African elephants are distributed widely over the southern half of the African continent where, it is generally agreed, they exist as two main subspecies; (i) Loxodonta africana africana, the most common Savannah or Plains elephant that inhabits the northern, eastern and southern states, from Kenya through Uganda, Tanzania, Zimbabwe, Mozambique, Botswana, Angola to South Africa and also includes the tall and more slender desert elephant, which has adapted wonderfully to survive in the very harsh and arid conditions of northern Namibia; (ii) Loxodonta africana cyclotis, the shy and appreciably smaller forest elephant that survives in the dense rainforests of The Congo and other Central and West African states (Roca et al. 2001).

Although there is a complete paucity of published data on reproduction in forest elephant, it seems reasonable to postulate that, with the likely exception of the stable equatorial ambient temperature and rainforest food supply of their habitat ensuring a more uniform age at puberty, and intercalving interval, no other significant differences would exist in basic reproductive processes between L. africana cyclotis and its bigger and more numerous cousin, L. africana africana. Hence, while the data presented in this paper is derived solely from L. africana africana, it is assumed to be pertinent for both subspecies of African elephant.

Four major post-mortem studies of population dynamics and reproductive strategies were undertaken in East Africa during the 1950s and 1960s, when elephant culling was being practised widely to prevent overcrowding and maintain biodiversity in existing game parks. They began with the groundbreaking collection and painstaking dissection of reproductive organs by Dr John Perry, working in Uganda (Perry 1953, 1964, 1974; Amoroso & Perry 1964) and were soon followed by extensive and meticulous surveys carried out by Dr Richard Laws, working with Dr Ian Parker and others, on three separate populations of elephants in Uganda, Kenya and Tanzania (Laws 1966, 1967a,b, 1968, 1969a,b; Laws & Parker 1968). Around the same time, Dr Ian Buss collected reproductive organs from elephant culled in Uganda (Buss & Savidge 1966; Buss & Smith 1966; Buss & Johnson 1967; Johnson & Buss 1967a,b) and Professor Roger Short and his then graduate student, Dr John Hanks, worked on culled elephant in Kenya and Zambia (Short & Buss 1965; Short 1966; Short et al. 1967; Hanks 1971, 1972; Hanks & Short 1972).

These pioneering investigations established the benchmarks of reproduction and population dynamics in the African elephant that have been supported and added to since by the long term behavioural observations of Cynthia Moss and Joyce Poole upon the closed, but expanding, herd of elephant in Amboseli Park in southern Kenya (Moss 1983, 1988; Poole 1989, 1996). More recently, Professor J. K. Hodges and colleagues carried out a range of in vitro endocrinological investigations utilizing tissues collected from elephant culled in Kruger National Park during the Early 1990s (Hodges et al. 1994, 1997; Heistermann et al. 1997a,b; Meyer et al. 1997; Greyling et al. 1997, 1998; Hodges 1998). Simultaneously, a number of valuable physiological and endocrinological observations and measurements have been made on oestrous cyclicity and pregnancy in captive African and Asian elephants in zoos (Ramsey et al. 1981; McNeilly et al. 1983; Brannian et al. 1988; Plotka et al. 1988; Brown et al. 1991; Taya et al. 1991; Niemuller et al. 1993, 1997; Olsen et al. 1994; Brown & Lehnhardt 1995; Kapustin et al. 1996; Meyer et al. 2004) and Professor Raman Sukumar has given a detailed account of the evolution, ecology and conservation problems associated with both Asiatic and African elephants (Sukumar 2003).

In his original post-mortem study, Perry (1953) calculated that young African elephants in the wild reached sexual maturity between 8 and 12 years of age (mean of 11 years) and males tended to pass through puberty 1–2 years after females. He concluded that mature females were polyoestrous, but showed no evidence of a breeding season, and, under normal conditions, exhibited a mean calving interval of 3.8 years. However, Laws (1969a), studying organs from five separate populations of elephant culled in Uganda, Kenya and Tanzania, with densities ranging from 2.5 to 10 elephant per square mile and experiencing big seasonal differences in rainfall and green feed, reckoned that both the age of sexual maturity and the intercalving interval were very density dependent. His mean age of puberty ranged from 12.5 to 14, and then on to 18, years with increasing population density in the three main populations studied, while the mean intercalving interval varied between 6.8 and 8.9 years. He also noted a definite seasonal pattern of reproduction, with biannual peaks in conceptions associated with the two rainy seasons, one of which appeared to be retarded in the high-density population in association with nutritional deficiencies. Furthermore, he detected definite ‘cycles of recruitment’ to the population at intervals of 6–8 years which correlated well with the level of rainfall at the time of the preceding conceptions. He concluded that ‘density dependent natural regulatory mechanisms are in operation in elephant populations, which involves changes in reproductive rate’.

