Neuroinflammation and psychiatric illness

Multiple lines of evidence support the pathogenic role of neuroinflammation in psychiatric illness. While systemic autoimmune diseases are well-documented causes of neuropsychiatric disorders, synaptic autoimmune encephalitides with psychotic symptoms often go under-recognized. Parallel to the link between psychiatric symptoms and autoimmunity in autoimmune diseases, neuroimmunological abnormalities occur in classical psychiatric disorders (for example, major depressive, bipolar, schizophrenia, and obsessive-compulsive disorders). Investigations into the pathophysiology of these conditions traditionally stressed dysregulation of the glutamatergic and monoaminergic systems, but the mechanisms causing these neurotransmitter abnormalities remained elusive. We review the link between autoimmunity and neuropsychiatric disorders, and the human and experimental evidence supporting the pathogenic role of neuroinflammation in selected classical psychiatric disorders. Understanding how psychosocial, genetic, immunological and neurotransmitter systems interact can reveal pathogenic clues and help target new preventive and symptomatic therapies.


Introduction
As biological abnormalities are increasingly identified among patients with psychiatric disorders, the distinction between neurological and psychiatric illness fades. In addition to systemic autoimmune diseases associated with psychiatric manifestations (for example, lupus) [1], more recently, patients with acute isolated psychosis were identified with synaptic autoimmune encephalitides (Table 1) [2][3][4][5][6]. These patients are often erroneously diagnosed with refractory primary psychotic disorders, delaying initiation of effective immune therapy (Table 1). Additionally, growing evidence supports the pathogenic role of anti-neuronal antibodies in neuropsychiatric disorders [7].
Separation of neurological and psychiatric disorders, supported by Descartes's conception of the 'mind' as an ontologically distinct entity and by the reproducibility of neuropathological abnormalities, dominated medicine in the 19th and early 20th centuries [8]. Since then, an expanding collection of reproducible biological causes, from neurosyphilis, head trauma, stroke, tumor, demyelination and many others caused symptom complexes that overlapped with classic psychiatric disorders [9][10][11]. More recently, neuroinflammatory and immunological abnormalities have been documented in patients with classical psychiatric disorders.
Medical conditions associated with chronic inflammatory and immunological abnormalities, including obesity, diabetes, malignancies, rheumatoid arthritis, and multiple sclerosis, are risk factors for depression and bipolar disorder [10,12,13,15,17,18]. The positive correlation between these medical conditions and psychiatric illness suggests the presence of a widespread underlying inflammatory process affecting the brain among other organs [10,19,20]. A 30-year populationbased study showed that having an autoimmune disease or a prior hospitalization for serious infection increased the risk of developing schizophrenia by 29% and 60%, respectively [16]. Further, herpes simplex virus, Toxoplasma gondii, cytomegalovirus, and influenza during pregnancy increase the risk of developing schizophrenia [16].
We first review the association between autoimmunity and neuropsychiatric disorders, including: 1) systemic lupus erythematosus (SLE) as a prototype of systemic autoimmune disease; 2) autoimmune encephalitides associated with serum anti-synaptic and glutamic acid decarboxylase (GAD) autoantibodies; and 3) pediatric neuropsychiatric autoimmune disorders associated with streptococcal infections (PANDAS) and pure obsessive-compulsive disorder (OCD) associated with anti-basal ganglia/thalamic autoantibodies. We then discuss the role of innate inflammation/autoimmunity in classical psychiatric disorders, including MDD, bipolar disorder (BPD), schizophrenia, and OCD.
The inflammation associated with anti-synaptic autoantibodies, particularly NMDAR-autoantibodies, is typically much milder than that associated with GAD-autoantibodies or anti-neuronal autoantibodies related to systemic autoimmune disorders or paraneoplastic syndromes [2,107].
PANDAS and pure obsessive-compulsive disorder associated with anti-basal ganglia/thalamic autoantibodies OCD often complicates neurological disorders involving the basal ganglia including Sydenham's chorea, Huntington's disease and Parkinson's disease. Antibasal ganglia antibodies are implicated in Sydenham's chorea [123]. PANDAS is characterized by acute exacerbations of OCD symptoms and/or motor/phonic tics following a prodromal group A β-hemolytic streptococcal infection. The pathophysiology is thought to involve cross-reactivity between anti-streptococcal antibodies and basal ganglia proteins [124]. The clinical overlap between the PANDAS and pure OCD suggests a common etiological mechanism [125].
Yet, in the same study only one of 19 sero-positive OCD subjects also had positive anti-streptolysin O antibody titers, suggesting that in pure OCD the anti-streptolysin O antibody seronegativity does not exclude the presence of anti-basal ganglia autoantibodies.
