Reelin and aromatase cooperate in ovarian follicle development

Reelin plays an important role in cerebral cortex development and synaptogenesis. In the hippocampus, the neurosteroid estrogen affects reelin expression. In this study we tested a potential crosstalk between estradiol and reelin, thus the possibility of a reelin-induced activation of the estradiol synthesizing enzyme aromatase. As a model system, we used ovaries, which express reelin and are a major source of estradiol. We found that in wild-type mice, reelin and aromatase are expressed in granulosa cells of growing follicles. The expression of reelin varies with the estrus cycle and is highest shortly before ovulation, when estradiol serum levels are at their maximum. In ovaries of reelin-deficient reeler mice, aromatase mRNA and protein are significantly reduced, as evidenced by real-time PCR, western blot analysis, and quantitative immunohistochemistry in granulosa cells of preovulatory follicles. In line with reduced estradiol synthesis, ovarian estrus cycle length is prolonged in reeler mice. Most importantly, treating cultured granulosa cells with recombinant reelin results in significant upregulation of aromatase mRNA and protein and increased secretion of estradiol into the supernatant. Our data provide evidence of a local increase of aromatase expression by reelin. Regarding reproduction, this crosstalk may contribute to follicular stability and counteract luteinization in ovaries.

estradiol synthesis 19 . To address this question of a direct crosstalk between reelin and aromatase we used ovaries and granulosa cell cultures, which have been shown to express reelin 20,21 and which are a major source of estradiol. Hence, the female gonads, and in particular granulosa cells, appear to be particularly suitable for investigating a potential crosstalk between aromatase and reelin.
The results of these experiments demonstrate that aromatase activity depends on reelin expression. It is more likely that this interaction is responsible for the disturbances in the estrus cyclicity in reeler mice than irregularities in the hypothalamo-hypophyseal-gonadal axis of the mutant. Together with our previous findings in the hippocampus 5 , the results point to an interdependence between reelin expression and aromatase activity.

Results
Reelin is expressed in ovaries of mice. Reelin could possibly act as a circulating protein from the blood stream that acts on the gonads, or it is expressed directly in the gonads. A prerequisite for reelin-induced aromatase expression is the presence of reelin protein in the ovaries. Western blot analyses of mouse ovarian tissue show that reelin protein is present in considerable amounts in ovaries and that the reelin protein shows the typical protein band of 180 kDa size, as numerously described in the literature for cortical tissue 3,5,22 (Fig. 1a). Subsequent RT-PCR analyses substantiate that reelin synthesis, in fact, occurs in mouse ovarian tissue (Fig. 1b). Using real time PCR, reelin expression in the ovaries and in the cerebral cortex of the same mouse were quantified; this showed that the amount of reelin mRNA is more than 5 times lower in the ovaries than in cortex, yet substantial expression of reelin mRNA in mice ovaries is clearly evident (Fig. 1c). In situ hybridization against reelin was carried out using ovarian tissue sections to identify the cell types that express reelin in the ovaries of 12-weekold mice. The highest signal intensity of reelin mRNA was found in granulosa cells of growing follicles (Fig. 1d).
Aromatase is downregulated in ovaries of reeler mice. Cyclic ovarian functions, such as ovulation, substantially depend on estradiol synthesis, and ovaries are the major source of estradiol in females. Accordingly, the enzyme aromatase cytochrome P450, which catalyzes the last step of estrogen biosynthesis and represents the rate-limiting irreversible aromatization of androgens to estrogens, is strongly expressed in ovaries, in particular in granulosa cells 23 .
