Neuroimmune nexus of depression and dementia: Shared mechanisms and therapeutic targets

Abstract

Dysfunctional immune activity is a physiological component of both Alzheimer's disease (AD) and major depressive disorder (MDD). The extent to which altered immune activity influences the development of their respective cognitive symptoms and neuropathologies remains under investigation. It is evident, however, that immune activity affects neuronal function and circuit integrity. In both disorders, alterations are present in similar immune networks and neuroendocrine signalling pathways, immune responses persist in overlapping neuroanatomical locations, and morphological and structural irregularities are noted in similar domains. Epidemiological studies have also linked the two disorders, and their genetic and environmental risk factors intersect along immune‐activating pathways and can be synonymous with one another. While each of these disorders individually contains a large degree of heterogeneity, their shared immunological components may link distinct phenotypes within each disorder. This review will therefore highlight the shared immune pathways of AD and MDD, their overlapping neuroanatomical features, and previously applied, as well as novel, approaches to pharmacologically manipulate immune pathways, in each neurological condition.

Linked Articles

This article is part of a themed section on Therapeutics for Dementia and Alzheimer's Disease: New Directions for Precision Medicine. To view the other articles in this section visit http://onlinelibrary.wiley.com/doi/10.1111/bph.v176.18/issuetoc

Abbreviations

ACC

anterior cingulate cortex

AD

Alzheimer's disease

CAMs

cell adhesion molecules

CRF

cortisol releasing factor

CRF₁ receptor

corticotrophin‐releasing hormone receptor 1

dlPFC

dorsolateral prefrontal cortex

GR

glucocorticoid receptor

GSK

glycogen synthase kinase

HPA

hypothalamus pituitary axis

ICAM‐1

intercellular adhesion molecule‐1

IDO

indoleamine 2,3‐dioxygenase

KYN

kynurenine

LFA‐1

lymphocyte function associated antigen 1

MDD

major depressive disorder

MMSE

Mini‐Mental State Examination

NbM

Nucleus Basilis of Meynert

OFC

orbitofrontal prefrontal cortex

TSPO

translocator protein

VCAM‐1

vascular cell adhesion molecule 1

1. INTRODUCTION

Activation of the immune system affords the host protection against pathological infiltrates and tissue injury. Unabated immune responses, however, indicate immune dysfunction either by insufficient negative feedback or persistent stimulation. Brains from patients with both Alzheimer's disease (AD) as well as major depressive disorder (MDD) can each exhibit ongoing maladaptive physiological events of immunological origin (Heppner, Ransohoff, & Becher, 2015; Miller & Raison, 2016). The immune effectors involved in these attacks exhibit a large degree of overlap between the two disorders and reflect adaptive responses to pathogenic or stressful insults. Such pathways include the activation of neuroactive cytokines and complement proteins, dysregulated leukocyte function and trafficking, maladaptive neuroendocrine signalling, and the release of endogenous stress products, which together facilitate a permissive para‐inflammatory state (Medzhitov, 2008), detrimental to homeostatic processes. Disequilibrium in immune activity has consequential neurophysiological effects, which manifest in both AD and MDD subjects. These effects include impaired neuronal plasticity and metabolism (Kennedy et al., 2001; Mosconi, 2005), altered cerebral haemodynamics (Takano et al., 2005), and degeneration of function‐specific circuit architecture. In AD and MDD, these impairments can manifest in overlapping limbic and cortical circuits responsible for emotional processing and executive functions (Drevets, 2001; Matsuda, 2001). Moreover, susceptibility to either disorder increases with certain risk loci in immune‐related genes, and following the exposure to environmental stimuli that are known immunological instigators (Carpenter et al., 2010; Perry, Cunningham, & Holmes, 2007; Figure 1). This review will therefore briefly highlight the intersecting immune pathways noted in AD and MDD, their overlapping neuropathological features, and previously attempted, as well as prospective, methods to target immunological effectors in both disorders.

Figure 1.

Risk factors activate common phenotypes for MDD and AD. Activation of immune‐inflammatory pathways in MDD and AD is dependent upon common risk factors. The exposure to psychological stress, poor diet, infection, or autoimmune disorders are environmental stressors that combined with genetic polymorphisms in immune‐related genes support an immunological component in both disorders

2. EPIDEMIOLOGICAL ASSOCIATIONS BETWEEN MAJOR DEPRESSION DISORDER AND AD

Depression, defined as major depression by symptomatic criteria of the Diagnostic and Statistical Manual of Mental Disorders (DSM)‐5 (APA), is a common disorder within the population, with one in five individuals experiencing an episode during their lifetime (Kessler et al., 2005). Based upon a growing number of epidemiological studies, a clinical relationship has been proposed to exist between depression and dementia. This suggestion is based on meta‐analysis studies concluding depressive episodes act as an independent risk factor for the development of AD‐dementia (Ownby, Crocco, Acevedo, John, & Loewenstein, 2006). Studies must distinguish risk parameters of early life versus late life depression because depression is often co‐morbid with AD‐dementia. Thus, for late‐life depression, deciphering whether depression precedes dementia or acts as a prodromal condition within the progression of dementia can be difficult. Indeed, other studies investigating temporal relationships between dementia and depression find increased risk for dementia is associated only with late‐life depression (Singh‐Manoux et al., 2017), or that depression acts solely as a prodromal component of dementia (Chen, Ganguli, Mulsant, & DeKosky, 1999). Some studies have found no relationship at all (Lindsay et al., 2002). These studies are contradicted by longitudinal studies of late‐life depression, which report that risk for developing dementia increases in proportion with the magnitude or number of depression symptoms (Barnes et al., 2012; Chen et al., 1999; Gatz, Tyas, St John, & Montgomery, 2005). In addition, longitudinal studies investigating the risk attributes of depression for up to 51 years found that depression was an independent risk factor for dementia, separate from prodromal dementia, and that early life episodes of depression can also exhibit a dosage relationship with dementia (Dotson, Beydoun, & Zonderman, 2010; Kessing & Andersen, 2004). If depression either increases the risk for dementia, or acts as a prodromal form, a common underlying mechanism may be responsible.

3. COMMON IMMUNE PATHWAYS

Exploring a potential a physiological explanation for the epidemiological link between AD and depression is complicated given their respective heterogeneities, as well as their complex genetic and environmental contributions. However, when considering their nongenetic risk factors, a number are associated with potent acute immunological responses. For example, autoimmune disorders and severe infections increase the risk to develop both mood disorders (Benros et al., 2013; Goodwin, 2011) as well as AD‐dementia (Nee & Lippa, 1999; Wotton & Goldacre, 2017). Psychological stress, well implicated in the aetiology of depressive episodes and capable of priming basal inflammatory activity throughout an individual's life (Carpenter et al., 2010; Steptoe, Hamer, & Chida, 2007), has been related recently to an increase for developing AD. Approximately a quarter of individuals who experience a significant life stressor develop depression (Van Praag, de Kloet, & van Os, 2004), while females who experience higher numbers of psychosocial stressors are at a higher risk to develop AD, but not vascular dementia (Johansson et al., 2010, 2013). Moreover, cardiovascular disease and diabetes, two conditions with significant inflammatory components, are often co‐morbid with depression and AD (Hao et al., 2015). Given the convergence of AD and MDD risk factors on inflammatory inducers, defining the types of and effectors in immune responses evoked in each disorder, can help in identifying overlapping pathways (Figure 2).

Figure 2.

Common immune pathways in MDD and AD. Samples of human plasma, CSF, and post‐mortem tissue, as well as in vivo imaging from patients with AD and MDD find activation of similar immune pathways (b). Shared neuroimmune activity (A1) includes a heightened basal level of inflammatory cytokines and chemokines, extravasation of T lymphocytes and neutrophils into brain extravascular regions, increased expression of endothelial ligation proteins that mediate leukocyte extravasation, and enhanced microglia reactivity. Certain endocrine and neuropeptide signalling pathways that influence immune activity, in particular those of the HPA, are also dysregulated in both disorders. Furthermore, impaired central signalling in peripheral immuno‐metabolism, often regulated by neuroimmune and neuroendocrine inputs (A2) that are shared include altered increased plasma Th17 and CD4⁺ immunophenotypes concurrent with decreased immune sensitivity (M: microglia; P: plasma; T: tissue)

3.1. Shared cytokines and chemokine profile

Inflammatory cytokines are the primary effectors of innate immune activation in response to infection or tissue injury. They are meant to promote clearance of damaged tissue and resolve infections; their collateral effects, however, include decreased synaptic strength, altered neuronal excitability, recruitment of neurotoxic secondary immune cells, and increased synthesis of senile plaques (Ghosh et al., 2013; Prieto et al., 2015). Comprehensive meta‐analysis studies indicate deregulated Th1 and Th17 inflammatory responses can exist as a physiological event in the plasma and CSF of a subset of patients with either AD or MDD. Cytokines generated in these responses regulate cell‐mediated immunity and phagocytic‐dependent inflammation in response to intracellular parasites and can promote autoimmunity (Calcagni & Elenkov, 2006). While the variance in the magnitude and types of overexpressed cytokines can be significant between trials and results can conflict (Orlovska‐Waast et al., 2018), a set of cytokines including IL‐1β, TNF‐α, IL‐6, IL‐18, IL‐17A, and IL‐10 do exhibit consistent up regulation in the plasma of untreated MDD patients (Hodes, Kana, Menard, Merad, & Russo, 2015; Syed et al., 2018). Similar alterations in the concentration of these cytokines are noted in their CSF (Young, Bruno, & Pomara, 2014). In comparison, meta‐analysis studies of patients with AD exhibit a similar inflammatory imbalance involving cytokines responsible for activating adaptive Th1 responses in both their plasma (Swardfager et al., 2010) and their CSF (Brosseron, Krauthausen, Kummer, & Heneka, 2014). A unique parabolic temporal pattern of cytokine expression appears to characterize the progression of inflammation in AD patients. While mild cognitive impairment (MCI) and early‐stage AD patients exhibit consistent overexpression of pro‐inflammatory cytokines (Brosseron et al., 2014), advanced AD patients often exhibit no differences in plasma cytokine levels (Motta, Imbesi, Di Rosa, Stivala, & Malaguarnera, 2007). These observations may help to account for the variability in cytokine profiling within the meta‐studies and suggest immune exhaustion may be a component of end‐stage AD. At one point in the progression of AD, however, a linear relationship does appear to exist between cytokine levels and AD symptomology as increased plasma levels of inflammatory cytokines can predict the conversion of MCI to AD (Tarkowski, Andreasen, Tarkowski, & Blennow, 2003). Moreover, post‐mortem samples from AD patients support the persistence of unresolved neuroinflammatory responses, as evident by increased expression of IL‐1β, IL‐18, and IL‐10R (Guillot‐Sestier et al., 2015; Ojala et al., 2009). Acute cytokine responses involving these dysregulated ILs are triggered by toll‐like and pattern recognition receptors in response to either endogenous or exogenous danger signals. Increased expression of these complexes is observed in MDD (Pandey, Rizavi, Bhaumik, & Ren, 2019; Pandey, Rizavi, Ren, Bhaumik, & Dwivedi, 2014) and AD (Walker et al., 2017) brains and components of these multiprotein complexes are critical for the seeding of extracellular plaques (Heneka et al., 2013). Increased CSF concentrations of certain pro‐inflammatory cytokines were also found to correlate with the neuropsychiatric symptoms associated with AD and MDD respectively, including lower scores on Mini‐Mental State Examination (MMSE) and Hospital and Anxiety Depression Scale (HADS) (Miranda et al., 2018; Wennstrom et al., 2015).

