Sensitive Periods for Hormonal Programming of the Brain

Abstract

During sensitive periods, information from the external and internal environment that occurs during particular phases of development is relayed to the brain to program neural development. Hormones play a central role in this process. In this review, we first discuss sexual differentiation of the brain as an example of hormonal programming. Using sexual differentiation, we define sensitive periods, review cellular and molecular processes that can explain their restricted temporal window, and discuss challenges in determining the precise timing of the temporal window. We then briefly review programming effects of other hormonal systems and discuss how programming of these systems interact with sexual differentiation.

1. Introduction

Hugo De Vries coined the term “sensitive period” to describe permanent effects of the environment on the development of poppy flowers (De Vries 1904). He observed that poppies grown under harsher conditions were smaller, less strong, and carried flowers with fewer pistils, which made them less ornamental than poppies grown under richer conditions. If he transferred plants grown under harsher conditions to richer soil when they were 5–6 weeks of age, the plants became large and strong but did not develop multi-pistillate flowers. Vice versa, if he made conditions harsher for plants initially grown under rich conditions, plants became weak and slender but still developed the multi-pistillate flowers seen in plants grown under richer conditions. The initial environmental signal had permanently set the flowers on a more or a less ornamental course. As De Vries put it, “The sensitive period has terminated.”

The phenomenon that environmental factors influence development of organisms during specific times has since been observed in a wide variety of organisms for a wide variety of processes including the development of the brain and behavior. Researchers often use the term “critical period” instead of ‘sensitive period.” The first papers to use the term “critical period” in their titles concerned social behavior of puppies, (Scott and Marston 1950; Scott 1957, 1958), which could be influenced by training only during the first 2 years of life. After those 2 years, the window for change closes. Ideally, the term “critical period,” should be reserved for an absolute window in time beyond which a specific developmental process cannot be influenced anymore. “Sensitive period” could also be used for windows in development during which a specific process can most easily be influenced without necessarily excluding the possibility that the same factor can have a weaker influence at other times. In most cases, it is rather difficult to determine the window for a critical period precisely, or to exclude the possibility that the same factor can influence the same process at later points in life. Therefore, we will preferentially use the term “sensitive period” for the remainder of this review.

1.1. External and Internal Factors Driving Sensitive Periods

The botanist De Vries argued that there must be a moment in development when the number of pistils is decided. As he could observe the “terminal flower” in 7-week-old plants, he argued that the sensitive period for environmental factors to influence this decision must occur earlier. Sensitive periods in brain development appear to follow similar rules. The idea that developing brains need specific types of external sensory information at restricted periods during development to grow fully functional sensory systems and species-typical patterns of behavior is now well established for all senses. Prime examples are the necessity of visual, auditory, gustatory, and olfactory input to develop proper vision, hearing, and smell (Zou et al. 2004; Hensch 2005; Dominguez 2011; Froemke and Jones 2011). During development, sensory stimuli find a plastic substrate ready to be programmed. A variety of cellular and molecular processes subsequently establishes a relatively stable state guided by growth factors and signaling mechanisms that control the level of plasticity (Berardi et al. 2000; Knudsen 2004).

Programming of neural substrate also takes place by internal factors generated elsewhere in the body. As hormones can reach and influence tissues throughout the entire organism, and as their levels are influenced by internal as well as external factors, they can function as an effective interface between the internal and external milieu, directing the organism to adapt to a variable environment. For the brain, one of the most thoroughly studied examples of hormonal programming is the role of gonadal hormones in sexual differentiation of the brain. In this review, we first discuss historic and current studies on sexual differentiation of the brain, then briefly review programming effects by hormones other than sex steroids. We also discuss interactions between different endocrine systems and explain how sensitive periods for these endocrine systems help the organism to adapt to the environment as well as make the development of the brain and behavior vulnerable to endocrine disruptors and other deleterious environmental factors.

2. Gonadal Steroids

2.1. A Brief History of Sexual Differentiation Research

The realization that gonadal secretions direct the development of body and brain can be traced back to the classical experiments of Professor Berthold in Gottingen, who reported that transplanting testes into castrated roosters prevented their development into capons by stimulating the development of spores, wattles, crowing, and feisty behavior typically displayed by intact male roosters (Berthold 1849). Professor Jost’s work later demonstrated that testes are essential for the development of the male reproductive organs and that, in their absence, bodies develop in a female direction (Jost 1947). He showed that testosterone replacement stimulated the development of internal and external male reproductive organs, but also that a second testicular product, later identified as anti-Müllerian hormone, inhibited the development of female internal reproductive organs (Jost 1970).

Currently, we have a pretty good grasp of the molecular and hormonal processes that underlie the development of male and female characteristics, also called masculinization and feminization, in body and brain. Briefly, in mammals, inheriting a paternal Y or X chromosome decides which of the two pathways is followed. Although most cells having either an XX or an XY chromosomal complement may in principle influence all tissues directly at any time during the lifespan (Arnold 2009), by comparison the path of hormonal programming of male and female characteristics is remarkably narrow at first. In fact, it may be restricted to only a few tissues at an equally restricted period during our life. For example, in mice, expressing only one Y chromosomal gene, the Sry gene, for only half a day (embryonic day 11 {E11}–E11.5), in only one cell-type, the Sertoli cell, launches a series of events that eventually leads to the male phenotype (Burgoyne et al. 1988; LovellBadge and Hacker 1995). The Sertoli cells secrete the aforementioned anti-Müllerian hormone and additional factors that direct the development of the primordial gonad into a testis. Subsequent surges of testosterone during perinatal development and later, during puberty and beyond, lead to the male phenotype. In absence of the Sry gene, the primordial gonad becomes an ovary (Loffler and Koopman 2002). Although ovarian secretions early in development are not believed to play a major role in sexual differentiation, the cyclic release of ovarian steroids from puberty onwards leads to the full expression of the female phenotype (Wilhelm et al. 2007).

