Experimental autoimmune encephalitis (EAE) and multiple sclerosis
Multiple sclerosis is a disease characterized by demyelination of axons as well as chronic inflammation. Multiple sclerosis exists in several forms; the majority of patients show a relapsing and remitting type of disease. These patients experience demyelination and inflammation but this resolves after some time. This process occurs multiple times during the course of disease, with each subsequent relapse being slightly worse, until they finally progress to secondary progressive multiple sclerosis [
108]. The observation that there is resolution suggests that M1/M2 dynamics might be relevant for this disease. Although the initial cause of inflammation is not clear, it has been observed that T cells, specifically Th1 and Th17 cells, are important contributors to multiple sclerosis pathology [
109]. As previously stated, Th1 cell-secreted IFNγ is a potent inducer of M1 cells, suggesting during the active phase that microglia are skewed towards M1 activation. Although T cells regulate the response, microglia and macrophages are the effector cells. Several groups have begun to examine these dynamics
in vivo using the multiple sclerosis animal model experimental autoimmune encephalitis (EAE).
An environment dominated by inflammatory cytokines favors polarization to M1 cells and inhibits an M2 switch. The consequences of this inhibition are not fully understood, but during EAE induction and progression, inflammatory factors have the potential to prevent recovery [
110]. Elevated levels of inflammatory cytokines are also observed in human multiple sclerosis [
105]. It is believed that inflammation contributes to axonal demyelination owing to neurotoxic cytokine effects on oligodendrocytes or inhibition of oligodendrocyte precursor cell proliferation and maturation [
110]. This places M1 cells as key contributors to multiple sclerosis pathogenesis. Indeed, mice lacking IL-4 or IL-4Rα showed significantly worse EAE pathology [
111]. The importance of IL-4 in EAE is also supported by observations that transduction with an IL-4 expressing viral vector reduced the symptoms of EAE [
93,
94]. Even though IL-4 has actions on other CNS cell types, its most potent effect is the induction of M2 microglia. Additionally, other M2 promoting cytokines, such as IL-33 [
98] and IL-10 [
96], have been shown to reduce the amount of demyelination [
96] and improve clinical scores [
98]. It is important to note that these EAE models were of the chronic variety, as opposed to other EAE models that only display a transient pathology. This demonstrates that altering the pro-M1 environment to one more conducive to M2 generation has beneficial effects in chronic diseases (Table
2). The mechanism behind the beneficial effects of M2 cells can be attributed to their production of neurotropic mediators that support remyelination and regeneration. Factors such as IGF1, PDGFα, TGFβ, and SPP1 are all upregulated in microglia during the recovery phase of disease [
112].
The beneficial effect of environmental modulation favoring M2 polarization can also be seen in human beings. The FDA-approved drug glatiramer acetate (GA), which has been shown to be useful in treating relapsing and remitting multiple sclerosis, works by inducing a Th1 to Th2 shift, resulting in the production of anti-inflammatory cytokines [
97]. Even though changing the environment seems to be beneficial in alleviating symptoms for the relapsing and remitting type of multiple sclerosis, the majority of patients ultimately experience progressive disease, suggesting that the environment is not the only factor in controlling inflammation. Since multiple sclerosis is a chronic disease that takes many years to progress, the continuous long-term activation of microglia has the potential to alter microglial function, either by making them less responsive to anti-inflammatory signals or less adept at phagocytosis. This potential failure of microglia to perform their proper function is also shared by other neurodegenerative diseases characterized by persistent, long-term inflammation. One of the best examples of this is Alzheimer’s disease.
Alzheimer’s disease
The idea that microglial activation states could impact Alzheimer’s disease has recently gained momentum. Alzheimer’s disease is the most common form of dementia and is characterized by the presence of neurofibrillar tangles of hyperphosphorylated Tau and extracellular deposits of the peptide amyloid β (Aβ), forming neuritic plaques. Another key feature of Alzheimer’s disease is the presence of prominent neuroinflammation [
113]. Interestingly, Aβ itself has been shown to have proinflammatory properties when injected into the CNS [
114,
115]. Amyloid β can bind to several innate immune receptors present on microglia, such as TRL2 [
114], TRL4 [
116], TLR6 [
116], and CD14 [
117], all of which can lead to activation when triggered. To that end, microglia surrounding Aβ plaques show elevated production of inflammatory factors [
118]. The inflammatory nature of amyloid has been recognized as a potential mechanism of disease progression. Amyloid β is generated from a parent protein called the amyloid precursor protein (APP), which is cleaved in two steps by the enzymes β then γ secretase, leading to release of the Aβ fragment [
119]. This is a normal physiological process that is modified in the disease state. Interestingly, inflammation can promote accumulation of Aβ by elevating APP levels and the activity of cleavage enzymes [
120]. These observations have led to the inflammatory cascade hypothesis of Alzheimer’s disease, which states that Aβ deposition induces neuroinflammation, which in turn generates more Aβ, resulting in a vicious cycle [
120].
