Alterations in Immune Cells and Mediators in the Brain: It's Not Always Neuroinflammation!

Abstract

Neuroinflammation was once a clearly defined term denoting pathological immune processes within the central nervous system (CNS). Historically, this term was used to indicate the four hallmarks of peripheral inflammaton that occur following severe CNS injuries, such as stroke, injury or infection. Recently, however, the definition of neuroinflammation has relaxed to the point that it is often now assumed to be present when even only a single classical hallmark of inflammation is measured. As a result, a wide range of disorders, from psychiatric to degenerative diseases, are now assumed to have an integral inflammatory component. Ironically, at the same time, research has revealed unexpected nonclassical immune actions of immune mediators and cells in the CNS in the absence of pathology, increasing the likelihood that homeostatic and adaptive immune processes in the CNS will be mistaken for neuroinflammation. Thus, we suggest reserving the term neuroinflammation for contexts where multiple signs of inflammation are present to avoid erroneously classifying disorders as inflammatory when they may instead be caused by nonimmune etiologies or secondary immune processes that serve adaptive roles.

Introduction

The number and range of disorders in which neuroinflammation is cited as an instigator and perpetuator of disease pathology is remarkable, ranging from psychiatric disorders (autism, schizophrenia, depression and obsessive compulsive disorder) to degenerative diseases [multiple sclerosis (MS), Parkinson's (PD) and Alzheimer's disease (AD)] 31, 39, 51, 67, 99. Superficially, this suggests a common, unifying pathology that could be treated with similar therapies. Yet, we know that it is not nearly so simple. Neuroinflammation means different things to different people (sometimes even within the same academic department). When such disparities arise, it is useful, first, to examine how the term is currently being applied; second, to define a minimal set of verifiable elements; and, third, to assess what tests can be used to validate the proposition. With these goals in mind, we will begin by reviewing the classical definition of neuroinflammation. We will then examine how this term has evolved in recent years to incorporate chronic disorders that meet a subset of these classical properties. And finally, we propose that this linguistic drift has, in some cases, obscured the complexity of immune molecules in physiological brain processes that extend beyond their classically defined roles and has led to assumptions about disease etiology that are incorrect.

Historical Perspective and Definitions

The Roman scholar Celsus first defined the constituents of inflammation—pain, heat, redness and swelling—in the first century AD 22. These cardinal signs specifically denote an acute inflammatory process triggered in response to invading pathogens, tissue damage or contact with an irritant. Celsus's initial definition of inflammation as a purely pathological process was subsequently revised to acknowledge its concomitant beneficial effects on tissue healing. Thus, far from being a discrete event, inflammation denotes a complex cascade of concurrent processes causing both tissue damage and repair.

Typically, inflammation is classified as either acute or chronic. While pathologically distinct, these forms of inflammation share four common cellular and molecular hallmarks: elevation in proinflammatory cytokines and chemokines, activation of macrophages, recruitment of leukocytes, and local tissue damage. Acute inflammation, due to infection or injury, begins with sentinel innate immune cells distributed throughout blood and tissues detecting pathogen or danger‐associated molecular patterns (PAMPs or DAMPs) with nonspecific pattern recognition receptors (PRRs) 13. This recognition initiates the release of inflammatory mediators, recruitment of leukocytes to the site of injury and activation of the complement cascade. Simultaneously, macrophages and dendritic cells present the self or foreign antigen, which drives release of proinflammatory cytokines and chemokines from local immune cells. While these same cells release anti‐inflammatory mediators to keep inflammation in check, the proinflammatory processes dominate until the inflammatory stimulus has been removed or sequestered. With the cessation of antigen presentation, the balance tips toward anti‐inflammatory processes, which promote tissue healing and repair. In most cases, these innate immune responses generate local inflammation that is quickly resolved with minimal secondary tissue damage 1.

