In 1995, we discovered a new pathway in the cat, which originates from the nucleus retroambiguus (NRA) and terminates in a distinct set of lumbosacral hindlimb, axial, and pelvic floor motoneuronal cell groups [VanderHorst VGJM, Holstege G (1995) Caudal medullary pathways to lumbosacral motoneuronal cell groups in the cat: evidence for direct projections possibly representing the final common pathway for lordosis. J Comp Neurol 359:457–475]. The NRA is a compact group of interneurons located laterally in the caudal medulla oblongata. Its projection to lumbosacral motoneurons is thought to represent the final common pathway for male mounting and for female receptive or lordosis behavior. However, females only display lordosis behavior when they are in estrus, which suggests that the NRA–lumbosacral pathway is only active during estrus. This raised the question of whether estrogen affects this pathway. The effect of estrogen on the NRA–lumbosacral projection was studied light microscopically, using wheat-germ agglutinin horseradish peroxidase (WGA-HRP) as a tracer. The rubrospinal pathway served as control. The density of labeled NRA fibers in their target hindlimb motoneuronal cell groups appeared abundant in estrous and very weak in nonestrous cats. Such differences were not found in the rubrospinal pathway. For electron microscopical study, the NRA projection to the semimembranosus motoneuronal cell group was selected. In this cell group, an almost ninefold increase of labeled profiles was found in estrous versus nonestrous cats. Moreover, the semimembranous motoneuronal cell group contained labeled growth cones in estrous, but not in nonestrous, cats. The present study is the first to show that estrogen induces axonal outgrowth of a precisely identified pathway in the adult mammalian central nervous system. The possible mechanisms underlying this outgrowth are discussed.
- sexual behavior
- spinal cord
- nucleus retroambiguus
- caudal medulla
- growth cone
- ventral horn
- lordosis behavior
- sex steroid
- pelvic floor
- adductor longus
- biceps femoris
The nucleus retroambiguus (NRA) is a compact group of interneurons in the lateral tegmentum of the caudal medulla and has been described in humans, cats, rats, hamsters, and birds (Olszewski and Baxter, 1954; Merrill, 1970; Paxinos and Watson, 1986;Holstege, 1989; Ellenberger and Feldman, 1990; Wild, 1993) (P. Gerrits, V. VanderHorst, and G. Holstege, unpublished observations). Anatomically as well as physiologically, the NRA has been shown to be involved in respiration, defecation, vomiting, and vocalization (Merrill, 1971, 1974; Feldman, 1986; Fukuda and Fukai, 1986; Miller et al., 1987, 1995; Holstege, 1989; Zhang et al., 1992). It receives projections from respiration-related neurons in the brainstem (Feldman, 1986; J. C. Smith et al., 1989; Gerrits and Holstege, 1996), and from the midbrain periaqueductal gray (PAG) (Holstege, 1989; VanderHorst and Holstege, 1996). The PAG plays a crucial role in the integration of survival behavior such as defensive and aggressive reactions as well as mating (Sakuma and Pfaff, 1979a,b; Bandler et al., 1991; Ogawa et al., 1991).
In the female as well as in the male cat (VanderHorst et al., 1994; VanderHorst and Holstege, 1995, 1996, 1997b), the NRA projects directly to lumbosacral motoneuronal cell groups innervating a distinct set of hindlimb, axial, and pelvic floor muscles. Combined action of this set of muscles in the female does not serve motor activities as stepping, jumping, scratching, running, or other daily activities, but underlies aspects of the receptive posture during mating. Such behavior consists of elevation of the lower back (lordosis), rhythmic movements of the hindlimbs (treading), and lateral deviation of the tail (Michael, 1960). This led us to postulate that in female cats, the NRA–lumbosacral motoneuronal projection forms the final common pathway for lordosis behavior (VanderHorst and Holstege, 1995). Electromyographic studies are currently undertaken by this lab to provide evidence for our hypothesis.
Female reproductive behavior is not displayed in the absence of estrogen, whereas in estrous animals, it is prominently present (Beach, 1948; Young, 1961; Clark and Mani, 1994). If the NRA–lumbosacral motoneuronal projection represents the final common pathway for lordosis behavior, the question arises of whether estrogen has an effect on this pathway. Therefore, the pathway was studied light and electron microscopically both in estrous and in nonestrous cats. The results indicate that estrogen induces axonal outgrowth of NRA fibers to their target motoneurons in the lumbosacral cord.
