Sex-specific disruption of murine midbrain astrocytic and dopaminergic developmental trajectories following antenatal GC treatment

All procedures were performed under a licence issued by the United Kingdom Animals Home Office. CBa/ca mice were used for the majority of experiments, whilst annexin A1 (ANXA1) null mice (Hannon et al. 2003) and C57BL/6 wild-type mice were also used, where specified. Mice were bred in-house and housed in the Imperial College facility, Hammersmith campus. The animals were maintained under controlled temperature (21–23 °C), humidity (63 %) and lighting (lights on 08.00–20.00 h), with free access to standard mouse chow and drinking water. Mice were mated overnight and matings were confirmed by the presence of vaginal plugs the following morning. The day following overnight mating was defined as embryonic day (E) 0. Pregnancy was confirmed approximately 10 days later via palpation. Pregnant mice were housed 5 per cage until gestational day (GD) 15, after which they were caged singly. The day of birth was defined as postnatal day 0 (P0). Offspring were weaned at P21, after which time the sexes were housed separately (5 animals per cage) and allowed to grow to adulthood, during which time they received only standard husbandry and no further treatment.

In accordance with previously published protocols (McArthur et al. 2005, 2006, 2007), a non-invasive method for late gestational administration of AGT was employed. The method, comprising of adding dexamethasone sodium phosphate (Faulding Pharmaceuticals Plc., Royal Leamington Spa, UK) to the drinking water of the pregnant mice, overcomes the potential confounding effects of stress associated with handling the animals and injection of drugs. The pregnant mice receiving AGT were divided into two groups, with one receiving a lower dose of dexamethasone (AGT_L, 0.5 µg/ml in the drinking water) from GDs 16–19, and the other receiving the higher dose (AGT_H, 1.0 µg/ml in the drinking water) for the same period of time. Dams in the control group received normal drinking water. There were no group differences in the volumes of water consumed, which were 15.54 ± 0.45, 12.00 ± 1.46 and 13.09 ± 1.5 ml (mean ± standard error of mean, SEM) for controls, AGT_L, and AGT_H, respectively, representing a daily dose of approximately 150–175 µg/kg (AGT_L) or 300–350 µg/kg (AGT_H). As discussed previously (McArthur et al. 2006, 2007), the levels of endogenous GCs that are required for normal lung maturation during late gestation are in the range of physiological ‘stress’ levels and, from a pharmacokinetic standpoint, the optimal dose of dexamethasone for mimicking this in rats has been estimated to be 288 µg/kg/day over GD 18–20 (Samtani et al. 2006a, b). Although equivalent pharmacokinetic data are not available for the mouse, the chosen dose of AGT_L and AGT_H span a relevant range and, moreover, are in the range used in clinical perinatal medicine (Jobe and Soll 2004; Yeh et al. 2004). This treatment had no significant effect on maternal behaviour or the offspring’s adult body weight (McArthur et al. 2007). When analysing the effects of AGT on the adult SNc and VTA, the potential effects of litter of origin were minimised using one male/female per litter per treatment group (n = 6 per group).

