Morphological studies on tissue resected from the human cortex provide important information regarding the structural features of distinct neuronal types and how these features may impart a specific function within the microcircuit it is embedded in. Each neuron has the capacity to perform complex computations which are reliant on its relative dendritic and axonal structure. Various 3D constructions of pyramidal neurons of resected human tissue have revealed great heterogeneity across and within cortical layers (Deitcher et al., 1991; Mohan et al., 2015; Beaulieu-Laroche et al., 2018; Kalmbach et al., 2018, 2021; Berg et al., 2021; Moradi Chameh et al., 2021). As a function of depth, extending from the pia to layer 6 (L6), human cortical neurons exhibit increases in morphological complexity, arising from their size, dendritic arborization, and cellular density that are unparalleled in the rodent cortex (Deitcher et al., 1991; Mohan et al., 2015; Kalmbach et al., 2018, 2021; Berg et al., 2021; Moradi Chameh et al., 2021). Importantly, these structural findings were shown to be unrelated to disease etiology, severity, or duration (Mohan et al., 2015), suggesting these patient specific factors did not impact the morphology of human cortical neurons resected from epilepsy patients.

When comparing the morphological properties of cortical neurons obtained from human tissue to those in well characterized preclinical rodent models, many species-related differences were identified (Mohan et al., 2015; Berg et al., 2020). As shown in transcriptomic studies, human cortical layer 2 (L2) and layer 3 (L3) have well-defined boundaries unlike the homogeneous presentation of rodent L2/3 which are practically indistinguishable (DeFelipe, 2011; Markram et al., 2015; Gouwens et al., 2019) (Figure 1). These transcriptomic findings have been further confirmed by morphological studies showing that human cortical pyramidal neurons in L2 and L3 can be subdivided based on their dendritic morphology: slim-tufted, which has a low density of tuft branches, and profuse-tufted, which has a high density of tuft branches (Deitcher et al., 1991). These two cells show no depth separation in L2 and L3 but have varying biophysical properties that underlie functional differences (Deitcher et al., 1991). Interestingly, there is no existing homologous equivalent of slim-tufted and profuse-tufted cells within the rodent L2/3 cortex (Deitcher et al., 1991). When performing analysis on morphological data, one study found that 88% of human L2 and L3 pyramidal neurons can be grouped into human-specific clusters based on limited shared features with other L2/3 neurons from species such as mice and macaque monkeys (Mohan et al., 2015). In addition, the dendritic architecture of the L2 and L3 pyramidal neurons of humans, as compared to L2/3 pyramidal neurons of mice, demonstrate increased length (three times larger) and advanced arborization (Mohan et al., 2015). These findings reflect the uniqueness of slim- and profuse-tufted neurons to the L2 and L3 of the human cortex (Figures 1A, C).

Similar to L2 and L3, human layer 5 (L5) cortical neurons can also be subdivided with respect to their long-range axonal projection targets: extra-telencephalic (ET), which project to areas within and outside the telencephalon, or intra-telencephalic (IT) which have projections that are restricted to the telencephalon (Baker et al., 2018). Unlike slim- and profuse-tufted neurons, both ET and IT cells are found in the L5 cortical layer of mice yet exhibit species-related differences (Kalmbach et al., 2021) (Figures 1B, D). One of the more prominent species-related differences between L5 of human and rodent tissue is the abundance and distribution of ET cells. In the human cortex, ET cells are relatively scarce and only account for 2–6% of all excitatory neurons in L5. In contrast, ET cells in mice are much more abundant, comprising 20–30% of all L5 excitatory neurons (Hodge et al., 2019; Kalmbach et al., 2021). Despite these discrepancies in abundance, the morpho-electric properties and gene expression profiles of ET cells in humans are just as distinctive as those of their rodent counterparts (Beaulieu-Laroche et al., 2018, 2021; Kalmbach et al., 2021). The exact proportions of ET cells within L5 of resections of human tissue is subject to natural variability between patients (Hodge et al., 2019; Kalmbach et al., 2021). With respect to dendritic length, the apical dendrites of ET cells extend from L5 to the pia, however human cortical L5 IT cells have dendrites that extend past L3 (Baker et al., 2018; Beaulieu-Laroche et al., 2018; Kalmbach et al., 2021). Originally, this finding sparked much debate about whether or not the dendrites of IT cells were being unintentionally truncated during human cortical slice preparation.