MATERIALS AND METHODS
Ovariohysterectomy and estrogen treatment
NRA series. Anterograde tracing experiments were performed in 17 adult female cats (Table 1). Eight females were ovariohysterectomized 4–5 weeks before the tracer injection. Five of them were estrogen-treated for 7–10 d before the tracing experiment, receiving daily subcutaneous injections of estradiol benzoate dissolved in oil (Mycofarm, 0.02 mg/kg/d). After 3 d of treatment, they started to display the typical lordosis behavior when tapping the lower back or presenting them to a male cat. Estrogen treatment was continued until to the day of perfusion. The remaining three ovariectomized control cats received no estrogen. Progesterone was not administered, because cats are reflex ovulators in which progesterone levels start to rise only after intromission by the male or intense vaginocervical stimulation (Dawson and Friedgood, 1940).
Of the remaining nine nonovariectomized cats, two were in full estrus, i.e., displayed the complete pattern of estrous behavior during 3–4 d before the tracer injection. The other seven cats showed no signs of estrous behavior in the 2 weeks before or the 3 d after the tracer injection.
Red nucleus control series. In four females, the rubrospinal projections in estrous and nonestrous females were studied (Table 1). Three cases were ovariectomized, of which two did and one did not receive estrogen treatment before the tracer injections. The remaining natural case did not display estrous behavior in the 2 weeks before or the 3 d after the tracer injection.
NRA series. The distribution of the NRA–lumbosacral neurons in the cat have been described by VanderHorst and Holstege (1995) (see also Fig. 1). Because the NRA extends rostrocaudally over a length of 6–7 mm, multiple needle penetrations were necessary to inject wheat germ agglutinin-horseradish peroxidase (WGA-HRP, Sigma) throughout its rostrocaudal extent. This reduced the possibility that differences in the injection site would result in differences in the NRA–lumbosacral projections among the cases. WGA-HRP was pressure-injected via a glass micropipette using a picopump after dorsal approach and exposure of the caudal medulla. Except for cases 2237, 2251, 2256, 2258, 2267, and 2271, the injections were preceded by an ipsilateral C2 hemisection, which interrupted all ipsilaterally descending, non-NRA pathways to the spinal cord (VanderHorst and Holstege, 1995). The hemisections were made by aspiration with a glass pipette.
Red nucleus control series. To rule out the possibility that estrogen has a similar effect on other descending pathways, the rubrospinal tract was chosen as a control pathway. In two cases (2245 and 2310), small WGA-HRP injections were made in the red nucleus, and in two other cases (2362 and 2363), the red nucleus was injected with similar volumes of tracer as used in the NRA-injected cases (Table 1). In case 2310, before the injection an ipsilateral C2 hemisection was made.
General surgical and histological procedures
The surgical procedures, pre- and postoperative care, and handling and housing of the animals were carried out according to the protocols approved by the Faculty of Medicine of the University of Groningen. The animals were anesthetized with an initial dose of ketamine, 0.1 ml/kg, i.m. (Nimatek, A.U.V.) and xylazine hydrochloride, 0.1 ml/kg, i.m. (Sedamun, A.U.V.). Subsequently, they were artificially ventilated under mixed halothane–nitrous oxide anesthesia. During surgery, ECG and temperature were monitored.
After 3 d survival time, the animals were initially anesthetized with ketamine and xylazine hydrochloride followed by an intraperitoneal injection of 6 ml of 6% pentobarbital sodium (Nembutal). Subsequently, they were transcardially perfused with 2 l of heparinized saline at 35°C, followed by 2 l of fixative containing 2% glutaraldehyde/1% paraformaldehyde in 0.1 m phosphate buffer, pH 7.2–7.4, at room temperature. Brains and spinal cords were removed and post-fixed for 1–2 hr. In the cases exclusively used for light microscopy, 4% sucrose was added to the fixative, and the tissue was stored overnight in 25% sucrose buffered phosphate at 4°C. After perfusion, brain and spinal cord were removed, and the injection site was cut on a freezing microtome into 40-μm-thick transverse sections and incubated using a standard diaminobenzidine (DAB, Sigma) procedure. The hemisected segments were processed to verify that the hemisections were complete. Finally, the sections were mounted on slides, dehydrated, and coverslipped with DePeX mounting medium (Brunschwig chemie).
