Could maternal testosterone levels govern mammalian sex ratio deviations?

Abstract

Although maternal dominance and good condition are frequently associated with raised offspring sex ratios in mammals, the key factor may be female testosterone, which not only underpins the behavioural indicators but could also provide a pathway to a possible proximate mechanism for sex determination. By taking into account the fact that female testosterone levels rise in response to environmental stressors, it is possible to re-interpret the findings of atypical sex ratios in mammals in a way that reconciles seemingly conflicting results and reveals instead what could be a coherent, adaptive system of sex allocation in mammals.

Introduction

At present, there is little consensus on whether or not mammals have adaptive control of the sex of their offspring. This is the case even though in some species (Conover and Van Voorhees, 1990; Shine, 1999), including some with chromosomal sex determination (Lovern and Wade, 2003; Komdeur et al., 1997, Komdeur et al., 2002), there is clear evidence of facultative adjustment of offspring sex ratios (West et al., 2002). Furthermore, there ought to be similar control in mammals since theoretically such control should greatly benefit parental fitness (Krackow, 1995a; Hardy, 1997). Nevertheless, many theorists conclude there is evidence for nothing except a Mendelian effect on the sex ratio in mammals (Williams, 1979; Charnov, 1982; Reiss, 1987).

Although sex allocation theory is based primarily on Fisher's (1958) model of frequency-dependent sex selection, the Trivers and Willard (1973) hypothesis provided a starting point for the investigation and interpretation of data showing atypical mammalian sex ratios by suggesting that mammalian mothers in good condition would derive a fitness advantage if they conceived and raised male offspring.

Although nearly a decade went by before the publication of the first research reports, the Trivers and Willard hypothesis subsequently engendered a substantial body of primary research (Cameron, 2004). In spite of this, reviews of the evidence (Clutton-Brock and Iason, 1986; Hiraiwa-Hasegawa, 1993; Hardy, 1997; Brown, 2001; Bercovitch, 2002; Brown and Silk, 2002; Krackow, 2002; Cameron, 2004; Sheldon and West, 2004) have brought little closure on the central problem, leading some theorists to suggest that further progress may only be achieved when more is known of a proximate mechanism for sex determination (Williams, 1979; Charnov, 1982; Reiss, 1987; Krackow, 2002).

In an attempt to solve some of the problems, I have drawn on recent data suggesting such a mechanism. This data is based on the hypothesis that mammalian females have a role to play in pre-determining the sex of their offspring by means of fluctuations in maternal testosterone levels. That is to say, the female mammal may have a direct influence on sex allocation. In a series of theory-driven observations, we have shown that the level of testosterone in a mammalian female's follicular fluid, prior to conception, predicts, at a statistically significant level, the sex of the subsequent embryo (Grant and Irwin, 2005; Grant et al., submitted). This suggests that the ovum could emerge, each oestrus or menstrual cycle, already adapted to receive an X- or a Y-chromosome-bearing spermatozoon.

In this paper, I explain the origins of this research and explore the implications of the results. In the first part, I describe the pathway by which the empirical hypothesis was devised. This includes Section 1.1 the rationale behind focusing on a maternal influence, Section 1.2 some clues from disparate lines of research on hormones and sex ratios and Section 1.3 the search for a link between the behavioural indicators of atypical sex ratios and a possible proximate mechanism. In Section 1.4, I describe the effect of stress on testosterone levels in mammalian females, and the effect of stress on developing males (male vulnerability), both of which have important implications for the new hypothesis.

In the second part of the paper, I describe how these observations and findings might help solve some of the problems in sex allocation theory. First, in Section 2.1, I relate the hypothesis to theories of sex allocation and sex ratio data. Next, in Section 2.2, I outline seeming inconsistencies in sex ratio data. In Section 2.3, I suggest some new categories for the analysis of sex ratio data and the reasons for them. And in Section 2.4, I demonstrate how the use of these categories renders sex ratio data compatible with the predictions of Fisher's frequency-dependent sex selection, the Trivers and Willard hypothesis and local resource competition (LRC). Finally, in Section 2.5, I show how mammals could, after all, have some adaptive control of the sex ratio of their offspring by means of stress-induced variation in maternal testosterone levels—an increase in male losses (due to male vulnerability) being offset by a testosterone-induced increase in the number of males conceived during stressful times.

