Wild adults from male and female P. obesus and G. tarabuli species were captured in the desert (near Beni Abbès, Southwest of Algeria) from March to June. Captured animals were housed in a research facility (Unité de Recherches sur les zones Arides de Beni Abbès) for three days before sacrifice. During the 3-day captivity, animals were exposed to a temperature of 22 ± 2 °C and a light-dark cycle (12L-12D) and fed ad libitum with chenopod plants (Psammomys) and barley seeds (Gerbillus). Seventeen G. tarabuli (35–50 g) and 15 P. obesus (80–100 g) were used for these studies. All manipulations conformed to local and international guidelines on ethical use of animals.

Animals were deeply anesthetized with urethane (1.5 mg/kg), injected intracardially with heparin and then perfused with fixative. One group of animals was perfused with solution composed of freshly prepared 4% paraformaldehyde and 0.1% glutaraldehyde in sodium phosphate buffer (0.1 M, pH 7.4, 200–300 ml during 20 min, at room temp). After postfixation in 4% paraformaldehyde overnight (4 °C), brains were removed and cut on vibratome to obtain serial frontal slices (about 30–50 μm) which underwent immunolabeling for light microscopy.

Another group of animals was perfused with a freshly prepared solution of 4% paraformaldehyde and 2.5% glutaraldehyde in 0.1 M sodium phosphate buffer (pH 7.4; 200–300 ml during 20 min, room temp). Following perfusion, brains were removed and postfixed for 2 h in the same solution. Blocks containing the SON were then dissected from frontal slices of the brain. These tissues were prepared further for electron microscopy.

For single immunolabeling, standard immunofluorescence and immunoperoxidase techniques were performed on free-floating sections of brains according to procedures described in detail in earlier studies [12–14]. Briefly, after rinsing in Tris-buffered saline (TBS), sections were treated with casein (0.5% in PBS) for 1 hr to block non-specific sites. They were then incubated for 24–48 hr at 4 °C with: (i) polyclonal rabbit antibodies against either VP (diluted 1:5000–1:10 000) or against OT (1:400–1:1000) [15] (gift of A. Burlet); or, (ii) monoclonal mouse IgM antibody that specifically recognizes PSA on NCAM (diluted 1:2000–1:8000) (gift of G. Rougon), (see [16] for production and specificity). Affinity-purified anti-rabbit or anti-mouse immunoglobulins (IgG) conjugated to fluorescein isothiocyanate (FITC, Biosys) (diluted 1:400, 2 hr, at room temperature) or FITC-conjugated anti-mouse IgM (diluted 1:100) were used. For immunoperoxidase, after incubation with anti-rabbit or anti-mouse Igs (diluted 1:200 or 1:400 respectively), sections were incubated with rabbit peroxidase-antiperoxidase (PAP, Dako) complexes (diluted 1:100) or with anti-mouse IgM conjugated to horseradish peroxidase (HRP, 1:50). Peroxidase reaction product was revealed either with 3, 3′-diaminobenzidine (DAB) and 0.01% H₂O₂ or with the more sensitive glucose oxidase-nickel-DAB method.

In some cases, double immunofluorescence was performed on free-floating sections incubated for 48 hr at 4 °C in mixtures of primary antibodies containing the polyclonal rabbit IgG anti-VP (diluted 1:5000) and monoclonal mouse antibodies specific for OT-linked neurophysin (OT-Np; diluted 1:500) [17] (gift of H. Gainer). After careful rinsing, the sections were incubated for 2 hr at room temperature in a mixture of fluorescent conjugates. Rabbit FITC-conjugated anti-rabbit Igs (diluted 1:400) were used to identify VP immunoreactivity, whereas anti-mouse Igs conjugated with Texas Red (TR) (Biosys, diluted 1:500) were used to visualize OT-Np immunoreactivities. The preparations were examined with light microscopy (Leica DMR) using bright field optics for HRP-containing sections and epifluorescence with appropriate filters for FITC- or TR-treated sections.

Controls: controls included omission of primary antibodies and incubation in normal rabbit serum or with irrelevant secondary antibodies. These sections did not show any labeling.

