Namaqua rock mice, red veld rats, gerbils and spiny mice (n = 33) (for details see Table 1) were live-trapped in early spring at Goro Game Reserve [Limpopo Province, South Africa; (22° 58′S, 22° 57′S; 29° 25′E, 29° 24′E)] using Sherman live traps (H.B. Sherman Traps Inc., FL, USA). Animal trapping and handling followed the guidelines of the American Society of Mammalogists (Gannon and Sikes, 2007) and was approved by the Animal Ethics Committee of the University of Pretoria, South Africa (Ethics Clearance Number EC028-07) and the CITES and Permits Management Office, Department of Environmental Affairs, Limpopo Province, South Africa (Permit number CPM-333-00002). Animals were identified morphologically (see below for a more detailed genetic identification procedure), sexed, and perfused immediately after trapping.

After deep anesthesia using Pentobarbital (50 mg/kg body weight), animals were perfused transcardially with heparinized phosphate buffered saline (PBS) followed by sodium sulfide solution and 4% paraformaldehyde (PFA) with 15% of a saturated solution of picric acid. Brains, femurs, and eye lenses were collected and post-fixed in 4% PFA overnight. Right brain hemispheres were transferred into 30% sucrose in PBS for immunohistochemistry. The left hemispheres were kept in fresh PFA until 2-hydroxyethylmethacrylat (HEM) embedding. Femurs and eye lenses were transferred into PBS/azide (500 mg/l) until further processing. Reproductive status of females was determined by inspecting uteri for pregnancy marks, and females where classified as nulliparous (no former pregnancy) or parous for females that had at least one litter before. None of the females were pregnant or lactating at the time of trapping.

For immunohistochemistry, 40 μm sagittal serial sections of frozen right hemispheres were collected and kept at −20°C in a cryo-protection solution till further processing. Free-floating sections were stained for DCX, Ki67 (Figure 1) and calbindin. Between all steps, sections were washed with Tris-Triton [TBS (pH = 7.4)/0.05% Triton] and, after incubation with primary antibody, with TBS only. For epitope retrieval, sections were transferred to citrate buffer (Dako) and either shortly microwaved (DCX and Calbindin) or heat-treated (40 min at 94°C for Ki67). To inhibit endogenous peroxidase activity, sections were incubated in 0.06% H₂O₂ for 15 min. After pre-incubation in Tris-Triton, 2% normal serum, 0.2% Triton, and 0.1% bovine serum albumin (BSA), sections were incubated in DCX (polyclonal goat IgG sc-8066, Santa Cruz Biotechnology, 1:5000), calbindin (polyclonal rabbit IgG CB-38a, Sigma, 1:10,000), or Ki67 (polyclonal rabbit NCL-Ki67p, Novocastra, 1:7000) antibodies at 4°C over night. Incubation in secondary antibody (rabbit anti goat 1:300, or goat anti rabbit 1:300, Vectastain) was followed by incubation with ABC solution (Vectastain). Finally 3,3′-diaminobenzidine (DAB) stained sections were mounted, DCX stained sections were counterstained with hematoxylin solution, and all sections were dehydrated and cover-slipped.

For the estimation of GC and apoptotic cell numbers, left hemispheres were embedded in 2-HEM (Technovit 7100, Kulzer GmbH). Hemispheres were washed with PBS and dehydrated in graded ethanols. After incubation in 1:1 ethanol and HEM solution, hemispheres were infiltrated in three consecutive HEM solutions for several days until final embedding. Twenty μm horizontal sections were cut and Giemsa-stained (Merck, Darmstadt, Germany) in 67 mmol KH₂PO₄ solution for 40 min at room temperature, differentiated in KH₂PO₄ for 90 s, dehydrated and cover-slipped. Mossy fiber terminals were Timm-stained following the procedure described before (Gatome et al., 2010).

The eye lenses were dried for 5 days at 80°C and weighted (Barker et al., 2003). Femurs were washed and immersed in a rapid decalcifying solution (Baker, Histo Grade) for 24 h. Using a digital calliper, a 3 mm segment was taken from the mid-diaphysis of the femur and infiltrated with HEM as described above. Twenty μm sections were hematoxylin-eosin stained. In the cross-sectioned femurs, adhesion lines (Figure 2) were counted in the inner and outer lamellar bone tissue using bright field and fluorescent illumination.

