Introduction
Microglia regulate white matter myelination in both developing and post-injury brains [
1,
2] and present a promising target for white matter repair [
3]. However, microglial cells are highly plastic and can acquire differential phenotypes across a range of spectrum [
3]. New single cell RNA-sequencing (scRNA-seq) studies have revealed the presence of various microglial subclusters within the brain after ischemic stroke, each displaying distinct transcriptomic profiles [
4‐
6]. Subclusters associated with resting microglia, proliferative microglia, or anti-inflammatory microglia were identified based on their expression patterns of homeostatic, inflammatory, or repair-promoting genes [
4‐
6]. Microglia can exert damaging effects post-stroke, by releasing inflammatory cytokines, hindering oligodendrocyte maturation, and promoting axonal degeneration [
7‐
9]. Alternatively, they can regulate adaptive functions by releasing restorative cytokines and growth factors, clearing tissue debris through phagocytosis, and promoting remyelination [
8,
10,
11]. The balance between detrimental and restorative state microglial subclusters is critical for initiating oligodendrocyte differentiation and the onset of remyelination [
10]. However, how to modulate specific microglial transition to anti-inflammatory and myelin-supporting functions is not well understood. A precise dissection of the specific roles of each cell type and their interactions facilitating white matter injury and myelin repair is also urgently warranted.
The Na
+/H
+ exchanger-1 (NHE1) serves as a key regulator for intracellular pH (pH
i) in microglia by extruding H
+ ions in exchange of Na
+ influx [
12,
13]. In recent studies, we reported that selective deletion of microglial
Nhe1 in the
Cx3cr1-CreER±;Nhe1flox/flox (cKO) mice reduced proinflammatory responses at 3 days post-stroke [
12] and improved oligodendrogenesis along with remyelination tested at 3, 14, and 28 days post-stroke in the
Nhe1 cKO mice [
13]. We detected enhanced microglial oxidative phosphorylation and phagocytosis of microglial cells in the
Nhe1 cKO mice at 3 days post-stroke [
13]. To further investigate the underlying mechanisms for these phenotypic changes, and to reveal whether the selective deletion of
Nhe1 in cKO brains causes uniform alterations among different microglial subpopulations, and whether it influences their functions and interactions with other cell types, we conducted scRNAseq transcriptome analysis at 3 days post-stroke timepoint to explore mechanisms between microglia-oligodendrocyte interactions. Our analysis revealed a total of 24 transcriptionally distinct cell populations within the white matter tissues, encompassing various cell types including oligodendrocytes, microglia, and astrocytes, etc. Notably, we identified 5 subpopulations for microglia (MG) and 5 for oligodendrocytes (OL), where MG3 and OL5 subpopulations were expanded in the
Nhe1 cKO white matter tissues. These subpopulations exhibited a unique genetic signature characterized by elevated lipid metabolism, enhanced phagocytosis, improved lysosome functions, and an array of myelin-supporting genes, as well as genes associated with lactate shuttling functions. In summary, our findings highlight the critical role of the pH-regulatory NHE1 protein in modulating of microglial phenotypes and microglia-oligodendrocyte interactions.