A homeostatic immune state is maintained through negative feedback inputs from immunosuppressive components of the adaptive and innate arms. An inability to recruit immunosuppressive effectors prevents resolution of initial acute inflammatory responses and can promote chronic inflammatory states consistent with those observed in AD and MDD (Fullerton & Gilroy, 2016). The multifunctional cytokine TGF‐β1 mediates immunosuppressive effects through its ability to regulate lymphocyte proliferation and resolve T‐cell inflammatory responses (Ishigame et al., 2013; Li & Flavell, 2008). The noted up‐regulation of CSF and plasma TGF‐β in AD patients (Swardfager et al., 2010), as well as in patients with MDD (Davami et al., 2016; Sutcigil et al., 2007), suggests that the magnitude of acute pro‐inflammatory responses may outweigh immunosuppressive feedback, or compromised immunosuppressive signalling. Evidence for depressed TGF‐β signalling in AD is suggested by the down‐regulation of TGFβ receptor 2 (TGFBR2) expression on cortical neurons (Tesseur et al., 2006). This reduction of neuronal TGFBR signalling correlated with the development of AD neuropathologies and describes how immunosuppressive signalling not only affects immune cells but also influences AD pathologies through receptor‐mediated signalling in neurons. Moreover, increased levels of TGF may have functional consequences on cognition in AD and MDD, as increased CSF TGF‐β correlated with both decreased MMSE and HADS scores, respectively (Miranda et al., 2018; Motta et al., 2007). Other cytokines that possess immunosuppressive capacities, in particular IL‐10, are also up‐regulated in MDD and AD (Guillot‐Sestier et al., 2015; Syed et al., 2018). This paradoxical persistence of both immunosuppressive and pro‐inflammatory cytokines in AD and MDD suggests both disorders involve unresolved adaptive responses that may similarly drive their neuropathologies.

Chemokines mediate the activation and transmigration of effector cells towards stressed tissue and synergize with cytokines to facilitate immune responses. Together, meta‐analysis of AD and MDD subjects reveals altered expression of chemokines in their plasma, suggesting an ongoing effort to recruit effector immune cells (Eyre et al., 2016; Leighton et al., 2018). The mobilization of monocytes appears to represent a physiological process in both AD and MDD, as monocyte recruiting factor chemokine ligand 2 (CCL2/MCP‐1) is the most consistently up‐regulated chemokine. Moreover, increased expression levels of CCL2 in the dorsal anterior cingulate from post‐mortem samples of depressed suicide subjects indicate the recruitment of monocytes originates from microglia inflammatory responses in the brain (Torres‐Platas, Cruceanu, Chen, Turecki, & Mechawar, 2014). As noted for cytokines, CCL2 levels in the plasma and CSF were increased in early‐stage AD and MCI, which predicted increased levels of β‐amyloid and phosphorylated tau (Correa, Starling, Teixeira, Caramelli, & Silva, 2011) that declined in patients with late‐stage AD (Brosseron et al., 2014). The similar temporal expression profile of chemokines and cytokines in AD may be a consequence of their overlapping transcriptional regulators. Moreover, as observed for inflammatory cytokines, positive correlations have been noted between increased CSF levels of CCL2 in MCI/early‐stage AD subjects, a lower 2 year MMSE score (Lee et al., 2018), and latency to develop AD‐dementia (Galimberti et al., 2006; Westin et al., 2012).

3.2. Leukocyte infiltration and T lymphocyte immunophenotypes

The differentiation of adaptive T lymphocytes into diverse subpopulations with distinct effector functions depends on a particular cytokine and chemokine milieu. A reflexive consequence of a persistent cytokine stimulation is therefore a trend towards disequilibrium of plasma leukocyte immunophenotypes (Li & Flavell, 2008). In the plasma of both AD and MDD patients, imbalances in the relative ratios are noted of T lymphocyte phenotypes. Patients with both disorders have exhibited a skew in both the number and differentiated state of plasma immunosuppressive T_reg lymphocytes (Li et al., 2010; Lombardi, Garcia, Rey, & Cacabelos, 1999; Maes et al., 1992; Oberstein et al., 2018; Tan et al., 2002), as well as increased numbers of Th17 lymphocytes (Chen et al., 2011; Patas et al., 2018; Oberstein et al., 2018). As differentiated T_reg lymphocytes are suggested to suppress Th17 differentiation (Chaudhry et al., 2009), impairment of this regulatory mechanism may in part contribute to increased Th17 activity in each disorder. Moreover, the ex vivo characterization of CD3⁺ T lymphocytes isolated from the plasma of AD and MDD patients further indicates deregulation of T lymphocytes may contribute to both disorders. In each case, isolated T‐cells exhibit increased affinity for and expression of inflammatory cytokine receptors (Bongioanni et al., 1997; Bongioanni, Boccardi, Borgna, & Rossi, 1998; Miller, 2010). Increased expression and affinity of cytokine receptors can indicate T lymphocyte exhaustion by chronic stimulation, as observed during chronic infections (Wherry, 2011). In further support of the principle that persistent immunological stimulation may promote exhaustion of specific T lymphocytes in both disorders is evidence that helper CD4⁺ T‐cells have a decreased telomere length (Karabatsiakis, Kolassa, Kolassa, Rudolph, & Dietrich, 2014; Panossian et al., 2003). A decrease in telomere length may reflect increased proliferative history of clonal ancestors induced by chronic immunogenic stimulation. This phenomenon would be consistent with the chronic inflammatory state underlying both disorders.

Under homeostatic conditions, the brain maintains an immunoprivileged environment, one that prevents potentially neurotoxic leukocytes from interacting with sensitive synaptic structures. However, decreased immunosuppressive activity can lead to aberrant leukocyte extravasation (Li & Flavell, 2008). Thus, the dysregulated plasma T lymphocyte profiles described may in part contribute to leukocyte infiltration into the brain noted in biopsied tissue from AD and MDD patients. Post‐mortem tissue from patients with confirmed AD found CD3⁺ T lymphocyte involvement in the medial temporal cortex, the hippocampus and the amygdala (Itagaki, McGeer, & Akiyama, 1988; Merlini, Kirabali, Kulic, Nitsch, & Ferretti, 2018; Togo et al., 2002). Supporting a unique immunological phenotype associated with AD‐dementia, parenchymal T‐cell infiltration was specific for AD‐dementia, while absent in Lewy body dementia, frontal temporal dementia, and Pick's disease (Togo et al., 2002). Moreover, post‐mortem tissue from patients with affective disorders, including unipolar depression, also exhibited increased densities of CD3⁺ T‐cells in limbic structures, such as the hippocampus and amygdala (Bogerts, Schlaaff, Dobrowolny, Frodl, & Steiner, 2018). While lymphocytes elicit cell‐mediated immunity, polymorphonuclear leukocytes such as neutrophils mediate their cytotoxic effects through phagocytosis or the release of potent neurotoxic effectors following degranulation. Neutrophil extravasation into the brain along with neurotoxic extracellular traps has been observed in post‐mortem brains of AD patients (Zenaro et al., 2015), while expression of myeloperoxidase, a marker of neutrophil activation, was up‐regulated in the plasma of patients with both MDD and AD (Tzikas et al., 2014; Vaccarino et al., 2008). Moreover, dendritic or macrophage monocytes have been shown to infiltrate the extravascular brain regions in models of stress‐induced depression and in a Tg3578 AD model, contributing to behavioural impairments and seizure susceptibility but also facilitating Aβ clearance (El Khoury et al., 2007; Koronyo‐Hamaoui et al., 2009; Menard et al., 2017; Wohleb, McKim, Sheridan, & Godbout, 2015). While T lymphocyte and polymorphonuclear infiltration is better characterized in AD and MDD, further experimental testing using genetic models (Lys‐M GFP⁺, Menard et al., 2017) can help to understand role and effects of monocyte trafficking in these disorders.

