Background
Microglia, myeloid-derived macrophage-like cells, are the resident immune cells of the central nervous system (CNS). Upon detection of an environmental disturbance, microglia become activated and may transition through several activation states in order to eliminate immunogens and/or promote neuroprotection. Indeed, microglia are highly active cells that exist in multiple states, constantly surveying the environment and responding to signals from neurons and other glial cells [
1,
2]. As pleiotropic cells, microglia constantly sense and respond differently to their environment depending on the stimuli they encounter. When responding to an insult, microglia release a number of factors that can be inflammatory or cytotoxic such as interleukin (IL)-1β, IL-6, tumor TNF-α, and a number of reactive oxygen and nitrogen species to neutralize immunogens. Although beneficial in the short-term, when prolonged, this form of microglial activation can also promote cellular stress and compromise the health of neural tissue, leading to neuronal damage, neurodegeneration, and subsequent deficits in cognitive or motor function.
To prevent a state of chronic inflammation, microglia are regulated by a number of factors including CD200, CD22, CD47, and fractalkine. Fractalkine (CX3CL1; FKN) is a chemokine that is expressed predominately by neurons in the CNS, with lower levels of expression occurring in astrocytes, and it has been shown to play an important neuroprotective role through the regulation of microglial activity [
3,
4]. Specifically, CX3CL1 is known to decrease microglial production of inflammatory mediators by binding to its cognate receptor (CX3CR1) on the surface of microglia (reviewed by [
5]). CX3CL1 is constitutively expressed as a membrane-bound protein, which can be cleaved by proteases, such as a disintegrin and metalloproteinase (ADAM) 10 or 17, to generate a diffusible, soluble form of the protein (sFKN [
6,
7];). Under normal physiological conditions in the periphery, membrane-bound CX3CL1 has been shown to play a role in the recruiting and adhesion of infiltrating leukocytes [
8]. sFKN, on the other hand, acts as both a chemoattractant involved in cellular migration and a neuroprotective signaling molecule that helps maintain microglia in a quiescent state [
9,
10]. While membrane-bound CX3CL1 may also bind receptors on the microglial cell surface, recent research suggests that the anti-inflammatory activity of CX3CL1 in the brain is mediated predominately by sFKN [
3,
9,
11].
The importance of the CX3CL1/CX3CR1 signaling axis for aging and disease is also well documented. Mice lacking the CX3CR1 receptor display enhanced susceptibility to inflammatory challenge with lipopolysaccharide (LPS) and 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP), chemical models of systemic inflammation, and Parkinson’s disease, respectively [
12]. Similarly, CX3CR1 knockout also accelerates disease progression in the G93A mutant Cu, Zn-superoxide dismutase (SOD1) mouse model of amyotrophic lateral sclerosis [
12]. We have also shown that mice deficient in CX3CR1 demonstrate several behavioral deficits in both hippocampal- and cerebellar-dependent learning tasks, such as contextual fear conditioning and rotarod, respectively [
13]. This effect correlates with increased levels of IL-1β and other inflammatory cytokines and can be blunted by blockade of IL-1β signaling. Furthermore, postnatal CX3CR1
−/− mice demonstrate significant abnormalities in synaptic function, such as an increased number of synapses, indicative of impairments in synaptic pruning by microglia, and electrophysiological disturbances consistent with immature synaptic function and impaired development of functional circuits [
14]. These alterations in brain connectivity carry into adulthood and correlate with behavioral deficits [
15]. Levels of CX3CL1 have been shown to be reduced in aged animals and CSF from Alzheimer’s disease patients [
16,
17]. This suggests that perturbations in the CX3CL1/CX3CR1 axis may have a significant impact on cognitive function in aging and disease, and restoring CX3CL1/CX3CR1 signaling may be a viable therapeutic approach to treat these conditions.
Indeed, intrastriatal administration of the chemokine domain of CX3CL1 significantly preserved tyrosine hydroxylase immunoreactivity following injection of 6-hydroxydopamine, and this preservation was accompanied by decreased numbers of activated microglia [
18]. Moreover, overexpression of the soluble form of CX3CL1 has been shown to mitigate neurodegeneration induced by overexpression of alpha-synuclein in a model of Parkinson’s disease, and has also been shown to reduce tau phosphorylation and improve cognition in a model of tauopathy [
11,
19,
20]. The benefits of CX3CL1 administration have also been observed more generally in aged animals, as treatment with recombinant CX3CL1, comprising only the chemokine domain, significantly increased neurogenesis, and reduced microglial activation in aged rats [
16]. The therapeutic potential of CX3CL1 in Alzheimer’s disease is more complex; however, knockout of CX3CR1 appears to be detrimental in mouse models of tauopathy, but beneficial in amyloid expressing mice [
21‐
24].
