NK cells are important players in immunity against pathogens and neoplasms. As a component of the innate immune system, they are one of the first effectors on sites of inflammation. Through their cytokine production capacities, NK cells participate in the development of a potent adaptive immune response. Furthermore, NK cells were found to have regulatory functions to limit and prevent autoimmunity via killing of autologous immune cells. These paradoxical functions of NK cells are reflected in CNS disorders. In this review, we discuss the phenotypes and functional features of peripheral and brain NK cells in brain tumors and infections, neurodegenerative diseases, acute vascular and traumatic damage, as well as mental disorders. We also discuss the implication of NK cells in neurotoxicity and neuroprotection following CNS pathology, as well as the crosstalk between NK cells and brain-resident immune cells.
The CNS has long been considered an immune-privileged site because the intracerebral injection of an Ag failed to generate a classical immune response. The brain parenchyma, the interstitial fluid, and the cerebrospinal fluid (CSF) are isolated from the blood by two barriers: the blood–brain barrier (BBB), which is localized at the level of the endothelial cells, and the blood–CSF barrier (BCSFB) at the epithelial layer of the choroid plexus. Both form complex morphological features, such as tight junctions between endothelial cells to restrict the diffusion of soluble molecules, pathogens, and immune cells. Despite these barriers, immune cells are detected in the brain parenchyma and in the CSF, where memory CD4+ T cells are predominant, supporting the hypothesis of immune surveillance of the CNS (1). In pathological conditions, the integrity of the BBB and the BCSFB can be disturbed, becoming permissive for the entry of inflammatory cells. Among them, NK cells were shown to be recruited to the CNS following several pathological conditions (2–6).
NK cells are the “founding” members of the innate lymphoid cell family. NK cells have been initially merely considered as nonspecific killers, whereas “primed” NK cells also abundantly produce cytokines and chemokines and likewise modulate the adaptive immune responses (7). Although NK cells can assist in dendritic cell (DC) maturation and T cell polarization, they can also prevent and limit immune responses via killing of APCs and lymphoid cells (7). These pleiotropic functions of NK cells might be related to their heterogeneity in phenotypes and functions. In mice, NK cells were shown to differentiate from CD11blow to CD11bhigh and could be further distinguished by their expression of CD27. The immature CD11blowCD27+ subset has the highest proliferative capacity and is mainly found in bone marrow and lymph nodes (LNs), whereas CD11bhighCD27+ and CD11bhighCD27− NK cells present lower rates of proliferation and are found in spleen and peripheral blood (8). In humans, NK cells can be subdivided into different populations based on the relative expression of the surface markers CD16 and CD56. The mature subtype CD56dimCD16+, found mainly in peripheral blood, has reduced proliferative capacity and produces negligible amounts of cytokines but is highly cytotoxic (9). In contrast, the more immature subset, CD56brightCD16−/dim, mainly present in peripheral lymphoid organs, is able to proliferate and to secrete a large range of cytokines, though it presents minimal cytotoxic capacity (10).
Although NK cell biology remains understudied in neurologic diseases, there are arguments to think that specific brain factors could modulate NK cell function and vice versa. Indeed, NK cell physiology is modulated following the circadian rhythm through the influence of neurotransmitters, such as norepinephrine, which inhibits cytotoxic function after engagement of β-adrenergic receptors (11). Alternatively, NK cells could impact CNS physiology by killing glial cells (12) or by secreting IFN-γ (13). Most knowledge about NK cells in the CNS is coming from studies of experimental autoimmune encephalomyelitis (EAE) where conflicting results are emerging, as NK cells have been shown to be either implicated in neuroprotection (14) or neurotoxicity (15). This is highlighting the dual nature of NK cells in CNS homeostasis as it was described for autoimmune diseases in general (16). Thus, NK cells could either promote or limit autoimmune disease by their intrinsic capacities to alter the issue of adaptive immune responses or preserve immune homeostasis (16). This Jekyll and Hyde face of NK cells was observed in several organs, including the uterus where they are essential for blood supply to the fetus but where they could also be implicated in pregnancy disorders (17). Although NK cell trafficking from blood to specific organs is well characterized, the mechanisms leading to the recruitment of NK cells to the CNS remain unclear (18).
