The studies described above establish beyond any doubt that many behavioural effects of T actually result from the action at the cellular level of endogenously produced metabolites, and in particular of oestrogens derived from T aromatisation. Several questions are raised by this observation: where does the T metabolism take place (brain vs periphery)? Is the metabolism a constant feature of the species or is it modulated by the age, sex and/or hormonal status of the subject? If T metabolism takes place in the brain, what is the anatomical specificity of the enzyme distribution at the macroscopic and at the cellular level?

These questions have been approached by two types of techniques. In vitro radioenzyme assays combined with a variety of micro-dissection techniques have been used to identify the presence of specific enzymes in a series of brain nuclei and to analyse the potential regulations of these enzymatic activities. More recently, immunocytochemical and in situ hybridisation techniques that permit the visualisation of the aromatase protein or of its messenger RNA have additionally become available and studies have been carried out to analyse the cellular distribution of this enzyme in the brain of a few avian species. The most extensive data have been collected in Japanese Quail and Ring Doves (Balthazart & Foidart 1993; Panzica et al. 1996; Balthazart 1997; Balthazart et al. 1996c; Hutchison 1991; Hutchison 1993; Hutchison et al. 1995) but a substantial amount of work has also been performed in one songbird species that breeds easily in laboratory conditions, the Zebra Finch.

High levels of aromatase were observed in the diencephalic and limbic nuclei that were classically known to express this enzyme in mammals and other avian species, namely the nucleus preopticus anterioris (POA), the nucleus periventricularis magnocellularis (PVM), the nucleus medialis hypothalami posterioris (PMH), the nucleus stria terminalis (nST) and the nucleus taeniae (Tn; homologue to parts of the mammalian amygdala).

Interestingly, these studies also reported the presence of significant levels of aromatase activity in telencephalic song control nuclei such as the area X (X), MAN, HVC and RA. They also indicated that the area parahippocampalis (APH), that had originally been selected as a control region, was characterised by an extremely high enzymatic activity larger than in any of the diencephalic or limbic nuclei that had been studied (Vockel et al. 1990a). Additional studies confirmed the presence of a very high aromatase activity in the dorsal telencephalon and showed that the enzyme is actually present in the entire roof of the forebrain in the neostriatum, hippocampal and parahippocampal region. This high level of telencephalic aromatase activity has been independently confirmed (Schlinger & Arnold 1991; Schlinger & Arnold 1992b) and a series of elegant studies based on the injection of radioactive androgens in the brain or periphery combined with blood collection in the jugulars or carotids has provided experimental evidence indicating that brain aromatase substantially contributes to the circulating levels of oestrogens in the Zebra Finch (Schlinger & Arnold 1992a; Schlinger & Arnold 1993).

The relationships between aromatase activity and oestrogen receptors can be evaluated in more detail and this comparison has revealed some striking discrepancies. Some brain areas such as the preoptic-hypothalamic nuclei (POA, PVM and PMH) and limbic regions (Tn and nSt) contain oestrogen receptors as demonstrated by both autoradiography (Nordeen et al. 1987) and immunocytochemistry (Gahr et al. 1987; Gahr & Konishi 1988; Gahr et al. 1993) as well as a high aromatase activity and a high density of aromatase-expressing cells. Some mismatches can however be observed. In the nucleus intercollicularis (ICo), identified in all studies as a site with dense oestrogen receptors (autoradiography: Nordeen et al. 1987, ICC: Gahr et al. 1987; Gahr et al. 1993, binding assays on punched nuclei: Walters et al. 1988) enzyme assays reveal no or barely detectable levels of enzyme activity and no aromatase-expressing cells can be detected by ICC or ISH. In contrast, high levels of aromatase activity are observed in a large portion of the dorsal telencephalon (hippocampus, area parahippocampalis, dorsal neostriatum) while these brain regions are not known as a classical target for steroids. Anatomical studies that have documented the distribution of aromatase and oestrogen receptors in these brain regions also confirm the presence of broad discrepancies between areas expressing the enzyme and the receptor. Oestrogen receptors have been demonstrated in a large neostriatal area located under the lateral ventricle at the level of HVC in the Zebra Finch brain (Gahr et al. 1987) where a high aromatase activity and numerous aromatase-expressing cells are detected. However aromatase-immunoreactive cells and cells expressing aromatase mRNA are found beyond this specific region of the neostriatum namely in the area dorsal to the lamina archistriatalis dorsalis and in the ventrolateral aspects of the neostriatum where no oestrogen receptors have been detected so far (Gahr et al. 1987). These facts therefore raise questions concerning how oestrogens locally produced in the dorsal telencephalon may affect directly or indirectly the control of reproduction and of singing behaviour.