Inflammatory processes unresolved by innate immune mechanisms lead to recruitment of increased numbers of leukocytes, which alert the adaptive arm of the immune system. Dendritic cells act as a communicative bridge, traveling from local sites of inflammation to lymph nodes where they engage B and T cells. In contrast to the initial innate response, adaptive immunity recognizes specific molecular structures requiring clonal expansion of antigen‐specific T and B cells: a process that takes several days and, therefore, requires longer durations of inflammation to initiate. In its most simplistic definition, chronic inflammation constitutes a persistence of inflammation beyond 6 weeks, but this is an arbitrary demarcation 23. At the mechanistic level, chronic inflammation implies a specific biological process in which inflammatory and repair mechanisms occur in parallel without resolution. Macrophage and T‐cell infiltration into tissues, the formation of granulomas, and generation of new connective tissue characterize this smoldering form of inflammation associated with numerous chronic conditions.

Inflammation within the brain, termed neuroinflammation, is definitionally much more murky, having undergone numerous reappraisals with the advent of technologies used to assess cellular and molecular processes 69. Until recently, the central nervous system (CNS) was considered “immune privileged.” Within this framework, the term neuroinflammation implied breach of the blood–brain barrier (BBB) and immune cell infiltration into the nervous system to induce pathological sequelae. Production of proinflammatory cytokines and chemokines, glial activation, and secondary cell death accompany such breaches in BBB integrity, and together with immune cell infiltration, recapitulate the four classic hallmarks of inflammation observed in the periphery 30. In cases of stroke, trauma and infection, all four criteria are present making a clear case for the dominance of neuroinflammatory processes in these situations with pathological consequences 16, 68, 106. Thus, historically, the term “neuroinflammation” was clearly defined and denoted immune‐driven pathology in the brain.

Unfortunately, this clarity has diminished in recent years. A perusal of current literature finds that measuring a single hallmark of neuroinflammation has become sufficient to define a process as inflammatory. In recent decades, the term neuroinflammation has been applied to conditions of the nervous system with no known causative insult and little change in BBB biology, leading to a wide range of neurodevelopmental, psychiatric, and neurodegenerative disorders being characterized as having an inflammatory component. However, the role of immune mediators and cells in these disorders is much more complex than simply mediating inflammation. The reason for this is best described by the notion of equifinality: all of these indicators, standing alone, can constitute evidence either of neuroinflammation or of homeostatic or adaptive responses. In fact, the field is just starting to appreciate the extent of the diverse noninflammatory roles of immune cells and mediators in the normal brain. These nonclassical roles are likely reflective of the unique immune environment in the brain characterized by an absence of dendritic cells, limited infiltration of peripheral leukocytes and distinct differentiation of microglia (in comparison with related peripheral macrophages). Together, these distinctions result in blunted inflammatory responses. Thus, it is becoming increasingly clear that all four signs of inflammation—increased cytokines, activated microglia, T‐cell recruitment and neurodegenerative tissue damage—should be assessed before applying the label of neuroinflammation to any neurological or psychiatric disorder.

Cytokines

In the past decade, detection of elevated levels of a few proinflammatory cytokines in the brain has often become the sole measure used to justify classifying a disease as inflammatory. Yet these cytokines, the communicative proteins of the immune system, are present in the normal brain throughout development, where they play myriad noninflammatory and even nonimmune roles in development, plasticity and function 34. While cytokines are thought of in terms of mediating (pro‐) or suppressing (anti‐) inflammatory responses in the body, this type of categorization may be misleading and/or incorrect when applied to their function in the brain. Here, we use studies of the archetypal proinflammatory cytokine, interleukin‐1 beta (IL‐1β), to illustrate the potential errors in assuming that elevations in proinflammatory cytokines are a proxy for neuroinflammation 36.