Discussion
White matter tissues are defined by dense insultation of axonal fibers by the wrapping myelin-sheath secreted by mature oligodendrocytes. However, these white matter tissues are also maintained by other support cells residing within them, including astrocytes and microglial cells. Though microglia were traditionally assumed to be a homogenous cell population, next generation sequencing analyses have demonstrated a spectrum of transcriptional states with distinct subclusters emerging under different conditions, such as early brain development (termed “axon tract-associated microglia”, ATM) [
20], Alzheimer’s disease pathogenesis (termed “disease-associated microglia, DAM) [
17], and other neurodegenerative diseases or aging (termed “microglial neurodegenerative phenotype”, MGnD) [
18]. scRNA-seq holds immense promise as a tool for investigating brain damage and recovery following ischemic stroke, not only by deepening our understanding of the complex genetic expression patterns and heterogeneity of cell responses, but also offering the potential to pave ways for future therapeutic developments with increased precision and less off-target complications. However, few studies had investigated brain cell heterogeneity in stroke conditions utilizing scRNA-seq [
4,
5]. A recent scRNA-seq study in the ischemic hemispheres of adult male C57BL/6J mice successfully identified distinct principal cell clusters including microglia, oligodendrocytes, astrocytes, neutrophils, and CNS-associated macrophages, unveiling their respective cell type subpopulations and differential gene expression patterns within 24 h after tMCAO [
5]. In another stroke scRNA-seq study, the authors dissected immune cell responses in the brain border compartments compared to the brain parenchyma and discovered a unique phenotype of myeloid cells involved in the inflammatory response to injury, termed “stroke-associated myeloid cells” (SAMC) [
47], which also presented internalized myelin droplets in the brains of human stroke patients. Despite these efforts, most of the studies were conducted in hemispheric brain tissues, and there is a lack of scRNAseq analysis specially in white matter tissues, where cellular compositions and responses to stroke injury can vary significantly compared to those in the gray matter.
In this study, we identified a novel population of microglia (MG3) that only appeared in the white matter tissues in response to stroke (in both WT and cKO) but not in the non-stroke white matter tissues (Fig.
3c). These stroke-associated white matter microglia (SAWM) showed lower expressions of the homeostatic microglial markers (
P2ry12, Tmem119, etc.) along with elevated expressions of the DAM- and/or MGnD-related genes (
Apoe, Trem2, Ctsb, Ctsd, etc.) [
17] (Fig.
2b). Most interestingly, compared to other microglial subclusters, these SAWM also exclusively increased all the signature genes for ATM [
20], highlighted by the high expressions of
Spp1, Gpnmb, Lgals3, and Lgals1 (Fig.
3b), where the ATM subcluster of microglia was reported to only appear during a restricted developmental window before myelination occurred, while the tissues in which they were concentrated eventually became heavily myelinated, such as the corpus callosum and cerebellum [
20]. In addition, among all cell types and/or subclusters, the SAWM population was the most highly involved in the SPP1 signaling pathway network, similar to those in the homeostatic MG1 subcluster (Fig.
8b), further validating the similarity to the physiological microglial phenotype in supporting white matter remyelination. In summary, here we not only revealed a complete landscape of the cell heterogeneity specifically in the white matter tissues after stroke, but also identified the novel SAWM population that could be critical to exert white matter repairing functions after stroke.
Our prior work found that the
Nhe1 gene plays a critical role in the fine-tuning of microglial profiles and that NHE1 protein inhibition promotes white matter remyelination after ischemic stroke or traumatic brain injury [
12,
13,
48]. Here we further investigated the mechanisms underlying these SAWM population transformation by deleting the
Nhe1 gene selectively in microglia. NHE1 protein mediates the electroneutral transport of H
+ efflux in exchange for Na
+ influx and is one of the major pH
i regulators in microglia [
49]. We identified changes to microglial acid–base homeostasis and transcriptomic profiles in the generic CD11b
+CD45
+ microglial population as well as the SAWM-specific CD11c
+ subpopulation in
Nhe1 cKO brains, including increased acidification and upregulated CREB signaling with elevated metabolism in these cells. Regarding how pH
i changes affect CREB signaling activation, studies characterizing the metabolic derangements in cancer cells have shown that decreased pH
i (as well as increased extracellular pH) promotes CREB signaling activation which in turn drives cellular metabolism across models of brain injury [
21,
50]. Our group has detected elevated oxidative phosphorylation energy metabolism in the
Nhe1 cKO microglia after stroke [
13]. Concurrently, we found significantly increased CREB1 activation/phosphorylation in the
Nhe1 cKO CD11b
+CD45
int cells compared to WT control. Moreover, a significantly higher proportion of p-CREB1-expressing CD11b
+CD45
int also increased BDNF expression, a growth factor proven to be essential for white matter integrity in human patients [
51,
52]. Although depletion of microglial BDNF and BDNF supplementation experiments generated different outcomes in different disease models, BDNF (and the upstream Akt/CREB signaling) appears to be a critical player in microglial activation and proliferation in response to stress/injury [
53‐
55]. Other possible mechanisms underlying CREB1 activation include reports that NHE1 protein activation can generate a periodic intracellular Ca
2+ increase (Ca
2+ oscillation), leading to Ca
2+/CaMKII-dependent CREB activation in brain pericytes [
56]. However, we previously reported that NHE1 blockade with either its potent inhibitor HOE642 [
49,
57] or global NHE1 knockout [
58] almost completely attenuated the Ca
2+ increase in cultured neurons, microglia, or astrocytes after in vitro ischemia, likely through blocking the reversed operation of Na
+/Ca
2+ exchanger (NCX
rev), which shall mitigate the Ca
2+/CaMKII-dependent CREB activation. Further investigation is needed to depict the specific mechanisms underlying the CREB1 activation in
Nhe1 cKO microglia/macrophages.
