Properties of human astrocytes and NG2 glia

Since animal models are inevitable for medical research, information on species differences in glial cell properties is critical for successful translational research. Here, we review current knowledge about morphological and functional properties of human astrocytes and NG2 glial cells and compare these data with those obtained for the comparable cells in rodents. Morphological analyses of astrocytes in the neocortex of rodents versus humans have demonstrated clear differences. In contrast, the functional properties of astrocytes or NG2 glial cells in these species are surprisingly similar. However, these findings should be interpreted with caution, as so far functional analyses of human cells are only available from neocortex and hippocampus, and it is known from rodent studies that the properties of astrocytes in different brain regions may vary considerably. Moreover, technical challenges render astrocyte electrophysiological measurements in situ unreliable, and human cell properties may be affected by medications. Nevertheless, based on the limited data currently available, there is substantial similarity between human and rodent astrocytes with regard to those functional properties studied to date. The unique morphological characteristics of astrocytes in human neocortex call for further physiological analysis. The basic properties for NG2 glia are even less completely evaluated with regard to the question of species differences but no glaring differences have been reported so far. In conclusion, it remains justifiable to employ mouse or rat models to investigate the etiology of human CNS diseases that might involve astrocytes or NG2 glia.

the functional properties of astrocytes or NG2 glial cells in these species are surprisingly similar. However, these findings should be interpreted with caution, as so far functional analyses of human cells are only available from neocortex and hippocampus, and it is known from rodent studies that the properties of astrocytes in different brain regions may vary considerably. Moreover, technical challenges render astrocyte electrophysiological measurements in situ unreliable, and human cell properties may be affected by medications. Nevertheless, based on the limited data currently available, there is substantial similarity between human and rodent astrocytes with regard to those functional properties studied to date. The unique morphological characteristics of astrocytes in human neocortex call for further physiological analysis. The basic properties for NG2 glia are even less completely evaluated with regard to the question of species differences but no glaring differences have been reported so far. In conclusion, it remains justifiable to employ mouse or rat models to investigate the etiology of human CNS diseases that might involve astrocytes or NG2 glia.

K E Y W O R D S
astrocyte, electrophysiology, human, morphology, mouse, NG2 glia, rodent, species differences 1 | INTRODUCTION Glial cells were first described by Rudolf Virchow in middle of the 18th century as "Nervenkitt" (nerve putty/cement/glue), because he assigned them merely structural functions (Virchow, 1856;Virchow, 1859). The term "astrocyte" was proposed almost 50 years later by Michael von Lenhossék and refers to the star-shaped morphology of the cells (Lenhossék, 1893). Today, we know that astrocytes comprise a morphologically and functionally highly heterogeneous class of cells, which makes their identification difficult (Griemsmann et al., 2015;Matyash & Kettenmann, 2010;. Common characteristics include the absence of electrical excitability, a high resting K + conductance and therefore a membrane potential close to the K + equilibrium potential, extensive connections to each other via gap junctions, expression of functional glutamate and GABA transporters, formation of numerous fine processes enwrapping synapses and blood vessels, and expression of intermediate filament proteins such as GFAP and vimentin (Verkhratsky & Parpura, 2015). Research over the past decades revealed that astrocytes are way more than "nerve glue" as they fulfill a plethora of vital physiological functions and, consequently, are also key players in neurological diseases (Parpura et al., 2012;Seifert, Schilling, & Steinhäuser, 2006;Verkhratsky & Parpura, 2015). Their functions include supply of nutrients to neurons, control of extracellular ion homeostasis, clearance of neurotransmitters, regulation of the bloodbrain barrier (BBB) permeability, promotion of synapse formation, and contribution to the immune response by release of proinflammatory cytokines or neurotrophic factors (Verkhratsky & Parpura, 2015). Importantly, astrocytes also directly modulate synaptic transmission by release, uptake, degradation, and recycling of transmitters (Araque et al., 2014;Verkhratsky & Nedergaard, 2018). To accomplish this, the fine processes of individual astrocytes may ensheath >100.000 synapses (Bushong, Martone, Jones, & Ellisman, 2002). The dynamic regulation of the presynapse and postsynapse by closely associated glial processes led to the concept of the "tripartite synapse" (Araque, Parpura, Sanzgiri, & Haydon, 1999). According to this concept, astrocytes not only detect and react to neuronal activity but also respond to and regulate neuronal activity and plasticity through the release of neuroactive substances (so-called gliotransmitters), such as glutamate, ATP, and D-serine (Bezzi et al., 2004;Bindocci et al., 2017;Halassa, Fellin, & Haydon, 2007;Henneberger, Papouin, Oliet, & Rusakov, 2010). The in vivo relevance of this process, termed gliotransmission, is, however, still matter of debate (Fiacco & McCarthy, 2018;Savtchouk & Volterra, 2018). A striking feature of astrocytes is that they are electrically and metabolically coupled to each other through connexin channels, leading to the formation of large syncytium-like functional networks. Due to this network organization, they are in a position to effectively control and synchronize large neuronal assemblies (Giaume, Koulakoff, Roux, Holcman, & Rouach, 2010).