There are several ways for Aβ to be removed from the brain. Amyloid β can be directly shuttled out of the brain via protein complexes such as LRP1 and apolipoprotein E, which can bind extracellular Aβ and transport them to the blood brain barrier, where they are then shuttled to the other side [
121]. Additionally, new observations suggest the existence of an alternate removal pathway, where extracellular Aβ in CNS interstitial fluid is moved into the cerebral spinal fluid in what is named the ‘glymphatic system’ [
122]. Another way to clear Aβ is via phagocytosis and degradation by resident CNS immune cells, such as microglia [
123], astrocytes [
124], and possibly neurons [
125]. This particular clearance pathway showcases the Janus face nature of microglia in that even though microglia are a primary source of inflammatory factors, they also represent a crucial element for removal of harmful material in the CNS [
123]. Thus, the failure of microglia to carry out homeostatic functions possibly underscores one mechanism of increased Aβ accumulation during disease.
When Aβ is injected into the rat CNS, microglia are observed to contain the injected peptide, demonstrating their ability to take up Aβ [
126]. However,
in-vitro evidence suggests that this phagocytic ability is inhibited during disease [
127]. The ability of microglia to phagocytize Aβ may depend on their phenotype. For example,
in-vitro treatment of microglia with the pro M1 activator lipopolysaccharide inhibited microglial phagocytosis of Aβ [
127]. Other proinflammatory cytokines, such as IFNγ and TNFα not only inhibited uptake of Aβ, but also prevented internalized Aβ degradation [
128,
129]. This demonstrates that M1 microglia might be less able to properly take up and degrade Aβ. While M1 microglia appear to be impaired in their ability to remove Aβ, M2 microglia have been demonstrated to be efficient phagocytes. Treatment with the pro M2 activating cytokine IL-4 can effectively block lipopolysaccharide-induced inhibition of Aβ phagocytosis [
127] and similar data have been obtained for IL-10 [
129]. This effect also extends to degradation of the internalized Aβ. Treatment with IL-4 or macrophage colony-stimulating factor can lower the pH of the phagosome and lysosome and allow for more efficient degradation of Aβ [
130,
131]. These
in-vitro M1 versus M2 observations are further supported by an ever-increasing set of
in-vivo data, as detailed next.
It appears that one of the reasons for Aβ accumulation in Alzheimer’s disease is the failure of microglia to react properly. As mentioned, inflammation limits the phagocytic potential of microglia [
127], but does this occur
in vivo? Early on in disease pathogenesis, there does seem to be an attempt to clear Aβ. Jimenez
et al.[
132] observed at 6 months, when Aβ begins to accumulate in the APP/PS1 Alzheimer’s disease mouse model, that there were YM1
+ cells present in the CNS; however, by 18 months YM1 mRNA levels decreased and there was a massive upregulation in inflammatory factors, suggesting a switch from M2 to M1 as pathology worsened [
132]. This is consistent with the idea that microglia become less responsive to M2 induction signals as they age, perhaps owing to an age-associated decrease in IL-4Rα levels [
133]. Correspondingly, in older, non-diseased mice there is downregulation of receptors associated with Aβ engulfment, such as scavenger receptor A and the Aβ degradation enzymes Nep, IDE, and MMP-9 [
134]. These observations suggest that the Aβ induced inflammatory environment, combined with age-associated effects on microglia, lead to a situation where M1 cells predominate and microglia lose the ability to switch phenotypes and mitigate damage.
Several groups have utilized animal models of Alzheimer’s disease to demonstrate that altering the microglial activation state can be beneficial (Table
2). As previously mentioned for multiple sclerosis, GA is a promising molecule that can alter the inflammatory environment by recruiting Th2 T cells to the CNS and induce production of IL-4. Since it has shown positive benefits in patients with multiple sclerosis, there is the potential that GA could be a useful Alzheimer’s disease treatment. Data from Michal Schwartz’s group demonstrated that GA treatment leads to increased Aβ clearance and elevated levels of neurotropic cytokines such as IGF-1 [
99]. The phenotype of microglia after treatment was similar to those treated with IL-4, suggesting that the GA effect could be IL-4 (and subsequently M2) dependent [
135]. PPARγ activation is another approach that can robustly induce the polarization of M2 microglia and may be a promising therapy for Alzheimer’s disease. Several groups have observed that prolonged treatment with PPARγ agonists can reduce Alzheimer’s disease pathology, demonstrating its potential therapeutic efficacy [
102,
136,
137]. In addition to reducing plaques, the PPARγ agonist pioglitazone increased mRNA levels of the M2 marker YM1 [
137] as well as the scavenger receptor CD36 [
102]. Recently, activation of the retinoid X receptor (RXR), which forms a heterodimer with PPARγ, was implicated as a promising therapeutic treatment for Alzheimer’s disease. Landreth’s group [
103] demonstrated that treatment with the FDA-approved RXR agonist bexarotene reduced CNS Aβ levels and improved cognition in an Alzheimer’s disease mouse model [
103]. Interestingly, PPARγ treatment has been demonstrated to polarize human monocytes to an M2 state [
138], which further supports the idea that manipulating proinflammatory M1 microglia to an M2 phenotype is a potentially viable therapeutic option.