Turning to ovarian function, Perry (1953) noticed that the ovaries of sexually mature elephants that were not in lactation anoestrus almost invariably contained more than one corpus luteum (CL). From this he concluded that the elephant was polyoestrous and polyovular. Furthermore, from his finding of multiple luteal structures of varying sizes in the ovaries of pregnant animals, he suggested that ‘the corpora lutea (CL) are replaced about mid-pregnancy by a second set which are formed by luteinization of all follicles with antra in both ovaries; some of the larger ones ovulate while many smaller ones do not’. However, Laws (1969a) disputed these interpretations. By serially slicing the ovaries of 109 pregnant females he noted that the actual number of CL ranged from as low as 2 to as high as 42, with a slight downward trend in number during gestation, from around 12 in early pregnancy to 9 near term. By measuring the cross-sectional area of the CL he calculated that the total mean weight of luteal tissue in both ovaries ranged from 19g in early pregnancy, up to 38g in mid-pregnancy and then down again to 22g near term. Thus, Laws (1969a) observed that the amount of luteal tissue in both ovaries remained relatively steady at around 15–40g throughout gestation and with an almost complete absence of mature-sized follicles (5–7.5mm diameter) in the ovaries in mid-pregnancy. He concluded that ‘the elephant in the wild is seasonally polyoestrous and polyovular, although it produces large numbers of small luteal structures in early pregnancy by luteinization of unovulated follicles’.

Short (1966), on the other hand, had the rare opportunity to post-mortem one mature female elephant a few days after she had been seen to be mated. He found one recent CL with an ovulation stigma and three older CL. Having observed as many as seven CL in the ovaries of another non-pregnant elephant he concluded that … ‘the elephant is a polyoestrous monovular animal that undergoes a number of sterile cycles before coming pregnant. These appear to be ovulatory cycles, judging by the appearance of ovulation stigmata, but it is not known if they are accompanied by overt signs of oestrus’. In response, Laws (1969a) summarized the available evidence and suggested that, in the wild, ‘the elephant is either: (i) polyoestrous and polyovular, occasionally monovular; (ii) polyoestrous and polyovular, but also produces large numbers of accessory CL by the luteinization of unruptured follicles or (iii) polyoestrous and monovular but forms large numbers of accessory CL’.

Moving on 30 years, reproductive interest in the African elephant in the 1990s concentrated more upon endocrinological aspects of the oestrous cycle and pregnancy. Serial serum progestagen and gonadotrophin assays carried out during a total of 38 oestrous cycles monitored in adult captive African elephants demonstrated that the whole cycle lasted 13.0–14.1 weeks and was made up of 4–5 weeks of follicular phase followed by 8.5–9.1 weeks of luteal phase. Two distinct surges in serum LH concentrations, three weeks apart, occurred during the follicular or inter-luteal phase, only the second of which resulted in ovulation and the expected rise in serum progestagen levels (Kapustin et al. 1996).

Hodges (1998) addressed the longstanding question of whether the elephant is monovular or polyovular. He recalled that the actual number of CL in the ovaries of wild females did not differ markedly between the pre-conception oestrous cycle and pregnancy and, based on the published figures of Smith & Buss (1975), de Villiers et al. (1989) and his own group (Heistermann et al. 1997a,b; Hodges et al. 1997), he suggested that 6–8 ‘luteal structures’ per animal would be a reasonable average. He discounted Short's hypothesis that the multiple CL encountered in pregnant and non-pregnant elephants might reflect accumulation of luteal structures from cycle to cycle in order to achieve a critical mass of productive luteal tissue to support pregnancy. He further argued that no evidence existed for any progressive increase in serum progestagen concentrations during successive non-fertile cycles and nor did serum progestagen concentrations correlate with the mass of CL tissue present in the ovaries (de Villiers et al. 1989; Hodges et al. 1997). The low incidence of non-pregnant, non-lactating adult female African elephants in the wild (Laws 1969a) and the high conception rates following observed matings (Moss 1983), seemed to argue against a lengthy process of accumulating luteal tissue in sequential 14 week ovarian cycles and Hodges (1998) therefore concluded that the formation of multiple CL, with and without ovulation stigmata, probably occurred in each cycle, with structural, but not functional, persistence into subsequent cycles as the most likely explanation of the visible evidence.