2 | MORPHOLOGICAL PROPERTIES OF HUMAN ASTROCYTES AND NG2 GLIA

| Astrocytes
Morphological studies revealed that astrocytes in the human cortex are larger, more complex and more diverse than those in rodents (Oberheim et al., 2009;Oberheim, Wang, Goldman, & Nedergaard, 2006;Vasile, Dossi, & Rouach, 2017). In addition to fibrous astrocytes in white matter and protoplasmic astrocytes in grey matter, two morphologically distinct subtypes of GFAP-positive cells, interlaminar and varicose projection astrocytes, are exclusively found in the cortex of higher primates (Colombo & Reisin, 2004;Colombo, Yáñez, Puissant, & Lipina, 1995;Oberheim et al., 2009;Sosunov et al., 2014). Human interlaminar astrocytes are solely found in the cortical layer I. They have small spheroid cell bodies with numerous short processes and one or two very long, tortuous processes that terminate in the neuropil or vasculature of cortical layers II to IV (Oberheim et al., 2009). The function of these cells is yet undefined, although it has been speculated that they are involved in information transfer between cortical layers (Oberheim et al., 2009). Varicose projection astrocytes, the second primate-specific subtype of astrocytes, are rather sparsely found in cortical layers V-VI. They typically exhibit several relatively short spiny processes and 1-to 5-mm-long unbranched processes with regularly spaced varicosities that extend within the deep cortical layers.
These cells were hypothesized to mediate long-distance communication across cortical layers or even between gray and white matter (Oberheim et al., 2009). Human cortical GFAP-positive protoplasmic astrocytes were reported to have 10 times more (37.5 vs. 3.8) and 2.6-fold longer (97.9 vs. 37.2 μm) primary processes than their rodent counterparts. Consequently, their domain diameter is about 2.6 times larger (142.6 vs. 56 μm), which corresponds to a 16.5-fold greater occupied volume. In contrast, the cell body of cortical astrocytes seems to be of comparable size in rodents and man (Oberheim et al., 2009;Oberheim Bush & Nedergaard, 2017), although we could not find where this data was originally published. Due to the larger diameter and the associated enhanced complexity of secondary process arborisation, human astrocytes were estimated to contact up to 2 million synapses, compared with up to 140.000 in rodents (Bushong et al., 2002;Oberheim et al., 2009). Like in rodents, cortical human astrocytes are organized in domains with little overlap, although the average area of overlap is greater in humans than in rodents (204.7 vs. 118 μm 2 ) (Oberheim et al., 2008;Oberheim et al., 2009). Human fibrous astrocytes in the white matter were also reported to be larger in the human versus mouse brain (183.2 vs. 85.6 μm) (Oberheim et al., 2009).
No morphometric analyses of astrocytes have yet been performed in human hippocampus. Thus, we have further evaluated GFAP/ S100β double staining performed previously in nonsclerotic hippocampal specimens from temporal lobe epilepsy (TLE) patients  and compared the data with those obtained from mouse hippocampus (Degen et al., 2012). The results revealed that soma volume and derived diameter of human hippocampal astrocytes are larger than those of mice ( Figure 1).

| NG2 glia
The morphological properties of human NG2 glia are less well characterized. In rodents, gray matter NG2 glia display a stellate appearance with a central round soma from which they extend several radial processes that are, however, much less branched than astrocytic processes (Jabs et al., 2005;Nishiyama, Yang, & Butt, 2005). In white matter the somata of NG2 glial cells are more elongated and the processes orientated along axonal projections (Chittajallu, Aguirre, & Gallo, 2004). In addition to the proteoglycan NG2, in rodents they coexpress the platelet-derived growth factor-α (PDGFα) receptor (Nishiyama, Lin, Giese, Heldin, & Stallcup, 1996).