The L3–S3 segments were cut on a freezing microtome into 40-μm-thick transverse sections, except in eight cases (2251, 2256, 2288, 2307, 2308, 2324, 2337, and 2353), which were cut on a vibratome into 60 μm sections. Every fourth section was processed using the tetramethylbenzidine (TMB, Sigma) procedure according to Mesulam (1982). Sections of the NRA-injected cases 2296 and 2299 and 2307 and 2308 and of the red nucleus cases 2362 and 2363 were processed in the same TMB procedure to eliminate differences in staining due to different reaction conditions. The density of anterograde labeling in the spinal cord was microscopically examined with a Zeiss Axioskop under combination of polarized light and dark-field condensor. Using a new, detailed overview of the spatial location of lumbosacral motoneuronal cell groups (VanderHorst and Holstege, 1997), it could be determined with great precision which of these cell groups received NRA afferents. Photomicrographs of representative sections were taken.
Cases 2251, 2256, 2288, 2307, 2308, 2324, 2337, and 2353 were examined light- and electron microscopically. The L3–S3 segments of these cases were cut on a vibratome. Every third section was incubated with TMB and ammonium heptamolybdate overnight (Olucha et al., 1985). The next day, they were processed using the slow-osmication method for post-fixation of Henry et al. (1985). The tissue was stained en bloc in 1% uranylacetate in aquabidest, and the slabs were dehydrated in graded series of alcohol and embedded in Epon between dimethyldichlorosilane-coated glass slides (Vinores et al., 1984).
Using a new, detailed overview of the spatial location of lumbosacral motoneuronal cell groups, it could be determined with great precision which of these cell groups received NRA afferents. A selection was made of sections containing anterogradely labeled fibers in the semimembranosus motoneuronal cell group. This is a compact cell group that does not overlap with other motoneuronal cell groups, except for its most rostral and most caudal poles (Romanes, 1951, 1964) (VanderHorst and Holstege 1997a), which made it possible to identify it in unstained sections. The selected tissue was cut into ultrathin sections (60 nm) and studied using electron microscopy. To determine the density of labeled terminal profiles per area, in each case the labeled terminal profiles were counted in 32 grid squares covering the semimembranosus motoneuronal cell group, each grid square measuring 10,000 μm2. The counts were made blind to the condition of the animals. The symmetry or asymmetry of the synaptic membrane specialization and the content of the labeled profiles were established. In cases 2324 and 2337, the perimeter of labeled terminal profiles and the length of postsynaptic densities were measured.
Location and size of the injection sites in the NRA
In all 17 cases, the injections involved the NRA over a considerable rostrocaudal extent (Fig. 2, Table 1). The hemisections were complete and did not extend across the midline. Examples of the DAB injection sites are shown in cases 2288 and 2324 (Fig. 3).
In the lumbosacral cord of all natural and ovariectomized nonestrous cases (Table 1, Fig. 2), the abdominal wall and pelvic floor motor nuclei received a substantial projection (Fig. 4) (see Holstege and Kuypers, 1982; Feldman et al., 1985; Holstege and Tan, 1987; Miller et al., 1989; Holstege, 1989). In addition, labeled fibers terminated in the hindlimb motoneuronal cell groups (Fig.5, left), but they were so sparse that in single sections it was not possible to determine which hindlimb motor nuclei were their main target. However, superposition of six consecutive sections, which were taken from a series in which every fourth section was collected, revealed a distinct projection pattern to the motoneuronal cell groups innervating the muscles iliopsoas, adductor longus, semimembranosus, semitendinosus, biceps femoris anterior and posterior, external anal and urethral sphincter, levator ani, abductor caudae internus, medial longissimus, and multifidi (VanderHorst and Holstege, 1995).