The other major contribution by Hodges and his colleagues in the 1990s was to confirm the original findings of Short & Buss (1965) and Smith et al. (1969) of the complete absence of progesterone in the peripheral blood and luteal tissue of cycling and pregnant elephants and its replacement by 5α-reduced progestagens, especially 5α dihydryoprogesterone (5αDHP). This important finding, coupled with the experimental observations of Meyer et al. (1997) and Greyling et al. (1998) that 5αDHP binds avidly to the endometrial progesterone receptor in the elephant, established the interesting likelihood that, as in the mare and other equids during the second half of gestation (Hamon et al. 1991; Holtan et al. 1991), 5αDHP and not progesterone, constitutes the biologically active progestagen which supports the pregnancy state in this species.

In terms of luteotrophic support during pregnancy, McNeilly et al. (1983) detected FSH and LH immunoreactivity in peripheral serum samples taken from pregnant wild African elephants at all stages of gestation, but the levels did not differ from those recorded in non-pregnant animals and nor was there any clear pattern related to gestational age. They, together with Hodges et al. (1987) and Brown & Lehnhardt (1995), did however, detect definite increases in serum prolactin concentrations from around 4–6 months of gestation in pregnant African and Asian elephants. Since prolactin was known to be luteotrophic in many other species, Hodges (1998) concluded that it might well perform the same function in the pregnant elephant, although he noted that the sharp rise around 4–6 months might be too late to prolong luteal lifespan beyond that of the cycle. Furthermore, there occurred no marked changes in the number, size or general appearance of the ovarian CL coincident with the prolactin rise. Thus, both the source and the precise role of prolactin during gestation in the elephant remain to be determined.

Most recently, in an excellent study, Meyer et al. (2004) measured the profiles of progestagens, prolactin, relaxin and cortisol in serial blood samples recovered throughout pregnancy from 19 Asian and eight African captive elephants maintained at zoos across America. They confirmed the marked rise in serum prolactin levels from around the fifth month of gestation as noted originally by McNeilly et al. (1983), recorded a similar rise in serum relaxin concentrations, also from around the fifth month, showed a dramatic rise in serum cortisol concentrations coincident with an equivalent fall in serum progestagen levels during the last 5–8 days before birth and, interestingly, showed a remarkably steady rise in serum progestagen concentrations from the time of ovulation which did not peak until as late as 5 months of gestation. Furthermore, they made the intriguing observation that, between 5 and 20 months of gestation, in Asian but not in African elephants, serum progestagen concentrations were significantly higher in mothers carrying male foetuses than in those carrying female foetuses. Finally, when viewing progestagen profiles in individual females, both Asian and African, there appeared to be a pronounced drop in levels around the end of the first month of gestation, followed by an equally pronounced secondary rise soon afterwards that continued unchecked until peak levels were achieved by month five (Meyer et al. 2004). This early gestation profile poses the fascinating question of whether the initial fall and secondary rise in progestagen concentrations around the end of the first month of gestation in the elephant is the result of one or more secondary or accessory ovulations or perhaps reflects a rather sudden enlargement and/or increase in steroid output by the existing multiple CL. And whichever mechanism is in play, what, if any, gonadotrophic stimulus is involved in the process and from where does it originate?

The anatomy of the elephant reproductive tract was described in detail by Perry (1964), Short (1966), Laws (1969a) and more recently by Allen et al. (2002a,b, 2003). Briefly, the uterus is bicornuate, with two relatively long horns that diverge laterally from each other about halfway along their length (figure 1a and figure 5 in the electronic supplementary material). The uterine body is very short and the posterior portions of the uterine horns are joined to one another by a short intercornuate ligament. In young non-pregnant adult elephants, the endometrium, myometrium and outermost serosa of the uterus are so collectively fibrous, tough and tonic that it is virtually impossible to insert a sharp knife or scalpel blade into the uterine lumen and slit open one horn along its length (see figure 6 in the electronic supplementary material). A transverse section across the whole uterus posterior to the lateral divergence of the two horns shows longitudinal folding of the endometrium to form a star-shaped uterine lumen (see figure 7 in the electronic supplementary material), as described originally by Amoroso & Perry (1964) and Perry (1974). Histological sectioning of the endometrium in this region demonstrates very tight apposition of the lumenal epithelia within the lateral clefts of the star (figure 1b), to give a picture more reminiscent of a rodent uterus than an animal as large as the elephant (Mossman 1987).

Figure 1 (a) The reproductive tract of a 4.5 month pregnant elephant viewed from the dorsal surface. The discrete conceptus bulge containing a foetus weighing 5.15g is situated at the lateral divergence of the right uterine horn. (b) Photomicrograph of (more ...)