That this process of sexual differentiation includes the brain became clear when Harris and Jacobson demonstrated that sexual differentiation of the hypothalamus, rather than sexual differentiation of the pituitary or peripheral structures, was responsible for sexual differentiation of the control of gonadal secretions, which was cyclic in female rats and more constant in males (Harris and Jacobsohn 1952). Two years later, Barraclough and colleagues demonstrated that single injections of testosterone right after birth caused infertility in female rats (Barraclough and Leathem 1954; Barraclough 1961). As lesioning the anterior preoptic area of the hypothalamus abolished cyclic gonadotropic hormone release (Barraclough and Gorski 1961), they suggested, “the anterior preoptic area is undifferentiated at birth with regard to its subsequent control of gonadotropin secretion,” and that, in the absence of androgen, the anterior preoptic area differentiates to sustain cyclicity (as in females) whereas in the presence of androgen it “becomes refractory to both intrinsic and extrinsic activation, and the more tonic type of male gonadotropin secretion is observed.” Although Barraclough and colleagues injected animals only once during development, later research suggested that the same effect could be obtained at other time points as well, but only if injections were made postnatally, which demarcated the sensitive period for programming effects of exogenously administered steroids on hypothalamic regulation of gonadal secretions (Maclusky and Naftolin 1981).

A paper often quoted as launching the study of hormonal control of sexual differentiation of the brain is the 1959 landmark paper of Phoenix, Goy, Gerall, and Young, who showed that the female offspring of guinea pigs given testosterone during pregnancy readily displayed male copulatory behavior when treated with testosterone as adults. Offspring receiving adult, but not prenatal, hormonal treatment did not show masculine behavior. The sensitive period for this effect in guinea pigs is prenatal, as females given testosterone neonatally did not display masculine behavior as adults (Phoenix et al. 1959). These authors stressed a dichotomy in the effects of gonadal steroids: during early development gonadal hormones permanently changed the responsiveness to hormones later on in life, whereas later in life the same hormones affect behavior only transiently. They called the developmental effects “organizational,” and the transient effects “activational.” The stimulating effects of testosterone on masculine sex behavior are a good example of the latter. Although later studies have questioned the absolute nature of this dichotomy (Arnold and Breedlove 1985), the heuristic value of this dichotomy has proved immensely helpful in delineating the various ways in which gonadal steroids shape brain and behavior. The terms, “organizational” and “activational” effects closely parallel the terms, “programming” and “acute effects” used in other fields of biology (Fig. 1).

Fig. 1.

The organizational/activational hypothesis of sexual differentiation as it applies to rats. The presence or absence of perinatal and pubertal surges in testosterone directs the brain in a male or female direction. From puberty onwards, gonadal hormones act on the now organized, or programmed neural substrate to stimulate the expression of sexually dimorphic behaviors, such as male or female sex behavior. In other cases, organizational and activational effects of hormones may reduce differences between males and females (De Vries 2004). The early organizational effects are permanent, and can be considered programming effects; the later activational effects are acute effects that will disappear after withdrawal of the hormones. Two extreme options of sensitive periods are depicted. The yellow columns suggest sensitive periods that coincide with the presence of gonadal steroids, first perinatally, and later during the onset of puberty. The green-skewed bell curve indicates a period of waxing and waning sensitivity to programming effects of endogenously produced or exogenously administered steroids

Phoenix et al. suggested that changes in the brain may underlie the programming effects of steroids on steroid responsiveness, but it took many years, before these suggestions were proven right. In 1966, Pfaff provided the first evidence for this: as adults, neonatally castrated rats showed morphologically different nucleoli (Pfaff 1966), and four years later, McEwen and his colleagues demonstrated that neonatal hormone manipulations changed the uptake of testosterone and estradiol by adult brains, presumably by binding to steroid hormone receptors (McEwen and Pfaff 1970; McEwen et al. 1970). One year later, Raisman and Field reported the first sex difference found in neural connectivity: the parastrial nucleus of the preoptic area of male rats had more synapses from non-strial origin on dendritic shafts and fewer of such synapses on dendritic spines than did females (Raisman and Field 1971). The same researchers showed that early exposure to different levels of gonadal hormones can have sexually dimorphic effects on brain structure that last into adulthood (Raisman and Field 1973). No attempt has been made to narrow down the time in which hormones can establish this sex difference.

2.2. Sensitive Periods in Sexual Differentiation of the Brain

A striking example of a sex difference for which a sensitive period was determined was found in the size of the sexually dimorphic nucleus of the preoptic area (SDN-POA) in rats, which is about five times larger in males than in females (Gorski et al. 1978). Gonadectomy or hormonal manipulations in adulthood did not affect the size of this difference, but neonatal castration reduced the size of the SDN-POA in males to a level intermediate of that in males and females (Gorski et al. 1978), suggesting that testicular hormones around the time of birth masculinized the SDN-POA. By injecting testosterone on different days between late gestation and early postnatal life in female rats, Rhees et al. (1990a, b) established that only injections between E18 and postnatal day 5 (P5) increased SDN-POA volume. These studies did not determine whether testosterone influenced the number or size of SDN-POA neurons, or both. It would be inappropriate to speak about a critical period, as later research showed that SDN-POA volume remains somewhat plastic. Although the initial report of the SDN-POA did not find any effect of adult hormone manipulation on SDN-POA volume, more recently castration of adult males was shown to reduce cell but not total nuclear volume, whereas testosterone treatment in ovariectomized females increased cell volume and total volume of the SDN-POA (Dugger et al. 2008). It is not known, however, whether these effects are as endurable as those of neonatal manipulations. Although the SDN-POA has been implicated in partner preference, and the size of the SDN-POA correlates with partner preference in a number a number of species (size in males corresponding to stronger attraction to same-sex vs. opposite-sex conspecifics) (Baum 2006), the causality of this relationship has not been explored, and it is still unclear what difference the higher number of cells in males make in terms of function (De Vries and Södersten 2009).

Another nucleus for which a sensitive period has been explored quite well is the spinal nucleus of the bulbocavernosus (SNB). This nucleus contains motor neurons that in males innervate the bulbocavernosus and levator ani muscles at the base of the penis. These muscles are absent or vestigial in females, and, correspondingly, females have many fewer motor neurons in their SNB (Breedlove and Arnold 1980). The function of this sex difference is so clear that most attention has been focused on the cellular and molecular mechanisms that lead to this difference. In fact, the SNB may very well be the most intensively studied neural sex difference (Sengelaub and Forger 2008). The sensitive period for this nucleus also appears to be around birth in rats, but depends on which feature of the SNB is studied (Breedlove and Arnold 1983). Testosterone injections from E17 to 22 (‘late prenatal’), or on P1–5 (‘early postnatal’) masculinized number as well as size of SNB motor neurons, but injections of testosterone from P7 to 11 (‘late postnatal’) masculinized SNB soma size only, suggesting different targets for testosterone’s programming effects on SNB number versus size. This study could not time the onset of the sensitive period, as no pups made it to term in dams injected from E11 to 16 (‘early prenatal’); in addition, it is not clear how long the window for testosterone’s programming effects on soma size extends beyond the ‘late postnatal’ period. The authors, however, demonstrated that the window does in fact close, as testosterone treatment of adult animals during the month before sacrifice did not erase the differences in SNB cell number or size between perinatally androgenized females and controls (Breedlove and Arnold 1983).