In the semimembranosus motoneuronal cell group of cases 2251, 2256, 2308, and 2324, a total number of 25, 9, 12, and 38 labeled terminal profiles were counted per 32 grid squares (320,000 μm2), respectively (Table 1). Synapses, which were all asymmetrical, were present in 22% of these profiles. Furthermore, 66% of the labeled profiles contained exclusively spherical vesicles, whereas 13% contained spherical as well as a few dense-cored vesicles (Table 2). Of the remaining 22%, the vesicle content could not be identified. Apart from spherical and dense-cored vesicles, labeled terminal profiles frequently contained some clustered mitochondria (Fig. 6, left). Labeled profiles with flattened or pleiomorphic vesicles were never observed.
In one case (2324), the average perimeter of labeled terminal profiles was determined and amounted to 7.41 μm (n = 38; ranging from 3.29 to 14.43 μm). On average, 6.9% of it was covered by synaptic junctions. Labeled terminal profiles predominantly contacted dendrites (Fig. 6, left) and only very occasionally neuronal somata.
In all natural and ovariectomized estrous cases, the NRA–lumbosacral projection was remarkably dense. In sharp contrast to the nonestrous cases, the projections to motoneuronal cell groups innervating the iliopsoas, adductor longus, semimembranosus, semitendinosus, biceps femoris anterior and posterior, and levator ani/abductor caudae internus were very prominent and could be easily discerned in single sections (Fig. 5, right). The differences between estrous and nonestrous cases in respect to the density of projections to the abdominal or pelvic floor motoneuronal cell groups were less apparent. These projections are prominent in estrous and in nonestrous cases (Fig. 4).
Among the cases of the estrous group, some differences in density of the NRA–lumbosacral projection were present. In natural estrous case 2288 and in ovariectomized estrogen-treated case 2361, the projections were extremely strong. In case 2320, with a small injection in the NRA, labeled fibers were not as numerous as in the other estrous cases, but still far outnumbered the NRA fibers in the nonestrous cases, including those with large NRA injections (Fig. 2, Table 1).
The number of labeled terminals per 320,000 μm2 in the semimembranosus motoneuronal cell group of estrous cases 2288, 2307, 2353, and 2337 amounted to 281, 112, 183, and 157, respectively (Table 1). The average number of labeled terminal profiles per grid square differed significantly between the group of estrous and the group of nonestrous cases (Wilcoxon–Mann–Whitney test,p < 0.025) (Fig. 7). In 14% of the labeled terminal profiles in the four estrous cases, synapses were observed, which were all asymmetrical. Of the labeled profiles, 63% contained spherical vesicles and 11% spherical as well as dense-cored vesicles. In the remaining 26%, the vesicle content could not be identified (Table 2). No labeled terminals with flattened pleiomorphic vesicles were observed. The labeled terminal profiles primarily contacted dendrites and only very occasionally neuronal somata.
In case 2337, the average perimeter of the labeled terminal profiles was 9.49 μm (n = 73; ranging from 5.25 to 22.25 μm), which is 1.28 times larger than in the nonestrous case 2324 (7.41 μm). In these labeled profiles, synaptic complexes on average formed 5.0% of the perimeter, which is less than in the nonestrous case 2324 (6.9%).
In the area under examination (the semimembranosus motoneuronal cell group), the absolute number of labeled terminals displaying synapses was much higher in the estrous cases as compared with nonestrous cases (100 and 18, respectively) (see Table 2). The percentage of labeled profiles showing one or more synaptic junctions was larger in nonestrous cases than in estrous cases (21 and 14%, respectively).
In all estrous cases, but not in nonestrous cases, very large labeled structures were found that contained large quantities of tightly packed mitochondria (297, 159, 86, and 32 mitochondria per 35.6, 25.4, 12.9, and 8.5 μm2), and some agranular reticulum, a few coated, dense-core, and large granulated vesicles, lysosomes, electron-dense particles, and microtubuli (Figs. 8, left,9). The mitochondria had a smaller diameter and were more elongated than the mitochondria in adjacent structures (Figs. 8,9). Similar structures have been described as the proximal or central part of neuronal growth cones in the developing CNS (Tennyson, 1970;Yamada et al., 1971; Bunge et al., 1973; Bridgman and Dailey, 1989;Davis et al., 1992), which leads to the conclusion that the large labeled profiles in the present material (4.6% of the labeled profiles in the four estrous cases) represent the proximal parts of growth cones.