Human NG2-positive glial cells were first described in tissue sections from adult neocortex and white matter by Chang, Nishiyama, Peterson, Prineas, and Trapp (2000) and Dawson, Levine, and Reynolds (2000). The former study reported that morphology and distribution of NG2 glia are similar in the human and rodent brain. The latter study, however, indicated that human NG2 glia have fewer processes than their rodent counterparts, an observation that apparently has not been quantified or confirmed later on (Dawson et al., 2000). Like in rodents, human NG2 glia consistently coexpress NG2 and PDGFα receptors but not GFAP Wilson, Scolding, & Raine, 2006). In contrast, the Ca 2+ binding protein S100β was expressed in almost all NG2 glial cells of human hippocampus , but only in a subset of mouse hippocampal NG2 glia (Jabs et al., 2005;Karram et al., 2008;Moshrefi-Ravasdjani et al., 2017). Interestingly, in the human nonsclerotic hippocampus the density of NG2 glial cells was higher than in the hippocampus of mice (~11 × 10 3 vs.~1.5 × 10 3 cells/mm 3 ). Moreover, the density of NG2 glia in the human tissue was similar to that of astrocytes, while mice display an NG2 glia/astrocyte ratio of about 1:4 in the same brain region Degen et al., 2012). It is, however, F I G U R E 1 Comparison of astrocyte and NG2 glia soma size in the human and mouse hippocampus. (a) The volume of human astrocyte soma was determined from three-dimensional GFAP/S100β double staining. Volume of human hippocampal NG2 glia was analyzed in stacks of PDGFRα staining. Specimens were obtained from two patients with lesion-associated TLE. The data of murine glia were determined in the hippocampus of mice with fluorescently doublelabeled cell types (Cx43kiECFP/NG2kiEYFP mice; Degen et al., 2012). Three-dimensional isosurface reconstructions were performed with IMARIS 8 software (Bitplane, Zürich, Switzerland) and used for calculation of soma volumes. (b) Assuming isovolumetric spheres, diameters were calculated for both cell types and species. Soma volume and diameter of astrocytes and NG2 cells were significantly larger in human hippocampus (hAstro: median, 648 μm 3 and 10.74 μm, n = 62; hNG2 cells: median, 891 μm 3 and 11.94 μm, n = 22) than in mice (mAstro: median, 319 μm 3 and 8.5 μm, n = 534; mNG2 cells: median, 449 μm 3 and 9.5 μm, n = 132; two-factor analysis of variance, p < .001, p = .91 for interaction; followed by Tukey's test, p < .001). In mice, soma volume and diameter of NG2 glia were larger than the corresponding parameters of astrocytes (p < .001), while in humans, the differences failed to reach statistical significance (p = .058). Data are presented as boxplots (box: first and third quantile; line: median; whiskers: 1.5 IQR, notches 95% confidence interval of the median; and circles: outliers). Note: nonoverlapping notches indicate significant difference possible that NG2 glial cell numbers in human tissue were influenced by seizures the patients experienced and, therefore, not representative for the situation in the healthy human brain. Indeed, seizureinduced proliferation of NG2 glial cells has been demonstrated in an experimental seizure model (Wennström, Hellsten, Ekdahl, & Tingström, 2003). As no information on the cell size of human NG2 glia is yet available, we further analyzed previously performed PDGFRα staining in the nonsclerotic human hippocampus  and compared the data with fluorescently labeled murine NG2 glia of the same region (CA1 stratum radiatum; Degen et al., 2012). Similar to astrocytes, the somata of human NG2 glial cells were larger as those of their rodent counterparts (Figure 1). Interestingly, the analysis further revealed that only in mice, the soma volume and diameter of NG2 glia exceeds those of astrocytes ( Figure 1).

