Scaled complexity of mammalian astrocytes: insights from mouse and macaque

Abstract

Astrocytes intricately weave within the neuropil, giving rise to characteristic bushy morphologies. Pioneering studies suggested that primate astrocytes are more complex due to increased branch numbers and territory size compared to rodent counterparts. However, there has been no comprehensive comparison of astrocyte morphology across species. We employed several techniques to investigate astrocyte morphology and directly compared them between mice and rhesus macaques in cortical and subcortical regions. We assessed astrocyte density, territory size, branching structure, fine morphological complexity, and interactions with neuronal synapses using a combination of techniques including immunohistochemistry, AAV-mediated transduction of astrocytes, diOlistics, confocal imaging, and electron microscopy. We found significant morphological similarities between primate and rodent astrocytes, suggesting that astrocyte structure has scaled with evolution. Our findings show that primate astrocytes are larger and more numerous than those in rodents but contest the view that primate astrocytes are morphologically more complex.

Graphical Abstract

We used a multi-technique approach (including immunohistochemistry, AAV-mediated transduction of astrocytes, diOlistics, confocal imaging, and electron microscopy) to label astrocytes in mice and rhesus macaques and conducted a comprehensive morphological analysis.

INTRODUCTION

Astrocytes have a complex architecture which allows them to directly interact with synapses, blood vessels, and other cells. Though astrocytes were thoughtfully described in the early studies of Rudolph Virchow and Ramon y Cajal, these cells were largely ignored and thought of as mere support for the better-studied neurons (Cajal, 1913; R. Virchow, 1846; Santiago Ramon y Cajal, 1911; Virchow, 1856). Only over the last few decades has interest in astrocytes grown substantially, resulting in a multitude of new tools to interrogate their function and characterize their morphology. Advancements such as chemogenetics and optogenetics combined with astrocyte-targeting transgenic mouse lines or adeno-associated virus (AAV) promoters allow for direct manipulation of astrocyte function (Adamsky et al., 2018; Mederos et al., 2019; Nagai et al., 2019; Yu et al., 2020). Techniques such as 3D electron microscopy have allowed for reconstruction of astrocyte processes and their interactions with neuronal synapses and blood vessels at fine resolution (Aten et al., 2022; Salmon et al., 2023). Many of these approaches are commonly used in mice, but rarely in other species (Colombo et al., 1997), which limits our knowledge of astrocyte morphology in primates.

Previous studies described significant differences in morphology between rodent and primate astrocytes (Colombo et al. 1995; Oberheim et al. 2006, 2009; Falcone et al., 2019).

While these studies present data supporting unique morphological features such as longer interlaminar astrocyte processes and the presence of varicosities on a subset of cortical hominid astrocytes (Colombo, 2001, 1996; Colombo et al., 1999, 1995), the underlying functional contributions, as well as prevalence of such features remain unknown. Astrocyte function in vivo has been primarily investigated in rodents, and thus any links between morphology and function are limited to morphological features that exist in rodent species. In comparing common morphological features, evidence demonstrating increased complexity of astrocytes in primate species is lacking. It has been established that primate astrocytes display larger territory areas (Oberheim et al., 2009). However, details regarding the cellular density, primary branching structure, fine process size and density, and interactions with neurons and synapses remain largely unexplored. Further, morphological complexity and cellular density can have significant functional implications, as they may determine the number of synapses an astrocyte can regulate. Despite the lack of data directly analyzing astrocyte complexity across species, papers commonly cite primate astrocytes as more complex than rodent (Allen and Eroglu, 2017; Vasile et al., 2017; Verkhratsky et al., 2018).

Although some studies have analyzed the morphology of astrocytes in human neurological disease or in non-human primates (Colombo et al., 2005, 2005, 1999; Reisin et al., 2004), an overwhelming majority of research is conducted using rodents as models for human disease. Given the increased interest in glial contributions to disease states, detailed characterization of species-specific astrocyte morphology and function must be undertaken. Such investigation will clarify the potential limitations or advantages of animal models in the context of glial contributions to disease.

Here, we used a series of techniques to systematically assess astrocyte morphology in mice and rhesus macaques across cortical and subcortical brain regions. Filtering down from overall astrocyte density to astrocyte territory size, individual astrocyte primary branching structure, and the fine process or perisynaptic astrocyte process (PAP) morphology, the data paint a comprehensive picture of astrocyte morphology across species. We show that while primate astrocytes are larger than their rodent counterparts, they display a comparable number of GFAP+ branches per unit area, and less densely packed GfaABC1D+/DiI processes in volume. Furthermore, we demonstrate that the degree of astrocyte-neuron interaction is scaled across species, despite previous claims. These findings offer novel insights on astrocyte morphology and complexity in rodent and primate species that counter long-standing views in the field. Our data suggest astrocyte complexity scales with size to optimally cover the neuropil in order to allow them to perform their essential physiological and pathophysiological roles across species.

METHODS

Experimental model and animal details

All procedures were approved by the Animal Care and Use Committee of Emory University and performed in accordance with the Guide for the Care and Use of Laboratory Animals (NRC, 2010) and the U.S. Public Health Service Policy on the Humane Care and Use of Laboratory Animals (revised 2015).

Rhesus Macaques

All rhesus macaques were sourced from the Emory National Primate Research Center (ENPRC). The rhesus macaques were maintained in a controlled environment with a 12-hour light/dark cycle, had free access to water, and were fed twice daily. Eleven rhesus macaques were used in this study (Table 1): Three animals received AAV injections (see below for details), tissue from two animals was used for diOlistic bombardment (DiI) and glial fibrillary acidic protein (GFAP) immunostaining experiments, and tissue from six animals was used in experiments to quantify cell density. The animals presented no clinical concerns, with the exception of the two rhesus macaques involved in DiI and GFAP studies. One rhesus macaque presented with heart murmur/possible hypertrophic cardiomyopathy (ID no. 4) and another rhesus macaque (ID no. 5) with chronic low weight without specific cause. As explained below, the results obtained from these animals are consistent with those obtained in other rhesus macaques in this study. Animals were included based on availability and ages correspond to adolescent (2–4 years), young adult (4–8 years), and mature adult (8–15 years) stages (Frings et al., 2015).

Table 1: Animals.

GfaABC1D indicates animals that received AAV5 injections. DiI and GFAP indicates animals whose tissue was subjected to diOlistic bombardment and GFAP staining. Cell counts indicates animals whose tissue was stained for astrocyte and neuronal markers. See main text for description of experiments.