Furthermore, labeled, large terminal-like profiles were observed that contained small-diameter elongated mitochondria, cisternae of agranular reticulum, spherical synaptic vesicles, dense-cored vesicles, large granulated vesicles, vacuoles, coated vesicles, and a few microtubules and neurofilaments (Figs. 8, 10). Similar assemblies of organelles have been described in growth cones and immature terminals (Tennyson, 1970; Yamada et al., 1971; Bunge et al., 1973; Vaughn and Sims, 1978; Knyihar-Csillik et al., 1986; Bridgeman and Dailey, 1989;Peters et al., 1991). The presence of large numbers of coated vesicles in the NRA terminals suggests a high level of membrane turnover (Rees et al., 1976), which is thought to be involved in the formation of axonal collaterals (Vaughn and Sims, 1978). At the postsynaptic site, mitochondria, multivesicular bodies, coated vesicles and polyribosomes were found frequently (Fig. 10). Postsynaptically located polyribosomes, which have been described to be particularly prominent during periods of synapse growth, might synthesize proteins that are important for axonal outgrowth or maturation of the synaptic junction (Steward and Falk, 1986, 1991).
In conclusion, many of the labeled NRA profiles in estrous cases contain growth cones and immature terminals, which indicates that outgrowth of NRA axons takes place when the animal is in estrus.
Rubrospinal control series
In four cases, WGA-HRP was injected in the red nucleus and surrounding tegmentum (Fig. 11). In all these cases, the density of anterogradely labeled fibers in the intermediate zone of the lumbar enlargement was denser than in the motoneuronal cell groups of NRA-injected cases and was easily visible using bright-field illumination. The two large injections resulted in a slightly stronger projection (nonestrous case 2362 and estrous case 2363) than the smaller red nucleus injections (nonestrous case 2245 and estrous case 2310). However, comparing the estrous with the nonestrous cases with similar injection sites did not reveal any difference in the density of labeling at the light microscopical level.
The present results demonstrate that the density of the pathway from the NRA to hindlimb motoneuronal cell groups shows significant estrogen-related differences in adult female cats. Electron microscopical results confirmed the light microscopical observations, demonstrating an almost ninefold increase in the number of NRA terminal profiles in estrous cases. These major differences cannot be fully explained by the small differences in the size of NRA profiles in estrous and nonestrous cases (estrous vs nonestrous = 1.28:1). The finding of labeled axonal growth cones in the semimembranosus motoneuronal cell group in estrous cats, which were never found in any of the nonestrous cats, demonstrates that the difference in number of labeled profiles is probably based on the formation of new NRA terminals. The finding that the percentage of labeled profiles with one or more synaptic junctions decreases in estrous as compared with nonestrous cases is in line with this hypothesis. Furthermore, electron microscopical examination of the semimembranosus cell group in estrous and nonestrous cases showed that the labeled profiles contained primarily spherical and dense-cored vesicles and formed asymmetrical synapses, suggesting a primarily excitatory role for the NRA–lumbosacral pathway (see Holstege, 1989). The NRA profiles can be classified as S- or NFs-type (Conradi, 1969; McLaughlin, 1972). Presently, a study is underway to find out whether the estrous increase in the number of labeled terminal profiles occurs at the expense of unlabeled ones or whether it is caused by a total increase in the number of terminal profiles (labeled and unlabeled) per unit area.
It is possible that estrogen increased the efficiency of WGA-HRP transport, thus resulting in densely labeled axons in estrous and weakly labeled axons in nonestrous animals. Although this possibility cannot be excluded, it is unlikely for several reasons. (1) In both nonestrous and estrous animals, a dense NRA projection was present to the abdominal wall and pelvic floor motoneuronal cell groups. These findings indicate that NRA neurons excellently transport the tracer along their axons throughout the length of the spinal cord in nonestrous cases. Moreover, the few dispersed NRA axons in nonestrous cases were very well labeled. It seems unlikely that steroids selectively affect transport efficiency of only a subpopulation of NRA axons. (2) Light microscopically, in the lumbosacral cord of the nucleus ruber-injected cases, no differences in the density of labeling were observed between nonestrous and estrous animals. (3) In all estrous cases, many growth cones (labeled and unlabeled) were observed in the semimembranosus motoneuronal cell group. In contrast, in all samples of the nonestrous cases, only one example of a growth cone (unlabeled) was found in the semimembranosus motoneuronal cell group. (4) Systemic administration of androgen, which induces outgrowth of motoneuronal dendrites in adult rats (see Kurz et al., 1986), does not affect transport of cholera toxin-HRP by motoneurons (Leslie et al., 1991).