Human cortical neurons are not simply scaled up compared to other species, but they also exhibit more complex dendritic arborization, with each branch capable of functioning independently (Deitcher et al., 1991; DeFelipe et al., 2002; Mohan et al., 2015; Gidon et al., 2020). This enhanced dendritic arborization allows human cortical neurons to make more extensive and widespread synaptic contacts with their various dendritic spines, thereby increasing their processing and computational abilities as compared to other homologous neurons (Spruston, 2008). Even when considering human cortical layers and their encompassing cell types, the abundance of dendritic spines are significantly greater than seen in mice (Figures 3A–D) (Elston et al., 2001; Eyal et al., 2018; Iascone et al., 2020). A typical human cortical neuron is expected to demonstrate ~25,000–30,000 spines that possess larger spine heads, longer neck length and diverse morphologies than their rodent counterparts (Benavides-Piccione et al., 2002; Ofer et al., 2022). In comparison to mice, human cortical neurons form three to four times more synapses and have a higher ratio of asymmetric (excitatory) to symmetric (inhibitory) synapses per neuron in all layers except layer 4 (L4) (DeFelipe et al., 2002; DeFelipe, 2011; Benavides-Piccione et al., 2013). Overall, the unique morphological properties of human neurons with respect to size, dendritic length and arborization, projection targets, and number of synapses likely underscores the distinctive circuitry connections in humans that impart advanced cognitive and executive functioning (Galakhova et al., 2022). The enhanced memory capacity of human cortical neurons and their networks also suggests that these differences may have important implications for human learning and memory processing (Schmidt and Polleux, 2022).

The evolution of the human brain, particularly the increase in size and complexity, implies that there is a functional specialization in neuronal electrophysiological properties in the human cortex (DeFelipe, 2011). Human cortical pyramidal neurons exhibit an astonishing degree of electrophysiological heterogeneity within each cortical layer, which is consistent with transcriptomic and morphological heterogeneity (Deitcher et al., 1991; Zeng et al., 2012; Mohan et al., 2015; Kalmbach et al., 2018; Tasic et al., 2018; Berg et al., 2020; Moradi Chameh et al., 2021).

The excitability of cortical pyramidal neurons in the L2 and L3 of the human cortex varies with depth and also between slim-tufted and profuse-tufted cells. Specifically, profuse-tufted cells have higher firing rate than slim-tufted cells (Deitcher et al., 1991; Kalmbach et al., 2018; Berg et al., 2020; Moradi Chameh et al., 2021) These biophysical differences allow these cells to process functionally independent information streams, with feedforward projections originating in L3 neurons and L2 neurons receiving feedback information (Markov and Kennedy, 2013). The differences in electrophysiological properties between human L2 and L3 neurons and rodent L2/3 neurons are likely functional manifestations of the increased dendritic length and arborization seen in humans (Deitcher et al., 1991; Mohan et al., 2015; Kalmbach et al., 2018; Berg et al., 2020; Moradi Chameh et al., 2021). As a result of the large size of human L2 and L3 neurons, these neurons undergo compartmentalization of the soma and dendrites that reduce electrical coupling between these areas. In turn, dendritic spiking does not necessarily yield somatic depolarization, which is commonly referred to as voltage attenuation (Beaulieu-Laroche et al., 2018; Gidon et al., 2020). To compensate for this compartmentalization and the resulting voltage attenuation, changes in membrane capacitance and channel expression are observed (Eyal et al., 2016; Kalmbach et al., 2018; Gidon et al., 2020). First, the membrane capacitance (Cm) of L2 and L3 pyramidal neurons in the human temporal cortex is approximately half of the commonly accepted value seen in rodent cortical neurons (~1 μF/cm²) (Eyal et al., 2016), however, see Beaulieu-Laroche et al. (2018). As a result, it is suggested that human L2 and L3 pyramidal neurons are able to transfer electrical signals along their dendritic arbors more efficiently, despite the increased length. Secondly, increases in the expression of hyperpolarization-activated cyclic nucleotide-gated (HCN) channels compensates for an attenuation of voltage through increasing HCN-mediated currents (Ih) and the reliability of transmission of electrical signals along long apical dendrites (Kalmbach et al., 2018). The expression of HCN channels in L2 and L3 of human cortical tissue demonstrates variability depending on depth, such that a higher expression of HCN channels is found in deep L3 (Kalmbach et al., 2018). This heterogeneous presentation of HCN currents is responsible for the functional specialization of neurons within these layers due to its role in altering input resistance, resting membrane potential, and sag voltage of neurons (Beaulieu-Laroche et al., 2018; Kalmbach et al., 2018; Moradi Chameh et al., 2021). Moreover, HCN channel expression and the consequent production of Ih is more prevalent in human cortical neurons than their rodent counterparts, indicative of species-dependent differences in neuronal function (Figures 2A, B) (Beaulieu-Laroche et al., 2018; Kalmbach et al., 2018). As an example, human pyramidal neurons in both deep L3 and L5 have a larger Ih than those in L2, which enhances their responsiveness to track the delta frequency compared to their rodent counterparts (Kalmbach et al., 2018; Moradi Chameh et al., 2021; Inibhunu et al., 2023). Finally, the attenuation of voltage in human cortical neurons is also compensated by the presence of calcium-mediated dendritic action potentials (dCaAPs) which act to enhance electrogenesis, the production of neuronal electrical activity, and excitability (Gidon et al., 2020) (Figures 2A, C). Overall, these alterations are unique to human cortical neurons for overcoming large dendritic distances.