Recent studies have reported a specialized group of oligodendrocytes, termed “disease-associated oligodendrocyte” (DAO), that participate in phagocytosis and lactate shuttling functions which support white matter remyelination and axonal regeneration [
41,
42] in the context of neurodegenerative diseases, such as Alzheimer’s disease and multiple sclerosis [
38‐
40]. We found that a subgroup of oligodendrocytes (OL5) exhibited increased expressions of these DAO-associated genes such as
Clu, Cd59a, Ecrg4, Ldha, Ldhb, etc. [
38‐
40] in both stroke and non-stroke hemispheres. Most interestingly, pathway analysis by IPA for the OL5 subpopulation predicted activated CREB1 signaling to be the most significant upstream regulator, which corroborated our microglial findings. In summary, our data suggest that selective deletion of microglial
Nhe1 could exert its effects on white matter myelination by stimulating microglial CREB1 pathway activation and BDNF secretion, which accelerated restorative microglia-oligodendrocyte communications.
In our study, it is worth noting that the trends of CREB1 signaling activation persisted to be significantly increased in both CD11b
+CD45
int microglia and CD11b
+CD45
hi microglia/macrophage populations in the post-stroke cKO brains. The pHrodo experiments revealed statistically significant acidification in the combined CD11b
+CD45
+ population, as well as in the CD11c
+ microglial populations (p < 0.05) in the cKO brains (CL and IL hemispheres), compared to the WT brains. But the changes in pH
i did not reach significance when CD11b
+CD45
int and CD11b
+CD45
hi microglia/macrophages in WT and cKO brains were separately analyzed. This could be due to large variability in the data with relative low n values. Nevertheless, as no differences were detected in pH
i in either CD11b
+/CD45
+ or CD11c
+ populations between sham-operated WT or cKO brains (Additional file
1: Figure S4a), the increased pHrodo fluorescent intensity (indicating acidified pH
i) in both CL and IL hemisphere of cKO brains are stroke-dependent. This is consistent with previous report that the non-lesioned CL hemispheres showed robust defect in electrophysiology recordings in human stroke patients, which correlate with poorer outcomes [
59]. In comparison, less acidification in the CD11c
+ microglia was detected in the non-stroke CL hemisphere of
Nhe1 cKO brains, where stroke was able to induce further acidification (Fig.
