C57BL/6 mice bred in our animal facility were housed under controlled photoperiod conditions (lights from 08:00 to 20:00 h) with commercial food and tap water supplied ad libitum, as previously described 1 2 . The day after overnight mating was counted as 0.5 day post-coitum (dpc). The animal facility is licensed by the French Ministry of Agriculture (agreement N°B92-032-02). All animal experiments were supervised by Pr. René Habert (agreement delivered by the French Ministry of Agriculture for animal experimentation N°92-191) in compliance with the NIH Guide for Care and Use of Laboratory Animals.

In this paper, we focussed on the ontogenesis of cytochrome P450 aromatase (accession # NP_031836 for protein and NM_007810.3 for mRNA), and named “aromatase” thereafter. To characterize aromatase expression during mouse testis development, its mRNA level was analyzed by RT-PCR from 12.5 dpc to birth. Three transcript variants, called T1, T2 and T3, were detected (Figure 1A-B). T1 and T2 were observed as early as 12.5 dpc. Sequencing confirmed that they were bona fide aromatase variants and showed that they corresponded to full-length aromatase (T1) and to the splicing variants without exon 3 (T2) and without exons 3 and 4 (T3) (see Additional file 1).

In order to quantify these variants during embryogenesis, we performed quantitative real-time RT-PCR (qRT-PCR) using the TaqMan method which allows the detection of each isoform using three different internal probes each one specific for one isoform (Figure 1A and Table 1 for sequences and conditions). We showed that T1 and T2 expression in the testis increased by around 20-fold between 13.5 and 17.5 dpc and thereafter started to progressively decrease (Figure 1C).

To determine if aromatase is translated in mouse testis, western blot analysis was performed using a specific anti-aromatase antibody (MCA2077T, Serotec, France) (Figure 1D). Two proteins around 54 kDa and one around 27 kDa were detected. The protein of 54 kDa was also present in the ovary extract and it approximately corresponds to the aromatase expected size. We thus suppose that the two heaviest proteins derived from the full-length form of aromatase (T1) with the highest form corresponding to a testis-specific post-translational modification that remains to be identified. In order to understand the origin of the 27 kDa protein, we analysed sequences of the T2 and T3 variants. It revealed that the splicing of exon 3 in T2 would change the ORF and create a precocious codon stop leading to a probably not detected protein of 6 kDa. Splicing of exons 3 and 4 in T3 would not change the ORF allowing in theory the synthesis of a truncated protein of 46 kDa. No protein at this expected size was detected in the western blot (Figure 1D). However, the use of an alternative start codon located later in T2 and T3 sequences may lead to a protein of 27 kDa containing the C-terminal part of aromatase.

These findings are different from those of the only previously published paper on this topic showing that, in the mouse, aromatase expression starts at 17.5 dpc and reaches the highest level at day 1 post-partum 3 . In our study, we detected aromatase expression as early as 12.5 dpc. This discrepancy probably results from the improvement of the methods of detection made since 1994. This is an important point because it shows that estrogens can be produced by mouse fetal testes very early and throughout development.

Our findings indicate that different aromatase transcripts are generated in fetal mouse testes. Previous studies in different mammalian species (including the mouse) reported that tissue-specific aromatase expression is driven by specific promoters 4 5 6 . Each tissue-specific promoter is associated with a specific untranslated first exon. In mice testis Golovine et al. have shown that aromatase transcripts may emerge from a specific promoter called Ptes 4 . Our study showed that aromatase expression is also regulated at a second transcriptional level generating two additional truncated variants T2 and T3 by mRNA splicing. Our results suggest that there are several forms of aromatase protein however the nature and the physiological function of these isoforms remain to be investigated.

Immunohistochemical analysis of aromatase localization in 17.5 dpc mouse testes using a specific anti-aromatase antibody (MCA2077T, Serotec, France) showed a strong staining in Leydig cells. Importantly, there was no detectable staining in Sertoli cells (Figure 2A). Conversely, previous studies in fetal and neonatal testes showed that aromatase was expressed in both Leydig cells and Sertoli cells in the rat 7 8 . This and other previous reports indicate that aromatase cell localization in fetal testis is quite variable in mammalian species. Indeed, aromatase is expressed in Leydig cells and not in Sertoli cells in the fetal testis of the Plains Vizcacha rodent 9 , is totally absent in the deer 10 , and is detected in both Sertoli cells and Leydig cells in fetal baboon and human testes 11 12 .

In addition, our immunohistochemical analysis showed that aromatase was also expressed in gonocytes, but the intensity of the signal was not uniform: in some cells the signal was very strong, whereas in others it was faint or undetectable (Figure 2A, arrowheads). Similar results were previously described for Retinoic Acid Receptor alpha 14 . As aromatase localization in germ cells was quite unexpected, aromatase immunostaining was also performed in enriched gonocyte cultures that were prepared from 17.5 dpc mouse testes as previously described 15 . Similarly, aromatase was detected in some germ cell VASA-positive cells, a germ cell-specific marker (Figure 2B). This result identifies a sub-population of gonocytes with endocrine function. Aromatase expression was previously reported in adult rat and human germ cells 16 17 and in pig gonocytes during development 18 . Aromatase expression was also detected in gonocytes of human fetal testes 12 .

In conclusion, aromatase cell localization in fetal testis appears to differ from one species to another and as consequence also the intracellular estrogen concentration. These differences should be taken into account to explain the variations in the susceptibility of fetal testis to estrogenic and anti-androgenic endocrine disruptors in different mammalian species that has been recently lightened 2 .