Similar to rodent neurons, the dendrites of human cortical L2 and L3 pyramidal neurons are capable of actively propagating action potentials in both directions and generating local sodium/calcium spikes exhibiting both action potential backpropagation and locally generated sodium/calcium spikes (Beaulieu-Laroche et al., 2018; Gidon et al., 2020). However, the ability of human cortical L2 and L3 neurons to evoke NMDA spikes was found to be significantly reduced compared to that of rodent neurons, likely due to the larger diameter of the dendrites in humans (Testa-Silva et al., 2022). Unlike L2 and L3 of the human cortex, which showed remarkable differences from their rodent counterparts, L5 cortical cells of humans share some electrophysiological properties with mice (Beaulieu-Laroche et al., 2018; Kalmbach et al., 2021; Moradi Chameh et al., 2021). Specifically, there are two types of pyramidal neurons found in L5 of human and rodent cortical tissue: ET and IT cells. In both species, ET and IT cells show distinct electrophysiological features, including differences in membrane properties related to Ih such as sag voltage, rebound depolarization and rebound spike, resting membrane potential, and input resistance (Baker et al., 2018; Kalmbach et al., 2021). ET cells in both human and rodents exhibit larger sag voltage and rebound depolarization, depolarized resting membrane potential, and smaller input resistance compared to IT cells (Figures 3E, F) (Kalmbach et al., 2021). Due to their larger sag voltage, ET cells can track subthreshold resonance frequencies in the range of 3–6 Hz, while IT cells respond to lower frequency oscillations (<2.5 Hz) (Beaulieu-Laroche et al., 2018; Berg et al., 2021). The differential expression of HCN channels in ET cells plays a crucial role in the integration of synaptic inputs. ET cells, with their lower input resistance, exhibit lower firing rates compared to IT cells and display bursting activity near their threshold, which is likely to be associated with dendritic calcium electrogenesis, as previously reported (Kalmbach et al., 2021). However, a recent study showed that human L5 neurons have lower electrogenesis in distal dendrites compared with L5 neurons of other mammalian species, including rodents, rabbits, marmosets, and macaques, and thus increased computational compartmentalization (Beaulieu-Laroche et al., 2021). In terms of morphology, ET and IT cells exhibit similarities in both human and rodents. However, when it comes to electrophysiology, there are notable differences between ET cells in humans and rodents. ET cells in humans exhibit fast action potential (AP) kinetics and display high-frequency AP burst firing near threshold. This firing pattern is associated with dendritic calcium ion (Ca²⁺) electrogenesis (Kalmbach et al., 2021). Interestingly, the human IT cells can be further subdivided into IT-like 1 and IT-like 2. Human IT-like 2 cells exhibit a two-fold higher input resistance and larger rebound depolarization than IT-like 1 cells. Furthermore, the activity of IT-like 2 neurons is more similar to that of ET neurons, both exhibiting fast and narrow action potentials (Kalmbach et al., 2021).

In other words, each layer of the human cortex has a diverse repertoire of neurons that show unique functional specializations based on their electrophysiological features, underscoring the divergence between neuronal networks of human vs. preclinical animal models. These findings reflect species-related differences in neuronal physiology and emphasize the necessity to perform electrophysiological research on resected human tissue.