During enzyme assays performed on cell fractions separated by differential centrifugation of the quail preoptic area (Schlinger & Callard 1989) or the Zebra Finch telencephalon (Schlinger & Arnold 1992b), it has been observed, that aromatase activity is found not only in the microsomal but also in synaptosomal (pinched-out nerve terminals) subfractions. The presence of aromatase-immunoreactive material has furthermore been confirmed in cell processes of the quail brain that have an axonal morphology (Foidart et al. 1994b) and, at the electron microscope level, it was confirmed that aromatase-immunoreactive material is present in high density at the level of presynaptic boutons (Naftolin et al. 1996). This local production of oestrogens in the close vicinity of synapses may provide steroid signals that could affect brain activity independent of the binding to nuclear receptors acting as transcription factors (i.e. by non-genomic mechanisms). It is known that oestrogens are able to modify the electrical activity of the neuronal membranes within seconds after their application which essentially rules out the possibility of a classical steroid receptor-mediated effect on protein synthesis (for additional discussion see: Blaustein & Olster 1989; Schlinger & Callard 1989; Schumacher 1990; McEwen 1994; Ramirez et al. 1996). Other effects of oestrogens on brain functioning that appear to be independent of their binding to intracellular receptors have also been described (e.g. Pasqualini et al. 1995; Thompson & Moss 1994; Mermelstein et al. 1996). Whether the presence of aromatase activity in brain areas such as part of the neostriatum that are apparently devoid of intracellular oestrogen receptors relates to these membrane effects of oestrogens is unknown at present.

In general, the partial mismatch between the T-metabolising enzymes and the steroid receptors raises the question of how T metabolites exert their action on the brain. It was classically assumed that all steroids act in an autocrine fashion and exert their biological effects through the binding with intracellular receptors located in the same cell as the enzymes. The presence of aromatase at the presynaptic level certainly suggests that a paracrine mode of action is also possible. The demonstration that infusion of radioactive androstenedione in the Zebra Finch telencephalon results in increased levels of radioactive oestrogen in the jugular blood (Schlinger & Arnold 1992a; Schlinger & Arnold 1993) even suggests an endocrine action for brain aromatase. This represents without a doubt one of the most exciting recent discovery in the field and its biological significance should be experimentally evaluated.

The specific role of oestrogens in the control of the sexual differentiation of the song control nuclei and in the activation of singing behaviour remains at this point unclear. Anatomical studies have revealed that the song control nuclei MAN, HVC and RA whose size and cell typology can be affected by oestrogens acting during ontogeny (Gurney 1981; Gurney & Konishi 1980; Konishi & Akutagawa 1988) appear to be devoid of aromatase-expressing cells, even if such cells are found in the close vicinity in the neostriatum. In addition, these three nuclei apparently contain no or very few oestrogen receptors (Gahr et al. 1987; Nordeen et al. 1987; Walters et al. 1988) at least in the adult bird. Immunoreactive oestrogen receptors could be present transiently at least in HVC during ontogeny (Gahr & Konishi 1988). It is therefore difficult to understand precisely how oestrogens affect the differentiation of these song control nuclei as well as the vocal behaviour in adulthood (Balthazart & Ball 1995; Arnold 1997; Schlinger 1998). It must be pointed also that the neostriatal production of oestrogen may relate to other functions such as the control of memory capabilities, in particular the spatial memory and encouraging results along these lines have been recently obtained (Schlinger 1997). The medial neostriatum adjacent to the lateral ventricle is also known as an area of active neurogenesis in the adult songbird brain (Alvarez-Buylla et al. 1990). A role of oestrogens in the survival, migration or differentiation of these newly formed neurons should therefore also be considered and experimentally tested. The biology of oestrogen action in the songbird brain is clearly far from being completely uncovered.

Biochemical studies carried out on microdissected brain regions or nuclei demonstrated that the activity of the T-metabolising enzymes in Zebra Finches is markedly affected by the age and sex of the birds (Balthazart et al. 1986; Vockel et al. 1988; Vockel et al. 1990a; Vockel et al. 1990b). As an example, the aromatase activity measured in limbic nuclei and in telencephalic areas centred on the song control nuclei (they included the nuclei but were probably contaminated by surrounding tissue; see above) of young (20 day old) and adult (>100 day old) males is shown in Fig. 3A.

As can be immediately seen, the enzymatic activity observed in our assay conditions was usually higher in the adult than in the juvenile birds. This was true for both the limbic and the song control nuclei. The interpretation of such a difference is, however, complicated by the fact that the apparent affinity (Km) of the aromatase for its substrate seems to vary with age in Zebra Finches (Vockel et al. 1988), contrary to what had been observed previously in quail (Hutchison & Schumacher 1986). The double reciprocal plots of a saturation analysis shown in Fig. 3B illustrates this fact. In very young birds (below 35 days post-hatch), a high hypothalamic aromatase activity was detected (Vmax = 350 fmol/mg protein/h) but it was associated with a low affinity of the enzyme for its substrate (Km =70 nM). As the birds became sexually mature (40-45 days post hatch), a shift in the enzyme affinity occurred and the Km of the aromatase for T reached its adult values that are typically below 30 nM. An increase in the enzyme maximal velocity was subsequently observed as birds grew older (80-85 day old group).

