Adult neurogenesis in the vomeronasal organ
To investigate the cellular basis of adult neurogenesis in the mouse vomeronasal epithelium, we used a range of molecular markers (Fig. 1a) expressed at different stages of neuron generation and maturation: Ki-67 and PCNA, which are nuclear proteins present during DNA synthesis in actively proliferating cells [19]; Dcx, which is expressed by neuronal precursor cells and immature neurons for 2–3 weeks after cell division [44]; OMP, which is expressed in all mature sensory neurons, and type-2 vomeronasal receptor (V2R2), present in mature neurons of the basal layer of the VNO [22, 45, 46]; bromodeoxyuridine, 5-bromo-2’-deoxyuridine (BrdU), a synthetic analog of thymidine that is incorporated into replicating DNA, was used to label cells that were actively proliferating at the time of administration [19]. To visualize these markers in the VNO epithelium we used either immunolocalization (Ki-67, PCNA, BrdU, OMP, and V2R2; Fig. 1b, top and central panels) or mouse strains expressing a fluorescent reporter under the control of a marker protein: Dcx-DsRed [47] and OMP-GFP mice [48] (Fig. 1b, bottom panels). Imaging on adult VNO tissue slices revealed that progenitor cells expressing Ki-67, PCNA, or BrdU (1 day post injection) are restricted to the marginal zones of the epithelium (Fig. 1a, b, d) whereas OMP-expressing mature neurons localize in both marginal and central zones (Fig. 1a, b). This is consistent with previous studies identifying these marginal VNO regions as active neurogenic sites (Fig. 1a) [11, 13–15, 17]. Immature neurons expressing Dcx-DsRed are predominantly distributed marginally although occupying a larger zone than Ki-67/PCNA-labeled cells (Fig. 1b), indicative of active migration to more central areas during cell maturation [49]. Higher magnification analysis at single cell-resolution revealed that fractions of Ki-67, PCNA, and BrdU positive cells co-express Dcx-DsRed (Fig. 1c). However, cells positive for these early proliferation markers do not co-localize with V2R2 and OMP+ cells (data not shown). Hence, Dcx-DsRed is widely expressed through neuron maturation from early proliferative stages to newly differentiated mature VSNs.
Next, we analyzed the time course of VSN generation and maturation in adult Dcx-DsRed female mice. We injected mice with 100 mg/kg BrdU and analyzed its incorporation in the VNO of four animal groups: 1, 7, 14, and 20 days post-injection (Fig. 1e). In addition to Dcx-DsRed reporter expression, VNO sections were immunolabelled for BrdU (green) and OMP (blue; Fig. 1d). BrdU+ cells were classified in three developmental stages: (1) postmitotic progenitor cells that are positive for BrdU and DsRed, but negative for OMP; (2) newly differentiated neurons positive for BrdU, DsRed, and OMP; and (3) mature neurons positive for BrdU and OMP, but negative for DsRed (Fig. 1e). Cells positive for BrdU and negative for DsRed and OMP probably account for either actively dividing progenitors or early differentiating glial cells. Cell quantification revealed that at day 1 after BrdU injection the most abundant BrdU immunoreactivity (62 %) corresponds to postmitotic progenitors (P <0.005), indicating a rapid proliferation cycle. These cells decreased sharply after 7–14 days, and by day 20 represented only ~5 % of the total BrdU+ cells. Newly differentiated neurons (positive for both DsRed and OMP) were nearly absent during the first day, but sharply increased to 28–35 % at 7–14 days, and were maintained stable (28 %) at 20 days. By contrast, fully mature neurons (OMP+, DsRed–) were absent during the first 7 days, rapidly growing to 38 % after 14 days, and becoming the most abundant cell type (55 %) at 20 days (P <0.005, n = 12; Fig. 1e). Thus, most VSNs maturate within a period of 14–20 days and only a very small fraction of progenitor cells remain undifferentiated during this time period.