Recent breakthroughs in transcriptomic techniques, including the development of single cell/nucleus RNA sequencing (scRNA-seq/snRNA-seq) and patch-sequencing, have allowed researchers to quantify the RNA expression of hundreds to thousands of individual cells simultaneously (Zeng et al., 2012; Bakken et al., 2016; Cadwell et al., 2016; Lake et al., 2016; Wu et al., 2017; Tasic et al., 2018; Tripathy et al., 2018; Hodge et al., 2019). Notably, this field of research has set the foundation for investigating the transcriptomics of human neurons, discovering the extent of within-cell heterogeneity that is present in traditionally defined cell types and subtypes, and comparing these unique genetic profiles to other species (Lake et al., 2016; Hodge et al., 2019). Through assessing the transcriptome of human cortical tissue resected from patients with epilepsy or tumor, cell types representing non-neuronal, excitatory, and inhibitory cells can be classified based on molecular constituents. Keeping a reductionist mindset, each individual human cortical neuron can be assigned into a hierarchical classification system: (1) class, (2) subclass, (3), type, (4) and subtype. Transcriptomic techniques have shown that even within the narrowest classification of subtype, there is both continuous and discrete variation with respect to RNA expression (Zeng et al., 2012; Lake et al., 2016; Berg et al., 2021; Scala et al., 2021). Such variability is increasingly being accepted as a design feature of the brain rather than biological “noise” (Cembrowski and Menon, 2018; Cembrowski and Spruston, 2019), with recent work suggesting that biophysical diversity can stabilize brain dynamics in face of a number of “insults” (Rich et al., 2022; Hutt et al., 2023). The human MTG, which is the most heavily characterized region of the human brain electrophysiologically, demonstrates remarkable transcriptomic diversity in excitatory and inhibitory neuron classes, with 45 inhibitory and 24 excitatory cell types (Hodge et al., 2019). Interestingly, excitatory, and inhibitory neuronal cell types do not adhere to the tight laminar organization of the cortex and instead are found to be widely dispersed across the layers of the human cortex (Hodge et al., 2019). Due to the extent of heterogeneity within neuronal classes, subclasses, types, and subtypes, analyzing the transcriptome of individual cells rather than using laminar location represents a more reliable method for predicting the identity of a neuron (Zeng et al., 2012; Hodge et al., 2019).

Through comparisons of the transcriptomic homology between human MTG and rodent cortex, 21% of genes within the human cortex show species-specific differential expression from homologous areas in mice (Zeng et al., 2012). However, all major classes and subclasses of cortical neurons can be aligned between species on the basis of shared marker genes, suggesting highly conserved transcriptomic organization (Zeng et al., 2012; Hodge et al., 2019). This species conservation begins to deteriorate when examining gene expression at the cell type level which reveals a mix of convergent and divergent genes (Zeng et al., 2012; Hodge et al., 2019; Kalmbach et al., 2021). The considerable differences in transcriptomic profiles between corresponding cell types of human and rodent tissue demonstrates species-differences in cell type genetic profiles, proportions, laminar distribution, and even function (Zeng et al., 2012; Hodge et al., 2019). For example, human cortical L2 and L3 demonstrate notable differences in neuronal size, density, morphology, and electrophysiological properties that can be attributed to transcriptomic differences (Mohan et al., 2015; Lake et al., 2016; Kalmbach et al., 2018, 2021; Berg et al., 2021; Moradi Chameh et al., 2021). In contrast, neurons within L2/3 of the rodent cortex are homogenous, lacking the well-defined boundaries seen in the human such that the layers are often considered inseparable (DeFelipe, 2011; Markram et al., 2015). To some extent, these species-related differences in laminar organization occur as a result of five transcriptomic cell types- LTK, GLP2R, FREM3, CARM1P1 and COL22A1- being present in L2 (LTK, GLP2R) and L3 (FREM3, CARM1P1 and COL22A1) of the human cortex while only three homologous counterparts have been identified in L2/3 of the rodent. There is no homolog of CARM1P1 and COL22A1 in mice (Berg et al., 2021). Furthermore, gene expression patterns and markers that are characteristic of L5 in the rodent are either eliminated, reduced, or shifted to L3 of the human (Zeng et al., 2012). Overall, transcriptomic data has illuminated species-specific similarities and differences in laminar organization as well as neuronal cell types, thereby emphasizing the importance of conducting research on the human brain in addition to preclinical models (Hodge et al., 2019; Scala et al., 2021).

