Heterogeneity and Molecular Markers for CNS Glial Cells Revealed by Single-Cell Transcriptomics

Abstract

Glial cells, including astrocytes, oligodendrocytes, and microglia, are the major components in the central nervous system (CNS). Studies have revealed the heterogeneity of each glial cell type and that they each may play distinct roles in physiological processes and/or neurological diseases. Single-cell sequencing (scRNA-seq) technology developed in recent years has extended our understanding of glial cell heterogeneity from the perspective of transcriptome profiling. This review summarizes the marker genes of major glial cells in the CNS and reveals their heterogeneity in different species, CNS regions, developmental stages, and pathological states (Alzheimer’s disease and spinal cord injury), expanding our knowledge of glial cell heterogeneity on both molecular and functional levels.

Introduction

The central nervous system (CNS) is mainly composed of neurons and glial cells. Microglial, astrocytic, and oligodendrocyte (OL) lineage cells comprise the major glial cells in the CNS. With new features such as secreting and uptaking discovered, glial cells are now considered critical for neuronal metabolism, regeneration, and homeostasis maintenance of the CNS, instead of solely supporting connective tissues.

Each type of glial cell has diverse states or functions. Microglia are macrophages that reside in the CNS, with properties of both glial and immune cells. In response to environmental changes, microglia can rapidly switch from a “resting state” to a “reactive state,” secreting cytokines and engaging in phagocytosis. Multiple subtypes of microglia and their corresponding marker genes have been identified in different states (Hammond et al. 2019; Masuda et al. 2019). “Reactive” cells are the ones with the defining feature of morphologically protruding branches (Kreutzberg 1996). Notably, the “reactive state” of microglia is dynamic and the complete molecular signatures are currently lacking.

As the most abundant cells in the CNS, astrocytes, identified by stellate projections on their surfaces, have diverse function. They can regulate synaptic transmission by tightly enwrapping neurons and communicate with fast and point-precise neural circuits relying on Ca²⁺ waves (Allen et al. 2012; Christopherson et al. 2005). They can also protect neurons by metabolic coordination (Ioannou et al. 2019; Rusakov 2015). In addition, astrocytes are important in blood–brain barrier (BBB) formation (Abbott et al. 2010), neuroinflammation (Linnerbauer et al. 2020; Xing et al. 2019), memory encoding (Santello et al. 2019), and generation of circadian rhythms (Hastings et al. 2018). Astrocytes exist as multiple subtypes in different brain regions, which may be the basis of their functional diversity (Batiuk et al. 2020).

OL lineage cells are the only myelin-forming cells in the CNS. The OL lineage includes OL precursor cells (OPCs, also called NG2-glia), committed OL precursors (COPs), mature oligodendrocytes (MOLs), and other intermediate state subtypes. The developmental lineage is OPCs-COPs-MOLs. Notably, a significant pool of OPCs is retained in the adult CNS and can differentiate into myelinating MOLs (Bercury and Macklin 2015). Interestingly, OLs also provide metabolic support to neurons (Philips and Rothstein 2017). Aberrant interaction between OPCs and vessels can disrupt the BBB in multiple sclerosis (Niu et al. 2019). Multiple markers in OL lineage cells have been elucidated (Marques et al. 2016), but the specific role of each subpopulation remains elusive.

The function and maintenance of homeostasis by glial cells are currently hot topics, and several recent reviews have provided incisive summaries of the progress in this field. Greenhalgh et al. summarized the function of CNS glial cells as immune cells (Greenhalgh et al. 2020), Allen et al. reviewed the structure and function of CNS glial cells (Allen and Lyons 2018), Akay et al. described the fates and function of OPCs in the brain (Akay et al. 2021), Liddelow et al. highlighted the role of the communication of microglia and astrocytes in homeostasis and disease (Liddelow et al. 2020), Bohlen et al. discussed how microglia affect brain development, homeostasis, and their role in neurodegeneration (Bohlen et al. 2019a, b), Madore et al. discussed the effects of lifestyle and environmental factors on microglia (Madore et al. 2020), Linnerbauer et al. reviewed the multifaceted role of astrocytes in CNS inflammation (Linnerbauer et al. 2020), and Elbaz et al. discussed the mechanisms of transcriptional control of the developmental lineage of OLs during myelination (Elbaz and Popko 2019). This article will fill a gap in the review literature by focusing on summarizing the progress on marker genes and cellular heterogeneity in glial cells.

Advances in Glial Marker Research

Cell identification and isolation are a prerequisite for cell function studies. Traditional methods rely on antigen–antibody reactions, such as immunohistochemistry (IHC) and fluorescence-activated cell sorting (FACS). Transgenic animals generated for specific cell type labeling have significantly accelerated the process of glial cell research. For example, Zhang et al. purified various types of neurons, glia, and vascular cells by FACS from different transgenic mice and generated a bulk RNA-seq database for each cell type (Zhang et al. 2014). However, these methods are limited in that they only achieve relatively crude labeling of the cell populations and do not accurately reflect the heterogeneity of the cells.