Following glutaraldehyde fixation and rinsing, tissues were postfixed in 1% O_SO₄ in Millonig's buffer (1 hr at room temp), dehydrated in ascending series of ethanol and flat embedded in Epon resin. During dehydration, they were treated with 1.5% uranyl acetate (in 50% ethanol) for 30 min en bloc to enhance contrast. After identification on semithin sections, silver-gray (about 60-80 nm) ultrathin sections were obtained on 300-mesh grids and contrasted further with lead citrate before examination on Philips CM10 electron microscope (SERCOMI Electron Microcopy Center, University of Bordeaux).

In comparison with laboratory rodents and other mammals [18–21], the HNS system of the two desert rodents, P. obesus and G. tarabuli, appeared highly developed. Whatever the immunocytochemical method that we applied, we observed many neurons immunoreactive for either of the HNS peptides in sections from both species (Figs. 1 to 3). Their neurohypophyses were also strongly immunoreactive for VP or OT.

In the hypothalamus of Psammomys and Gerbillus, magnocellular VP- and OT-positive neurons were visible in a number of areas (Figs. 1 to 3). Strong immunolabeling characterized their somata that were round to ovoid (20–30 μm in diameter) and possessed two to three primary dendrites and an axon with varicosities (insert, Fig. 1; Fig. 2c). The fibers were also strongly immunoreactive. As the other species, the magnocellular neurons in desert rodents accumulated in the well-delineated paired PVN and SON (Figs. 1, 2), and in the latter, they were particularly striking in its retrochiasmatic portion (RC) (Figs. 2, 4). In addition, they were easily detectable in the hypothalamic region between the magnocellular nuclei, including in those areas referred to as accessory nuclei (anterior commissural, perifornical and circularis nuclei) (Figs. 1, 3). A semi-quantitative evaluation revealed that VP neurons dominated in the SON (about 2 times more in its principal part and four times in its retrochiasmatic part) and in accessory nuclei. In the PVN, the number of VP and OT neurons appeared equivalent. Bundles of vasopressinergic and oxytocinergic nerve fibers were visible in the internal portion of the median eminence.

Double immunofluorescence revealed that many neurons displaying strong immunoreactivity for VP also showed reaction for OT (Fig. 2). Doubly labeled cells were detected in both magnocellular nuclei; they were particularly visible in the SON, mainly in its RC, (Fig. 2a and c). A semi-quantitative evaluation, on five to ten frontal sections of hypothalamus containing the SON in both species, revealed that immunoreactivities coexisted in about 20–30% of magnocellular neurons.

Activity-dependent neuronal-glial and synaptic changes have been detected and analyzed with electron microscopy in the rat [8] and in the mouse [9]. In the present study, therefore, we examined the SON of desert rodents with electron microscopy and found that the magnocellular nuclei in these species displayed ultrastructural features characteristic of the activated HNS in laboratory animals. As in other species [8], the SON of desert rodents is relatively homogeneous and is composed essentially of magnocellular somata, dendrites, astrocytes and their processes as well as other neuropile elements. As in chronically dehydrated rats [8] and mice [9], the SON of the two desert rodents displayed many directly juxtaposed neuronal profiles, not separated by the fine astrocytic processes that normally separate neuronal elements (Fig. 4). The neuronal juxtapositions were visible between adjacent magnocellular somata and between dendrites (Fig. 4a) that often appeared as bundles (Fig. 4b). In addition, as in laboratory rodents in which HNS secretion is strongly stimulated [1], we found many axonal buttons making synaptic contact with two or more somatic and/or dendritic profiles in the same plane of section (‘multiple’ synapses, Fig. 4c).

Using an antibody that specifically recognizes PSA on NCAM [16], and immunoperoxidase and immunofluorescence techniques, we detected high levels of PSA-NCAM immunoreactivity in the hypothalamus of the two desert rodents.

In hypothalamic sections, a striking immunoreactivity was noted in both the PVN (Fig. 5a) and SON (Fig. 5c). The immunoreactive signal was particularly strong in the ventral portions of the SON and in its ventral glial lamina (Fig. 5c), where neurosecretory dendrites and astrocytic processes accumulate. In the PVN and SON, it filled the neuropile surrounding the magnocellular somata, which were immunonegative (Fig. 5b and d). As in the stimulated rat SON [10,12], there was no PSA immunoreactivity around clusters of closely juxtaposed neuronal somata (Fig. 5b and d).