Outside of the brain, IL‐1β is a multifunctional regulator of the inflammatory response. Among its myriad actions, IL‐1β drives secretion of acute phase proteins from the liver and amino acid release from muscle, recruits neutrophils, drives proliferation and differentiation of B and T cells, and stimulates hormone release from the hypothalamus–pituitary–adrenal axis 38. IL‐1β is capable of eliciting all of the classical signs of acute peripheral inflammation when levels increase to a mere 10 ng/kg in circulation 25, 37. Thus, within the brain, elevations in IL‐1β should be a reliable biomarker for neuroinflammation. Indeed, measurements of IL‐1β, along with a handful of other proinflammatory cytokines, including tumor necrosis factor α (TNF‐α), IL‐6 and IL‐12, have been cited as evidence of pathological neuroinflammation in a spectrum of disorders, including autism, schizophrenia, AD, MS and depression 3, 5, 8, 10, 14, 47, 56, 60, 64, 81, 83, 100, 103, 104.

Within the healthy brain, IL‐1β serves a diverse array of noninflammatory functions throughout brain development. During prenatal development, IL‐1β acts as a neurotrophic factor inducing neuronal progenitor differentiation in the dorsal region of the spinal cord and inhibiting differentiation in the ventral region 28. In this context, IL‐1β's actions are indistinguishable from any other trophic factor patterning the dorsal–ventral axis. In later development, IL‐1β and its receptors are expressed by both neurons and glia throughout the brain, with levels peaking during periods of heightened synaptogenesis and neuronal network reorganization 21. Two IL‐1 family receptors also act as trans‐synaptic adhesion molecules inducing excitatory synaptic differentiation in the cortex and inhibitory differentiation in the cerebellum 43, 102, 108, 109, 110, thereby performing essential nonimmune roles as synaptic organizers 98. In the mature brain, long‐term potentiation (LTP), a form of synaptic plasticity that increases the information storage capacity of neuronal networks and is hypothesized as the molecular substrate of learning and memory, leads to a sustained increase in IL‐1β gene expression 82. IL‐1β levels also increase in the mouse hippocampus in response to contextual learning, and intracerebroventricular (ICV) administration of IL‐1β can actually improve memory 44, 45, 92. Moreover, learning and memory are impaired following pharmacological blockade of IL‐1 signaling with the IL‐1β endogenous competitive inhibitor, IL‐1ra 45, 46, 107. Consistent with these findings, animal models with genetic deficiencies in IL‐1 signaling by either targeted deletion of the IL‐1β receptor IL‐1R1, or overexpression of IL‐1ra, have impairments in spatial memory and fear conditioning, no LTP, and reduced short‐term plasticity 6, 93. Taken together, these results indicate that altering IL‐1β levels would have dramatic impacts on neural development and function that are distinct from any inflammatory role.

Even in the context of neuronal dysfunction, changes in IL‐1β levels or signaling cannot be assumed to be detrimental, which is implied by the term neuroinflammation. Although toxic (micromolar) levels of IL‐1β released by glia and invading peripheral immune cells exacerbate excitotoxic brain damage following ischemia 48, exogenous application of nanomolar concentrations (1000‐fold above basal brain levels) of IL‐1β has a neuroprotective effect in the face of excitotoxic exposures 97. While it is unknown how elevations in IL‐1β exacerbate brain damage in ischemic models, recent studies suggest an indirect mechanism through IL‐1β mediated recruitment of neutrophils 40. Yet sustained overexpression of IL‐1β for 2 months in the dentate gyrus with accompanying neutrophil infiltrates resulted in no evidence of neuronal or synaptic damage 89. Moreover, exogenous application of high concentrations of IL‐1β to dissociated neuronal cultures does not affect viability nor does genetic overexpression of IL‐1β in the hippocampus 20, 87, 97. While IL‐1β alterations can exert deleterious effects (as can any nonimmune protein with integral roles in brain development and function), these depend on specific—as yet unknown—combinations of dose, timing and context.

Similar misinterpretations of increases in proinflammatory cytokines as indicators of pathology can be found in models of neurodegenerative disease. For example, elevations of IL‐1β in postmortem brain tissue from AD patients has consistently been interpreted as evidence of neuroinflammatory pathogenesis 47. However, chronic overexpression of IL‐1β in the hippocampus of a mouse model of AD reduces both plaque pathology, insoluble amyloid peptide, and enhanced microglial phagocytosis of Aβ 88. This could mean that inflammation has positive, neuroprotective effects or that a change in IL‐1β signaling does not, by itself, indicate an inflammatory process.