Subject ID	Species	Sex	Age at perfusion	Experiment
1	Rhesus macaque	Male	4 years	AAV5-GfaABC1D
2	Rhesus macaque	Female	4 years	AAV5-GfaABC1D
3	Rhesus macaque	Male	5 years	AAV5-GfaABC1D
4	Rhesus macaque	Male	6 years	DiI, GFAP
5	Rhesus macaque	Female	4 years	DiI, GFAP
6	Rhesus macaque	Female	9 years	Cell counts
7	Rhesus macaque	Male	11 years	Cell counts
8	Rhesus macaque	Male	3 years	Cell counts
9	Rhesus macaque	Female	12 years	Cell counts
10	Rhesus macaque	Female	6 years	Cell counts
11	Rhesus macaque	Male	5 years	Cell counts
1	Mouse	Female	9 months	AAV5-GfaABC1D
2	Mouse	Female	9 months	AAV5-GfaABC1D
3	Mouse	Female	9 months	AAV5-GfaABC1D
4	Mouse	Female	8 months	AAV5-GfaABC1D
5	Mouse	Female	8 months	AAV5-GfaABC1D
6	Mouse	Male	6 months	DiI, GFAP
7	Mouse	Male	6 months	DiI, GFAP
8	Mouse	Male	7 months	DiI, GFAP
9	Mouse	Male	7 months	DiI, GFAP
10	Mouse	Male	7 months	DiI, GFAP
11	Mouse	Male	7 months	DiI, GFAP
12	Mouse	Male	7 months	DiI, GFAP
13	Mouse	Male	5 months	Cell counts
14	Mouse	Male	5 months	Cell counts
15	Mouse	Male	5 months	Cell counts
16	Mouse	Male	3 months	Cell counts
17	Mouse	Female	3 months	Cell counts
18	Mouse	Female	3 months	Cell counts

Mice

Mice (all C57BL/6J) were fed regular chow and water ad libitum and maintained at a 12-hour light/dark cycle. Eighteen mice were included in this study: 5 received AAV injections, 7 for diI and GFAP experiments, and 6 for cell counts. Animals were included based on availability and ages correspond to young (3–6 months) to mature (6–10 months) adult stages (Radulescu et al., 2021). The number of mice included in each experiment is summarized in Table 1 and listed in the figure legend.

AAV injections in mice and rhesus macaques

Prior to enrollment in the study, blood samples from all rhesus macaques were tested by the EPC Viral Core and confirmed sero-negativity for AAV5 antibodies. In two of the rhesus macaques and all mice, the pZac2.1 vector backbone and AAV5 serotype was used to express the fluorescent tag TdTomato under the GfaABC1D promoter (Addgene viral prep # 44332-AAV5 ; http://n2t.net/addgene:44332; RRID:Addgene_44332)(Shigetomi et al., 2013). In the third rhesus macaque, the pZac2.1 vector backbone and AAV5 serotype was used to express the membrane-tethered fluorescent tag GFP under the GfaABC1D promoter (Addgene viral prep # 105598-AAV5; RRID:Addgene_105598)(Shigetomi et al., 2013). The virus solutions were stored at −80°C and thawed the day of the injection. The details regarding injection procedure for two rhesus macaques have been previously described(Heffernan et al., 2022). Briefly, rhesus macaques underwent chamber placement surgery or intracerebral injections of the virus. They were initially sedated with ketamine (10 mg/kg) and intubated for isoflurane anesthesia which was maintained at 1–3% throughout surgery. Post-operatively, rhesus macaques were treated with buprenorphine (0.03 mg/kg) and banamine (1 mg/kg) as analgesics, and Rocephin (25 mg/kg) to prevent infection. Rhesus macaques that underwent chamber placement surgery then received intracerebral injections of the virus guided by electrophysiology. Rhesus macaques were perfused 4–9 weeks following injection.

Mice were anesthetized with 3–4% isoflurane and maintained at 1.5–2.5% during surgery. Mice were given buprenorphine ER (1mg/kg) and 0.2% lidocaine (4mg/kg) and head fixed on a stereotaxic frame (Kopf). Craniotomies were performed bilaterally above primary motor cortex. In each hemisphere, 0.5 μl of virus was deposited in motor cortex (in mm from Bregma: AP +1.1, ML ± 1.5, DV 0.75) and striatum (AP +1.1, ML ± 1.5, DV 2.75) using a nanoinjector (Nanoinject II, Drummond Scientific). Mice were perfused 3–4 weeks following surgery.

General methods related to immunohistochemistry

For all methods below except for DiI labeling and GFAP staining, rhesus macaques and mice were transcardially perfused with Ringer’s solution followed by a fixative solution containing 4% PFA and 0.1% glutaraldehyde. Brains were post-fixed in 4% PFA for 24–48 hours at 4°C and then sectioned coronally at 60 μm on a vibratome (Leica). Prior to staining, all sections were first treated with 1% sodium borohydride, to reduce free aldehydes and Schiff bases resulting from glutaraldehyde fixation and rinsed in phosphate buffered saline (PBS; 0.01M, pH 7.4). For DiI labeling and GFAP staining, rhesus macaques and mice were transcardially perfused with Ringer’s solution. Whole mouse brains and 4 mm blocks of rhesus macaque brains were immersion-fixed in 4% PFA for 1 hour at room temperature and then sectioned coronally at 300 μm on a vibratome (Leica). Sections were stored in an anti-freeze solution at −20°C until further processing. All details for primary and secondary antibodies are listed in Table 2.

Table 2.

Antibodies.

Antibody	Source	Identifier
mCherry mouse	Takara	Cat. No. 632543/ RRID: AB_2307319
mCherry rabbit	Abcam	Cat. No. AB167453/ RRID: AB_2571870
GFP chicken	Millipore	Cat. No. 06-896/ RRID: AB_11214044
NeuN rabbit	Millipore	Cat. No. ABN78/ RRID: AB_10807945
Sox9 rabbit	Millipore	Cat. No. AB5535/ RRID: AB_223976 1
GFAP rabbit	Abcam	Cat. No. AB7260/ RRID: AB_305808
vgat11 mouse	R&D	Cat. No. MAB6847/ RRID: AB_2814814
Fab donkey anti-mouse IgG	Jackson Lab	Cat. No. 715-007-003/ RRID: AB_2307338
DAPI	Invitrogen	Cat. No. D-1306/ RRID: AB_2629482
Donkey anti-mouse FITC	Jackson Lab	Cat. No. 715-095-150/ RRID: AB_2340792
Donkey anti-mouse RRX	Jackson Lab	Cat. No. 715-295-150/ RRID: AB_2340831
Donkey anti-rabbit FITC	Jackson Lab	Cat. No. 711-095-152/ RRID: AB_2315776
Donkey anti-rabbit RRX	Jackson Lab	Cat. No. 711-295-152/ RRID: AB_2340613
Donkey anti-chicken FITC	Jackson Lab	Cat. No. 703-095-155/ RRID: AB_2340356
Biotinylated goat anti-rabbit	Vector	Cat. No. BA1000/ RRID: AB_2313606
Biotinylated horse anti-mouse	Vector	Cat. No. BA2000/ RRID: AB_2313581
Biotinylated goat anti-chicken	Vector	Cat. No. BA9010/ RRID: AB_2336114

Antibody characterization

Both mCherry antibodies (Takara and Abcam, respectively) were protein A-purified and raised against mouse (Takara) or rabbit (Abcam) mCherry and recognized a single band of approximately 30–35 kDa (Takara) or 30 kDa (Abcam) on western blots of transfected HEK 293 cells (manufacturers’ datasheets). The antibodies were used to amplify endogenous expression from viral injection. Both stained a pattern of cellular morphology and distribution in the mouse and rhesus macaque brain that was consistent with endogenous fluorescence observed in unstained tissue from the same animals (that received AAV5-pZac2.1-gfaABC1D-tdTomato injection). No staining was observed in animals lacking mCherry expression (uninjected animals).

The GFP antibody was purified and raised against chicken GFP and recognizes a single band of approximately 30 kDa on western blots of rat brain lysate containing purified GFP (manufacturer’s datasheet). The antibody was used to amplify endogenous expression from viral injection. It stained a pattern of cellular morphology and distribution in the mouse and rhesus macaque brain that was consistent with endogenous fluorescence observed in unstained tissue from the same animals (that received AAV5-pZac2.1-gfaABC1D.PI.Lck-GFP.SV40 injection). No staining was observed in animals lacking GFP expression (uninjected animals).

