Male Wistar rats were used throughout the study. Anesthesia was induced with isoflurane and maintained with phenobarbitone (1.25 mg/kg of bodyweight; i.p.). The animals were perfused transcardially with 0.9% saline solution for 3 min, followed by 15 min fixation with a fixative consisting of 1% paraformaldehyde, 15% v/v saturated picric acid, in 0.1 M phosphate buffer (PB), pH 7.4. This fixation protocol allowed for the visualization of immunoreactivity for both NG2 and synaptically localized proteins. The brains were kept in the same fixative solution overnight at 4°C. An important aspect of the study was to determine the association between NG2 cell profiles and LC noradrenergic somata and dendrites. Since the LC noradrenergic dendrites project preferentially in the rostro-caudal plane (Shipley et al., 1996; Travagli et al., 1996), the LC was thus sectioned in the horizontal plane. A Vibratome was used to prepare 70 μm thick tissue sections which were then stored in 0.1 M PB containing 0.05% sodium azide.

The non-specific binding of secondary antibodies was blocked by incubating sections with 20% normal horse serum, diluted in TRIS-buffered saline containing 0.3% Triton-X100 (TBS-Tx) for 2 h at room temperature. Tissue sections were incubated with a range of primary antibodies documented in Table 1. All antibodies were diluted in TBS-Tx, and incubated for 24 h at 4°C. After washing with TBS-Tx, sections were incubated in a mixture of appropriate secondary antibodies conjugated with either Alexa Fluor 488 (Invitrogen, Eugene, OR, USA), indocarbocyanine (Cy3; Jackson ImmunoResearch), and indodicarbocyanine (Cy5; Jackson ImmunoResearch) for 2 h at room temperature. Sections were then washed in TBS-Tx and mounted in Vectashield (Vector Laboratories, Burlingame, CA, USA). Method specificity was also tested by omitting the primary antibodies in the incubation sequence. To confirm the absence of cross reactivity between IgGs in double and triple immunolabeling experiments, some sections were processed through the same immunocytochemical sequence, except that only an individual primary antibody was applied with the full complement of secondary antibodies.

Sections were examined with a confocal laser-scanning microscope (LSM710; Zeiss, Oberkochen, Germany) using either a Plan Apochromatic 63x DIC oil objective (NA1.4) or a Plan Apochromatic 100x DIC oil objective (NA1.46). Z-stacks were used for routine evaluation of the labeling. All images presented represent a single optical section. These images were acquired using sequential acquisition of the different channels to avoid cross-talk between fluorophores, with the pinholes adjusted to one airy unit for all channels. Images were processed with the software Zen2008 Light Edition (Zeiss, Oberkochen, Germany) and exported into Adobe Photoshop. Only brightness and contrast were adjusted for the whole frame, and no part of a frame was enhanced or modified in any way.

The LC plays a central role in mediating the stress response at the CNS level and modulating adaptive behavioral responses to future stressors (Valentino et al., 1983; Valentino and Van Bockstaele, 2008). In addition, LC dysregulation following exposure to severe, chronic stress is also implicated in maladaptive responses to future stressors and the development of mental illnesses such as anxiety and depression (Harro and Oreland, 2001; Berridge and Waterhouse, 2003). It is currently unclear whether the principal noradrenergic neurons within the LC are solely engaged in LC-stress responses or other cell-types, such as non-noradrenergic neurons or glia within the LC are implicated. We therefore investigated whether exposure to a mild stressor influenced the expression of NG2 immunoreactive profiles within the LC. A deliberately mild stress protocol was used which has been shown not to induce any anxiogenic behavior, thus suggestive of an adaptive response to future stressors (Dhabhar et al., 1997), with a view to associating NG2 cell plasticity with such cellular mechanisms. We therefore used a repeated, variable restraint stress protocol (Buynitsky and Mostofsky, 2009) since restraint stress has been shown to robustly engage not only the peripheral but also central stress pathways by increasing the expression of proteins mediating the stress response (Inoue et al., 1993). A total of 12 (six control and six stress) male rats, aged PND 60 were used in this part of the study. Stress animals were exposed to a variable restraint protocol in order to prevent habituation. On the first 2 days, the animals were placed in a rodent Plexiglas restrainer (Harvard Apparatus) for 30 min. On the third day, the animal was placed in the restrainer without securing the fastener, thus allowing the animal to escape with a certain amount of effort. Once emerged from the Plexiglas restrainer, the animal was left in the cage with the restrainer for the remainder of the 30 min. On the fourth day, the animal was placed in the test cage for 30 min together with the restrainer, but not restrained. This sequence was then repeated for 10 days. One day after the last stressor, the animals were assessed for measures of anxiety and locomotor activity using the elevated plus maze (EPM) according to standard protocols (Walf and Frye, 2007). The time spent in the open and closed arms, as well as the number of entries into each arm was quantified within a 5 min exposure to the EPM. Differences between the mean time spent in each arm and the number of entries into each arm between control and stress cohorts were assessed for statistical significance using the Mann-Whitney test. The animals were then prepared for histological analyses as below.