Immunoreactive cells were visualised at 15× magnification using a Nikon TE2000U epifluorescence microscope fitted with red and green band-pass filters, a Plan Achromat 10× objective lens, and a 1.5× magnifying lens, linked to a Hamamatsu C4742-95 CCD camera (Hamamatsu Photonics UK, Welwyn Garden City, UK) and an Apple Macintosh G5 computer, running Openlab 5.5 software (Perkin-Elmer, Coventry, UK). Images were coded and stored in a way such that the person performing the cell analysis was unaware of the treatment groups. Each section was projected onto a computer monitor and the contours of the SNc and VTA were delineated by the TH-IR neuronal groups and by referring to visual anatomical landmarks and to the mouse brain atlas of Franklin and Paxinos (Franklin and Paxinos 2008). The SNc can be clearly distinguished from the surrounding non-DA regions of the thalamus dorsally, and the substantia nigra pars reticulata (SNr) ventrally by the presence of TH-IR cells, and from the adjacent VTA by the third cranial nerve tract that runs between the two nuclei. The VTA was delineated by the TH immunopositive region bordered medially by the interfascicular nucleus and the interpeduncular nucleus, and dorsally by the parabrachial-pigmented nucleus. Where the SNr was analysed as a control region, it was defined as the region of tissue ventral to the SNc and dorsal to the glia limitans on the ventral surface of the brain. To detect any regional differences in volume/shape and cell distribution through the SNc/VTA, sections containing the nuclei were divided into four levels (A–D), each spanning 200–250 µm, in a manner analogous to that which we have used previously for analysing the rat SNc/VTA (McArthur et al. 2007), beginning at the anatomical level represented at around −2.7 mm relative to bregma, where the SNc TH-IR cells first appear. Representative images at each level are shown in Fig. 1. Level A was defined solely as the SNc, due to difficulties in distinguishing the very rostral parts of the VTA, which begin to mix with the SNc at this level; level D was defined as the most caudal level for analysis to avoid the problem of distinguishing the dopaminergic populations of the retrorubral field, which mix with the caudal SNc (Fu et al. 2012). Using this delineation, the majority of the TH-IR cells bodies, which are used to define the contours of the nuclei within which the astrocytes (GS-IR cells) are counted, are included in the 4 levels of our analysis, and the TH-IR at levels more caudal to level D is largely due to cell fibres, rather than perikarya, that slowly coalesce to form the MFB.

Regional volumes were calculated using Cavalieri’s method (McArthur et al. 2007) where the cross-sectional area for each level (A–D) of the SNc/VTA, defined by the presence of TH-IR cells, was measured, multiplied by 20 (the section thickness before staining of sections cut at 20 µm) and then quadrupled (to account for staining every 4th section); the overall volume was the sum of the volumes for levels A–D (SNc) and B–D (VTA). The process of fixing, staining and mounting results in a degree of tissue shrinkage, but as this value is constant across all treatment groups, it was not accounted for in the analysis (McArthur et al. 2007).

To estimate the total cell population per level per animal, IR cells within the whole region of interest were counted manually using the count function plug-in available in ImageJ (Rasband, US National Institutes of Health, version 1.4., http://​rsb.​info.​nih.​gov/​ij/​). This two-dimensional counting technique is most appropriate for relatively small nuclei such as the SNc and VTA because it allows for the inclusion of all cells across the whole of the coronal plane, providing a large sampling window for accurate assessment (McArthur et al. 2007; Virdee et al. 2014). It has also been demonstrated that such empirically derived counts of dopaminergic neurons in the SNc, using serial reconstruction, are not significantly different from those obtained using stereological methods (Baquet et al. 2009).

To estimate TH-IR cell size, six such cells per level per animal were randomly chosen from digital images of coded sections, selecting only cells in which the nucleus could be readily identified. From such images, the cross-sectional area was calculated (McArthur et al. 2007). The mean cell areas per level per animal were pooled to form the group means (n = 6).

A z-stack of optical sections was captured from each mouse brain tissue slice that had been counterstained with DAPI, using a Leica TCS SP5 inverted confocal laser scanning imaging system (Leica Microsystems, Heidelberg, Germany) equipped with a Leica DFC 320 digital camera, and powered by a Chameleon Ultra-II MP laser (Coherent Inc., CA, USA) and visualised with red–green–blue RGB lasers (Leica Microsystems). The confocal z-stacks were taken in 0.78 μm steps with an apochromat 63×/1.4 NA oil-immersion objective lens, in sequential scanning mode. All parameters (pinhole, contrast, gain and offset) were held constant for all sections from the same experiment. The captured images were saved with Northern Eclipse software (Empix Imaging Inc., Mississauga, Canada), exported as TIFF files, and processed post-capture using Adobe Photoshop software (version 7.0), adjusting them for levels of brightness and contrast. Images were merged automatically using the built-in functions available in Photoshop software.