Selective leukocyte extravasation is dependent upon their interaction and capture by reactive endothelial cell adhesion molecules (CAMs), selectins, and integrins. Under basal physiological conditions, their expression along the endothelium is negligible but can be induced to increase in response to cytokine and chemokine stimuli. Endothelial CAMs involved in cell trapping of leukocytes are considerably up‐regulated in post‐mortem samples of AD and MDD patients. Notably, the expression of the intercellular adhesion molecule‐1 (ICAM‐1) was significantly higher in the plasma and in the grey and white matter of the dorsolateral prefrontal cortex (dlPFC) in patients with depression (Dimopoulos et al., 2006; Thomas et al., 2000). Moreover, while patients with AD also exhibit heightened plasma expression of ICAM‐1, neuronal expression was located to senile plaques and along the brain endothelium (Frohman, Frohman, Gupta, de Fougerolles, & van den Noort, 1991; Verbeek et al., 1994). The expression of the endothelial ligand vascular cell adhesion molecule 1 (VCAM‐1) is increased in plasma of AD patients, a level that correlated with both decreases in MMSE and volume reductions of white matter (Huang et al., 2015). Patients with late‐life depression have also exhibited increased plasma concentrations of VCAM‐1 (Dimopoulos et al., 2006). The respective primary integrin ligand‐binding partners for ICAM‐1 and VCAM‐1 are the α4β1 integrin (VLA‐4), expressed primarily by monocytes and lymphocytes to facilitate their binding to endothelial cells, and the lymphocyte function associated antigen 1 (LFA‐1). Both LFA‐1 and VLA‐4 are up‐regulated in brain leukocytes and microglia from samples of AD post‐mortem tissue (Akiyama et al., 1993; Frohman et al., 1991) as well as in their plasma (Rentzos et al., 2004). The up‐regulation of integrin and CAM proteins in both disorders furthers the argument that an inflammatory environment conducive to leukocyte extravasation similarly contributes to their neuropathologies.

3.3. Microglia activation

As the principal neuroimmune cell, microglia translate a variety of endogenous and exogenous danger signals into immunological responses. Their activity and morphological state—either as surveillant or activated—also influences dendritic spine morphology and synapse steady‐state densities, thus effecting neuronal function and plasticity in not only an activity (Schafer et al., 2012) but also an immune‐dependent manner. Alterations to steady‐state metaplasticity and impaired neuronal conductance are implicated in both early‐stage AD and MDD (Styr & Slutsky, 2018), suggesting microglia processes may contribute to early cognitive impairments in each disorder. In addition, microglia phagocytic potential is regulated in a region‐dependent manner (Grabert et al., 2016), potentially reflecting the activity and metabolic demands of regional neural networks. Consistent with this, microglia activation patterns are region specific in both AD and MDD. While a range of markers associated with microglia activation (CD68, MHCII, ionized calcium‐binding adapter molecule 1, IL‐1 receptor, and CD11c) are broadly overexpressed in the frontal, orbitofrontal, cingulate, and temporal cortices of biopsied AD brains (Hopperton, Mohammad, Trépanier, Giuliano, & Bazinet, 2018), their activation pattern (IL‐1 receptor) in the parahippocampal cortex has laminar specificity, particularly in layers III–IV (Sheng, Griffin, Royston, & Mrak, 1998). Characterization of microglia activation in AD patients using PET imaging with ¹⁸FEPPA binding to mitochondrial translocator protein (TSPO) also supports the presence of inflammatory or phagocytic microglia in the anterior cingulate cortex (ACC), temporal cortices, multiple frontal gyri, and striatal regions (Edison et al., 2008). Several limitations in the validation of TSPO as a biomarker for microglia activation include its multicellular expression and the presence of TSPO polymorphisms that influence binding affinity. Moreover, post‐mortem samples from suicidal depressed patients also show increased expression of microglia activation markers in the frontal cortex, particularly the dlPFC, the ACC, and the hippocampus (Steiner et al., 2008; Torres‐Platas et al., 2014). Furthermore, PET studies involving TSPO ligands in untreated depressed patients also found heightened reactivity in the dlPFC, ACC, and the insula cortex (Holmes et al., 2018; Setiawan et al., 2015, 2018). For both AD and MDD, the severity of psychiatric symptoms in each disorder could be predicted by binding intensity of TSPO, and in patients with mild cognitive impairment, considered a prodromal feature of AD, TSPO binding could predict conversion to AD (Cagnin et al., 2001; Hamelin et al., 2016; Kreisl et al., 2013; Varrone et al., 2015). Further defining how microglia activity contributes to both disorders and their overlapping features is an important line of future investigations and may assist in defining at risk patients.

Signalling by the classical complement pathway is critical for synapse maturation and the maintenance of activity‐dependent synapse plasticity by microglia (Paolicelli et al., 2011; Schafer et al., 2012). While basal expression levels of complement proteins are critical for synapse homeostasis, heightened expression can promote excessive synaptic phagocytosis via microglia. The C1q protein is the central mediator of complementary cascades. Following its release into the extracellular space, C1q is targeted towards postsynaptic ligating partners, where it acts as a scaffold for subsequent complement reactions, which recruit phagocytic microglia (Perry & O'Connor, 2008). Using post‐mortem tissue from both MCI/AD and MDD patients, there is evidence that overactive complement signalling may facilitate synapse deterioration (Afagh, Cummings, Cribbs, Cotman, & Tenner, 1996; Fonseca, Kawas, Troncoso, & Tenner, 2004; Pasinetti, 1996; Veerhuis, Janssen, Hack, & Eikelenboom, 1996). In AD patient's deposits of C1q, other complement proteins colocalize with fibrillary Aβ, suggesting their heightened activity is in part meant for facilitating phagocytosis of extracellular β amyloid. Moreover, increased expression of the downstream C3 complement component has been noted in the orbitofrontal cortex (OFC) of suicide victims (Thalmeier et al., 2008). Combined, heightened microglia activity and increased complement pathway signalling may contribute to the region‐specific synaptic impairments of AD and MDD.

3.4. Hypothalamus pituitary axis and neuropeptides

Persistent activation of noradrenergic pathways by stress stimulates the hypothalamus to release cortisol releasing factor (CRF), which acts in an endocrine manner to promote the release of adrenocorticotropic hormone from the anterior pituitary, and in turn corticosteroids from the adrenal cortex. Patients with depression (Burke, Davis, Otte, & Mohr, 2005) and in dementia (Popp et al., 2015; Rasmuson, Näsman, Carlström, & Olsson, 2002; Swaab et al., 1994; Zvěřová et al., 2013) present with increased serum and CSF levels of cortisol. However, patients with non‐AD type MCI did not exhibit increased CSF cortisol levels, suggesting increased activity of the hypothalamus pituitary axis (HPA) is unique to AD and creates an intriguing relationship between the role of increased cortisol production in depressed patients and its potential as an effector of AD neuropathologies (Figure 2a). Increased cortisol levels in patients with depression may be an effect of decreased expression of the glucocorticoid receptor (GR; Anacker, Zunszain, Carvalho, & Pariante, 2011). Consequently, insufficient GR signalling may fail to complete the negative feedback inhibition of cortisol releasing hormone in the hypothalamus and may potentiate its production.

In addition to stimulating the hypothalamus to produce cortisol, CRF acts as a peptide neurotransmitter within circuits that mediate emotional processing. Its receptor, corticotrophin‐releasing hormone receptor (CRF₁ receptor), a G‐protein, has high expression patterns on neurons that innervate the amygdala, and stress models, as well as genetic studies (Davis et al., 1999) have shown CRF signalling by CRF₁ receptors in amygdala neurons plays a key role in gating the valence of anxiety in response to stress (Savarese & Lasek, 2018). While models suggest CRF contributes to anxiety‐type depression and is consistently up‐regulated in CSF of depressed patients (Hartline, Owens, & Nemeroff, 2018; Nemeroff et al., 1984), its precise physiological contribution to depressive symptoms remains elusive because of confounding clinical studies involving its primary receptor. While expression of CRF₁ receptors is decreased in the frontal cortex of suicidal depressive patients (Leake, Perry, Perry, Fairbairn, & Ferrier, 1990; Nemeroff, Owens, Bissette, Andorn, & Stanley, 1988), other post‐mortem studies involving tissue from depressed patients show increased CRF₁ receptor expression in neurons of the hypothalamic periventricular nucleus (Raadsheer, Hoogendijk, Stam, Tilders, & Swaab, 1994). In contrast, CRF‐CRF₁ receptor signalling appears to play a protective role in the development of AD, as decreased CSF levels of CRF correlated with the onset of cognitive impairments (Behan et al., 1995; Pomara et al., 1989). In addition, in AD, a physiological mechanism to compensate for decreased CRF in the frontal, occipital, and temporal cortex of AD brains appears to be an increase in CRF₁ receptor expression in the same cortical regions (De Souza, Whitehouse, Kuhar, Price, & Vale, 1986; Leake et al., 1990). It may be that CRF‐CRF₁ receptor signalling exerts neurotropic effects, but persistent stimulation of the CRF‐CRF₁ receptor pathway, over time, fails to generate sufficient CRF.

Alterations to glucocorticoid signalling and circuitry of the HPA have implications for extrahypothalamic neural networks, as well as immune activity. First, unabated corticosteroid production can be a reinforcing process. Corticosterone itself promotes region‐specific degeneration in the hippocampus and the amygdala (Lupien et al., 1998), and because both structures negatively regulate the secretion of CSF from the hypothalamus (Buijs & Van Eden, 2000), degeneration of these structures exacerbates cortisol production. Second, while glucocorticoids are intuitively immunosuppressive and indeed are when above basal levels (Sorrells, Caso, Munhoz, & Sapolsky, 2009), acute episodes of glucocorticoid exposure prime innate inflammatory responses, particularly in the CNS (Kelly et al., 2018; Kelly, Miller, Bowyer, & O'Callaghan, 2012). In addition, certain metabolic deficits in AD and MDD may be related to increased plasma glucocorticoid levels, which has been shown to promote glucose intolerance (Adam et al., 2010).