Single-cell RNA sequencing (scRNA-seq) technology is a powerful approach to decipher cellular heterogeneity, which reveals genome-wide gene expression at single-cell resolution (Papalexi and Satija 2018). This technique was first developed in 2009 by Tang et al. and was employed on a large scale by 2014. The applications of this technique have increased exponentially (Svensson et al. 2018). Cell markers are the basis for cell type identification and single-cell analysis. In Table 1, we summarize the markers for major classes of glial cells.

Table 1.

Marker genes used to identify major classes of glial cells in scRNA-seq

Cell type	Marker genes	Ref.
Microglia	Cx3cr1, Trem2, P2ry12, Fcrls, C1qa, CD45, CD11b	Blum et al. (2021), Hammond et al. (2019), Keren-Shaul et al. (2017), Masuda et al. (2019), Rosenberg et al. (2018), van der Poel et al. (2019), Zeisel et al. (2018)
Astrocyte	Gfap, Aqp4, Aldh1l1, Slc1a3	Batiuk et al. (2020), Blum et al. (2021), Rosenberg et al. (2018), Xu et al. (2018), Zeisel et al. (2018)
OPC	Pdgfra, Cspg4, Ptprz1	Blum et al. (2021), Floriddia et al. (2020), Marques et al. (2016), Rosenberg et al. (2018), Xu et al. (2018), Zeisel et al. (2018)
OL	Mbp, Mobp, Olig2, Mag, Mog	Blum et al. (2021), Floriddia et al. (2020), Marques et al. (2016), Rosenberg et al. (2018), Xu et al. (2018), Zeisel et al. (2018)
OL lineage	Aspa, Sox10	Floriddia et al. (2020)

The table shows the marker genes most commonly used in glial cells. These genes are used to define cell types in single-cell sequencing studies. Representative references are listed in the last column

Microglia

Multiple markers of microglia have been identified in past studies. CD45 and CD11b (gene name Itgam) are the most common cell surface markers of microglia, and using these markers for FACS, microglia can easily be sorted from other glial cells (Garcia et al. 2014). In pathology studies, ionized calcium-binding adapter (induction of brown adipocytes 1, Iba-1) is the most widely used activated microglial cell marker. Subsequent studies have shown that microglia specifically express the cell surface marker transmembrane protein 119 (Tmem119) (Bennett et al. 2016). With the development of antibodies targeting exposed epitopes in the N-terminal Tmem119 protein, microglia have been successfully purified (Bennett et al. 2018; Bohlen et al. 2019a, b). Chemokine (C-X3-C motif) receptor 1 (Cx3cr1) is a specific receptor for the neural actin CX3C chemokine fractalkine, which serves as a checkpoint for microglia activation. Genetic deletion of Cx3cr1 has uncovered multiple functions of microglia in the CNS (Jung et al. 2000; Li et al. 2020; Parkhurst et al. 2013; Wu et al. 2020). Except Cx3cr1, recent scRNA-seq papers reveal that many genes, including Trem2, P2ry12, Fcrls, C1qa, CD45, and CD11b, can be used to recognize microglia (Blum et al. 2021; Hammond et al. 2019; Keren-Shaul et al. 2017; Masuda et al. 2019; Rosenberg et al. 2018; van der Poel et al. 2019; Zeisel et al. 2018). These marker genes were located in the plasma membrane or extracellular space, responsible to the rapid activation and secretion of microglia. Caution is needed when applying specific markers of microglia, because microglia and peripheral macrophages share a few markers (e.g., Fcrls, P2ry12, Tmem119, and Trem2) in early development and chronic neuroinflammation. (Grassivaro et al. 2020).

Astrocytes

Unlike microglia, astrocytes are isolated by the immune panning method, in which other glial cells are removed by specific antibodies, leaving behind the pure astrocytes (Foo 2013). Recently, an immunoaffinity-based method was developed for sorting ultrapure astrocytes based on the high affinity of the acetyl–CoA synthetase (Acsa-2) antibody to the astrocytic surface marker Atp1b2 (a Na + /K + transporting ATPase, beta 2 polypeptide) (Batiuk et al. 2017; Kantzer et al. 2017). Glial fibrillary acidic protein (Gfap), an intermediate filament protein, is the best known marker for astrocytes. It plays a role in almost all known functions of astrocytes, including proliferation, migration, myelination, nerve growth, and blood–brain barrier permeability (Middeldorp and Hol 2011). Genetic manipulations of Gfap and another astrocytic marker aldehyde dehydrogenase 1 family member L1 (Aldh1l1) have been widely used in transgenic mice for astrocyte labeling and functional studies (Messing et al. 1998; Srinivasan et al. 2016; Zhuo et al. 2001). Detailed descriptions of the experimental tools used to study astrocytes and microglia have been published (Guttenplan and Liddelow 2019; Yu et al. 2020).