In pure OCD, sero-positivity for anti-basal ganglia/ thalamic antibodies is associated with increased levels of CSF glycine (P = 0.03) [88], suggesting that these antibodies contribute to hyperglutamatergia observed in OCD [84,88,131]. The improvement of infection-provoked OCD with immune therapies supports the pathogenicity of these autoantibodies [132]. A large NIH trial assessing the efficacy of intravenous immunoglobulin (IVIG) for children with acute onset OCD and anti-streptococcal antibodies is ongoing (ClinicalTrials.gov: NCT01281969). However, the finding of slightly higher CSF glutamate levels in OCD patients with negative CSF anti-basal ganglia/thalamic antibodies as compared to those with positive CSF antibodies suggests that non-immunological mechanisms may play role in OCD [84]. Other mechanisms, including cytokinemediated inflammation (Table 2), are also hypothesized.

Psychiatric disorders associated with innate inflammation
Disorders of innate inflammation/autoimmunity occur in some patients with classical psychiatric disorders. We discuss innate inflammation-related CNS abnormalitiesincluding glial pathology, elevated cytokines levels, cyclooxygenase activation, glutamate dysregulation, increased S100B levels, increased oxidative stress, and BBB dysfunction-in MDD, BPD, schizophrenia, and OCD. We also describe how innate inflammation may be mechanistically linked to the traditional monoaminergic and glutamatergic abnormalities reported in these disorders (Figures 1 and 2). The therapeutic role of antiinflammatory agents in psychiatric disorders is also reviewed.
Animal studies are consistent with human studies showing astroglial loss in MDD. Wistar-Kyoto ratsknown to exhibit depressive-like behaviors-revealed reduced astroglial density in the same areas as observed in humans [40]. Administration of the astroglial-toxic agent, L-alpha-aminoadipic acid, induces depressivelike symptoms in rats, suggesting that astroglial loss is pathogenic in MDD [41].
Post-mortem studies of MDD subjects documented reduced oligodendroglial density in the prefrontal cortex and amygdala [54][55][56][57]66], which may correlate with brain MRI focal white matter changes occasionally noted in some MDD patients [57]. However, microvascular abnormalities may also contribute to these changes [57].
In contrast to neurodegenerative disorders that are commonly associated with astroglial proliferation [136], psychiatric disorders are instead associated with either reduced or unchanged astroglial density [138]. The lack of increased glial density in early-onset psychiatric disorders [44,138] may reflect the slower rate of degenerative progression in psychiatric illnesses [138].
We postulate that degenerative changes associated with psychiatric disorders are subtler and not severe enough to provoke astroglial intracellular transcription factors that positively regulate astrogliosis, including signal transducer activator of transcription 3 and nuclear factor kappa B (NF-κB) [136].
While the majority of post-mortem studies focused on the alteration of glial density in MDD, BPD, and schizophrenia, others described alteration of glial cell morphology, with mixed findings. In MDD and BPD, glial size is either increased or unchanged [55]. One study found reduced glial size in BPD and schizophrenia but not in MDD [43]. A post-mortem study of depressed patients who committed suicide found increased astroglial size in the anterior cingulate white matter but not in the cortex [158]. One study in schizophrenic subjects found markedly decreased astroglial size in layer V of the dorsolateral prefrontal cortex, notwithstanding that astroglial density is double that of controls in the same layer [48]. The mixed results may partially reflect earlier studies of glial alterations in psychiatric illnesses that did not specify astroglia versus oligodendroglia [148].
Glial loss in psychiatric illnesses may contribute to neuroinflammation through several mechanisms, including abnormal cytokine levels (see Cytokine section), dysregulated glutamate metabolism (see Glutamate section), elevated S100B protein (see S100B section), and altered BBB function (see Blood brain barrier section), resulting in impaired cognition and behavior [44,45,54,133,159].

Microglial histopathology
Microglia are the resident immune cells of the CNS. They provide ongoing immune surveillance and regulate developmental synaptic pruning [160,161]. CNS injury transforms ramified resting microglia into activated elongated rod-shaped and macrophage-like phagocytic amoeboid cells that proliferate and migrate towards the site of injury along chemotactic gradients (that is, microglial activation and proliferation (MAP)) [161]. Human microglial cells express NMDARs that may mediate MAP leading to neuronal injury [162].
In MDD, BPD and schizophrenia, the results of postmortem studies investigating the presence of MAP are mixed. Post-mortem studies revealed elevated MAP in only one out of five MDD subjects [67]. In some BPD disorder patients, increased human leukocyte antigen-DR-positive microglia displaying thickened processes were documented in the frontal cortex [69]. In schizophrenia, while some studies reported elevated MAP relative to controls, others showed no difference between groups [22,67,70]. In a post-mortem study assessing MAP in MDD and BPD; quinolinic acid-positive microglial cell density was increased in the subgenual anterior cingulate cortex and anterior midcingulate cortex of MDD and BPD patients who committed suicide relative to controls [53]. Post-hoc analysis revealed this increased MAP was solely attributable to MDD and not BPD, since the positive microglial immunostaining in MDD subjects was significantly greater than that in the BPD subgroup in both the subgenual anterior cingulate and midcingulate cortices, and since the microglia density was similar in both BPD and control groups [53]. A study comparing all three disorders (nine MDD, five BPD, fourteen schizophrenia, ten healthy controls) demonstrated no significant difference in microglial density across the four groups [68].