In agreement with the data in the literature, our immunocytochemical stainings of ovarian tissue sections showed a strong signal in the stratum granulosum of advanced antral follicles (Fig. 2a, white arrow), whereas in corresponding ovarian follicles of reeler mice, only very faint staining intensities were detectable (Fig. 2b, white arrow). Quantification of immunohistochemical stainings by confocal microscopy and subsequent image analysis of advanced antral follicles revealed a significant reduction of aromatase signal intensity (0.46 ± 0.05 × -fold of control; U-test; *p = 0.04) in reeler when compared to wild-type mice (Fig. 2c). We then prepared ovarian whole cell lysates of wild-type and reeler mice for western blot analyses to quantify aromatase protein levels (Fig. 2d,e). Figure 2d demonstrates the signal intensities of aromatase protein bands at a predicted size of 55 kDa of three different wild-type mice as compared to those bands of three different reeler mice of the same western Gel bands of PCR amplification products of reelin mRNA with commercially available reelin primers in ovarian tissue compared to cortical tissue of the same mouse. (c) Real time quantification of reelin mRNA in cortical and ovarian tissue of 12-week-old mice with commercially available reelin primers demonstrated that both tissues contain reelin mRNA. However, expression is less prominent in the ovaries as compared to cortical tissue. (d) Non-radioactive in situ hybridization revealed distinct somatic signal in granulosa cells of small follicles (left panel) and of the cumulus oophorus (middle panel) in ovaries of 12-week-old mice, if antisense-probes against reelin mRNA were applied (top). No signal was detected in corresponding sections treated with the sense-probes (bottom). In corpora lutea (right panel), no difference between antisense-and sense-probe-treated sections was detectable.
Coordinated expression of reelin and aromatase during the estrus cycle. We further studied cyclic expression of reelin and aromatase mRNA in ovaries of 12-week-old mice at defined stages of the estrus cycle in order to address a potential crosstalk between aromatase and reelin. Figure 3a shows the expression of reelin mRNA during the estrus cycle, from diestrus (di) to proestrus (pro), followed by the estrus (es) and finally the metestrus (met) stage. Interestingly, we found an increase in reelin mRNA during the proestrus phase of approximately 8-fold compared to the diestrus phase. This increase in reelin mRNA correlates with an increase in aromatase mRNA during the proestrus stage, as well as a strong decrease in reelin mRNA and aromatase mRNA in the estrus phase compared to the proestrus stage ( Fig. 3a; U-test; p = 0.01, 3b; U-test; p < 0.001). Data from the literature confirms that aromatase mRNA peaks during the proestrus phase 24,25 . Our finding would support the notion that estradiol increases reelin expression, confirming our previous results in hippocampal tissue 5 . Estrus cyclicity is disturbed in reeler mice. Since ovarian cyclicity is widely reflected by cyclic estradiol synthesis, and aromatase is downregulated in the ovaries of reeler mice, we studied the length of the estrus cycle in reeler mice by inspecting vaginal swabs. In 6 mutants and 5 wild-type mice the cyclic phase was determined daily over an averaged time period of 13 ± 9 days in the wild-type and 20 ± 10 days in the reeler. Based upon these data, the lengths of the various cyclic phases in reeler were compared to those of their wild-type siblings  Figure 3c shows the frequency distribution of the cycle phases within different stages of the estrus cycle. The frequency distribution of the cycle phase is highest for wild-type animals during the estrus phase (45% ± 4.9; Fig. 3c), but shifts in reeler to be highest during metestrus (31% ± 4.9; Fig. 3c), whereas other cycle stages such as proestrus and diestrus are not altered (Fig. 3c). In particular, the metestrus phase, when estradiol synthesis starts to increase to finally reach the preovulatory peak, is prolonged in reeler mice, and the later estrus phase is accordingly shorter than in wild-type. Hence, reeler mice show a disturbed gonadal cycle time, with a metestrus being extended from 14% (wild-type) up to 31% (reeler mutant, Fig. 3c) and the estrus is shortened from 45% (wt) to 27% in the reeler (Fig. 3c). The average cycle length in the wild-type is 3.5 days, with a maximum length of five and a minimum of two days (Fig. 3d), which is consistent with the data in the literature 26,27 . In contrast, reeler mice show an average cycle length of 5.3 days, with a maximum of 11 and a minimum of two days (Fig. 3d). Hence, the reeler has irregular cyclic phases.