Similar to Gfap and Aldh1l1, Aqp4 and Slc1a3 are also common astrocytic markers, which have been validated by both bulk RNA-seq and scRNA-seq (Batiuk et al. 2020; Blum et al. 2021; Rosenberg et al. 2018; Xu et al. 2018; Zeisel et al. 2018; Zhang et al. 2014). The aquaporin Aqp4 can coordinate water balance and osmotic pressure in the CNS (Vandebroek and Yasui 2020). Slc1a3 encodes a high affinity glutamate transporter protein, which can regulate excitatory synapse activity by neuronal glutamate uptake (Nagelhus and Ottersen 2013). However, few reports are available about their specific roles in astrocytes. Notably, astrocytes resemble satellite glial cells (SGC) in the peripheral nervous system, although expression of Kcnj10/Kir4 (potassium inwardly-rectifying channel, subfamily J, member 10) and Fabp7 (fatty acid binding protein 7) differs between these two cells (Avraham et al. 2020).

OLs

The immune panning method has been used to isolate OPCs [platelet-derived growth factor receptor alpha (Pdgfra) positive], myelinating MOLs [myelin oligodendrocyte glycoprotein (Mog) positive], and post-mitotic OLs [galactosylceramidase (GalC) positive] with cell surface markers bound to specific antibodies (Emery and Dugas 2013). By gene manipulation, transgenic animals targeting sex determining region Y box 10 (Sox10), oligodendrocyte transcription factor 2 (Olig2), and Pdgfra have been generated to specifically label all OL lineage cells, MOLs, and OPCs, respectively (Kang et al. 2010; Kessaris et al. 2006; Lu et al. 2016). Generally, myelin basic protein (Mbp), myelin-associated oligodendrocytic basic protein (Mobp), Olig2, myelin-associated glycoprotein (Mag), and Mog are used to identify MOLs, and Pdgfra, chondroitin sulfate proteoglycan 4 (Cspg4), and protein tyrosine phosphatase receptor type Z1 (Ptprz1) are used for OPCs (Blum et al. 2021; Floriddia et al. 2020; Marques et al. 2016; Rosenberg et al. 2018; Xu et al. 2018; Zeisel et al. 2018). COPs are OPC-like cells which express Neu4 and genes maintaining the undifferentiated state (e.g., Sox6, Bmp4, and Gpr17), but unlike OPCs, they lack Pdgfra and Cspg4 (Marques et al. 2016). In line with these, marker genes in MOL are critical for myelination, axon supporting, and signal conduction (Philips and Rothstein 2017); marker genes in OPCs act to maintain the proliferation characteristics of OPCs (Dang et al. 2019; Kucharova and Stallcup 2018; Fujikawa and Noda 2016). Notely, all Sox10-positive cells express aspartoacylase (Aspa) during development and thus Aspa and Sox10 are pan-markers of the OL lineages (Floriddia et al. 2020).

Species Heterogeneity of Glial Cells

Rodents are a common animal model for glial cell research, and despite the highly conserved nature of their genes and those of humans, the genetic consequences of subtle differences cannot be ignored. One study pointed out that heterogeneity among cohorts is widespread even for the same species and the same disease (Neff et al. 2021). Microglia have significant heterogeneity in different genetic backgrounds, including differences in the expression of several genes associated with Alzheimer’s disease (AD) (e.g., Apoe, Trem2, and Sorl1) and in the number of interferon-responding microglia (Yang et al. 2021). The idea of comparing multiple model organisms has been used in regenerative medicine research (Jorstad et al. 2017; Letelier et al. 2018; Wang et al. 2020). However, for mechanistic studies and drug screening, few, if any, have considered the heterogeneity of glial cells in different species or strains.

It is almost certain that differences in the subtypes and transcriptional profiles of human and murine glial cells exist (Han et al. 2013; Hodge et al. 2019; Song et al. 2020). Xu et al. compared the gene expression profiles of several common glial cells in mouse and human brain using scRNA-seq, noting that humans and mice share several common glial cell markers (see Table 1) but they are differentially expressed, and some non-marker genes are also differentially expressed between similar cells of different species (Table 2) (Xu et al. 2018). Another study found limited overlap in gene expression between mouse and human microglia during aging, with overlapping genes containing microglia marker genes, such as Cx3cr1, P2ry12, and CD11B. A number of immune-related receptor genes, including TLR, Fcγ, and SIGLEC, and proliferation and cell cycle regulators TAL1 and IFI16 were up-regulated in human but not mouse microglia (Galatro et al. 2017). In summary, humans and mice have transcriptomic differences, but their marker genes are identical.

Table 2.