The ovary of the adult elephant is very small in relation to its body size and it is fibrous and nodular in general appearance (figure 1c), not unlike that of the pig. In situ it is completely enveloped by a distinct serosal pouch which has a luscious red mucosa (figure 1c) and which constitutes the infundibulum of the oviduct incorporated in the ovarian bursa (Amoroso & Perry 1964). The mature pre-ovulatory Graafian follicle in the elephant ovary reaches a maximum diameter of only 0.5–1.0cm (see figure 8 in the electronic supplementary material; Hodges 1998) yet in every one of 58 pregnant uteri examined by the author over a 3 year period, from the very earliest stages of gestation to near term, between two and eight large, plum-like CLs measuring 3–6cm in diameter (figure 1d, and figure 9 in the electronic supplementary material) were always present, usually, although not invariably, clustered together on the one ovary situated ipsilateral to the gravid uterine horn (Allen et al. 2003). When sectioned they exhibited a homogeneous pale brown or khaki-coloured luteal parenchyma with relatively few fibrous trabeculae or other supportive elements (figure 2c). Thus, a marked discrepancy seems to exist between the small size of Graafian follicles in the ovaries of non-pregnant and pregnant elephants and the much larger, multiple CL that are always present in the ovaries of pregnant animals throughout gestation. The question of when and how these very large multiple CL of pregnancy arise remains an intriguing mystery.

Figure 2 (a) An opened conceptus sac that contained a 1.6g foetus (four months gestation; Craig 1984). Note the pressurized bulging of the pale, tough endometrium with the attached embryonic membranes. The pale outline of the progenitor placental band (more ...)

This position of the essentially spherical, fluid-filled conceptus in the uterus, which is itself suspended in the abdomen directly beneath the rectum, makes the pregnant elephant an ideal candidate for transrectal ultrasonography. Dr Thomas Hildebrandt and his colleagues from the Institute of Wildlife Biology in the University of Berlin have, over the past 10 years, successfully performed transrectal ultrasonographic examinations in over 300 Asiatic and African elephants in captivity, using a conventional linear array transducer attached to a ‘rigidly flexible’ extension arm (Hildebrandt et al. 1998, 2000a,b; Hermes et al. 2004). In addition to making very accurate diagnoses of pregnancy, they have also highlighted a worrying incidence of leiomyomata in the uteri of these captive females (Montali et al. 1997). Pathology of this type, combined with abnormal or absent oestrous cyclicity and a depressing array of behavioural stereotypies, no doubt underlies the very high levels of infertility exhibited by female elephants in captivity (Kurt & Khyne 1996; Schmidt & Khyne 1996; Schmid 1998; Taylor & Poole 1998; Wiese 2000; Brown 2000; Olson & Wiese 2000).

Between 4 and 6 months of gestation, when the foetus weighs 1–5grams, it is possible to discern the pale, ribbon-like thickening in the equatorial region of the ovoid conceptus that constitutes the developing placental band (figure 2a–c) and it is also possible to distinguish the division of the allantois into four separate compartments (figure 2b), as described originally by Amoroso & Perry (1964).

The placental band broadens as a result of lateral growth and it becomes increasingly reddish-brown in colour (figure 2c, and figures 11 and 12 in the electronic supplementary material). Usually, it forms an unbroken ribbon of tissue around the equatorial region of the conceptus, with uniformly brown-coloured lateral edges indicating the development of the blood-filled haemophagous zones between the placenta and the endometrium (figure 2c,d). However, some conceptuses show pronounced narrowing or even breaks of the placental band in some regions (see figure 11 in the electronic supplementary material). In the final stages of gestation, the placental band grows to as much as 12inches in width and 5–6inches in depth in the central region (see figures 13–19 in the electronic supplementary material). Its cut surface reveals well-defined areas of homogeneous reddish-brown placental parenchyma separated by fibrous trabeculae. The lateral edges of the band are still much darker brown in colour due to staining by the accumulated maternal blood in the haemophagous zone (see figures 12–19 in the electronic supplementary material; Allen et al. 2003).

A surprising feature of the gravid elephant uterus is the relative narrowness of the placental hilus that attaches the conceptus to the endometrium (figure 2c,d). The width of this vital connecting bridge is limited to the 1–2cm breadth of the original upgrowth of stromal villi from the lumenal surface of the endometrium during initial development of the placental band and it must carry the plexus of maternal endometrial blood vessels which vascularize the band (figure 2d). However, despite the relatively enormous expansion of the placental band throughout gestation, the hilus can gain little additional width and it remains an increasingly narrow channel or tract through which a growing concentration of ever enlarging uterine blood vessels must pass to reach the placental zone. This attachment pedicle can be easily severed using ordinary laboratory scissors which allows the conceptus to simply roll out of the uterus (figure 2c; Allen et al. 2003).

