Immunolocalization of polysialic acid in the median eminence and neurointermediate hypophysial lobe of adult rats

Iota'halof

Chemicalm Neuroanatorr~ ELSEVIER

Journal of Chemical Neuroanatomy 8 (1994) 33-45

Immunolocalization of polysialic acid in the median eminence and neurointermediate hypophysial lobe of adult rats G. Alonso* INSERM U 336, D~veloppement, Plasticit~ et Vieillissement du Systbrne Nerveux, Universit~ de Montpellier H, Place E. Bataillon, 34095 Montpellier cedex 5, France Accepted 7 September 1994

Abstract

Polysialic acid (PSA) is abundant on growing axons during brain development and down regulated on maturation. However, high amounts of this carbohydrate polymer have been found to persist in some regions of the adult rat brain including the mediobasai hypothalamus. In this study, confocal laser scanning microscopy combined with double fluorescence immunostaining was used to characterize the cellular localization of PSA throughout the median eminence and neurointermediate hypophysial lobe of adult rats. In these regions, polysialic acid-immunoreactivity (PSA-IR) generally appeared associated with fiber-like structures. Double immunostaining experiments demonstrated that, in addition to large axons of the neural lobe immunoreactive to vasopressin or oxytocin, PSA was constantly associated with fibers projecting into the intermediate hypophysial lobe immunoreactive to either 7-aminobutyric acid (GABA) or tyrosine hydroxylase. Similarly, PSA-IR was detected on most, but not all the fibers immunoreactive to GABA or tyrosine hydroxylase dispersed throughout the neural lobe and the different layers of the median eminence. On the other hand, no PSA-IR was detected on axons immunoreactive to somatostatin or to corticotropin releasing hormone projecting throughout the median eminence, or on glial cell bodies and processes immunoreactive for glial fibrillary acidic protein (GFAP) or for vimentin dispersed throughout the median eminence and the neural lobe.

Keywords: Hypothalamus; Adult rat brain; 7-Aminobutyric acid (GABA); Dopamine; Neural cell adhesion molecules (NCAM); Confocal microscopy

1. Introduction

During development of the nervous system, the structuring of neuronal groups and the establishment of neuronal connections are dependent on a series of complex mechanisms that induce and regulate axonal outgrowth. During recent years, emphasis has been placed on several adhesion molecules involved in such mechanisms. Among them, a group of high molecular weight glycoproteins belonging to the immunoglobulin superfamily and collectively known as neural cell adhesion molecules (NCAM), have been shown to promote cell-cell adhesion through homophilic interactions (for review see Rutishauser et al., 1988; Rutishauser and

*Tel.: +33 67 14 33 90; Fax: +33 67 14 33 18. 0891-0618/94/$07.00 © 1994 Elsevier Science B.V. All rights reserved SSDI 0891-0618(94)00027-Q

Jessel, 1988; Edelman and Crossin, 1991). N C A M exists in several structurally distinct isoforms that contain various amounts of c~-2,8-1inked polysialic acid (PSA) on the extracellular domain of the molecule (Finne et al., 1983). The lengths of the chains of PSA residues is developmentally regulated: during the late embryonic and early postnatal period, highly polysialylated N C A M (PSA-NCAM) is abundant throughout the nervous system, after which its expression rapidly decreases with the establishment of neuronal connections (Rothbard et al., 1982; Rougon et al., 1982; Chuong and Edelman, 1984). However, large amounts of PSA have been found to persist in some regions of the adult brain such as the olfactory bulb, the mediobasal hypothalamus or the hippocampus, that are known to undergo morphological remodelling in response to physiological or experimental stimulations (Miragall et al., 1988;