The NeuN antibody was purified and raised against rabbit NeuN and recognizes a band of approximately 48/42 kDa on western blots of mouse E16 brain lysate (manufacturer’s datasheet). The immunogen is the GST-tagged recombinant fragment corresponding to the first 97 amino acids from the N-terminal region of murine NeuN. It stained an expected pattern of cellular morphology and distribution in the mouse and rhesus macaque brain. No staining was observed in white matter.

The Sox9 antibody was purified and raised against rabbit Sox9 and recognizes a band of 56–65 kDa in HepG2 cell lysate (manufacturer’s datasheet). The immunogen is the KLH-conjugated linear peptide corresponding to the C-terminal sequence of human Sox9. Ablation of Sox9 in the developing retina resulting in no staining with this Sox9 antibody, supporting antibody specificity (Poché et al., 2008).

The GFAP antibody was purified and raised against rabbit GFAP and recognizes a band of 55 kDa and a GFAP derived 48 kDa band. The immunogen is a recombinant full-length protein corresponding to human GFAP. Immunohistochemical staining with the antibody shows patterns consistent with the increased RNA expression of GFAP (assessed by RT-PCR) and increased protein expression of GFAP (assessed by western blot) over developmental time points in the mouse brain, supporting antibody specificity (Yoon et al., 2017).

The vesicular GABA transporter 1 (vgat11) antibody was purified and raised against mouse vgat11 and detects human vgat11 in direct ELISAs (manufacturer’s datasheet). The immunogen is E. coli-derived recombinant human VIAAT/SLC32A1. The observed pattern of staining was comparable to previous reports (Swain et al., 2020).

For studies involving mouse primary antibodies on mouse tissue, we utilized Fab blocking of endogenous mouse IgG. We performed control experiments in which the primary antibody was omitted, and no staining was observed.

Cell counts: neurons and astrocytes

For each animal, one 60 μm section was selected. All sections were located approximately within −0.1 to −0.46 mm from bregma in mouse and 16.05 to 13.8 from the interaural line in rhesus macaque according to the Paxinos brain atlases (Paxinos et al., 2000; Paxinos and Franklin, 2019). To prevent non-specific binding and permeabilize the tissue, sections were incubated in a solution of 1% donkey serum, 1% bovine serum albumin (BSA), and 0.3% Triton X-100 in PBS. Sections were then incubated overnight in the blocking/permeabilization solution containing antibodies for neurons (NeuN 1:10000) and astrocytes (Sox9 1:5000). Table 2 contained details of antibodies used in this study. Binding sites were revealed using a fluorophore-conjugated secondary antibody (all 1:100). Sections were incubated in the secondary solution overnight and rinsed in PBS. Sections were incubated in PBS containing DAPI (5μg/mL) for 10 minutes. Sections were then mounted with Vectashield (Vector Laboratories) on slides and coverslipped.

Slides were imaged using a confocal microscope (Leica DM5500B) equipped with a CCD camera (Orca R2; Hamamatsu). Regions of interest (primary motor cortex layers 2/3 and 5, globus pallidus (external for rhesus macaque), and dorsal putamen (rhesus macaque) or dorsolateral striatum (mouse) were identified by pattern of DAPI staining and Z-stacks were obtained at 20x with a 1 μm step size. From each z-stack, maximum intensity projections of 11 images (10 μm) were generated. The 10 μm thickness of the stack was selected to avoid variability resulting from differences in staining intensity throughout the section. Neurons and astrocytes were counted using the CellCounter plugin in ImageJ (Schindelin et al., 2012). Total cellular densities were estimated by automatically counting DAPI+ nuclei using CellProfiler (Stirling et al., 2021).

GFAP staining to investigate primary branching

Stored 300 μm thick sections from mouse and rhesus macaque brains were stained for glial fibrillary acidic protein (GFAP). Sections were incubated in a solution of 1% donkey serum, 1% BSA, and 1% Triton X-100 in PBS. Sections were then incubated for 3 days at 4°C in the blocking/permeabilization solution (with the Triton X-100 concentration was lowered to 0.3%) containing the GFAP antibody (1:5000). Binding sites were revealed using a fluorophore-conjugated secondary antibody (all 1:100). Sections were incubated in the secondary solution for 3 days at 4°C and rinsed in PBS. Sections were incubated in PBS containing DAPI (5μg/mL) for 10 minutes. Sections were then mounted with Fluoromount-G (Southern Biotech) on slides with 2 stacked 0.12mm deep coverslip spacers (Grace Bio-Labs) to reduce tissue compression.

GFAP-labeled astrocytes were imaged using a Leica SP8 confocal microscope. Z-stacks were obtained with a 40x (NA 1.3) objective, 1024×1024 scan format, 2x optical zoom, and 0.35 μm z-steps. Stacks were converted to maximum intensity z-projections in ImageJ. The image was thresholded using the automated “Default” selection and binarized. An outlined area overlay of the cell territory was generated by using the “Analyze Particles” tool with a minimum particle size of 5 μm². The area outside of the territory was then “cleared” and the binarized image was subjected to Sholl analysis with the Simple Neurite Tracer plugin. The center of the soma was marked as the starting radius with a step size of 4 μm.

Territory size and neuropil infiltration volume (NIV) in astrocytes transduced with AAV5-GfaABC1D

Animals received viral injections and were perfused as described above. Sections were stained following the same procedure above described for cell counts using primary antibodies against either mCherry (1:1000 mouse or 1:2500 rabbit) or GFP (1:2000), revealed with the appropriate fluorophore-conjugated secondary antibody, and counterstained with DAPI. Sections were then mounted with Vectashield (Vector Laboratories) on slides and coverslipped.

Astrocytes were imaged using a microscope (Leica DM5500B) equipped with a spinning disk confocal (VisiTech International) and CCD camera (Orca R2; Hamamatsu). Z-stacks were obtained 63x with a 0.1 μm step size with a resulting x-y pixel size of 0.105 by 0.105 μm/pixel. Astrocytes were selected for imaging if they 1) were well-isolated from other labeled cells 2) displayed adequate signal to noise ratio 3) resided with layer 5 of primary motor cortex or post-anterior commissural putamen in rhesus macaques (as determined by distance to brain surface and pattern of DAPI staining) or dorsolateral striatum in mice and 4) had its nucleus visible within the slice. Images were analyzed in ImageJ. Stacks of 89–537 images (8.9–53.7 μm) were converted to maximum intensity z-projections. The image was thresholded and binarized. Threshold parameters varied by cell due to differences in the level of tag protein expression and background. An outlined area overlay of the cell territory was generated by using the “Analyze Particles” tool with a minimum particle size of 5 μm². The overlay generated area and Feret’s maximum diameter (the longest edge to edge distance of the overlay).

The raw z-stacks of cortical astrocytes were also loaded into Imaris (v 9.8.0). For each astrocyte, three ROIs measuring 10 μm × 10 μm × 10 μm were created. All ROIs were placed to avoid the soma and major branches. The surface function of Imaris was used to render 3D reconstructions of the astrocyte branches within the ROI boundary. The volumes of each ROI, termed neuropil infiltration volume (NIV) were recorded(Baldwin et al., 2021; Stogsdill et al., 2017). Three ROIs were averaged for each cell, resulting in 90 ROIs and 30 NIV averages for mouse and 42 ROIs and 14 NIV averages for rhesus macaque. One rhesus macaque which received a viral injection with the membrane tethering Lck sequence (GfaABC1d-Lck-GFP) was not included in NIV analysis.