Species-specific genes of glial cells identified by scRNA-seq

Cell type	Marker genes	Ref.
Microglia		No publish
Astrocyte	RGCC/human	Xu et al. (2018)
	Mybpc1/mouse
OPC	SCN9A/human	Xu et al. (2018)
	Lpcat2 1/mouse
OL	GLDN/human	Xu et al. (2018)
	Kcnk13/mouse

For the same type of glial cells, the genes in the table are only expressed in one species of human and mouse. The species-specific genes of microglia were not shown because there are no published studies

Temporal Heterogeneity of Glial Cells

The heterogeneity of CNS glial cells throughout development has been a major area of scientific inquiry. Recent scRNA-seq studies have done much to advance knowledge in this field. Here, we summarize the latest progress in understanding the heterogeneity of glial cells from developmental initiation through the embryonic (E), postnatal (P), juvenile, adult, and elderly stages (Table 3).

Table 3.

Time-specific gene expression characteristics of glial cells identified by scRNA-seq

Cell type	Marker genes and stages	Ref.
Microglia	Ms4a7, Ccr1, Ms4a6c/embryonic	Hammond et al. (2019), Masuda et al. (2019)
	Spp1, Gpnmb, Igf1/postnatal day 4 and 5
	Ctsb, Ctsd, Lamp1/embryonic
	CST3 −, SPARC − high percentage/embryonic
	Tmem119, Selplg, Slc2a5/postnatal
Astrocyte	No publish
OPC	Ptprz1 high percentage/postnatal day 20	Floriddia et al. (2020)
	Ptprz1 high percentage/postnatal day 60
OL	Ptgds high percentage/postnatal day 60	Floriddia et al. (2020)

“+” and “−” represent the positive and negative signal of immunofluorescence imaging, respectively; “high/low percentage” indicates that the proportion of cells with this characteristic in this type of glial cells is high or low compared with other times

Microglia

In vivo lineage tracing studies in the mouse model have confirmed that microglia originate from primitive myeloid progenitor cells that appear before embryonic day 8 and later transfer to central maturation (Ginhoux et al. 2010). Unlike other glial cells, microglia have no other precursor cells throughout their lifespan (Gomez Perdiguero et al. 2015; Grassivaro et al. 2020; Kierdorf et al. 2013). Li et al. found high microglial cell heterogeneity in the early postnatal period, compared with little heterogeneity in adult homeostasis-state microglia (Li et al. 2019). Hammond et al. used scRNA-seq to analyze the transcriptional profiles of microglia throughout the lifespan of the mouse brain (E14.5, P4/5, P30, P100, P540) and developed the gene expression profile data for each stage into a visual website (Hammond et al. 2019). In this study the authors found at least nine distinct subpopulations of microglia. In the embryonic period, microglia are characterized by Ms4a7-specific expression; in P4/5, microglia specifically express Spp1; in the early postnatal brain, microglia are characterized as metabolically active and proliferative due to high expression of Birc5, Hist1h2bc, Mcm6, Rrm2, and other genes; and finally, microglial aging is characterized by increased inflammatory and interferon responses and an increased proportion of Ccl4-positive cells (Hammond et al. 2019). Another study reported that high expression of Tmem119, Selplg, and Slc2a5 is a feature of juvenile and adult but not embryonic microglia; Ctsb expression is a feature of microglia in the embryonic period; and lack of Cst3 and Sparc expression is also a feature of microglia in the embryonic period (Masuda et al. 2019). One reason why the development-specific markers of microglia found in the above two studies are not consistent is that the mice used were not exactly the same age, which suggest that the study of microglia developmental heterogeneity may require a more detailed time window.

Astrocytes

Two pathways can renew the mature astrocyte pool; one is the differentiation of radial glial cells acting as precursor cells and the other is the self-renewal of mature astrocytes through their own division and differentiation (Chen et al. 2020; Ge et al. 2012; Morel et al. 2017; Noctor et al. 2004). There has been no clear check point for the developmental initiation and maturation process of astrocytes (Molofsky and Deneen 2015; Molofsky et al. 2012). In the developing spinal cord, as early as E12.5, astrocytes are demarcated by induction of Slc1a3 (GLAST, glutamate/aspartate transporter), Aldh1l1, and Nfia (Cahoy et al. 2008; Deneen et al. 2006). Transcriptome sequencing of astrocytes located in the mouse spinal cord at E13.5–18.5 combined with biochemical experiments revealed that Sox9 promotes astrocyte and OL maturation by regulating Nfe2l1 (Molofsky et al. 2013). Transcriptome sequencing of samples at multiple developmental time points from E12.5 to P7 demonstrated that Asef, Tom1l1, Mfge8, and Gpr37l1 can serve as novel markers for intermediate stages of spinal astrocyte development (Chaboub et al. 2016). Boisvert et al. analyzed gene expression profiles of astrocytes in 4-month-old and 24-month-old mice and found that aging astrocytes exhibited significant molecular differences with increased expression of inflammatory and synapse elimination pathway genes (Boisvert et al. 2018). The above findings were obtained from bulk RNA sequencing in astrocytes isolated by FACS. No single-cell resolution study that accurately reflects the heterogeneity of astrocytes in a temporal time scale is available, which may be an important future direction to explore.