A total of six rats, three aged postnatal day (PND) 3 and three aged PND 60 were used to estimate the density of NG2 cells within the LC during postnatal development. Tissue from five control and five stress animals was used to quantify the effect of stress on the density of NG2 cells within the LC. Quantification of NG2 cell density was performed according to previously published methods (Corteen et al., 2011). Briefly, serial, horizontal sections of entire LC nuclei were prepared (70 μm-thick sections) using a Vibratome. In our pilot experiments we found no differences in the fluorescence intensity of NG2 immunoreactivity throughout the rostro-caudal and dorso-ventral extent of the LC nucleus. Therefore, 4 tissue sections per animal were used for counting NG2 cell numbers within the LC. TH-immunoreactivity was used to delineate the LC. Only those NG2 cells located within the nuclear core region of the LC were counted. At the magnification used, several fields of view (FOV) were required to image the full area of the LC nuclear core. For a FOV, confocal Z-stacks of TH, NG2 and NeuN immunoreactivity were acquired from the top to the bottom surfaces of the tissue section using a Plan Apochromatic 40X DIC oil objective (NA1.3) with each optical section within the Z-stack measuring 236 μm × 236 μm × 5 μm (X, Y, Z). All NG2-immunopositive cells within an optical section, as well as whether they expressed NeuN or were located within TH-immunopositive somatic indentations, were manually counted using ImageJ software (NIH). The numbers of NG2 cells for each optical section within a field of view were combined and the mean density ± SD for all FOV within and between sections were compared for statistical differences using Kruskal–Wallis one-way analysis of variance. These values were then pooled since there were no statistical differences between FOV, between tissue sections and animals of the same age. The density analysis is presented as the number of cells per 100,000 μm² and the N values refer to the number of animals as a function of either age or exposure to stress. The differences of the mean NG2 cell densities between animals aged PND3 and 60 or between control and stress treatment was assessed for statistical significance using the unpaired Student’s t-test.

The quantitative method used is according to (Corteen et al., 2011) using N = 4 animals and, three sections per animal. The following immunohistochemical reactions were performed: (1) TH- NG2-VGLUT2; (2) TH-NG2-gephyrin; (3) TH-NG2- GABA_AR alpha3 subunit. Three FOV were randomly selected within the LC nuclear core region of each tissue section. A Z-stack consisting of three optical sections was acquired for each FOV with a Plan Apochromatic ×100 (NA1.4) DIC oil immersion objective. The dimensions of the optical sections were 84.94 μm × 84.94 μm in the X and Y planes and 1 μm thick in the Z plane. Optical sections were spaced 5 μm apart in the Z plane. In all cases, triple immunofluorescence was acquired using sequential acquisition of the different channels. The number of clusters within an optical section, either alone or in contact with NG2 cell profiles was manually counted using ImageJ software or expressed as the number of clusters per 10,000 μm².

Our initial investigations focused on the arrangement or location of NG2 cells in relation to the principal noradrenergic neurons of the LC (Figure 1A1). At low magnification, in adult tissue, visualization of pontine NG2 immunoreactivity revealed an enrichment of the signal within the region occupied by the LC, identified by tyrosine hydroxylase immunoreactivity, compared to neighboring brainstem nuclei, such as the mesencephalic trigeminal nucleus and Barrington’s nucleus (Figure 1A2). NG2 cells in adulthood are purported to serve as reservoir of OPCs and thus are primarily involved in myelination (Kang et al., 2010). However, LC noradrenergic neurons are thought to be un-myelinated (Aston-Jones et al., 1980; Olschowka et al., 1981). In addition, this study demonstrated a poor overlap of immunoreactivity for NG2 with myelin basic protein (MBP; Figure 1A3). Indeed, the region of the LC which expressed the highest levels of NG2 immunoreactivity, namely the nuclear core, showed only sparse labeling for MBP (Figure 1A3), which raises the question of the potential role of NG2 cells within this nucleus. NG2 immunoreactive cells had relatively small somata from which highly ramified processes emanated (Figure 1B2). A striking arrangement within the nuclear core of the LC was the location of a sub-population of NG2 cells within indentations (Figure 1B1) of tyrosine hydroxylase (TH) immunopositive somata suggesting a highly intimate relationship between LC noradrenergic neurons and a sub-population of NG2 cells (Figure 1B3). A further population of NG2 immunopositive cells which were not located in somatic indentations were randomly scattered throughout the extent of the LC with their processes located in close apposition to either TH immunopositive profiles or profiles immunopositive for glial fibrillary acidic protein (GFAP), a marker of astrocytes (Figure 1C). Collectively, these localization data suggest that NG2 cells are positioned to interact with the various cell-types contained within the LC nucleus.