Among reproductive physiologists it is widely thought that the sex of the offspring in mammals is a matter of chance, depending on whether an X- or a Y-chromosome-bearing spermatozoon from the male arrives at the ovum first. However, in evolutionary biology, there is a large number of published studies reporting the occurrence of statistically significant atypical secondary sex ratios in a wide variety of mammals, and in almost all these studies, atypical offspring sex ratios have been associated with maternal characteristics.

Since these two models of sex determination do not provide overall consistency, efforts have been made to reconcile them. While accepting the chance model of sex allocation at conception, many evolutionary biologists justify their focus on maternal characteristics by arguing that differential maternal investment post-conception provides a sufficient explanation for atypical secondary sex ratios (Hrdy, 1999). Here, the focus is on the mother because atypical sex ratios arise from differential mortality by sex, both pre- and post-natally (Eberhard, 1996). Several researchers have described pre-natal effects on the sex ratio (Dittus, 1979; Gosling, 1986; Jones, 1988, Pratt and Lisk, 1989), (although since the same environmental effects pertained both before and after conception it is difficult to establish whether the atypical sex ratios had their origin at conception or during pregnancy, or both—see Fiala, 1980). Nevertheless, notwithstanding theoretical objections, (Myers, 1978) there do appear to be post-conceptual pathways (even in placental mammals, including humans, see Hassold et al., 1983) to adjustment of the sex ratio, in spite of the putative loss in reproductive potential. However, none of these necessarily excludes an additional method of adaptive control at, or before, conception. Furthermore, evolutionary biologists recognize that the mammalian female's contribution to nurturing her young is more costly than the male's, both pre- and post-natally, so perhaps it would make intuitive sense for her to have at least some input into sex allocation.

Whatever the reason, almost all studies of the secondary sex ratio in mammals, by evolutionary biologists, refer only to the characteristics and/or behaviour of the mother, and not the father. Although Trivers and Willard (1973) wrote of “parents” in the earlier part of their paper, they referred mainly to “maternal” characteristics, suggesting several different ways in which a mother might influence the sex ratio of her offspring. They acknowledged the role of X- and Y-chromosome-bearing spermatozoa, suggesting that the mother's control of the sex ratio must therefore involve differential mortality by sex at an early stage in development.

At the theoretical level, sex allocation theorist Reiss (1987) is one of the few to have seriously considered the possibility of pre-conceptual maternal control of the sex ratio. Although his main focus was an analysis of potential evolutionary conflict between parents and zygotes, he also explored a model of sex ratio adjustment in which the mother had the primary role. Suppose, he argued, that a mother was “capable of informing sperm or ova of the preferred sex of her offspring, which seems theoretically possible”. Reiss (1987) demonstrated mathematically that there could be no a priori reason to eliminate a role for the mammalian mother in the determination, or pre-determination, of the sex of her offspring; rather the contrary might apply, that there was every reason to explore the possibility of a maternal influence, especially if it could be activated earlier rather than later in the reproductive process. Hence, there appears to be no fundamental theoretical reason why the mammalian mother should not have a role in the predetermination of the sex of her offspring.

One of the earliest suggestions for solving the puzzle of atypical sex ratios and conflicting results was that “a hormonal mechanism, mediated by environmental factors” could provide “a plausible explanation of many trends” (Clutton-Brock and Iason, 1986). In the same year, James began building a strong case for a hormonal influence on the secondary sex ratio (James, 1986, James, 1987, James, 1989, James, 1990, James, 1992). Working from a comprehensive array of secondary sources, he postulated both paternal and maternal hormonal effects on the sex ratio. For example, on the basis of reviews of atypical offspring sex ratios among men with various illnesses, or exposed to adverse chemicals, he suggested that low paternal testosterone could result in female-biased offspring sex ratios. (For a recent summary see James, 2006.) On the basis of full reviews of all the relevant, published evidence, both animal and human, James concluded that “adaptive sex ratio variation seems to be controlled by the gonadal hormones of both parents: and (some) non-adaptive sex ratio variation by maternal adrenal hormones” (James, 2006).

However, there is almost no evidence of a paternal effect on the sex ratio in non-human mammals. An exception is data presented by Gomendio et al. (2006) If, as is being suggested, an ovum is produced already adapted to receive an X- or a Y-bearing spermatozoon, there would need to be some process of recognition. Thus it is likely that some sperm characteristics would be implicated in sex determination. Putative paternal effects on human secondary sex ratios could also be the result of second-order effects due to assortative mating (Grant and Metcalf, 2003).