In recently post parturient uteri a pronounced, trench-like scar can be observed passing around the circumference of the previously gravid horn (see figure 20 in the electronic supplementary material). It clearly indicates the site of attachment of the placental hilus of the previous conceptus and, in the uteri of very aged females, it is possible to discern multiple parallel circumferential scars in the endometrium of both uterine horns, each indicating the attachment site of previous a pregnancy (figure 2e; Laws 1967b).

Figure 3 (a) High power photomicrograph of the trophoblast–endometrium interface in the every early stages of gestation (estimated 15–40 days, no embryo visible). A blunt projection of trophoblast cells (T) is forcing its way beneath the more darkly (more ...)

Figure 4 Diagrammatic representation of significant morphological changes that occur during the development of the zonary placental band on the elephant conceptus. (a) Replacement of the lumenal epithelium of the endometrium by trophoblast in the equatorial region (more ...)

By 3–6 months of gestation the placental zone is now visible to the naked eye as a pale, circumferential band, approximately 1–2cm wide, passing around the equatorial region of the still spherical conceptus (figure 2b). The upgrowths of trophoblast-invested endometrial stroma have now grown into a discrete line of simple frond-like protrusions of stroma, each still closely invested by its layer of trophoblast cells and all rising vertically upwards from the endometrial surface somewhat like a closely planted bed of new asparagus shoots (figure 3c). They begin to branch at their tips (figure 3d) and, at this early stage, an appreciable amount of endometrial stroma still persists between the investing trophoblast cells and the endothelium of the maternal blood vessels contained within the core of each villus (figure 4c). With advancing gestation, the stromal lamellae become thinner and more stretched in appearance thereby allowing ever closer apposition of foetal trophoblast to maternal endothelium (figures 3e (also figure 18 in the electronic supplementary material) and 4d). Coincidentally, the lengthening placental lamellae at the edge of the placental band lay over laterally towards the endometrium to form a blind cleft between the endometrial and placental surfaces (figure 4d, and figure 17 in the electronic supplementary material). The lumenal epithelium within these lateral clefts becomes convoluted and the epithelial cells themselves assume a tall columnar configuration with blunt pseudopodia protruding from their apical surfaces. Erosions of the endothelium of the maternal vessels in the region cause leakage of maternal blood into both the clefts and the core of the placental lamella (figure 4d). The trophoblast cells bathed in the extravasated blood likewise become taller and flocculent in appearance and they actively phagocytose the red blood cells and other blood components (figure 3f).

With advancing gestation the placental lamellae increase steadily in length, breadth and branching complexity, while coincidentally becoming progressively folded or pleated within the confines of the placental band, to maximize the surface area contact between trophoblast and maternal vasculature (figure 4e). The trophoblast comes into ever more intimate contact with the endothelium of the maternal vessels and these and their foetal equivalents, indent deeply into the trophoblast layer to reduce the interhaemal distance between the foetal and maternal circulations to only 2–3μm (Wooding et al. 2005).

The lateral spreading of the placental lamellae shows that the increasing size, complexity and exchange area of the placental interface stems solely from elongation, secondary and tertiary branching and, finally, tight folding of the original stromal upgrowths within their envelope of allantochorion rather than from any significant increase in either the number or the basic bulk of the primary endometrial protrusions (Allen et al. 2003). The width of the placental hilus attaching the conceptus to the endometrium hardly increases at all and, within the placental band, the trophoblast layer remains unicellular throughout, with no sign whatever of syncytial formation (Wooding et al. 2005).

In the maternal ovaries, 2–8 large plum-like CL, usually clustered on the one ovary ipsilateral to the gravid uterine horn (figure 1d), appear in the very early stages of pregnancy and persist throughout gestation (Smith & Buss 1975; Hodges et al. 1997; Hodges 1998; Allen et al. 2002a). These pregnancy CL are much larger than the maximum diameter of a preovulatory follicle (Allen et al. 2003) and their homogeneous luteal tissue contains high concentrations of 5α-dihydroprogesterone and other 5α-pregnane derivatives, but negligible amounts of native progesterone (Smith et al. 1969; Hodges et al. 1997).