Because important changes in enzyme affinity are observed with the increasing age, it is likely that the changes in maximum velocity reflect more than a simple change in enzyme concentration. Changes in the apparent enzyme affinity (Km) of aromatase can reflect modifications of the enzyme (its maturation) or the presence of a new form of enzymatic protein or the modulation of the enzymatic activity by new regulatory molecules (co-factors, inhibitors). Discriminating between these alternative explanations would require detailed biochemical and enzymatic studies. Extrapolating the in vitro assays to the physiology in the intact bird is also difficult and, in particular, it is hard to determine, based on these data, the actual activity level of the brain aromatase in the differentiating embryo or neonate. It must also be stressed that these kinetic data are based on studies of hypothalamic homogenates and it is not known whether similar changes would be detected in the telencephalic areas.

The fact that a major shift in the kinetic characteristics of the Zebra Finch aromatase occurred at the age of 40-45 days when the birds begin their sexual maturation as evidenced by molting into adult plumage and change in beak colour from black to red (Pröve 1983) suggests that these changes are induced by steroids, a conclusion that was indirectly supported by the observation that aromatase activity was higher in the hypothalamus of adult males compared to adult females (Vockel et al. 1988).

A different pattern emerged however in the telencephalic samples. In areas such as the dorsal neostriatum or the APH, no clear sex difference in enzyme activity could be detected in sexually mature, gonadally intact birds. Furthermore, gonadectomy associated or not with a chronic treatment with exogenous T did not seem to affect much enzyme activity (no change at all in APH, small numerical increase in the neostriatum) but the changes failed to reach statistical significance; (Fig. 4B). The data therefore suggested that telencephalic aromatase is not sexually differentiated and not regulated by T. A lack of a sex difference in the telencephalic aromatase activity has also been reported in young Zebra Finches (Schlinger & Arnold 1992b).

It must also be mentioned here that in the original papers reporting on these data, a ‘reverse’ sex difference in aromatase activity had been indicated for the song control nuclei HVC and RA (higher enzyme activity in females than in males; Vockel et al. 1990a) and this difference did not appear to be caused by a differential activation by steroids because it was not affected by gonadectomy and T treatment (Vockel et al. 1990b). These studies were however carried out when it was technically impossible to localise aromatase-containing cells at the cellular level so that microdissections could only be guided by the tissue appearance in Nissl stain and not by the actual distribution of the enzyme. As noted above, when anatomical methods for the visualisation of aromatase became available, it became clear that HVC and RA actually contain little or no aromatase so that the enzyme activity that had been measured in these nuclei presumably resulted from a contamination by surrounding positive tissue. Because HVC and RA are significantly larger in males than in females and punches of the same size had been used in both sexes to measure aromatase activity, it now appears likely that the sex difference that had been noted resulted from a more important contamination by surrounding tissue in females than in males.

It remains true, however, that there appears to be a differential regulation of aromatase in the telencephalic area and in the limbic structures. This suggests that different forms of enzyme may be present in these two parts of the brain. Recent studies utilising the tools of molecular biology actually bring some support to this idea (Ramachandran et al. 1997) but it is beyond the scope of the present study to analyse the molecular biology of aromatase regulation (see: Simpson et al. 1991; Bulun & Simpson 1994; Simpson et al. 1994; Lephart 1996; Lephart 1997 for recent reviews).

The data summarized above demonstrated that aromatization of T is a key factor in the regulation of reproductive behaviour in one oscine species, the Zebra Finch, like it is in mammals and several previously investigated avian species. They also showed that in hypothalamic and limbic areas the Zebra Finch aromatase is distributed and regulated by steroids like in quail and doves. Interestingly, however, an extremely active broadly distributed aromatase was found in the Zebra Finch telencephalon that seems to distinguish this species from other avian species that had been previously considered. This telencephalic aromatase also does not seem to be controlled by steroids like the hypothalamic enzyme.

It is quite tempting to generalize these features of the Zebra Finch aromatase to the entire songbirds order (passeriformes) but until quite recently data supporting this generalization were critically missing with the exception of limited studies devoted to a few species such as the Brown-headed Cowbird Molothrus ater, White-crowned Sparrow Zonotrichia leucophrys, House Sparrow Passer domesticus, or Island Canary Serinus canaria (Saldanha & Schlinger 1997; Schlinger et al. 1992; Schlinger 1996). Such generalisations may appear quite hazardous for two reasons. On the one hand, most Zebra Finches studied in the laboratory are derived from semi-domesticated stocks and they may consequently exhibit significant differences by comparison with the wild species but these are difficult to appreciate. On the other hand, the Zebra Finch is a non-photoperiodic species that inhabits the australian deserts and its use as a model species for the songbird group that contains many photoperiodic species originating from the temperate zone appears questionable. For these reasons, it appeared desirable to investigate aromatase distribution, regulation and function in a broader variety of songbird species. Additionally, it was also of interest to research whether the dramatic changes in hypothalamic aromatase activity that had been observed following castration and/or treatment with exogenous T would also be detected after more physiological changes in plasma T such as those that are normally observed during a breeding cycle .

