Morphometric analysis of astrocytes in brainstem respiratory regions

Abstract Astrocytes, the most abundant and structurally complex glial cells of the central nervous system, are proposed to play an important role in modulating the activities of neuronal networks, including respiratory rhythm‐generating circuits of the preBötzinger complex (preBötC) located in the ventrolateral medulla of the brainstem. However, structural properties of astrocytes residing within different brainstem regions are unknown. In this study astrocytes in the preBötC, an intermediate reticular formation (IRF) region with respiratory‐related function, and a region of the nucleus tractus solitarius (NTS) in adult rats were reconstructed and their morphological features were compared. Detailed morphological analysis revealed that preBötC astrocytes are structurally more complex than those residing within the functionally distinct neighboring IRF region, or the NTS, located at the dorsal aspect of the medulla oblongata. Structural analyses of the brainstem microvasculature indicated no significant regional differences in vascular properties. We hypothesize that high morphological complexity of preBötC astrocytes reflects their functional role in providing structural/metabolic support and modulation of the key neuronal circuits essential for breathing, as well as constraints imposed by arrangements of associated neurons and/or other local structural features of the brainstem parenchyma.

Astrocytes are critically important for glutamate re-cycling. Since glutamate-mediated transmission is critical for the generation of the inspiratory rhythm (Feldman et al., 2013;Hayes, Wang, & Del Negro, 2012;Koizumi et al., 2016;Koshiya & Smith, 1999), preBötC astrocytes could potentially modulate the activities of respiratory rhythm generating neurons via control of glutamate re-cycling (Hülsmann et al., 2000). In addition, preBötC astrocytes directly modulate inspiratory circuit activity through the release of gliotransmitters, particularly ATP/adenosine (Huxtable et al., 2009;Lorier et al., 2007;Rajani et al., 2017;Sheikhbahaei et al., 2018), prostaglandin E2 (Forsberg, Ringstedt, & Herlenius, 2017), and D-serine (Beltrán-Castillo et al., 2017). However, morphological arrangements of astrocytes that may reflect the complexity and functional significance of neuroglial interactions in respiratory regions and other brainstem areas have not been investigated. Considering the critical role of the preBötC, we hypothesized that preBötC astrocytes and neurons may have special structural arrangements. Recent comparative genomic, morphological, and physiological studies assessing regional transcriptional, structural and functional properties of astrocytes have provided clear evidence for regional functional diversity and specialization of astrocytes (Chai et al., 2017;Kasymov et al., 2013;Schnell et al., 2015;Turovsky et al., 2016). In this study, we reconstructed morphology and performed morphometric analyses of immuno-labeled glial fibrillary acidic protein (GFAP) astrocytes residing within the preBötC region of adult Sprague-Dawley rats. For comparison, we analyzed morphology of astrocytes in two other brainstem regions at the preBötC coronal plane: the nucleus tractus solitarius (NTS), and an intermediate reticular formation (IRF) region located dorso-medial to the preBötC in which some of the neurons have been shown to have respiratory activity and a premotor function (Koizumi et al., 2008(Koizumi et al., ,2013Revill et al., 2015). The data obtained suggest that astrocytes in the preBötC are structurally more complex, which may reflect their functional role in providing structural/metabolic support and modulation of the key neuronal network essential for breathing. This complexity may also reflect constraints imposed by arrangements of associated neurons and/or other local structural features of the brainstem parenchyma. Committees. Animals were housed in a temperature-controlled facility with 12h-12h light-dark cycle (lights on at 7:00 A.M.). Tap water and regular laboratory rodent food were provided ad libitum.