Although programming effects of gonadal steroids on neural structures have been described for many sex differences, relatively few studies have tried to narrow down the sensitive periods for these effects. The ones described above form a sizable part of them (Cooke et al. 1998; Simerly 2002). But even in cases where sensitive periods were explored, often no clear answers were found. For most sex differences, no attempt was made to test whether they depend on programming or acute effects of gonadal steroids or on other factors, such as hormone-independent effects of sex chromosomal complement as described in (McCarthy et al. 2012).

2.3. Challenges in Determining Sensitive Periods

There are several reasons why it may be difficult to determine the sensitive period for programming effects of gonadal hormones using exogenously administered steroids. One is that, it is nearly impossible to generate steroid levels that mimic the developmental profile of endogenous levels precisely. For example, a one time injection of testosterone will likely target cells that are “ready” for testosterone’s programming action. Such cells are in the right developmental stage, express hormone receptors at the right level, etc. For other cells, however, it may be the wrong time during development or even the wrong time of day. Such a one time injection is, therefore, unlikely to produce complete masculinization of the system under study, and, if the actual window encompasses a time of waxing and waning sensitivity, may not readily produce effects at the beginning and end of the sensitive period robust enough to detect. This would lead to an underestimation of the window. This appears to be the case for the SDN, where only testosterone injections between E18 and P5 could increase SDN volume (Rhees et al. 1990a, b). Castration studies, however, suggest that the window for programming effects of endogenously produced gonadal steroids is much wider, as removing the testes even as late as P29 reduced the size of SDN-POA (Davis et al. 1995).

Removing the gonads at different times after birth can also cause underestimation of the sensitive window for sexual differentiation. For example, castration at a time when the testes do not secrete hormones in levels sufficient to permanently influence differentiation may lead to the wrong conclusion that testosterone can no longer influence the system. Recent studies suggest indeed that for some sex differences sensitive periods for sexual differentiation extend well into puberty, even though earlier studies had suggested the sensitive window was strictly perinatal. For example, male hamsters castrated just before puberty show less male sexual behavior in response to testosterone treatment in adulthood than do hamsters castrated in adulthood (Sisk and Zehr 2005). Although this might mean that the programming effects of perinatal surges of testosterone are followed by a second wave of programming effects of gonadal steroids secreted around puberty, it does not necessarily mean that the sensitive period opens around birth, closes, then opens again around puberty. More likely the entire period from before birth to around puberty forms one extended sensitive period during which exogenously administered testosterone can modify sexual behavior in adulthood, as injections of testosterone just prior to puberty are more effective in masculinizing behavior than injections during puberty (Schulz et al. 2009) (Fig. 1).

There are additional problems in removing the source of a hormone. The field of sexual differentiation often assumes tacitly that castration removes testosterone but not anything else that can explain the effects on brain and behavior. Recent data, however, suggest that another testicular product anti-Müllerian hormone, which as mentioned above inhibits the development of female reproductive organs in males, may play a role in sexual differentiation of the brain and behavior in mice as well as humans (Morgan et al. 2011; Wittmann and McLennan 2011, 2013a, b; Pankhurst and McLennan 2012). For example, the number of calbindin-immunoreactive (IR) cells in the preoptic area and bed nucleus of the stria terminalis (BNST) is higher in male than in female mice (Gilmore et al. 2012). In mutant mice that do not express anti-Müllerian hormone, this sex difference is markedly reduced, but not absent, suggesting that testosterone as well as anti-Müllerian hormone programs calbindin expression in the medial preoptic area (Wittmann and McLennan 2013a, b).

Even determining the developmental profile of hormone receptor expression is problematic if used to define sensitive periods. A number of excellent studies have documented the ontogeny of, for example, androgen and estrogen receptor expression in the BNST and MPOA (DonCarlos and Handa 1994; DonCarlos 1996; McAbee and DonCarlos 1999), areas, which as we indicated above show marked sex differences in neuroarchitecture and which also have been implicated in sexually dimorphic functions (Cooke et al. 1998; Simerly 2002). Such studies are typically better in suggesting the onset of hormone sensitivity for a certain area as, in many cases, hormone expression extends beyond the sensitive period. An exception is the rat cortex, which shows a transient expression of relatively high levels of estrogen receptor expression in the first 2 postnatal weeks. In this case, however, the precise role of these receptors is unknown and, therefore it is unclear what sensitive period this transient expression may define. An additional pitfall of relying on developmental expression profiles for hormone receptors is that hormones may indirectly influence the sexual differentiation of a specific system, which is a problem if the systems mediating differentiating effects are unknown.

2.4. Cellular Determinants of Sensitive Periods

Sex differences have been found for almost any cellular feature imaginable, macroscopically from the size of brain areas and brain tracts to microscopically, the size, number, morphology, and molecular make-up of neurons as well as glia. The latter category is the widest and encompasses sex differences in expression levels of neurotransmitters, neurotransmitter and hormone receptors, intracellular signaling molecules, all the way down to epigenetic marks in the chromatin (McEwen 2001; Simerly 2002; Cahill 2006; McCarthy 2008; McCarthy et al. 2009a, b). This suggests that hormones act on a wide variety of molecular and cellular processes to direct sexual differentiation of the brain. Some of these processes take place only during development and can therefore, better explain sensitive periods than processes that take place over the entire lifespan.

A clear example of the former is developmentally programmed cell death, the bulk of which in rodents takes place around birth (Forger 2006, 2009; Ahern et al. 2013). Differential effects on this process will typically be permanent, as dead neurons are unlikely to be replaced by new ones, in almost all brain regions. The programming effects of testosterone on cell number in the SDN and SNB fit in this category. Although testosterone affects the number of dying cells in these regions during development (Nordeen et al. 1985; Davis et al. 1996; Chung et al. 2000), the hypothesis that differential cell death leads to sex differences in cell number in these areas has only been tested directly for the SNB, using mutant mice in which developmental cell death was markedly reduced or absent. Whereas wild-type mice showed a sex difference in cell number in the SNB, this difference was absent in mice over-expressing cell death-reducing factor, Bcl-2, and in mice with a null mutation in the gene encoding the cell death factor, Bax (an obligate factor for neuronal developmental cell death) (Zup et al. 2003; Forger et al. 2004). The same mice could not be used to test whether differential cell death underlies sexual differentiation of the SDN, as mice do not show a similar group of cells as the SDN in conventionally stained microscopic sections.