Fractal dimension and NIV in astrocytes labeled by DiI

A diOlistic device was constructed using pipe fittings, a pressure gauge, solenoid, relay, push button, and battery-operated air pump (see Heffernan and Galvan, 2023 for complete details). To prepare carriers of DiI-coated gold particles, 50 mg of gold particles (0.8–1.3 μm in diameter) were mixed with a 1.5 mg DiI/100% ethanol solution. The solution was spread even on a coverslip, allowed to dry, and scraped into a tube containing distilled water. The tube was sonicated with intermittent vortexing for 15 minutes. The solution was then pipetted onto parafilm-covered #8 washers and allowed to dry. As described in Heffernan and Galvan, 2023, the parameters used for the diOlistic device included 100 psi of pressure, 3 cm distance from top of brain section, and 20 μm filter paper. Sections were “shot,” rinsed with PBS and left to incubate for 24 hours at room temperature. Sections were incubated in PBS containing DAPI (5μg/mL) for 10 minutes. Sections were then mounted with Fluoromount-G (Southern Biotech) on slides with 2 stacked 0.12mm deep coverslip spacers (Grace Bio-Labs) to reduce tissue compression.

DiI-labeled astrocytes imaged using a Leica SP8 confocal microscope. Z-stacks were obtained with a 40x (NA 1.3) objective, 1024×1024 scan format, 2x optical zoom, and 0.35 μm z-steps resulting in an x-y pixel size of 0.13 μm × 0.13 μm. Stacks were converted to maximum intensity z-projections in ImageJ. The image was thresholded using the automated “Huang” method and binarized. An outlined area overlay of the cell territory was generated by using the “Analyze Particles” tool with a minimum particle size of 5 μm². The overlay generated area and Feret’s maximum diameter (the longest edge to edge distance of the overlay). The binarized image was additionally subjected to box counting via the ImageJ FracLac plugin. The fractal dimension is defined as the slope of $\frac{\text{log}\left(N\right)}{\text{log}\left(\text{r}\right)}$ where N is the number of boxes and r is the inverse of box size. The raw z-stacks were also loaded into Imaris. The NIV analysis is described above for analysis of virally labeled astrocytes. Three ROIs were averaged for each cell, resulting in 30 ROIs and 10 NIV averages for mouse and 36 ROIs and 12 NIV averages for rhesus macaque.

Density of inhibitory terminals within astrocyte territories

To reduce autofluorescence in the tissue, all sections were subjected to “photobleaching” prior to incubations. The sections were placed in 6-well transparent plates containing PBS in a refrigerator (4°C) equipped with broad-spectrum white lights. The well-plates were placed on a rack directly above the lights and remained exposed to light for 4 days. Following photobleaching, sections were stained and coverslipped using the same procedure described for cell counts and virally labeled astrocytes. Transduced astrocytes were stained for mCherry (1:1000 rabbit) and GABAergic synaptic terminals for vgat1 (1:7500).

Astrocytes were imaged using a Leica SP8 confocal microscope. Z-stacks of 5.4 μm were obtained with a 63x (NA 1.4) objective, 1024×1024 scan format, 1.8x optical zoom, and 0.3 μm z-steps resulting in an x-y pixel size of 95 nm × 95 nm. Laser settings were kept identical across all images. Stacks of 6 images (1.8 μm) were converted to maximum intensity z-projections in ImageJ, resulting in 3 maximum intensity z-projections per astrocyte. The channel containing the astrocyte was thresholded and binarized. An outlined area overlay of the cell territory was generated by using the “Analyze Particles” tool with a minimum particle size of 5 μm². The channel containing the vgat1 underwent background subtraction with a rolling ball radius of 50 pixels and was thresholded using the “Moments” method in ImageJ (Tsai, 1985). The number of terminals was then determined by using the “Analyze Particles” tool with a particle size range of 0.2–6 μm². The number of terminals within the astrocyte territory and area of the territory was recorded. To reduce signal and noise variability across samples, all samples were processed and imaged in the same batch.

Immunohistochemistry: mCherry/GFP immunoperoxidase for brightfield and electron microscopy

To verify the location of AAV injections, sections were incubated in a solution of 1% normal serum, 1% BSA, and 0.3% Triton X-100 in PBS. Sections were then incubated overnight in the blocking/permeabilization solution containing antibodies for mCherry (1:2500 mouse or 1:1000 rabbit) or GFP (1:2000). Binding sites were revealed using a biotinylated secondary antibody (see Table 2, all 1:200). Sections were incubated in the secondary solution for 90 minutes and rinsed in PBS. Sections were then incubated in an ABC solution (1:200, Vector Laboratories) containing 1% BSA and 0.3% Triton X-100 for 90 minutes. Followed by rinses in PBS and TRIS buffering solution (0.05M, pH 7.6), sections were incubated in TRIS buffer containing 0.025% 3,3’-diaminobenzidine tetrahydrochloride (DAB; Sigma), 10 mM imidazole, and 0.006% hydrogen peroxide for 10 minutes, immediately followed by PBS rinses. Following mounting, sections were dehydrated in a graded series of alcohol and toluene, then coverslipped. Slides were scanned with an Aperio Scanscope CS system (Leica).

For electron microscope analysis, sections were first treated with 1% sodium borohydride as above, before being placed in a cryoprotectant solution, frozen at −80°C to maximize antibody penetration and washed in serial dilutions of cryoprotectant in PBS. After blocking with 1% normal serum and 1% bovine serum albumin in PBS, the sections were incubated for two days at 4°C in the primary antibody. The avidin-biotin complex (ABC) method was used to amplify signal. The sections were incubated in biotinylated secondary antibodies for 2 hours (see Table 2, all 1:200), and in the ABC solution (1:100; Vectastain Standard kit, Vector) for 90 minutes. Sections were rinsed in PBS and TRIS buffer (0.05 M, pH 7.6) before being placed in DAB solution for 10 minutes. Next, sections were rinsed in PBS and kept at 4°C overnight. The sections were then placed in PB for 10 min to desalt and postfixed in osmium tetroxide (1% in PB). They were washed in PB and dehydrated in a series of ethanol and propylene oxide. The 70% ethanol solution contained 1% uranyl acetate. The sections were embedded in resin overnight and placed on slides before being placed in the oven for 48 hours at 60°C. Blocks from primary motor cortex were cut from each animal and glued on top of resin blocks. After facing, serial ultrathin sections were cut on an ultramicrotome (Leica Ultracut T2, Leica, Nussloch, Germany). Sixty nanometer thick sections from the surface of the block were collected on copper single slot grids and stained with lead citrate for 5 minutes.

Stained grids were examined on an electron microscope (EM Model 1011; Jeol, Peabody, MA, USA) coupled with a CCD camera (Gatan Model 785; Warrendale, PA, USA). Each section was scanned at 5,000x to identify immunoperoxidase labeling. Peroxidase-labeled elements were imaged at 40,000x. Using ImageJ, elements were identified according to ultrastructural features.(Peters et al., 1991) Images containing peroxidase-labeled astrocytes and synapses were categorized based on their spatial relationship with labeled astrocyte processes. Astrocyte processes which touched the axonal terminal were considered “pre” contacts, astrocytes which touched a dendrite or dendritic spine were considered “post” contacts. If the astrocyte touched both the pre- and post-synaptic elements, it was considered a “cleft” contact. Additionally, the widths of cleft-associated astrocyte processes were measured in ImageJ. At the cleft, the astrocyte process was measured at an angle approximately parallel to the visible synapse. If the membrane integrity of the astrocytic process making cleft contact was compromised, the process was not included in the width measurements. We additionally classified synapses into symmetric, asymmetric based on ultrastructural properties (Peters et al., 1991). Those synapses that we could not identify were termed ‘unknown’.