These mixed results may be attributed to variable microglial immunological markers used among different studies [70] and/or the failure to control for disease severity [22,53,68]. Notably, three post-mortem studies of MDD and schizophrenic subjects documented a strong positive correlation between MAP and suicidality in the anterior cingulate cortex and mediodorsal thalamus, independent of psychiatric diagnosis [22,53,68]. Thus, MAP may be a state rather than a trait marker for MDD and schizophrenia.
In OCD, animal models suggest that dysfunction and reduction of certain microglial phenotypes, such as those expressing the Hoxb8 gene, which encodes homeobox transcription factor, can cause OCD-like behavior [71,72].
Hoxb8 knockout mice exhibit excessive grooming behavior and anxiety in association with reduced microglial density [71,72]. This excessive grooming behavior resembles the behavioral characteristics of human OCD. Hoxb8 injection in adult Hoxb8 knockout mice reverses microglial loss and restores normal behavior [71,72]. The role of these specific microglial phenotypes in human OCD is unclear.
In vitro animal studies suggest that the ratio of harmful versus neuroprotective microglia can be influenced by the net effect of inflammatory counter-regulatory mechanisms [15,74,164,166]. These mechanisms include the number of neuroprotective CD4 + CD25 + FOXP3 + T regulatory cells ((T regs) Figure 1) [15,74,164,166] and brain cytokine levels; low IFN-γ levels may promote neuroprotective microglia (Figure 2) [166], whereas high levels can promote the harmful phenotype [166].
In MDD, the most recent meta-analysis (29 studies, 822 MDD, 726 healthy controls) of serum proinflammatory cytokines confirmed that soluble IL-2 receptor, IL-6 and TNF-α levels are increased in MDD (trait markers) [91], while, IL-1β, IL-2, IL-4, IL-8 and IL-10, are not statistically different from controls [91]. In a primary cytokine study comparing MDD subgroups (47 suicidal-MDD, 17 non-suicidal MDD, 16 health controls), both sera IL-6 and TNF-α were significantly higher, while IL-2 levels were significantly lower in MDD subjects who committed suicide relative to both other groups [96]. This finding suggests that IL-6 and TNF-α are also state markers of MDD [96]. The decrease of serum IL-2 levels associated with acute suicidal behavior may reflect increased binding to its upregulated receptor in the brain; parallel to the aforementioned meta-analysis showing increased soluble IL-2 receptor in MDD [91]. Studies investigating the clinical significance of cytokines in MDD showed that serum cytokine levels are elevated during acute depressive episodes [171,172] and normalized following successful, but not failed, treatment with antidepressants [17] and electroconvulsive therapy [29]; these findings suggest a possible pathogenic role for cytokines.
In BPD, serum cytokine alterations were summarized in a recent review; TNF-α, IL-6 and IL-8 are elevated during manic and depressive phases, whereas IL-2, IL-4 and IL-6 are elevated during mania [92]. Other studies showed that sera IL-1β and IL-1 receptor levels are not statistically different from healthy controls [92], although tissue studies documented increased levels of IL-1β and IL-1 receptor in the BPD frontal cortex [69].
In schizophrenia, results from studies investigating cytokine abnormalities are conflicting ( Table 2). While some studies found both decreased serum proinflammatory (IL-2, IFN-γ) and increased serum and CSF antiinflammatory cytokines (IL-10) [52], others found elevated serum proand antiinflammatory cytokines, with a proinflammatory type dominance [22,173,174]. One cytokine meta-analysis (62 studies, 2,298 schizophrenia, 858 healthy controls) showed increased levels of IL-1R antagonist, sIL-2R and IL-6 [174]. However, this study did not account for the use of antipsychotics, which is thought to enhance proinflammatory cytokine production [52]. A more recent cytokine meta-analysis (40 studies, 2,572 schizophrenics, 4,401 controls) that accounted for antipsychotics, found that TNF-α, IFN-γ, IL-12 and sIL-2R are consistently elevated in chronic schizophrenia independent of disease activity (trait markers), while IL-1β, IL-6 and transforming growth factor beta positively correlate with disease activity (state markers) [173]. Cell cultures of peripheral blood mononuclear cells (PBMC) obtained from schizophrenic patients produced higher levels of IL-8 and IL-1β spontaneously as well as after stimulation by LPS, suggesting a role for activated monocytes/macrophages in schizophrenia pathology [175].
In OCD, results from a random survey of sera and CSF cytokines, and LPS-stimulated PBMC studies, are inconsistent [93][94][95][176][177][178][179]. There is a correlation between OCD and a functional polymorphism in the promoter region of the TNF-α gene [34], although low-powered studies did not confirm this association [180]. Therefore, the mixed results from studies documenting either increased or decreased TNF-α cytokine levels [93,[176][177][178]] may reflect their variable inclusion of the subset of OCD subjects with this particular polymorphism in their cohorts.