4b, c). Thus, we cannot rule out that additional mechanisms may be involved in regulating pH in the CD11c
+ microglia. It is reported that Hv1 only mediated large proton currents in microglia, but not in neurons or astrocytes, indicating a relative selectivity of Hv1 expression in microglia [
60]. Global knockout of Hv1 reduced brain damages and conferred neuroprotection after ischemic stroke [
60], spinal cord injury [
61], or traumatic brain injury [
62] via similar mechanisms of preventing oxidative damage [
63]. While
Hvcn1 (encoding Hv1) gene expression in MG3 subpopulation is relatively low compared to other microglia/macrophages (MG1, 2, 4, and 5; 2.13-fold decrease, p < 0.0001), we found a small but significant elevation of
Hvcn1 gene expression in the MG3 of cKO brains post-stroke (1.32-fold increase, p < 0.01; Additional file
1: Figure S7), which shall alkalinize pH
i. Other potential pH
i regulating mechanism includes the Na
+/HCO3
− cotransporter (NBCe1), which facilitates HCO3
− outward transport (coupled with Na
+ efflux) to buffer extracellular H
+ loads in response to neuronal activity [
64], or stimulates inwardly directed HCO3
− transport (coupled with Na
+ influx) in ischemic stroke conditions [
65]. We detected similarly low levels of
Slc4a4 (encoding NBCe1) gene expression in MG3 compared to other microglia/macrophage subpopulations, but stroke triggered a small but significant increase of
Slc4a4 gene expression in the MG3 of cKO brains (1.36-fold increase, p < 0.01; Additional file
1: Figure S7). Whether the elevated gene expression translates to a higher NBCe1 protein expression, and whether the NBCe1 activity is in outward mode (in compensation of NHE1 blockade) or inward mode (triggered by stroke) remain unknown and warrant further exploration. Overall, our new findings support a novel mechanism of pH regulation in promoting restorative microglial transformation and growth factor release, shedding light on pH regulation as a novel therapeutic approach in stroke treatment.
There are several limitations in this study. First, our overview UMAP reveals the presence of neurons in the dissected tissues. During our tissue dissection process, white matter tracts (CC and EC) were visually identified and manually dissected from brain sections at 4 different bregma levels (Additional file
1). Therefore, there are inevitably some contamination of non-white matter tissues (cortex and/or hippocampal tissues) in the single cell suspension. We focused our analysis on the MG and OL populations, where neuronal contamination imposed minimum effects. Secondly, astrocytes are the most abundant glial cell type in the human brains, however the transcriptome changes of astrocytes and their roles in white matter damage/repair are not fully explored here. Astrocytes can contribute to white matter damage through either direct cell–cell communication with oligodendrocytes, or by indirectly affecting microgliosis through protein accumulation, unbalanced secretion of a variety of molecules, including extracellular matrix proteins, pro- and anti-inflammatory cytokines and chemokines [
66]. Astrocyte may also alter the gap junctional network and change ionic and nutrient homeostasis [
66]. Therefore, the astrocyte-oligodendrocyte interactions warrant further investigation in future studies. Lastly, despite the significant improvements in white matter myelination in the post-stroke
Nhe1 cKO brains [
12,
13], the only significantly altered subgroup of oligodendrocytes in the
Nhe1 cKO white matter tissues compared to the WT white matter, was OL5. While we saw several folds of expansion of the OL5 subcluster in the
Nhe1 cKO white matter (6.3-fold in the CL and 3.5-fold in the IL hemispheres), the cell numbers of the OL5 subgroup were relatively small, only consisting of less than 3% of total oligodendrocytes. However, this is consistent with other report showing that the distinct DAO population does not exceed 5% of all oligodendrocytes [
67], which is also consistently shared across multiple pathologies, including different models of Alzheimer’s disease (5xFAD amyloidosis model, P301L pure tauopathy model, and PS2/APP/P301L combined tauopathy/amyloidosis model), a multiple sclerosis model of experimental autoimmune encephalomyelitis, or following LPS-induced brain inflammation [
67], with signature genes involved in phagocytosis and lactate shuttling functions [
38‐
40]. Further in-depth studies are required to determine the roles and mechanisms of this small OL population in relation to disease progression/repair. In addition, the genetic profiles within each OL subgroups may not be homogenous between the WT and cKO groups, despite their similar cell numbers. Thus, further analyses on the other OL subgroups (OL1-4) are also warranted. Overall, our scRNAseq transcriptome study revealed new possible cellular mechanisms between microglia-oligodendrocyte interactions, which will be further validated using microglia-oligodendrocyte in vitro co-culture system in future studies.