To date, most studies assessing CX3CL1 activity as it relates to aging and disease have focused predominately on sFKN signaling or signaling of a truncated soluble form containing only the chemokine domain; however, more recent work has begun to take into account the activity of full-length membrane-bound CX3CL1 as well [
9,
11,
25]. Emerging evidence suggests that these two forms may display differential effects on aging and neurodegenerative disease processes, and that these differences in activity may be highly context-specific. For example, it has been reported that, similar to CX3CR1 deficiency, amyloid precursor protein/presenilin-1 (APP/PS1) mice lacking CX3CL1 showed reduced amyloid pathology, and expression of obligate sFKN in this context did not improve or exacerbate this effect [
25]. Furthermore, deficiency in membrane-bound CX3CL1 specifically resulted in enhanced microglial activation and tau phosphorylation. On the other hand, in models of Parkinson’s disease, it was shown that membrane-bound CX3CL1 had no effect on overall disease progression and pathology while sFKN administration significantly improved motor function and preserved TH-positive neurons in the
substantia nigra [
9,
11]. These studies suggest that membrane-bound CX3CL1 and sFKN may display differing degrees of therapeutic efficacy depending on disease context; however, the individual roles of membrane-bound CX3CL1 and sFKN on motor function and cognition in a normal physiological setting have not yet been elucidated and may shed light on the therapeutic benefits and functions of each form of CX3CL1.
In this study, we confirm that CX3CL1 deficiency was sufficient to induce cognitive impairment. Furthermore, we used CX3CL1 knockout (CX3CL1−/−) mice to evaluate the differential abilities of both a mutated, obligate membrane-bound form of CX3CL1 (mFKN) and sFKN to rescue deficits caused by suppressed CX3CL1 signaling. To our knowledge, our results are the first to demonstrate that loss of CX3CL1 leads to significant cognitive impairment, in good agreement with our previous observations for CX3CR1, and to define the differing activities of mFKN and sFKN in this context.
Discussion
Our previous work was the first to demonstrate that CX3CR1 plays a physiological role in cognition and memory [
13]. In the current study, we further demonstrate that genetic knock out of its ligand, CX3CL1, also produces physiological consequences similar to those observed with receptor knockout. In particular, CX3CL1 knockout significantly impaired hippocampal dependent learning and memory processes in both a fear conditioning task for long-term memory, and a Barnes maze task to assess spatial learning and memory (Figs.
2 and
3). Interestingly, in the Barnes maze, the deficit was a widening of the spatial search pattern during the probe trial, indicating impaired spatial mapping that is correlated with activity in the dentate gyrus [
31]. Impaired memory correlated with deficits in hippocampal neurogenesis as demonstrated by a significant decrease in both Ki67+ and DCX+ neurons within the SGZ of CX3CL1
−/− animals in comparison to their WT counterparts (Fig.
6). Moreover, recordings from isolated hippocampal slices indicated that CX3CL1
−/− mice display a marked deficit in LTP maintenance (Fig.
5). Collectively, these data suggest that cognitive dysfunction in these mice may be the result of impaired synaptic plasticity and reduced neurogenesis. While the phenotype caused by CX3CL1 knockout seems to be somewhat different than that demonstrated by CX3CR1
−/− mice, these data are in excellent agreement with our previous findings and those of others.
In addition to examining the consequences of CX3CL1 knockout on cognitive function, we also evaluated the ability of both mFKN and sFKN to rescue the effects of CX3CL1 deficiency. Using rAAV to express different forms of the CX3CL1 protein, we determined that sFKN showed a definitive trend towards improving performance in a contextual fear conditioning task for long-term memory while mFKN treatment did not alter the effects of FKN knockout in this test (Fig.