In summary, the number of probable excitatory NRA profiles in the semimembranosus motoneuronal cell group increases enormously in estrous cases as compared with nonestrous cases. The presence of labeled growth cones in estrous (natural as well as ovariectomized estrogen-treated) and the absence of such structures in nonestrous cats (natural as well as ovariectomized) suggest that the difference in density in the NRA–lumbosacral projection is based on estrogen-induced outgrowth or sprouting of NRA axons (Fig. 12).
Possible mechanisms underlying sprouting of NRA axons
The finding that estrogen induces outgrowth of the NRA–lumbosacral pathway leads to the question of which mechanisms are involved in this process. The following possibilities underlying the axonal growth are discussed: activity of the NRA–lumbosacral pathway, effects of estrogen on induction of protein synthesis by genomic activation via intracellular estrogen receptors in neurons, muscles, and astroglia, and changing the membrane excitability via estrogen receptors in the cell membrane.
Activity of the NRA–motoneuronal pathway
Theoretically, activity of the NRA–motoneuronal pathway might induce the growth in this pathway. This is not a likely option, because estrogen administration alone does not activate this pathway. Only after adequate sensory stimuli, which are naturally evoked by a mounting male, does the female display the receptive posture. This activation only occurs when the female is in full estrous and not at all when she is in nonestrous. Thus, increased activity of the NRA pathway to hindlimb motoneuronal cell groups appears to take place after the outgrowth has taken place.
Genomic activation via intracellular estrogen receptors
Estrogen-induced protein synthesis is a relatively slow response (hours to days) (Pfaff and McEwen, 1983; Clark and Mani, 1994; Pfaff et al., 1994). It is mediated via intracellular receptors, which when bound to estradiol, activate a specific DNA target (Halachmi et al., 1994). The resulting newly synthesized proteins are transported down the axon (Pfaff and McEwen, 1983; Pfaff et al., 1984; Mobbs et al., 1988), where they are thought to be involved in plastic changes (seePfaff et al., 1994). Estrogen-related plasticity has been described in cell populations containing intracellular estrogen receptors (Pfaff and Keiner, 1973; Stumpf et al., 1975; Stumpf and Sar, 1976; Rees et al., 1980). These cell groups have a facilitating effect on lordosis behavior in adult mammals. Examples are the lateral septum (Miyakawa and Arai, 1987), ventrolateral part of the ventromedial hypothalamus (Carrer and Aoki, 1982), and PAG (Chung et al., 1988, 1990). Such plastic changes occur within 24–48 hr, in parallel with the estrous cycle in the rat and hamster (Olmos et al., 1989; Frankfurt et al., 1990; Meisel and Luttrell, 1990; Frankfurt and McEwen, 1991; Langub et al., 1994).
The question is whether a similar mechanism underlies axonal outgrowth in the NRA–lumbosacral pathway. Neither the NRA nor its target motoneuronal cell groups contain estrogen receptors or concentrate estrogen (Stumpf et al., 1975; Stumpf and Sar, 1976; Rees et al., 1980;Morrell et al., 1982) (E. Meijer, V. VanderHorst, and G. Holstege, unpublished observations). However, in the rat, the lateral PAG contains estrogen-concentrating neurons, some of which project to the caudal medulla oblongata (Corodimas and Morrell, 1990). In the cat, estrogen receptor-containing PAG neurons target the NRA (V. VanderHorst, E. Meijer, and G. Holstege, unpublished observations). Other estrogen-concentrating structures in the forebrain, such as the ventrolateral part of the ventromedial hypothalamus, medial preoptic area, and amygdala, do not have direct connections with the NRA interneurons (Holstege, 1991). Possibly, the estrogen-concentrating neurons in the PAG play a major role in preparing the NRA–lumbosacral pathway for its specific action during lordosis behavior.