Reelin, but not GnRH, upregulates aromatase in granulosa cells. Our data on downregulated aromatase expression and disturbed estrus cyclicity in reeler mice suggest that reelin regulates estradiol synthesis. GnRH has to be considered as a further player. Due to the migratory deficit of GnRH neurons in reeler mice 6 , estradiol synthesis might be disturbed by lack of GnRH in reeler mice and accordingly, aromatase expression would be downregulated.
In agreement with previously published findings 14,28,29 , we found expression of GnRH-R mRNA in the ovaries of both WT and reeler mice, supporting the idea that ovaries are in fact responsive to GnRH (Fig. 4a). Remarkably, GnRH-R expression was even significantly higher in reeler mice ( Fig. 4a; U-test; **p < 0.001). In order to investigate whether GnRH induces aromatase expression, we stimulated granulosa cells with GnRH or with recombinant reelin protein. The presence of the reelin receptors ApoER2 (LRP8) and VLDRR has been shown in primary granulosa cell cultures 19 , and in this study we show that this also holds true for KGN cells (Fig. 4b,c). No significant effect on ApoER2 and VLDRR mRNA expression was seen after treatment with GnRH or reelin (Fig. 4b,c). However, after application of reelin to the medium of KGN cells for 2 hours, an increase in aromatase mRNA (CYP19A1) was detected in KGN cells and in primary granulosa cells, as revealed by real-time PCR, whereas no effect was seen after treatment with GnRH (Fig. 4d,e). The increase in aromatase mRNA was significant in KGN cells ( Fig. 4d; Kruskal-Wallis ANOVA **p < 0.001, post-hoc Bonferroni **p < 0.001) and was  seen in tendency in primary granulosa cells (Fig. 4e). At the protein level, we were also able to demonstrate that 2 h incubation of KGN cells with recombinant reelin protein resulted in a significant increase in aromatase protein ( Fig. 4f; T-test; **p = 0.009).
Measurements of estradiol content in the medium using either RIA (KGN cells; Fig. 4g) or an ELISA-kit (primary granulosa cells; Fig. 4h) underscore that reelin affects estradiol synthesis in KGN cells and in primary granulosa cells. The amount of estradiol in the supernatant of KGN cells was significantly elevated after reelin treatment for 24 h (Fig. 4g; U-test; *p = 0.048). In granulosa cells, reelin treatment doubled the amount of estrogen in the supernatant from 47 pg/ml to approximately 100 pg/ml (Fig. 4h, T-test; *p = 0.036). Testosterone levels were not significantly affected in response to application of reelin for 24 h (Fig. 4h, T-test; p = 0.71).

Discussion
In this study, we demonstrate that the expression of reelin and aromatase is coordinated in the ovaries, and most importantly, that reelin activates estradiol synthesis in ovarian granulosa cells. Reduced estradiol levels, and as a potential consequence, irregularities in the estrus cyclicity in reeler, may contribute to reduced fertility of these mutants. Impaired fertility in reeler mice. It has long been known that the fertility of reeler mice is reduced 30 .
Difficulties in breeding reeler, together with the finding that in reeler the number of GnRH neurons in the hypothalamus is reduced 6 , suggests that irregularities in the hypothalamo-hypophyseal-gonadal axis account for the disturbed reproductive functions in reeler mice. At a first glance, morphological alterations in the testes support this view. Cariboni et al. 6 found a reduction in the number of seminiferous tubules in male gonads. Hyperplasia of Leydig cells, a typical feature after insufficient GnRH secretion 31 , however, was not seen. Thus, the reduction in seminiferous tubule number is not likely to result from GnRH deficiency, but may result from local disturbances of spermatogenesis in reeler mice. Regarding the existence of GnRH-Rs, which we found in ovaries of both reeler and wild-type mice, and which have previously been reported as being present in granulosa cells 14,28 , a direct response of ovarian tissue and granulosa cells to GnRH might be assumed. In contrast to previous reports 15, 16 , however, this was not the case, but reelin that was found to have an effect on aromatase expression in granulosa cells.