OLs

OL development involves several well-defined steps, starting with the differentiation of neural progenitor cells into proliferating OPCs, moving on to differentiation into COPs, and finally developing into MOLs with myelinogenic functions (Richardson et al. 2006). In the brain, the origin of OPCs is restricted to 3 waves in mice, occurring at E11.5–12.5, E16.5, and after birth. Spatially, these OPCs develop in the ventricular zone of the medial ganglionic eminence and anterior entopeduncular area, the Gsh2-expressing ventricular zone of the lateral and central ganglionic eminences, and the cerebral cortex, respectively (Kessaris et al. 2006). In the spinal cord, OPC formation initiates in the ventrally localized motor neuron domain (pMN domain) at E12.5 and in the dorsal precursor region of the spinal cord at E15.5, but by birth the ventral region rather than the dorsal is the primary source of OPCs (Fogarty et al. 2005; Ravanelli and Appel 2015; Richardson et al. 2000). OPCs first appear in humans at approximately 10–15 weeks of gestation (Jakovcevski et al. 2009). OL maturation and myelin formation in mice begin at E18.5 and continue into adulthood (Akay et al. 2021; Michalski and Kothary 2015). The mechanisms of developmental initiation and myelin formation in OPCs have been summarized in detail (Akay et al. 2021; van Tilborg et al. 2018).

Temporal heterogeneity of OPCs and OLs were investigated by multiple studies. Branco’s group sequenced single cells from Pdgfra-positive populations in adolescent and adult mice and found that OPCs that developed at different times are not transcriptionally heterogeneous, but they can differentiate into MOL populations with temporal and spatial specificity (Marques et al. 2016). Similarly, results from study did not report significant transcriptional program differences in OPCs isolated from embryonic and postnatal periods (Marques et al. 2018). However, Spitzer et al. studied the function of mouse OPCs at several time points from E13 to 9 months and found that although OPCs originated from homogeneous populations, they were spatially and temporally different in electrophysiological properties (Spitzer et al. 2019).

Spatial Heterogeneity of Glial Cells

Although glial cells exist throughout the entire CNS, differential developmental origins can lead to regionalization of glial cell populations. Anatomically, numerous neurological diseases have foci that occur only in specific brain regions, implying that the spatial heterogeneity of glial cells is closely related to disease progression. This section provides a summary of recent advances in glial cell spatial specificity. We summarized the progress of spatial heterogeneity of glial cells identified by scRNA-seq (Table 4).

Table 4.

Spatial-specific gene expression characteristics of glial cells identified by scRNA-seq

Cell type	Marker genes	Ref.
Microglia	CST3 + SPARC + high percentage/cortex	Masuda et al. (2019)
	CST3 + SPARC + low percentage/cerebellum
	CST3 + SPARC − low percentage/cortex
	CST3 + SPARC − high percentage/cerebellum
Astrocyte	Frzb, Ascl1/cortex	Batiuk et al. (2020)
	Unc13c/hippocampus
OPC	No publish
OL	Klk6 high percentage/dorsal spinal cord	Floriddia et al. (2020)
	Klk6 low percentage/cortex and corpus callosum

“+” and “−” represent the positive and negative signal of immunofluorescence imaging, respectively; “high/low percentage” indicates that the proportion of cells with this characteristic in this type of glial cells is high or low compared with other times

Microglia

Grabert et al. proposed that microglia in different brain regions (striatum, hippocampus, cerebral cortex, and cerebellum) of adult mice have distinct transcriptomic profiles in terms of immune response and energy metabolism as determined by microarray analysis (Grabert et al. 2016). Ayata et al. demonstrated that microglia in the cerebellum have stronger scavenging activity compared to the cortex and striatum, due to polycomb repressive complex 2 (PRC2), which limits the expression of genes with scavenging activity in the striatum through an epigenetic mechanism (Ayata et al. 2018). ScRNA-seq transcriptome of microglia in eight anatomical regions at three developmental time points revealed strong variation between different regions of the CNS in microglia (Masuda et al. 2019). However, it has also been argued that the spatial heterogeneity of microglia is not absolute. One study used scRNA-seq to analyze the transcriptome of microglia from different brain regions at multiple developmental time points. The authors found that the transcriptomes of adults, but not embryonic and postnatal microglia, are very similar regardless of the brain region (Li et al. 2019). Another study performed scRNA-seq on CD45-positive cells isolated from several CNS compartments (including the leptomeninges, the perivascular space and parenchyma, and the choroid plexus) and found that the disease-associated subpopulation had stronger heterogeneity in microglia than the homeostatic subpopulation. Although the major marker genes were expressed population-wide, the disease-associated microglia clusters showed lower expression of P2ry12, Maf, and Slc2a5 and higher expression of Ccl2, Cxcl10, Ly86, and Mki67; in inflammatory diseases and spinal cord injury, expression of P2ry12, Tmem119, and Selplg was strongly down-regulated and Ly86 expression was up-regulated (Jordao et al. 2019).