We found that the accuracy of delineating the SNc and VTA at P2 was relatively poor, so the region was measured as a single entity. Apoptotic profiles were quantified by scanning the entire SNc/VTA for each section. TH-IR neurons were classified as apoptotic when their DAPI-stained nuclei appeared condensed, round, brighter and/or fragmented, compared to healthy nuclei (Ahmadi et al. 2003). Moreover, apoptotic nuclei were identified by detection of nuclear chromogen condensation, a morphological alteration indicative of apoptosis (Oo and Burke 1997; Blomgren et al. 2007). Other studies have confirmed that apoptotic profiles assessed using similar methods represent unbiased counts appropriate for the quantification of natural cell death in dopaminergic neurons (Oo and Burke 1997; Oo et al. 2003; Jackson-Lewis et al. 2000). Images were analysed post-capture using ImageJ software. Nuclear profiles surrounded by TH-IR cytoplasm were counted in 5 tissue sections per animal, and the proportion of cells exhibiting chromatin condensation was calculated. Group means were calculated from these values. Confirmation of apoptotic cell death was made by co-localisation of activated caspase-3 IR within TH-IR cells.

The overall shape of the SNc was similar in control male and female mice, but the sum of the volumes at each level was significantly greater in females compared with males by 31 %, with the major difference occurring at level A (Fig. 2a, b, open bars). AGT_L had no effect in either sex (Fig. 2a, b, grey bars). In males, AGT_H induced a marked overall change in the shape of the nucleus, as seen by an increase in volumes at levels C and D, resulting in a net increase (28 %) in overall volume (Fig. 2a, black bars). In contrast, in females, AGT_H decreased overall SNc volume by 17 % due to a substantial effect only at level A (Fig. 1b, black bars).

The overall VTA volume was similar in control male and female mice although a significantly greater volume at level D in females indicates a sex difference in the overall shape (Fig. 2c, d open bars). As for the SNc, AGT_L had no effect in either sex, whereas AGT_H increased total volume in males by 64 % due to significant effects at C and D, resulting in a marked change in the overall shape of the nucleus (Fig. 2c, black bars). However, in females there was no overall effect of AGT_H (Fig. 2d, black bars).

The GS-IR cell count and distribution within the SNc were similar in control males and females (Fig. 3a, b, open bars). In both sexes, AGT_L resulted in a two- to threefold increase in the estimated total GS-IR cell count, as well as the counts at individual levels (Fig. 3a, b, grey bars). Females were affected to a greater extent than males, leading to a significant sex difference. This effect was sustained in males after AGT_H, but in females AGT_H had no significant effect (Fig. 3a, b, black bars).

The estimated total GS-IR cell counts made within the VTA were also similar in control males and females, although a notably lower count at level B in females would indicate a significant sex difference in the distribution of astrocytes throughout the VTA (Fig. 3c, d, open bars). In both sexes, AGT_L induced a remarkable three- to fourfold increase in the estimated total adult GS-IR cell numbers, due to a particularly marked effect at level B in males and at both levels B and D in females (Fig. 3c, d, grey bars). There was thus a sex difference in regional responsiveness of the astrocytes, which led to the estimated total GS-IR cell count in the adult female VTA being double that seen in the male after exposure to AGT_L. AGT_H also greatly increased in the estimated total GS-IR cell numbers in males, and, in contrast to AGT_L, significant effects were seen at all 3 levels, indicating dose effects on regional responsiveness. In females, however, AGT_H was without effect. This sexually dimorphic pattern of responses in the VTA mirrored that seen in the SNc.

To evaluate the relationship between regional volumes (Fig. 2) and astrocyte numbers (Fig. 3), GS-IR cell density is presented in Table 1. This reveals that exposure to AGT_L caused a two- to threefold increase in astrocyte density at all levels of the SNc and VTA in both sexes, except for the VTA at level C in females, where an almost sevenfold increase led to a significant sex difference. In contrast, AGT_H failed to alter GS-IR cell density in any region or sex, except for the female VTA at level C, where the increase in density (threefold) was approximately half that seen with the lower dose of AGT. Notably, neither AGT_L nor AGT_H affected adult GS-IR cell density in the neighbouring SNr in either sex (Table 1), indicating that the effect in the SNc and VTA was not a uniform, non-specific effect throughout the brain.