Other examples of processes that take place during restricted periods in development are the massive waves of neurogenesis (Altman and Bayer 1995), migration of cells from germinal zones to their final destination (Hatten 1999; Ayala et al. 2007), and the colonization of the brain by monocytes and their subsequent transformation into microglia (Ginhoux et al. 2010; Harry and Kraft 2012), all of which, in rodents, take place mainly prenatally. Neurogenesis has often been dismissed as an important factor in sexual differentiation as for most sexually dimorphic structures neurogenesis is over by the time that testosterone exerts its differentiating effects (Cooke et al. 1998; McCarthy et al. 2009b). However, gonadal steroids continue to influence neurogenesis in hypothalamic, amygdalar, and hippocampal areas of adult animals (see, e.g., Tanapat et al. 1999; Fowler et al. 2008; Brock et al. 2010). It is not difficult to imagine that long-term exposure to gonadal hormones may contribute to sex differences observed in these areas. In these cases, the sensitive period would remain open. As far as we know, this has never been tested systematically.

Apart from a handful of studies (for review, see Tobet et al. 2009), there is not much evidence that differential migration plays a major role in sexual differentiation. This lack of evidence does not mean that neuronal migration is not an important part of sexual differentiation. Research on neuronal migration is technically challenging, which may be the reason that only very few groups have ventured to study the role of this process in sexual differentiation.

The notion that glial cells modulate gonadal steroid action on the brain is well established (Jordan 1999). This includes glial activity during development, as hormonally induced sex differences in, for example, astroglia can be found already at P1 in rats (Mong et al. 1996). Recently, it has become clear that microglia are also dimorphic and play a role in steroid-driven sexual differentiation. The numbers of microglial cells shows sex differences in many areas that vary depending on the age (Schwarz et al. 2012; Lenz et al. 2013). These microglia appear to mediate some of the early effects of steroids on sexual differentiation of the brain, as microglial inhibition prevented estradiol-driven masculinization of spine density and sexual behavior (Lenz et al. 2013). It is not clear whether factors intrinsic to microglia contribute to the length of sensitive periods, e.g., for masculinization of sexual behavior, or whether this depends on features of the cells with which they interact.

2.5. Sexual Differentiation in Neuronal Phenotype: Lessons Learned from Vasopressin

It is more difficult to explain sensitive periods for steroid effects on processes that can be expressed throughout life, such as dendritic arborization, synaptic spine formation, and expression of specific genes. Presumably some processes, such as placement of enduring epigenetic marks, stabilizes the state induced by exposure to gonadal steroids (McCarthy et al. 2009a, b). An example of this may be the vasopressin innervation of the brain.

Vasopressin is produced in a number of distinct areas in the brain, most prominently, the paraventricular, supraoptic, and suprachiasmatic nucleus, as well as the BNST and medial amygdala (MA) (De Vries and Miller 1998). Projections from the BNST and MA latter are highly dimorphic, with males showing more vasopressin-expressing cells and denser projections from these cells to fore-, mid-, and hindbrain areas (De Vries and Miller 1998). This difference may very well be the most consistently found sex difference, as it has been reported in many species within all classes of vertebrates with the exception of fish (De Vries and Panzica 2006).

Determining a sensitive period for hormone action on this system turned out to be tricky. When we first tested whether gonadal hormones direct sexual differentiation of these projections during a restricted period in development, we castrated males and administered testosterone to females at different times after birth. We then assessed vasopressin innervation on P28, when the sex difference in the lateral septum, one of the main targets of these projections, is most extreme (De Vries et al. 1981). We found that vasopressin innervation of neonatally castrated males resembled that of control females, whereas males castrated at 2 weeks of age had a fiber innervation similar to control males. Males castrated at P7 had a fiber density intermediate to that of control males and females, suggesting that under physiological conditions, gonadal hormones direct the differentiation of this system around P7, which roughly matched the timing of sexual differentiation of the model systems mentioned above. Testosterone injections into females or neonatally castrated males, however, fully masculinized vasopressin innervation whether they were given in the first, second, or even third week of life (De Vries et al. 1983). This did not fit with emerging ideas on the timing of the sensitive periods reported for other sex differences. What we did not know at the time is that this system remains exquisitely sensitive to acute effects of hormones throughout life, and that the effects of testosterone on this system persist for some time after its removal: the BNST and MA cease producing vasopressin mRNA within days after gonadectomy, but it takes weeks to months before vasopressin-immunoreactivity disappears (De Vries et al. 1984; Miller et al. 1992). If one would not take these dynamics into account, one would erroneously conclude that the sensitive period for programming effects of testosterone extends into the third week.

To deal with this issue, we studied the effects of neonatal manipulations on vasopressin expression in 3-month-old rats in which testosterone levels had been equalized for all groups for 4 weeks before sacrifice. (A similar design is commonly used to test the perinatal effects of gonadal steroids on masculine sexual behavior, which also requires acute stimulation by gonadal hormones to be seen.) We found that males castrated as adults had more vasopressin-IR cells in the BNST and a higher density of vasopressin-IR fibers in the lateral septum than did neonatally castrated male rats even though both groups were treated with the same dose of testosterone for 4 weeks before sacrifice. Neonatally castrated rats did not differ from female rats, whether these had been ovariectomized at birth or as adults. This suggested that there was a sensitive period during which postnatal testicular secretions masculinize the vasopressin innervation. To narrow down the window for this sensitive period, we castrated males at different times after birth. Males castrated at the day of birth or at P7 had less pronounced vasopressin innervation than rats castrated at 3 weeks of age or as adults, suggesting that testicular secretions masculinize vasopressin-IR projections until around P7. We confirmed this by treating neonatally gonadectomized male and female rats with testosterone propionate at P7, which masculinized the vasopressin innervation (Wang et al. 1993).