Given the cryptic nature of some of the southern African murid rodent species, murid samples were genetically identified. This was critical for the red veld rat (Aethomys chrysophilus), which is morphologically indistinguishable from the Tete veld rat (Aethomys ineptus) (Skinner and Chimimba, 2005). In addition, the highveld gerbil (Tatera brantsii) is morphological similar to the recently re-classified Gerbilliscus leucogaster (formerly Tatera) which have not previously been genotyped from the sampling site. Extracts from bone tissue of all specimens of the subfamily Murinae, along with voucher specimens of A. ineptus, M. namaquensis (formerly Aethomys), and the re-classified G. leucogaster (subfamily Gerbillinae), were analysed molecularly by means of D-Loop amplification and sequence analyses. Primers L14925 and H16499 were used to target a 460 bp region of the mitochondrial genome, amplifying a segment of the D-loop gene (Bastos et al., 2011). Genomic amplification was performed in a 50 μl reaction volume while the thermal profile followed an initial denaturation step at 96°C for 20 s, 2 cycles of denaturation at 96°C for 12 s, annealing at 49°C for 30 s, and extension at 72°C for 2 min, 3 cycles of denaturation at 96°C for 12 s, annealing at 48°C for 30 s, and extension at 72°C for 2 min and 35 cycles of denaturation at 96°C for 12 s, annealing at 47°C for 30 s, and extension at 72°C for 2 min. All samples were purified using PCR Product Purification Kit (Roche), and cycle sequenced using BigDye v. 3.1 terminator cycle-sequencing kit (Perkim-Elmer, Foster City, USA). Samples were run on an ABI 3130 sequencer and the resulting sequence chromatograms were viewed and edited in Chromas program embedded in Mega 4 (Tamura et al., 2007). A BLAST nucleotide search (www.ncbi.nm.nih.gov/blast) was performed to identify the rodent reference species with the highest sequence identity. Additionally, phylogenetic analyses were inferred in Mega 4, in order to molecularly identify specimens according to their relationship with the voucher specimens. Of the 33 sampled animals, genotyping identified 22 as Namaqua rock mice, six as red veld rats, and three as highveld gerbils. The two spiny mice of the subfamily Deomyinae were easily morphologically identified without genotyping due to the presence of dorsal spines (Table 1).

The estimation of cell numbers followed the rules of design-based stereology. The optical fractionator (West et al., 1991) was applied using StereoInvestigator software (MicroBrightField, Inc., Williston, USA). Cells were counted using a 100× oil immersion lens (NA = 1.3). For DCX, on every 6th section (spiny mice every 5th section) the counting frame was set as 30 × 30 μm and step size as 150 × 150 μm. For GCs, on every 6th section the counting frame was set as 10 × 10 μm and step size as 210 × 210 μm.

Ki67-positive and pyknotic cells were counted exhaustively (asf = 1, tsf = 1) using a 100× oil immersion lens (NA = 1.3). Only cells within the GC layer and the subgranular cell layer were considered. Positive cells in the top focal plane were not counted. Every 6th section (spiny mice every 5th section) was analysed. GC size was measured using the Nucleator (MigroBrightField) with four test lines as described previously (Amrein and Slomianka, 2010).

Comparing brain weight [F_{(3, 29)} = 27.57] and body weight [F_{(3, 29)} = 20.84] between species, Namaqua rock mice and red veld rats do not differ from each other (P_brainweight = 0.91; P_bodyweight = 0.47), otherwise, weights differ between all species (P_brainweight < 0.001; P_bodyweight < 0.001) (Table 1). Within Namaqua rock mice, there is no difference in these measurements between sexes [F_{(1, 20)} = 2.65; P_brainweight = 0.12; F_{(1, 20)} = 0.62; P_bodyweight = 0.44]. Due to a small sample size, gender differences in the other species were not evaluated statistically. In the hippocampal formation, the number of GCs [F_{(3, 27)} = 19.92] does not differ between Namaqua rock mice and red veld rats (P = 0.55), between Namaqua rock mice and highveld gerbils (P = 0.30), and between red veld rats and highveld gerbils (P = 0.20). Similarly, within Namaqua rock mice, there is no gender difference for total GCs [F_{(1, 18)} = 0.04; P = 0.84].