After myelin-competent oligodendrocytes, astrocytes comprise the largest class of glial cells in the mammalian CNS (Pelvig et al., 2008; Herculano-Houzel, 2014). They have key roles in maintaining the blood-brain barrier (Abbott et al., 2006), regulating regional blood flow (MacVicar and Newman, 2015), providing trophic, antioxidant, and metabolic support to neurons (Allaman et al., 2011), neurotransmitter uptake and recycling (Simonato et al., 2014; Weber and Barros, 2015), and regulating synaptogenesis and synaptic transmission (Araque et al., 2014; Chung et al., 2015). While these functions appear generally conserved across vertebrates, being equally documented in the murine and human brain (Vasile et al., 2017; Bedner et al., 2020), there is the emerging recognition of distinct human features for their cellular and molecular underpinnings (Oberheim et al., 2006; de Majo et al., 2020). As human neurons are larger than their rodent counterparts (Mohan et al., 2015), so are human astrocytes. Human astrocytes in the white matter have a maximum linear dimension of 183.2 μm, which is 2.1 times larger than the size observed in mice (85.6 μm) (Oberheim et al., 2009). Similarly, the extension of astrocyte branching domains in the human gray matter exhibits characteristic domain diameters about 2.6 times larger (142.6 vs. 56 μm), corresponding to a 16.5-fold greater occupied volume. The larger size of human astrocytes can be attributed to their more elongated and complex arborization, featuring 10 times more primary processes (37.5 vs. 3.8) that are, on average, 2.6-fold longer than mice (Oberheim et al., 2009). This increased complexity potentially reflects unique functional principles governing the organization of neural circuits in the human brain. Notably, a single astrocyte in the human brain can interact with millions of synapses, which is at least 10 times more than in mice (Bushong et al., 2002; Oberheim et al., 2009).

Human astrocytes can be classified into four main categories based on their morphology. These categories include interlaminar astrocytes found in Layer I, protoplasmic astrocytes present in Layers I to VI, varicose projection astrocytes located in deep Layers V to VI, and fibrous astrocytes found in the white matter (Figure 4) (Colombo and Reisin, 2004; Oberheim et al., 2009; Shapson-Coe et al., 2021). Among these subtypes, varicose projection and interlaminar astrocytes seem bearing uniquely hominid features pinpointing to the existence of unique neuron-astrocyte circuits in Hominidae (Oberheim et al., 2012). Varicose projection astrocytes have indeed only been observed in humans and higher primates, differently from interlaminar cells, which could also be found in the murine cortex (Degl'Innocenti and Dell'Anno, 2023). Interlaminar astrocytes in humans and higher primates however show increased complexity in terms of higher number of branches, with multiple processes infiltrating deeper cortical layers as opposed to the rudimental anatomy of their murine counterpart which are instead generally contained within their layer of origin (Colombo and Reisin, 2004; Falcone et al., 2021). Furthermore, the positional arrangement of astrocytes in the human cortex is likely to have distinct characteristics. Ultra-resolution microscopy of nonepileptic cortical tissue obtained from MTLE patients supports this notion revealing densely packed aggregates of intermingling protoplasmic astrocytes throughout all cortical layers (Shapson-Coe et al., 2021). This finding contrasts sharply with the observation that astrocyte domains are minimally overlapping in the mouse brain instead (Bushong et al., 2002; Ogata and Kosaka, 2002; Abdeladim et al., 2019).

Our understanding of the biophysical basis for the morphological specialization of human neuron-astrocyte circuits is still in its infancy. The notable size difference between human and rodent astrocytes explains the larger capacitance of the former (44.7 pF vs. ~10 pF−25 pF) (Bordey and Sontheimer, 1998; Amzica and Neckelmann, 1999). However, the membrane resistance of human astrocytes can range from values as low as 4.5 ± 3 MΩ (Bedner et al., 2020), similar to that observed in rodents, to 64-fold higher values (288 MΩ) (Bordey and Sontheimer, 1998). This suggests that the molecular composition of human astrocyte's membranes could differ substantially from that of mice. There may also be significant differences in intracellular properties. One crucial intracellular signal in astrocyte physiology is calcium (Ca²⁺), which plays a role in ion homeostasis, synaptic activity sensing, and its modulation (Navarrete et al., 2012; De Pittà et al., 2015; Semyanov et al., 2020). Interestingly, several genes associated with Ca²⁺ homeostasis, namely RYR3, MRVI1, and RGN, were found to be more abundant in human astrocytes than in mice (Zhang et al., 2016). Furthermore, the average propagation velocity of Ca²⁺ signals through astrocyte networks is considerably higher in humans (43.4 μm/s) compared to their murine counterparts (8.6 μm/s) (Oberheim et al., 2009).