| Methods to study human glial cells
Data on functional properties of human glial cells are sparse due to the limited availability of healthy human tissues and technical difficulties with recordings from human slices. Most analyses have been performed on acute brain slices surgically resected from patients with medically intractable TLE. Here, the experiments were either performed in acute slices from cortical specimens resected to gain access to the epileptogenic area (Navarrete et al., 2013;Oberheim et al., 2009), or in nonsclerotic hippocampal slices from patients with "lesion-associated" TLE that lacked significant histopathological hippocampal alterations Heinemann et al., 2000;Hinterkeuser et al., 2000;Jauch, Windmüller, Lehmann, Heinemann, & Gabriel, 2002;Kivi et al., 2000;Navarrete et al., 2013;Seifert et al., 2002;Seifert, Hüttmann, Schramm, & Steinhäuser, 2004). In the latter, seizures are generated by focal lesions in the temporal lobe, such as tumors (e.g., ganglioglioma or dysembryoplastic neuroepithelial tumors), malformations of cortical development (mainly focal cortical dysplasia) or vascular malformations (Blümcke et al., 2017). Although these specimens display anatomically preserved structures and lack significant neuropathological changes, it must be considered that seizure activity and/or medication may have affected cellular properties.
Besides tissue from epilepsy surgery, biopsy samples from tumor patients without epileptic seizures were used for functional analysis of cortical astrocytes (Bordey & Sontheimer, 1998;Oberheim et al., 2009;Picker, Pieper, & Goldring, 1981). However, such specimens are rare, and though cells in tissue outside the tumor margins are considered functionally unaffected, tumor-or medication-associated alterations cannot be completely ruled out. Moreover, biopsies are obtained from different cortical subareas, which may limit comparability of the results.

| Electrophysiological properties of human astrocytes
In rodents, adult astrocytes are characterized by a high resting K + conductance, giving rise to a very low input resistance (R i ) (<5 MΩ), a resting potential close to the equilibrium potential of K + (E K ) and almost linear ("passive") current to voltage relationships (Jabs, Seifert, & Steinhäuser, 2008;Verkhratsky & Parpura, 2015). First whole cell patch-clamp recordings from glial cells in acute human cortical and hippocampal specimens were performed by Bordey and Sontheimer (1998) and Hinterkeuser and colleagues .
However, the two studies cannot be directly compared because (a) different K + concentrations ([K + ]) were used in the pipette and bath solutions, which affects resting potentials and (b) the age of the patients analyzed varied greatly (4 months to 14 years vs. mean age of 34 years), which may explain the higher R i of cells in the former study (R i = 288 MΩ vs. 140 MΩ) (for developmental changes in R i , see Kressin, Kuprijanova, Jabs, Seifert, & Steinhäuser, 1995). Referring to these two studies, it has been speculated that an apparently higher R i in human versus rodent astrocytes represents an evolutionary adaptation to the larger size of astrocytes, as it results in an increased length constant (Oberheim Bush & Nedergaard, 2017;Vasile et al., 2017).
However, the whole cell current patterns shown in the papers by Bordey and Sontheimer (1998) and Hinterkeuser et al. (2000) indicate that the authors had analyzed NG2 glial cells and not astrocytes. In fact, at the end of the 1990s and beginning of the 2000s, glial cells with time-and voltage-dependent transmembrane currents were still erroneously regarded as immature astrocytes (discussed in Bergles et al., 2010). This conclusion is strongly supported by a later study that systematically compared properties of hundreds of bona fide astrocytes and NG2 glial cells in "control-like" (i.e., nonsclerotic) hippocampi from patients with pharmacoresistant TLE. In this study, human astrocytes always showed predominating passive currents lacking time-or voltage-dependence, similar to their rodent counterpart . Importantly, by further evaluating recordings from our latter study, we determined an average R i of human astrocytes of 4.6 ± 3 MΩ (n = 21 unpublished), which is in the range of R i of rodent astrocytes (juvenile: 8.1 ± 6.5 MΩ, n = 38, Jabs et al., 2005; adult: 3.3 ± 0.9 MΩ, n = 63; unpublished). The resting potential of human astrocytes was −67.9 ± 4.5 mV (liquid junction potential not compensated for, unpublished observation), which is comparable to that in mouse astrocytes of the same region (juvenile: −82 ± 4 mV, is 13 mV more negative. The finding that C m of human (70 ± 22 pF, n = 21) and murine (74 ± 17 pF, n = 63; unpublished data) astrocytes did not differ suggests that in the same subregion (CA1 stratum radiatum), astrocytes from both species have a similar cell size. The resting K + conductance in astrocytes of rodent hippocampus is mainly mediated by Kir4.1 channels (Seifert et al., 2009;reviewed by Seifert, Henneberger, & Steinhäuser, 2018) and Kir4.1 encoding transcripts have also been detected in human hippocampal astrocytes . Together, these findings indicate that the membrane properties of astrocytes are largely conserved between rodents and humans.