Quantification and statistical analysis

Data were first tested for normality using the D’Agostino-Pearson test. If the data was normally distributed, two group comparisons were made using two-tailed unpaired t-tests. Welch’s correction was implemented for any analyses in which there were significantly unequal variances between groups. For data that was not normally distributed, two group comparisons were made using unpaired Mann-Whitney tests. For analyses with more than two groups, multiple unpaired t-tests with Holm-Sidak correction for multiple comparisons or Chi-Square/Fisher’s exact were run. To evaluate the relationship between two variables, linear regression was performed. The results of statistical analysis, n, and p-values are included in figure legends. The results of statistical analysis, n, and p-values are included in figure legends. N is defined as number of animals or cells depending on analysis and is described in the figure legend. Statistics were run using GraphPad Prism 9. Data are represented as mean ± SD. In all figure plots, p-values are represented by asterisk(s): * for p < 0.05; ** for p < 0.01; *** for p < 0.001; **** for p < 0.0001. All statistical test results as well as means and SD values shown in plots are provided in a supplementary file.

RESULTS

Strategy

We studied astrocytes from 4 functionally connected regions of the adult rhesus macaque and mouse brain: putamen/dorsolateral striatum (STR), external globus pallidus/globus pallidus (GPe), and primary motor cortex (MCX). To assess potential cortical layer-specific differences, layers 2/3 and 5 were examined. We first investigated astrocyte density, neuron density, and astrocyte-to-neuron ratio in all 4 regions. We then investigated astrocyte primary branching structure using glial fibrillary acidic protein (GFAP) in layer 5 of primary motor cortex. To further investigate astrocyte morphology, we utilized an astrocyte-targeting adeno-associated virus (AAV) as well as diOlistics in striatum and cortex. Lastly, we characterized astrocyte ultrastructural relationships with synapses using 2D electron microscopy in primary motor cortex.

Astrocyte density is conserved across species and brain regions while neuron density varies considerably

We first examined neuron and astrocyte densities across different brain regions in mouse and rhesus macaque. Neurons and astrocytes were labeled by immunohistochemistry (IHC) for selective cell-type markers: NeuN (neurons), Sox9 (astrocytes)(Sun et al., 2017), and counterstained with DAPI (Figure 1A–B). Neuron densities varied by brain region in both species, though in all examined regions, densities were lower in rhesus macaques (STR: p=0.0022; GPe: p=0.0022; MCX L2/3: p=0.0022; MCX L5/6: p=0.0022; Multiple Mann-Whitney tests; Figure 1C). The increased neuronal density in mice also resulted in increased proportions of neurons relative to total cellular density (STR: p=0.0022; GPe: p=0.0022; MCX L2/3: p=0.0043; MCX L5/6: p=0.0022; Multiple Mann-Whitney tests; Figure 1D). Previous reports have found that neuronal density decreases with increased brain size which is consistent with our findings (Herculano-Houzel, 2014). Differences in astrocyte densities among species were region specific, with a lower density for rhesus macaques in GPe and MCX L2/3 (STR: p=0.2229; GPe: p=0.0022; MCX L2/3: p=0.0087; MCX L5/6: p=0.8506; Multiple Mann-Whitney tests; Figure 1E). Despite these differences, proportions of astrocytes relative to total cellular density were significantly different across most brain regions examined, with higher astrocyte proportions in rhesus macaques (STR: p=0.0022; GPe: p=0.5887; MCX L2/3: p=0.0022; MCX L5/6: p=0.0022; Multiple Mann-Whitney tests Figure 1F). Although total cellular densities were similar across brain regions within species, they were lower in rhesus macaque (p=0.0022 for all regions; Multiple Mann-Whitney tests; Figure 1G). Taken together, these data suggest that rodents display higher proportions of neurons, while primates display higher proportions of astrocytes relative to other cell types.

Figure 1. Astrocyte and neuron densities across the rhesus macaque and mouse brain.

(A) Representative example of sampled brain regions from mouse (left) and rhesus macaque (right). MCX, motor cortex; STR, striatum; GP, globus pallidus. Scalebars 2mm. (B) Representative examples of astrocyte (red) and neuron (green) labeling in striatum and motor cortex in mouse (left) and rhesus macaque (right), scalebars 20 μm. (C) Neuron density measurements (STR: p=0.0022; GPe: p=0.0022; MCX L2/3: p=0.0022; MCX L5/6: p=0.0022; Multiple Mann-Whitney tests). (D) Percent of total nuclei that are NeuN+ (STR: p=0.0022; GPe: p=0.0022; MCX L2/3: p=0.0043; MCX L5/6: p=0.0022; Multiple Mann-Whitney tests). (E) Astrocyte density measurements (3STR: p=0.2229; GPe: p=0.0022; MCX L2/3: p=0.0087; MCX L5/6: p=0.8506; Multiple Mann-Whitney tests). (F) Percent of total nuclei that are Sox9+ (STR: p=0.0022; GPe: p=0.5887; MCX L2/3: p=0.0022; MCX L5/6: p=0.0022; Multiple Mann-Whitney tests). (G) Total nuclei density assessed by DAPI staining (p=0.0022 for all regions; Multiple Mann-Whitney tests). All panels: inter-species comparisons assessed by multiple Mann-Whitney tests. All individual data are shown with lines representing means and error bars (panel G) as standard deviation (n_mouse = 6, n_{rhesus macaque} = 6).

Astrocyte primary branching scales with territory size

Individual astrocytes can interact with multiple neurons within their territory. The results examining neuron and astrocyte densities suggest that the territory of individual astrocytes in mice survey a greater number of neurons than in rhesus macaques. Next, we studied the gross and fine structure of astrocytes to examine how the morphology of these cells may relate to the cellular and chemical composition of their environment in each species. Subsequent analyses were conducted in primary motor cortex and/or striatum to address whether species differences in astrocyte morphology are present in a functionally connected area cortical and subcortical structures.

To assess the primary branching structure of astrocytes, we used GFAP immunohistochemistry in mouse and rhesus macaque primary motor cortex. We chose layer 5 to avoid layer 1 interlaminar astrocyte processes which traverse layers 2–4 in rhesus macaque.(Falcone et al., 2021b) We did not include striatal astrocytes in this analysis because the striatum does not express an abundance of GFAP (Chai et al., 2017). Using 2D analyses, we determined the area and maximum diameter of the GFAP-immunoreactive astrocyte territory (Figure 2A–C). When compared to mouse, rhesus macaque astrocytes displayed a 2x larger territory area (p<0.0001, t-test with Welch’s correction) and 1.4x larger diameter (p<0.0001 t-test) (Figure 2D–E). These data are consistent with previous studies (Oberheim et al., 2009). Sholl analysis revealed that rhesus macaque astrocytes displayed a higher number of GFAP+ branches (Figure 2F)(Sholl, 1953). We also quantified the number of primary branches for each astrocyte by counting those that arose from the soma (Figure 2G). Rhesus macaque astrocytes displayed 7 primary branches on average, versus 4 in mice (p<0.0001, t-test with Welch’s correction). However, when accounting for the species differences in territory area, the maximum number of GFAP+ Sholl intersections was not significantly different between species (Figure 2H). Further, the number of Sholl intersections increased with territory size within each species (Figure 2I).

