Therefore, T cells from FAD may express a higher level of APP.
Furthermore, with the characteristics of a cell adhesion molecule, APP can bind to extracellular matrix components including collagen and laminin (Sondag & Combs, 2006), implying a role for APP in cell adhesion, cell-extracellular matrix contact, or cell-cell contact during T-cell recruitment and infiltration.
Strikingly, data obtained in vitro have shown that synthetic APP peptides stimulate the proliferation of resting lymphocytes from young and old healthy donors, which correlates with IL-2 expression (Trieb et al., 1996). It is a consistent finding that Aβ stimulation in vitro also significantly enhances T-cell proliferation derived from AD patients and healthy elderly individuals (Jozwik et al., 2012).

| Alpha secretases regulate T-cell function
Alpha secretases (α-secretases) are members of the ADAM (a disintegrin and metalloproteinase) family, which cleave within the Aβ peptide to produce sAPPα and C83 in the non-amyloidogenic pathway (Zhang et al., 2011). sAPPα, but not secreted ectodomain APPβ (sAPPβ), protects neurons against Aβ-induced cytotoxicity and is thought to be a neurotrophic and neuroprotective factor (Tackenberg & Nitsch, 2019). Therefore, α-secretases can facilitate AD prevention, not only by competitive cleavage of APP to preclude the formation of Aβ peptide but also by providing neuroprotective agents. Additionally, decreased activity of α-secretase was observed in AD patients (Colciaghi et al., 2002;Kim et al., 2009). Therefore, pharmacological intervention targeting α-secretase may provide a potential therapy for AD.
ADAM9, ADAM10, and ADAM17 (tumor necrosis factorα-converting enzyme, TACE) have been identified as having αsecretase activity (Buxbaum et al., 1998;Lammich et al., 1999). It has recently been shown that ADAM9 drives Th17-cell development by producing bioactive transforming growth factor β1 (TGFβ1), and that T cells lacking ADAM fail to induce Th17-dependent experimental autoimmune encephalomyelitis (Umeda et al., 2021). Besides, ADAM10 and ADAM17 were shown to regulate T-cell function via the cleavage of lymphocyte activation gene 3 (LAG3), which must be cleaved to allow normal T-cell activation . ADAM10 and ADAM17 were also identified as major sheddases of T-cell im-  (Umeda et al., 2021). Indeed, further analysis revealed that the transcription factor inducible cAMP early repressor (ICER) can directly bind to ADAM9 promoter and to promote ADAM9 transcription during T-cell activation (Umeda et al., 2021). Moreover, TCR signaling transduction can also induce enzymatic activity of ADAM17 in a PKCθ-dependent manner . Given that ADAM exhibits neuroprotective properties in AD, T-cell-induced modulation of ADAM activity could be an alternative target for AD treatment.

| T-cell function is associated with β-secretase
Beta-secretase, known as β-site amyloid precursor protein cleaving enzyme 1 (BACE1), cleaves the extracellular domain of APP to produce Aβ. The concentration and enzymatic activity of BACE1 in the CSF and blood of AD patients are significantly higher than that of control participants, indicating that BACE1 is a promising candidate biological marker of AD (Ewers et al., 2011;Hampel & Shen, 2009;Shen et al., 2018). Based on the amyloid hypothesis, Aβ is considered one of the leading potential causes of AD, so inhibition of BACE1 to reduce the production of Aβ is considered to be an effective strategy for AD treatment (Yan & Vassar, 2014 (Meakin et al., 2018).
In addition to the nervous system and hepatic metabolism, there are also immune-system-related substrates of BACE1. For example, it has been shown that P-selectin glycoprotein ligand-1 (PSLG-1) is a substrate for BACE1 and the cleavage site has been identified using mass spectrometry (Lichtenthaler et al., 2003). It is expressed in endothelial cells and leukocytes, including T cells, and binds to Lselectin, E-selectin, and P-selectin to mediate monocyte adhesion during inflammation (da Costa Martins et al., 2007;Moore, 1998).
Surprisingly, recent studies identified that CD4+ T cells highly expressed BACE1, and BACE1 mediated T-cell activation in the EAE and AD mouse model Hernandez-Mir et al., 2019).
In conclusion, these findings implicate an unexpected relationship between BACE1 and T cells. This relevance is bilateral because, on the one hand, changes of T-cell-related activity regulate BACE1 expression and activity. On the other hand, BACE1 may modulate T-cell function via cleavage of various substrates expressed on leukocytes, including T cells.