Therefore, in a first set of studies, we researched whether the high level of telencephalic aromatase that had been detected in Zebra Finches would also be found in a variety of wild European songbird species caught in the wild during the breeding season. Males from four different species of songbirds belonging to four different families were for this purpose caught by mist nests in southwest Sweden at a field site close to Göteborg. These included 7 chaffinches (Fringilla coelebs, Fringillidae; caught between May 29 and June 6 with fully developed testes), 7 Pied Flycatchers (Ficedula hypoleuca, Muscicapidae; caught between May 28 and June 7 while defending secondary territories), 7 great tits (Parus major, Paridae; caught between June 5 and 26 while feeding nestlings [n=6] or when the female was incubating [n=1]) and 6 willow warblers (Phylloscopus trochilus, Sylviidae; caught on May 30). Aromatase activity was measured in the brain of these birds and compared to the activity found in previously studied species, namely the Zebra Finch (Taeniopygia guttata, Estrildidae, 6 two-year old males raised in the laboratory), Japanese Quail (Coturnix japonica, Phasianidae, 5 adult sexually mature males raised in the laboratory) and ring dove (Streptopelia roseogrisea [= risoria], Columbidae, 2 males raised and kept in an outdoor aviary and killed in December).

As can be observed in Fig. 5A, this study confirmed that in Zebra Finches, contrary to quail and doves, the aromatase activity, expressed per mg protein is larger in the telencephalon than in any other brain area including the diencephalon. The same situation was observed in the flycatcher but not in the other songbird species that were considered (chaffinch, tit, warbler), therefore questioning the generalization of this pattern to the passeriformes order. One must however consider that the telencephalon is at least one order of magnitude larger than the diencephalon. When aromatase activity is therefore recalculated for the entire brain area, the activity difference between the telencephalon and diencephalon becomes more prominent in all songbird species and a 5 to 20 fold excess is found in the former by comparison with the latter area (see Fig. 5B). With this expression of activity, the aromatase activity is however still lower in the telencephalon than in the diencephalon of quail but not of doves. It must be noted that the Ring Doves used here for comparison were killed outside breeding season when diencephalic aromatase was presumably minimal. Assays performed on reproductively active birds would probably yield a pattern of activity that would favor the diencephalic aromatase in this species also.

In conclusion, these data support the notion that songbirds are characterized by a high level of telencephalic aromatase activity by comparison with other avian groups. The difference is however not as dramatic as previously thought and some overlap can be observed especially if enzyme activities are normalized for the weight of brain areas and expressed per mg of protein. It should also be pointed out that aromatase activity is present in the telencephalon of all species studied so far including non songbirds. The evolutionary significance of the high or low ratio between telencephalic and diencephalic aromatase activity remains unclear at present and will probably be impossible to specify until the enzyme has been studied in a larger number of species originating from a broad range of families and displaying a variety of ecological life styles.

As mentioned above, most studies on brain aromatase in birds have so far been concerned with domestic or semi-domestic species such as the ring dove, quail, chicken or Zebra Finch and little information is available on the anatomical distribution and on possible changes in brain aromatase activity during the normal reproductive cycle in free-living birds. In these semi-domestic species, the regulation and neuroanatomical distribution of aromatase has to this date been studied by product formation assays, immunocytochemistry and molecular biology (assays of mRNA concentration or their visualisation by in situ hybridisation). As reviewed above, these studies have revealed that the aromatase activity is high in the preoptic area and hypothalamus (Callard et al. 1978; Steimer & Hutchison 1981a; Schumacher & Balthazart 1986) that are implicated in the activation of male reproductive behaviour (Panzica et al. 1996). Moreover, the preoptic aromatase activity is strongly increased by T in several bird species including the quail, dove and Zebra Finch (Steimer & Hutchison 1981a; Schumacher & Balthazart 1986; Vockel et al. 1990b) which supports the notion that this enzyme plays a critical role in the control of male sexual behaviour (see: Balthazart 1989; Balthazart et al. 1996b for review). Additional studies indicate that, in the few species studied so far the T-induced changes in preoptic aromatase activity result from changes in the concentration of the enzyme caused by an increased transcription of the corresponding mRNA (see: Balthazart & Foidart 1993; Foidart et al. 1995b for additional discussion). High levels of aromatase activity are also present in the telencephalon of songbirds (Vockel et al. 1990a; Vockel et al. 1990b; Schlinger & Arnold 1991) but the control of aromatase by T that is present in the preoptic area and anterior hypothalamus is surprisingly not observed in the telencephalon of the only songbirds species that has been studied in this respect, the Zebra Finch (Vockel et al. 1990b).

Seasonally breeding species that live in the temperate zone show marked seasonal changes in testis weight and in testosterone (T) plasma levels. These variations are paralleled by pronounced behavioural modifications. In many species, males defend a territory and vigorously attack any intruder, they court a female, and eventually help feeding the young. The high level of aggressiveness that characterizes the establishment of the territory is generally lost during the period of incubation and feeding of nestlings (Silverin & Deviche 1991; Silverin 1993). The endocrine mechanisms that control these behaviour changes have been investigated in a variety of species and it is now widely accepted that both aggressive and reproductive male behaviours are activated by T in many vertebrate species (Silver et al. 1979; Balthazart 1983; Wingfield 1985).