Figure 2. Astrocyte primary branching with GFAP.

(A) Representative maximum intensity z-projection of GFAP+ astrocyte in layer 5 of mouse motor cortex. (B) Representative maximum intensity z-projection of GFAP+ astrocyte in layer 5 of rhesus macaque motor cortex. All scalebars 20 μm. (C) Steps of processing: binarization, territory outline, and Sholl analysis. (D-E) Territory area (p<0.0001, t-test) and Feret’s maximum diameter (p<0.0001, t-test) of mouse and rhesus macaque GFAP+ astrocytes (N_{rhesus macaque}= 2, N_cells= 21 N_mouse= 6, N_cells= 32). (F) Sholl analysis of individual astrocyte morphologies. Data are represented as mean ± SD. (G) Number of GFAP+ primary branches in individual astrocytes (p<0.0001, t-test). Data are represented as mean ± SD. (H) Complexity ratio: territory area/maximum number of intersections (p=0.0754, t-test). (I) Linear regressions of maximum number of intersections and area.

Using GFAP immunostaining, we also looked for the previously described and morphologically distinct subtype of primate astrocytes, the varicose-projection astrocyte (VPA). VPAs have been previously identified in humans and apes (Falcone et al., 2021a; Oberheim et al., 2009). We identified the presence of varicosities on GFAP+ astrocytes in cortical layer 5 of one rhesus macaque (Figure S1). We were unable to trace the process to the cell body of origin.

Primate astrocyte territory areas are larger with lower neuropil infiltration volumes

Although GFAP is widely used to identify astrocytes and assess morphology, it is not expressed in the finer branches and processes of astrocytes (Bushong et al., 2002). To assess species differences in fine morphological detail beyond primary branching structure, we injected AAV5-GfaABC1D-tdTomato or Lck-GFP (Figure S2). Since the promoter had not been used in primates before, we recently characterized the specificity and efficiency in the rhesus macaque. As described (Heffernan et al., 2022), the promoter has high specificity for astrocytes in the rhesus macaque brain with limited transduction in neurons and oligodendrocytes. The activity of the GfaABC1D promoter in mice has been previously well-described (Lee et al., 2008; Yu et al., 2020).

Following staining with antibodies against mCherry (to detect tdTomato) or GFP, astrocytes were identified and imaged. For subsequent analyses, astrocytes selected for imaging had to 1) be well-isolated from other labeled cells, 2) display adequate signal-to-noise ratio, 3) reside within layer 5 of primary motor cortex or putamen in rhesus macaques or dorsolateral striatum in mice, and 4) have their nuclei visible within the slice. Since brain sections were 60 μm thick, we did not acquire stacks containing entire astrocytes in the z-dimension. We verified that the image depth acquired in the z dimension did not determine astrocyte territory volume or Feret’s diameter (Figure S3C–D). The theoretical 3D shape of an astrocyte is approximately spherical with the soma in the center though this may vary by brain region, and thus, by ensuring that all measured astrocytes had a visible soma in the acquired stack, we effectively measured the largest possible territory in the x-y dimensions for each cell. In addition, we verified that the territory areas did not differ between the two fluorophores used (one rhesus macaque received AAV with GFP-Lck, while all other animals received tdTomato, Figure S3B). Compared to mice, astrocytes in primate primary motor cortex also displayed a 1.5-fold larger territory area (p<0.0001, t-test with Welch’s correction) and 1.3-fold larger Feret’s diameter (p<0.0001, t-test with Welch’s correction)(Figure 3B, top, 3C, 3D). Similarly, astrocytes in primate striatum displayed a 1.5-fold larger territory area (p<0.0001, t-test with Welch’s correction) and 1.3-fold larger diameter (p<0.0001, t-test with Welch’s correction) compared to rodent striatal astrocytes (Figure 3B, bottom, 3E, 3F). To assess the distribution and density of fine branches within an astrocyte territory, we measured the neuropil infiltration volume (NIV) (Baldwin et al., 2021; Stogsdill et al., 2017). NIV analysis assesses small regions of interest (ROI) and may vary considerably based on placement of the ROI. We carefully placed ROIs to avoid the soma and primary astrocyte branches, though we interpret this data cautiously. Remarkably, despite larger territories, we found that the NIV of rhesus macaque astrocytes was 1.8-fold smaller than mouse in both primary motor cortex (p<0.0001, t-test with Welch’s correction) and striatum (p<0.0001, t-test)(Figure 3G–I).

Figure 3. Comparison of astrocyte territories after viral transduction using the GfaABC1D promoter.

(A) Schematic of experimental procedure. (B) Example maximum intensity z-projections or astrocytes in mouse and rhesus macaque motor cortex layer 5 and striatum (putamen in rhesus macaque and dorsolateral striatum in mouse). All scalebars 20 μm. (C-D) Territory area (p<0.0001, t-test with Welch’s correction) and Feret’s maximum diameter (p<0.0001, t-test with Welch’s correction) of cortical astrocytes (N_{rhesus macaque}= 3, N_cells= 25, N_mouse= 5, N_cells= 30). (E-F) Territory area (p<0.0001, t-test with Welch’s correction) and Feret’s maximum diameter (p<0.0001) of striatal astrocytes (N_{rhesus macaque}= 2, N_cells= 41, N_mouse= 5, N_cells= 22). (G) Representative examples of sites for neuropil infiltration volume (NIV) analysis in mouse (left) and rhesus macaque (right) in primary motor cortex (top) and striatum (bottom). Scalebar 10 μm on overview image and 2 μm on inset. (H) Neuropil infiltration volume (p<0.0001, t-test with Welch’s correction) of cortical astrocytes (N_{rhesus macaque}= 2, N_cells= 14, N_mouse= 5, N_cells= 30). (I) Neuropil infiltration volume (p<0.0001, t-test) of striatal astrocytes (N_{rhesus macaque}= 2, N_cells= 15, N_mouse= 5, N_cells= 22). Only animals injected with AAV5-GfaABC1D-tdTomato were included in NIV analysis.

Using GFAP immunoreactivity, we established that primary branching structure of the astrocyte is conserved between mouse and rhesus macaque. Further, with viral-mediated expression of fluorescent tag proteins, we found that the territory and diameters of primate astrocytes are larger in rhesus macaques than mice, while mice have increased NIV compared to rhesus macaques. Given that astrocyte branching structure may be severely affected by chemical fixation (Korogod et al., 2015), and viral labeling methods to transduce astrocytes may trigger a reactive phenotype, it was possible that our observations resulted from species differences in fixation and astrocyte reactivity. Thus, to complement the viral-mediated labeling technique, we forwent transcardial fixation for a short immersion fixation period and utilized a diOlistic labeling method (Figure 4) to assess cortical astrocyte morphology. Consistent with our results in virally labeled and perfusion-fixed tissue (Figure 3), diOlistically-labeled cortical astrocytes in mice had a smaller territory area and diameter when compared to rhesus macaques (p<0.0001 and p<0.0001, t-tests, Fig. 4C–D). However, astrocytes in mice and rhesus macaque displayed similar overall complexity in 2D as measured by fractal dimension (Figure 4E). In addition, similar to our results obtained with the AAV labeling method, rhesus macaque astrocytes displayed sparser astrocytic infiltration of the neuropil (p=0.0164, t-test)(Figure 4G).