Comparison of the sex difference in vasopressin expression between species suggests that there can be much variation in how steroids generate sex differences, even if the nature of the difference (male[female) does not differ that much between species. The mechanisms in which these sex differences are generated can be diametrically opposed. In rats, for example, neonatal treatment with estradiol [which in this species is the testosterone metabolite responsible for masculinization of many aspects of behavior (McCarthy 2008)] masculinizes vasopressin expression; in quail, however, increasing estrogen levels during development results in the female phenotype, and lowering them produces a male phenotype with regard to the homologous vasotocin innervation (Panzica et al. 1998). This inter-species variation may extend to sensitive periods as well. For example, in voles, testes play an essential role in masculinization of vasopressin innervation. However, administering testosterone perinatally to females or to neonatally castrated males does not masculinize vasopressin expression (Lonstein et al. 2005). This means that either the testis does not use testosterone to masculinize vasopressin expression, or that the sensitive period occurs at a different time than in rats, most likely later, as the prenatal treatment of females was initiated before the testis was likely to secrete testosterone.

Although the specific target for programming effects of gonadal steroids on vasopressin cells has not been identified, several possibilities can be excluded. As the number of vasopressin cells differ in the BNST and MA, steroids could have affected cell birth, migration, cell death, or cellular phenotype. The first two possibilities are not likely candidates, as vasopressin cells are born on E12 and 13 (Al-Shamma and De Vries 1996; Han and De Vries 1999), which is prior to the differentiation of the gonads and well before the sensitive period for programming effects of gonadal steroids for this system.

Differential cell death can be eliminated as a factor as well. To do so, we determined vasopressin cell number in mice over-expressing the pro-survival protein, Bcl-2 or in mice that lacked the pro-death protein, Bax. Both mice strains have greatly reduced neuronal cell death, so if differential cell death underlies the sex difference in vasopressin cell number, the difference should be decreased or absent in these mice. We did find an increase the total number of cells that produce vasopressin in both sexes, but the sex difference in cell number remained intact (De Vries et al. 2008). Therefore, cell death determines the number of cells that have the potential of expressing vasopressin, but it is not responsible for the sex difference. This leaves sexual differentiation of cellular phenotype as the only remaining plausible mechanism for sexual differentiation of vasopressin expression.

The most likely mechanism underlying sexual differentiation appears to be epigenetic modification of gene expression. Recently, Forbes-Lorman et al. (2012) showed that reducing expression of the DNA methyl binding protein, MeCP2, in the MA neonatally eliminated the sex difference in vasopressin expression in the rat MA and BNST, which points at a critical role for DNA methylation in sexual differentiation of vasopressin expression. Whether DNA methylation is altered on the vasopressin gene itself or on some other gene that then influences vasopressin expression remains an open question. As the majority of neural sex differences found concern cellular morphology or neurochemistry rather than cell number, epigenetic mechanisms may be responsible for the lion’s share of sex differences in the brain (De Vries and Simerly 2002; McCarthy et al. 2009a, b).

3. Non-gonadal Hormones

3.1. Hypothalamic–Pituitary–Adrenal Axis

Similar to gonadal steroids, adrenal steroids influence neural development, and thereby adult behavior. In rodents, a clear sensitive period is the latter stage of pregnancy, when elevations in glucocorticoid levels as a consequence of exogenous administration or maternal stress program the hypothalamic–pituitary–adrenal (HPA) axis [reviewed in Harris and Seckl (2011), Xiong and Zhang (2013)]. Such elevated levels lead to a hyperactive HPA axis in adulthood characterized by increased circulating ACTH and corticosterone levels as well as elevated hypothalamic CRH mRNA, under nonstressed as well as stressed conditions (Henry et al. 1994; Barbazanges et al. 1996; Welberg et al. 2001; Shoener et al. 2006). Hippocampal glucocorticoid and mineralocorticoid receptors are decreased in the adult offspring indicating that this hyperactivity is due, in part, to reduced negative feedback of corticosteroids on the HPA axis (Henry et al. 1994; Welberg et al. 2001; Shoener et al. 2006). A hyperactive HPA axis likely contributes to altered behavioral responses to stress in rats and a higher prevalence of mental health disorders and behavioral problems in humans exposed to elevated glucocorticoids prenatally (Welberg et al. 2001; Van den Bergh and Marcoen 2004; Gutteling et al. 2005a, b; Salomon et al. 2011). In addition, excess prenatal glucocorticoids is associated with decreased birth weight and increased risk of cardio-metabolic disease later in life (Reviewed in Cottrell and Seckl 2009).

Under normal conditions, the fetus is protected from maternal glucocorticoids by placental 11β-hydroxysteroid dehydrogenase type 2 (11β-HSD2), which catalyzes ~90 % of maternal cortisol and corticosterone to their inactive 11-keto metabolites (Benediktsson et al. 1997). Inhibition of placental 11β-HSD2 by itself is sufficient to permanently alter HPA function and anxiety-like behavior (Welberg et al. 2000). Stress-induced increases of maternal glucocorticoids may overwhelm the protective capacity of 11β-HSD2 and in addition reduce placental 11β-HSD2 activity (Mairesse et al. 2007), thereby further increasing corticosteroid exposure of the fetus. Notably, maternal malnutrition also down-regulates 11β-HSD2, suggesting that excess glucocorticoid exposure may contribute to fetal programming from other factors as well (Bertram et al. 2001; Lesage et al. 2001). Hence, elevated glucocorticoids may be a general signal from the mother to the fetus of adverse environmental conditions.

Studies that tried to delineate the sensitive period prenatally found that, in rats, adrenal steroids program the HPA axis in the third week of pregnancy, but not earlier. For example, Nyirenda et al. (1998) found that daily dexamethasone injections in the third, but not the first or second, week of pregnancy reduced birth weight and increased hyperglycemia and hyperinsulinemia in adulthood. This suggests that as with gonadal steroids and the gonads, the sensitive period for adrenals does not start before the fetal adrenal starts secreting corticosteroids, which in rats happens in the last week of pregnancy (Wotus et al. 1998). The sensitive period may extend into the postnatal period, as brief handling of pups, separation of mother from pups, and maternal licking and grooming of pups during this period can also program the adult HPA axis (Levine 2005; Zhang et al. 2013). However, whether this is the result of altered glucocorticoid exposure has not been determined. While glucocorticoid programming of the HPA axis is not believed to occur in adult animals, we do not know of studies that tested the effects of long-term perturbations of adrenal steroids.