Another possibility is that estrogen exerts an effect on the NRA-motoneuronal pathway via the muscles involved (for review, seeHaydon and Zoran, 1994). In this respect, in humans it has been demonstrated that the pelvic floor muscle levator ani, but not the rectus abdominis muscle, contain estrogen receptors (Smith et al., 1990, 1993). This would correspond with the finding that, at least in the cat, the motoneuronal cell groups innervating pelvic floor muscles receive NRA afferents (Holstege and Tan, 1987; VanderHorst and Holstege, 1995), whereas the rectus abdominus motoneurons are not targeted by the NRA (Holstege, 1989). Possibly, under high levels of estrogen, estrogen receptor-containing muscles send a retrograde signal to their motoneurons, which in turn induce outgrowth of NRA axons. This option, in which the NRA target motoneurons play a role in the outgrowth of NRA axons, is supported by the finding that the outgrowth is directed to only these motoneurons.
In addition to its action via estrogen receptors in neurons or muscles, estrogen might also have an effect on axonal growth and the formation of new terminals via astroglia (see Theodosis and Poulain, 1993). Intracellular estrogen receptors have been demonstrated in astrocytes in the medial preoptic area and median eminence (Langub and Watson, 1992). In the present study, astrocytes were frequently observed in close proximity to labeled profiles (see Figs. 8, 9).
Estrogen effects on membrane excitability
Estrogen has also been shown to change the membrane excitability of neurons within minutes (see Alcaraz et al., 1969; Yagi, 1970; Kelly et al., 1977; Dufy et al., 1979; Levesque and Di Paolo, 1988; Schumacher, 1990; Smith, 1994). Such rapid effects have been described for numerous cell groups such as the anterior hypothalamus (Kawakami et al., 1970; Cross and Dyer, 1972; Alcaraz et al., 1969), medial amygdala (Nabekura et al., 1986), hippocampus (Wong and Moss, 1991, 1992), neostriatum (Levesque and Di Paolo, 1988; Thompson and Moss, 1994; Mermelstein et al., 1996), and cerebellum (S. S. Smith et al., 1987, 1989). This effect seems to be mediated via cellular membrane receptors (Schumacher, 1990) that affect L-type calcium channels (Mermelstein et al., 1996). The extent to which this mechanism plays a role in the NRA axonal outgrowth is not known.
Functional implications of estrogen-induced axonal sprouting in the NRA–lumbosacral pathway
The female sex steroid estrogen has been shown to induce dendritic growth or synaptic plasticity in mammalian mid- and forebrain structures (see, for example, Matsumoto and Arai, 1979; Carrer and Aoki, 1982; Miyakawa and Arai, 1987; Chung et al., 1988, 1990; Langub et al., 1989, 1994; Frankfurt et al., 1990; Meisel and Luttrell, 1990;Frankfurt and McEwen, 1991; Matsumoto, 1991; Woolley and McEwen, 1992a,b, 1994; Garcia-Segura et al., 1994) as well as in non-neuronal structures such as the uterus (Burrows, 1949; Reynolds, 1951) (for review, see Clark and Mani, 1994). In the adult mammalian spinal cord, male sex steroids are known to increase the length of motoneuronal dendrites (Kurz et al., 1986). These changes in dendritic length are acompanied by differences in synaptic input of unidentified origin (Leedy et al., 1987; Matsumoto et al., 1988). The present study is the first to show that the female sex steroid estrogen induces axonal outgrowth of a precisely identified pathway in the adult mammalian CNS. This outgrowth takes place in a long brainstem–spinal motor tract, which is thought to represent the final common pathway for lordosis. The effect of estrogen on this pathway seems to be rather specific, because no estrogen-related differences could be detected in the rubrospinal motor tract.
These findings are in line with the notion that estrogen or other sex steroids are necessary for activation of the reproductive neural circuitry, which appears to be latently present in nonestrous animals.
We thank Henk de Weerd for his work on the electron microscopic study; Ellie Meijer and Klaas van Linschoten for their histotechnical support; Peter van der Sijde for photographic assistance; and Han van der Want (Laboratory for Cell Biology and Electronmicroscopy, University of Groningen) and Jan Carel Holstege (Department of Anatomy, Erasmus University, Rotterdam) for their advise on the electron microscopic study.
Correspondence should be addressed to Dr. Veronique G. J. M. VanderHorst, Department of Anatomy, 0ostersingel 69, 9713 EZ Groningen, The Netherlands.