Disruptions in cytokines classically defined as “anti‐inflammatory” also perturb brain development and function, further complicating the attribution of pathology solely to elevations in “proinflammatory” cytokines. Transgenic mice constitutively overexpressing the archetypal “anti‐inflammatory” cytokine IL‐10 show aberrant behaviors and learning deficits when tested for spatial exploration and associative learning 63. Moreover, increases in IL‐10 lead to significant decreases in fetal brain expression of IL‐1β and TNF‐α. Since IL‐1β and TNF‐α play critical roles in neurogenesis, migration, and adaptive plasticity, these decreases could lead to pathogenesis. Thus, alterations in pro‐ and anti‐inflammatory brain cytokines can signify homeostatic, adaptive or pathological processes, and cannot clearly be attributed to neuroinflammation without meeting additional criteria.

Activated Microglia

Microglial activation, with or without detection of the other hallmarks of inflammation, is another commonly cited marker of pathogenic neuroinflammation. In the case of classical neuroinflammatory diseases such as stroke and trauma, microglia contribute to pathogenesis when activated, which is typically assessed by morphology and immunohistochemistry, using specific markers. Microglia are widely assumed to function in two dichotomous and morphologically distinct states: static (ramified) or inflamed (amoeboid, with elevated HLA‐DR levels). Recent work indicates, however, that microglia (like cytokines) participate in numerous noninflammatory roles in the brain characterized by a wide range of intermediate morphologies and markers 2, 59, 85. This extraordinary variety in phenotype is just beginning to be uncovered by fractal and transcriptome analysis and likely reflects equal diversity in function 27, 54, 91.

What we have learned in recent years about the function of microglia in the brain suggests a reappraisal of the immunocentric view of these cells 79. Once thought to be synonymous with tissue macrophages, microglia have a distinct transcriptome signature when compared to other cells of the myeloid lineage, reflecting their unique nonclassical functions in the brain 11. Contrary to widely held assumptions, microglia can display conventionally “activated” profiles in the absence of any neuroinflammation 24, 26. In fetal development and early postnatal life, the majority of microglia show amoeboid morphology and markers of activation while in the cortical ventricular and subventricular zones, where they function to phagocytose the final set of neural progenitors to control the end of neurogenesis 26. In the cerebellum, spinal cord, and retina, microglia also instruct naturally occurring apoptosis of developing neurons 42, 49, 61 by employing classic inflammatory mediators (reactive oxygen species and TNF‐α). Conversely, microglia also appear to preserve layer‐specific neuronal populations through secretion of insulin‐like growth factor (IGF‐1) 17, 101. Thus, in early development, microglia regulate physiological processes while displaying morphology and markers typically associated with neuroinflammation.

At later stages of development, microglia extend their processes in apposition to synapses where they participate in activity‐dependent synaptic pruning 80. Microglia‐mediated synaptic pruning involves communication with neurons and astrocytes as well as activation of another classical inflammatory immune mediator—the complement system. The complement protein C3b is the canonical “eat me” signal to phagocytic cells in the periphery and is thought to act similarly in the brain by tagging weak synapses. Live imaging studies show that synapses with high levels of complement undergo microglia‐mediated phagocytosis, a process that requires microglial expression of complement receptor 3 (CR3) 80, 95, 96. Like the culling of neuronal excess, this reduction in overelaborated synaptic connections is necessary for the proper activity‐dependent wiring of the cortex and does not reflect neuroinflammation.

Finally, microglia also play essential roles in excitatory synaptic maturation and function. Mice lacking the chemokine receptor Cx3CR1, which is required for microglia to properly migrate to the CNS during development, show increases in synaptic transmission in the developing hippocampus 71, 111. Later in life, these increases are reversed to cause deficits in hippocampal‐dependent learning and memory, motor and contextual learning, and novel object recognition 72. Together, these results suggest that microglia can act in both immune and nonimmune contexts, roles associated with overlapping morphology and markers of activation. These nonimmune functions are necessary for the development and maintenance of functional brain connectivity and, as commonly studied, could easily be mistaken for signs of neuroinflammation. Validated markers specific for microglia that are repurposing inflammatory mediators for these physiological roles, as well as microglia that may be fueling destructive inflammatory processes, will be essential to distinguish these divergent roles of microglia in the future.