Th1 and Th2 predominant immunophenotypes have been proposed for MDD and schizophrenia, respectively, based on peripheral, rather than CNS, cytokines patterns [52,173]. We believe that peripheral cytokines patterns are unreliable surrogate markers of those in the CNS. Indeed, peripheral cytokine levels can be influenced by many extra-CNS variables, which are not consistently controlled for in several of the peripheral cytokines studies, including: 1) age, body mass index, psychotropic medications, smoking, stress and circadian fluctuations; 2) the influence of disease activity/state on the production of selected cytokines synthesis [95,173]; and 3) the effects of psychotropic agents on cytokines production [52]. The short half-lives and the rapid turnover of serum cytokines [181] (for example, 18 minutes for TNF-α [182] versus 60 minutes for IL-10 [183]), may further limit the reliability of interpreting their levels measured from random sera sampling.
In MDD, there is a consensus that a proinflammatory Th1 immunophenotype response predominates (Table 2) [17,29]. High levels of quinolinic acid in post-mortem MDD brains [53], suggest the presence of an upregulated Th1 response (Figure 1) [21,29]. Elevated CNS quinolinic acid can promote calcium influx mediated apoptosis of human astroglia [184], which hypothetically may blunt the Peripheral resting T lymphocytes constitutively express mGluR5. Activated T lymphocytes, but not resting T lymphocytes, can cross the BBB. In the animal models, the interaction between TCR of activated T lymphocytes and their cognate antigen presenting cells downregulates mGluR5 and induces mGluR1 expressions. Experimental data suggest that excess glutamate can bind to lymphocytic mGluR1 receptors, promoting production of Th1 cytokines. Hypothesis: In some MDD patients, parallel to experimental data, binding of excess CNS glutamate to induced lymphocytic mGluR1 receptors may contribute to an excess Th1 response, including IFN-γ. We further hypothesize that IFN-γ in a small quantity, similar to its in vitro effects on microglia, may induce microglial expression of MHC-II and EAAT-2, allowing microglia to serve as cognate antigen presenting cells and to provide glutamate reuptake function, thereby transforming harmful microglia into neuroprotective phenotype that participate in eliminating excess extracellular glutamate and reducing its excitotoxicity. Therefore, we hypothesize that excess Th1 response in some MDD patients is a double-edged sword; promoting harmful inflammation and serving as a beneficial counter-regulatory mechanism that may limit excess glutamate-related neuroexcitotoxicity? AMPA, 2-amino-3-(5-methyl-3-oxo-1,2-oxazol-4-yl)-propanoic acid; APC, antigen presenting cell; BBB, blood-brain barrier; CNS, central nervous system, EAAT, excitatory amino acid transporter; IDO, indoleamine-2,3-dioxygenase; IFN-γ, interferon gamma; IL, interleukin; KMO, kynurenine 3-monooxygenase; mGluR1/5, metabotropic glutamate receptors 1 and 5; MHC II, major histocompatibility complex class 2; NMDA, N-methyl-D-aspartate; NO, nitric oxide; NR1, glycine site; QA, quinolinic acid; TCR, T-cell receptor; Th, T-helper; TNF-α, tumor necrosis factor alpha.
CNS hyperglutamatergia may also contribute to an excess Th1 response in the brain (Figure 2). An in vitro study suggests that the peripheral resting T lymphocytes constitutively express metabotropic glutamate receptor 5 (mGluR5) [164], whose binding to glutamate inhibits lymphocytic IL-6 release, thereby downregulating autoreactive T-effector cell proliferation [164]. Activated T lymphocytes, but not resting T lymphocytes, can cross the BBB [37].
Experimental data suggest that the interaction between T cell receptors of activated T lymphocytes and their cognate antigen presenting cells can downregulate mGluR5 and induce mGluR1 expressions [164]. In animal models, binding of excess glutamate to lymphocytic mGluR1 receptors promotes production of Th1 cytokines, including IFN-γ [164].
We hypothesize that in some MDD patients, parallel to experimental data [164], the binding of excess CNS glutamate to induced lymphocytic mGluR1 receptors may contribute to an excess Th1 response, including IFN-γ ( Figure 2). We speculate that IFN-γ in a small quantity, similar to its in vitro effects on microglia [166], may induce microglial expression of MHC-II and EAAT2 [163,166], allowing microglia to serve as cognate antigen presenting cells and to provide glutamate reuptake function [163,164,166], thereby transforming harmful microglia into neuroprotective phenotype [163,166] that participate in eliminating excess extracellular glutamate [163,164,166]. Therefore, we also hypothesize that excess Th1 response in subgroups of MDD patients is a double-edged sword, promoting harmful inflammation and serving as a beneficial counterregulatory mechanism that may limit excess glutamaterelated neuroexcitotoxicity (Figure 2).
In schizophrenia, while some peripheral cytokine studies suggest the predominance of an antiinflammatory Th2 immunophenotype/response [52], others refute this [173,174]. However, we agree with the authors who hypothesized that the Th2 response is the dominant phenotype in schizophrenia [52]. Elevated brain, CSF, and serum levels of KYNA [21,52] suggest downregulation of microglial IDO and KMO, which is a function of Th2 response that shifts astroglial KYN catabolism towards KYNA synthesis ( Figure 1) [21,52]. Reduced KMO activity and KMO mRNA expression in post-mortem schizophrenic brains [73] is consistent with excess Th2 response (Figure 1). Increased prevalence of Th2-mediated humoral immunity abnormalities in subgroups of schizophrenia patients-as evidenced by increased B cell counts [21,76], increased production of autoantibodies including antiviral antibodies [76] and increased immunoglobulin E [52]-adds further support to the Th2 response dominance hypothesis.