Monitoring NK cell subsets as surrogate markers for disease activity as well as understanding the mechanism of recruitment and the function of brain-infiltrating NK cells are interesting goals for the development of new NK cell–based therapies. Most studies are focusing on the modification of systemic immunity following CNS disorders. In this review we first discuss the impact of brain disease on circulating NK cell physiology. We then focus on the recruitment of NK cells within the healthy and pathological brain, which is so far understudied.
NK cell modifications associated with CNS disorders
There is increasing knowledge about the bidirectional interconnection between the nervous system and the immune system (19); nonetheless, data are patchy and complex. Furthermore, it is not clear whether the modifications of the immune system (including NK cells) that are described in several CNS pathologies are a cause or a consequence of the disease.
Primary brain tumors
Patients with glioma are known to have impaired immune function, parallel to an increase of anti-inflammatory molecules in plasma (20). Fadul et al. (21) observed decreased absolute numbers of NK cells in PBMCs isolated from glioblastoma multiforme (GBM) patients who were treated with combined radiation therapy and temozolomide. Furthermore, NK cell frequency and their cytotoxic activity (NKCA) were shown to be affected in the blood of brain tumor patients as well as in animal glioma models (22, 23). Several factors could mediate this effect on NK cell functions. The plasma levels of anti-inflammatory molecules, such as TGF-β, were shown to be elevated in the context of brain tumors and could inhibit NK cell functions, as well as the level of expression of the activating receptor NKG2D (24). Thus, decreased expression of NKG2D was shown on NK cells that were isolated from GBM patients (24). This downmodulation of NK cell functions in brain tumors reflects a common mechanism of immune escape found in other solid tumors (25).
Infections of the CNS
Shortly after they were discovered, NK cells were shown to play an important role in antiviral immune responses. The rare patients completely devoid of NK cells were extremely susceptible to infections by viruses of the herpes family (26, 27). Five consecutive children with herpetic encephalitis were shown to have functional NK cell deficiencies (27). By using a mouse model of HSV1 encephalitis, Adler et al. (28) directly demonstrated that NK cells have an important role in the defense against this severe form of infection. Nonetheless, TAP-deficient patients known to have hyporesponsive NK cells rarely present with viral encephalitis (29). In contrast, TLR3–deficient patients are susceptible to HSV encephalitis, and they present impaired NK cell IFN-γ secretion following stimulation with polyinosinic-polycytidylic acid, but normal production against K562 (30), suggesting that TLR3 dysfunction could be at the origin of this disease phenotype. Additionally, it has been described that NK cells are important actors of the antiviral immune response in mouse cerebral infections by Theiler’s murine encephalomyelitis virus (31), in mouse hepatitis virus (32), and in Semliki Forest virus (4), with a rather detrimental cytotoxic function in the latter case. Regarding SIV infection in macaques (33), it has been shown that the intensity of NK cell cytotoxicity was inversely related to the severity of CNS lesions.
Bacterial infections of the CNS.
Only a few studies addressed the role of NK cells in bacterial infections of the CNS. Hayashi et al. (34) developed a nonlethal mouse model of CNS listeriosis. Under these conditions, NK cells were identified as the main source of bacterial growth-limiting IFN-γ in the early phase of infection. Mice are relatively protected from Listeria neuroinvasion as long as they have NK cells and can respond to IFN-γ (35). In contrast, NK cell–derived IFN-γ was found to be responsible for the immune pathology associated with Streptococcus pneumoniae meningitis, whereas IFN-γ KO mice are resistant to severe forms of this disease (36).
In 1986, Stach et al. (37) described that NK cell cytotoxic function was normal in benign uncomplicated malaria, whereas it was severely decreased in children with cerebral malaria. In a mouse model of malaria induced by Plasmodium berghei ANKA, it was shown that the NK cell complex genetically influences pathogenesis, that NK cells are necessary for both disease pathogenesis and control of parasitemia, and that NK cells infiltrate the brain of mice with cerebral malaria (38). The same group demonstrated that a reciprocal activation of NK cells and DCs is implicated in the development of inflammatory pathways during malaria (39). With IFN-γ being known as a crucial factor for the development of cerebral pathology in malaria, Villegas-Mendez et al. (40) identified NK cells as the early source of this cytokine, before T cells take center stage. Overall, in CNS infections, NK cells can be beneficial or detrimental according to the experimental model used. An emerging concept states that neuroinflammation following infection may play an important role in the etiology of neurodegenerative diseases (41).