The central conversion of T into estradiol (E2), or aromatization, appears to be a critical step in the activation of male copulatory behaviour and also possibly of parts of the aggressive repertoire (Adkins & Pniewski 1978; Balthazart 1989; Balthazart & Foidart 1993; Schlinger & Callard 1990) but to this date, little or no data are available concerning the neuroanatomical distribution and functional regulation of the aromatase enzyme in the brain of free-living species. We therefore decided to initiate studies on this topic in male Pied Flycatchers Ficedula hypoleuca. The Pied Flycatcher is a migratory species common in northern Europe during the spring breeding season. It is a poly-territorial, hole-nesting bird that breeds preferentially in deciduous forests. The Pied Flycatcher is a polygamous species in which the male leaves his home territory usually on the day his female lays her first egg and establishes a secondary territory some 100-400 m away where he tries to attract a second female. He will return to the home territory during the later part of the incubation period of the first female and participate in the feeding of the nestlings. The second female will incubate and feed her nestlings alone and these young will therefore have a lower survival success (Silverin 1983b; Silverin 1983a).

Male Pied Flycatchers arrive in the breeding area and occupy territories before the females arrive. When a female settles in a territory, she starts nest-building more or less immediately. Within just a day or two, the male is found to have drastically increased luteinizing hormone (LH) and T titers that remain elevated during the early part of the nest-building period (Silverin & Wingfield 1982). The Pied Flycatcher is also known for its early and extremely rapid gonadal regression starting during the later part of the incubation period. All males have basal T levels at the time of egg hatching and, just a few days later, the gonads are completely regressed (Silverin 1975; Silverin 1983b; Silverin & Wingfield 1982). Immunocytochemical and biochemical studies were initiated in this species to describe the distribution of aromatase-containing cells in the brain and research whether changes in enzyme activity are associated with the natural changes in plasma T that physiologically occur during the reproductive cycle (see Foidart et al. 1998 for a full description of these results).

One set of birds (seven males defending a secondary territory and six males feeding nestlings) were sacrificed by decapitation and brains were immediately dissected out of the skull and frozen on powdered dry ice. Aromatase activity was then measured by the radioenzyme assay quantifying the release of tritiated water described above in 6 brain regions: the telencephalon (Tel.), the anterior diencephalon (ant.Di., corresponding to the preoptic area and rostral hypothalamus including the POM), the posterior diencephalon (post.Di., including the ventromedial nucleus of the hypothalamus [VMN] and the tuberal region), the optic lobes (O.L.), the cerebellum (Cereb.), and the brain stem (B.S.). Results of these assays expressed as fmol of water (or oestrogen) produced per hour and per mg protein are summarised in Fig. 6.

As already reported above, the highest level of aromatase activity was observed in the telencephalon (Fig. 6A) where enzyme activity clearly exceeded the activity detected in any other brain area including the anterior or posterior parts of the diencephalon. Low to non detectable levels of enzyme activity were observed in the other brain regions. The aromatase activity per mg protein was approximately twice higher in the telencephalon than in the diencephalon, a difference that, given the large size of the telencephalon, translates into a 20 fold excess when total activities are computed for the entire area (see also Fig. 5).

Analysis of these data by a mixed two-way ANOVA with the two periods of the reproductive cycle as an independent variable and the 6 brain regions as a repeated measure clearly indicated the heterogeneous distribution of aromatase activity between the 6 brain regions (P < 0.001) as well as the overall significant difference between the two experimental groups (P = 0.002) . Interestingly, a strong interaction between these two variables was also detected (P = 0.002) and additional post hoc t-tests using the relevant mean square of the ANOVA as the basis for comparison showed that aromatase activity is significantly higher in the secondary territory males than in males feeding nestlings in the anterior (P < 0.05) and posterior (P < 0.01) diencephalon but this difference is not present in the other brain regions including the telencephalon (P > 0.05 in each case).

These data showed that territorial males who have higher plasma T levels also have a higher aromatase activity in the diencephalon as compared with males feeding nestlings. Because previous studies in other avian species have shown that T increase aromatase activity (e.g. Steimer & Hutchison 1981a; Schumacher & Balthazart 1986), we further analysed the relationship between the individual variations in these two variables in the different brain regions for the two groups of birds considered together or separately. In the pooled data from all subjects studied at the two points in the reproductive cycle a linear regression analysis demonstrated the existence of a significant correlation between plasma T level and aromatase activity per mg protein in the anterior and posterior diencephalon (ant.Di.: r = 0.887, P > 0.001; post.Di.: r = 0.910, P > 0.001) but not in the telencephalon (P > 0.05; see Fig. 6B). These significant correlations obviously originated in part in the seasonal variation and in part in the individual differences at one given stage of the reproductive cycle. To separate these two sources of variance, correlations between the same variables were computed again for the territorial birds and the birds feeding nestlings separately. All these correlation coefficients were then found to be non significant but in birds defending a territory, correlations fell short of significance (statistical tendencies) for the relationship between plasma T levels and aromatase activity in the anterior and posterior diencephalon (r = 0.754, P = 0.083 and r = 0.773, P = 0.071 respectively) while no relationship was seen between plasma T levels and aromatase activity in the telencephalon (P > 0.10). Similarly no relationship could be detected between plasma T levels and aromatase activity in all brain areas of the birds feeding nestlings (P > 0.10 in each case).