3.5. Altered kynurenine signalling via cytokine signalling

The conversion of tryptophan to either the kynurenine (KYN) precursor compound N‐formylkynurenine or the neurotransmitter 5‐HT is dependent on a number of neuroimmune inputs including type 1 cytokines up‐regulated in both depression and AD (Section 3.1). Inflammatory conditions shift the reaction equilibrium towards N‐formylkynurenine by increasing transcription of the rate‐limiting enzyme indoleamine 2,3‐dioxygenase (IDO), thus promoting tryptophan metabolism away from the neurotransmitter 5‐HT. Regulation of 5‐HT by immune pathways neatly addresses inflammation‐associated symptoms and the monoaminergic hypothesis of depression (Dantzer, O'Connor, Lawson, & Kelley, 2011; Zunszain et al., 2012). Evidence of altered KYN dynamics in both AD and MDD is supported by an up‐regulation of IDO1 in microglia of AD patients (Bonda et al., 2010), as well as evidence that severity of depressive episodes varies as a function of plasma IDO1 activity (Capuron et al., 2002; Wichers et al., 2004). Additionally, altered profiles of IDO1 pathway metabolites are noted in the plasma of both AD and MDD patients. Patients with AD have up‐regulation of plasma KYN (Chatterjee et al., 2018), and the benzol ring derivative of KYN, kynurenic acid, was decreased in serum of both AD and MDD patients (Savitz et al., 2015; Wurfel et al., 2017). Moreover, quinolinic acid is also up‐regulated in the microglia (Bonda et al., 2010; Steiner et al., 2011) in AD and MDD patients, and in the plasma of AD patients (Gulaj, Pawlak, Bien, & Pawlak, 2010). In both cases, the magnitude of quinolinic acid increase was inversely related to cognitive function.

3.6. Amyloid plaques

While deposition of neurotoxic β‐amyloid (Aβ) is a dominant neuropathological feature of AD, the manner and extent to which it contributes to AD continues to be tested. Interestingly, this amyloid species is a pathological feature in certain patients with a history of MDD, while its extraneuronal presence facilitates depressive like behaviour in murine model systems (Ledo et al., 2013). To non‐invasively understand the evolution of amyloid plaques in AD, novel radioligand imaging techniques have been recently developed. The PET tracer C¹¹‐labelled PiB, which binds with high affinity to insoluble fibrillary Aβ, is one such ligand that has been used to illustrate a central role of Aβ in the progression of AD. The magnitude of PiB signal can accurately predict both the extent of neurodegeneration in AD patients as well as an individual's susceptibility to develop dementia following symptomatic mild cognitive impairment (Jack et al., 2010). A second set of ¹⁸F‐labelled tracers bind with the insoluble fibrillary Aβ species, and their binding scores have also predicted the magnitude of neurodegeneration in histologically confirmed AD cases (Sabri et al., 2015). While such radioligands help in diagnosing AD patients, their use in depressed subjects help to suggest Aβ may have a physiological function in certain affective disorders. Patients with a history depression, but who were not positive for either MCI or AD, exhibit heightened ¹⁸F uptake in a number of cortical areas (Harrington et al., 2017; Wu et al., 2014). Based upon these observations, an argument may be made that the development of neurotoxic plaques in AD may be the consequence of an instigator for depressive behaviours, or of one during depressive episodes. Interestingly, both increased production of cytokines and cortisol are aspects of depression that each promote Aβ plaque production in mouse models of AD (Dong & Csernansky, 2009; Ghosh et al., 2013).

The amount and ratio of certain Aβ isoforms in human CSF can also reflect levels of cortical Aβ and thus may serve as a non‐invasive biomarker. Specifically, both increased CSF levels of Aβ₄₂ and the Aβ₄₀:Aβ₄₂ ratio have been correlated to increased cortical ¹⁸F emission levels in AD patients (Leuzy et al., 2016; Palmqvist et al., 2014; Sun et al., 2008). This correlation between CSF Aβ and brain Aβ deposition is important to consider for MDD, as cognitively intact individuals with MDD also exhibit decreased concentrations of CSF Aβ₄₂, as well as an increased Aβ₄₀:Aβ₄₂ ratio (Pomara et al., 2016)_. In addition, patients with a lifetime history of depressive episodes who later developed MCI exhibited increased bilateral frontal cortical Aβ burdens, as compared to MCI individuals without a history of depression episodes (Chung et al., 2015).

4. NEUROANATOMICAL CORRELATES OF DEPRESSION AND AD

4.1. Region‐specific neuropil reductions

The neuropil is characterized as components of cortical tissue, excluding blood vessels, cell bodies, and myelinated axons. The remaining dendritic spines, nonmyelinated axons, and axon terminals gate interneuronal communication, signal valence, and thus circuit‐specific functional and cognitive outputs. These components exhibit a high degree of plasticity and are thus particularly malleable to stress, inflammatory effectors, and region‐specific microglia activity (Shatz, 2009; Stevens et al., 2007). Given that AD and MDD can share microglia activation patterns in overlapping domains (Section 3.3), this biological mechanism may contribute to the synapse degeneration observed in similar domains in patients with AD and MDD (Table 1). Individuals with early‐stage AD present with reductions in layer III and V dlPFC (BA 9) synapse numbers (DeKosky & Scheff, 1990; Scheff, DeKosky, & Price, 1990), while post‐mortem tissue from patients with MDD show synaptic reductions in layers III‐V of the dlPFC (BA 9; Kang et al., 2012). As depression shares certain neuropsychiatric deficits with AD, including impaired working memory and attention, altered sleep patterns, and reduced psychosocial interactions, the reduction of projection neuron synapses in BA9 of the dlPFC, responsible for working memory (Goldman‐Rakic, 1995) and involved in cortico–cortico connections with the superior temporal gyrus in primate (Petrides & Pandya, 1999), may be both a common neuropathology linking their cognitive symptoms and a stage in the pathological progression of each disorder.

Table 1.

Neuroanatomical features of MDD and AD

	Depression	AD‐dementia	Reference
Neuron and glial density	↓ Glial Density: ofPFC1 (IIIa,V‐VI); dlPFC1(III, V); AMY1; ACC1 (VI)	↓Glial Density: AMY1;	Cotter et al. (2002); Cotter et al. (2001); Rajkowska et al. (1999); Scott et al. (1992)
	↓ Neuron Size: mOFC1 + rOFC1 (II‐IV); dlPFC1 (II‐IV); ACC1 (VI) ↓ Interneuron Density: Calbindin, dlPFC1 (II,III)	↓ Neuron Density: mOFC2 (III, V); dlPFC6 (III, IV); mTC6 (III, IV); NbM1; AMY1	Cotter et al. (2001, 2002); Hof et al. (2018); Rajkowska et al. (1999); Rajkowska et al. (2007); Scott et al. (1992); Van Hoesen et al. (2000)
Volumetric analysis	↓Volume: HIPP5, AMY5, dmPFC5, dlPFC5, ofPFC5, rACC5	↓Volume: NbM5, AMY5, HIPP5, mTC5	Bora et al. (2012); Campbell et al. (2004); Eustache et al. (2016); Hamilton et al. (2008); Jones et al. (2006); MacQueen et al. (2003); Poulin et al. (2011)
Synaptic density	↓ Synapse Density: dlPFC4 (III‐V) ↓ Synaptic Gene Expression: HIPP3, ACC3, PFC3	↓ Synapse Density: dlPFC4 (III, V), TC (II‐III)4, HIPPdg4, AMY4, BF4	Davies et al. (1987); DeKosky and Scheff (1990); Duric (2013); Kang et al. (2012); Samuel et al. (1994); Scheff et al. (1990); Scheff et al. (2006); Scheff et al. (2011); Zhao et al. (2012)

Note: In addition to shared immune activity, the brains of MDD and AD patients may share region‐specific neuroanatomical alterations. Notable synapse loss and cell loss have been noted in similar limbic and cortical structures, which can be illustrated through multiple imaging techniques. Roman numerals correspond to cortical layer, when applicable (I–VI). See Abbreviations list for Region Abbreviations.¹Light Microscopy Direct Three‐Dimensional Counting, ²Stereological Counts of NFT Positive Neurons, ³Gene Expression of Synaptic Proteins, ⁴ Electron Microscopy Stereology, ⁵Voxel‐Based MRI, ⁶Stereological Counts of SMI32 Neurons.

Knowledge of further synapse and circuit impairments in depression is limited by an undefined site of pathology and likely involvement of interconnected parallel circuits (Nestler et al., 2002). However, reductions in the gene expression of pre‐ and postsynaptic receptors and neurotransmitter vesicles suggest that synaptic reductions in the hippocampus and the ACC are also pathological processes in depression (Duric, 2013; Zhao et al., 2012). In contrast, brain biopsies from AD patients illustrate definite synapse loss in long‐term memory domains such as in layer III of the inferior temporal gyrus and the outer molecular layer of the hippocampal dentate gyrus (Davies, Mann, Sumpter, & Yates, 1987; DeKosky & Scheff, 1990; Scheff, Price, Schmitt, & Mufson, 2006; Scheff, Price, Schmitt, Scheff, & Mufson, 2011). Hippocampal synapse loss in AD is also confirmed by glycoprotein 2a PET imaging, in which the magnitude of synapse loss correlated with lower MMSE scores (Chen et al., 2018). Moreover, decreased synaptic numbers have also been observed in forebrain structures in early‐stage and late‐stage AD, including in cholinergic rich nuclei of the basal forebrain (Samuel, Terry, DeTeresa, Butters, & Masliah, 1994).

4.2. Region‐specific neuron and glial degeneration

Synaptic loss observed in biopsied samples from patients with AD and MDD may reflect either loss of cortical neurons or loss of cortically projecting axons and therefore decreased dendritic arborization. Identifying which are the susceptible neuronal populations in each disorder may help establish the involvement of overlapping functional domains. Towards answering this question, morphometric examinations of the dlPFC in patients with MDD find reduced neuronal size in layers VI and III and reduced calbindin interneuron densities in layers II and III (Cotter et al., 2002; Rajkowska et al., 1999; Rajkowska, O'Dwyer, Teleki, Stockmeier, & Miguel‐Hidalgo, 2007). In addition, this study found reduced glial densities in the dlPFC, which may adversely affect fitness of surrounding neurons. Since the dlPFC, in primate (Selemon & Goldman‐Rakic, 1988), contains extensive connections with the OFC, subgenual PFC and the ACC, the effects of impaired circuit architecture in the dlPFC may diffuse into these interconnected regions that regulate emotional responses. Indeed, the rostral and medial OFC (mOFC) in these same MDD samples exhibit decreased cortical thickness that occurs simultaneously with a loss of large pyramidal neurons in upper layers II–IV and a loss of glial cells in layers III and IV (Rajkowska et al., 1999). A similar reduction in glial density and neuronal size is noted in layers V–VI and layers II–IV, respectively, of the caudal OFC in these same samples. Notable alterations determined by stereological examination observed in layer VI of the ACC include reduced glial density and neuronal size (Cotter, Mackay, Landau, Kerwin, & Everall, 2001), while subgenual PFC (sg24) was found only to have reduced glial density (Ongur, Drevets, & Price, 1998). The OFC or dlPFC, in primate, also share reciprocal connections with both the basolateral nucleus of the amygdala and hippocampus (Aggleton, Burton, & Passingham, 1980; Carmichael & Price, 1995): limbic structures that regulate stress responses. Decreased glial density, but not changes to neuronal size, has been observed in the amygdala of MDD patients while the glia to neuron ratio in the entorhinal cortex of MDD patients trended upward but did not meet significant numbers (Bowley, Drevets, Öngür, & Price, 2002; Hamidi, Drevets, & Price, 2004).

