Late fetal and prepubertal ovarian morphology
The foregoing described the morphological development and differentiation of the ovaries of the African elephant and the numbers of small follicles they contain, from 11 months of gestation to a prepubertal age of 9 years. Although each measurement and description of the ovaries were independent observations made in different individuals, the pattern of change over time is suggestive of development and in the absence of longitudinal studies provides the only currently available data.
At 11 months gestation very few naked oogonia remain identifiable within the cortex of the fetal ovary and nearly all the germ cells are present as meiotic oocytes arrested at the dictyate stage of prophase I  and surrounded by a single layer of granulosa cells which vary in shape from flat to cuboidal. These follicles within the cortex appear to form the follicle reserve for future reproductive life. The many follicles observed within the medulla of the elephant fetal ovary do not share this future and are destined for growth, beginning around 12 months gestation, to small antral stages before the onset of atresia which may occur at any developmental stage thereafter up to mid antral size. Fetal ovarian weight and volume increase from 11 months with the onset of interstitial cell hyperplasia and the growth of medullary antral follicles. By 15 months, as described previously by Allen et al. (2005), the increase is significant. The maximum number of antral follicles is observed around 16 months following which they decline to occupy <2% of ovarian volume at birth. Meanwhile, the volume of interstitial tissue continues to grow toward a maximum point thought to occur just prior to birth (22 months), although this could not be confirmed in the present study as the oldest set of fetal ovaries collected was at 20.2 months of gestation. Loss of small follicles and pre-antral follicles has been observed in other mammalian species during fetal life . The reasons for this loss are speculated upon [25, 26] and are now generally referred to as natural wastage. The present findings suggest, however, that the elephant fetus may employ its medullary follicles for a useful purpose; ie the production of greater amounts of interstitial tissue as discussed below. In turn, the rising steroid production within the fetal ovary may halt the further development of antral follicles.
How and why the small follicles within the cortex are protected from recruitment to pre-antral growth and the fate of atresia is not known, nor is it known in other species what causes activation of some primordial follicles and not others during fetal and post natal life . There is, however, evidence that either primordial follicles  or the local environment  may be involved. What seems likely from this elephant study is that recruitment is intimately related to local environment, this being stimulative in the medulla and restrictive in the cortex during the second half of fetal life.
During postnatal life a remarkable feature of the elephant ovary is the persistence of interstitial tissue within the medulla, otherwise it develops in a manner similar to the bovine  and other mammals.
Interstitial cells within the mammalian ovary are not static components . They differentiate from stromal fibroblasts, either independently (primary) or when they become associated with growing follicles [2, 32] and they revert to this cell type, either independently or following follicle atresia .
In equids, interstitial cells in both the fetal testis and fetal ovary begin to multiply rapidly in the absence of any follicle or seminiferous tubule growth from around day 80 of gestation. This interstitial cell hyperplasia and hypertrophy continues unabated to days 220–250 of gestation when the fetal ovaries weighing 50–100 g each are considerably heavier than the now-inactive maternal ovaries [19, 33]. The tightly packed epithelioid interstitial cells secrete large quantities of 19-carbon androgen precursors, including androstenedione and dehydroepiandrosterone (DHEA), [34, 35] and the very unusual 3-ß hydroxy-5,7 pregnanedien-20-one and 3-ß hydroxy-5,7 androstadien-17-one  which are then aromatised by the placenta to produce the relatively enormous (μg/ml) quantities of both phenolic (oestrone and oestradiol) and Ring B unsaturated (equilin and equilenin) oestrogens that are present in the blood and urine of pregnant mares during the second half of gestation [37, 38] and which appear to be important for growth and development of the very precocious equine fetus at birth . Similar interstitial tissue development results in ovarian hypertrophy in the Grey and Common seals, also without antral follicle formation . In both horses and seals the interstitial tissue declines rapidly in late gestation so that the gonads have shrunk to their normal prepubertal size at birth [19, 20, 40].