The overall pattern of distribution of ARO-ir cells was similar in males defending secondary territories and in males feeding nestlings and strongly resembled the distribution previously reported for ARO-ir cells and cells expressing aromatase mRNA in the Zebra Finch brain (Balthazart et al. 1990c; Balthazart et al. 1996a; Shen et al. 1995). In the anterior preoptic area, a medial cluster of immunoreactive cells adjacent to the third ventricle was observed in a ventral position under the tractus septomesencephalicus and extended through the rostro-caudal extent of the preoptic area to end up in a position just ventral to the anterior commissure (CA). A similar group of ARO-ir cells has been described in Zebra Finches (Balthazart et al. 1996a) and identified as the homologueue of the medial preoptic nucleus (POM) of the quail (Panzica et al. 1991; Panzica et al. 1996). The same terminology will be used here (Fig. 7A).

At the same level, a small group of ARO-ir neurons is observed between the ventral edge of the lateral ventricles and the dorsal edge of the CA (Fig. 4B). This cluster has been identified as the rostral part of the nucleus striae terminalis (nST) in our work on Zebra Finches (Balthazart et al. 1996a) but is located in a slightly more ventral position than the nST as indicated in the stereotaxic atlas of the canary brain (Stokes et al. 1974). A small number of weakly labelled cells was in addition observed at the lateral tip of the lateral ventricles (Fig. 7A) in a position identified as the nST in the canary brain atlas (Stokes et al. 1974).

In sections immediately caudal to the CA, the groups of ARO-ir neurons located in the nST and in the caudal POM appear to merge and form a V-shaped group of immunopositive cells identified as the caudal part of the nST in the Zebra Finch and quail (Balthazart et al. 1996a; Foidart et al. 1995a). At these rostro-caudal levels, another more ventral group of ARO-ir cells is observed at the level of the nucleus ventromedialis hypothalami (Fig. 7A; VMN; = posteromedial hypothalamus, PMH in the canary brain atlas). This group of immunoreactive cells is present without discontinuity throughout the mediobasal hypothalamus but the number of positive cells progressively decreases in the most caudal part of the hypothalamus. Finally, a small number of weakly labelled cells is also found ventral to the VMN in the nucleus tuberis (Tu).

In addition to the ARO-ir cells observed in the diencephalic and limbic areas, large numbers of immunoreactive cells were seen in the flycatcher telencephalon with a pattern of distribution very reminiscent of the distribution reported for in the Zebra Finch telencephalon (Balthazart et al. 1996a; Shen et al. 1995). These telencephalic ARO-ir cells appeared in general to be less numerous in flycatchers than in Zebra Finches but it is impossible at this stage to know whether the difference in abundance reflects a true species difference or is related to technical differences (different cross-reactivity with antibodies, different levels of optimization for the immunocytochemical technique). In the flycatcher, telencephalic ARO-ir cells were mainly found in the neostriatum (Fig. 7B). At the most rostral levels, the positive cells were exclusively located in the medial neostriatum and in the ventral hyperstriatum. In more caudal sections, this cell cluster progressively extended ventrally along the lateral ventricle until the point where this ventricle crosses the lamina medullaris dorsalis. In even more more caudal levels, the immunopositive cells became more numerous and were spread over a broader area that covered at its maximal extent a large area dorsal to the lamina archistriatalis dorsalis and a large part of the caudal neostriatum. A small number of immunopositive neurons were also observed in dorsomedial position throughout the hippocampus and the area parahippocampalis. No ARO-ir cells could be detected in the nucleus taeniae (Tn), nor in the telencephalic nuclei of the song control system: the high vocal centre (HVC) and nucleus robustus archistriatalis (RA).

Surprisingly however, the analysis by the same technique of the numbers of ARO-ir cells counted at five successive rostro-caudal levels in the VMN identified an effect of the position in the rostro-caudal axis but also an overall difference between the experimental groups and an interaction between these two variables. The number of ARO-ir cells was unexpectedly higher in the posterior diencephalon of the males feeding nestlings than in territorial birds, whereas aromatase activity was higher in territorial birds (see: Foidart et al. 1998 for the detail of these results).