Another noteworthy example is the co-expression of GDH1 and GDH2 genes in human astrocytes but not in rodents (Zhang et al., 2016). These genes encode glutamate dehydrogenase, which converts glutamate into α-ketoglutarate as the first step in the GABA/glutamate-glutamine cycle underpinning neurotransmitter recycling at central inhibitory and excitatory synapses (McKenna et al., 2016). Accordingly, it was shown that GDH1 and GDH2 co-expression enables human astrocytes to better withstand increased metabolic demands (Nissen et al., 2017), such as those arguably arising from their extensive synaptic interactions. Altogether, these findings indicate that human astrocyte diversity involves unique molecular characteristics, such as distinct Ca²⁺ signaling and enhanced metabolic capabilities, which may contribute to the functional complexity of the human brain.

Recent advancements in transcriptomics have provided unparalleled insights into the molecular characteristics of human astrocytes. Genetically, human astrocytes exhibit a distinct set of expressed genes that are either absent or significantly reduced in rodents (Miller et al., 2010; Zhang et al., 2016). Some of these genes' expression strongly correlates with the morphological features observed in human astrocytes. One such gene is PMP2, which encodes a cytoplasmic fatty-acid binding protein that appears to contribute to the larger size of human astrocytes (Kelley et al., 2018). Another example is CD44, which encodes a cell-surface glycoprotein involved in cell-cell interactions, cell adhesion, and migration. CD44 is prominently expressed by varicose projection and intralaminar astrocytes, potentially explaining their morphological signature characterized by long unbranched processes that extend across adjacent astrocytes (Sosunov et al., 2014). Human astrocytes' spatial location and transcriptional identity are determined by their embryonic site of origin, independent of neurons (de Majo et al., 2020). Interestingly, cultures of human embryonic stem cells can be patterned to generate astrocytes with specific regional identities even without inductive cues from neurons (Krencik et al., 2011). This could be explained by the existence of human-specific astrocyte progenitors that might account for uniquely human astrocyte populations (Hansen et al., 2010; Pollen et al., 2015), although the mechanisms of differentiation of astrocyte lineages from single progenitors remain to be explored (Beattie and Hippenmeyer, 2017; Ostrem et al., 2017). On the other hand, astrocytes, in the presence of multiple extracellular signals originating from the surrounding neuropil, can undergo a wide range of intracellular and transcriptional changes that can influence or maintain their heterogeneity (Haim and Rowitch, 2017). Correlational analysis of RNA-seq data from numerous post-mortem human brain specimens supports this scenario, revealing that, in addition to the presence of unique genes, astrocyte transcripts co-vary with the cellular composition of the brain area they live in (Kelley et al., 2018).

Besides anatomic distribution, age and pathology are crucial in determining astrocyte transcriptional and functional identity. In line with the observed limited response of the white matter to injury in fetal brains, human fetal astrocytes exhibit significantly lower expression levels of pro- and anti-inflammatory microRNAs (MiRs) compared to adult human astrocytes (Rao et al., 2016). Similarly, MiR-124, the regulator of the type 2 excitatory amino acid transporter (EAAT2), shows low or undetectable levels in fetal astrocytes but is abundant in adult astrocytes (Fine et al., 1996; Liang et al., 2002; Rao et al., 2016). These findings align with the understanding that EAAT2 is predominantly functionally expressed by mature cells (Sloan et al., 2017). The onset of pathology can elicit a wide range of molecular, transcriptional, and morphological changes in astrocytes, collectively referred to as reactive astrogliosis (Sofroniew, 2015; Escartin et al., 2021). However, the extent to which astrocyte functional diversity is maintained under pathological conditions remains unclear, as both toxic gains and loss of function have been proposed (Liddelow and Sofroniew, 2019). For instance, the decreased functional expression of the astrocytic enzyme glutamine synthetase (GS) in patients with MTLE has long been associated with elevated extracellular glutamate levels in this condition. The loss of GS activity slows down the conversion of glutamate to glutamine in astrocytes, contributing to the extracellular accumulation of glutamate (Eid et al., 2004). However, recent experiments have revealed that this loss of GS expression may not correlate with astrocyte reactivity. Instead, it may result from subcellular down-regulation or redistribution of GS while still maintaining detectable regional GS expression (Papageorgiou et al., 2018). This evidence supports the idea that, beyond inherited transcriptional factors, human astrocyte identity and functional specialization is ultimately tightly regulated by a diverse set of intrinsic and extrinsic factors within the specific neural circuits they inhabit (Belin et al., 1997; Eid et al., 2013; Haim and Rowitch, 2017).