Mouse model Findings
Over past decades, it was elucidated that the essential function of Notch signaling is mediated by γ-secretase in the biological processes of T cells. Treatment with γ-secretase inhibitors or deletion of Notch in hematopoietic progenitors and common lymphoid precursors impair the development of T cells (Hadland et al., 2001;Radtke et al., 1999;Wilson et al., 2001); conversely, retroviruses can induce continuous expression of Notch1 in hematolymphoid progenitors leading to thymic-independent T-cell development (Pui et al., 1999).  Graham et al., 2007), has been identified as a novel substrate of γsecretase in vitro (Lammich et al., 2002), implying a novel mechanism of γ-secretase in the regulation of T-cell function.
In summary, the role of γ-secretase in most biological T-cell processes suggests that γ-secretase inhibitors should be used with caution to avoid affecting normal physiological T-cell function.
Conversely, γ-secretase inhibitor treatment may be an effective strategy for the rescue of abnormal T-cell function under certain conditions.

| T cells contribute to Tau pathology
Tau is a microtubule-binding protein abundant in neurons and is mainly localized on axons and dendrites (Hirokawa et al., 1996;Ittner et al., 2010;Konzack et al., 2007;Utton et al., 2002). The primary role of tau in neurons is to modulate the stability of axonal microtubules by interacting with tubulin. While phosphorylation is a common post-translational modification of tau (Cleveland et al., 1977;Ksiezak-Reding et al., 2003;Sengupta et al., 1998), abnormal hyperphosphorylation of tau leads to NFT formation and is neurotoxic in neurodegenerative diseases, including AD (Kenessey et al., 1995;Kopke et al., 1993;Ksiezak-Reding et al., 1992). Under pathological conditions, excessive or abnormal phosphorylation of tau which aggregates into NFTs is responsible for synaptic dysfunction and neuronal death (Guo et al., 2017).
Tau-specific CD4+ T cells are widely distributed in the peripheral blood from the general population (Lindestam Arlehamn et al., 2019).
These cells collected from either young or healthy elderly donors exhibited reactivity to tau-and p-tau-derived peptides associated with IL-5 and IFNγ. The presence of the tau-autoreactive T cells indicates that the thymic selection of CD4+ T cells is not sufficient to eliminate these cells. Furthermore, extravascular T cells are observed in the brains, specifically in the hippocampus, of AD patients, most of which are CD8+ T cells. These extravascular T cells are correlated with tau pathology but not with Aβ pathology (Merlini et al., 2018), suggesting that T cells could be critical for driving the tau-dependent phase of the AD pathology.
Interestingly, T-cell infiltration has been discovered positively correlated with p-tau load in the inferior parietal lobule, middle temporal gyrus, and medial frontal gyrus of AD patients (Zotova et al., 2013). In addition, T-cell infiltration was also observed in the cortex of frontotemporal dementia patients with P301L tau mutation and the hippocampus of THY-Tau22 mouse model (Laurent et al., 2017).
These data support an instrumental role of tau-driven pathophysiology in brain T-cell infiltration. However, the underlying molecular mechanisms are not clear.
All these findings together suggest that T cells can respond to, as well as contribute to, tau-driven pathology. Given the high levels of tau aggregation in the elderly, T-cell responses to tau may contribute to the progression of neurological diseases, including AD and PD. As tau-driven pathology in AD induces synaptic loss, cytoskeletal dysfunction, and impair axonal transport, tau-targeted therapeutic strategies have been proposed. Immunization against tau has shown great potential to treat tau pathology by inhibiting tau transmission and aggregation (Kfoury et al., 2012;Yoshiyama et al., 2013). Furthermore, it has been demonstrated that passive immunization may be more effective and safer than active immunization (Spillantini & Goedert, 2013). It is worth pointing out that it is not known whether tau-targeted treatment is involved in the regulation of T-cell response and whether this occurs without breaking down the functional immune homeostasis.