The sudden appearance and persistence of multiple large CL in the maternal ovaries during pregnancy indicates the possibility of the production of some sort of placental gonadotrophin in the elephant. Furthermore, the enlargement of the foetal gonads during the second half of gestation in the elephant, which parallels similar foetal gonadal hypertrophy in equids (Cole et al. 1933; Hay & Allen 1975), likewise suggests the involvement of a gonadotrophin of possibly placental origin. It also indicates the possibility of fetoplacental cooperation in the synthesis of oestrogens, as demonstrated in the mare (Bhavnani et al. 1969, 1971; Bhavnani & Short 1973a,b; Cox 1975; Tait et al. 1983, 1985) and in women (Diczfalusy 1964). However, closer investigation has so far failed to support either possibility.

Minced placenta collected freshly from 11 elephants carrying foetuses ranging in gestational age from 5.5 to 20.6 months (Craig 1984) and incubated in vitro with ³H cholesterol, ³H pregnenolone and ³H androstenedione, failed to synthesize any progestagens, androgens or oestrogens. However, similar incubation of minced gonad recovered from six foetuses aged 13–21 months, but not from three younger foetuses, with ³H cholesterol or ³H pregnenelone, resulted in appreciable conversion of the labelled precursors to 5α-dihydroprogesterone and other 5α-reduced progestagens (Allen et al. 2002a). This was supported by strong immunohistochemical labelling of the interstitial cells of the enlarged gonads with an antibody specific for 3β hydroxysteroid dehydrogenase, the enzyme responsible for converting pregnenolone to progesterone and other progestagens (figure 26 in the electronic supplementary material). In parallel experiments, saline homogenates of fresh placenta recovered from five conceptuses aged 4.5–11.0 months (Craig 1984) submitted to a mouse oocyte recovery rate bioassay for total gonadotrophins (Adams 1982), a pig testis radioreceptor assay for FSH (Maghuin-Rogister et al. 1978) and an amplified enzyme-linked immunoassay for equine eCG/LH (Allen et al. 2002b), all failed to show any hint of gonadotrophic activity (Allen et al. 2002a).

Thus, like the zonary placentae of felids and canids, it seems highly unlikely that the elephant placenta secretes any sort of gonadotrophic hormone and the source and timing of the gonadotrophic and/or luteotrophic stimuli which result in the multiple CL of pregnancy and the enlargement and steroidogenic activity of the foetal gonads remains an intriguing mystery. The rise in serum prolactin levels that has been reported to occur in elephants between the fourth and sixth months of gestation (Hodges et al. 1987; Meyer et al. 2004) might well supply luteotrophic support to ensure persistence of the CL throughout gestation. But it is unlikely to be involved in the initial formation of these luteal structures or in the enlargement and steroidogenic activity of the foetal gonads.

The African elephant (Loxodonta africana) is easily the largest of the world's land-based mammals yet, curiously, embryological evidence of functional nephrostomes persisting during kidney development in utero (Gaeth et al. 1999), the existence of the elongated and highly mobile proboscis or trunk and the absence of a pleural cavity (Short 1962), all point to a relatively recent evolutionary emergence of the elephant from an aquatic ancestry (Gaeth et al. 1999).

The many well intentioned and well executed post-mortem studies performed on culled elephant in Africa over the past half century, combined with more recent endocrinological investigations carried out on living animals in zoos and small game parks, have established the basic parameters of reproduction in L. africana. The female passes through puberty between 10 and 18 years of age depending upon population density and nutritional availability (Perry 1953; Laws 1969a), she should show regular oestrous cycles of 14–15 weeks duration when unmated (Kapustin et al. 1996), she conceives readily a singleton pregnancy from the two or three matings she undergoes by the musth bull when she is in oestrus (Poole 1996), but she very rarely conceives twins, despite adequate room in her uterus to gestate twin conceptuses.

The young elephant embryo develops a zonary endotheliochorial placenta, first by eroding the lumenal epithelium of the endometrium and replacing it by trophoblast and then by inducing the underlying endometrial stroma and blood vessels to commence an ever increasing upward growth above the surface of the endometrium to form the highly crimped and folded lamellae that provide steadily increasing placental exchange throughout the 22 months of gestation. Typical haemophagous zones develop beneath the lateral edges of the placental band in which leaked maternal erythrocytes, ferric iron and no doubt many other essential metals, vitamins and minerals are actively taken up by specialized trophoblast cells in the region. And in the centre of the ever widening placental band, a remarkably narrow hilus, through which all the endometrial blood vessels traverse to vascularize the placenta, attaches the whole conceptus to the endometrium. Severance of this attachment pedicle at term, during or soon after calving, is associated with considerable intrauterine haemorrhage which may contribute to the 4 year or greater intercalving interval exhibited by most African elephants (Laws 1969a; Moss 1983; Allen et al. 2003).