As noted in MDD tissue samples, reductions in neuronal density within layers III and IV are similarly noted in the dlPFC and in the inferior temporal cortices of patients with AD (Hof, Morrison, & Cox, 2018). A surrogate mode to characterize glial and neuron densities is through the staining of neurofibrillary tangles, which has been well characterized, and directly correlates with reduced glial volume and neuron death (Hof et al., 2018; Whitwell et al., 2008). Within all cases of AD, the pathology of neurofibrillary tangles in the OFC had a unique bilaminate distribution, which was most pronounced in layers III and V of the mOFC (Van Hoesen, Parvizi, & Chu, 2000). As mentioned earlier, patients with MDD exhibit decreased neuronal size in upper layers of the mOFC. The potential progression of cytostructural abnormalities in upper cortical layers of the mOFC, suggested by separate biopsied tissue samples, may reflect a persistent neurotoxic physiological response. It also suggests an uncertain relationship: whether stress responses mediated by the reciprocal connections of the amygdala and the mOFC (Carmichael & Price, 2018) drive mOFC neuropathologies in both AD and MDD.

Loss of cholinergic nuclei in the Nucleus Basilis of Meynert (NbM) is a leading neuropathology associated with AD (Schmitz & Nathan Spreng, 2016) and has been confirmed by histological staining (Geula & Mesulam, 1996; Mesulam, Shaw, Mash, & Weintraub, 2004; Sassin et al., 2000) and in vivo neuroimaging studies (Grothe, Heinsen, & Teipel, 2012, 2013). The Ch4p region of the NbM, which is often the first nuclei of cholinergic cells to degenerate (Liu, Chang, Pearce, & Gentleman, 2015; Whitehouse et al., 1982), contains connections with the amygdala, in primates (Smiley & Mesulam, 1999). This is notable because, as observed in MDD (Bowley et al., 2002), reductions to neuronal and glial densities are also observed in the amygdala of patients with AD (Scott, DeKosky, Sparks, Knox, & Scheff, 1992) and provide a putative circuit through which maladapted amygdala networks generated during stress may influence cholinergic nuclei. In addition, the NbM shares reciprocal connections with other limbic structures implicated in emotional regulation, including the nucleus accumbens (Záborsky & Cullinan, 1992) and the ventral hippocampus (Leranth, Deller, & Buzsaki, 1992), confirmed in rat and primate respectively.

Volumetric neuroimaging studies of AD and MDD patients exhibit volume loss in the brain regions found to have morphological alterations in biopsied tissue. For limbic structures, significant reductions in hippocampal and amygdala volume are noted in prodromal AD, AD (Eustache, Nemmi, Saint‐Aubert, Pariente, & Peran, 2016; Poulin, Dautoff, Morris, Barrett, & Dickerson, 2011), and in nonmedicated MDD subjects (Campbell, Marriott, Nahmias, & MacQueen, 2004; Hamilton, Siemer, & Gotlib, 2008). The co‐morbidity of anxiety disorders in patients with MDD potentiated reductions in amygdala volume, highlighting the potential influence of stress limbic circuitry (Bora, Fornito, Pantelis, & Yücel, 2012). While the number of depressive episodes predicted hippocampal volume reduction in AD patients (MacQueen et al., 2003; Videbech & Ravnkilde, 2004), the magnitude of amygdala reduction may play more of an effector role in the progression of AD. Reductions in amygdala volume in AD patients predicted both hippocampal volume loss (Poulin et al., 2011) and an individual's susceptibility to convert to AD from MCI (Prieto del Val, Cantero, & Atienza, 2016). Cortical regions innervated by the amygdala also exhibit morphometric volume reductions in MDD and AD. A meta‐analysis of voxel‐based morphometry studies showed grey matter reductions in the rostral ACC, the dorsolateral, and the dorsomedial PFC in MDD patients with multiple episodes (Bora et al., 2012). Moreover, AD patients may also exhibit reductions in cingulate volumes, with a particular susceptibility for degeneration in the posterior cingulate gyrus (Jones et al., 2006). Degeneration of the medial temporal lobe is also observed by volumetric analysis in AD patients; which is particularly apparent in early‐stage AD (Bakkour, Morris, Wolk, & Dickerson, 2013; Erkinjuntti et al., 1993; Jobst et al., 1992). Moreover, the involvement of the temporal lobe in the progression of degeneration may be a threshold event for AD, as onset of its degeneration has predicted conversion to AD in patients with MCI (Visser, Verhey, Hofman, Scheltens, & Jolles, 2002). In support of an immune link to the aforementioned conditions with MDD, a subset of depressed patients co‐morbid with high inflammation and anxiety‐like symptoms, exhibited decreased connectivity between the right amygdala and the left ventromedial prefrontal cortex; reduced connectivity in this subset of depressed patients was correlated with anxiety‐like symptoms (Mehta et al., 2018).

5. GENETIC CORRELATIONS

The most prevalent mutations observed in patients with early‐onset AD involve proteins in the amyloid pathway: PSEN1, PSEN2, and amyloid precursor protein (APP). The only genes associated with depressive‐like phenotypes and that exhibit Mendellian inheritance patterns are DTCN1 and DYT1. However, mutations in these genes characterize Perry Syndrome and dystonia respectively (Flint & Kendler, 2014). The other endophenotypes of AD and MDD both exhibit polygenic forms of inheritance, with identified risk loci found in immune related genes. While genome‐wide association meta‐analyses indicate both AD and MDD share a risk loci in the major histocompatibility complex gene and in the complement receptor 1 (CR1) gene (Hamilton et al., 2012; Lambert et al., 2009; Wray et al., 2018), smaller targeted genetic studies in AD and MDD have suggested polymorphisms in inflammatory proteins also increase disease susceptibility. Most notably, polymorphisms in the promoter region of the IL‐1β gene can increase the risk to develop MDD, as well as AD (McCulley, Day, & Holmes, 2003; Murphy et al., 2001; Payao et al., 2012). Physiological effects of IL‐6 polymorphisms depend on its form and location: those in promoter regions that promote IL‐6 expression can increase susceptibility to develop MDD (Bull et al., 2008) and late‐onset AD (Licastro et al., 2003), while those that reduce its activity may provide protection against AD (Papassotiropoulos et al., 2001). Allelic variations in chemokines, cytokines, neuropeptide receptors, and other inflammatory mediators are noted in meta‐analysis studies to influence susceptibilities to either disorder (Bufalino, Hepgul, Aguglia, & Pariante, 2013; Davis et al., 2018; Gonda et al., 2018). Moreover, polymorphisms in gene networks associated with Th1 inflammatory responses both increased the risk for a diagnosis of MDD and predicted responses to 5‐HT (serotonin) reuptake inhibitor (SSRI) antidepressant therapy (Wong, Dong, Maestre‐Mesa, & Licinio, 2008). Although accumulating evidence reveals genetic susceptibilities to AD and MDD can be harboured in immune‐related genes, a genetic correlation study from genome‐wide association meta‐analyses statistics did not find statistically significant genetic correlations between the two disorders (Bulik‐Sullivan et al., 2015). In addition, one large comparative study from the Generation Scotland Scottish Family Health Study and UK Biobank also did not find overlapping polygenetic architectures (Gibson et al., 2017). The authors did note, however, limitations in their study that may have prevented the identification of overlapping genetic risk factors, such as that the heterogeneity of depression phenotypes may have led to an underpowered study.

6. PHARMACOLOGICAL APPROACHES FOR NEUROIMMUNE ACTIVATION

The shared immunological effectors identified in MDD and AD present a potential window to leverage similar therapeutic approaches for both disorders. Implementation of such an approach would be limited, however, by heterogeneity in the pathophysiology of both depression and AD. Moreover, heightened immune activity is not a necessary event for depressive‐like behaviour, and not all AD patients have inflammatory profiles. However, given the potential epidemiological link between AD and depression, as well as their shared immune profiles, targeting a common immune mechanism may benefit both disorders. To further examine this claim, the successes and failures of immune‐modulators previously tested either clinically or preclinically for AD and MDD will be reviewed.