3.2. Metabolic Signals

David Barker was one of the first to suggest that fetal weight and metabolic syndrome later in life are related (Hales and Barker 2001). Low birth weight due to reduced maternal nutrition, high birth weight due to maternal obesity, and poor postnatal nutrition strongly predict obesity, and metabolic disease later in life (Hales and Barker 2001). Barker’s “thrifty phenotype hypothesis,” which later became the “Barker hypothesis,” posits that adverse perinatal environments program the development of multiple tissues, including the brain, thereby potentially setting up the organism for metabolic dysfunction later in life. Although the mechanisms underlying programming of feeding and energy homeostasis have yet to be fully understood, but recently important inroads have been made in understanding how hypothalamic circuits may be programmed by perinatal nutritional conditions (Simerly 2008; Breton 2013).

Leptin appears to play a central role in these programming events. Leptin is a hormone that is secreted by adipose tissue and controls food intake by acting on the arcuate nucleus of the hypothalamus (ARH) (Zhang et al. 1994). Adult homozygous leptin-deficient mice (ob/ob) are obese, however, they do not differ in weight from wild-type neonatally and only begin to weigh more than wild-type mice during the second week (Mistry et al. 1999). During development, leptin levels surge in the first 2 postnatal weeks, peaking around P10 (Delahaye et al. 2008). The size of the surge and the timing of the peak depend on nutritional status. The surge is much smaller and the peak occurs several days earlier in pups with food-restricted mothers (Delahaye et al. 2008). In contrast, the surge is much higher and prolonged in the offspring of obese mothers (Kirk et al. 2009). Precisely, how maternal nutritional information is transferred is unclear, as leptin surges independent of changes in fat mass (Ahima et al. 1998).

Leptin released during the developmental surge may play a programming role. Artificial modulation of the size and timing of a neonatal surge of leptin between P4 and 16 suggest that variability in the surge affects adult metabolic status; for example, blocking lepin activity at P9 led to a relative increase in body weight at 3 months of age (Granado et al. 2011). In addition, disrupting neonatal endogenous leptin induces, whereas neonatal leptin treatment to pups born to undernourished dams prevents, developmental programming of metabolic dysfunction in response to a high-fat diet (Vickers et al. 2005; Attig et al. 2008).

Leptin released during the surge may play a neurotrophic role. In a very elegant series of experiments, in many ways analogous to the neonatal castration experiments and testosterone replacement described above, Richard Simerly and colleagues tested the effects of leptin on the development of hypothalamic circuits that have been implicated in feeding behavior and energy homeostasis. They noted that leptin-deficient ob/ob mice have in comparison to wild-type mice a lower density of output projections from the ARH to all of its target sites (including the paraventricular nucleus and the dorsomedial nucleus of the hypothalamus, which function as satiety centers, and the lateral hypothalamic area, which is a hunger-promoting area) (Bouret et al. 2004). If they injected ob/ob mice daily with leptin from P4 to 16, they noted an increase in density of ARH projections; the sensitive period for this leptin effect appears limited to development as treating adult ob/ob mice with leptin did not normalize these projections (Bouret et al. 2004). In addition to altered projections, adult ob/ob mice display more inhibitory synapses onto orexigenic alpha-MSH/POMC neurons and more excitatory synapses on anorexigenic AgRP/NPY neurons than WT mice. When 8-week mice were intraperitoneally injected with leptin, this excitatory/inhibitory balance normalized within 6 h, even before leptin’s effects on feeding behavior were noticeable, suggesting that the window within which leptin may program food intake and energy homeostasis may extend well into the second month of life (Pinto et al. 2004).

Similar to gonadal hormones, the sensitive period for metabolic programming occurs during the development of hypothalamic centers that regulate food intake. Both AgRP/NPY1 and alpha-MSH/POMC neurons develop around E17 and 18 in rats, however, projections from the ARH to all of its target projections do not fully mature until P18 (Bouret and Simerly 2007). During this time, perinatal nutrition can influence the differentiation and proliferation of these neurons and their projections or influence the levels of gene expression of their principle neuropeptides. Indeed, offspring of food-restricted mothers exhibit an elevated NPY mRNA expression as weanlings (Cripps et al. 2009). When placed on a catch-up high-fat diet, these offspring continue to exhibit elevated NPY levels as adults compared to offspring of ad libitum fed rats (Ikenasio-Thorpe et al. 2007). Conversely, postnatal overnutrition induced by small litter size reduces adult NPY levels (Ferretti et al. 2011). Interestingly, a 50-day high-fat diet during adulthood also reduces NPY levels to the same degree as observed after high caloric intake; however, whether the effects persist have not been tested (Ferretti et al. 2011). Offspring of protein-restricted (Coupe et al. 2010) and calorie-restricted mothers (Delahaye et al. 2008) have reduced alpha-MSH IR fibers innervating the PVN, but when they are fasted as adults, they do not demonstrate reductions in alpha-MSH projections to the PVN (Breton et al. 2009).

Both neurogenesis and epigenetic modifications have been implicated as possible cellular and molecular mechanisms that determine the sensitive period for developmental metabolic programming. Proliferation of NPY and alpha-MSH neurons in the ARH is decreased in offspring of dams that were food-restricted during the first two weeks of gestation (Garcia et al. 2011). Conversely, rats born to mothers kept on a high-fat diet during the last two weeks of pregnancy have a greater number of NPY1-expressing neurons (Chang et al. 2008). Perinatal nutritional manipulations also resulted in epigenetic modifications of the POMC: it is hypermethylated in “overfed” weanling rats reared in small litters (Plagemann et al. 2009) and hypomethylated in the offspring of dams fed a low-protein diet (Coupe et al. 2010). These epigenetic modifications may represent resilient programming events to the proteome of ARH neurons, which remains sensitive to significant nutritional manipulations in adulthood.

3.3. Inflammatory Signals

Perinatal infections program the brain, behavior, and immune system. Prenatal infection has long been known to be associated with increased risk of Schizophrenia. Rodents exposed prenatally to the influenza virus exhibit several neuropathological signs in their brains and adult behavior (e.g., decreased sensorimotor gating and exploratory behavior), which are thought to be relevant to schizophrenia and other psychotic illnesses (reviewed in (Meyer 2013)). Many of these behaviors are similarly impacted by exposure to proxies of viral and bacterial infections (e.g., poly(I:C) and lipopolysaccharide (LPS), respectively), indicating they are the result of the mother’s inflammatory response rather than the infectious agent itself (for reviews see Bilbo and Schwarz 2012; Meyer 2013).