Astrocytes

Previous findings provide evidence that astrocytes exhibit functional and molecular heterogeneity in a region-dependent manner (Bayraktar et al. 2014; Zhang and Barres 2010). Subsequent studies have also analyzed in detail the region-specific nature of astrocytes at the transcriptional, translational, and epigenetic levels as well as by novel FACS methods (Chai et al. 2017; Herrero-Navarro et al. 2021; John Lin et al. 2017; Morel et al. 2017; Sharma et al. 2015). The rehabilitative effect of certain activities, such as exercise, on astrocytes also occurs in a region-dependent manner (Lundquist et al. 2019). Not only that, several studies imply that regional specificity leads to functional limitations of astrocytes, e.g., astrocytes from different regions cannot compensate for each other’s functions (Tsai et al. 2012), and astrocytes from different regions have varying abilities to differentiate into neurons (Mattugini et al. 2019). Advances have been made in identifying spatially specific subpopulations of astrocytes using scRNA-seq. Batiuk et al. analyzed the astrocyte single-cell transcriptome in the hippocampus and cortex of adult mice and found significant heterogeneity in astrocytes in both regions, with the hippocampus but not the cortex containing subpopulations associated with neural regeneration; more than 70% of the enriched genes were subtype specific (Batiuk et al. 2020). Moreover, Bayraktar et al. performed a novel high-resolution spatial transcriptome analysis and found differential laminar distribution and varying transcriptome characteristics of astrocytes in the mouse cortex (Bayraktar et al. 2020).

OLs

Results from recent studies have shown significant molecular heterogeneity and spatial preference in MOLs (Floriddia et al. 2020; Marques et al. 2016). However, it is controversial whether or not OPCs originating from different regions are heterogeneous. Using scRNA-seq, Floriddia et al. found that OPCs from different regions show similar transcriptomic features during embryonic development (Floriddia et al. 2020). However, inconsistent results have been observed: cells of the OL lineages generated by separate OPC pools during development have varying responsiveness to demyelination in vivo and show heterogeneous migration and differentiation abilities in vitro (Crawford et al. 2016). The most recent view is that the functional heterogeneity of MOLs is independent of OPCs and may depend on the spatial environment or neuron type in which they are located and this may partially explain the controversy (Zonouzi et al. 2019). The results of Floriddia et al. also supported the idea that the intrinsic controls of OPCs do not determine the heterogeneity of MOLs (Floriddia et al. 2020). Sun et al. proposed a molecular mechanism that could explain the region- and time-specific myelin formation in the CNS, i.e., cell-autonomous programmed death of MOL subpopulations mediated by the transcription factor EB (TFEB) (Sun et al. 2018). However, the results of Floriddia et al. indicated that certain MOL subpopulations in the spinal cord are not produced by TFEB regulation (Floriddia et al. 2020). Therefore, more evidence is needed to resolve these controversies.

Heterogeneity of Glial Cells in the Pathogenesis of AD and Spinal Cord Injury (SCI)

CNS lesions are accompanied by changes in the number and gene expression of glial cells (Greenhalgh et al. 2020). Recently, advances have been made in understanding the molecular and cellular complexity of glial cells in diseases. Using AD and SCI as a representative neurodegenerative diseases and CNS mechanical injury, respectively, we summarize the progress of research in glial cell heterogeneity in disease.