34

G. Alonso /.L Chem. Neuroanat. 8 (1994) 33-45

Theodosis et al., 1991; Bonfanti et al., 1992; Le Gal La Salle et al., 1992). In the hippocampus, moreover, high levels of PSA have been shown to be expressed transiently by newly generated granule cells of the dentate gyrus (Seki and Arai, 1993). As clearly shown in a series of biochemical and immunocytochemical studies, PSA detected in the brain of adult mammals is essentially associated with NCAM (Aaron and Chesselet, 1989; Theodosis et al., 1991; Bonfanti et al., 1992; Le Gal La Salle et al., 1992). Since cell-cell adhesion promoted by NCAM or by other adhesion molecules is attenuated by PSA-NCAM (Hoffman and Edelman, 1983; Sadoul et al., 1983; Rutishauser and Landemesser, 1991; Landmesser, 1992), the hypothesis has been proposed that the persistence of large amounts of PSA within these specific regions of the adult brain is specifically associated with mechanisms of morphological plasticity (Miragall et al., 1988; Aaron and Chesselet, 1989; Theodosis et al., 1991; Miller at al., 1993). Within the mediobasal hypothalamus PSA-NCAM has recently been shown to be associated with the hypothalamic neurons projecting to the neural lobe (Theodosis et al., 1991; Kiss et al., 1992) that are well known to undergo considerable morphological reorganization in response to physiological or experimental stimuli (Tweedle and Hatton, 1982, 1987). It is possible, however, that the high levels of PSA detected within the mediobasal hypothalamus and the neurohypophysis of adult mammals are also associated with other neurons projecting to this region that maintain a remarkable capacity for morphological plasticity throughout adulthood (Perera et al., 1993). In order to better establish a functional role of PSA in the mediobasal hypothalamus of the adult rat, the present study was undertaken to characterize more thoroughly the cellular localization of this molecule throughout the median eminence and the neurointermediate lobe of the hypophysis. Confocal laser scanning microscopy combined with double immunofluorescence immunostaining was used to determine the cellular association of PSAimmunoreactivity with specific neuronal or glial constituents of the neurohypophysis of the adult rat. 2. Material and methods

2.1. Animals Male adult Sprague-Dawley rats were used. They were kept in light (12 h light/12 h darkness) and temperature (24° ± 1°C) controlled rooms and had free access to standard dry food and tap water. 2.2. Preparation of tissues Animals (n = 10) were anaesthetized and perfused through the ascending aorta with phosphate buffered saline (PBS), followed by 500 ml of fixative composed of 4% paraformaldehyde and 0.5% glutaraldehyde in 0.1 M phosphate buffer at pH 7.4. The forebrain was dissected

and fixed by immersion in the same fixative without glutaraldehyde for 2-4 days. The ventromedial hypothalamus and neurointermediate lobe of the hypophysis were then cut frontally or sagittaly with a vibratome into 30-40 #m thick sections. These were carefully rinsed with PBS and subsequently treated for double fluorescence immunostaining.

2.3. lmmunocytochemistry Vibratome sections were treated for double immunofluorescence labeling by incubating them with two primary antibodies including an lgM and an IgG mouse monoclonal antibody ascites, or a mouse monoclonal antibody (IgM or IgG) and a rabbit polyclonal antibody. The antibodies used included: (i) a mouse IgM monoclonal antibody against c~-2,8-linked PSA (kindly provided by Dr. Rougon, Lumigny, France; dilution 1:2000), (ii) mouse IgG monoclonal antibodies against GABA (Chemicon, dilution 1: t000) or vimentin (VIM) (Sigma, dilution 1:1000), and (iii) rabbit polyclonal antibodies against corticotropin releasing hormone (kindly provided by Pr. C. Oliver, Marseille, France, dilution 1:1000), tyrosine hydroxylase (Jacques Boy Laboratories, Reims, France, dilution 1:5000), glial fibrillary acidic protein (GFAP) (Dako laboratories, dilution 1:1000), anti-vasopressin, anti-oxytocin, and anti-somatostatin (raised in our laboratory, dilution 1:1000). During the immunocytochemical procedures, antibodies were diluted in phosphate buffered saline (PBS) at pH 7.4 containing 0.1% saponin and 1% normal goat serum, and incubated for 48 h at 4°C. For each neurointermediate lobe treated, some sections were additionally incubated with the PSA antibody and a mixture including vasopressin and oxytocin antibodies. After rinsing in PBS, they were incubated for 2 h at 4°(2 with two secondary antibodies corresponding to the primary antibodies used (i.e. antibodies to mouse IgM or IgG, or to rabbit IgG), respectively, conjugated with rhodamine and fluorescein (Sigma). The secondary antibodies were diluted 1:200 in PBS containing 1% of normal goat serum and 0.1% saponin. After careful rinsing, sections were mounted in Mowiol (Calbiochem, La Jolla, USA). The specificities of the antibodies used have been described and documented elsewhere (see Alonso, 1988 for anti-somatostatin, -oxytocin and -vasopressin; Alonso et al., 1988 for anti-corticotropin releasing hormone; Rougon et al., 1986 for anti-PSA; Arluison et al. 1984 for anti-tyrosine hydroxytase; Viale et al., 1991 for antiGFAP; Szabat et al., 1992 for anti-GABA; Muijen et al., 1984 for anti-vimentin). The method controls consisted of omitting the primary antibodies or applying each primary antibody sequentially and then reacting them with the inappropriate secondary antibody. 2.4. Confocal microscopy Mounted sections were observed under a MRC-600