The role of S100B S100B is a 10 kDa calcium-binding protein produced by astroglia, oligodendroglia, and choroid plexus ependymal cells [196]. It mediates its effects on the surrounding neurons and glia via the receptor for advanced glycation end-product [196]. Nanomolar extracellular S100B levels provide beneficial neurotrophic effects, limit stress-related neuronal injury, inhibit microglial TNF-α release, and increase astroglial glutamate reuptake [196]. Micromolar S100B concentrations, predominantly produced by activated astroglia and lymphocytes [196,197], have harmful effects transduced by receptor for advanced glycation end product that include neuronal apoptosis, production of COX-2/PGE-2, IL-1β and inducible nitric oxide species, and upregulation of monocytic/microglial TNF-α secretion [21,196,198].
Serum and, particularly, CSF and brain tissue S100B levels are indicators of glial (predominantly astroglial) activation [199]. In MDD and psychosis, serum S100B levels positively correlate with the severity of suicidality, independent of psychiatric diagnosis [200]. Post-mortem analysis of S100B showed decreased levels in the dorsolateral prefrontal cortex of MDD and BPD, and increased levels in the parietal cortex of BPD [196].
Meta-analysis (193 mood disorder, 132 healthy controls) confirmed elevated serum and CSF S100B levels in mood disorders, particularly during acute depressive episodes and mania [201].

Neuroinflammation and increased oxidative stress
Oxidative stress is a condition in which an excess of oxidants damages or modifies biological macromolecules such as lipids, proteins and DNA [206][207][208][209]. This excess results from increased oxidant production, decreased oxidant elimination, defective antioxidant defenses, or some combination thereof [206][207][208][209]. The brain is particularly vulnerable to oxidative stress due to: 1) elevated amounts of peroxidizable polyunsaturated fatty acids; 2) relatively high content of trace minerals that induce lipid peroxidation and oxygen radicals (for example, iron, copper); 3) high oxygen utilization; and 3) limited anti-oxidation mechanisms [206,207].
In MDD, increased superoxide radical anion production correlates with increased oxidation-mediated neutrophil apoptosis [206]. Serum levels of antioxidant enzymes (for example, superoxide dismutase-1) are elevated during acute depressive episodes and normalize after selective serotonin reuptake inhibitors (SSRIs) treatment [206]. This suggests that in MDD, serum antioxidant enzyme levels are a state marker, which may reflect a compensatory mechanism that counteracts acute increases in oxidative stress. [206]. In schizophrenia by contrast, CSF soluble superoxide dismutase-1 levels are substantially decreased in early-onset schizophrenic patients relative to chronic schizophrenic patients and healthy controls. This suggests that reduced brain antioxidant enzyme levels may contribute to oxidative damage in acute schizophrenia [210], though larger studies are needed to confirm this finding.
Several additional experimental and human studies examined in more detail the mechanisms underlying the pathophysiology of increased oxidative stress in psychiatric disorders . In animal models of depression, brain levels of glutathione are reduced while lipid peroxidation and NO levels are increased [206,262].
Postmortem studies show reduced brain levels of total glutathione in MDD, BPD [206] and schizophrenic subjects [206,207]. Fibroblasts cultured from MDD patients show increased oxidative stress independent of glutathione levels [262], arguing against a primary role of glutathione depletion as the major mechanism of oxidative stress in depression.
Microglial activation may increase oxidative stress through its production of proinflammatory cytokines and NO [206][207][208][209]. Proinflammatory cytokines and high NO levels may promote reactive oxygen species (ROS) formation, which in turn accelerates lipid peroxidation, damaging membrane phospholipids and their membrane-bound monoamine neurotransmitter receptors and depleting endogenous antioxidants. Increased ROS products can enhance microglial activation and increase proinflammatory production via stimulating NF-κB [208], which in turn perpetuates oxidative injury [208], creating the potential for a pathological positive feedback loop in some psychiatric disorders [206][207][208][209]. Although neuroinflammation can increase brain glutamate levels [85,86], the role of glutamatergic hyperactivity as a cause of oxidative stress remains unsubstantiated [207].
Mitochondrial dysfunction may contribute to increased oxidative stress in MDD, BPD and schizophrenia [206]. Postmortem studies in these disorders reveal abnormalities in mitochondrial DNA, consistent with the high prevalence of psychiatric disturbances in primary mitochondrial disorders [206]. In vitro animal studies show that proinflammatory cytokines, such as TNF-α, can reduce mitochondrial density and impair mitochondrial oxidative metabolism [211,212], leading to increased ROS production [206,213]. These experimental findings may imply mechanistic links among neuroinflammation, mitochondrial dysfunction and oxidative stress [206,213], meriting further investigation of these intersecting pathogenic pathways in human psychiatric disorders.