For programming of the immune system, both prenatal and early postnatal immune challenges lead to decreased responses to immune challenges in adulthood (Ellis et al. 2005, 2006; Williams et al. 2011). Here too, challenges by both bacterial (e.g., LPS) and viral (e.g., Poly(I:C)) mimetics are effective, but the programming appears to be challenge-specific: perinatal LPS injections alter later responses to LPS, but not Poly(I:C) and vice versa (Ellis et al. 2005, 2006). Reactivity of the HPA axis is also impacted by early life immune challenges and may play a role in decreased immune reactivity in adulthood (Ellis et al. 2005). However, the direction of the programming effect on the HPA response is not always consistent across treatments and may depend upon the type of immune challenge (poly(I:C) vs. LPS), timing of the initial immune challenge (prenatal vs. postnatal), and/or age that the HPA axis is assessed (neonatal or adult). For example, prenatal LPS exposure decreases, whereas postnatal LPS exposure increases the HPA response to a stressor (Ellis et al. 2005; Hodyl et al. 2008).

The sensitive period for early life infection has not been as systematically studied as sensitive periods in sexual differentiation. Reasons why this is more challenging than determining sensitive periods in sexual differentiation are the large variety of infectious agents as well as the multitude of signaling molecules involved. Furthermore, the bioactivity of the two most commonly used immune challenges, Poly(I:C) and LPS, can vary across orders from the same supplier, making it even more difficult to compare doses across studies unless a biomarker is reported (reviewed in Harvey and Boksa (2012)). The multitude of protocols that differ in timing and dose of injections as well as in endpoints measured makes it difficult to compare across studies. Therefore, while it is clear that early immune activation triggers a number of changes in the immune system as well as in the brain, the mechanisms and timeline of these actions are not well defined.

3.4. Thyroid Hormones

In addition to their role in the regulation of metabolism, the thyroid hormones, triiodothyronine (T₃), and thyroxine (T₄), are essential for normal brain development. Thyroid hormones influence myelination and neuronal and glial differentiation, processes that are critical in brain organization (Bernal 2005). T₄ is the primary form of thyroid hormone found in blood circulation and is converted into T₃ by type II 5^′ iodothyronine deiodinase (D2) once it reaches the desired cells (de Escobar et al. 2004). T₃ is the form of thyroid hormone that acts on the fetal brain during development, as evidenced by the presence of T₃ in the cortex at gestation week 12. Before 16–18 weeks of gestation, the fetus is incapable of producing its own thyroid hormone (Kester et al. 2004; Obregon et al. 2007) and relies on the maternal supply of T₄ for development. This can lead to various types of maternal-based hypothyroidism, including maternal hypothyroxinemia, the timing of which can predict the type of neurological deficit seen in the offspring. This dependence on maternal T₄ is true in rodent models as well, but because the rodent brain at birth is analogous to a 6-month-old fetus brain, maternal hypothyroidism may not produce as severe of an effect in the rat (Horn and Heuer 2010).

Hypothyroidism prenatally affects balance, attention, spatial memory, and motor activity in adulthood, but precise timing of thyroid hormone effects on these functions are still unclear. The few studies on developmental timing of hypothyroidism on behavioral output were extensively reviewed by Zoeller and Rovet (2004), who suggest that prenatal thyroid hormone insufficiency results in increased motor activity and attention deficit, whereas insufficiency between birth and weaning reduces motor activity in adult rats. However, a more recent study by Negishi et al. (2005) suggests that dams treated with propylthiouracil (PTU), a thyroid hormone blocker, from gestation day 3 to postnatal day 20 produce offspring that show shortened attention spans and hyperactivity in adulthood. Notably, perinatal hypothyroidism differentially impacts the cognitive development of male and female rats as females, but not males, exhibit impaired spatial memory (van Wijk et al. 2008).

The behavioral consequences seen in hypothyroidism may occur through deficits in neuroanatomical development. Thyroid hormones are necessary for the proper development of spine density of pyramidal neurons, cortical layer organization, proliferation of cerebellar granule cells, and other neuroanatomical functions (Ahmed et al. 2008). In the rat, the cerebellum develops within the first three postnatal weeks, which allows for a great experimental model to test the effects of thyroid hormones on cerebellar development (Xiao and Nikodem 1998). The cerebellum develops by proliferation of cells in the external granule layer (EGL); those cells migrate to the internal granule layer (IGL), making the EGL ultimately disappear (Zoeller and Rovet 2004). Hypothyroidism leads to a higher peak and extended period of apoptosis in the IGL (Xiao and Nikodem). This results in a smaller IGL in adulthood, and also affects the reduction in the EGL. In addition, hypothyroidism results in morphological changes of almost every cell-type in the cerebellum, which can affect the connectivity and function of the neurons (Horn and Heuer 2010). These effects may be caused by first- or second-order growth factor responses to thyroid hormone or through mediating transcription factors responsible for controlling neuronal development gene expression (reviewed in Horn and Heuer 2010). Thus, thyroid hormone acts in a time-dependent manner, but there is not one sole sensitive period for its action on the developing brain.

More evidence for the timing of thyroid hormone-sensitive periods comes from clinical studies, reviewed extensively in Zoeller and Rovet (2004) and Ahmed et al. (2008). Maternal hypothyroxinemia provides a natural way of looking at thyroid hormone insufficiency during the first half of pregnancy. Numerous clinical studies, reviewed by Henrichs et al. (2013), demonstrate that maternal hypothyroxinemia, if left untreated, can impair intellectual abilities and psychomotor skills of the offspring. However, thyroid hormone insufficiency in the second half of gestation provides mixed results, with some studies demonstrating cognitive defects and others showing no differences from controls. Interestingly, the only randomized clinical trial study on the effectiveness of prenatal treatment with levothyroxine in thyroid hormone-insufficient women at gestation week 13 demonstrated that thyroid hormone treatment did not improve cognitive function in children at three years of age (Lazarus et al. 2012). This study shows that despite the mothers having low T₄ levels in pregnancy, supplementing their T₄ levels during the sensitive period does not provide a significant change in cognitive deficits in the offspring.