| Tissue processing and immunohistochemistry
Five male Sprague Dawley rats (3-4 months old, 350 g) were terminally anesthetized with an overdose of urethane (3g/kg) and perfused transcardially with 250 ml phosphate-buffered (PB, 0.1 M) solution and then with 4% paraformaldehyde (PFA) in PB solution. The brains were subsequently removed and post-fixed in 4% PFA for 3-5 days.
Brainstems were isolated and then cryoprotected at 4 C in 30% sucrose (in 0.1 M PB-saline solution) over 2-3 days and sectioned coronally at 40-50 μm with a freezing microtome (Leica). Floating sections (1 in 4 series) were quenched in PBS containing 10% methanol and 3% H 2 O 2 to suppress background fluorescence. Antigen retrieval was performed in 1% citrate buffer warmed to 80 C to unmask the proteins. Free-floating tissue sections were incubated for 1-3 days at 4 C with primary antibodies for GFAP (Table 1), choline acetyltransferase (ChAT ; Table 1) to label motoneurons, and/or the rat endothelial cell antigen-1 (RECA-1; Table 1) to label vascular endothelium. The sections were subsequently incubated in specific secondary antibodies conjugated to the fluorescent probes (each 1:250; Lifescience Technologies) for 1.5 h at room temperature. Sections were mounted on slides and covered with an anti-fading medium (Fluoro-Gel; Electron Microscopy Sciences). Tiled images of several medullary regions of interest were obtained automatically under low magnification (10x) using an inverted confocal laser scanning microscope (Zeiss LSM 510). For morphological reconstruction and analysis of astrocytes from the selected regions, confocal image stacks of the GFAP-positive astrocytes within the preBötC, IRF and the NTS were obtained from the same brainstem section (see Figure 1) using a high magnification, oil immersion objective (40×/1.2 NA) applying the same image acquisition settings with 1024 × 1024 pixel resolution. In order to avoid possible variations in GFAP expression during the circadian cycle (Prolo, Takahashi, & Herzog, 2005), and to minimize differences in the background fluorescence as well as in the immunostaining of astrocytes in tissue sections from different animals, all brains were fixed simultaneously using identical protocols and solutions. One investigator sectioned all the brains at the same time, and all the tissue sections were immunostained with identical solutions and processed by the same investigator.
These two anti-GFAP antibodies were found to reveal a very similar pattern of labeling in brainstem astrocytes of adult rats (see Aves Labs, catalog #GFP-1020, RRID: AB_10000240) was obtained from animals that were immunized with recombinant GFP protein.

| 3D reconstruction of brainstem vasculature
The image stacks of regional microvasculature were imported into Neurolucida 360 (MBF Bioscience) and reconstructed using the software's tracing tools. Vessels labeled with RECA-1 were traced throughout the entire thickness of the sections by one investigator and completeness of the tracing was verified by a second investigator. process and sphere at a given radius), and process lengths out to a given radius not including the volume of any smaller radius shells (i.e., total length of processes passing through a shell). 3D convex hull analysis, which analyses the volume enclosed by and surface area of a polygon that joins terminal points of the processes, was used to estimate the volume occupied by the astrocytic process field and surface area of the encased region occupied by an astrocyte (see Figure 3f ).

| Morphometric analyses of astrocytes
For normalization and comparison of cell process complexity between astrocytes from different regions, we used Complexity Index (CI), which was originally developed for the analysis of neuronal dendrites (Pillai et al., 2012) and adopted here for the analysis of astrocyte morphology. CI was defined and computed automatically from the morphometric data by the Neurolucida Explorer software (MBF Bioscience) using the following formula: (Σ terminal orders + number of terminals) × (total process length/number of primary branches), where the number of "terminal orders" for each terminal point is calculated as the number of branches that appear in proceeding backward from the defined terminal to the cell soma. Astrocyte "terminals" were defined as the smallest GFAP-immunostained processes clearly identifiable from the last branching point in high resolution confocal images.

| Brainstem vasculature
After the microvascular 3D tracing data was imported into the Neurolucida Explorer software, the number of microvascular segments (a segment was defined as a section between two vessel branch points), total length of vessels, as well as the total volume and surface area of all the reconstructed microvessels contained within the scanned volume of each region were calculated. The volume occupied by identified blood microvasculature was normalized with respect to the total scanned volume of the region.

| Morphological arrangements of brainstem astrocytes
Processes of the parenchymal astrocytes residing near the ventral medullary surface below the preBötC projected extensively into the dorsal aspect of the brainstem (Figure 2c). This organization became less apparent moving rostrally, as an extra layer of thin astrocytic processes appeared at the RTN level between the ventral surface pial membrane and the parenchyma (Figure 2f,g). Cell bodies of these relatively sparse laminar astrocytes were found to be located close to the pia mater and have numerous long processes coursing parallel to the ventral surface in the medio-lateral plane, creating a prominent network of astrocytic fibers (Figure 2g). These GFAP-positive processes are straighter than those of astrocytes residing within preBötC, IRF, or NTS ( Figure 2g). This dense overlap of GFAP-positive fibers was not observed in any other brainstem regions surveyed and represents a feature unique to the juxta ventral surface region of the medulla oblongata at the level of the RTN.
In the preBötC, IRF, and NTS regions where the astrocytic processes were found to be less densely arrayed, individual astrocytes and their branched processes could be readily distinguished from patterns of GFAP immunostaining. We therefore selected IRF and NTS regions at the same medullary level with preBötC for detailed reconstruction and comparative analysis of astrocyte morphology. The densely intermingled GFAP-stained processes of RTN astrocytes did not allow accurate tracing of processes of individual astrocytes, and, therefore, the morphology of RTN astrocytes was not assessed further.  Figure 4) of IRF astrocytes (n = 10, from five different rats). Error bars represent SEM

| Morphometric features of brainstem astrocytes
Sholl analysis was applied to the reconstructed astrocytes from the preBötC, IRF, and NTS regions (five medullary sections in total, analyzed from 5 adult rats, see Section 2.6). The average number of branch points, number of process intersections, the total length of processes (Figures 4-6), number of process terminals (Figure 7), as well as the convex hull volume and surface area of the reconstructed astrocytes from these brainstem regions were compared (Figure 7).