2). This may suggest that sFKN activity is particularly important for executing hippocampal-dependent associative learning and memory tasks. Further, when spatial learning and memory were assessed by Barnes maze, we similarly noted that animals treated with sFKN displayed a significantly altered search pattern when seeking the target hole during the probe trial than their GFP-treated counterparts. Indeed, these mice tended to spend more time in the target zone and significantly less time searching zones adjacent to the target when compared to GFP-treated mice (Fig.
3), signifying that sFKN activity may be broadly important for hippocampal-dependent memory tasks. Additionally, although it did not improve performance in contextual fear conditioning, administration of rAAV expressing mFKN did enhance performance in the Barnes maze in a manner similar to that of sFKN (Fig.
3). This observation could indicate a specific role for mFKN in spatial memory formation, such as an ability to enhance function within the dentate gyrus, independent of neurogenesis.
While both sFKN and mFKN appear to display some activity in enhancing hippocampal-dependent functions in CX3CL1
−/− mice, it is noteworthy that only sFKN appears to rescue hippocampal neurogenesis in these animals. Treatment with AAV-sFKN partially restored expression of both Ki67 and DCX in the SGZ, indicative of increased neurogenesis, suggesting that its ability to mitigate cognitive deficits in CX3CL1
−/− mice could be dependent on this activity. However, mFKN did not appear to have any effect on neurogenesis in this region despite its ability to improve spatial memory (Fig.
6). In spite of this discrepancy, both sFKN and mFKN appear to partially restore LTP (Fig.
5). While it has been established that neurogenesis can play an important role in facilitating LTP in mice, it has also been observed that mice deficient in hippocampal neurogenesis develop compensatory mechanisms to sustain LTP [
32]. Although, mFKN-treated CX3CL1
−/− mice showed LTP maintenance, examining a single signal (Fig.
5b) that showed the varied fEPSPs post theta burst indicating inconsistent maintenance of LTP. Collectively, this could suggest that mFKN may play a role in facilitating the formation of such compensatory mechanisms in CX3CL1
−/− mice in order to partially restore LTP; however, mFKN signaling may not be sufficient in and of itself to reliably sustain this function.
While our observations on the effect of CX3CL1 knockout and restoration on cognitive function seem to be in good agreement with our prior findings in CX3CR1
−/− mice, the effects of CX3CL1 knockout on motor learning and function were different from those previously observed in CX3CR1
−/− mice. Indeed, CX3CL1 knockout appeared to enhance motor performance as assessed by a rotarod task in direct contrast to knockout of CX3CR1, which significantly impairs motor function [
13]. Despite improved motor performance, however, motor learning, which was measured as the rate at which mice improved in the rotarod task over time, did not show any differences when comparing CX3CL1
−/− mice to their WT counterparts (Fig.
4b). Indeed, although CX3CL1
−/− animals showed a slight trend towards increased motor learning, this difference did not reach significance when compared to the other treatment groups.
The effects of restoring mFKN and sFKN signaling on rotarod performance were also evaluated in CX3CL1
−/− mice. Interestingly, all mice, regardless of treatment, learned the task at a similar rate, indicating no differences in motor learning between groups (Fig.
4b); however, mFKN and sFKN treatment had opposing effects on the overall motor performance. While mFKN-treated mice behaved more similarly to GFP-treated CX3CL1
−/− mice, displaying significantly enhanced motor performance and endurance in comparison to WT mice, particularly on day two of the rotarod task, treatment with AAV-sFKN altered motor performance such that it was more similar to that of WT controls (Fig.
4a). Open field observations did not reveal any significant differences in spontaneous locomotor activity between mFKN-treated, sFKN-treated, CX3CL1
−/−, or WT mice, suggesting that improvements in motor performance were not due to hyperactivity (Fig.
4c). While the process underlying these unexpected results is not clear, it is possible that enhanced motor performance may be due to peripheral effects on tissues outside the CNS such as skeletal muscle or circulating macrophages. Additionally, it is also possible that loss of CX3CL1 activity during development in specific areas of the brain involved in motor function, such as the striatum and cerebellum, may have consequences for overall motor performance. More specifically, CX3CL1 is expressed in high levels within the striatum [
33], and loss of this protein may impact the development and maturation of neural pathways associated with motor function. In this context, restoring sFKN signaling may normalize the function of such pathways, while mFKN signaling does not appear to influence this process. In similar fashion, it has recently been observed that CX3CL1 signaling is necessary for activity-dependent synaptic remodeling to occur in the cortex following sensory lesion induced by whisker cutting in mice, and that inhibition of ADAM10, one of the proteases responsible for cleavage of full-length CX3CL1 into sFKN in the brain, impairs this process [
30]. This supports the idea that membrane-bound forms of CX3CL1 may not play a significant role in the remodeling of synaptic circuits, a process that could be involved here in normalizing motor function, and suggests instead that this synaptic remodeling may be predominately mediated by sFKN signaling.