AD

AD is the most common neurodegenerative disease. Extensive work has been done on the pathogenesis, clinical phenotypes, research methods, biomarkers, etc. of the disease (Lashley et al. 2018; Mantzavinos and Alexiou 2017; Penney et al. 2020; Scheltens et al. 2016). However, it is unclear on the molecular level how individual glial cell types are associated with AD. Four representative studies that use single-cell nuclear sequencing (snRNA-seq) to analyze the transcriptomes in different regions of the cerebral cortex of postmortem AD patients have not only provided new insights into cellular heterogeneity in the AD process but more importantly have generated a valuable data resource (Del-Aguila et al. 2019; Grubman et al. 2019; Mathys et al. 2019; Zhou et al. 2020). Pathology-associated glial cell types and genes were identified by scRNA-seq in AD patients: the AD-pathology-associated OL lineage cells characterized by high expression of CRYAB or QDPR, GLUL, and CLU are preferentially expressed in AD-pathology-associated astrocyte subtype (Mathys et al. 2019). These genes have the potential to become markers of AD pathology-associated glial cell subtypes. It is noteworthy that the heterogeneity and function of disease-related reactive astrocytes have also been implicated in a variety of neurodegenerative diseases (Escartin et al. 2021). In addition to the discussion of the abundance and differentially expressed genes of glial cells in AD patients, there are several interesting findings from these studies. Zhou et al. noted that changes in the expression of disease-related genes of glial cells in AD mouse models do not match these in AD patients (Zhou et al. 2020), Grubman et al. reported that the AD risk gene Apoe was specifically repressed in OPCs and astrocytes and up-regulated in disease-specific microglia in AD patients (Grubman et al. 2019), and Mathys et al. found significant gender differences in AD-associated transcriptomic alterations, with a higher proportion of AD-associated glial cell subtypes in female patients (Mathys et al. 2019). It has been shown that p2y12 is a marker of homeostatic microglia (Mildner et al. 2017); however, Walker et al. observed high expression levels of p2y12 in brain sections from patients with severe AD (activated microglia), emphasizing the heterogeneity of microglia in AD patients (Walker et al. 2020). Two studies focused on microglial heterogeneity in two separate AD mouse models using scRNA-seq and identified new disease-associated microglia subtypes as well as marker genes and reprogramming trajectories associated with these subtypes (Keren-Shaul et al. 2017; Mathys et al. 2017). Although both astrocytes and OL lineages have been reported to play a role in the AD process (Carter et al., 2019; Zhang et al. 2020), there are no studies on these two cell types in mouse models of AD with single-cell resolution, which would likely be a fruitful avenue of future research. We summarized the Alzheimer’s disease-specific differentially expressed genes of glial cells identified by scRNA-seq (Table 5).

Table 5.

Alzheimer’s disease-specific differentially expressed genes (DEGs) of glial cells identified by scRNA-seq

Cell type	Marker genes	Ref.
Microglia	RPS19, CD163/up-regulated	Grubman et al. (2019)
	CX3CR1, CD86, P2RY12/down-regulated
Astrocyte	MAP1B, S100A6/up-regulated	Grubman et al. (2019)
	APOE, NRXN1, GABRB1, GRIA2/down-regulated
OPC	APOE/down-regulated	Grubman et al. (2019)
OL	IL1RAPL1/up-regulated	Grubman et al. (2019)
All types	MT-ND1, CRYAB, HSPA1A/up-regulated	Grubman et al. (2019)

The DEGs in this table come from a single-cell sequencing study using AD patients as samples. “All types” means that these genes have consistent expression characteristics in all types of glial cells

In significant recent research, Jiang et al. integrated AD-related scRNA-seq databases and developed the scREAD dataset. scREAD provides a comprehensive analysis of 73 databases from 10 brain regions, including the construction of control maps, cell type prediction, identification of differentially expressed genes, and identification of cell type-specific regulators (Jiang et al. 2020). In addition, several new algorithms were developed to construct complex molecular networks of AD-related glial cells, which could help reveal new therapeutic targets for AD (Wang et al. 2016; Xu et al. 2021).

SCI

The intrinsic growth capacity of neurons and the injury microenvironment regulated by glial cells are two major determinants of the process of neural circuit reconstruction after SCI (He and Jin 2016). Spinal cord regeneration is extremely difficult in mammals, and glial scars formed by reactive astrocytes, fibroblasts, and microglia/macrophages are thought to be important in limiting spinal cord regeneration (Tran et al. 2018; Yang et al. 2020). However, many new findings have challenged this popular opinion. In the early stages of SCI, rapidly proliferating and activated microglia promote the formation of glial scars that are beneficial for repair, which helps to stop the spread of the lesion. This was confirmed by the results of subsequent experiments using transgenic mice or specific drugs depleting microglia, resulting in inhibition of spinal cord regeneration (Bellver-Landete et al. 2019; Fu et al. 2020). Li et al. used scRNA-seq and bioinformatic analysis to identify a subtype of microglia suitable for promoting scarless wound healing and suggested that fibronectin derived from this population of microglia may act as a bridge at the injury to promote wound healing (Li et al. 2020). Astrocytes are the main component of glial scars, and it is controversial whether they form a “wall” or a “bridge” for axonal regeneration after SCI. Astrocytes can produce chondroitin sulfate proteoglycans (CSPGs) that inhibit axonal growth (Silver and Miller 2004). With a rigorous experimental design, Anderson et al. demonstrated that scar-forming astrocytes allow, but do not prevent, CNS axon regeneration with appropriate stimulation (Anderson et al. 2016). There is widespread heterogeneity in astrocytes, but no study has yet indicated which subtypes have regeneration-inhibiting properties.