In addition to pharmacological methods to induce M2 activation, direct use of anti-inflammatory cytokines can lead to Aβ removal. As previously stated, the definitions for multiple phenotypes of alternatively activated macrophages originate from the periphery and therefore might not be completely applicable to the CNS. However, microglia treated with different anti-inflammatory cytokines exhibit unique activation states and functions. This topic is actively being pursued, in order to better understand differences between states [
54]. For the sake of simplicity and a point of reference, we will refer to the activation states that are known for peripheral macrophages. TGFβ is a cytokine with potent anti-inflammatory properties that polarizes microglia to an M2c phenotype. Importantly, mice overexpressing TGFβ have reduced plaque loads [
139] and mice deficient for TGFβ signaling showed elevated pathology [
140].
In-vitro cultures also confirm that TGFβ treatment can enhance microglial uptake of Aβ [
141]. Interleukin-4, the prototypic M2 inducing cytokine, has been shown in several cases to mitigate Alzheimer’s disease pathology. Acute injection of 100 ng of IL-4 decreased Aβ levels in just a few days [
32]. The Aβ decrease was correlated with an increase in pro Aβ phagocytic and degradation enzymes CD36 and neprilysin that colocalized to YM1 and Arg1
+ M2 cells. Using an adeno-associated virus type 2 vector to provide sustained IL-4 expression, Kiyota
et al. observed a reduction in gliosis, decreased Aβ, and improved spatial memory [
100]. Interestingly, this same group attempted to replicate these results with IL-10 and only observed an increase in neurogenesis [
101]. This discrepancy between M2-inducing cytokines suggests that different subtypes of alternatively activated microglia have unique functions. Interleukin-4-induced M2a microglia seem to be better in terms of engulfing Aβ, while IL-10-induced M2c microglia might play more of a supportive function.
Interestingly, and somewhat surprisingly, several groups, using both immunohistochemistry and ELISA as measurements, have observed that injection of inflammatory cytokines can also result in decreased Aβ. Work from our laboratory, in which the APP/PS1 Alzheimer’s disease mouse [
142] was crossed with a mouse that conditionally expressed IL-1β, demonstrated that four weeks of sustained inflammation led to decreased Aβ plaque deposition as opposed to enhanced pathology [
143]. Other groups have observed similar effects with different proinflammatory mediators and cytokines, such as lipopolysaccharide, IFNγ, TNFα, and IL-6 [
144‐
147]. At first, this does not seem consistent with the
in-vitro observations of inflammatory cytokines impairing Aβ clearance; however, one important distinction is that the
in-vitro experiments exist in a closed system. Single treatments with cytokines
in vitro impair phagocytic functions of microglia but do not take into account other cells and how they react to the inflammatory state
in vivo. As previously mentioned, there is an established pattern of immune cell activation during inflammation. Initially there is a proinflammatory response, which gives way to an anti-inflammatory response that mitigates and repairs damage. Cells capable of secreting Th2 cytokines, such as T cells [
148‐
150] and mast cells [
151,
152] migrate to the inflamed area and are potential sources of anti-inflammatory cytokines. However, the presence and function of these cells during Alzheimer’s disease is still debated. Even though the activity of peripheral cells is not clear, endogenous cells like astrocytes and microglia have been observed to secrete IL-4 or IL-10 during pathological conditions [
153]. This important distinction between
in-vivo and
in-vitro data needs to be kept in context when observing how inflammatory stimuli affect the CNS.
Alternative activation in human beings?
The recent advances in our understanding of alternatively activated microglia and their potential efficacy in treating disease have led to greater interest in translational human studies. Unfortunately, the leap to human M2 cells is not without its own problems. As noted in Table
1, several commonly used markers, such as Arg1 and YM1 are not expressed in human myeloid cells [
33], which limits the ability to identify distinct human microglial phenotypes. However, other markers appear to be consistent. CD163-positive cells have been observed in stroke and multiple sclerosis, providing a means to identify M2 microglia in human diseases [
39,
157]. Furthermore, mutations in TREM2, a molecule implicated in both human and murine M2 microglial function, are associated with increased risk of Alzheimer’s disease [
158]. In relation to this idea, levels of M2-inducing molecules, such as resolvin D1 [
21,
159] and IL-10, were reduced in patients with Alzheimer’s disease [
160]. Moreover, resolvin D1 levels significantly correlated with worse Mini-Mental State Examination scores [
160], suggesting that the lack of factors to induce M2 polarization has potential functional relevance in human disease. To that end, efforts have been taken to examine different populations of M1 or M2 markers in human Alzheimer’s disease patients. Using several different mRNAs for M2 markers, two different populations can be identified: Alzheimer’s disease brains that are skewed towards M1 and those with an M2a bias [
161]. Furthermore, the different populations are associated with different disease stages. In what appears to be early Alzheimer’s disease, there is an M1 bias, while patients with later stage Alzheimer’s disease have an M2a bias. This suggests a potential functional relevance of different microglia populations. However, whether the transition from M1 to M2a is related to disease progression or is simply a response to the enhanced pathology is yet to be understood. Obviously, more work is needed to determine just what these populations represent.