Consistently, estrus cyclicity, which we tested by monitoring the estrus cycles of reeler mice, is disturbed in reeler. Estrus cyclicity widely depends on cyclic synthesis of progesterone and estradiol, the latter being preferentially synthesized in granulosa cells. Our data show that the average cycle length is prolonged, and as a consequence, the frequency of ovulations must be reduced. Increased synthesis of LH, as shown by Lombadero et al. 32 , which may represent a feedback response to low estradiol levels, is obviously not able to compensate for reduced estradiol synthesis due to lack of reelin in the ovaries of reeler mice. Hence, reelin deficiency in the reeler mutant and lack of reelin-induced aromatase expression very likely account for irregular estrus cyclicity and may contribute to impaired fertility.
Reelin signaling to induce estrogen synthesis. This is the first study to show that reelin influences estradiol synthesis. However, reelin is an ECM protein, and numerous earlier studies have shown that ECM proteins are able to influence steroidogenesis in the gonads and in gonadal cells in vitro [33][34][35][36] ; for review see 37,38 .
It was recently shown that reelin signaling promotes granulosa cell proliferation in chicken ovarian follicles 19 via activation of ApoE2 and VLDR receptors and subsequent dab1 phosphorylation. Ovarian follicle development, however, is also highly dependent on estradiol synthesized in granulosa cells 23 . Since reelin stimulates estradiol synthesis, a synergistic action of reelin and estradiol in follicle development is probable. Follicle development is largely a function of granulosa cell proliferation and differentiation. Since aromatase is downregulated in single granulosa cells of growing follicles in reeler, as compared to WT animals, and reelin increases aromatase mRNA and protein, the proliferative effect of reelin appears to be mediated by enhanced aromatase expression and estradiol synthesis.
Regarding the expression pattern of reelin, it is seen in granulosa cells but not in corpora lutea; this confirms previous studies by Fayad et al. 39 and Eresheim and coworkers 19 . It appears that reelin plays a role in follicle development by stabilizing the follicle. Reelin binds to ApoE2R and VLDR receptors, which results in the phosphorylation of dab1, an adapter protein interacting with the cleaved Notch intracellular domain 19,[40][41][42] . Notch signaling is required for follicle development and fertility and is, similar to reelin, expressed in the cells of growing follicles. Several lines of evidence indicate that Notch signaling in the ovary is necessary for granulosa cell proliferation, as shown for reelin 19 and normal follicle development [43][44][45][46][47] . In the context of steroidogenesis, however, it was shown that Notch signaling stimulates progesterone production in luteal cells 48 , while steroidogenesis in granulosa cells of small follicles is suppressed, and it was proposed that Notch signaling inhibits FSH-induced steroidogenic gene expression 49 . In contrast, in this study we found that aromatase mRNA is upregulated in granulosa cells, which results in enhanced estradiol secretion, suggesting that reelin-induced Notch signaling is unlikely to participate in the control of estradiol synthesis in granulosa cells. Alternatively, dab1 phosphorylation induced by reelin binding to ApoER2 and VLDLR activates PI3K and LIM kinase1, which in turn results in the phosphorylation of cofilin 50 . Phosphorylation renders cofilin unable to depolymerize F-actin, and active cofilin/ADF is involved in ovary development and oogenesis [51][52][53] . Karlsson et al. 54 showed that active cofilin is required to initiate luteinization, thus the transformation of the follicular wall into a corpus luteum, which is paralleled by a rise in progesterone secretion and substantial rearrangement of cytoskeleton 47 . This rearrangement of the cytoskeleton includes activation (dephosphorylation) of cofilin. Importantly, inactive cofilin caused a 70% decrease in LH-stimulated progesterone secretion into the supernatant. Based upon these results, it seems likely that reelin-induced phosphorylation of cofilin in growing follicles prevents luteinization until the LH surge is triggered by increasing estradiol synthesis. The finding that ApoER2/ LRP8, which binds reelin, is maximally expressed in granulosa cells of dominant follicles and is downregulated upon hCG/LH stimulation, supports this hypothesis 21 . In addition, the downregulation of ApoER2 is paralleled by decreased aromatase expression 55 . Finally, ApoE deficient mice, mutants that lack the natural ligand of the ApoER2, show decreased aromatase expression 56 . Hence, reelin by phosphorylating cofilin, likely counteracts luteinization by stabilizing granulosa cells.