Newly born vomeronasal cells give rise to functional sensory neurons
We next asked whether newly generated cells become functional sensory neurons in the vomeronasal neuroepithelium. First, we studied the general morphology and axon conveyance of Dcx-DsRed+ cells. To provide a direct and simultaneous comparison of Dcx and OMP labeling, we crossed Dcx-DsRed mice with OMP-GFP mice (Dxc-DsRed/OMP-GFP; Fig. 2). Utilizing en face confocal imaging in a VNO whole-mount preparation [50], we visualized VSN knobs at the dendritic tips of Dxc-DsRed/OMP-GFP mice (Fig. 2a). The en face view of the vomeronasal epithelium reveals a homogeneous and punctate distribution of OMP-GFP-expressing VSNs (Fig. 2a). Notably, Dcx-DsRed+ knobs were also present at dendritic tips of a substantial number of VSNs (Fig. 2a). OMP-GFP+ knobs were 5–10-fold more abundant than Dcx-DsRed+ knobs (Fig. 2b; GFP knob density: 11.25 knobs/100 μm2; DsRed knob density: 1.41 knobs/100 μm2; n = 6) and the level of overlap between OMP-GFP and Dcx-DsRed represented ~12 % of GFP+ knobs (Fig. 2c; DsRed-GFP/GFP ratio = 12.14 %, 2,649 double positive knobs). Furthermore, imaging of VNO axons and the AOB nerve layer illustrates strong enrichment of Dcx-DsRed axons in both structures (Fig. 2a). Dcx-DsRed+ axons deriving from the nerve layer terminate into AOB glomeruli that were identified by co-localization with OMP-GFP (Fig. 2a, bottom). Virtually all glomeruli in the anterior and posterior AOB show Dcx-DsRed+ fibers. These data show that Dcx-DsRed+ VSNs can extend their dendrites into the VNO lumen and send axonal projections that reach the AOB glomeruli.
Since maturating VSNs have access to chemostimuli in the VNO lumen and can deliver the olfactory information to the AOB, we next asked whether these cells are functional sensory neurons capable of detecting chemosignals. We analyzed stimulus-induced activity in the VNO of Dcx-DsRed mice using ratiometric Ca2+ imaging on freshly dissociated VSNs [51, 52]. We compared the response pattern of Dcx-DsRed-positive vs. -negative cells in the same VSN preparation in response to three stimuli: (1) the major histocompatibility complex-peptide SYFPEITHI (10−11 M); (2) fresh adult male C57Bl/6 urine (diluted 1:300); and (3) its high molecular weight fraction (HMW; 1:300), containing major urinary proteins [51, 52]. These stimuli have been previously established as VSNs stimuli [51, 53]. Activation by these three stimuli was detected in ~1–3 % of the 6,186 cells screened, taken from 25 mice. Importantly, all three stimuli induced similar levels of cell activation in both Dcx-DsRed+ and Dcx-DsRed– cells (P = 0.85; Fig. 3a), suggesting that Dcx-DsRed+ cells exhibit functional sensory properties.
To assess a more precise maturity status of Dcx-DsRed + cells, we examined the expression of OMP, V2R2, and Ki-67 using post-hoc immunostaining of responsive VSNs directly in the Ca2+ imaging recording chamber [51, 52] in independent parallel experiments. To determine the maturation stage, we classified the cells based on the combination of the specific markers at the time: cells positive for either OMP or V2R2 or negative for Ki-67 reached a late maturation stage, whereas cells positive for Ki-67 or negative for OMP and V2R2 were in an early stage (Fig. 3b). We imaged 2,173 (Ki-67), 2,715 (OMP-GFP) and 2,233 (V2R2) cells taken from 10, 7, and 7 mice in three parallel experiments and calculated the number of Dcx-DsRed+ cells activated by the three stimuli. We observed near-maximum rates of activation for SYFPEITHI, HMW, and whole urine in cells labeled with late maturity markers V2R2 and OMP and negative for Ki-67 (Fig. 3b, c). Immature cells that were either negative for OMP or positive for Ki-67 showed no responses to any of the three pheromone stimuli. Ki-67+ cells are a minority (~2 %) within the total pool of Dcx-DsRed+, and therefore this lack or reduced responsiveness remains unnoticed when averaging all Dcx-DsRed+ responses (Fig. 3a). Cells negative for V2R2 showed a 2- to 5-fold reduction in cell activity (Fig. 3c, d), suggesting intermediate maturation. Stimulation with depolarizing high K+ (90 mM) solution confirmed the presence of voltage-dependent Ca2+ channels in these cells (Fig. 3c, d). All cell groups exhibited activation to high K+ (27 % or higher) except for Ki-67+ cells (0 %), confirming that neuron-like features are not yet fully developed in these cells. Together, these results indicate that newly generated Dcx-DsRed+ VSNs become chemosensitive neurons and that this process occurs at a late maturation stage subsequent to the expression of OMP and V2R2, but not in immature Ki-67+ cells.