In cortical brain regions, heterogeneous populations of NG2 cells are evident based on not only their neurochemistry or progeny (Trotter et al., 2010) but also their functional characteristics, in particular, their capability of generating electrical activity reminiscent of neurons (Belachew et al., 2003; Karadottir et al., 2008). We used a range of neurochemical markers to investigate the molecular profiles of NG2 cells located within the LC. A sub-population of NG2 cells within the LC expressed the neuronal marker NeuN (Figure 2A; see Figure 3 for quantification). Notably, all NeuN immunopositive NG2 cells were located within indentations of TH immunopositive somata although not all those NG2 cells which were located within indentations of TH immunopositive somata were NeuN immunopositive (see Figure 3 for quantification). Under our experimental conditions, we did not detect any NG2 cells expressing NeuN in other cortical brain regions apart from isolated cells within the piriform cortex, in agreement with (Rivers et al., 2008; Figure 2B). NG2 cells within the LC were also immunopositive for doublecortin, a marker of migratory neuronal progenitors (Figure 2C). Such NG2-doublecortin immunopositive cells have been previously described in cortical brain regions (Tamura et al., 2007; Guo et al., 2010). Collectively, these data suggest that a select population of NG2 cells located within indentations of noradrenergic neurons express the molecular phenotypes of neurons.

The LC undergoes extensive functional and morphological plasticity during postnatal development (Nakamura et al., 1987; Bezin et al., 1994). Furthermore, NG2 cell numbers in various brain regions have been shown to vary dynamically during brain maturation (Dawson et al., 2003). To determine whether NG2 cell numbers within the LC change during postnatal development, we examined NG2 immunoreactivity in the LC of animals aged PND 3 and 60 and quantified the mean NG2 cell density at these ages. Qualitatively, the level of LC NG2 immunoreactivity was strikingly more intense in tissue of animals aged PND 3 compared to that of PND 60, when the respective tissue sections were reacted and imaged under identical conditions (Figures 3A,B). Quantification of the density of NG2 cells located within the LC revealed a significant decrease during the postnatal period (mean ± SEM; PND 3, 11.4 ± 2 NG2 cell per 100,000 μm² versus PND 60, 5.6 ± 0.4 NG2 cell per 100,000 μm² (N = 4 animals for each age; P < 0.0001, unpaired Student’s t-test; Figure 3C1). This equates to a ~50% decrease in NG2 cells density during postnatal development. The density of those NG2 cells located within indentations of TH-immunopositive somata also decreased significantly during development (mean ± SEM; PND 3, 8.1 ± 1.7 NG2 cell per 100,000 μm² versus PND 60, 2.5 ± 0.3 NG2 cell per 100,000 μm², P < 0.0001, unpaired Student’s t-test; Figure 3C2). Thus, approximately 75% of LC NG2 cells were located within indentations of TH-immunopositive somata at PND 3 compared to only 45% at PND 60. Finally, the density of NeuN immunopositive NG2 cells also decreased postnatally (mean ± SEM; PND 3, 3.7 ± 0.7 NG2 cell per 100,000 μm² versus PND 60, 1.4 ± 0.1 NG2 cell per 100,000 μm², P < 0.0001, unpaired Student’s t-test; Figure 3C3). This equates to 34 versus 25% of NG2 cells expressing NeuN immunoreactivity at PND 3 and 60 respectively.