Alzheimer disease and Parkinson disease.
Alzheimer (AD) and Parkinson (PD) diseases are age-related neurodegenerative conditions characterized by neuronal loss in hippocampus and substantia nigra, respectively. There are concepts supporting that AD and PD pathogenesis could be originally related to systemic immune system dysfunction (42). The frequency of NK cells seems unmodified in the blood of AD patients (43, 44). Recently, it was described that NK cells from these patients had increased expression of the receptor 5-HT2C for the neurotransmitter serotonin (45), which was shown to be implicated in either increase or decrease of NKCA (46, 47). However, patients with PD show an increased expression of the inhibitory receptor NKG2A (48, 49) and, depending on the study, no modification (48) or an increased (49) expression of NKG2D on NK cells. NKCA was shown to be either diminished (50) or equal (51) in PD patients and controls. Furthermore, ex vivo activation of NK cells with IL-2, IFN-α, or IFN-β (51) resulted in increased NKCA in patients compared with control donors. NKCA is positively correlated with PD severity (48). Moreover, Solerte et al. (52) showed that purified NK cells from blood of PD patients spontaneously secreted more IFN-γ and TNF-α than did healthy donors. All of these modifications of NK cell physiology suggest that they might play a role in the pathogenesis of PD. Nonetheless, further investigations are required for delineating the possible mechanisms.
Multiple sclerosis (MS) is the most common disabling neurologic disease in young adults. MS is a neuroinflammatory autoimmune disease affecting the CNS, characterized by widespread inflammation with infiltration of immune cells and demyelination. The involvement of NK cells in the pathogenesis of MS has been shown by several approaches: MS relapses were correlated with a deficit of NK cells, which were proposed to modulate the autoimmune response and to participate in the repair of neurologic damage (53); the functional activity of NK cells decreases during active phases of the disease and is restored during remission (54); and the treatment of MS patients with IFN-β (55) or daclizumab (56) results in decreased disease flares accompanied by a growing population of regulatory CD56bright NK cells. Moreover, during pregnancy, MS activity is reduced and the regulatory NK cell population is increased in blood (57). Interestingly, a reduction of blood NK cells expressing CX3CR1 correlated with disease severity (54). Because CD56bright NK cells were shown to be CX3CR1− in normal donors, it would be interesting to further look at the capacity of daclizumab to modulate CX3CR1 expression on NK cell subtypes (58). Manipulation of NK cells in animal models of EAE leads to contradictory results. According to some authors, NK cells play a deleterious role (15) during myelin oligodendrocyte glycoprotein EAE in C57BL/6 mice. On the contrary, numerous studies attribute, using the same EAE model, a neuroprotective role to NK cells. Depletion of NK cells, by injecting anti-NK1.1 or anti–asialo GM1 Abs, induced worsening of the disease, associated with increased proliferation of CD4+ Th17 cells and high production of cytokines (14). Overall, this suggests that NK cells are a potential target for future MS therapy.
Acute stroke and trauma
Acute CNS diseases, such as stroke and traumatic brain injury (TBI), have been associated with increased risk of infectious complications (59). It is thought that CNS injuries break the delicate interplay between the immune system and CNS, resulting in secondary immunosuppression (60). Many studies have shown a reduced NKCA of blood NK cells of patients with spinal cord injury and TBI (61–63). Although the CD56bright population stayed unchanged, the CD56dim subtype was decreased, and this correlated with the occurrence of infections after TBI (64). Alternatively, in the context of stroke, the NK cell counts showed a nonsignificant decrease (65–67). Nonetheless, there is a decrease of NK cells expressing IFN-γ and perforin, consistent with the reduction of their NKCA in the acute phase of ischemic stroke. Interestingly, others reported that deficiency in perforin expression was negatively correlated with Fas ligand expression by NK cells in individual patients (67). The decrease in IFN-γ secretion was also observed in a mouse model of middle cerebral artery occlusion leading to spontaneous septicemia and pneumonia (68). The adoptive transfer of NK cells diminished bacterial load in the lung. These authors also showed evidence that the sympathetic nervous system participates in the immunosuppression induced by middle cerebral artery occlusion, as its blockage by a β2-adrenergic receptor antagonist prevents the defective IFN-γ response and the occurrence of bacterial infections.