| Regional blood microvessel morphology
Brainstem astrocytes, similar to astrocytes residing within the other brain regions, make extensive contacts with all parenchymal blood (e,f ) Summary data of the convex hull volume (e) and surface area (f ) of astrocytes in the preBötC, IRF, and the NTS. PreBötC astrocytes have longer processes, more branch points and terminals, and greater convex hull volume and surface area, compared to IRF and NTS astrocytes. Data sets without p values indicated are not significantly different vessels (Figure 10a,b). Regional differences in the morphology and complexity of astrocytes may reflect differences in the arrangements of local cerebral vasculature. Therefore, we morphometrically assessed preBötC, IRF, and NTS microvasculature within the same medullary sections (total of five sections at the same medullary level from five different animals analyzed). Figure 10c,d illustrates the 2D arrangement, represented by the maximum projection from a 3D rendered confocal image stack and reconstruction of microvessels in the preBötC region. There were no differences in the average number of blood vessel segments (91 ± 3, 81 ± 5, and 84 ± 5, n = 5, p = 0.30) and the total vascular length (1.02 ± 0.06 μm, 0.90 ± 0.05 μm, and 0.88 ± 0.06 μm, n = 5, p = 0.20), in the preBötC, IRF, and NTS regions, respectively. Moreover, the average total volume occupied by the parenchymal blood vessels was similar (p = 0.60) in the preBötC (8.8 ± 0.3 μm 3 , n = 5), IRF (8.7 ± 0.1 μm 3 , n = 5) and NTS (7.8 ± 0.6 μm 3 , n = 5) regions ( Figure 10e). There were also no differences in the average total surface area of parenchymal vessels in the preBötC (1.94 ± 0.02 μm 2 , n = 5), IRF (1.92 ± 0.03 μm 2 , n = 5), and NTS (1.74 ± 0.08 μm 2 , n = 5) regions (p = 0.06) (Figure 10f ). While it has been proposed that astrocytes may have neural circuitspecific structural and functional properties (Chai et al., 2017), the morphological features of astrocytes residing in functionally distinct brainstem respiratory regions have not been examined. In this study, we used computer-based 3D reconstruction of GFAP-immunoreactive astrocytes in several brainstem regions at the medullary level of pre-BötC respiratory circuits of adult rats to characterize morphology of mature brainstem astrocytes.