NG2 cells have been shown to receive synaptic input from neurons (Bergles et al., 2000; Ge et al., 2006) suggesting communication between such cells and neighboring neurons. To gain a perspective, at the light microscopical level, on the potential mechanisms by which NG2 cells might communicate with neighboring cells within the LC, we used a range of synaptic marker proteins to examine the proximity of NG2 cell processes in relation to excitatory and inhibitory synapses within the LC. Clusters immunoreactive for the vesicular glutamate transporter 2 (VGLUT2), a protein expressed selectively in glutamatergic axon terminals, decorated NG2 cell processes (Figure 4A). We found only sparse evidence of vesicular GABA transporter (VGAT) immunoreactive clusters apposed to NG2 cell profiles. NG2 cell processes were closely apposed to clusters immunoreactive for gephyrin (Figure 4B), a protein which functions primarily to anchor glycinergic and GABAergic receptors at inhibitory synapses with its expression thus predictive of the location of such synapses (Tyagarajan and Fritschy, 2014). In addition, NG2 cell processes were also closely apposed to puncta immunopositive for the GABA-A receptor alpha3 subunit (alpha3-GABA_AR; Figure 4C), the principal GABA_AR subunit in noradrenergic neurons of the LC (Corteen et al., 2011). We found no convincing evidence of gap-junction expression between NG2 and TH-immunopositive profiles. While ultrastructural evidence using transmission electron microscopy is imperative for unequivocal confirmation, the location of NG2 cell processes in close proximity to clusters immunoreactive for synaptic proteins suggests a degree of synaptic input onto NG2 cells and synaptic contact with LC noradrenergic neurons.

To investigate the comparative association of these excitatory and inhibitory synaptic proteins with NG2 cell and noradrenergic profiles at the light microscopical level, we quantified the density of the contacts between the respective immunoreactivity profiles (Figure 5). The mean ± SEM density of VGLUT2 immunoreactive clusters within the nuclear core of the LC was 603 ± 11 clusters per 10,000 μm² whereas the density of VGLUT2 immunoreactive clusters which contacted NG2 cell processes was 86 ± 5 clusters per 10,000 μm² (Figure 5A) which suggests that approximately 14% of VGLUT2-containing glutamatergic axon terminals within the LC contact NG2 cells. The mean ± SEM density of gephyrin immunoreactive clusters within the nuclear core of the LC was 138 ± 3 clusters per 1000 μm² whereas the density of gephyrin immunoreactive clusters which contacted NG2 cell processes was 15 ± 2 clusters per 1000 μm² which equates to approximately 9% of total gephyrin immunoreactive clusters (Figure 5B). The mean ± SEM density of alpha3-GABA_AR immunoreactive clusters within the nuclear core of the LC was 112 ± 4 clusters per 1000 μm² whereas the density of alpha3-GABA_AR immunoreactive clusters which contacted NG2 cell processes was 13 ± 1 clusters per 1000 μm² which equates to approximately 13% of total alpha3-GABA_AR immunoreactive clusters (Figure 5C).

NG2 cell density within the brain is highly dynamic since such cells continue to proliferate throughout adulthood (Dawson et al., 2003) with such proliferation accelerated in response to physical brain trauma (Levine et al., 2001). However, it is currently unclear whether NG2 cell density fluctuates in response to emotional trauma such as that which is known to contribute to mental illnesses, for example, environmental stress. If so, such NG2 cells plasticity is likely to occur in brain regions principally involved in the processing of stressful stimuli, such as the LC. The stress hormone, CRH directly innervates LC noradrenergic neurons (Valentino et al., 1992) and is thus the central mediator of the LC-stress system (Valentino et al., 1993). Surprisingly, we found CRF-immunopositive varicosities closely apposed to NG2 cells profiles (Figure 6A). Exposure of animals to 10 days of repeated restraint stress (30 min per day) resulted in a noticeable increase in the level of LC NG2 immunoreactivity in tissue from stress animals compared to tissue from controls, reacted and imaged under identical conditions (Figures 6B,C). In order to determine the reason for the stress-induced increase in LC NG2 cell immunoreactivity, we quantified the density of NG2 cells within the LC of control and stress animals. The density of LC NG2 cells was significantly higher in tissue from stress animals compared to control (mean ± SEM; control, 5.3 ± 0.4 NG2 cells per 100,000 μm² versus stress, 10.8 ± 2 NG2 cells per 100,000 μm², N = 5 animals; P < 0.05, unpaired Student’s t-test; Figure 7A). In addition, stress significantly increased the density of the subpopulation of NG2 cells which were located within indentations of TH immunopositive somata (mean ± SEM; control, 2.6 ± 0.3 NG2 cells per 100,000 μm² versus stress, 3.7 ± 0.2 NG2 cells per 100,000 μm², N = 5 animals; P = 0.0178, unpaired Student’s t-test, N = 5 animals in each group). However, the density of the subpopulation of NG2 cells which were also NeuN immunopositive was significantly less in tissue from stress animals compared to control (mean ± SEM; control, 1.7 ± 0.2 NG2 cells per 100,000 μm² versus stress, 1.1 ± 0.1 NG2 cells per 100,000 μm², N = 5 animals; P = 0.0185, unpaired Student’s t-test, N = 5 animals in each group). Finally, there did not appear to be any difference in the proportion of NG2 cells which were also immunopositive for doublecortin in tissue from control and stress animals (mean ± SEM; control, 5.1 ± 0.5 NG2 cells per 100,000 μm² of which, 4.1 ± 0.5 were immunopositive for doublecortin, that is ~80% versus stress, 10.6 ± 0.8 NG2 cells per 100,000 μm² of which, 8.6 ± 0.7 were immunopositive for doublecortin, that is ~ 81%.