As described by Hanks  and Allen et al., and confirmed in the present study, gonadal enlargement also occurs in the elephant fetus during the second half of gestation due to a similar hyperplasia and hypertrophy of primary interstitial cells augmented by secondary interstitial tissue that persists following atresia of antral follicles. And as also demonstrated in the present study, these hypertrophied primary interstitial cells stain for the steroidogenic enzyme, 3-β hydroxysteroid dehydrogenase (3β–HSD) which indicates that they are capable of synthesising progestagens. Indeed, Allen et al. demonstrated that slices of elephant fetal gonad incubated with tritium-labelled cholesterol or pregnenelone secreted appreciable quantitites of 5-∂ dihydroprogesterone (5-∂ DHP) and other 5-∂ pregnane derivatives into the culture medium. More recently, Yamamoto et al. demonstrated the secretion of placental lactogen (elPL) by elephant trophoblast tissue and speculated that this chorionic hormone may be the essential luteotrophic stimulus for the enlargement of the fetal gonads and their synthesis of progestagens to assist the accessory corpora lutea in the maternal ovaries to maintain the pregnancy state. The highly vascularised nature of the fetal ovary from 18 months of gestation to term would help to transport the progestagens being synthesized by the interstitial tissue to the fetal, and hence to the maternal circulation to boost the supply of progestagens for pregnancy maintenance .
An unexpected and interesting finding in the present study was the persistence of 3β–HSD positive nests of interstitial cells in the ovaries of female elephant calves after birth. Such steroid-secreting tissue accounted for some (30–40%) of the total volume of the ovary during the first 6 months of life and it declined slowly thereafter to disappear completely only at around 4.5 years of age. It seems reasonable to speculate that continued secretion of progestagens by these interstitial cells during early post natal life may act locally to suppress any significant growth of antral follicles during the period, and indeed, very little antral follicle development occurs in the first 1–2 years of life in the elephant calf. A few antral follicles develop later in the second year of life, coincidentally with the disappearance of the interstitial tissue and the rate of antral follicle growth increases markedly from 4.0 to 4.5 years onwards when the whole ovary begins to increase in size. Further research is planned to explore the steroid output of these post-natal cells.
Also of interest was the 3β–HSD positive staining of the granulosa cells of small follicles from 16 months of gestation till the oldest studied sample at 5 years of age (Figure 7e) as indeed do the Sertoli cells of the testis at the same ages (F J Stansfield and W R Allen unpublished data). Light staining for 3β–HSD has also been described in pig , sheep  and human  fetal ovaries. It is highly irregular to obtain such a precise and specific staining for 3β–HSD in the granulosa cells of SF during what is considered to be a gonadotrophin independent stage of follicle development ). In the current study, the granulosa cells of follicles beyond the secondary stage of development did not stain positively. It is normal in mammalian ovaries for the production of androgens to take place exclusively in the theca cells of the developing follicle  and theca cells are usually first observed when the follicle has >2 rows of granulosa cells. In the rat these theca cells are capable of producing steroids just prior to antrum formation [49, 50]. In this study 3β–HSD stained theca cells of small antral follicles starting around 400 μm in diameter were observed from 16 months of gestation to 5 years of age.
The stereological measurement of the numbers of small follicles in the ovaries of 29 prepubertal elephant calves and 6 late gestation fetuses revealed a wide between-animal variation in numbers (range 303 084–2 456 741) during the period; the close agreement in follicle numbers between the two ovaries of each animal gave great confidence as to the validity of the counting method . Natural variation in small follicle numbers between animals of similar age is commonly observed in many mammalian species. For example, Schmidt et al. revealed variations of more than two orders of magnitude following cortical biopsies of human ovaries and Hansen  showed similar results in a review of small follicle numbers in 122 women from birth to 51 years of age.