Figure 4. Morphology of fine astrocyte processes using diOlistics.

(A) Experimental overview. (B) Representative examples of diI-labeled astrocytes in mouse (left) and rhesus macaque (right). Maximum intensity z-projections (top) and individual slices (middle and bottom) of each cell. Scalebars 20 μm. (C-D) Territory area (p<0.0001, t-test, N_mouse=6 N_cells= 10, N_{rhesus macaque}=2 N_cells= 12) and Feret’s diameter (p<0.0001, t-test) of cortical diI-labeled astrocytes. (E) Fractal dimension box counting (p=0.4961, t-test). (F) Representative examples of neuropil infiltration volume (NIV) sites and surface reconstructions in mouse (left) and rhesus macaque (right). (G) Neuropil infiltration volumes (p=0.0164, t-test).

Primate and mouse astrocyte territories cover the same number of axon terminals

Since we found a significant difference in astrocyte territory between species, we next investigated whether mouse and rhesus macaque astrocyte territories covered a comparable number of axonal terminals in primary motor cortex and striatum. We focused on inhibitory terminals because of the extensive literature on interneuron expansion across evolution, and thus suspected that they would reveal the greatest potential species differences (Krienen et al., 2020; Tepper and Bolam, 2004; Wildenberg et al., 2021). The analysis was conducted in virally labeled astrocytes (Figure 5). The territory of individual astrocytes in primary motor cortex in mice and rhesus macaques encompassed a similar number of vgat1+ axon terminals per territory (p=0.7120, t-test). The similarity of terminal numbers between species, despite larger territory size in the rhesus macaque, was due to a lower density of inhibitory terminals in rhesus macaque (p<0.0001, t-test)(Figure 5C–D). Interestingly, in the striatum, both vgat1+ terminal density (p<0.0001, t-test with Welch’s correction) and the number of vgat1+ terminals per astrocyte territory (p<0.0001, Mann-Whitney)(Figure 5E–F) were lower in rhesus macaque.

Figure 5. Astrocyte coverage of neuronal GABAergic terminals.

(A) Representative examples of image processing from right to left: tdTomato+ astrocyte, vgat1 labeling, outlining of the astrocyte territory, and quantification of the vgat1+ terminals within the astrocyte territory. (B) Representative images of astrocytes and vgat1 staining in mouse (left) and rhesus macaque (right) primary motor cortex (top) and striatum (bottom). (C) Density of vgat1+ terminals per 100 μm² in layer 5 of mouse and rhesus macaque motor cortex (N_mouse=4, N_astrocytes=19, N_images=57; N_{rhesus macaque}=2, N_astrocytes=8, N_images=24; t-test, p<0.0001). (D) Number of vgat1+ terminals per astrocyte territory in primary motor cortex (t-test, p=0.712). (E) Density of vgat1+ terminals per 100 μm² in striatum of mouse and rhesus macaque (N_mouse=4, N_astrocytes=12, N_images=36; N_{rhesus macaque}=2, N_astrocytes=8, N_images=24; t-test with Welch’s correction, p<0.0001). (F) Number of vgat1+ terminals per astrocyte territory in striatum (Mann-Whitney, p<0.0001). All scalebars 20 μm.

Astrocyte-synapse relationships are the same in mouse and rhesus macaque

To assess spatial relationships between astrocytes and neuronal synapses, we prepared blocks of primary motor cortex from AAV5-GfaABC1D-injected animals for transmission electron microscopy. Astrocyte processes in contact with pre and/or post-synaptic elements were categorized as pre-synaptic, post-synaptic, or cleft-associated astrocytes (Figure 6B–D). The proportions of astrocyte contact types were similar between mouse and rhesus macaque with the majority (around 75%) of astrocyte processes contacting the synaptic cleft (Figure 6E). Given that the viral method used to label astrocytes may not transduce all astrocytes, we did not assess the proportion of synapses not contacted by astrocytes. Thus, our results do not inform about the proportion of synapses with or without astrocyte contacts.

Figure 6. Spatial relationship between astrocyte processes and neuronal synapses.

(A) Schematic of experimental procedure. (B-D) Representative examples of astrocyte-synapse interactions. Astrocyte (red), axon terminal (blue), dendrite or dendritic spine (yellow). The type of synapses formed by the blue terminals are asymmetric in B and D, and symmetric in C. Scalebars are 1μm. (E) Distribution of the types of astrocyte-synapse contacts in mouse and rhesus macaque motor cortex (N_mouse=3, N_synapses=166; N_{rhesus macaque}=2, N_synapses=141; Chi-square on raw counts, p=0.8256). (F) Cleft-associated astrocyte processes were measured parallel to the synapse (N_mouse=3, N_synapses=115; N_{rhesus macaque}=2, N_synapses= 94; Mann-Whitney, p=0.017). (G) Distribution of asymmetric and symmetric synapses of all astrocyte-synaptic contacts. (H) Distribution of astrocyte contact types for identified asymmetric synapses (N_mouse=3,N_synapses=76; N_{rhesus macaque}=2,N_synapses=62; Fisher’s exact on raw counts, p=0.2853). (I) Distribution of astrocyte contact types for identified symmetric synapses (N_mouse=3, N_synapses=18; N_{rhesus macaque}=2, N_synapses=23; Fisher’s exact on raw counts, p=0.8163).

Since we found that primate astrocytes displayed lesser neuropil infiltration volumes, we sought to determine whether primate astrocytes had smaller fine processes which would explain the disparity. Astrocyte processes are not simple cable structures, and thus we needed to determine a consistent method to measure processes by standardizing the location on the astrocyte and angle for the measurement. We measured cleft-associated astrocyte processes at the interface that touched the synaptic cleft and at an angle approximately parallel to the synapse. We found that the smallest astrocytic units, perisynaptic astrocytic processes (PAPs), are slightly smaller in rhesus macaques than in mice (p=0.0170, Mann-Whitney). Rhesus macaque PAPs measured 222.3 nm on average compared to 258 nm for mouse (Figure 6F). The small differences in astrocyte PAP size may partly contribute to the species differences in NIV.

We also assessed the relationship between astrocytes and synapse types in mouse and rhesus macaque. The proportion of synapse types (asymmetric and symmetric) were similar between species (Figure 6G). However, a fraction of synapses was not identifiable by ultrastructural properties alone, and thus there may be a difference in the proportions of synapse types which were not captured by our analysis. The astrocyte contact type did not differ between asymmetric (p=0.2853, Fisher’s exact) and symmetric synapses between species (p=0.8163, Fisher’s exact), with the overwhelming majority of synapses contacted at the cleft (Figure 6H–1).

DISCUSSION

A growing interest in glial cells has generated a myriad of new studies focused on the role of astrocytes in health and disease. While the literature concerning murine astrocytes has grown substantially, investigation of primate astrocytes has been stifled due to technical and resource limitations. By using a variety of techniques to investigate astrocyte morphology in mice and rhesus macaques, we found a great deal of similarity between species.

We found that mice and rhesus macaques displayed differences in the proportions of neurons and astrocytes despite comparable astrocyte densities. Mice had higher densities of neurons across brain regions when compared to rhesus macaques. However, the density of astrocytes remain stable between species. In assessing the proportions of astrocytes and neurons within the total number of cells (as determined by DAPI), mice had a higher proportion of neurons, whereas rhesus macaques had higher cellular proportions of astrocytes. Increased proportions of astrocytes may suggest decreased functional load or stress on an individual primate astrocyte compared to rodent, as we found both the proportion and density of neurons are lower in primates. Thus, an individual rhesus macaque astrocyte governs a lower number of neuronal cell bodies than a mouse astrocyte.