In the context of this stress-induced increase in LC NG2 cell density, we next explored whether the baseline behavior of the animals was altered using the elevated plus maze to assess their levels of anxiogenic behavior and locomotor activity. There were no significant differences between the times spent in both the open (mean ± SEM; control, 30 ± 16 s versus 23 ± 7 sec; P = 0.42, N = 6 animals; unpaired Student’s t-test) and closed arms (mean ± SEM; control, 134 ± 18 s versus stress, 151 ± 14; P = 0.465, N = 6 animals; unpaired Student’s t-test; Figure 7B) as well as the total number of entries into each arms (mean ± SEM; control, 14 ± 1 entries versus stress 14 ± 1 entries; P = 0.42, N = 6 animals; unpaired Student’s t-test) of the elevated plus maze by control and stress animals (Figure 7C).

To further investigate this stress-induced plasticity of NG2 cell profiles in the context of LC cellular networks, we evaluated whether there were any changes in the degrees of association between NG2 cell profiles and the synaptic marker proteins shown in Figures 4 and 5. There was a striking increase in the intensity of VGLUT2 immunoreactivity within the LC following the 10 days of the repeated restraint stress protocol (Figures 8A,B). Quantification of the total density of VGLUT2 immunoreactive clusters revealed a significant increase in tissue from stress animals compared to control (mean ± SEM; control, 647 ± 38 clusters per 10,000 μm² versus stress, 1040 ± 32 clusters per 10,000 μm²; P = 0.0014, unpaired Student’s t-test, N = 5 animals in each group). Accordingly, there was also a higher proportion of VGLUT2 immunoreactive clusters which contacted NG2 immunoreactive profiles in tissue from stress animals compared to control (mean ± SEM; control, 100 ± 7 clusters per 10,000 μm² versus stress, 529 ± 40; P = 0.004, unpaired Student’s t-test, N = 5 animals in each group).

In contrast to this stress-induced increase in the glutamatergic innervation of the LC, the level of immunoreactivity for the GABA_AR alpha3 subunit, the principal GABA_AR subtype within the LC and thus a key component of GABA_AR-mediated inhibition of LC neuronal activity was significantly decreased (Figures 8C,D). Quantification of the total density of alpha3 subunit immunoreactive clusters revealed a significant decrease in tissue from stress animals compared to control (mean ± SEM; control, 130 ± 9 clusters per 10,000 μm² versus stress, 82 ± 8 clusters per 10,000 μm²; P = 0.0136, unpaired Student’s t-test, N = 5 animals in each group). Accordingly, there was also a lower proportion of alpha3 subunit immunoreactive clusters which contacted NG2 immunoreactive profiles in tissue from stress animals compared to control (mean ± SEM; control, 18 ± 2 clusters per 10,000 μm² versus stress, 8 ± 2; P = 0.0139, unpaired Student’s t-test, N = 5 animals in each group). No significant differences in the total density of gephyrin immunoreactivity (mean ± SEM; control, 125 ± 8 clusters per 10,000 μm² versus stress, 132 ± 9 clusters per 10,000 μm²; P = 0.5565, unpaired Student’s t-test, N = 5 animals in each group) or the proportion of gephyrin immunoreactive clusters contacting NG2 immunoreactive profiles (mean ± SEM; control, 15 ± 1 clusters per 10,000 μm² versus stress, 16 ± 2 clusters per 10,000 μm²; P = 0.6918, unpaired Student’s t-test, N = 5 animals in each group) were detectable in control and stress tissue.