VNO neurogenesis is enhanced during pregnancy
In mammals, adult stem cell division is physiologically stimulated during pregnancy in the forebrain SVZ and haematopoietic tissues which seem to favor the display of postpartum and maternal behaviors [28, 29, 31, 54]. To test whether VNO neurogenesis is upregulated during pregnancy, we examined the abundance of Ki-67, PCNA, and Dcx-DsRed+ cells in the VNOs of pregnant females at gestation days (GDs) 19–20 and non-pregnant controls of the same age (Fig. 4a). We found that pregnant mice had significantly more cells positive for the three markers in the VNO after 19–20 days of gestation (Fig. 4b and Additional file 1: ratio of DsRed+ cells pregnant/control 1.75 ± 0.26; *P <0.05; **P <0.01, n = 3 pregnant and 13 control mice; PCNA: control 1,256 ± 361, n = 3; pregnant 5,838 ± 1,090, n = 3 *P <0.05; Ki-67, control, 2,860 ± 734, n = 6; pregnant, 12,111 ± 2,680, n = 4 **P <0.01), indicating a pregnancy-induced increase in neurogenesis. The increase on Dcx-DsRed+ cells occurred in the whole VNO neuroepithelium with no obvious apical/basal gradients. Female mice used in these experiments were previously exposed to adult males in order to be inseminated (see Methods). To verify that contact to male odors does not influence VNO stem cell proliferation, we exposed control-naïve females to male bedding for 20 days and counted the number of Dcx-DsRed cells (Fig. 4b). Bedding-exposed females showed control levels of Dcx-DsRed cells in the VNO (P = 0.873; n = 9 mice) indicating that male-derived odors did not increase the generation of new VSNs. We further determined whether high rates of immature neurons stay stable over time or decrease after parturition. We quantified the number of Dcx-DsRed+ cells 3 days after delivery (Fig. 4b) and observed no significant differences compared to controls (P = 0.77; n = 3 mice), indicating that the number of Dcx-DsRed+ returns to pre-pregnancy levels after parturition.
The increased neurogenesis could result from higher proliferation or, alternatively, from a higher cell survival rate. To distinguish between these possibilities, we injected females with BrdU at GD0 and examined immunoreactivity at GD20 (Fig. 4c, d). VNOs from pregnant females displayed lower numbers of BrdU+ cells at GD20, compared to non-pregnant controls (**P <0.01, n = 4 mice per group; Fig. 4c, d), indicating a decrease in cell survival of cells produced 20 days earlier. Therefore, the increase in neurogenesis (Fig. 4a, b, e, f) seems to be proportionally higher than the decrease of cell survival during late pregnancy (Fig. 4c, d). BrdU+ cell numbers at day 1 were similar in both pregnant mice and controls (P = 0.91; Fig. 4d), showing that proliferation is similar at this stage and mating has no major effect. Together, these results suggest that the increase of immature VSNs during pregnancy is caused by both higher proliferation and cell turn-over, but not by enhanced cell survival. This might be indicative of a cell-selective differentiation/survival process in pregnant females toward specific neuronal cell types.
We next sought to establish when pregnancy modulates neurogenesis in the maternal VNO by determining in which gestational period cell proliferation occurs. We therefore mated females with C57Bl/6 males (GD0) and collected maternal VNOs at GD1, 7, 14, and 20. We measured VNO cell proliferation following embryo implantation using immunodetection of Ki-67 (Fig. 4e, f). In these experiments, we observed a significant increase in the number of Ki-67+ cells only at GD20 (Fig. 4e, f). By contrast, maternal VNOs that were collected at GD1, GD7, or GD14 showed no significant increase in Ki-67+ cells compared to control female VNOs (Fig. 4e, f).
The VNO transcriptomes of pregnant vs. non-pregnant females
To further characterize the molecular and cellular processes occurring in the VNO during pregnancy, we analyzed its transcriptome in pregnant (GD20) and non-pregnant control female mice via high-throughput RNA-sequencing (RNAseq). We sequenced five pregnant and four age-matched control VNO samples at high depth using the Illumina HiSeq platform (Additional file 2). To ensure that our RNA samples were captured from pregnant mice at the peak of VNO neurogenesis, we first analyzed the levels of Dcx expression (Fig. 5a). We found that three of the five pregnant females displayed ~1.5-fold increases in Dcx expression compared to the control mice, consistent with the observed increase in neural proliferation (Fig. 4). The other two females had no increase in Dcx expression. Interestingly, the embryos from these mice appeared to be 1–2 days younger than the other three litters, suggesting the development of their pregnancy was slightly delayed and the peak of neurogenesis had not yet been reached. We therefore excluded these two samples from our subsequent analyses.
To compare the transcriptomes of the remaining three pregnant females to the controls, we first performed a differential transcriptome-wide expression analysis. In total, 101 genes were identified as significantly differentially expressed (DE) with a false discovery rate of 5 %; 59 of these were upregulated in the pregnant samples, whereas 42 were downregulated (Fig. 5b, Additional file 2). A gene ontology analysis on the DE genes revealed a significant enrichment of terms related to cell differentiation, regulation of transcription, cell proliferation, growth, and apoptosis, as well as central nervous system neuron development and axonogenesis (Additional file 2). Many of these DE genes have been reported to interact (Fig. 6) and form a regulatory network likely to underlie the observed alterations in proliferation and cell turn-over.