G. Alonso/J. Chem. Neuroanat. 8 (1994) 33-45 confocal laser scanning microscope (Biorad) equipped with a krypton/argon mixed gas laser. Two laser beams emitting wavelenghts of 488 nm and 568 nm were used, respectively, for exciting the fluorescein and the rhodamine-conjugated secondary antibodies providing a minimum overlap of the emission spectra of the two fluorochromes. The organization of the two immunofluorescent labelings was studied on optical sections of thickness varying from 5 to less than 1 /~m, depending on both the numerical aperture of the objective lens used (from 0.5 to 1.4) and the size of the confocal aperture (see Wall6n, 1991). The background noise of each confocal image was reduced by averaging 6 image inputs.

3. Results The organization of the different types of axonal fibers and of the different types of glial cells present in diverse portions of the neurohypophysis conformed to previous observations (represented in Fig. 1). 3.1. PSA-imrnunostaining As previously described (Theodosis et al., 1991; Bonfanti et al., 1992) intense PSA-immunostaining was detected throughout the mediobasal hypothalamus including the periventricular mediobasal area, the median eminence, and the neuro-intermediate lobe of the

35

hypophysis (Figs. 2-3). In contrast to various structures stained by other antibodies throughout the neurohypophysis, the outlines of PSA-IR structures were generally diffuse (Fig. 2A). When observed at high direct magnification, however, i.e. on confocal images corresponding to thinner optical sections, such immunostaining frequently appeared associated with homogeneously stained fiber-like structures (Fig. 2B,C). In the neural lobe, PSA-IR was associated either with thin, elongated fiber-like structures that appeared homogeneously stained, or with large axonal profiles with a preferential localization of the immunostaining to the periphery of these axonal dilatations (Figs. 3D, 5A,C). Additionally, large, homogeneously stained PSA-IR fibers were frequently observed to project from the periphery of the neural lobe towards the intermediate lobe of the hypophysis (Fig. 5B,D). 3.2. Double-irnmunostaining for PSA and for neuropeptides Experiments of double-immunostaining failed to demonstrate any clear association between PSA-IR and the numerous peptidergic axons innervating the external layer of the median eminence. All along the rostrocaudal extension of the median eminence, PSA-IR fibers were localized to regions that contained numerous fibers immunoreactive to somatostatin or corticotropin releas-

1

int V

..4 ,/ext

S~

*

6../. f N L

'IL. Fig. 1. Schematic representation of a sagittal section through the neurohypophysisshowing the anatomical organization of the different types of neurohypophysial structures considered in the present study. 1: Vasopressinergicor oxytocinergicaxons originating in the magnocellularneurons of the supraoptic and paraventricular nuclei: they form a dense pathway running all along the internal layer of the median eminence and terminate around the blood vesselsof the neural lobe. 2: Somatostatin-, or corticotropin releasing hormone-lR axons, respectively,originating in parvicellular neurons of the periventricular and paraventricular nuclei: they terminate in the external layer of the median eminence, all along its rostro-caudal portions. 3: GABAergicand/or dopaminergicaxons originating in parvicellularneurons of the arcuate nucleus: they either terminate in the external layer of the median eminence and in the neural lobe, or project within the intermediate hypophysial lobe. 4: VIM-IR tanycytes, the cell bodies of which are located all along the ventricularborder of the median eminence,and whose basal processesradiate into the underlying neuropil towards perivascularexternal layer of the organ. 5: GFAP-IR classical astrocytes whose cell bodies and short radiating processes are essentially located in the internal median eminence layer. 6: GFAP-IR astrocyte-likepituicytes whose cell bodies and processesare dispersed throughout the neural lobe. by: blood vessels; ext and int: external and internal layers, respectively, of the median eminence; IL: intermediate lobe of the hypophysis; ME: median eminence; NL: neural lobe of the hypophysis; V: third ventricle.