6.1. Direct suppression of cytokines and chemokines

Recent attention has been given to targeting the up‐regulation of cytokine and chemokine effectors evident in both AD and depression (Pfau, Ménard, & Russo, 2018). While clinical attempts using this strategy in either AD or MDD are limited, cytokine‐targeting therapy for other primary autoinflammatory indications has improved depression symptoms and suggests their potential application (Table 2). For example, targeting of TNF‐α‐induced inflammation with an anti‐TNF‐α antibody in psoriasis patients improved their Hamilton Rating Scale for Depression and Beck depression inventory scores with twice weekly treatment (Tyring et al., 2006). Groups similarly pursued immunotherapy‐mediated anti‐TNF‐α treatments for mild–severe AD patients and found while once weekly perspinal injection caused significant improvements in both MMSE and ADAS‐Cog scores (Tobinick, Gross, Weinberger, & Cohen, 2006), a once weekly subcutaneous injection of the same antibody caused no significant changes in cognition endpoints (Butchart et al., 2015). Although the immunotherapy treatment crosses the blood–brain barrier (BBB), inconsistent results suggest peripheral and central inflammatory pathways may differentially contribute to cognitive aspects of AD. A second inflammatory cytokine with peripheral and central effects in MDD that has been targeted in autoinflammatory conditions is IL‐6 (Hodes, Ménard, & Russo, 2016). In support of the therapeutic potential of IL‐6, a clinical trial involving an immunotherapy against IL‐6 in patients with psoriasis improved their depressive symptoms as a secondary outcome measure (Hsu et al., 2015). How IL‐6 physiologically mediates depressive‐like effects may also depend on its central versus peripheral activity. While increased plasma IL‐6 in response to stress impairs BBB structural integrity, artificial overexpression of IL‐6 in astrocytes caused interneuron degeneration and astrogliosis (Campbell et al., 1993). Complementing the neurotoxic features of IL‐6 observed in this artificial system, increased plasma levels of IL‐6 correlated with accelerated cortical thinning within the inferior bilateral temporal lobes (McCarrey et al., 2014). These results further support the prospect of therapeutically targeting IL‐6 as a prophylactic approach against neurodegeneration. The overexpression of IL‐1β also exerts pleiotropic physiological effects including activation of the HPA, activation of endothelial cells, impairment of synaptic connections, and up‐regulation of other proinflammatory cytokines and chemokines (Shaftel, Griffin, & O'Banion, 2008). Immunotherapy against IL‐1β is applied for inherited autoinflammatory disorders, but no studies have attempted such therapies for AD or measured its efficacy in attenuating depressive symptoms. However, preclinical studies in murine stress‐induced depression and AD model systems involving anti‐IL‐1β approaches suggest it may benefit both. For example, using the 3xTg AD mouse model, the administration of an IL‐1 receptor antibody rescued cognitive deficits, suppressed microglia activation and the autocrine/paracrine production of IL‐1β and TNF‐α in the brain, and attenuated tau pathologies (Kitazawa et al., 2011). Moreover, in a stress‐induced depression murine model, inhibition of the IL‐1 receptor with a monoclonal antibody prevented a set of depression‐associated pathologies, including increased BBB permeability, and neuroinflammatory responses associated with exposure to stress (Arakawa, Blandino, & Deak, 2009; Koo & Duman, 2008). Unlike TNF‐α and IL‐6, activation of IL‐1β requires proteolytic cleavage of its zymogen form via inflammasome complexes, preferably by the NLRP3 inflammasome (Latz, Xiao, & Stutz, 2013). This enzymatic platform may therefore serve as a therapeutic target. Preclinical studies involving an inhibitor of NLRP3 inflammasome assembly prevented cognitive deficits, tau pathologies, and amyloid plaque burdens in AD models (Heneka et al., 2013). In addition, genetic ablation of nlrp3 provided resilience against stress‐induced depression and anxiety symptoms in a murine model (Alcocer‐Gomez et al., 2016). Other dysregulated cytokines in both AD and MDD that can be pharmaceutically manipulated include IL‐12/23, IL‐18, and the anti‐inflammatory cytokine IL‐10. Immunotherapies targeting the activity of these effector cytokines have been tested in both murine model systems of AD and MDD, with encouraging results (Guillot‐Sestier et al., 2015; Langley et al., 2010; vom Berg et al., 2012), further indicating a functional role of these cytokines in each disorder.

Table 2.

Clinical and preclinical targeting of common immune mediators in MDD and AD

Target	Depression approaches		Alzheimer's approaches
	Model and protocol	Results	Model and protocol	Results
Cytokines and pattern recognition receptors
TNF‐α	Clinical: Etanercept ×2 weekly for 12 weeks (Tyring et al., 2006)	Increased Ham‐D and BDI in patients with psoriasis	Clinical: S.Q/I. S injection of etanercept (Butchart et al., 2015; Tobinick et al., 2006)	Intraspinal injection improved AD cognitive decline; S.Q did not improve cognitive symptoms
IL‐1β	Preclinical: I.C.V of IL‐1Ra before stressor (Arakawa et al., 2009)	Prevented reduction in social behaviour and memory impairment	Preclinical: I.P. IL‐1R mAb in 3xTg AD mice (Kitazawa et al., 2011)	Reduced activated microglia, pTau, GSK3β, fibrillar Aβ and rescued cognitive deficits
IL‐6	Clinical: S. C IL‐6 mAb for R.A25mg q4wks (Hsu et al., 2015)	Significant improvements in PDMA symptoms at week 12	GFAP‐IL6 astrocyte expression (Campbell et al., 1993)	Astrogliosis, neovascularization, neurodegeneration of parvalbumin neurons
IL‐10	Preclinical: IL10−/− mice exposed to stress (Mesquita et al., 2008)	Increased susceptibility to stress‐induced depression; rescued with IL‐10 administration	Preclinical: IL10−/− mice cross with APP/PS1 (Guillot‐Sestier et al., 2015)	Preserved synaptic integrity and mitigated cognitive disturbance
IL‐12/23	Clinical: 52 week of IL‐12/23 antibody (Langley et al., 2010)	Patients with psoriasis experienced improvements in HADS	Preclinical: Antibody against IL‐12/23 p40 subunit to APP/PS1mice (vom Berg et al., 2012)	Reduced soluble Aβ and cognitive declines
TGF‐β	N.A	N.A	Preclinical: TβRIIΔk/hAPP (Tesseur et al., 2006)	Reduced TGFβ signalling increases Aβ, myeloid deposition and dendritic degeneration
NLRP3	Preclinical: NLRP3−/− mice with immobilization stress (Alcocer‐Gomez et al., 2016)	Prevented stress‐induced depressive behaviours, microglia activation, and neuroinflammation	Preclinical: Inhibition of NLRP3 or NLRP3−/− in APP/PS1 mice (Dempsey et al., 2017; Heneka et al., 2013)	Reduced Aβ accumulation, improved cognitive function, attenuates microglia activation and neuroinflammation.
TLR4	Preclinical: TLR4−/ ‐ mice during stress (Cheng et al., 2016)	Resilient to learned helplessness and neuroinflammation	Preclinical: TLR4−/− mice (Bhaskar et al., 2010)	Mice exposed to LPS had reduced MAPT phosphorylation
Microglia activity
CX3CR1	Preclinical: CX3CR1−/− mice exposed to psychological stress (Wohleb et al., 2013)	Prevented against stress induced monocyte recruitment into the brain and corresponding anxiety‐like symptoms	Preclinical: CX3CR1−/− x 5xTgAD (Fuhrmann et al., 2010)	Prevented layer III neuron death; no change in β amyloid; knockout increased plasma levels of inflammatory cytokines and increased cognitive decline
Microglia activation (NF‐κB)	Clinical: 200 mg·kg−1·day−1 minocycline for 12 weeks I (Dean et al., 2017; Husain et al., 2017)	Improvement in HAM‐D score and CGI	MADE trial: Mild AD treated for 24 months 400 mg minocycline	T.B.D
C1q/C3	Preclinical: C3aR−/− mice during CUS (Crider et al., 2018)	Prevented stress induced depression, monocyte infiltration, and neuroinflammation	Preclinical: In APPPS1 mice, antibody to C1q and APP/PS1:C1q−/− (Hong et al., 2016)	Prevented against Aβ‐mediated synapse loss
Immuno‐metabolism
GR receptor	Clinical: Treatment of PMD with mifepristone for 7 days (Belanoff et al., 2002; Belanoff et al., 2001)	Decline in BPRS and HAM‐D‐21 scores following short term use; long term did not find benefit	Clinical: 200 mg Mifepristone daily for 6‐weeks (Pomara et al., 2002)	In patients with mild to moderate AD no improvement in baseline ADAS‐Cog score, HDRS score, MMSE
IDO	Preclinical: IDO1 Inhibitor during stress (Laugeray et al., 2016)	Reduced peripheral inflammatory cytokines and alleviated depressive symptoms	N.A	N.A
CRF1 receptor	Clinical: Antagonists of CRF1 receptor (Binneman et al., 2008 ; Coric et al., 2010)	Failed to Improve HAM‐D over baseline	Preclinical: CRF1 receptor antagonism in PS 19 and Tg2576 mice (Carroll et al., 2011; Dong et al., 2014)	Prevented stress induced tau and Aβ aggregation, memory impairments, and neurodegeneration
Treg depletion	Preclinical: anti‐CD25+ antibody prior to stress (Kim et al., 2012)	Increased susceptibility to inflammation, depressive symptoms	Preclinical: Foxp3+ depletion in 5x FAD (Baruch et al., 2015)	Enhanced Aβ clearance, suppressed neuroinflammation, and cognitive decline
Integrin and cell adhesion molecules
α4‐Integrin	N.A	N.A	Preclinical: Treatment of 3xTg‐AD mice with mAb (Piacentino et al., 2016)	Improved cognitive deficits, reduced pTau, restored synaptic protein expression
LFA‐1	Clinical: Efalizumab ×1 per week for 12 weeks in psoriasis patients (Ortonne et al., 2005)	Improvements in SF‐36 emotional and psychological stress scores	Preclinical: I.V. LFA‐1 mAB or Itgal−/− in 3x FAD mice (Zenaro et al., 2015)	Prevented transmigration of neutrophils into parenchyma, decreased cognitive deficits, microgliosis, and tau phosphorylation
GSK3	Preclinical: Pretreatment with GSK3 Inhibitor prior to stress (Cheng et al., 2016)	Reduction in stress‐induced cytokines in hippocampus and depressive behaviours	Clinical: Tideglusib (GSK3b Inhibitor) once daily oral (Lovestone et al., 2015)	Failed to meet primary cognition endpoints, but participants with mild AD showed significant improvement on ADAS‐cog15 and MMSE scores

Note: Substantial clinical and preclinical investigations have attempted to target activated neuroimmune pathways shared between AD and MDD. While a number of clinical trials designed to suppress cytokines for other primary indications improved depression symptomology as a secondary outcome, targeting cytokine expression in AD in the clinic has been limited to TNF‐α, in spite of preclinical evidence that supports further investigations. Inhibition of pattern recognition proteins such as the NLRP3 inflammasome or TLR receptors hint at their clinical potential in both AD and MDD. Suppression of microglia activation has been attempted with minocycline and improved clinical depression scores, while a large clinical trial of minocycline in AD is currently in progress. Preclinical murine models also support targeted approaches against complement pathway proteins involved in synaptic pruning. Approaches that target immuno‐metabolism pathways, such as antagonism of GR or CRF₁ receptor, showed promise in both AD and MDD in the short term and in smaller studies but failed to improve cognition scores or prevent disease progression over longer periods. Targeting integrin and cell adhesion molecules involved in leukocyte migration showed improved depressive symptoms when treated for other primary indications and exhibits clinical potential based on preclinical studies using genetic and antibody models.