| Electrophysiological properties of human NG2 glia
The functional properties of NG2 glial cells are unique among glia, as they express various types of ligand-and voltage-gated ion channels typically attributed to neurons Larson, Zhang, & Bergles, 2016). In rodents, the resting potential of NG2 glia is very negative (i.e., close to E K ), due to a high K + conductance at negative voltages (Larson et al., 2016). However, the density of these channels seems to be lower than in astrocytes, as NG2 glia have a substantially higher R i . The resting K + conductance (mainly mediated by Kir4.1 channels; Tang, Taniguchi, & Kofuji, 2009) increases during postnatal development (Bordey & Sontheimer, 1997;Kressin et al., 1995;Maldonado, Vélez-Fort, Levavasseur, & Angulo, 2013), producing more passive whole cell current patterns and a lower membrane resistance in adult rodents (~30-50 MΩ in adults vs.~100-500 MΩ in juveniles; Kressin et al., 1995;Lin & Bergles, 2004;Kukley et al., 2008;Haberlandt et al., 2011;Braganza et al., 2012;Maldonado et al., 2013;Larson et al., 2016;Passlick, Trotter, Seifert, Steinhäuser, & Jabs, 2016). Although NG2 glial cells receive synaptic input and express almost the same set of voltage-gated ion channels as neurons, they are not able to fire action potentials, due to the relatively low density of voltage-gated Na + (Na v ) channels (De Biase, Nishiyama, & Bergles, 2010;Larson et al., 2016). Hinterkeuser et al. (2000) have functionally characterized human NG2 glia in the histopathologically intact human hippocampus resected from patients with lesion-associated TLE (although at that time, these cells were still thought to be immature astrocytes; see Section 3.2). The human cells exhibited mean R i (140 MΩ) and C m (26 pF) values similar to those determined in rodent NG2 glia (Larson et al., 2016). About 75% of the human cells displayed TTX-sensitive Na + currents, with an average density of 17 pA/pF, also closely matching data reported from mouse (10-30 pA/pF Steinhäuser, Kressin, Kuprijanova, Weber, & Seifert, 1994;De Biase et al., 2010) or rat hippocampus (16 pA/pF; Xie et al., 2007). These current densities are about an order of magnitude lower than those typically seen in neurons. Current injections into human NG2 glia never produced action potentials, similar to rodent NG2 glia. Hinterkeuser et al. (2000) and Schröder et al. (2000) furthermore demonstrated that human NG2 glial cells express voltage-gated A-type (K A ) and delayed rectifier (K DR ) K + channels as well as inward-rectifier Kir4.1 K + channels, similar to what has been found in rodents (Larson et al., 2016). Although Hinterkeuser et al. (2000) performed the measurements in hippocampal NG2 glia from adult patients (average age 34 years), the density of inwardly rectifying K + currents was apparently lower than in cells from adult mouse hippocampus (Braganza et al., 2012). However, this difference might have been due to different protocols used for Kir current isolation (leak current subtraction has been performed in the former study). A large number of NG2 glia in the nonsclerotic human hippocampus was recently characterized by Bedner et al. (2015). The authors demonstrated that human hippocampal NG2 glia (a) lack gap junction coupling, (b) express ionotropic glutamate receptors, and  et al. (2015) displayed no developmental changes and were similar to those reported by Hinterkeuser et al. (2000).
Together, these data demonstrate that basic membrane properties, that is, ion channel expression profiles, and its developmental regulation, of NG2 glia in rodent and human hippocampus do not differ.

| Glutamate sensitivity of human astrocytes and NG2 glia
One of the major roles of astrocytes is clearance of excessive glutamate from the synaptic cleft, a pivotal mechanism for normal excitatory neurotransmission and protection against excitotoxicity (Schousboe, Scafidi, Bak, Waagepetersen, & McKenna, 2014). This is accomplished via specific transporters, called excitatory amino acid transporters (EAATs), which are enriched in perisynaptic astrocyte membranes and utilize the electrochemical gradient of Na + and K + as a driving force for transmembrane movement of glutamate. Astrocytes express EAAT1 and EAAT2, often referred to as glutamateaspartate transporter 1 (GLAST1) and glutamate transporter 1 (GLT-1) in rodents (Vandenberg & Ryan, 2013). Although several studies have investigated expression of EAAT protein and transcripts in the human brain (e.g., Bjørnsen et al., 2007;Roberts, Roche, & McCullumsmith, 2014), their functionality in situ has been examined only recently . In the latter study, rapid application of glutamate to outside-out patches excised from the soma of astrocytes in nonsclerotic human hippocampus evoked transient inward currents at negative membrane potentials, which were completely inhibited by the glutamate transporter blocker DL-TBOA, but were insensitive to the AMPA/kainate receptor antagonist NBQX, indicating that the observed currents were due to glutamate uptake through EAAT transporters ( Figure 2; Bedner et al., 2015).