A growing body of research has found that mental disorders may influence somatic health and disease by altering immunity. During acute stress, such as public speech, there is a rapid mobilization of NK cells, with a CD56dimCD16+CD62L− phenotype, from various reservoirs into the blood (69). Acute stress correlates with increased NKCA (70). The NK cell redistribution was proposed to be coordinated via secretion of catecholamines from the sympathetic nervous system into the blood and further NK cell mobilization regulated through β2-adrenergic receptors (71). In contrast, in the context of chronic stress or depression, NKCA is markedly reduced compared with controls (72). Diminished NKCA was associated with a reduction of the CD16+ NK cell subset in blood of depressed patients (73). Additionally, several studies have found that antidepressants that block the reuptake of serotonin are associated with upregulation of NK cell functions in depressed individuals (74). The mechanism by which mental dysfunction acts on NK cells is still debated. Many hypotheses are proposed, including the deregulation of the hypothalamic/pituitary/adrenal axis leading to elevated levels of blood glucocorticoids, which further inhibit NKCA through epigenetic processes (75). Several other mental disorders are correlated with a diminution of NK cell functions, such as obsessive-compulsive disorder (76) and autism (77). In the latter study, the authors proposed that NK cell impairment in early life could impede neurogenesis, resulting in subsequent mental dysfunction, especially because neuronal stem cells were shown to express NKG2D ligands (78).
Phenotype and physiology of brain NK cells in health and disease
NK cells have been found in most organs, where they present various phenotypes and functions related to the organ-specific environment (18). Although they could be recruited to the brain in response to several CNS disorders, there are sparse data about the phenotype of NK cells in the nascent CNS. In the C57BL/6 mouse model, NK1.1+αβTCR− NK cells represent 9.74% of total lymphoid cells within the brain before infection with P. berghei (38). We confirmed the presence of NK1.1+NKp46+ NK cells in the brain of C57BL/6 mice where they represented 11.1 ± 0.4% (n = 3) of CD45high cells (Fig. 1D). These NK cells were in the majority immature CD11blowCD27+ and CD11bhighCD27+ cells and in the minority mature CD11bhighCD27− cells, representing 58.9 ± 4.3, 27.0 ± 3.9, and 5.7 ± 1.3%, respectively (Fig. 1E). This profile corresponds to the one in LNs (Fig. 1F), which contrasts with a higher percentage of mature NK cells in spleen or blood (Fig. 1G, 1H).
We demonstrated the presence of immature NK cells in the brain of healthy mice. We excluded a contamination by blood NK cells, as animals were extensively perfused with PBS and because they display different phenotypes. To our knowledge, there is no evidence for the presence of NK cells in nonpathological human brain (NPB) parenchyma because of obvious ethical considerations. Nonetheless, we observed rare NK cells in the CSF of patients with either peripheral neuropathy or MS, exhibiting mainly CD56bright phenotypes (Fig. 2A). This repartition is typical of the phenotype of NK cells found in LNs and is different from blood NK cells (Ref. 10 and Fig. 2B). Recently, Hamann et al. (79) observed an overall diminished frequency of NK cells in CSF compared with blood with a CSF/blood ratio of 1:4.
The validity of comparing between human and mouse NK cell subsets and their maturation is still debated, despite that CD27high mouse NK cells show some functional and phenotypic similarity to the CD56bright human subset such as the receptor expression pattern. However, the CD27high mouse subset is highly cytotoxic as opposed to the CD56bright human NK cells (8). Nonetheless, it has been proposed that human NK cell maturation could be further delineated according to the expression of CD11b and CD27 (80), such as in the mouse. The same team demonstrated that CD56bright NK cells are mainly CD27+, indicating that human brain NK cells may display an immature phenotype, although a deeper analysis is needed to confirm this.