As systemic immune activity can influence cognition and neuronal health, manipulating activity of effector cells that regulate immune equilibriums may also benefit both disorders. The ratio of plasma T_reg:T_eff lymphocytes is an immunological parameter that directly influences systemic inflammatory states (Chaudhry et al., 2009). Manipulating this ratio through pharmacological or genetic techniques to improve AD or MDD symptoms has been investigated, producing results that highlight not only the complexity of T lymphocyte function but the uncertainty in how they contribute to each condition. Using the 5x FAD model, the selective depletion of immunosuppressive Foxp3⁺ T_regs facilitated amyloid‐β plaque clearance, suppression of neuroinflammatory responses, and the reversal of cognitive decline (Baruch et al., 2015). Meanwhile, depletion of CD4⁺ CD25⁺ T_regs with an anti‐CD25 antibody in mice exposed to physical stressors potentiated inflammatory effectors in the plasma, as well as depressive‐like symptoms (Kim et al., 2012). Rather paradoxically, depletion of T_regs in AD appeared to resolve AD pathologies, while T_reg depletion increased the susceptibility to depression phenotypes. One hypothetical explanation for these divergent results is that pro‐inflammatory responses are critical for resolving AD‐specific insults, but their collateral effects involve depressive like behaviours. In addition, dysregulated T_reg activity may reflect immune exhaustion, in which they adopt a more neurotoxic immunophenotype. This phenomenon would be consistent with their heightened inflammatory responses, and their shortened telomere length (Section 3.2).

Immune exhaustion has been a recent target for novel therapeutic approaches in AD. Antagonists of the programmed cell death protein 1/programmed death‐ligand (PD‐1/PD‐L1) pathway were originally developed to increase the susceptibility of cancer cells to T‐cell‐mediated cytotoxicity. Further investigations into how PD‐1/PD‐L1 regulates T‐cell function demonstrate its ability to mediate immunological exhaustion (Wherry, 2011). Administration of PD‐1 antagonists stimulated IFN‐γ T‐cell responses, which facilitated the clearance of cerebral amyloid β and improvements in cognitive symptoms (Baruch et al., 2016). Conflicting studies, however, find that while PD1 engagement stimulated systemic immune activity; it did not lead to improvement in amyloid pathology or cognitive impairments (Latta‐Mahieu et al., 2018). In addition, PD‐1 immunotherapy in cancer patients also elicits pronounced immune activation, but patients commonly develop depression symptoms and cognitive impairments (Bower et al., 2009). While T‐cell modulation offers promise for regulating immune function, how it may be applied for altered immunological activity in AD and MDD may require a deeper understanding of T‐cell dynamics.

6.2. Modulation of neuroendocrine and metabolic pathways

Metabolic activity of the kynurenine pathway influences peripheral and central immunophenotypes and function (Baban et al., 2009). Certain metabolites of the pathway can also influence neuronal conductance, by acting as NMDA agonists or antagonists. Alterations to the equilibrium of these kynurenine metabolites noted in brains of MDD and AD patients may distort the balance of excitatory to inhibitory inputs, which can increase the susceptibility for excitotoxicity‐induced neurodegeneration. Selective targeting of rate‐limiting enzymes in this pathway may therefore prove beneficial in either condition. Attempts to target IDO1, responsible for converting L‐tryptophan to N‐formylkynerine, as well as the production of other KYN pathway metabolites in murine and drosophila model systems illustrate how this pathway may contribute to degeneration and synapse loss noted in both AD and depression, as well as their therapeutic potential (Breda et al., 2016; Laugeray et al., 2016).

Increased activity within the hypothalamus‐pituitary axis and increased cortisol production elicits physiological effects consistent with AD and depression. These include selective hippocampal degeneration, primed inflammatory responses, altered metabolic activity, and amyloid pathologies. Strategies that suppress effectors within the pathway may therefore provide therapeutic benefit and suppress systemic inflammation. Glucocorticoid receptors exhibit complex regulatory dynamics: physiological effects of cortisol fit a parabolic trend. At low and very high concentrations cortisol impairs memory function and promotes inflammation, while basal concentrations support cognitive function and act in an anti‐inflammatory manner (Frank, Thompson, Watkins, & Maier, 2012). This pattern may in part be an effect of environment‐induced DNA methylation, as suicidal victims who experienced childhood abuse had increased methylation of the NR3C1 promoter region, accompanied by decrease GR mRNA expression (McGowan et al., 2009). Moreover, suppression of cortisol signalling with GR antagonists has been pursued in therapeutic settings, with mixed results. While short‐term treatment of psychotic major depression via GR inhibition improved their scores on their brief psychiatric ratings scores (Belanoff, Flores, Kalezhan, Sund, & Schatzberg, 2001; Belanoff et al., 2002; Spiker et al., 1985), a longer and higher powered phase III trial failed to show significant differences between the GR inhibitor and placebo groups (Corcept Therapeutics Press Release, March 29, 2007). While preclinical models support a GR‐targeted strategy for rescuing episodic memory deficits and impaired LTD in an AD model system (Lanté et al., 2015), treatment of advanced AD with a potent GR inhibitor failed to show statistically significant improvements in MMSE or ADAS‐Cog subtests but trended positively (Pomara, Doraiswamy, Tun, & Ferris, 2002). This was a pilot study, and the small sample size together with the short duration of treatment limits interpretation of the results.

The dysfunctional CRF₁ receptor‐CRF signalling noted in both AD and MDD patients (Section 3.4) may not only act as a biomarker for both disorders, but may also be a potentially modifiable component of disease symptomology. While recognized features of CRF₁ receptor‐CRF extrahypothalamic pathway signalling involve its regulation of behavioural adaption to stress (Smith et al., 1998), CRF₁ receptor‐CRF innervation of the hypothalamic paraventricular nuclei maintains a reciprocal relationship with immune and neuroendocrine responses (Venihaki, Dikkes, Carrigan, & Karalis, 2001; Webster, Torpy, Elenkov, & Chrousos, 1998). Altered CRF₁ receptor‐CRF signalling may therefore similarly contribute to both cognitive and immune aspects of AD and MDD. However, clinical studies involving CRF₁ receptor antagonists in major depression did not provide definitive therapeutic insight. While administration of CRF₁ receptor antagonists provided a dose‐dependent improvement in both anxiety (HAMA) and depression (Hamilton Rating Scale for Depression) levels (Zobel et al., 2000), subsequent larger studies failed to replicate statistically significant improvements in depressive or anxiety symptomology (Binneman et al., 2008; Coric et al., 2010; Ising et al., 2007). Inconsistent results may be attributed to regional specificity of CRF₁ receptor modulation in the brain, as selective deletion of CRF₁ receptor in mice forebrain glutamatergic circuits reduced anxiety, while selective deletion in midbrain dopaminergic neurons increased anxiety like behaviours (Refojo et al., 2011). Investigations that explore the therapeutic potential of CRF₁ receptors in AD has been limited so far to murine models, which illustrate that CRF₁ receptor antagonism can attenuate stress‐induced development of AD‐associated pathologies (Carroll et al., 2011; Dong et al., 2014; Filipcik et al., 2012). The potential to treat either AD or MDD through CRF₁ receptor antagonism may be limited when considering the kinetics of GPCRs like CRF₁ receptors, in which there is rapid desensitization and internalization.

6.3. Leukocyte and monocyte transmigration and BBB structural integrity

The infiltration of bone marrow‐derived leukocytes into the immune‐privileged CNS may represent a pathophysiological component of both in MDD and AD (Section 3.2). How this phenomenon contributes to symptoms of each disorder in humans is under investigation, but in murine models of AD and stress‐induced depression, extravasation of peripheral monocytes and leukocytes can facilitated anxiety symptoms via activation of neuroinflammation (Wohleb et al., 2015). As access into the immune‐privileged CNS requires the recognition of leukocyte‐specific CAMs by complementary integrin proteins expressed along the brain endothelium, targeting integrin or CAM proteins is a potential means to suppress the collateral neurotoxicity of invading immune cells. A clinical trial that tested a monoclonal antibody against the CD11a domain of LFA‐1 in patients with psoriasis observed improvement in mental health secondary endpoints (Ortonne, Shear, Shumack, & Henninger, 2005). While in a 3xAD murine model, both an antibody against LFA‐1 and a conditional knockout of Itgal, the gene containing the LFA‐1 protein each prevented the intraperanchymal extravasation of neutrophils (Zenaro et al., 2015). Prophylactic treatment with the antibody prior to symptom onset in this study also prevented cognitive decline, the release of intraparenchymal neurotoxic extracellular traps, and reduced gliosis throughout the brain. Another leukocyte integrin involved in transmigration is α4β1 (VLA‐4), which assists in capturing peripheral cells and is up‐regulated in both AD and MDD patients, as described in Section 3.2. Targeting with an anti‐α4β1 antibody in a triple transgenic AD model reduced tau hyperphosphorylation and restored synaptic protein expression (Piacentino et al., 2016). As further proof of principle, VLA‐1 inhibition prevented brain extravasation of bone marrow‐derived leukocytes (Haanstra et al., 2013) and dendritic cells (Jain, Coisne, Enzmann, Rottapel, & Engelhardt, 2010) in the experimental autoimmune encephalomyelitis model of multiple sclerosis. One serious side effect that may limit implementation of anti‐integrin/CAM approaches is increased susceptibility to immunodeficiency and progressive multifocal leukoencephalopathy.