the

36

G. Alonso/,L Chem. Neuroanat. 8 (1994) 33-45

Fig. 2. Confocal images of sections through the median eminence immunostained for PSA. A: Confocal image obtained at low magnification (objective 20 × with a numerical aperture of 0.5), showing the distribution of intense PSA-immunostaining throughout the different layers of the median eminence. B and C: The observation of thinner optical sections corresponding to confocal images obtained by using objective lens with increasing numerical aperture (B: objective 40 × with a numerical aperture of 1.0; C: objective 60 × with a numerical aperture of 1.4) shows that, in the external layer of the median eminence, immunostaining is mostly associated with fiber-like structures. V: third ventricle; bv: blood vessel. A: × 200; B: × 400; C: × 600. Bar = 50 #m.

Fig. 3. Confocal paired irmq~ of sections through the median eminence and the neural lobe immmmstained for PSA (A-D) and for conicotropin hormone (A' and B'), somatosUttin (C'), or vasopressin + oxytocin (D'). A-C: External layer of the n~dian emigrate. A confocal image obtained at low ~ a t i o n (A, objective 20 × ) shows that although the structures immunostained for PSA or corticotropin releasing hormone appear closely intermingled, a number of corticotropin releasing hormone-lR axons located in the most external layer appears to be deprived of PSA-IR (arrows A'). Confocal images obtained at higher mngl~c,ation (B and C, objective 60 × ) clearly show that the PSA-IR fiber-like structures detected in the perivasc,ular layer are distinct from fibers immunoreactive to corticotropin releasing hormone (B') or to somatostatin (C') present m the same resion. D: Neural lobe. PSA-IR is assogiated both with thin fibers that appear vasopressin- and oxytocin-nesative (arrows D), and with most of the large axonal profiles irmmmoreactive to vasopressm or oxytocin (double arrows D - D ' ) . Note that, in contrast to the elongated PSA-IR fibers detected in the external median eminence (see Figs. 2C, 3B,C), PSA-IR frequently appears localized at the periphery of the vasopressin- or oxytocin-IR axons, by: blood vessel; A × 300: B, C and D: × 600. Bar = 50 ~tm.

I

s~

38

G. Alonso /J. Chem. Neuroanat. 8 (1994) 33-45

Fig. 4. Confocai paired images of sections through the median eminence immunostained for PSA (A-D) and for tyrosine hydroxylase (A ', B') or GAB.A (C ', D'). A-D: External layer of the median eminence. Low magnification (objective 20 x ) shows that the distribution pattern of PSAIR structures appears very similar to that of structures immunoreactive to tyrosine hydroxylase (A') or to GABA (C'). Higher magnification (objective 60 x ) shows that PSA-IR is associated with the majority of fibers exhibiting tyrosine hydroxylase-IR (B') or GABA-IR (D'). Note that whereas the large majority of PSA-IR fibers also exhibit GABA-IR ( D - D ' ) , a number of PSA-IR fibers appear tyrosine hydroxylase-negative (arrows B). A and C: x 300; B and D: × 500. Bar = 50 ~,m.

G. Alonso /J. Chem. Neuroanat. 8 (1994) 33-45

39

Fig. 5. Confocal paired images of sections through the neurointermediate lobe immunostained for PSA ( A - D ) and for tyrosine hydroxylase (A ', B ') or GABA (C ', D'). A and C: Neural lobe (objective 60 x ). PSA-IR is associated both with some fibers immunoreactive to tyrosine hydroxylase (arrows A - A ' ) or to GABA (arrows C - C ' ) , and with large axonal profiles that appear tyrosine hydroxylase- or GABA-negative (arrow heads A and C). B and D: Intermediate lobe (objective 60 x ). Intense PSA-IR is associated with large fibers which are systematically immunoreactive to tyrosine hydroxylase ( B - B ' ) or to GABA ( D - D ' ) . A - D : x 600. Bar = 50 tzm.