Spiny mice have exceptionally high numbers of GCs which differ from all other species (P < 0.001). This unexpected high number of GCs was verified by an independent second investigator, yielding results within 10% of the original estimate, i.e., within the margin of error defined by the CE estimates. To analyse this difference further, we estimated GC size. The cell diameter of spiny mice GCs is on average 6.8 μm, whereas GCs in Namaqua rock mice, red veld rats, and highveld gerbils measure on average between 7.9 and 8.2 μm in diameter.

Aging wild-trapped small mammals is not trivial, and, without a calibration curve, as used by Epp et al. (2009), it can only be relative. We measured lens weight and total bone lines for each individual animal (Figure 2) as suggested by Barker et al. (2003) and Epp et al. (2009). The data were recalculated as a percentage of the mean of the respective measurements within species and summarized in an averaged numerical rank. The transition into an estimated age in months is based on the dataset of Namaqua rock mice. Assuming that all animals have wintered, we gave the animal with the lowest rank the estimated age of 4 months. Twenty-one out of twenty-two animals showed a continuous rank increase, indicating that these animals were born during the preceding season; the oldest animal of this continuous rank line is therefore set to be 12-months old. A linear function was calculated and applied to transform ranks into estimated age in months in all animals. By so doing, 31 animals had an estimated age of between 4 and 12 months, one red veld rat and one Namaqua rock mouse were likely to be older than 1 year.

Despite the morphological differences summarized above, the total number of proliferating cells (Ki67; Table 1, Figure 3), once normalized to the resident GCs [F_{(3, 27)} = 7.91], do not differ significantly between Namaqua rock mice and red veld rats (P = 0.55), between highveld gerbils and Namaqua rock mice (P = 0.40), and between highveld gerbils and red veld rats (P = 0.73). Only spiny mice show a higher percentage of proliferating cells than all other species (P < 0.001). The percentages of differentiating cells of neuronal lineage and dying cells within the dentate gyrus do not differ between species [DCX: F_{(3, 27)} = 0.71; P values between 0.83 and 0.15; dying cells: F_{(2, 26)} = 0.16; P values between 0.76 and 0.59; the number of dying cells in spiny mice was not estimated].

Across all species, there was no correlation between estimated age and normalized proliferating cells [r₍₂₉₎ = −0.18, P = 0.33], dying cells [r₍₂₇₎ = −0.30, P = 0.11], or total GCs [r₍₂₉₎ = 0.11, P = 0.56], whereas normalized numbers of DCX-positive cells show a negative correlation with age [r₍₂₉₎ = −0.41, P = 0.02]. Within the largest animal sample (n = 22), the Namaqua rock mice, we tested cell counts without normalization and found a weak negative correlation between age and DCX-positive cells [r₍₂₀₎ = −0.41, P = 0.06], and there were no correlations for proliferating cells [r₍₂₀₎ = −0.30, P = 0.17], apoptotic cells [r₍₁₈₎ = −0.23, P = 0.32], or total GCs [r₍₁₈₎ = −0.21, P = 0.38] with age.

Female Namaqua rock mice with previous pregnancies (parous) harbor less proliferating cells than males [F_{(2, 19)} = 5.78; P = 0.004], and similar numbers as nulliparous females (P = 0.14). Similarly, the number of DCX-positive cells in parous females was reduced compared with males [F_{(2, 19)} = 5.37; P = 0.006], but not different from nulliparous females (P = 0.30) (Figure 4). Nulliparous females did not differ from males for either proliferating cells (P = 0.19) or DCX-positive cells (P = 0.11). For the number of dying cells [F_{(2, 19)} = 1.66], the only trend is found in parous females compared to nulliparous females (P = 0.09). The number of total GCs did not differ between males, nulliparous, and parous females.

The four southern African rodent species exhibited a different pattern of AHN compared to the four rodent species from Europe (Figure 5). In rodents from the southern African temperate climate, characterized by dry winter and hot summer (Köpper–Geiger climate classification, Cwa), normalized proliferation is decreased [c²(1, N = 59) = 13.24, P = 0.0003]. Normalized number of DCX-positive cells is increased [c²(1, N = 59) = 8.47, P = 0.0036] compared to rodents from the European cold climate (Köpper–Geiger climate classification Dfb). There is no habitat by age interaction for proliferation [c²(1, N = 59) = 2.57, P = 0.1087] and DCX-positive cells [c²(1, N = 59) = 1.34, P = 0.2469].