NG2 cells in rodent hippocampus express functional AMPA and GABA A receptors and receive direct synaptic input from neurons (Bergles et al., 2000;Jabs et al., 2005;Lin & Bergles, 2004). AMPA receptors are tetramers formed by the subunits GluA1-4 which, dependent on their subunit composition, considerably vary in their functional properties. In mouse hippocampal NG2 glia, transcripts of all four subunits were detected, with co-expression of GluA1, GluA2, and GluA4 being most frequent (Seifert, Weber, Schramm, & Steinhäuser, 2003;Seifert, Zhou, & Steinhäuser, 1997). In the juvenile hippocampus, a mixture of AMPA receptors with high and low Ca 2+ permeability coexist in individual cells, while receptors in NG2 glia from older animals are more uniform and display a lower divalent cation permeability . AMPA receptor subunits occur in two splice variants, called flip and flop, which determine the gating properties of the receptors. The GluA2 flip variant is upregulated during maturation of hippocampal NG2 glia . It has been shown that sustained activation of AMPA receptors in NG2 glial cells of mouse hippocampus inhibits Kir currents (Schröder, Seifert, Hüttmann, Hinterkeuser, & Steinhäuser, 2002).
NG2 glia in human hippocampal tissue resected from patients with intractable TLE also express functional AMPA receptors Seifert et al., 2002;Seifert et al., 2004). Indeed, bath application of the receptor agonist kainate in situ produced receptor currents in the human cells that were completely inhibited by the specific AMPA receptor antagonist, GYKI53655. This antagonist also blocked the glutamate-induced responses in acutely isolated cells, indicating that human NG2 glia, like their rodent counterparts (Matthias et al., 2003), selectively express ionotropic glutamate receptors of the AMPA subtype . In accordance with findings in rodents, AMPA receptors in human NG2 glia exhibit an intermediate Ca 2+ permeability, express mainly GluA1, GluA2, and GluA4, and their activation inhibits K + currents . Species-dependent differences were detected in AMPA receptor desensitization kinetics, which was faster in human than in mouse NG2 glia. Variable splicing of the receptor subunits might account for this difference, as the GluA1 flip/flop splice variant ratio was considerably lower in human versus mouse NG2 glia Seifert et al., 2004). In a later study, the glutamate sensitivity of human hippocampal NG2 glial cells was investigated through flash photolysis of caged glutamate in F I G U R E 2 Characterization of astrocytes and NG2 glial cells in the nonsclerotic human hippocampus. (a) The whole-cell current pattern of an astrocyte (left; 50-ms voltage steps ranging from −160 to +20 mV; 10 mV increments; V hold = −80 mV) was dominated by a passive resting conductance. Rapid application of glutamate to an outside-out patch failed to induce outward currents at positive voltages (middle and right), indicating the absence of ionotropic receptors. The inward currents were due to glutamate uptake. (b) Gap junction coupling between hippocampal astrocytes visualized by diffusion of biocytin from a single cell, filled with the tracer through the patch pipette during 20 min of whole cell recording. Scale bar = 100 μm. (c) Typical whole-cell current pattern of an NG2 glial cell (left). Depolarization and hyperpolarization activated time-and voltage-dependent outward and inward currents. Flash photolysis of caged glutamate activated currents with a linear current/voltage relationship (middle and right) indicating expression of ionotropic receptors and lack of glutamate transporters. (d) Biocytin filling of NG2 glial cells revealed lack of tracer coupling. Scale bar = 25 μm. Modified from Bedner et al. (2015), reproduced with permission situ or fast glutamate application to excised patches . Both methods revealed rapidly decaying currents with linear current to voltage relationships and a reversal potential close to zero ( Figure 2). Irrespective of the age of the patients (1-45 years), the currents were completely blocked by a nonselective ionotropic glutamate receptor antagonist, kynurenic acid, and by the AMPA/kainate receptor antagonist NBQX, further supporting previous observations. It is, therefore, reasonable to assume that, like their rodent counterparts, NG2 glial cells in the human hippocampus receive synaptic inputs from glutamatergic neurons. This assumption, however, still requires experimental validation.