NK cells are known to actively screen nonlymphoid and lymphoid organs in search of infected targets and transformed cells, and the brain appears as a “new” compartment for future investigation of their trafficking. In steady-state conditions, CD4+ T cells were found to patrol from the blood through the choroid plexus into the CSF to perform CNS immunosurveillance (1). Thus it is conceivable that NK cells follow the same route to scan the CNS milieu for pathological components either directly through their surface receptors or by interaction with brain-resident immune cells (Fig. 3). NK cell subsets express a different array of chemokine receptors in humans and mice. For example, human CD56bright NK cells were shown to express high levels of CCR7 and CXCR3, so they can migrate in response to CCL19 and CXCL10, which are constitutively expressed in the CNS (81). Moreover, TGF-β is constitutively present in the CSF and was found to modulate macrophages (82). Because TGF-β was recently shown to negatively regulate maturation of NK cells, it might be relevant to investigate ex vivo whether particular CSF constituents might modify NK cell phenotypes (83). The reason for the presence of immature NK cells in the healthy CNS needs further experiments to delineate whether the brain influx of NK cells is dependent on their phenotype or whether it is the CNS milieu that further influences such influx.
Although the possible immune surveillance of the CNS by NK cells is still enigmatic, their infiltration following brain pathology is more characterized. Nonetheless, their mode of entry into CNS and the molecules leading to this recruitment are poorly understood. This recruitment could be coordinated by CNS resident cells, such as microglia, astrocytes, and neurons, that secrete chemokines involved in NK cell migration, such as CCL2, CXCL10, and CX3CL1 (3, 18, 38).
In intracranial tumors, NK cells account for a minor part of infiltrating leukocytes as shown in GBM, pilocystic astrocytomas, and brain metastases of carcinoma (2, 84–86). According to our observations, NK cells comprise 2.11 ± 0.54% of infiltrating immune cells in GBM (J. Kmiecik, A. Poli, N.H.C. Brons, A. Waha, G.E. Eide, P.O. Enger, J. Zimmer, and M. Chekenya, unpublished observations). These tumor-infiltrating NK cells were shown to be inhibited by glioma-derived factors but also by other local immune cells with anti-inflammatory properties (20). Nonetheless, immunotherapies are emerging to increase this NK cell recruitment and function. As an example, oncolytic HSV infection of GBM in mice resulted in a massive influx of NK cells. Interestingly, although NK cells recruited to tumors in control mice were mostly CD11bhighCD27low, oncolytic HSV administration resulted in recruitment of CD11bhigh/lowCD27+ NK cells (87). This subpopulation was able to activate macrophages/microglia through secretion of IFN-γ. In a rat xenograph model of GBM, we also observed an increase of proinflammatory macrophages/microglia following intracranial injection of activated NK cells leading to beneficial outcome (A. Poli, J. Wang, O. Domingues, C.B Rygh, T. Yan, F. Thorsen, J. Planagumà, E. McCormack, F. Hentges, P.H. Pedersen, J. Zimmer, P.O. Enger, and M. Chekenya, submitted for publication). Thus, brain NK cells may influence the disease outcome by interaction with other CNS immune cells.