Previous studies in Japanese Quail showed that changes in aromatase activity as a function of the sex of the subjects, gonadal state (castrated vs gonadally intact) or endocrine treatment (replacement therapy with androgens and/or oestrogens) are almost always paralleled by differences in the number of ARO-ir cells, suggesting that the changes in enzyme activity are caused largely if not entirely by a change in enzyme concentration (Balthazart & Foidart 1993; Panzica et al. 1996). Why such correlations between enzymatic activity and numbers of ARO-ir cells were not clearly observed in flycatchers remains partly unknown. Territorial males appeared to have more ARO-ir cells in the POM than birds feeding nestlings, especially in the caudal part of the nucleus. This result fits in well with previous data in quail showing that a treatment with exogenous T increases the number of ARO-ir cells specifically in the caudal part of the POM (Foidart et al. 1994a; Balthazart et al. 1996d) and with the enzyme assay data in flycatchers showing a decrease in aromatase activity in birds feeding nestlings by comparison with territorial birds. The seasonal changes in number of ARO-ir cells in POM did not reach statistical significance presumably because a relatively small sample size was available (6 -7 subjects further decreased by technical problems so that 4 pairs of birds only could be used in the general ANOVA comparing all rostro-caudal levels in all subjects).

A host of studies have shown in Zebra Finches, quail and doves that T aromatisation is a critical step in the induction of most behavioural effects of the androgen and in particular, it has been shown that , in two species of song birds (Zebra Finch and red-winged blackbird), song is activated by a synergistic action of oestrogenic and androgenic metabolites of T. The role of the central T metabolism in the control of vocal behaviour remains, however, partly unclear. If it has been possible to demonstrate clear effects on song production of systemic treatments that affect the availability of T metabolites (Harding et al. 1983; Walters & Harding 1988), the specific role played by the enzymes in each brain area is completely unknown. In particular, the specific implication of the high telencephalic aromatase activity identified in songbirds in the control of vocalisations remains unknown given that ARO-ir cells do not appear to overlap with song control nuclei (see: Balthazart et al. 1996a for further discussion on this topic).

The anatomical mismatch between aromatase, oestrogen receptor and song control nuclei in the Zebra Finch central nervous system questions the way in which steroids affect behaviour. An autocrine mode of action has always been assumed. Oestrogens derived from intracellular aromatisation would act in the cell where they are produced and these should therefore contain both oestrogen receptors and aromatase. The recent anatomical (immunocytochemistry and in situ hybridisation) and physiological evidence suggest the possibility of paracrine or even endocrine effects for the oestrogens produced in the brain.

The data presented in this paper indicate that, as could be suspected based on the limited amount of available data (Saldanha & Schlinger 1997; Schlinger et al. 1992; Schlinger 1996), the existence of high levels of aromatase activity in the telencephalon can be considered as a typical feature of songbirds, even if the difference with other avian orders is of a smaller magnitude than could have been expected based only on the study of the zebra finch. The anatomical distribution of this telencephalic aromatase or of its messenger RNA has now been mapped with some detail in three songbird species, the zebra finch, brown-headed cowbird and Pied Flycatcher. A great degree of similarity was observed in the hypothalamic and limbic distribution of the enzyme in these species and all studies agree to localize the majority of telencephalic aromatase-containing cells in the neostriatum, ventral hyperstriatum, hippocampus and area hippocampalis. In all studies also, little or no aromatase was detected in the song control nuclei HVC and RA.

In Pied Flycatchers like in Zebra Finches (Vockel et al. 1990a; Schlinger & Arnold 1991; Balthazart et al. 1996a; Shen et al. 1995), a high level of aromatase activity was observed in the telencephalon and this enzymatic activity corresponded to the presence of broad populations of ARO-ir cells in the neostriatum, hyperstriatum, hippocampus and area parahippocampalis. These ARO-ir cells had remained undetected in our original study of the Zebra Finch brain (Balthazart et al. 1990c) but they have been more recently identified by immunocytochemistry with the help of a new more sensitive antibody or by pretreating the subjects with a non aromatizable inhibitor (Balthazart et al. 1996a). Their distribution has been confirmed by in situ hybridisation (Shen et al. 1995). It was thought originally that the presence of aromatase in the telencephalon was a characteristic of songbirds but this conclusion was only based on studies in a few species (see above). The present study brings further support to this idea with a broader set of data.

It must also be noted that a significant level of aromatase activity was recently detected in the telencephalon of the Japanese Quail, in parallel with the existence of large numbers of weakly immunoreactive aromatase cells (Balthazart et al. 1998). Therefore, the presence of aromatase in the telencephalon is probably not a unique characteristic of songbirds as opposed to other avian species but it should rather be considered that songbirds differ from other avian species by the higher level of telencephalic expression of the enzyme. When expressed per mg of protein, aromatase activity appears to be twice higher in the telencephalon than in the hypothalamus in Zebra Finches or Pied Flycatchers but this ratio is more than 10 times lower in quail (Balthazart et al. 1998). Because the telencephalon weight is about one order of magnitude larger than the diencephalon weight, the total amount of enzyme activity in the telencephalon remains important by comparison with the hypothalamic aromatase activity even in species where activity per mg protein is relatively modest (e.g. quail). Based on the anatomical studies described above, it appears that the aromatase activity in the telencephalon of songbirds may not be directly related to the control of song expression because ARO-ir cells and cells expressing the ARO mRNA do not appear to be present in the song control nuclei HVC and RA (Shen et al. 1995; Balthazart et al. 1996a). Indirect functional relationships cannot be ruled out at present but the possibility that telencephalic aromatase could be related to completely different biological functions is certainly supported by the presence of a significant level of aromatase activity in avian species that do not possess a song control system such as quail and doves.