SCI leads to demyelination of nerve axons, and subsequent remodeling of spinal cord circuits requires remyelination. Information on how OL lineage cells change after SCI, as well as their function and ultimate fate has been summarized previously (Duncan et al. 2020). Spectral tracking experiments after SCI showed that OPCs are the main source of myelinogenic OLs (Assinck et al. 2017; Bartus et al. 2019) and that OPCs can also differentiate into astrocytes to facilitate glial scar formation (Hackett et al. 2018; Sellers et al. 2009). However, these results are insufficient to confirm whether there is heterogeneity in (1) the OPCs after injury and (2) between the original resident OLs and the newly formed OLs. Floriddia et al. used scRNA-seq to identify the effects of SCI on the transcriptome of MOLs and found that different subtypes of MOLs from healthy mice responded differently to SCI, but there was no evidence to indicate which MOL subtype promoted spinal cord regeneration (Floriddia et al. 2020).

In summary, glial cells dynamically change after SCI and contain complex lineages (Barnabe-Heider et al. 2010). Little is known about the heterogeneity of glial cells after SCI, and strategies to activate beneficial subtypes and inhibit deleterious ones might be utilized in the future to meet the challenges of mammalian spinal cord regeneration.

Conclusions and Perspectives

Recently, especially since the broad application of scRNA-seq technology, our knowledge of the molecular characteristics and spatiotemporal heterogeneity of glial cells has been greatly expanded. This will accelerate the breakthroughs in drug development targeting glial cells and BBB permeability and will lay a solid foundation for precision therapies (Yao et al. 2020). The enormous amount of data that can be generated by single-cell sequencing is a treasure trove to be mined and only a tiny fraction of it has been utilized so far. Here, we present a summary of representative scRNA-seq online data retrieval sites related to glial cells (Table 6), which is useful for understanding of the molecular basis and the biology of many glial cell-related diseases.

Table 6.

Online retrieval sites of visualized single-cell transcriptional atlas

Website	Cell type	Species	Ref.	Keywords
www.microgliasinglecell.com	Microglia	Mouse	Hammond et al. (2019)	Time specific; brain injury
www.brainrnaseq.org	Microglia	Mouse	Li et al. (2019)	Time and region specific
https://holt-sc.glialab.org/	Astrocyte	Mouse	Batiuk et al. (2020)	Region specific
https://ki.se/en/mbb/oligointernode	MOL	Mouse	Floriddia et al. (2020)	Region specific; spinal cord injury
http://linnarssonlab.org/oligodendrocytes/	OL	Mouse	Marques et al. (2016)	Time and region specific
http://spinalcordatlas.org	Neuron	Mouse	Blum et al. (2021)	Spinal cord
http://mousebrain.org	All CNS cell	Mouse	Zeisel et al. (2018)	Region specific
http://ngi.pub/snuclRNA-seq	All CNS cell	Human	Del-Aguila et al. (2019)	Alzheimer’s disease
http://adsn.ddnetbio.com	All CNS cell	Human	Grubman et al. (2019)	Alzheimer’s disease
https://bmbls.bmi.osumc.edu/scread/	All CNS cell	Human and mouse	Jiang et al. (2020)	An integrated database for Alzheimer’s disease

This table summarizes the online database of single-cell sequencing related to CNS glial cells. The information in the table includes website address, cell type, species, references, and database topics

There are complex interactions between glial cells and between glial cells and neurons that together maintain CNS homeostasis (Konishi et al. 2020; Patel et al. 2019). The massive amount of single-cell sequencing data has led to increased research efforts in bioinformatics, and a variety of cell–cell communication analysis methods based on ligand and receptor interaction databases have been proposed (Armingol et al. 2021; Jin et al. 2021; Noel et al. 2021). Notably, Clark et al. developed a viral vector-based scRNA-seq technology, RABID-seq, which enables effective molecular target identification of cell-to-cell interactions and will enrich the database of ligand–receptor interactions (Clark et al. 2021).

At present, the mechanisms regulating the generation, maintenance, and transformation of glial cells in temporal, spatial, and species specificity are not well understood. Some studies have been undertaken to examine these (Huang et al. 2020; Lee et al. 2008; Sun et al. 2018), but none of them are at single-cell levels. Single-cell ATAC-seq (assay for transposase-accessible chromatin with high-throughput sequencing) can obtain the chromatin accessibility profiles of individual cells and can resolve the transcriptional data of key genes in cells (Buenrostro et al. 2015; Cusanovich et al. 2018). This technology has the power to yield new breakthroughs in understanding glial cell heterogeneity in the near future.

Author Contributions

SJ, XL, and WL provided ideas and wrote the paper. SJ, SY, CZ, QJ, and ZS were responsible for the retrieval of papers. All authors read the manuscript.

Declarations

Conflict of interest

The authors have read this review and state that there was no conflict of interest in the publication of it.

Consent for Publication

Written informed consent for publication was obtained from all participants.