The number of small follicles fell during late fetal life, as observed in many mammalian species . These follicles were evenly distributed between the left and right ovaries as observed previously  and follicle density decreased coincidentally with the age-associated increase in cortical volume of the ovary (Figure 1d). No true primordial follicles were observed so, as in the previous studies [6, 54], the small follicle pool was taken as being composed basically of early primary and true primary follicles.
The unexpected finding that the numbers of early primary follicles and the numbers of small follicles in calves aged 4.5–9 years was higher than those in younger calves raises a number of interesting questions as to how such an unusual situation might occur. Firstly, it should be acknowledged that, given the small number of calves included in the study and the large variation in follicle numbers among individuals, the difference, although statistically significant, may have been due to a Type 1 statistical error, brought about by coincidentally using a group of younger calves with lower numbers of follicles and (or) coincidentally using a group of older calves with higher numbers of follicles. Other possible explanations include genetic diversity and post-natal oogenesis. The additional possibility of error in calculating volume was also examined closely.
Relatively few animals could be included in the study (n = 6 fetuses and 29 calves) and biological variation may indeed have accounted for the apparent increase in numbers of follicles counted after birth [55, 56]. The selection of the family group to be shot was based on their presence in the culling area on the day, there was no prior knowledge of the group and all members of the group were killed. The chosen culling area was based on high elephant numbers and was deemed to be a random selection of the female family groups within SVC. It is also noteworthy that nearly all of the original 670 elephant which constituted the founding population of the SVC were introduced during 1991 and 1992 from a closed population in nearby Gonarezhou National Park and no further translocations have been made during the ensuing 20 years while population size has increased by natural breeding alone to an estimated 1 500 animals. Predominantly family groups were introduced originally to the Conservancy so the population is biased towards females with few mature mating bulls. Knowing that in women the age of menopause, brought on by exhaustion of the follicle reserve is highly heritable , the degree of genetic diversity of the elephant population sampled could be less than that of a truly wild population.
With regard to the experimental protocol, the calculation of the volume of the cortex may be suspected as a source of error to produce the increasing small follicle numbers. In mammals, the division between the cortex and the medulla of the ovary is fluid  and the line of differentiation is set halfway between the occurrence of follicles in the cortex and the blood vessels in the medulla . The volume calculations for the cortex and other compartments within the ovary in the present study of fetal and calf ovaries to 4.5 years of age were made using Cavalieri’s Principle as discussed by Gundersen  and have been described previously . These calculations were made on the same slides from which the follicle densities were obtained.
The division between the medulla and cortex is particularly blurred during fetal life. In the youngest fetal ovaries in which follicle counts were made in the present study (15.2 months gestation) the cortex formed a very narrow and distinct band just under the surface epithelium of the ovary, which then contained only small follicles. The medulla was composed predominantly of interstitial tissue and growing small follicles and antral follicles; all follicles would normally be included in the cortical region of the ovary of a calf or adult. The percentage volume of the cortical region containing the small follicles was similar in the fetal and calf ovaries (Figure 1b) despite the very considerable fall in ovarian weight and volume during the neonatal period. This suggests strongly that the cortical volume measurements were accurate in both developmental stages. Post-natally, it is generally accepted  that there is no further development of new antral follicles in the medulla and all subsequent follicles are derived from the cortical reserve.
The third possibility of some form of post natal oogenesis occurring in the elephant is particularly intriguing. As mentioned in the Introduction, the localization of oogonial stem cells (OSC) of adult mouse and human ovaries [4, 60] and the birth of pups following transplantation of female germ-line stem cells to the ovaries of irradiated mice  has raised significant, although still controversial, doubts about the finality of meiotic arrest in the ovaries of all mammals during fetal life. It remains possible that OSC may persist in, or migrate to, the epithelium covering the outermost surface of the cortex of the elephant ovary during early post natal life which could, due to some hitherto unknown stimulus, multiply mitotically within the cortex before entering meiotic arrest as in fetal life prior to acquiring an outer layer of persisting granulosa cells to form new SF and so boost the reserve of these structures.