In comparison with the rat and other mammalian species [18,19,21,28], the hypothalamic magnocellular system of the two desert rodents, P. obesus and G. tarabuli, which is constituted by the neurons secreting VP and OT is very developed. These neurons, which appear far more numerous than those in the laboratory rat, for example, accumulate in well defined areas, the paired supraoptic, PVN nuclei and a number of accessory nuclei. As in the other rodents, VP and OT neurons in the desert species display loose homotypic segregation in the SON, PVN and accessory nuclei but there is no strict separation of the two populations and the two types of neurons intermingle. In our models, the vasopressinergic neurons clearly predominated in the SON whereas in the rat, within the SON, neurons producing VP are found in approximately equal number with OT neurons [18,19]. The retrochiasmatic part of the SON is also much extended largely, composed of VP producing neurons. All these observations reveal the importance of VP in these desert species. That, the vasopressinergic system is well developed in these desert animals is not surprising when one considers that the neurohormone serves to increase the permeability of the renal ducts and thus to concentrate urine, a function of obvious value for survival of a desert rodents that do not drink free water and that must conserve all water obtained from its food and cell metabolism.

In addition, the strong immunostaining of VP and OT observed in SON and PVN of desert rodents was due to an accumulation of the immunoreactive material occupying all the cytoplasm and dendrites. Furthermore, as shown by electron microscopy in desert rodents, the cytological features of the magnocellular somata were clearly those of stimulated neurons (enlarged Golgi apparatus and rough endoplasmic reticulum) involved in hyperactive neurosecretion. In contrast to the rat, in which peptide stores are depleted under strong dehydrating stimuli [29,30], we found strong immunostaining for VP or OT in neurohypophysial axons. The axons were also filled with secretory granules, as seen with electron microscopy. Both observations indicate then that the storage of HNS peptides is great in the neural lobes of both Psammomys and Gerbillus, and are in accordance with previous results [5,7,31–33], demonstrating important stores of VP in the neural lobe of desert rodents, presumably necessary to allow rapid adaptation of such species to the strong dehydrating conditions of the desert environment. The AVP storage in the hypothalamus is also increased, making it possible to restore pituitary AVP pools, thereby replacing the AVP released into the blood during long-term water restriction.

Earlier studies showed that plasma VP concentrations were higher in desert rodents than in laboratory rats [6,33]. In our study, we detected high immunoreactivity not only for VP but for OT as well, in the hypothalamus of the two desert rodents examined. This is not too surprising in view of the fact that OT also contributes to hydromineral regulation, as earlier described under osmotic stimulation or chronic dehydration [34,35]. Previous reports suggested that OT can bind to the V2 receptor in the kidney, thereby stimulating levels of aquaporin 2 and inducing antidiuresis [36]. Such activation of V2 receptors by OT then could contribute to water conservation in these species.

Another novel finding of this study was that we detected numerous HNS neurons that displayed both VP and OT immunoreactivities. As in the dehydrated rat [24], immunocytochemical colocalization of the two peptides was observed in SON and mainly in its retrochiasmatic part and in PVN of Psammomys and Gerbillus. Colocalization was noted not only in neuronal somata but in axons as well, indicating simultaneous transport of both peptides to the neurohypophysis. Whether OT neurons are recruited into VP expression upon prolonged dehydration challenge to compensate the deficit of VP, or whether VP neurons begin to express OT under such strong conditions of stimulation remains to be determined.

Since the HNS in the both desert rodents, P. obesus and G. tarabuli shows many features of increased neurosecretion, it was not too unexpected to find that its neuronal-glial interrelationships are those characterizing the HNS of rodents like the rat and mouse, under various conditions of heightened HNS secretion [1]. In particular, the SON of Psammomys and Gerbillus displayed a morphology similar to that of dehydrated rats [37] and mice [9]. With electron microscopy, we noted that in both desert rodents, the surfaces of many neurosecretory neurons were directly juxtaposed, without glial interposition, at the level of somata and dendrites. The reorganization of the system was also accompanied by synaptic changes, particularly well illustrated by the presence of numerous ‘multiple’ synapses coupling the neurosecretory neurons [1,38]. Such synapses were also a common feature of the magnocellular nuclei of the desert rodents.