Although architecturally straightforward and apparently hormonally inert (Allen et al. 2002a), the elephant placenta nonetheless leaves a number of puzzling questions concerning its wider role in both the induction and maintenance of the pregnancy state. For example, when and how do the large CL of pregnancy develop, why are they so much bigger and more numerous than the ovarian follicles from which they presumably originate, is there one or more secondary or accessory ovulations around the end of the first month of gestation as the recent results of Meyer et al. (2004) would seem to suggest, and do these pregnancy CL, either primary or secondary, derive any gonadotrophic or luteotrophic support from the placenta? Does the 5αDHP and other progestagens secreted by the enlarged foetal gonads during the second half of pregnancy add significantly to those secreted by the large pregnancy CL in terms of maintenance of the pregnancy state and what, if any, is the nature and source of the gonadotrophic stimulus which drives this foetal gonadal enlargement? Could it stem from activity of the foetal pituitary gland during the second half of gestation or might the increased rate of secretion of prolactin from around the fifth month of gestation (Hodges et al. 1987; Meyer et al. 2004), from the maternal pituitary gland or perhaps even the placenta, act in a quasi-gonadotrophic manner to bring about the observed changes? There is much yet to be learned about the physiology of pregnancy in this magnificent species.

With the high level of fertility existing in both the large mating bulls and the oestrous cows leading to high conception and pregnancy rates (Moss 1988), the 22 month gestation period in which early and late pregnancy losses seem to occur infrequently (Laws 1969a; Moss 1983, 1988), and a high postnatal survival rate combined with general longevity (Moss 1988), it seems that the African elephant, if not killed by its only predator, man, has the inherent fertility mechanisms to increase its population size by a remarkable 8% per annum (Whyte 2001). Inevitably, this high reproductive rate leads to overpopulation in secure and well-managed game parks with the accompanying remorseless loss of biodiversity as the elephants destroy the habitat, together with the risk of horrendous mass die-offs in times of drought. Yet, coincidentally and sadly, with the burgeoning human population and associated demand for arable land in other parts of Central and Eastern Africa, combined with the ever present pressure of illegal poaching in these less closely policed areas, whole populations of elephants are becoming increasingly endangered and could even face complete extinction in coming years (Sukumar 2003).

The resulting paradox of ‘too many here and too few there’ is not easily solvable by seemingly simple procedures, such as physical translocation of whole families of elephant or the creation of migratory corridors between overstocked and understocked parks. For example, darting and extensive translocation of almost 1000 elephant from Kruger National Park over the past decade has completely filled any requirements for elephant in smaller private game reserves throughout South Africa and neighbouring countries (Whyte 2003). Similarly, a recent attempt to remove much of the elephant-proof fence between Kruger Park and the neighbouring very understocked Limpopo National Park in Mozambique, followed by physical translocation of whole families of Kruger elephant into Limpopo Park, resulted in remarkably rapid migration of the elephant along the persisting fence and back to their home range in Kruger Park (Ian Whyte 2004, personal communication).

Since routine annual management culling of elephant ceased in Kruger Park a decade ago in 1995, the total elephant population in that park has risen from 8000 to just over 14000 (Rogers 2005). And in Hwange Park in northern Zimbabwe, where management culling ceased in 1985 and in the adjoining and contiguous Chobe National Park in neighbouring Botswana, where elephant culling has never been practised, the last aerial census in 2001 estimated a total of 140000 elephant in the two parks, at a density of 2.9 elephant for km², which greatly exceeds the widely accepted sustainability norm of 0.5 elephant for 1km² (Foggin 2003; WWF-SARPO Report 2003; D. Gibson 2004, personal communication). Thus, the timebomb of elephant overpopulation exists in these two lovely game parks at an alarming level and a desperate and urgent need exists to re-commence organized professional culling in all three parks, Hwange, Chobe and Kruger, to preserve habitat, reap the benefits of commercial utilization of the elephant products (meat, skin and ivory) and, perhaps most important, avert the awful and protracted suffering of the drought-driven mass die-off that must inevitably occur otherwise.