Peripheral Immune Cell Infiltration

A third hallmark of neuroinflammation is the recruitment of peripheral T cells to sites of brain injury. The neurodegenerative consequences of T cells in the brain due to BBB rupture accompanying insult or autoimmunity is widely acknowledged, and until recently, any presence of adaptive immunity in the CNS was viewed as injurious 105. However, this view has undergone significant revision in recognition of the essential role for CD4+ T cells in CNS repair and cognition 84. Experiments of central nerve damage using a model of optic nerve crush showed that contrary to expectations, depleting T cells results in increased secondary degeneration and poorer outcomes 66. Even more surprisingly, boosting the T‐cell response with adoptive transfer of autoimmune T cells recognizing myelin basic protein (T_MBP) attenuates secondary damage and increases nerve regeneration 65. These autoimmune T cells appear to drive the resolving arm of the inflammatory response by promoting a neuroprotective phenotype in local and infiltrating myeloid populations, which mediate neuro‐ and gliogenesis, axonal regeneration, and remyelination 18. Thus, the presence of T cells, even those implicated in autoimmune pathogenesis, can be therapeutic in specific contexts.

Like cytokines and microglia, T cells also play roles in many aspects of healthy brain development and function. This is best exemplified using mouse models of immunosuppression, which exhibit unexpected cognitive deficits and altered behaviors, including impairments in hippocampal‐dependent learning and memory, increased repetitive behaviors, and increased anxiety 57, 58, 75. Replenishment of immune function in these mice by adoptive transfer of WT splenocytes or by bone marrow reconstitution resolves cognitive deficits and repetitive behaviors in adult animals, indicating the importance of ongoing immune support for healthy brain function 7, 15, 32, 78. IL‐4 secreting CD4+ T cells mediate these procognitive and behavioral effects 32, 74 by homing to meningeal spaces, rather than entering the brain parenchyma as they do in acute injury 33. Within the meninges, they remotely influence brain plasticity through secretion of cytokines and neurotrophic factors. These T cell‐derived soluble factors, in turn, promote microglial phenotypes associated with neurogenesis, synaptic plasticity, and neuroprotection. Increased numbers of meningeal T cells, as seen in the T_MBP transgenic line, boost hippocampal neurogenesis and performance on cognitive tasks 17, 112. Conversely, chronic inflammation that skews T cells to a proinflammatory phenotype in the periphery or CNS induces recruitment of interferon gamma (IFN‐γ)‐expressing T cells to the meninges that impair hippocampal neurogenesis 17. Thus, T cells in the CNS can promote or impair cognitive processes depending on their subtype. Moreover, increased recruitment to meningeal spaces or the parenchyma can represent either an enhancement of adaptive physiological processes or a maladaptive turn with neurodegenerative consequences. Developing assays to discern between these two possibilities in the context of disease is an unmet challenge and currently impedes using the presence of infiltrating leukocytes as demonstrative of neuroinflammation.