TH-IR cell counts within the adult SNc were similar in control males and females (Fig. 4a, b open bars). AGT_L had no overall effect in male mice, but significantly increased the estimated total cell count in females due to a predominant effect at level A. In contrast, AGT_H increased counts at C and D in male mice, leading to an overall significant increase (32 %) in the estimated total TH-IR cell count, whereas females were unaffected by AGT_H.

In the VTA, TH-IR cell counts were similar in control males and females, but at the individual levels C and D counts were significantly lower or higher, respectively, in females compared with males, indicating a sex difference in the distribution of dopaminergic neurons throughout the nucleus (Fig. 4c, d). In males AGT_L significantly increased counts at level B and decreased them at C, with no net overall effect, whereas AGT_H increased estimated total counts by 62 % due to effects at C and D. Females, in contrast, responded only to AGT_L with significantly increased counts at B and C.

Table 2 shows that, with only one exception, neither dose of AGT affected regional TH-IR cell density in the adult SNc and VTA of either sex. The one exception was the female VTA at level C, where TH-IR cell density was increased approximately fourfold or threefold after AGT_L or AGT_H, respectively. There are thus sex effects on regional responsiveness.

Measurement of cross-sectional areas of TH-IR cell bodies established that there were regional or sex differences, with the average respective values (mean ± s.e.m, µm²) for males and females being 113.5 ± 2.6 and 112.1 ± 2.3 for the SNc and 125.3 ± 7.3 and 114 ± 2.9 for the VTA. No significant interactions of treatment groups were seen. Thus, regional changes in volume or cell density cannot be attributable to any effects of AGT on TH-IR cell sizes.

As a peak of programmed cell death (PCD) via apoptosis has been reported at P2 (Burke 2004), the effects of AGT_L on apoptotic markers in TH-IR were assessed at this time point. Figure 5a provides an example of chromogen condensation (blue DAPI fluorescence) in the nucleus of a TH-IR neuron (red fluorescence), which is also co-localised with active caspase-3 (green fluorescence), as confirmed by the merged image. At a lower magnification, Fig. 5b illustrates that the cells exhibiting nuclear chromatin condensation are also positive for TH-IR, and all analyses indicated that apoptotic markers were restricted to TH-IR cells, with no evidence of apoptosis in glial cells. In both wild-type and ANXA1-null mice, AGT_L markedly suppressed apoptosis in the male and female SNc/VTA, indicating no role for ANXA1 in programmed cell death of the midbrain dopaminergic neurons (Fig. 5c, d). At P2, the density of GS-IR cells was significantly increased in both sexes by AGT_L in the SNc/VTA region (Fig. 6a), in the SNr (Fig. 6b) and also in the red nucleus (data not shown).

A qualitative summary of the main effects of the two doses of AGT on the estimated numbers and density of astrocytes and dopaminergic neurons in the SNc and VTA of adult male and female rats, as well as neonatal apoptosis in these regions is provided in Table 3.

The present study revealed an intrinsic sex difference in the overall volume (SNc) and shape (SNc and VTA) of the midbrain nuclei in control animals. As sex differences in regional brain volumes strongly associate with behavioural differences in humans and rodents (Cahill 2006; Cosgrove et al. 2007; Swaab 2008; Forger 2009), these findings provide structural evidence in support of functional sex differences (behavioural, neurochemical and neurodegenerative responses) that have been reported for these regions in the normal brain (Gillies et al. 2014a, b; Becker and Hu 2008). The effects of AGT on regional volume appeared to be dose-dependent, with only the group exposed to the higher dose showing neuroanatomical disturbances. Notably, however, the effect in males was to increase the overall SNc and VTA volumes (effects at caudal levels C, D), whereas in females overall volume fell in the SNc (changes rostrally at level A), with no effects in the VTA, thereby exaggerating the sex dimorphisms seen in control animals. To date, structural changes induced by early life GC exposure have been reported for the hippocampus, amygdala, prefrontal cortex and hypothalamus, with evidence for different effects in males and females as well as for links to neuropsychiatric disorders (Uno et al. 1994; Vythilingam et al. 2002; Gilbertson et al. 2002; Sapolsky 2002; Coe et al. 2003; Geuze et al. 2005; Buss et al. 2007, 2012; Lupien et al. 2009; Rao et al. 2010; McEwen and Gianaros 2010; Charil et al. 2010; Garcia-Caceres et al. 2010). The present study therefore highlights the potential for SNc/VTA volume changes to be used as a sensitive, non-invasive biomarker for disorders where early life environmental challenges, as well as dopaminergic malfunction, contribute to their aetiologies.