In conclusion, NG2 glia and astrocytes in the human and mouse hippocampus display the same segregated expression of glutamate receptors and transporters (Matthias et al., 2003).
Cx30 and Cx43 protein and transcripts are also expressed in the human (Aronica et al., 2001;Collignon et al., 2006;Deshpande et al., 2017;Elisevich, Rempel, Smith, & Edvardsen, 1997;Fonseca, Green, & Nicholson, 2002;Naus, Bechberger, & Paul, 1991). However, the presence of protein does not necessarily mean that functional channels are formed. The first characterization of functional gap junctionmediated coupling between human astrocytes has recently been performed by Bedner et al. (2015) in acute slices from nonsclerotic hippocampus of TLE patients. In this study, biocytin, a low molecular weight tracer that easily crosses gap junction channels, spread from an injected cell into almost 100 neighboring cells (Figure 2; Bedner et al., 2015). This extent of coupling is in the same range as in adult rodent hippocampus (Gosejacob et al., 2011;Griemsmann et al., 2015;Wallraff et al., 2004). The relative contribution of Cx43 versus Cx30 to intercellular coupling was not determined in human hippocampus, but in a subsequent study, the Western blot analysis revealed similar protein expression levels of the two isoforms in man and mouse (about 10-fold more Cx43 than Cx30 in both species ;Deshpande et al., 2017), suggesting that coupling between astrocytes in human hippocampus is also mainly accomplished by Cx43. Coupling of astrocytes seems to be a key prerequisite for maintaining ion homeostasis and proper neuronal signaling, because its loss was suggested to cause epilepsy .

| K + buffering by human astrocytes
During neuronal activity, K + is released into the extracellular space where its concentration transiently increases. Since extracellular accumulation of K + would lead to neuronal depolarization and hyperexcitability, tight control of K + homeostasis has been hypothesized as a major function of astrocytes (Kofuji & Newman, 2004;Walz, 2000).
Astrocytes balance extracellular K + levels mainly by two mechanisms, K + uptake and spatial buffering. Net uptake of K + is mainly mediated by Na + /K + ATPases and Na + -K + -Cl − cotransporters and accompanied by cell swelling and local depolarization of astrocytes (D'Ambrosio, Gordon, & Winn, 2002;Kofuji & Newman, 2004;Ransom, Ransom, & Sontheimer, 2000;. A very effective and energy independent mechanism of glial K + clearance describes the spatial buffering hypothesis, according to which excessive extracellular K + is taken up by astrocytes at sides of high neuronal activity, and then redistributed through the astrocytic gap junctioncoupled network to be released at distant regions of lower [K + ] o . Here, uptake and release of K + occur passively, via diffusion through Kir4.1 channels (Kofuji & Newman, 2004;Orkand et al., 1966;Walz, 2000).
The ability of human astrocytes to buffer extracellular K + through mechanisms involving Kir channels has already been investigated almost 20 years ago (Heinemann et al., 2000;Jauch et al., 2002;Kivi et al., 2000). In these studies, the authors used ion-selective microelectrodes to assess the effect of Ba 2+ , which blocks glial Kir channels at sub-mM concentrations, on rises in [K + ] o , induced either by repetitive alvear stimulation or by iontophoretic application of K + in hippocampal tissue from patients with intractable TLE. In the nonsclerotic human hippocampus, even μM concentrations of Ba 2+ substantially augmented [K + ] o rises induced by both methods, suggesting a significant contribution of glial Kir channels to K + clearance (Heinemann et al., 2000;Jauch et al., 2002;Kivi et al., 2000). Together with the above-mentioned findings that human astrocytes possess a high resting K + conductance and are abundantly interconnected through gap junctions, these results argue in favor of a crucial involvement of spatial K + buffering in removal of external K + in the human hippocampus.

| Ca 2+ signaling and gliotransmission in human astrocytes
Intracellular [Ca 2+ ] elevations in astrocytes that propagate through the astroglial network have initially been described in rat hippocampal slices (Porter & McCarthy, 1996) and later also in rat and mouse in vivo (Kuga, Sasaki, Takahara, Matsuki, & Ikegaya, 2011;Kurth-Nelson, Mishra, & Newman, 2009). Intercellular Ca 2+ waves can occur spontaneously or in response to a variety of stimuli, such as mechanical stimulation or neurotransmitters released from synaptic terminals.