| Immunohistochemical labeling and reconstruction of astrocyte morphology
We employed GFAP immuno-labeling to delineate key features of astroglial structure allowing anatomical reconstruction. GFAP belongs to the family of intermediate filament proteins that are mainly expressed in protoplasmic and specialized CNS astrocytes (Lawrence F. Eng, 1985;Jessen, Thorpe, & Mirsky, 1984). This structural protein is one of the fundamental components of the astroglial cytoskeleton and plays a critical role in the formation of complex processes of astroglia (Fuchs & Weber, 1994;Gomi, Yokoyama, & Itohara, 2010;Middeldorp & Hol, 2011;Weinstein, Shelanski, & Liem, 1991).
Although GFAP immunostaining does not reveal the entire structural volume of astrocytes, this labeling approach can be used for comparative analysis of key morphometric properties of astrocytes (Eilam, Aharoni, Arnon, & Malach, 2016;Saur et al., 2014).
Other astroglial markers such as S100β, vimentin, glutamine synthetase, and glutamate transporters (such as GLAST or GLT) have also been used to study astroglial properties (Catalani et al., 2002). However, S100β and glutamine synthetase immunostaining is mainly localized in the cytoplasm of astrocytes, and only weakly label cellular processes (Wu, Zhang, & Yew, 2005). Moreover, it was reported that glutamine synthetase is expressed in oligodendrocytes and neurons (Bernstein et al., 2014;Tansey, Farooq, & Cammer, 1991). Although vimentin is also a good marker for analyzing astrocytic morphology, it is FIGURE 8 Complexity metrics of preBötC, IRF, and NTS astrocytes. Summary data comparing the measures of structural complexity of astrocytes obtained from the Complexity Index formula (see Section 2.6) applied to reconstructed astrocytes from the preBötC, IRF, and NTS regions. When compared to astrocytes from the other brainstem regions, preBötC astrocytes exhibit a significantly higher complexity index primarily expressed in developing (i.e., immature) glia cells (Dahl, Rueger, Bignami, Weber, & Osborn, 1981;Pixley & de Vellis, 1984). GLAST or GLT immunostaining is also not suitable for morphometric analysis of astrocytic processes (Saur et al., 2014), since only low quality images can be acquired (M. Zhang et al., 2011). SOX9 is another astrocyte specific marker that can be used to identify astrocytes in the adult brain (Sun et al., 2017), but SOX9 only labels the cell nucleus.
There is evidence that in hippocampal astrocytes filled with lipophilic dyes (which reveal the fine cellular processes) or immunostained with GFAP antibody (which does not delineate the finest processes), there were no significant differences between measured values of astrocyte diameter as well as the longest and thickest processes (Oberheim et al., 2008). Thus, GFAP immunostaining of astrocytes appears to be a reliable method to identify the major cellular processes of mature astrocytes. In this study, for comparative analyses of GFAPlabeled astrocytes, the brains were fixed with the identical protocol and solutions, processed at the same time, and developed in the identical immunostaining solutions for the same period of time to standardize labeling. In addition, images used for morphological reconstruction were acquired for the different regions of interest from a single medullary section at the same level to assure standardized conditions for both immunostaining and image acquisition.  processes, a 3D convex hull analysis was performed to provide a metric of the volume occupied by the astrocytic process fields, which should encase much of the field of fine processes not stained by GFAP (Supplementary Figure 1). Other approaches such as genetically-driven expression of fluorescent proteins or injections of fluorescent dyes that have been used to label leaflets of astrocytic processes (Grosche et al., 1999;Miller & Rothstein, 2016) combined with super resolution microscopy or serial electron microscopy would ultimately be required to assess the entire structural volume of astrocytes.

| Astroglial morphometric properties
Our data suggest that preBötC astrocytes are larger (higher convex hull volume) and structurally more complex (higher Complexity Index) than astrocytes residing within the other functionally distinct brainstem regions (IRF and NTS). Specifically, preBötC astrocytes have longer processes, more branch points and terminals, and greater convex hull volume and surface area compared to IRF or NTS astrocytes.
The data obtained also suggested that GFAP-labeled processes of nearest neighboring astrocytes residing within the preBötC or IRF exhibit relatively little spatial overlap. However, astrocytes in the NTS and especially the RTN appear to have overlapping domains. In the rodent hippocampus and cortex, protoplasmic astrocytes residing in the gray matter occupy distinct spatial domains, with little overlap (less than 5%) (Bushong et al., 2002;Halassa et al., 2007;Livet et al., 2007;Oberheim et al., 2008;Ogata & Kosaka, 2002), though this notion has been recently challenged by the studies conducted using human (Oberheim, Wang, Goldman, & Nedergaard, 2006) and ferret (López-Hidalgo, Hoover, & Schummers, 2016) cortical tissue. The extent of spatial overlap of astrocytic processes may have implications for neighboring astrocytes to form networks and interact functionally (Ma et al., 2016;Xu, Wang, Kimelberg, & Zhou, 2010). Structuralfunctional imaging data (Chai et al., 2017) revealing the active domains is needed to confirm that preBötC and IRF astrocytes occupy nearly exclusive territories.

| Morphometry of brainstem microvasculature
The observed differences in regional astrocytic morphological features could reflect constraints imposed by arrangements of associated neurons and/or blood vessels. We have not assessed regional somatodendritic morphology of neurons specifically in relation to astrocyte morphology, but we analyzed morphology of local microvasculature since it is well known that astrocytes are intimately associated with brain parenchymal blood vessels. We found, however, that the arrangement of the microvasculature in terms of average number of vessel segments, total vascular length, vascular volume and surface area in the preBötC, IFR, and NTS regions were not different.

| Concluding remarks
Astrocytes play an important role in modulating the activity of the respiratory rhythm-generating circuits of the preBötC . Here we show that preBötC astrocytes are structurally more complex than those residing within the functionally distinct neighboring IRF region, or the NTS located in the dorsal medulla oblongata. We hypothesize that this morphological complexity of pre-BötC astrocytes reflects their functional role in providing structural/ metabolic support and modulation of the key neuronal network essential for breathing, and possibly also reflects constraints imposed by arrangements of associated neurons and/or other local structural features of the brainstem parenchyma.