Subfamily membership (Murinae: including all species from Europe and Namaqua rock mice and red veld rat from southern Africa; Deomyinae: the spiny mice; and Gerbillinae: highveld gerbils) did not yield significant effects [DCX: c²(2, N = 59) = 0.22, P = 0.896; Ki67: c²(2, N = 59) = 1.06, P = 0.588].

The precision of the stereological estimates was calculated for the data on Namaqua rock mice, red veld rats, and highveld gerbils. The CE of total GC numbers, proliferating cells (Ki67), and cells of neuronal lineage (DCX) ranged between 0.06 and 0.1, indicating that the estimation procedures generated a variance that amounts to only 6–10% of the mean. Due to the low numbers of dying cells, CEs for these estimates ranged between 0.3 and 0.36. For all estimates, the methodologically introduced variance is a minor contributor to the total variance of the group means (CE²/CV² < 0.14).

As described above, we did not observe a quantitative difference in brain weight, body weight, or any of the assessed cell numbers between red veld rats and Namaqua rock mice. Histologically, both species show a reflected blade (CA4) of the CA3 pyramidal cell layer as it inserts into the dentate hilus and joins the infrapyramidal limb of the polymorphic cell layer of the hilus (Figure 6). Timm-stained intra- and infra-pyramidal mossy fiber buttons are largely restricted to CA4-like pyramids. Both Namaqua rock mice and red veld rats were also characterized by a distinct differential distribution of calbindin within CA1 pyramidal cells along the proximo-distal axis of the cell layer, and an increase in the apparent number and staining intensity of calbindin-positive cells toward the temporal hippocampus (not illustrated). Interestingly, a heterogenous distribution of calbindin in GCs (Figure 7) were observed in Namaqua rock mice, red veld rats (both stratified), and spiny mice (mosaic).

Namaqua rock mice and the red veld rat are considered phylogenetically closely related (Mills and Hes, 1997; Skinner and Chimimba, 2005). None of the AHN-associated cell estimates differ between the two species. No obvious qualitative morphological differences were observed in the hippocampal histoarchitecture, the distribution of mossy fibers or calbindin. The hippocampus shows a reflected blade of the CA3 cell layer into the hilar region [CA4 after Lorente de Nó's terminology (Lorente de No, 1934)], a feature present in, for example, primates (Rosene and Van Hoesen, 1987), rabbits and guinea pigs (Geneser-Jensen et al., 1974; Geneser, 1987), or foxes (Amrein and Slomianka, 2010). A CA4-like cell layer has not previously been observed in murine rodents. This CA4-like cell layer is accompanied by suprapyramidal and infrapyramidal mossy fiber projections that are typical for the most proximal CA3 pyramidal neurons in other murine rodents (Schwegler and Lipp, 1983). Further investigation is necessary to determine if this field in Namaqua rock mice and red veld rats shares structural and connective characteristics that define CA4 in other species.

All animals were sampled after winter and before the seasonal breeding starts (Withers, 1983; Muteka et al., 2006), and it is therefore highly unlikely that any animal would be younger than 4 months. By comparing the range of lens weights in Namaqua rock mice, we observe a 63% increase in lens weight between the presumably youngest to the oldest animal. As this value reflects a single animal that most likely overwintered twice, we tested for lens weight differences in the animals that overwintered once. We observe a 37% increase in lens weight between the presumably youngest to the oldest animals in this cohort, which would correspond to an age difference of more than 4 months in adult Sprague-Dawley rats (Epp et al., 2009). We therefore expect to have considerable age variation in our sample of Namaqua rock mice. Testing for age-dependent regulation of AHN, we found that the number of DCX-positive cells in the southern African rodents declines with age, as it has previously been reported for many mammals (Kuhn et al., 1996; Amrein et al., 2011). However, the number of proliferating cells showed no correlation with age. It is not clear if this correlation has been obscured by variance associated with small cell numbers and a tentative age ranking. Alternatively, proliferation in these species may be low, but stable in animals born in the same season as it has previously been shown in wood mice (Amrein et al., 2004b).