The vulnerability of neural tissue to oxidative damage varies among different psychiatric disorders based on the neuroanatomical, neurochemical and molecular pathways involved in the specific disorder [207]. Treatment effects may also be critical, as preliminary evidence suggests that antipsychotics, SSRIs and mood stabilizers possess antioxidant properties [206,207,262]. The therapeutic role of adjuvant antioxidants (for example, vitamins C and E) in psychiatric disorder remains to be substantiated by highpowered randomized clinical trials. N-acetylcysteine shows the most promising results to-date, with several randomized placebo-controlled trials demonstrating its efficacy in MDD, BPD and schizophrenia [207].

Blood-brain barrier dysfunction
The BBB secures the brain's immune-privileged status by restricting the entry of peripheral inflammatory mediators, including cytokines and antibodies that can impair neurotransmission [214,215]. The hypothesis of BBB breakdown and its role in some psychiatric patients [60,214,216,217] is consistent with the increased prevalence of psychiatric comorbidity in diseases associated with its dysfunction, including SLE [97], stroke [11], epilepsy [218] and autoimmune encephalitides (Table 1). An elevated 'CSF:serum albumin ratio' in patients with MDD and schizophrenia suggests increased BBB permeability [214].
In one study (63 psychiatric subjects, 4,100 controls), CSF abnormalities indicative of BBB-damage were detected in 41% of psychiatric subjects (14 MDD and BPD, 14 schizophrenia), including intrathecal synthesis of IgG, IgM, and/or IgA, mild CSF pleocytosis (5 to 8 cells per mm 3 ) and the presence of up to four IgG oligoclonal bands [216]. One post-mortem ultrastructural study in schizophrenia revealed BBB ultrasructural abnormalities in the prefrontal and visual cortices, which included vacuolar degeneration of endothelial cells, astroglial-end-footprocesses, and thickening and irregularity of the basal lamina [60]. However, in this study, the authors did not comment on the potential contribution of postmortem changes to their findings. Another study investigating transcriptomics of BBB endothelial cells in schizophrenic brains identified significant differences among genes influencing immunological function, which were not detected in controls [217].
Oxidation-mediated endothelial dysfunction may contribute to the pathophysiology of BBB dysfunction in psychiatric disorders. Indirect evidence from clinical and experimental studies in depression [219] and, to a lesser extent, in schizophrenia [220] suggests that increased oxidation may contribute to endothelial dysfunction. Endothelial dysfunction may represent a shared mechanism accounting for the known association between depression and cardiovascular disease [219,221], which may be related to decreased levels of vasodilator NO [221][222][223]. Experimental studies suggest that reduced endothelial NO levels are mechanistically linked to the uncoupling of endothelial nitric oxide synthase (eNOS) from its essential co-factor tetrahydrobiopterin (BH4), shifting its substrate from Larginine to oxygen [224][225][226]. Uncoupled eNOS promotes synthesis of ROS (for example, superoxide) and reactive nitrogen species (RNS) (for example, peroxynitrite; a product of the interaction of superoxide with NO) [227] rather than NO, leading to oxidation-mediated endothelial dysfunction [224][225][226].
Animal data showed that SSRIs could restore deficient endothelial NO levels [219], suggesting that anti-oxidative mechanisms may contribute to their antidepressant effects. In humans, L-methylfolate may potentiate antidepressant effects of SSRIs [228], putatively by increasing levels of BH4, which is an essential cofactor for eNOS recoupling-mediated anti-oxidation [229], as well as for the rate-limiting enzymes of monoamine (that is, serotonin, norepinephrine, dopamine) synthesis [228].
Taken together, both the recent work emphasizing the role of uncoupled eNOS-induced oxidative stress in the pathogenesis of vascular diseases [230,231] and the epidemiological studies establishing depression as an independent risk factor for vascular pathologies, such as stroke and heart disease [219,221], add further support to the clinical relevance of uncoupled eNOS-mediated endothelial oxidative damage in depression. Despite abundant evidence for cytokine abnormalities in human psychiatric illnesses and the experimental data showing that proinflammatory cytokines can reduce eNOS expression [212] and increase BBB permeability [215], human evidence that directly links excess proinflammatory cytokines to eNOS dysfunction and/or BBB impairment is lacking.

Imaging neuroinflammation in situ
Clinically, neuroinflammation imaging may prove to be crucial for identifying the subgroup of psychiatric patients with neuroinflammation who would be most likely to respond favorably to immunomodulatory therapies. Additionally, such imaging may allow clinicians to monitor neuroinflammation-related disease activity and its response to immune therapy in psychiatric patients. Imaging inflammation in the human brain has traditionally depended upon MRI or CT visualization of extravagated intravenous contrast agents, indicating localized breakdown of the BBB. Gadolinium-enhanced MRI occasionally demonstrates such breakdown in the limbic regions associated with emotional processing in patients with psychiatric disorders attributable to paraneoplastic or other encephalitides [107,109,113]. To our knowledge, however, abnormal enhancement has never been demonstrated in any classical psychiatric disorder [21,214,232], despite functional [214,216] and ultrastructural BBB abnormalities [60].