As the multi-drug resistance protein, P-glycoprotein, extrudes dexamethasone from the adult brain, access of peripherally administered dexamethasone to the brain has been questioned. However, this protein has not been detected in the rodent brain before P7 (Matsuoka et al. 1999). Additionally, GRs are expressed in the basal ganglia, including astrocytes, from E15 in rodents, and GCs can up-regulate GS expression during development (Vardimon et al. 1999; Zschocke et al. 2005; Diaz et al. 1998). Together, these observations support the view that AGT may directly target the cells of the SNc and VTA during development.

In providing a direct link between GC actions in the developing brain and cytoarchitectural disturbances in the adult brain, the current study suggests novel mechanisms which may underpin the strong association between early environmental challenges and vulnerability to neuropsychopathologies, such as schizophrenia, attention deficit/hyperactivity disorder, autism spectrum disorders, substance abuse, Parkinson’s disease, anxiety and depression (Ben Amor et al. 2005; Szpir 2006; Khashan et al. 2008; Raikkonen et al. 2008; Thomas et al. 2009; Hu et al. 2004; Barlow et al. 2007), which involve the ascending DA systems and typically show a sex bias (Aleman et al. 2003; Becker and Hu 2008; Biederman et al. 2002; Baron-Cohen et al. 2005; Gillies et al. 2014a, b). In support of this, we have recently demonstrated that specific behaviours that have psychopathological relevance and are known to involve midbrain dopaminergic systems in numerous species, including rats, mice and humans, are differentially affected by AGT in male and female rats (Virdee et al. 2014). These relate to female behaviours (motivation, arousal), which are altered in depression [more prevalent in women (Kessler 2003)] and male behaviours (pre-attentional processing/pre-pulse inhibition of the startle response), which are affected in male, but not female, schizophrenic subjects (Kumari et al. 2008). Given our evidence for the similarity in AGT programming of the SNc/VTA dopaminergic neurons between mice and rats, it would seem reasonable to suggest that cytoarchitectural re-programming would lead to functional consequences in mice, just as it did in rats. Nonetheless, future studies will be required to ascertain this, and to consider the alternative hypothesis that astrocyte plasticity may represent compensatory mechanisms to preserve function, which is a characteristic of the midbrain dopaminergic systems (Golden et al. 2013; Virdee et al. 2014).

Although precise information on the differential pharmacokinetics between humans, rats and mice is lacking, certain considerations, discussed in detail elsewhere (McArthur et al. 2006, 2007), support the view that the doses of dexamethasone selected in the current study hold physiological and translational relevance. Specifically, they were in the low clinical range used in perinatal medicine (Ballard and Ballard 1995; Jobe and Soll 2004) and also fall within the range required for inducing normal lung maturation in the rat (Samtani et al. 2006a, b). Furthermore, treatment was administered at a stage in mouse brain development which approximates to the latter part of the second trimester of human gestation (Clancy et al. 2001), when AGT is administered in cases of threatened premature birth (Liggins 1994). Our findings therefore highlight the sensitivity of the SNc/VTA neural systems to AGT and underscore the concerns for incurring neurological deficits after repeated courses of AGT, which remain in practice, despite recommendations to the contrary and the lack of evidence for any proven clinical benefits (Baud and Sola 2007; Seckl and Holmes 2007; Murphy et al. 2008). As GC programming mechanisms are thought to be similar in humans and experimental species (Seckl and Holmes 2007), the data also support the validity of our model for predicting potential neuropathological changes after early inappropriate exposure to GCs, and for testing strategies to reverse such effects.