As shown in this study, estradiol and reelin may act synergistically by converging at the same intracellular cascades, thus producing similar effects. Thus, expression of both reelin and aromatase is highest during proestrus of WT mice. Estradiol and reelin share common signaling molecules such as Notch 5,57 , cofilin 58,59 and PI3K/ GSK3 60,61 . In this study we provide evidence that reelin increases aromatase expression and estradiol synthesis, hence pointing to a positive feed-back mechanism, which needs to be analyzed in detail in future comparative studies on the ovaries and hippocampus.

Material and Methods
Animals. Reeler mice (B6C3Fe-a/a-relnrl; Jackson Laboratory) and sibling wild-type (wt) mice were reared in accordance with the animal care guidelines of the University of Hamburg. Genotypes were verified by PCR analysis of genomic DNA. Ovaries of 12-week-old reeler mice and sibling wild-type mice were used for real time PCR, western blot and for immunostaining of tissue sections. Twelve-week-old Wistar rats (Institute of Anatomy, Hamburg, Germany) were maintained under controlled conditions; water and food were available ad libitum and taken for in situ hybridization experiments. All experiments were carried out in accordance with the German law on the use of laboratory animals (German Animal Welfare Act; project number ORG 470 by the "Amt für Verbraucherschutz, Lebensmittelsicherheit und Veterinärwesen"). All methods described below were accomplished in accordance with the relevant guidelines and regulations. Estrus cyclicity is mirrored by changes of the cellular composition in vaginal smears 24,25 . The cyclic stage was determined by a vaginal smear that was stained using standard procedures 62,63 . When comparing reeler with wild-type animals, only animals in the proestrus stage were taken for experiments. Firstly, these animals show the highest expression of aromatase in the ovaries 24 and secondly, these animals therefore possess a comparable hormonal status (Figs 2, 4). Sexual maturity and cyclic regularity were controlled, at the earliest, 8 weeks postnatal, again by vaginal smear.
Cell culture. The human ovarian granulosa-like tumor cell line, designated KGN 64 , was taken for aromatase mRNA studies or western blot experiments. KGN cell cultures were cultured as described by Nishi and colleagues 64 with DMEM/F-12 (5%) (HAM), Pen-Strep (1%) and incubated at 37 °C and 5% CO 2 in a humidified incubator. The first experiment was performed after three times passage of KGN cells, and the cells were then constantly kept in culture for following experiments. For each experiment, cells were seeded initially with a ratio of 1:200.000 cells per dish in a 6-well plate and stimulation was done on DIV 3.
Primary granulosa cell cultures were prepared from 12-week-old female mice. The ovaries were removed, cut open and follicles picked out without discrimination of developmental age. The follicles were washed in PBS and then digested with 1 mg/ml collagenase type IV (Sigma) for 15 min at 37 °C in DMEM, 10% FCS and Pen-Strep (1%). Afterwards, the granulosa cell suspension was centrifuged, washed with PBS three times and resuspended in DMEM, 10% FCS, Pen-Strep (1%) and seeded in six-well dishes. On DIV 3, stimulation with reelin was performed for 24 h for estradiol determination in the supernatant, or stimulation on DIV 4 for 2 h with reelin for aromatase RNA quantification.
Cell cultures stimulated with GnRH1 (Sigma-Aldrich, USA) were used in a final concentration of 10 −7 M in the culture medium.