| T-cell activity is associated with neuronal loss
Brain atrophy caused by neuronal and synaptic loss is one of the definitive pathological lesions observed in AD. Several neuronal death mechanisms have been determined in AD (Cotman & Su, 1996;Niikura et al., 2006). Substantial evidence suggests that Aβ plays a significant role in initiating neurotoxicity and neuronal dysfunction and results in neuronal death. Accumulated Aβ-initiated toxicities are characterized by mitochondrial dysfunction, oxidative stress, and calcium dyshomeostasis in neurons (Canevari et al., 2004;Caughey & Lansbury, 2003;LaFerla, 2002). Aβ also alters the acetylcho- cyclin-dependent kinase-5 (Cdk5) activation, which thereafter mediates neuronal apoptosis via induction of p53 phosphorylation (Lapresa et al., 2019). In addition to the toxic effects of Aβ, inhibition of the proteasome is sufficient to induce neuronal apoptosis by an increase in poly-ADP-ribosylation, elevated activation of caspase-3like proteases, and accelerated amylospheroid (Keller & Markesbery, 2000;Komura et al., 2019;Qiu et al., 2000). Widespread proinflammatory factors in the brain such as TNFα, IFNγ, and IL-1β also have significant detrimental effects on neurons (Barker et al., 2001;Brown & Neher, 2010;Combs et al., 2001;Rothwell, 2003).
The migration of T cells into the CNS parenchyma during the pathological process of AD has attracted attention due to the crosstalk between neurons and T cells. Nitsch and colleagues were first to present the process of direct physical contact between neurons and T cells in living brain tissue (Nitsch et al., 2004). Notably, they showed  (Liu et al., 2017).
Most importantly, T cells are indispensable for spatial learning and the maintenance of neurogenesis under physiological situations in adulthood (Ziv et al., 2006). More work is required to discriminate the contributions from different T-cell subtypes to neurons in neurodegenerative disorders and to determine their pathogenic and neuroprotective properties. Understanding the particular effects of T cells in different diseases may provide important information to be carefully considered when developing therapeutic strategies.

| T cells contribute to neuroinflammation
Pathogenesis of AD is not limited to neuronal loss but also extends to extensive glial cell activation (Lee & Landreth, 2010 et al., 1996). In AD, neuroinflammation is a common phenomenon activated by amyloid plaques and NFTs. It contributes to pathogenesis just as much as plaques and tangles, perhaps even more so, and there is evidence to suggest that neuroinflammation is a critical modulator of AD development (Heneka et al., 2015;Jiang et al., 2014;R. Li et al., 2004;Van Eldik et al., 2016;L. Yang et al., 2002).
Microglia, the most abundant resident innate immune cells in the brain, are vital cellular mediators for initiating neuroinflammatory responses. Various immune-related receptors are expressed on microglia cell membranes, including scavenger receptors, chemokine receptors, cytokine receptors, and pattern-recognition receptors (PRRs), which can bind with proinflammatory mediators to trigger microglial activation (Kierdorf & Prinz, 2013). Furthermore, Genome-Wide Association Studies (GWAS) have also determined many gene mutations related to an elevated risk of late-onset AD (LOAD) and most are expressed abundantly in microglia, such as triggering receptor expressed on myeloid cells 2 protein (TREM2), complement receptor type 1 (CR1) and CD33 (Karch & Goate, 2015).
Recently, high-throughput sequencing methods at the single-cell level (scRNA-seq) have been applied to study microglial function and heterogeneity throughout microglial lifespan and AD (Hammond et al., 2019;Keren-Shaul et al., 2017;Q. Li et al., 2019;Mathys et al., 2017;Zeisel et al., 2015). Mathys and colleagues first used the scRNA approach to track activation of microglia during the neurodegenerative processes in CK-p25 mice (Mathys et al., 2017), which mimic the major pathological hallmarks of AD, such as impaired synaptic plasticity, upregulated Aβ, and neuronal death. They found that proinflammatory factors TNFα and macrophage migration inhibitory factor (MIF) were upregulated in the early microglial response state (1-week post p25 induction), suggesting that inflammation occurs as early as Aβ production and may initiate cascading effects that ultimately lead to neuronal loss and cognitive dysfunction.
Astrocytes are another key regulator of neuroinflammation. During neuroinflammation, destructive signaling pathways in astrocyte are triggered by IL-17, sphingolipids, and LacCer, followed by activation of the NF-κB-dependent and STAT3-dependent transcription of proinflammatory factors, which finally contributes to neuroinflammation and promotes neurodegenerative disorders.