We found no significant difference in the expression of the VR gene repertoire when considered together (paired t-test, P = 0.08); however, the VNO contains hundreds of different sub-types of VSNs, each defined by the expression of a given VR, or combination of VRs. If the neurogenesis or turn-over of neurons during pregnancy disproportionally affected specific VSN subtypes, the expression of the VR(s) that define those sub-types may be altered relative to the others. To test this, we normalized for the number of total mature neurons in each sample using the expression of genes specifically expressed in mature VSNs (see Methods for details). We identified 24 VR genes that were significantly DE; most (83.33 %) were proportionally upregulated in the VNOs of the pregnant mice (Fig. 5c). From these, 17 are V1R genes and 7 are V2R genes (Fig. 7), distributed across a number of VR subfamilies (Fig. 5d). Together our results indicate that pregnancy induces observable differences in gene expression indicative of alterations in cell proliferation and cell turnover in the VNO of pregnant mice. We find evidence of small biases towards a few subtypes of VSNs, but the majority of the VSN repertoire remains unchanged.
Estrogen stimulates VNO neurogenesis
Neurophysiological signals that arise during pregnancy emerge in response to either embryo implantation or circulating maternal hormones. Given that the neurogenesis increase appears in the last part of gestation, we reasoned that circulating hormones might have a more central role than implantation. Using our RNAseq data, we analyzed the gene expression levels of hormone receptors in the VNO (Fig. 8a). We observed high levels of expression of the progesterone receptors membrane-component 1 and 2 (Pgrmc1 and Pgrmc2), androgen receptor (Ar), prolactin receptor (Prlr), and estrogen receptor α (Esr1). Expression of the classical progesterone receptor (Pgr), as well as estrogen receptor β (Esr2), G-protein coupled estrogen receptor 1 (Gper1/Gpr30), oxytocin receptor (Oxtr), luteinizing hormone/choriogonadotropin receptor (Lhcgr), gonadotropin-releasing hormone receptor (Gnrhr), and follicle-stimulating hormone receptor (Fshr) was lower or not present (Fig. 8a). No significant hormone receptor expression differences were found in controls vs. pregnant females (Additional file 2).
The levels of prolactin, oxytocin, progesterone, and estrogen are known to be affected during pregnancy. Prolactin and oxytocin remain at low levels during mid-pregnancy and rise sharply near the onset of parturition and lactation [55], making them unlikely candidates for mediating the neurogenic response observed here. By contrast, circulating levels of two other hormones – progesterone and estradiol – are found at high levels during the second half of the pregnancy. Progesterone circulating levels increase following mating and drop in the last 2–3 days before parturition, and estrogen levels rise and stay high during the last half of the pregnancy [55, 56]. Pgmrc1 has been shown to be expressed in mature VSNs by immunohistochemistry [41]. Esr1, but not Esr2, protein and mRNA expression has also been reported in the VNO [57]. Our RNAseq data revealed that Esr1 is expressed in the VNO of control and pregnant mice at similar levels and, on average, 8-fold higher than Esr2 (Fig. 8a). We therefore analyzed the pattern of Esr1 expression in the VNO using Esr1 immunoreactivity on WT C57Bl/6 female mice. Consistent with previous studies [58], Esr1 was clearly detected in the non-sensory part of the VNO epithelium (Fig. 8b). Additionally, we detected high Esr1 immunoreactivity at the lateral corners of the VNO epithelium where neuronal progenitors congregate (Fig. 8b, arrowheads). Double-labeling experiments showed that this Esr1 immunoreactivity is co-localized with both proliferating Ki-67 cells and immature Dcx-DsRed+ neurons (Fig. 8c). Esr1 immunoreactivity was absent in mature OMP+ neurons (Fig. 8c). These results strongly suggest that estrogen could exert its effect in the VNO on neuronal progenitors by activation of Esr1.
To establish a functional link between high levels of circulating estrogen/progesterone present during late pregnancy and VNO neurogenesis, we tested the effects of estrogen and progesterone treatment on stem cell proliferation in the VNO. We performed six subcutaneous injections with estradiol benzoate (100 ng/0.1 mL/day) for a total period of 9 days into 8- to 10-week-old, ovariectomized Dcx-DsRed females. On the last day of hormone treatment we administered BrdU and extracted the VNOs after 8 h to monitor proliferation. A 9-day estradiol treatment increased BrdU-labeled cells in the VNO neuroepithelium, indicative of an increase in neurogenesis (Fig. 8d). This effect was highly specific as progesterone treatment (1 mg/0.1 mL/day) failed to induce a significant increase on the number of BrdU+ cells, comparable to that in vehicle-treated females (Fig. 8d). Thus, sustained high levels of circulating estrogen are sufficient to augment the proliferation of VNO neural precursors in ovariectomized females.