Collectively, our data suggest that CX3CL1 signaling plays an important role in maintaining normal cognitive function in mice and demonstrates that a loss of CX3CL1 signaling could underlie the development of cognitive impairment. Moreover, we demonstrate that membrane-bound CX3CL1 and sFKN display differential activities on cognitive function that could affect their suitability as therapeutic targets. As perturbed CX3CL1 signaling has been observed in both aging and disease, the CX3CL1/CX3CR1 axis has garnered significant attention as a potential target for the treatment of several neurodegenerative diseases including Alzheimer’s disease, Parkinson’s disease, ALS, and multiple sclerosis, as well as ischemic stroke [
9,
11,
12,
17‐
25,
29,
34‐
36]. However, until recently, the importance of considering the differential functions of different forms of the CX3CL1 protein had not been taken into account. With this in mind, our data could have significant implications for the development of treatments targeting the CX3CL1/CX3CR1 axis as it suggests that sFKN has a much greater and more consistent impact on mitigating cognitive deficits than membrane-bound forms of the CX3CL1 protein, and thus may be a better therapeutic candidate for treating diseases with a significant cognitive component. However, there are still significant gaps in our knowledge regarding the functions of these two forms of the CX3CL1 protein as well as limitations to the current study that must be taken into account.
For example, the differences observed between sFKN and mFKN activity in hippocampal-dependent behavioral tasks could be due to differences in overall availability of the two proteins. As CX3CL1 is not expressed by every neuron following rAAV-PhP.B administration, the inability of mFKN to freely diffuse to nearby neurons could significantly impact its activity in comparison to sFKN. Moreover, although equal concentrations of mFKN and sFKN rAAV were administered to the mice, expression of the sFKN protein we observed was approximately 2-fold higher than that observed for mFKN (Fig.
1). Thus, it cannot be ruled out that higher expression levels of sFKN contributed to its greater and more consistent effects on cognitive function, neurogenesis, and LTP in comparison to mice treated with mFKN, which showed expression comparable to WT levels (Fig.
1). This could also suggest that over-expression of sFKN is required to induce positive alterations in cognition in the context of aging or disease; however, further study is needed to test this hypothesis. Furthermore, it is important to note that levels of sFKN detected in WT brain tissue are higher than those observed for membrane-bound forms of the protein, suggesting that sFKN could be more biologically active even at physiological concentrations [
9]. As the ELISA used to detect FKN from brain homogenates in this study that detects both forms of the protein present in WT animals, the data presented here for WT mice include the total levels of both sFKN and membrane-bound CX3CL1, with sFKN being the predominate species, while the data for both sFKN- and mFKN-treated animals represent levels of only one form of the protein. Therefore, it is likely that mFKN levels in rAAV-treated animals are also over-expressed in comparison to those of membrane-bound CX3CL1 found in WT animals, though not to the extent of sFKN. With this consideration in mind, it is possible that even over-expression of mFKN may not be sufficient to positively and consistently influence cognition in the context of CX3CL1 depletion. Furthermore, while sFKN shows appeal as a therapeutic agent given its broad range of effects on cognitive processes, several studies to date have highlighted the need to better characterize the effects of sFKN, which contains the entirety of the mucin-like stalk, versus truncated versions of the soluble CX3CL1 ligand comprising only the chemokine domain [
37]. These studies have indicated that different versions of the soluble ligand can produce vastly different outcomes in the context of both neuropathic pain and Alzheimer’s disease that is likely linked to their ability to elicit different changes in microglial phenotype [
19,
20,
25,
38,
39]. This distinction has likely been a source of significant variation among studies evaluating the potential of the CX3CL1/CX3CR1 axis as a target for therapeutic development and illustrates the need for further study to define the differential roles of all forms of the CX3CL1 protein.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.