Should sound animal husbandry principles and political backbone ever prevail to enable a resumption of professional cropping of elephant in the parks most threatened by overpopulation pressures, a concerted international scientific effort must be mounted to make best possible use of the veritable wealth of unique biological specimens that would arise as spin-offs from the cull. All aspects of whole body physiology and pathology should be covered by appropriate recording and sampling of significant tissues and body fluids and these samples should be analysed by recognized medical, veterinary and biological specialists in each particular field. Concerning reproduction especially, the many unanswered questions highlighted in this review and elsewhere (see Hodges 1998) with regard to follicular and luteal development in both pregnant and non-pregnant adult females, should be pursued vigorously, by giving special attention to the number, sizes and tissue-progestagen contents of follicles and/or any CL found in the ovaries of still non-pregnant lactational anoestrous females with 1–4 year old calves at foot, and, of course, during any observed periods of oestrus and in the very earliest stages of pregnancy. Similarly, in seeking a possible source of gonadotrophic stimulus for both the maternal ovaries during the first half of gestation and the foetal gonads during the second half of gestation, placental tissue and maternal blood samples in very early gestation, together with foetal blood and foetal pituitary tissue samples during the second half of gestation, should be assayed for gonadotrophin, steroid and prolactin contents. And the endometrium of all adults, both pregnant and non-pregnant, should be examined grossly and histologically for signs of age-related degenerative, and/or pathological, changes which may yield clues to the incidence and significance of such changes observed ultrasonographically in captive female elephant (Montali et al. 1997; Hildebrandt et al. 2000a; Hermes et al. 2004). Finally, further attention should be paid to early post-partum uteri to attempt to unravel the cause and significance of the excessive amounts of intrauterine blood seen in such uteri to date (Allen et al. 2003).

Coincidentally with renewed culling and the associated collection and analyses of samples, it is clear that further veterinary and behavioural studies are required to ascertain the underlying causes of the very high levels of infertility currently recognized in captive Asian and African elephants (Sukumar 2003; Hermes et al. 2004). Very good between-zoo management collaborations, breeding policies and veterinary investigative techniques exist at the present time to undertake such studies, including the training of female elephants to accept serial blood sampling and transrectal ultrasonographic examinations of their reproductive tracts (Hildebrandt et al. 2000a), and the training of the adult males to accept vigorous transrectal massage of the accessory sex glands to induce ejaculation and thereby provide a representative sample of semen, both for analysis of sperm quantity and quality and for possible artificial insemination of oestrous females (Hildebrandt et al. 2000b). In the meantime, it seems reasonable to propose that more concerted management efforts to house, feed and generally manage captive elephants under conditions that more closely resemble the environmental, nutritional and hierarchical family traditions and behaviour patterns which pertain in the wild, and which have been catalogued and described so eloquently by Moss (1983, 1988) and Poole (1989) in their prolonged study of the Amboseli Park elephant population in southern Kenya would be most likely to bear the best fruit.

For example, the provision of very much larger enclosures containing appropriate browsing vegetation and walking trails to provide regular exercise, all in an attempt to limit weight gain, general body maturation and the onset of puberty in young female captives to more closely equate with those parameters in young wild equivalents (Laws 1969a; Stevenson 2005). And of equal or greater importance, to run young pre-pubertal female and male elephants together, as occurs in the wild, and to permit mating or insemination of the females as soon as possible after they pass through puberty and begin to display regular and hopefully ovulatory and fertile, oestrous cycles. Oestrus and associated ovulation may be presumed to be rare occurrences for wild females due to the constant succession of pregnancy followed by a 2–3 year post-partum lactation anoestrus followed again by renewed pregnancy as soon as mating recurs. Perhaps it is the repeated 14–16 week ovarian cycles experienced by captive adult females (Kapustin et al. 1996), which are probably ‘silent’ in terms of expressed oestrous behaviour due to the absence of an adult musth male, that are most directly related to the endometrial cystic degeneration and leiomyomata formation observed ultrasonographically in relatively young captive females by Hildebrandt and his colleagues (Montali et al. 1997; Hermes et al. 2004). Even if they must be forced to live in zoos in unnatural climates and eating very artificial diets, let elephants be pregnant elephants together from a young age and normal reproductive patterns and fertility will very likely be resumed and maintained.

The author expresses sincere gratitude to Dr Susanna Stout for her enthusiastic support in collecting and analysing placental samples from culled elephant in Kruger National Park, South Africa and to Professors John Skinner and Rudi van Aarde of The Mammal Research Institute, University of Pretoria for laboratory accommodation and logistic support. Also to Dr Ian Whyte, Mrs Colleen Wood and the members of the culling team in Kruger Park for much practical assistance and great courtesy and kindness. The Bernard Sunley and Sir Philip Oppenheimer charitable trusts kindly provided generous financial support.