Whether or not subtler neuroinflammation can be visualized in vivo in classical psychiatric disorders remains unknown. One promising technique is positron emission tomography (PET) using radiotracers, such as C11-PK11195, which bind to the translocator protein, previously known as the peripheral benzodiazepine receptor, expressed by activated microglia [233,234].
Using this method, patients with schizophrenia were shown to have greater microglial activation throughout the cortex [235] and in the hippocampus during acute psychosis [236]. One study (14 schizophrenia, 14 controls) found no significant difference between [11C] DAA1106 binding in schizophrenia versus controls, but a direct correlation between [11C] DAA1106 binding and the severity of positive symptoms and illness duration in schizophrenia [236].
Investigators from our institution utilized C11-PK11195 PET to demonstrate bi-hippocampal inflammation in a patient with neuropsychiatric dysfunction, including psychotic MDD, epilepsy, and anterograde amnesia, associated with anti-GAD antibodies [237]. However, PK11195 PET has low signal-to-noise properties and requires an on-site cyclotron.
Accordingly, research is being devoted to developing improved translocator protein ligands for PET and SPECT. Future high-powered post-mortem brain tissues studies utilizing protein quantification aimed at elucidating metabolic and inflammatory pathways, CNS cytokines and their binding receptors, in psychiatric disorders are needed to advance our understanding of the autoimmune pathophysiology.
Several human studies showed that COX-2 inhibitors could ameliorate psychiatric symptoms of MDD, BPD, schizophrenia and OCD (Table 3) [248]. By contrast, adjunctive treatment with non-selective COX-inhibitors (that is, non-steroidal antiinflammatory drugs (NSAIDs)) may reduce the efficacy of SSRIs [249,250]; two large trials reported that exposure to NSAIDs (but not to either selective COX-2 inhibitors or salicylates) was associated with a significant worsening of depression among a subset of study participants [249,250].
In the first trial, involving 1,258 depressed patients treated with citalopram for 12 weeks, the rate of remission was significantly lower among those who had taken NSAIDs at least once relative to those who had not (45% versus 55%, OR 0.64, P = 0.0002) [249]. The other trial, involving 1,545 MDD subjects, showed the rate of treatmentresistant depression was significantly higher among those taking NSAIDs (OR 1.55, 95% CI 1.21 to 2.00) [231]. The worsening of depression in the NSAID groups may not be mechanistically linked to NSAID therapy but instead related to co-existing chronic medical conditions [10,[12][13][14][15][16][17][18] that necessitate long-term NSAIDs and which are known to be independently associated with increased risk of treatment-resistant depression [249,251]. Future studies investigating the impact of NSAIDs on depression and response to antidepressants in humans are needed.
In other experimental studies utilizing acute-stress paradigms to induce a depression-like state in mice, citalopram increased TNF-α, IFN-γ, and p11 (molecular factor linked to depressive behavior in animals) in the frontal cortex, while the NSAID ibuprofen decreased these molecules; NSAIDs also attenuated the antidepressant effects of SSRIs but not other antidepressants [249]. These findings suggest that proinflammatory cytokines may paradoxically exert antidepressant effects despite overwhelming evidence from human studies to the contrary (as reviewed above), which can be attenuated by NSAIDs [249]. At least two considerations may account for this apparent paradox: 1) under some experimental conditions, proinflammatory cytokines have been associated with a neuroprotective role, [251; (for example, IFN-γ in low levels can induce neuroprotective microglia ( Figure 2) [163,166,251]); and 2) whether these responses observed in the context of an acute stress paradigm in an animal model are applicable to endogenous MDD in humans remains unclear [251]. The therapeutic effects of COX-2 inhibitors in psychiatric disorders may involve modulation of biosynthesis of COX-2-derived prostaglandins, including proinflammatory PGE2 and antiinflammatory 15-deoxy-Δ 12,14 -PGJ 2 (15d-PGJ 2 ) [252,253]. COX-2 inhibitors can reduce PGE2mediated inflammation, which may contribute to the pathophysiology of psychiatric disorders [252,253]. They may also alter the levels 15d-PGJ 2 , and the activity of its nuclear receptor peroxisome proliferator-activated nuclear receptor gamma (PPAR-γ) [252,253].
Several studies suggest that 15d-PGJ 2 and its nuclear receptor PPAR-γ can serve as biological markers for schizophrenia [253]. In schizophrenic patients, serum PGE2 levels are increased, whereas serum levels of 15d-PGJ 2 are decreased, as is the expression of its nuclear receptor PPAR-γ in PBMC [252]. While COX-2 inhibitors may limit the potentially beneficial antiinflammatory effects of the COX-2-dependent '15d-PGJ 2 /PPAR-γ pathway' , they may advantageously reduce its harmful effects, including 1) the increased risk for myocardial infarction and certain infections (for example, cytomegalovirus and Toxoplasma gondii) in schizophrenic patients [254] and 2) its pro-apoptotic effects observed in human and animal cancer tissue [255]. Other potential mechanisms of COX-2 inhibitors therapeutic effects may involve their ability to reduce proinflammatory cytokine levels [163], limit quinolinic acid excitotoxicity (as in MDD) and decrease KYNA levels (as in schizophrenia) [128].