Within Namaqua rock mice, we observed that females that had previously experienced at least one pregnancy (parous) show lower AHN than males. The small sample of nulliparous females do not differ from either males or parous females. Our observations are in general agreement with previous studies, in which reproductively active female meadow voles showed lower proliferation compared to both males and reproductively inactive females (Galea and McEwen, 1999; Ormerod and Galea, 2001). Studies in laboratory rodents reported a decrease in neurogenesis during gestation and the post-partum period, with a recovery of AHN to baseline levels after weaning (Leuner et al., 2007; Pawluski and Galea, 2007; Kim et al., 2010). Our animals were sampled at the beginning of the breeding season, and the last post-partum period must have passed ~3 months before. A persisting decrease in AHN associated with a life history event has previously been shown in shrews, in which AHN completely ceases after overwintering (Bartkowska et al., 2008).

Behavioral flexibility which would allow quick adaptations to environmental challenges might benefit from AHN (Amrein and Lipp, 2009; Garthe et al., 2009; Amrein et al., 2011). If the types and frequencies of such challenges are rather stable in a habitat, one may expect that there is little need for an extensive modulation of the extent of AHN. AHN in wild or wild-derived wood mice and house mice is indeed relatively stable (Hauser et al., 2009; Klaus et al., 2012; Schaefers, 2013). In the sympatric southern African rodents presented here, the stability in proliferation and neuronal differentiation is maintained across three murid taxonomic units. By comparing the southern African rodents with mice from a cold climate, we show here that rodents from different habitats show distinct patterns. Southern African rodents generate less new cells, but in proportion to proliferation, harbor more DCX-positive cells than rodents from Europe. The European species presented here are monophyletic (i.e., all species belong to the subfamily Murinae). It cannot be ruled out that phylogenetic inertia is partially responsible for similarities between these species (Harvey and Pagel, 1991; Blomberg and Garland, 2002). The southern African Murinae (Namaqua rock mouse and red veld rat) do however not cluster with the European ones, and phylogenetic inertia can therefore only be responsible for similarities in one of the two murine species groups. The habitat-dependent differences are also found in the species pair Namaqua rock mouse and long-tailed wood mouse, even though they are considered to be ecological equivalents in their respective habitats. Low proliferation rates, but high numbers of DCX-positive cells have previously been observed in other species (Amrein and Slomianka, 2010). At the cellular level, this can be achieved by either increasing survival rate of the cells, extending the maturation phase of young neurons, or a combination thereof. In our study, apoptotic cells occur less frequently in the African rodents than in the European species, but the variance in the apoptosis data is too high to draw firm conclusions on the survival of newly born cells. Species-specific differences in maturation time have previously been reported. Young neurons mature slower in mice than in rats (Snyder et al., 2009), and the maturation of newly born hippocampal neurons in primates may take more than 6 months (Ngwenya et al., 2006; Kohler et al., 2011). Independent of the regulatory mechanisms, southern African rodents in this study harbor more DCX-positive cells in the hippocampus than rodents from Europe. Behavioral tests in laboratory mice indicate that AHN is not necessary for learning per se (Jaholkowski et al., 2009) but required for fast behavioral changes that are interpreted as increased cognitive flexibility (Garthe et al., 2009; Kempermann, 2012). We previously suggested that neuronal differentiation can mediate species-specific adaptations to habitat requirements (Amrein et al., 2011). The obvious question is therefore which habitat factor drives the development of the observed differences? Based on the Köpper–Geiger habitat classification, which includes temperature and rainfall, the seasonally high temperatures and water shortages in the southern African habitat clearly differ from the colder European habitat with no shortage in water. Environmental challenges for small mammals in the hot, dry conditions of deserts exceed the ones for animals living in cold climates, e.g., the upper critical temperature for an euthermal mammal is less variable than the lower critical temperature, and also water conservation requires extensive adaptations on the physiological and behavioral level (Merritt, 2010). If and how these factors might interact with AHN is unclear. Many parameters other than temperature and metabolic water homeostasis that are linked or independent from climatic conditions, such as food availability or predator risk, could play a role in regulating AHN. We provide evidence that the habitat can shape two of the stages that comprise AHN, which implies that the specific pattern expressed provides a selective advantage. Further studies are needed to define the interactions between habitat and hippocampal function that provide this advantage.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.