It is clear that T cells infiltrate the CNS and promote neuroin-
flammation during the pathogenesis of AD (Hoppmann et al., 2015;Laurent et al., 2017;Mietelska-Porowska & Wojda, 2017;Raveney et al., 2015;Wimmer et al., 2019;Yu et al., 2015). Th1 and Th17 cells significantly accumulate in the brains of APP/PS1 mice; however, only Aβ-specific Th1 cells adoptively transplanted to APP/PS1 mice lead to a deficit in cognitive function . These Aβspecific Th1 cells promote microglia activation and neuroinflammation via IFNγ production, and administration of a neutralizing IFNγ antibody reverses the outcomes of Th1 cells on microglia activation and Aβ deposition. In addition, microglial cells co-cultured with Aβspecific Th1 cells or Aβ-specific Th17 cells induce proinflammatory cytokine production in microglia, which can be attenuated by Th2 cells (McQuillan et al., 2010), indicating that the regulation of microglia activation by T cells occurs in a cell-type-dependent manner.
Surprisingly, T cells, especially brain-resident CD4+ T cells, can regulate microglia activation and are required for microglia maturation (Pasciuto et al., 2020). The absence of CD4+ T cells traps microglia in a fetal-like transcriptional state and results in defective synaptic pruning and depression-like mouse behavior. Besides, T cells can positively contribute to astrocyte activation and then exacerbate neuroinflammation. T-cell-derived IFNγ induces astrocyte proliferative response in vitro and promotes brain reactive astrogliosis (Yong et al., 1991). Moreover, IL-17, produced by Th17 cells, has been repeatedly identified as an effective astrocyte stimulator. IL-17 stimulates inducible nitric oxide synthase activation (Trajkovic et al., 2001), regulates macrophage inflammatory proteins-1α (MIP-1α) expression via PI3K/Akt and NF-κB pathways (Yi et al., 2014), and enhances the IL-6 signaling pathway  in astrocytes.
In addition to activating microglia and astrocytes, T cells may also promote brain inflammation by inducing myeloid cells, including dendritic cells (DCs) and macrophages associated with secretion of TNFα, IL1β, and IL-6 (Town et al., 2005b). More critically, T cells not only contribute to neuroinflammation but also initiate neuroinflammation in an MHCII-dependent manner. In this way, conventional DCs process and present myelin antigen to parenchymal T cells and then trigger T cells to infiltrate the CNS to initiate neuroinflammation (Mundt et al., 2019).
Of particular note, T cells also exhibit neuroprotective properties by regulating the trophic/cytotoxic glial cell balance and restoring glial cell activation (Beers et al., 2008). CD4+ T-cell-derived IL-10 and IL-4 are two immunoregulatory factors with important neuroprotective properties, involving the inhibition of microglia with subsequent reduction of nitric oxide and TNFα levels (Chao et al., 1993;Frenkel et al., 2005).
Summarizing, these results highlight T cells as an essential modulator in mediating neuroinflammation, which is achieved by activating microglia and astrocytes and releasing proinflammatory factors, implicating T cells as a potential immunotherapy target for neuroinflammation in neurodegenerative disease.

| CON CLUDING REMARK S
In this article, we first reviewed the abnormal behavior of T cells in the progression of AD. Although the significance of T cells in AD pathogenesis is still hotly debated, there is convincing evidence from pre-clinical, epidemiological, and genetic studies which indicate that the immune system is closely involved in AD and that T cells con- In conclusion, we summarized that AD risk factors and hallmarks can modulate T-cell function, and abnormal activation of T cells in AD can also act on these critical factors, ultimately exacerbating AD pathology ( Figure 2). Thus, understanding these causal associations may provide important insights into developing effective therapeutics. We also proposed that targeted treatment based on these risk factors and hallmarks may cause changes in the normal T-cell phenotypes and peripheral immune responses under the physiological state. Therefore, the critical question is how to identify and limit the potential side effects of AD-related factor-based therapies on the normal T-cell function.

CO M PE TI N G I NTER E S TS
All authors declare that they have no competing financial interests.

AUTH O R CO NTR I B UTI O N S
Linbin Dai and Yong Shen made substantial contributions to the review including writing the manuscript, preparing the figures and tables, and discussing the content.

DATA AVA I L A B I L I T Y
The data that support the findings of this study are available from the corresponding authors upon reasonable request.