Omega-3 fatty acid effectiveness in psychiatric disorders is unclear [248]. In a 2011 meta-analysis of 15 randomizedcontrolled trials (916 MDD), omega-3 supplements containing eicosapentaenoic acid ≥60% (dose range 200 to 2,200 mg/d in excess of the docosahexaenoic acid dose) significantly decreased depressive symptoms as an adjunctive therapy to SRIs (P <0.001) [246]. A subsequent metaanalysis, however, concluded that there is no significant benefit of omega-3 fatty acids in depression and that the purported efficacy is merely a result of publication bias [247]. A 2012 meta-analysis of 5 randomized-controlled trials including 291 BPD participants found that depressive, but not manic, symptoms were significantly improved among those randomized to omega-3 fatty acids relative to those taking placebo (Hedges g 0.34, P = 0.025) [257]. In a randomized controlled trial of schizophrenic subjects followed up to 12 months, both positive and negative symptom scores were significantly decreased among the 66 participants randomized to long-chain omega-3 (1.2 g/day for 12 weeks; P = 0.02 and 0.01, respectively) [258]; the authors concluded that omega-3 augmentation during the early course of schizophrenia can also prevent relapses and disease progression [258].
A 2012 meta-analysis of seven randomized-controlled trials assessing omega-3 augmentation in 168 schizophrenic patients found no benefit of treatment [259]. The authors of this meta-analysis specifically stated that no conclusion could be drawn regarding the relapse prevention or disease progression endpoints [259]. Experimental data suggest that eicosapentaenoic acid and docosahexaenoic acid mediate their antiinflammatory effects by promoting synthesis of resolvins and protectins, which can inhibit leukocyte infiltration and reduce cytokine production [248].
Neurosteroids, including pregnenolone and its downstream metabolite allopregnanolone, may have a beneficial role in some psychiatric disorders [248,260]. In MDD, several studies found decreased plasma/CSF allopregnanolone levels correlating with symptom severity, which normalized after successful treatment with certain antidepressants (for example, SSRIs), and electroconvulsive therapy [261]. In schizophrenia, brain pregnenolone levels can be altered [248] and serum allopregnanolone levels may increase after some antipsychotic drugs (for example, clozapine and olanzapine) [260]. In three randomized-controlled trials (100 schizophrenia (pooled); treatment duration, approximately nine weeks) positive, negative, and cognitive symptoms, as well as extrapyramidal side effects of antipsychotics were significantly improved in one or more trials among those randomized to pregnenolone relative to those receiving placebo [248]. In one trial, the improvement was sustained with long-term pregnenolone treatment [248]. Pregnenolone can regulate cognition and behavior by potentiating the function of NMDA and GABA A receptors [248]. Furthermore, allopregnanolone may exert neuroprotective and antiinflammatory effects [248]. More RCT studies are needed to confirm the beneficial role of neuroactive steroids in early-onset psychiatric disorders in humans.
We are awaiting the results of several ongoing clinical trials investigating the therapeutic effects of other anti-inflammatory agents, including salicylate, an inhibitor of NF-κB (NCT01182727); acetylsalicylic acid (NCT01320982); pravastatin (NCT1082588); and dextromethorphan, a non-competitive NMDAR antagonist that can limit inflammation-induced dopaminergic neuronal injury (NCT01189006).

Future treatment strategies
Although current immune therapies (for example, IVIG, plasmapheresis, corticosteroids and immunosuppressive agents) are often effective for treating autoimmune encephalitides wherein inflammation is acute, intense and predominately of adaptive origin, their efficacy in classical psychiatric disorders wherein inflammation is chronic, much milder, and predominately of innate origin, is limited [2]. Development of novel therapeutics should aim at reversing glial loss [46,138], down-regulating harmful MAP, while optimizing endogenous neuroprotective T regs and beneficial MAP, rather than indiscriminately suppressing inflammation as occurs with current immunosuppressive agents. Additionally, development of potent co-adjuvant antioxidants that would reverse oxidative injury in psychiatric disorders is needed.

Conclusions
Autoimmunity can cause a host of neuropsychiatric disorders that may initially present with isolated psychiatric symptoms. Innate inflammation/autoimmunity may be relevant to the pathogenesis of psychiatric symptoms in a subset of patients with classical psychiatric disorders. Innate inflammation may be mechanistically linked to the traditional monoaminergic and glutamatergic abnormalities and increased oxidative injury reported in psychiatric illnesses. VGKC: Voltage-gated potassium channel; X AG -: Glutamate aspartate transporter; X c -: Sodium-independent astroglial glutamate/cystine antiporter system.

Competing interests
The authors declare that they have no competing interests.
Authors' contributions SN and DMP performed an extensive literature review, interpreted data, prepared the manuscript, figures, and tables. KA prepared the section pertaining to oxidative mechanisms and contributed to the manuscript revisions. AN and OD critically-revised and improved the design and quality of the manuscript. All authors read and approved the final manuscript.