~h

G. Alonso/.L Chem. Neuroanat. 8 (1994) 33-45

41

ing hormone, and some fibers immunoreactive to vasopressin. However, although both types of immunostained structures frequently appeared tightly intermingled, peptidergic fibers were generally found to be deprived of PSA-IR (Fig. 3 A - A ' , B - B ' , C - C ' ) . In the neural and intermediate hypophysial lobes PSA-IR was clearly found to be associated with two types of axonal fibers including: (i) elongated fibers that were homogeneously stained for PSA but were not stained for vasopressin or oxytocin, and (ii) large varicose axonal profiles that were immunostained for vasopressin or oxytocin and on which PSA-IR generally appeared localized to the periphery of the axon (Fig. 3 D - D ' ) .

not found to be associated with any of the neurohypophysial glial cell types including: (1) intensely vimentin-IR tanycytes that extend throughout the different layers of the median eminence, and (2) GFAP-IR classical astrocytes and astrocyte-like pituicytes, respectively, located in the internal layer of the median eminence and in the neural lobe (Fig. 6B-D). By contrast, all along the ventral surface of the hypothalamus lateral to the median eminence, intense PSA-IR was associated with astrocytes immunoreactive to G F A P constituting the glia limitans (Fig. 6 A - A ' ) .

3.3. Double-imrnunostaining for PSA and for GABA or tyrosine hydroxylase Throughout the different regions of the neurohypophysis, the organization pattern of PSA-IR structures was very similar to that of fibers immunoreactive to tyrosine hydroxylase or to GABA (Fig. 4). When observed on thinner optical sections at high direct magnification, PSA-IR was frequently found to be colocalized with GABA- or tyrosine hydroxylase-IR within fibers innervating the different neurohypohysial regions (Figs. 4, 5). Remarkably, within fibers innervating the median eminence and the neural lobe, the colocalization of PSA-IR and GABA-IR appeared to be more frequent than that of PSA-IR and T H (Figs. 4 B - B ' , D - D ' , 5 A - A ' , C - C ' ) . On the other hand, PSA-IR was systematically colocalized with either tyrosine hydroxylase-IR or GABA-IR within fibers innervating the intermediate hypophysial lobe (Fig. 5 B - B ' , D - D ' )

4.1. Methodological considerations During the last few years, combined immunocytochemical labeling techniques have been extensively used for characterizing the colocalization of various neuronal markers (including neurotransmitters, neuropeptides and enzymes) within the same neuronal systems (for review see H6kfelt et al., 1986). The most critical requirements of these techniques are the absence of cross-reaction between the different antibodies used and the clear discrimination of the different immunostainings obtained. In the present study, such conditions were insured by using: (i) two primary antibodies that were not immunologically related, and (ii) a confocal microscope equiped with a krypton/argon mixed gas providing two laser lines at 488 and 568 nm that unambiguously discriminate between the two secondary antibodies, respectively, labeled with fluoresceine and rhodamine. Moreover, the use of a laser scanning confocal microscope allowed us to simultaneously detect the two immunostainings on 30-40 ~m thick immunostained vibratome sections, thus avoiding the need to freeze the tissue for cryostat sectioning, which could alter the cellular morphology. Differential penetrations of the various types of immunostaining within the thickness of vibratome sections may represent a drawback of such an approach (Piekut, 1983). lmmunostaining penetrations can be, however, easily controlled under the confocal microscope, by scanning through the successive vertical planes. Using this method, all the IgG antibodies were found to penetrate the whole thickness of the vibratome sections under the conditions used here, whereas the penetration of the IgM molecules of the anti-PSA (which are considerably larger in size than IgG) was generally limited to the 10 to 15 t~m-thick

3.4. Double immunostaining for GABA and for tyrosine hydroxylase Experiments combining GABA and tyrosine hydroxylase antibodies confirmed data obtained by others (Vuillez et al., 1987; Meister and H6kfelt, 1988; Schimchowitsch et al., 1991). Both immunostainings were colocalized within most, but not all of the fibers innervating the median eminence and the neural lobe, whereas the colocalization of both immunostainings was constant within fibers innervating the intermediate hypophysiai lobe (not shown). 3.5. Double immunostaining for PSA and for vimentin or GFAP In all the neurohypophyses examined, PSA-IR was