Ca 2+ signaling in human astrocytes was first studied in acute slices from the neocortex (Oberheim et al., 2009). The authors showed that Ca 2+ signals evoked in a single cortical astrocyte by photolysis of caged Ca 2+ may induce subsequent Ca 2+ wave propagation. Interestingly, the speed of Ca 2+ wave propagation was about 5 times faster in human compared to rodent cortex. This study also revealed that [Ca 2+ ] i elevations in human astrocytes can be triggered by ATP and glutamate, which is similar to rodent astrocytes (Oberheim et al., 2009). In a later study, Navarrete et al. (2013) reported that human astrocytes in acute hippocampal and cortical slices from TLE patients exhibit spontaneous Ca 2+ transients that were independent of neuronal activity. Moreover, astrocyte Ca 2+ elevations could be stimulated by local application of glutamate, cannabinoid, and purinergic receptor agonists as well as by electrical stimulation, indicating that they are induced by synaptic activity (Navarrete et al., 2013). Intriguingly, the authors could also show that astrocyte [Ca 2+ ] i transients stimulated by local ATP application increased the frequency of NMDA-mediated currents in both cortical and hippocampal neurons, suggesting that human astrocytes, like their rodent counterparts, are able to modulate synaptic activity through Ca 2 + -dependent release of glutamate. These data provide evidence for the presence of gliotransmission and bidirectional neuron-astrocyte signaling in the human cortex and hippocampus, and it is, therefore, reasonable to assume that the concept of the "tripartite synapse" proposed for rodents applies also to the human brain (Navarrete et al., 2013).

| CONCLUSIONS
The important role of glial cells for proper brain signaling is increasingly recognized, and evidence emerges suggesting that dysfunctional glial cells are key players in the etiology of neurological diseases (Parpura et al., 2012;Seifert et al., 2006). As medical research relies on animal models, knowledge about species differences in the properties of glial cells is vital to predict the translatability of animal data to humans. Rodents are the most widely used experimental model organisms to investigate human diseases. Surprisingly, careful comparison reveals that astrocytes and NG2 glia in rodents and human share many similar properties.
Morphological analyses revealed a higher glia-to-neuron ratio in the human versus rodent neocortex (1.6 vs. 0.3; Verkhratsky & Nedergaard, 2016). Human cortical astrocytes are larger, contact more synapses, and are more diverse than their rodent counterparts (Colombo et al., 1995;Oberheim et al., 2009). Intriguingly, results available so far demonstrate that the functional properties of astrocytes and NG2 glial cells are amazingly similar between rodents and humans. However, the current data situation must be interpreted with caution. First, functional data are only available from two human brain regions, neocortex and hippocampus, and it is unknown whether this similarity also applies to other areas of the CNS. Notably, rodent studies have revealed a high degree of functional heterogeneity among astrocytes, both within a given and across different brain regions. Second, in neurosurgically resected "control-like" human brain tissue, even if lacking obvious anatomical alterations, medication may have affected cellular properties. Third, reliable analysis of astrocyte function and its interactions are principally hampered by the very low electrical compactness of these cells and their fine distant processes, which fall below current optical resolution (Ma, Xu, Wang, Enyeart, & Zhou, 2014;Seifert et al., 2009). This limitation applies to both rodent and human astrocytes and hampers precise biophysical analyses and comparison. Finally, it should be considered that in many studies, different experimental conditions were chosen (K + concentrations in the solutions used, temperature, age of animals, and so on), which limits the comparability of published data. Previous reviews even compared properties of human NG2 glia with rodent astrocytes, leading to wrong conclusions as to apparent species-dependent differences. Despite these limitations, it can be concluded that based on the data currently available, human and rodent glial cells share many functional properties, providing justification for using mouse or rat models to investigate causes of human brain diseases.

ACKNOWLEDGMENT
The work of the authors quoted in this review is currently supported by grants from the EU (ITN project EU-GliaPhD) and BMBF (CONNEXIN). We thank Larissa Schmitz-Ullrich for help with the morphological analysis.

CONFLICT OF INTEREST
The authors declare no conflict of interest.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.