Immunohistochemistry. After removal, the ovaries were placed in liquid nitrogen and cut frozen. 12 µm sections were cut on a cryomicrotome (HM560) and fixed afterwards with acetone and subsequently, after washing, with 4% paraformaldehyde. Fixation, as well as the following steps, were performed on a microscopic slide inside a humid chamber. Tissue sections were blocked with 10% NGS-PBS (Sigma-Aldrich, USA) for 30 min, ScIentIfIc REPORtS | (2018) 8:8722 | DOI:10.1038/s41598-018-26928-x washed, and then incubated overnight in the primary antibody dissolved in a 10% NGS-PBS solution. After washing with PBS, secondary antibody binding (dissolved only in PBS) was performed for one hour on a shaker at room temperature. Finally, after washing, the nuclei were stained with Dapi (4′,6-Diamidino-2-phenylindol) and cryosections were mounted in DAKO (Agilent Technologies, USA). Images were captured with a confocal Zeiss LSM (Axiovert 100, MikroSystems LSM) by using a Zeiss Plan-Neofluar (40x, numerical aperture 0.75) objective lens. Altogether, we analyzed 47 optical slices from (in the mean) 8 different preovulatory follicles of 3 reeler mice and (in the mean) 38 optical slices from (in the mean) 7 follicles of 3 different wild type mice. Quantification of signal intensity (densitometry) was achieved using ImageJ software (NIH, USA). For calculation of relative values, the mean of the wild-type values was set at 1 and the values determined in the reeler were related to it and presented as x-fold values of control.
Primer. For a qualitative detection of reelin mRNA in mice ovaries, mice-specific TaqMan probes (Thermo Fisher, USA) for reelin (RELN; Mm00465200) were used.
Radioimmunoassay for estradiol. KGN cultures were preincubated for 15 min with 10 −7 testosterone (Sigma) as a substrate for granulosa cell aromatase. Culture supernatants of the medium treated with reelin and mock-treated KGN cultures were collected after 24 h of treatment. 5 ml of the culture supernatant were collected for each condition. A commercially available radioimmunoassay kit (DIA Source: E2-RIA-CT KIP0629) was used for measurement of estrogen (E2). For RIA, 50 µl of culture supernatant were applied and E2 concentrations were calculated from the results of these measurements and presented as x-fold values of control. RIA was analyzed on a gamma counter (RIASTAR). To calculate the relative values, the mean of the steroid concentrations determined in the medium collected from control cultures was set at 1 and the values determined in the treatment groups ScIentIfIc REPORtS | (2018) 8:8722 | DOI:10.1038/s41598-018-26928-x were related to it and presented as x-fold values of control. The mean value for estrogen determined in the supernatant of the KGN cells was in the range of 581 pg/ml under control conditions (n = 9) and 699 pg/ml after reelin stimulation (n = 19).
ELISA for estradiol and testosterone content determination. Primary granulosa cell cultures from 12-week-old mice were preincubated for 15 min with 10 −7 testosterone (Sigma) as a substrate for granulosa cell aromatase. Culture supernatants of the medium treated with reelin and mock-treated granulosa cell cultures were collected after 24 h of treatment. 2 ml of the culture supernatant were collected for each condition. For determination of estrogen or testosterone in the supernatant, a commercially available ELISA kit (17β-estradiol Saliva ELISA RE52601, IBL, Hamburg, Germany; Tetosterone Saliva ELISA RE52631, IBL, Hamburg, Germany) was used. Photometric analysis was done with the ELISA-Reader Dynatech Laboratories MRX (Dynex-Technologies, Denkendorf, Germany).

Statistics. The results
show mean values ± SEM. Means were tested for normal distribution and variance homogeneity. Depending on the experimental group, the unpaired t-test was chosen using SPSS software. In the case of non-normal distribution of the raw data, the Mann-Whitney-U-Test or Kruskal-Wallis ANOVA, with a following Bonferroni post-hoc test, was used for statistics. For all statistical test procedures, the level of significance was set at p < 0.05.