During the last few years, the activity of T-metabolising enzymes has been measured in the brain of a few avian species including one songbird, the Zebra Finch. These studies have revealed the presence of large enzymatic differences related to the sex, age or endocrine status of the birds. In some cases, the aromatase activity appears to be controlled by the adult endocrine environment but in others cases it is not, suggesting the existence of differential regulation mechanisms in different species and neuroanatomical loci. The nature of these regulatory mechanisms is currently being uncovered largely with the help of molecular biology techniques.

Studies carried out almost exclusively in laboratory species had shown that experimentally-induced changes in plasma T are rapidly reflected (within a few hours or days) in changes preoptic-hypothalamic aromatase activity. We have now demonstrated that the testicular collapse and dramatic fall in plasma T that take place during the natural breeding cycle of Pied Flycatchers when birds cease to defend a territory and start feeding young nestlings are also paralleled by a similar drop in the preoptic and hypothalamic aromatase activity. In contrast, the telencephalic aromatase appears to be insensitive to this control by T in flycatcher, just like it was not affected by castration in Zebra Finches (Vockel et al. 1990b).

These data strongly argue for the presence of different regulatory mechanisms for aromatase activity and presumably aromatase synthesis in these two brain areas. In all species that have been studied so far in some detail, a good parallel has been observed between T-induced changes in the aromatase activity, in the number of ARO-ir cells and in the aromatase mRNA concentration in the preoptic area and hypothalamus (Steimer & Hutchison 1981a; Schumacher & Balthazart 1986; Vockel et al. 1990b; Balthazart & Foidart 1993; Panzica et al. 1996; Balthazart et al. 1996b). These data strongly suggest that T enhances the transcription of aromatase in the diencephalon, a fact that is consistent with the recent discovery of an androgen responsive element at the 5’-end of the first untranslated brain-typical exon of the aromatase gene in rat (Kato et al. 1997). Why this control mechanism does not appear to be present in the telencephalon has remained mysterious so far but the analysis of the aromatase gene by molecular biology techniques is beginning to provide a detailed answer to this question (Simpson et al. 1991; Bulun & Simpson 1994; Simpson et al. 1994; Lephart 1996; Lephart 1997). The control of aromatase activity by different hormones or growth factors in different tissues appears to be related to the fact that aromatase transcription is regulated by different promoters located in or at the 5’-end of different tissue-specific untranslated exons 1 (Simpson et al. 1997). A specific exon 1 (called exon 1f) has been isolated and sequenced in the rat brain and presumably explains the peculiar regulation of the enzyme synthesis in the brain compared to peripheral tissues in this species (Yamada-Mouri et al. 1997; Kato et al. 1997; Lephart 1996). Interestingly, a recent report in abstract form indicates that different types of exon 1 called exons 1a and 1b are also present in the Zebra Finch brain (Ramachandran et al. 1997) and these could be differentially expressed in the diencephalon and telencephalon. These anatomically specific first exons associated with different regulatory elements could explain why aromatase transcription is regulated in a different fashion within specific brain areas. Detailed studies of the telencephalic and diencephalic exons 1 and of their 5’-end should help explaining why aromatase activity is controlled by T (Zebra Finches) or changes with season (Pied Flycatcher) in one brain area but not in the other. Although we did not formally test here the idea that the seasonal change in diencephalic aromatase activity is related to the changes in plasma T levels in Pied Flycatchers, the fact that levels of aromatase activity are closely correlated with plasma T in the two diencephalic regions but not in the telencephalon strongly supports this idea.

Many studies in birds and other vertebrates demonstrate that T is a key endocrine factor responsible for the activation of sexual and aggressive behaviour. The endocrine changes seen in the Pied Flycatcher when males begin to feed nestlings are therefore presumably responsible for the dramatic behavioural metamorphosis seen at that time: males stop defending a territory, they no longer show aggressive reactions towards territory intruders, they no longer try to attract additional females and they start helping the primary female in feeding her young nestlings (Silverin 1993). In flycatchers, the diencephalic aromatase is located in nuclei that are obviously related to the control of various aspects of reproduction (POM, VMN, nST) and the observation that this enzymatic activity decreases when birds switch from territorial defence to caring for nestlings certainly reflects the underlying control of these behavioural activities. It is known for example that oestrogens contribute to the activation of singing and possibly of aspects of aggressive behaviours in songbirds (Harding et al. 1988; Harding et al. 1983; Walters & Harding 1988; Walters et al. 1991). The diencephalic drop in aromatase activity may therefore be implicated in the seasonal control of these behavioural changes. More studies should however be carried out in a variety of avian species to investigate whether control mechanisms for steroid-metabolising enzymes that have originally been discovered in (semi-) domestic species also apply to free-living birds. Additional regulatory mechanisms may be uncovered as suggested here by the discrepancy between seasonal changes in aromatase activity and in the number of aromatase-immunoreactive cells in the posterior hypothalamus of flycatchers.

ACKNOWLEDGMENTS