At the level of an individual astrocyte, rhesus macaque astrocytes displayed greater GFAP branching. However, the density of GFAP+ branches within an astrocyte territory were comparable between mouse and rhesus macaque due to the increased territory size of primate astrocytes. Taken together, our data shows that the primary branching of astrocytes is similar between mouse and rhesus macaque. Our data also suggests that rhesus macaque astrocytes show a greater degree of territory overlap due to a similar density of astrocytes across species but larger astrocyte territory area in rhesus macaque. Our results are consistent with the finding that human astrocytes show a greater degree of overlap than mouse (Oberheim et al., 2009). Further studies are needed to investigate the potential molecular underpinnings of astrocyte territory organization in the primate. A wealth of genes that regulate territory formation have been identified in mouse astrocytes (Baldwin et al., 2021; Endo et al., 2022). It remains unexplored whether a different or additional set of genes may be involved in territory regulation in the primate since our morphological data suggest greater territory overlap in primate astrocytes.

We measured neuropil infiltration volumes (NIVs) to compare the degree of fine process branching and extension into the surrounding neuropil (Baldwin et al., 2021; Stogsdill et al., 2017). Unexpectedly, our results demonstrate a lower NIV in rhesus macaque cortex and striatum compared to rodent. While primate astrocytes possess a greater number of GFAP+ primary branches, the lower NIV suggests that these primary branches may support fewer fine process extensions per unit volume. However, we do not interpret this lower NIV as a decrease in complexity in rhesus macaque astrocytes. We propose that the degree of fine process branching between mouse and rhesus macaque astrocytes is the same (as indicated by similar fractal dimension in both species in Fig. 4E), but due to the increased territory area in rhesus macaques, the spacing between branches is greater when measured in volume. Differences in NIV were not entirely explained by differences in astrocyte process size. By measuring the dimension of perisynaptic astrocyte processes (PAPs) parallel to the adjacent synapse in electron micrographs, we found that the distribution of PAP sizes were only marginally smaller in rhesus macaques compared to mice. Thus, the smallest fundamental unit of the astrocyte is similar across species.

There may be functional implications from the assessed morphological features of astrocytes in this study. Astrocytes are known modulators of neuronal synapses, whether through the secretion of synaptogenic molecules (Allen and Lyons, 2018), clearance of neurotransmitters (Murphy-Royal et al., 2017), or release of neuroactive species (Araque et al., 2014). Alterations in the distance of PAPs to synapses could impact such functions (Bernardinelli et al., 2014; Kater et al., 2023). We therefore investigated if the spatial relationship between astrocytes and synapses were different between species in primary motor cortex. Synapses in rhesus macaque and mouse were contacted by PAPs in a similar fashion. The majority of synapses, around 75%, were contacted at the cleft, while few synapses were only contacted at their presynaptic or postsynaptic elements. Cleft contact remained the predominant spatial relationship between astrocytes and synapses when classifying synapses into asymmetric and symmetric groups in both species. Identifying astrocyte processes without labeling in 2D electron microscopy can be extremely challenging and unreliable, thus we relied on expression and immunoperoxidase labeling of tag proteins (tdTomato and GFP) driven by the GfaABC1D promoter. Since the expression of tag proteins can vary in an area of tissue and even within an individual cell, any assessment of synapses without astrocyte processes could have been due to a lack of transduction or lower levels of expression, rendering false-negative results. Therefore, it is possible that a different proportion of synapses are not contacted by PAPs across species. Additionally, we did not characterize symmetric versus asymmetric synapses, thus there could be species differences in the spatial relationship with a specific synapse type.

We determined that astrocytes in mouse and rhesus macaque have common spatial relationships with neuronal synapses, but we also investigated whether they interacted with a comparable number of synapses. It was proposed that primate astrocytes govern a significantly larger number of neuronal synapses, implying a greater functional capacity and complexity than rodents (Oberheim et al., 2009). However, based on estimated synapse densities in human and mouse (10.34 and 23.28 ×10⁸ per mm³ of cortical layer 5, respectively)(DeFelipe et al., 2002), we hypothesized that the synapse density inversely scaled with astrocyte territory size such that the number of synapses governed per astrocyte territory would be comparable between species. We focused on inhibitory synapses given the literature on interneuron expansion across evolution, and thus suspected that they would reveal the greatest potential species differences (Krienen et al., 2020; Wildenberg et al., 2021). Our data shows that the density of inhibitory synapses in primary motor cortex is higher in mice, but that mouse and rhesus macaque astrocytes govern a similar number of inhibitory synapses by territory area. The differences in synapse density between species are consistent with previous studies (DeFelipe et al., 2002; Wildenberg et al., 2021). Interestingly, the number of inhibitory synaptic terminals governed by an astrocyte in the striatum is significantly lower in rhesus macaques than mice, suggesting possible region-specific differences in astrocyte morphology and function across species.

Our findings highlight the similarities between mouse and rhesus macaque astrocytes and add much needed clarity to the commonly held belief that primate astrocytes are more complex than rodent. We find that while primate astrocytes comprise a larger proportion of the primate cellular milieu compared to rodents, and territory sizes are larger in primate astrocytes, their morphology and synapse interaction negatively scale with increased size. Thus, the computational load of astrocytes, if any, may be equivalent across species. Our study raises the question whether astrocyte territory size scales with increases in brain volume across all mammals, or if this increase is primate specific. Additionally, while the larger size of primate astrocytes does not result in increased synapse interaction and neuronal modulation, greater size could contribute to greater vulnerability to oxidative stress (Li et al., 2021). It is conceivable that the core functions of astrocytes are shared across species, but the larger size of astrocytes in primates and perhaps humans make them more susceptible to insult. The functional relationship between astrocyte size and vulnerability to disease should be resolved in future studies. Overall, we provide a comprehensive assessment of astrocyte morphology in mouse and rhesus macaque, with significant implications for astrocyte morphology across primates. Our data inform future disease-related, physiological, and computational studies of astrocytes.

LIMITATIONS OF THE STUDY

We would be remiss if we did not discuss the limitations of this study. First and foremost, while we speculate that the morphological differences and similarities across species found in this study is applicable to human astrocyte morphology, we do not assess human astrocyte morphology. We also cannot address any species sex differences due to animal availability. We used a mix of male and female animals across experiments, but some individual experiments (viral injections) only had female or male mice. We also included rhesus macaques spanning adolescence to mature adulthood based once again on availability. While we did not see substantial differences between individuals of the same species but different ages, we cannot discount the potential influence of age on results in this study. Due to limitations on rhesus macaque availability, two animals with non-neurological clinical issues were included in the study. However, we found that the territory area and diameter of astrocytes in these animals were comparable to that observed in animals without clinical concerns. Our techniques did not allow us to study interlaminar astrocytic processes, which are known to exist in the cortex in rhesus macaque (Falcone et al., 2021b, Colombo, 2001, 1996; Colombo et al., 1999, 1995). Lastly, as this is a morphological study, we can only hypothesize about functional differences or similarities between species. Our findings nonetheless provide the basis to address these limitations in years ahead.

DATA AVAILABILITY

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Adriana Galvan (agalvan@emory.edu).