6.4. Complement‐mediated synaptic degeneration

Aberrant expression of complement proteins in prefrontal cortical neurons of MDD and early‐AD brains (Section 3.3) may prevent synaptic loss and therefore offer a promising therapeutic target. In the APP/PS1 AD model, region‐specific elevation of C1q in the frontal cortex and hippocampus preceded plaque development while a conditional C1q knockout or an antibody targeting C1q prevented oligomer Aβ‐mediated synaptic loss (Hong et al., 2016). The downstream complement component C3 is responsible for opsonization of synapses. In both an APP/PS1 murine model, as well as in a stress‐induced depression paradigm, inhibition of C3 prevented synapse loss, the onset of cognitive impairments, monocyte infiltration, and neuroinflammation (Crider et al., 2018; Hong et al., 2016). While preventing complement‐mediated removal of synapses is an attractive therapeutic strategy, it's implementation may be limited by the fact that, normal synaptic pruning mediated by the complement cascade is critical for maintaining excitatory to IPSP equilibriums. Highlighting barrier to such an approach, pharmaceutical disruption of different complement proteins lead to improper synaptic connectivity, and an increased propensity for epileptic phenotypes (Chu et al., 2010; Schafer et al., 2012). While considering the limitations of a complement approach, an antibody against C1q is currently being investigated in patients with AD (NCT03010046).

The Fractalkine receptor (CX₃CR1) expressed in peripheral leukocytes mediates their migration, while expression of CX₃CR1 in microglia mediates their recruitment to injured neurons secreting CX₃CL1 (Fractalkine) (Harrison et al., 1998). Preventing CX₃CR1‐mediated synapse engulfment may similarly attenuate synapse loss in AD and MDD. Genetic ablation of Cx3cr1 in the murine 5xTg AD model prevented layer III neuron death, while not effecting levels of Aβ (Fuhrmann et al., 2010). In contrast, a neuroprotective role for CX₃CR1 in AD model systems has also been described whereby genetic removal of Cx3cr1 enhanced tau pathology, cognitive deficits, and neurotoxic effects of inflammatory cytokines (Cho et al., 2011). The disparity between these studies may reflect a role for CX₃CR1 in facilitating the removal of damaged neurons. Moreover, in murine models of stress‐induced depression, CX₃CR1 appears to in part mediate behavioural impairments and inflammatory neuropathologies (Hellwig et al., 2016; Wohleb, Powell, Godbout, & Sheridan, 2013). Other non‐targeted approaches to suppress an activated microglia phenotype involved the tetracycline antibiotic minocycline, which prevents polarization of microglia towards a reactive morphology associated with increased synaptic pruning and inflammation (Kobayashi et al., 2013). Clinical studies utilizing minocycline in neurodegenerative disorders show reduced microglia activation but persistence of degenerative neuropathologies (Kim & Suh, 2009; Scott et al., 2018).

6.5. Upstream effectors of immune and inflammatory stimulators

Initiation of inflammatory and immune responses depends upon an integrated network of intracellular proteins that tightly regulates transcriptional networks, cellular differentiation, and effector products. Dysregulation of these networks results from both stochastic events and genetic variables and increases susceptibility for heightened immunological activity. The prototypical inflammatory signalling pathway mediates its effects through NF‐κB signalling. The archetypal activators of NF‐κB transcription involve the evolutionarily homologous IL‐1, TNF, and toll‐like receptor family, which stimulate NF‐κB inflammation through different kinase intermediaries. Proteins regulating NF‐κB activity exhibit different expression patterns in AD and MDD and may therefore serve as a therapeutic target. The protein glycogen synthase kinase‐3 (GSK), with α and β isoforms, regulates glycogen synthase, but GSK3β also interacts with NF‐κB to promote inflammatory cytokine expression, which could be prevented by selective GSK3β inhibition (Martin, Rehani, Jope, & Michalek, 2005). Biopsies from patients with depressive suicide support this approach as increased GSK activity was noted in their ventral prefrontal cortex (Karege et al., 2007). In addition, increased GSK activity is evident in the frontal cortex of patients with AD (Leroy, Yilmaz, & Brion, 2007) and is up‐regulated in their hippocampus (Blalock et al., 2004). Translating the principle of GSK3β inhibition into murine model of stress‐induced depression illustrated a critical role for GSK3β in mediating depressive like behaviours (Cheng et al., 2016). Interestingly, lithium ions act as noncompetitive inhibitors of ATP for GSK3β (Cross et al., 2008), which may in part be responsible for its anti‐depressive traits. An improvement in behavioural and neuropathological endpoints was also observed by inhibiting GSK in AD murine models (Licht‐Murava et al., 2016). However, in a phase 2a dose‐escalating trial, once daily treatment with a GSK3β inhibitor failed to meet primary cognition endpoints (Lovestone et al., 2015). A noted improvement on ADAS‐cog15 and MMSE scores in participants with mild AD in this study may support prophylactic suppression of inflammatory pathways to delay the onset of AD.

Another mediator of inflammatory responses is PPAR‐γ, which is a nuclear transcription factor that has increased expression patterns in macrophages and monocytes and is responsible for their activation and differentiation (Clark, 2002). Ligand stimulation of PPAR‐γ is also associated with increased secretion of pro‐inflammatory cytokines including IL‐1β, IL‐6, and TNF‐α, while agonism of PPAR‐γ inhibits their production in monocytes (Jiang, Ting, & Seed, 1998). Increased activity of PPAR‐γ may contribute to inflammation in AD and MDD, as its expression is up‐regulated in brains from these patients (Kitamura et al., 1999). In contrast to its activity in immune cells, increased expression of PPAR‐γ in neurons may be a transcriptional anxiolytic response mechanism, as stimulation of PPAR‐γ in mouse amygdala neurons prevented the onset of stress‐induced depressive symptoms (Domi et al., 2016). In an open label trial involving PPAR‐γ agonists, noted improvements in depression symptoms were observed in bipolar patients, whose response to treatment was dependent upon reductions in inflammatory cytokines (Kemp et al., 2014). In contrast, PPAR‐γ agonism in two double‐blind studies in subjects with mild‐to‐moderate AD failed to improve any primary endpoints (Harrington et al., 2011).

GATA1 is a key synaptic transcriptional regulator that can suppress IL‐6‐mediated responses, by binding to IL‐6 promoters, and down‐regulates a variety of synapse genes known to be down‐regulated in depression (Cole et al., 2010). Thus, GATA1 may be a component of a negative feedback loop to down‐regulate IL‐6‐mediated responses in a pro‐inflammatory environment, at the consequence of synaptic plasticity and density. Post‐mortem samples from AD brains similarly exhibited increased expression of GATA1 (Wang, Zhao, Freire, Ho, & Pasinetti, 2014), GATA1 also increases transcription of γ‐secretase activating protein, a key protein involved in generating β‐amyloid plaques (Chu, Wisniewski, & Pratico, 2016). Thus, pharmaceutically targeting GATA1 represents a viable approach towards suppressing inflammation‐induced decreases in synaptic density, as well as increased amyloid load in AD.

6.6. Immunosuppressive features of current treatments

The prevailing treatment paradigms for AD and MDD involve acetylcholinesterase (AChE) inhibitors (AChEI) and (5‐HT) serotonin reuptake (SERT) inhibitors respectively. Beginning with physostigmine treatment (Davis et al., 1978) and evolving with select AChEIs (Rogers, Farlow, Doody, Mohs, & Friedhoff, 1998), the clinical application of AChEIs is founded on the principle that deficits in cholinergic signalling mediate cognitive impairments in AD (Coyle, Price, & DeLong, 1983). The temporary functional benefits of AChEIs in AD patients are therefore considered an effect of increased intrasynaptic ACh levels. However, another physiological effect of ACh is its influence on immune activity. Recent literature describing how AChE inhibition suppresses systemic cytokine expression in AD patients (Pavlov et al., 2009; Reale et al., 2004), which indicates their efficacy may in part be derived from anti‐inflammatory effects. Moreover, in one study, donepezil treatment in MCI patients slowed their conversion to AD only when subjects with MCI were co‐morbid with depression (Lu et al., 2009). An implication of these results is that increased ACh levels mediate their effects through a mechanism conserved between depression and AD. Whether this resiliency results from anti‐inflammatory effects or increased cholinergic signalling in the brain is uncertain.

Meanwhile, leading treatments for depression involve inhibition of monoamine oxidases, or inhibition of noradrenaline and 5‐HT reuptake proteins. Like for AChEI treatment, the therapeutic benefit of SSRIs is predicated on increasing intrasynaptic neurotransmitters (Thase, Entsuah, & Rudolph, 2001). However, SSRIs also exhibit anti‐inflammatory properties. Clinical studies with SSRIs find that clinical symptomatic relief occurs in parallel with a decrease in plasma pro‐inflammatory cytokines and an increase in plasma immunosuppressive T_reg cells (Hannestad, DellaGioia, & Bloch, 2011; Himmerich et al., 2010; Maes et al., 1999). Further investigations should identify whether the anti‐inflammatory properties of SSRIs are an effect of direct interactions between increased 5‐HT and immune effectors or because of normalized brain 5‐HT signalling, which suppresses neurogenic sources of inflammation.

7. CONCLUSION

Alterations in neuroimmune activity can figure prominently in the diagnoses of depression, as well as dementia. By nature of their overlapping immune profiles, shared genetic and environmental risk factors, and neuropathological correlates, there is substantial evidence to make an argument that an immunological link may similarly confer susceptibility to either Alzheimer's and depression. Up until recently, a pharmacological approach to suppress inflammation and immune activity was limited by the drug targets available. With new insights into proteins that act as key activators of inflammation and new neuroimmune pathways, novel therapeutic approaches can be developed, or drugs with other primary indications can be repurposed for use as a preventative treatment of AD and MDD.

7.1. Nomenclature of targets and ligands

Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY (Harding et al., 2018), and are permanently archived in the Concise Guide to PHARMACOLOGY 2017/18 (Alexander, Kelly et al., 2017a,b).

Herman FJ, Simkovic S, Pasinetti GM. Neuroimmune nexus of depression and dementia: Shared mechanisms and therapeutic targets. Br J Pharmacol. 2019;176:3558–3584. 10.1111/bph.14569