Neurodegeneration

In addition to the classic signs of inflammation in the periphery, neural degeneration is also often assumed to signify an inflammatory pathogenesis. Yet there is ample evidence contradicting this view. Immune activation concurrent with neurodegeneration in different contexts may signify either an adaptive response or a result of dysfunction in the nonimmune roles of microglia and immune mediators. For example, in a mouse model of AD, triggering an acute inflammatory response within the CNS paradoxically reduces amyloid‐beta (Aβ) and plaque levels 35, 50, 73. These neuroprotective effects appear to be mediated by increased microglial phagocytic activity and secretion of neurotrophic factors 17. Strikingly, the amyloid‐forming small heat shock proteins (sHSP)—usually synonymous with inflammation—are therapeutic in animal models of MS 70. Thus, inflammatory mediators that are pathological in one disease context can be therapeutic in another. Another example of this complexity is the case of the proinflammatory cytokine TNF‐α in MS. Although elevated levels of TNF‐α are found in the cerebrospinal fluid and active lesions in MS, consistent with pathogenic neuroinflammation 53, 76, 86, 90, neutralizing TNF‐α in a clinical trial resulted in a dose‐dependent increase in MS attack frequency, duration, and severity 113. Subsequent studies in animal models suggest that the increases in TNF‐α observed in MS patients represent an adaptive response promoting remyelination through oligodendrocyte regeneration 4, 55, 62. Similarly, increases in TNF‐α in the brains of epileptic patients were once thought to propagate seizures. However, TNF‐α regulates neuronal network homeostasis under physiological conditions 94. Thus, elevations in TNF‐α may represent an attempt to scale down network activity in response to noninflammatory disease processes. Consistent with this hypothesis, increased brain levels of TNF‐α significantly inhibit seizures in mice 9. Together, these examples illustrate the rapidly accumulating evidence that enhanced immune responses typically associated with inflammation can, in some contexts, function paradoxically to attenuate neurodegenerative disease pathogenesis.

Another way that immune cells and cytokines could contribute to neurodegeneration without causing neuroinflammation is through dysfunction in their noninflammatory and nonimmune roles. In this context, reactivation in the mature brain of microglial‐mediated synaptic pruning could result in neurodegeneration. Evidence of increased C1q expression in postmortem brain tissue of AD patients supports this possibility. Moreover, C1q deficiency in a mouse model of AD attenuates neurodegeneration 41. It remains to be seen under what set of conditions microglia engage in pathogenic synaptic striping. Until then, it is unclear whether signs of microglial activation coincident with neurodegeneration represent activation of their inflammatory, or failure of their noninflammatory, functions or both. Parsing this ambiguity is essential for development of therapeutic interventions targeting microglial function.

Finally, microglial signaling can also have paradoxical effects that are dependent on context. In models of ischemia, which exhibit all four hallmarks of inflammation, increased expression of CX3CR1 is associated with microglial activation and recruitment to sites of injury and precipitates secondary neurodegeneration 19, 29, 52. However, reduced microglial expression of CX3CR1 increases microglial‐mediated neurotoxicity in animal models of MS and PD. The latter effect may be the result of altered noninflammatory roles of microglia. Since CX3CR1 interaction with neuronal CX3CL1 facilitates mature synaptic function under physiological conditions, dysfunction in this microglia–neuron communication could lead to neuronal network instability with neurotoxic consequences 77. Again, we currently lack the tools to distinguish between these possibilities, emphasizing the importance of increasing our precision in describing changes in immune cell and mediator signaling and reserving the term neuroinflammation for situations where all four hallmarks of neuroinflammation are present.

Concluding Remarks

“If many remedies are prescribed for an illness,” Chekov wrote, “you may be certain that the illness has no cure.” Here, we face an inversion of Chekov's cautionary aphorism. The term neuroinflammation is applied so broadly that in its current form, it ceases to impart useful information about the illnesses it describes, and in some cases, it can be entirely misleading. We suggest reserving the term neuroinflammation for conditions that meet all four hallmarks: elevations in proinflammatory cytokines, microglial activation, infiltration of peripheral leukocytes, and secondary degeneration. This restrictive definition then compels us to examine the ways in which conditions currently lumped under this umbrella term differ from classic neuroinflammation. The past decade has revealed an unexpected repurposing of immune mediators and cells for homeostatic and adaptive functions in the CNS with little resemblance to their classical pro‐ and anti‐inflammatory roles in the periphery. The therapeutic possibilities of tapping the neuroprotective functions of altered immune signaling in the brain abound as many immune molecules traverse the BBB with ease and modulate neuronal network activity. However, before we can harness potentially powerful immunomodulatory therapeutics, we must distinguish pathogenic from adaptive and homeostatic functions.