4. Interactions Between Sensitive Periods

As the timing and targets of sensitive periods of different hormonal systems often overlap, interactions between programming effects of hormones are likely. Interactions between stress during development and sexual differentiation have been especially well studied. For example, pre- but not postnatal stress reduced the propensity of intact males to display male sexual behavior while increasing their propensity to display female sexual behaviors (Ward 1972). Effects of stress on sexual differentiation of behavior have been confirmed in many other studies (Charil et al. 2010; Bale 2011) including those investigating non-reproductive behaviors. For example, prenatal stress reduces play behavior in male but not female rats (Ward and Stehm 1991). The HPA axis is an obvious candidate for mediating the effects of stress on sexual differentiation. A higher activity of the HPA axis has been linked to lower plasma levels of testosterone, modified testosterone metabolism, and altered expression of steroid hormone receptors in the developing brain (Weisz et al. 1982; Ward and Weisz 1984; Weinstock 2007; Llorente et al. 2011). All of these changes are likely to change the programming effects of testosterone and thereby sexual differentiation of neural circuits. For example, stress in the last week of pregnancy reduces the size of the sex difference in SDN-POA volume (Anderson et al. 1985).

Stress effects on the brain, however, are likely to be multifactorial and mediated by more systems than just the HPA axis. Stress, for example, increases the secretion of proinflammatory cytokines and lymphokines (Black 2002), which, as we discussed above, program the brain as well. Cytokines and lymphokines have become an area of focus to understand the etiology of disorders such as autism, schizophrenia, and depression. Activation of the immune system during pregnancy increases the frequency of diagnoses for all these disorders (Brown and Derkits 2010; Patterson 2011). Interestingly, all these disorders show marked sex differences in onset, course, and incidence. For example, schizophrenia is more common in men (Abel et al. 2010), and autism more in boys (Fombonne 2009).

Interactions between programming effects of inflammatory signals and gonadal hormones may contribute to such sex differences. We recently studied interactions between inflammation and the development of social play behavior and vasopressin expression. Play behavior is interesting as it is one of the earliest social behaviors to emerge before puberty (Pellis and Pellis 1998). Vasopressin is interesting as it has been implicated in social disorders such as autism (Ebstein et al. 2010; Green and Hollander 2010) as well as in normal aspects of human social behavior (Walum et al. 2008; Meyer-Lindenberg et al. 2009). Evidence for involvement of vasopressin in social behavior is even stronger for experimental animals (Caldwell et al. 2008; Donaldson and Young 2008; Goodson 2008).

We found that treating rats with LPS on day 15 of pregnancy reduced social play of male, but not of female, offspring, thereby erasing a sex difference in play displayed by control rats (Fig. 2). The same treatment also reduced vasopressin expression in the BNST and MA, again in males but not in females, thereby reducing the sex difference in vasopressin mRNA expression seen in control rats (Fig. 2). As sex differences in play behavior and vasopressin expression depend on programming effects of testosterone (Meaney et al. 1983; Meaney and McEwen 1986; Wang et al. 1993; Han and De Vries 2003), LPS may have interfered with these effects. Our data, however, do not support a general effect of LPS on sexual differentiation, because the sex difference in SDN-POA volume was unaffected by LPS treatment.

Fig. 2.

Effects of a prenatal immune challenge on the sexual differentiation of juvenile play behavior and vasopressin mRNA expression of the brain. Pregnant rats were treated with the bacterial endotoxin, LPS, on gestational day 15, and offspring were tested for juvenile play behavior between postnatal days 26–40. a The mean ± SEM number of total play events displayed in a 10-min testing period differed between male and female control juveniles, with males playing more than females (white bars). Treatment with LPS reduced play only in males, thereby eliminating the sex difference in those animals exposed to a prenatal immune challenge (black bars). b Following play testing, brains were collected and the number of cells expressing vasopressin mRNA in the BNST was determined by in situ hybridization. There was a sex difference in vasopressin cell number in control animals (white bars). Prenatal LPS challenge decreased vasopressin cell number only in males, thereby reducing the sex difference (black bars). Reprinted with permission of BioMed Central from (Taylor et al. 2012)

LPS treatment may have affected sexual differentiation indirectly. For example, humoral factors generated as a result of LPS treatment may have had differential access to male and female fetuses. Mueller and Bale (2008) showed that stress early in pregnancy significantly changes the expression of genes implicated in the hypoxic response, cell differentiation, and metabolism in male but not in female placentas. Like stress, immune challenges may have similar dimorphic effects on the placenta, thereby differentially affecting the exchange of nutrients, metabolic waste products, and signaling molecules across the placental barrier, which may affect brain development differentially in males and females.

Finally, LPS treatment may also have interfered with specific aspects of sexual differentiation. Sexual differentiation of specific neural circuits share components of signal transduction pathways used by inflammatory processes (Amateau and McCarthy 2004; Bale 2009; Petersen et al. 2012). If vasopressin cells in the developing BNST respond to inflammatory signals as they do in adult animals (Pittman et al. 1998), there may be cross-talk between sexual differentiation and immune signaling pathways in these cells that impact their proliferation, apoptosis, or neuronal phenotype preferentially in males.

5. Conclusion

During restricted periods, hormones direct the development of neural circuitry involved in specific functions and behaviors. It is relatively easy to understand how such programming effects of hormones can be time-restricted if it targets processes that take place during restricted periods in development, such as the massive wave of developmental cell death. Sensitive periods are more difficult to understand if the targets involve processes that take place throughout life. Long-term modifications of the chromatin may be an explanation in those cases. As this latter type of programming appears to constitute the lion’s share of developmental effects of hormones, interactions between hormones may take place at the epigenetic level as well. Focusing our attention on such interactions may help us understand how, for example, adverse conditions affect males and females differently. This may uncover novel therapeutic targets for behavioral and neurological disorders, the vulnerability for which has been shown to be influenced by programming effects of hormones.

Although this review focused mainly on hormonal programming during the pre- and postnatal period, hormonal programming may occur even earlier. Stress and nutritional state have effects that cross multiple generations, and epigenetic modifications are suspected to mediate such transgenerational effects (Junien et al. 2005; Dunn et al. 2011; Matthews and Phillips 2012). It is not outlandish to assume that hormones are early mediators of the effects of stress and nutrition on the ancestors’ physiology. If the ripples of those effects end up modifying the same mechanisms that are later targeted by the same hormones during progeny’s development, that would mean that the sensitive period for programming effects of these hormones extends into earlier generations, or, as Hugo De Vries stated it in developing the concept of sensitive periods, “[variability between organisms of the same strain] may for a large part, and perhaps wholly, be the result of the life-conditions of their parents and grandparents” (De Vries 1904).