Brain NK cells were observed in greater abundance following CNS infection (6, 38). However, NK cell recruitment could be either beneficial or detrimental depending on the infectious agent. Thus, although the recruitment of NK cells was shown to be essential for controlling mouse hepatitis virus and Semliki Forest virus replication within the CNS through an IFN-γ–dependent mechanism (3, 4), their cytolytic activity (Fas and granule exocytosis dependent) was shown to potentiate the immunopathology, such as increased animal death and brain demyelination (Fig. 4) (4). Alternatively, NK cells were shown to be essential for controlling parasitemia; however, their depletion protects C57BL/6 mice against P. berghei–mediated cerebral malaria. It seems that NK cells stimulated DCs to produce IL-12 required for T cell priming, leading to subsequent cerebral disease (Fig. 4) (38, 39). In neuroborreliosis, NK cell numbers in the CSF are significantly higher than in tick-borne viral encephalitis, in correlation with the level of IFN-γ found in CSF (88). The outcome of CNS infections varies according to the type of pathogens, and it is also dependent on interindividual genetic and environmental factors. The immune response following infection can contribute to fatal disease, but is also necessary for recovery, and the specific role of immune cells, including NK cells, is not well described (89). Furthermore, many teams propose that neurodegenerative diseases could be a consequence of CNS infections (89). Thus, repeated infection by HSV was proposed to be at the origin of AD and MS emergence, as pathological features of this CNS infection share similarity with neurodegenerative diseases, such as neuronal death (89). Moreover, several models of CNS infection in mice were associated with parenchymal demyelination (4). The recruitment of immune effectors, including NK cells, following infection leads to the release of cytotoxic agents, such as granzymes, perforin, and proinflammatory molecules, which could generate direct or indirect cellular damages leading to neurodegenerative foci (89) (Fig. 4). As a comparable example, liver NK cells were implicated in the emergence of fulminant hepatic failure following mouse hepatitis virus infection, whereas the depletion of NK cells limited liver damage and increased animal survival (90). Alternatively, IFN-γ was also associated with both neuronal protection and repair by modulation of microglia properties (91), so we can propose that NK cells secreting IFN-γ following infection may have a neuroprotective function (Fig. 4). Furthermore, NK cells were shown to secrete IL-10 following viral infection of the liver, and it could be interesting to investigate this in the context of CNS infection and neuroprotection (92). All of these facts indicate why NK cells could be implicated in neurodegeneration and/or neuroprotection (Fig. 4).
During the last decade, the major increase in knowledge about CNS NK cells arose from studies of MS patients and animal models of EAE. The EAE model revealed that NK cell homing to the CNS is essential to control neuroinflammation (Fig. 4). Indeed, recruited NK cells were shown to kill microglia responsible for the priming of autoreactive T cells (5, 14). Furthermore, there is an exacerbation of brain lesions and a concomitant reduction of brain NK cells during EAE in CX3CR1-deficient mice. This emphasizes that CX3CR1+ NK cells are essential for brain homing to further modulate EAE pathogenesis (93). In contrast, in a postmortem study of MS patients, Durrenberger et al. (94) detected rare NKp46+ NK cells in demyelinated areas. Additionally, in the CSF, Hamann et al. (79) described a reduction of NK cell frequency in the MS patients compared with NPB samples. However, a preliminary investigation of four patients with suspected MS described an increased absolute NK cell number in their CSF compared with patients with other neurologic defects (unpublished data). These NK cells display an immature phenotype with high CD56 (Fig. 2C) and low CX3CR1 (79) expression. Overall, the mechanism leading to the recruitment is not clear; nonetheless, it seems that the genetic background might influence this phenomenon. Thus, Nimer et al. (95) found that the influx of CD3−CD161bright NK cells in experimental traumatic brain contusion in DA rats was more marked than in PVG rats. This interindividual genetic heterogeneity might influence the outcome and treatment response following CNS pathology and partially explain the differences observed between studies.
In this review we have discussed the systemic NK cell phenotypic and functional alterations involved in the most common brain disorders. Note that the characterization of NK cells was often not well performed, as many studies used single staining of CD56+ or CD57+ cells (44, 45, 85, 86, 96). Both of these molecules can also be expressed by CD3+ T cells and should by consequence be excluded. Furthermore, there are many unanswered questions, and the future challenge is to delineate the precise influence of NK cells on CNS pathology, as well as the characterization of brain-resident NK cells, which have so far been understudied. Thus, it is not well understood whether brain NK cells are the origin or the consequence of CNS pathogenesis. Some evidence suggests that NK cells can control CNS autoimmunity and inflammation by their capacity to kill autoreactive or proinflammatory microglia (5, 12, 14). Alternatively, NK cell response in CNS following infection may be deleterious (39). Future studies providing novel insights into the role of NK cells in brain pathogenesis might result in the development of NK cell–based approaches to cure patients with inflammatory and autoimmune disorders of the CNS.
This work was supported by the Ministry of Higher Education and Research of Luxembourg, the Centre National de la Recherche Scientifique, the Bergen Medical Research Foundation, the Meltzer Fund, and the Norwegian Cancer Society.
Abbreviations used in this article:
blood–cerebrospinal fluid barrier
experimental autoimmune encephalomyelitis
NK cell cytotoxic activity
nonpathological human brain
traumatic brain injury.
M.B. is an employee of Innate Pharma. The other authors have no financial conflicts of interest.