Given the predominance (in reproductive physiology) of the chance model of sex determination, it is not surprising that evolutionary biologists looked for post-conception effects on the sex ratio. And since theoretically the longer it took for these to be implemented, the more wasteful, researchers sought mechanisms which might operate immediately post-conception. Reproductive physiologists had already documented a possible maternal effect of timing of insemination (Guerrero, 1974) with conceptions taking place earlier in the menstrual cycle being more likely to be male (Weinberg et al., 1995). Krackow (1995b) argued that just such a mechanism could be used by the mammalian female to adaptively alter the sex ratio by controlling the relative time of insemination within the oestrus cycle. Added to this was new information on differential developmental rates by sex of zygote, with males developing faster than females from insemination onwards (Mittwoch, 1993).

Krackow (1995b) argued that maternal hormone levels could influence both uterine responsiveness and developmental asynchrony, thus exerting an influence on differential survival of blastocysts. This reasoning was also consistent with sex-specific resorption and sex-specific embryonic mortality already observed in some mammals (Krackow, 1995b).

Thus it is likely there are a number of different, but inter-related physiological mechanisms post-conception that provide the means for a modification or refinement of conception sex ratios. Another example is the exploration of the physiological correlates of good condition and the suggestion that a maternal influence on sex determination might involve variations in the levels of maternal glucose in the uterine environment immediately post-conception (Larson et al., 2001; Cameron, 2004). This hypothesis was further supported by the fact that the addition of glucose to in vitro culture mediums often resulted in enhanced development of males, but not females (Rosenfeld and Roberts, 2004).

Thus, the developmental asynchrony hypothesis, now supported by a growing number of different potential mechanisms, describes the means by which a mother could influence the sex ratio of her offspring. However, although these mechanisms may play an important role in refining the sex ratio, they may not be the earliest means of establishing it.

Aside from the theoretical position that a standard method of sex ratio adjustment that occurs post-conception could be too wasteful, there have been a few hints of a maternal influence on the primary sex ratio occurring at, or prior to, conception in some species. For example, when Van Schaik et al. (1989) found that high ranking female macaques produced more sons than lower-ranking females, their analysis of inter-birth intervals made them suspect that “the deviation in birth sex ratios is already established at conception”. And Perret, 1986, Perret, 1990 reported a statistically significant difference in offspring sex ratio in mouse lemurs according to whether the mothers were housed in groups or in isolation prior to mating. Since the differences in offspring sex ratios must have been established prior to conception, Perret suggested that they occurred in relation to isolation-induced hormonal differences in the mothers.

While some researchers concentrated on measures of weight and food, others noticed that in many species, the animals in good condition tended to be the dominant animals, so they measured dominance as an indicator of good condition (Clutton-Brock et al., 1984). Although at first sight, it might appear that body condition would be the characteristic more likely to lead to a physiological measure with the potential to link to a proximate mechanism for sex determination, it now appears that dominance could provide the necessary link. The argument is that under normal or good environmental conditions, good body condition is a consequence of having high dominance (because it has led to priority access to resources) and that dominance in turn is subsumed by the hormone testosterone (see below), which has the potential to have a direct effect on the allocation of sex.

Dominance as a characteristic requires precise definition (Grant, 1998). Underlying the observed dominant behaviours it is likely that there is such a thing as inherent dominance (Bernstein, 1976; Dunbar, 1988) intrinsic to an individual and having a biological basis (Sapolsky and Ray, 1989; Beehner et al., 2005). This biologically based characteristic then predisposes the individual to behave in particular ways, giving rise to the expression of observable, quantifiable, dominant behaviours. Depending on the dominance of others in the group, these behaviours may be expressed in some interactions and not in others, leading to a dominance ranking for the individual, and the existence of a dominance hierarchy within the group (Simpson and Simpson, 1982). Dominance rankings are usually calculated by observing the number of times a particular animal displaces another in the competition for desired objects, most frequently food (e.g. Clutton-Brock et al., 1984). Thus, often, the very techniques used for measuring dominant behaviours showed that the dominant animals were physiologically advantaged (better nourished), and thus in better condition, as a result of their dominant behaviour.

The idea that intrinsic dominance could have a biological basis was not a new one. A number of studies (see below) had already demonstrated that normally distributed serum testosterone could provide the link to the range and intensity of individual differences in these particular behavioural characteristics. Although the dominance-testosterone relationship has been explored most fully in human males (Mazur and Booth, 1998) it has also been demonstrated in non-human mammals both male (Rose, 1974; Rose et al., 1975; Mazur, 1976; Sapolsky and Ray, 1989; Ruiz-de-la-toree and Manteca, 1999) and female (Bouissou, 1978; Bouissou and Vandeheede, 1996; Beehner et al., 2005). The higher the serum testosterone levels, the more likely the individual will exhibit dominance behaviours, and the more likely it will be higher in a dominance hierarchy.

Although mammalian females have only about one-tenth the amount of serum testosterone as males (Ehrenkranz et al., 1974; Purifoy and Koopmans, 1979; Haning et al., 1993; Christiansen, 1998; Grant and France, 2001) it appears to play a similar role in underpinning dominant behaviour (Grant, 2005). In addition, it provides a potential link between dominance-related atypical sex ratios and a proximate mechanism for sex predetermination.

For such a link to be theoretically plausible, there would need to be some means by which testosterone levels changed over time, in response to events in the environment. As it happens, there is just such a mechanism, one that for physiological reasons is much more finely attuned to environmental changes in females than it is in males. Under conditions of chronic stress, male testosterone levels decrease (Kreuz et al., 1972; Opstad, 1992; Christiansen, 1998) but female testosterone levels increase (Grey, 1992). This difference is explained, at least in part, by where and how testosterone is manufactured in the body. Whereas a male's testosterone comes from the testes, the production of a female's testosterone is closely associated with the adrenal cortex, which in turn is also responsible for the increase in cortisol in response to stress (Powell et al., 2002). As Mazur et al. (1997) described it, in females, but not in males, the adrenal cortex produces testosterone and cortisol simultaneously. Testosterone levels in females appear to rise and fall in synchrony with levels of stress-induced cortisol, because the adrenal cortex is the primary source of both hormones in females (Mazur et al., 1997).

Thus there are likely to be several factors influencing the level of a female's testosterone at any one time. One factor is of genetic origin, contributing to a female's position on a normally distributed baseline level of testosterone. Additional influences arise from age, place in the hierarchy, state of health, and time since joining the group (Wilson, 1975). Environmentally induced rises and falls in testosterone levels and dominance status appear to be part of the normal regime for both males and females. In addition to this intra-individual fluctuation, there are environmental influences on the group as a whole (e.g. lack of resources, depleted habitat). (This point is discussed in more detail below, see Section 2.4.) While retaining its normal distribution it is likely the whole range of testosterone values (and its associated dominance behaviours) rises when females are in a chronically stressful environment. Thus, in females, stressful conditions impose a unidirectional (upwards) effect on both dominance behaviours and testosterone levels (see Fig. 1).

Suppose the level of maternal testosterone were related to offspring sex allocation, so that mothers with higher (above the mean) testosterone levels were more likely to conceive male offspring, and mothers with below-the-mean testosterone levels were more likely to conceive female offspring. During stressful times, a small number of mammalian mothers, whose normal levels of testosterone were immediately below the level required for conceiving a male, would be tipped above it—the hypothesized critical cut-off point for conceiving a male. The number of male conceptions would thus rise, temporarily shifting the primary sex ratio.

Potentially, female testosterone could influence the actual process of sex determination in a variety of ways. Of these, one currently being explored (Grant and Irwin, 2005) is by means of inter- and intra-individual variations in the levels of testosterone in the female's follicular fluid (the fluid that surrounds the developing ovum as it matures during each oestrous or menstrual cycle). In at least some mammals, the levels of testosterone in the follicular fluid vary considerably (Henderson et al., 1982; Meinecke et al., 1987) and exceed those found in serum by 10,000–30,000 times (Greenspan and Baxter, 1994).

Using the cow as our animal model, we took ova from bovine follicles and fertilised them in vitro. At the same time, each matched sample of follicular fluid was assayed for testosterone. We found that the level of testosterone in the follicular fluid prior to conception predicted the subsequent sex of the offspring at a statistically significant level (Grant et al., submitted). If replicated, it would appear that higher concentrations of testosterone in the ovarian follicle could provide a means by which the ovum is modified or adapted, each oestrus or menstrual cycle, to receive an X- or a Y-chromosome-bearing spermatozoon. Such a mechanism, involving as it would, differential access for X-and Y-spermatozoa, (see, for example, Saling, 1991) could provide a potential pathway to sex predetermination under the control of mammalian mothers.