4. Discussion

Fig. 6. Confocalpaired imagesof sections immunostainedfor PSA (A-D) and for GFAP (A ', B' ) or Vimentin(C ', D' ). Low magnification(objective 20 x) shows that in the ventral hypothalamusjust lateral to the median eminence, intense PSA-IR is associated with GFAP-IR astrocytic processesof the glia limitans located at the ventral surface of the lateral hypothalamus(A-A'), whereasstructures immunostainedfor PSA exhibit patterns that clearly differ from that of the GFAP-IR pituicytes of the neural lobe (B-B') or Vimentin-lR tanycytesof the median eminence (C-C'). Highermagnification(D-D', objective60 x ) clearly shows that in the external median eminence,PSA-IR is essentiallyassociated with fiber-likestructures that are clearlydifferentfrom the Vimentin-IRtanycyticprocessespresent in this region, by: blood vessel. A, B and C: × 300: D: x 600. Bar = 50 g.m.

42

G. Alonso /J. Chem. Neuroanat. 8 (1994) 33-45

superficial layers of the sections. In doubly immunostained sections, comparison of the localization of the two immunostainings was, thus, carefully limited to the superficial layers of the sections exhibiting intense PSAimmunostaining. A major advantage of confocal microscopy is the considerable improvement of the optical resolution in both the lateral dimension and the depth. Indeed, by using objective lens at the highest numerical aperture (i.e. here by using a 60 x objective with numerical aperture 1.4) and by decreasing the size of the confocal aperture, it is possible to observe optical sections less than 1 t~m thick (Wall6n, 1993). This greatly diminishes the risk of superimpositional artifacts that may occur when studying the colocalization of two immunocytochemical markers within axons, especially in the median eminence which contains a large variety of closely intermingled axons. On the other hand, however, decreasing the size of the confocal aperture considerably decreases the amount of fluorescence detected, thus hindering the visualization of slightly immunoreactive structures. 4.2. Identity of the PSA-IR structures As reported in previous studies (Theodosis et al., 1991; Bonfanti et al., 1992), the present observations showed that PSA-immunostaining frequently appeared as intense labeling diffusely distributed throughout the different neurohypophysial regions. However, although the cellular limits of PSA-IR structures were barely identifiable on 30-40/zm thick vibratome sections, the observation of confocal images at high direct magnification (i.e. in conditions providing the thinnest optical sections) allowed the clear association of PSA-immunostaining with specific immunocytochemical markers. Such an approach indicates that the high levels of PSAIR detected throughout the different neurohypophysial regions are not associated with the various glial cell types immunoreactive to GFAP and/or to VIM, present all along the neurohypophysis. This appears in contradiction to previous studies based on electron microscope immunocytochemistry (Theodosis et al., 1991; Kiss et al., 1992), which reported the association of PSA-IR with the membranes of astrocyte-like pituicytes of the neural lobe. It is possible that the present negative immunostaining of neurohypophysial glial cells results from a lower sensitivity of the immunofluorescent labeling depending on both the small size of the confocal aperture used here (see above), and the use of high concentration of glutaraldehyde in the fixative (0.5% vs. 0.1% in other studies). Nevertheless, since the conditions used here allow the visualization of intense PSA-IR on GFAP-IR astrocytes all along the glia limitans (Fig. 6A-A '), and on GFAP-IR reactive astrocytes occurring around various types of surgical lesions in the adult rat brain (Alonso and Privat, 1992), it can be assumed that only very low amounts of PSA are associated with the

neurohypophysial glial cells. The observations of double immunostained sections also indicate that the axonal fibers immunoreactive to somatostatin or to corticotropin releasing hormone, which project all along the external median eminence exhibit little if any PSAIR. On the other hand, the intense PSA-IR detected throughout the neurohypophysis of adult rats appears essentially associated with two types of neurohypophysial axonal fibers including: (l) vasopressin- and oxytocin-IR fibers terminating in the neural lobe, and (2) GABA- or tyrosine hydroxylase-IR fibers innervating the median eminence and the neuro-intermediate hypophysial lobe. The present observations that PSA-IR is associated with both vasopressin- and oxytocin-IR axons terminating in the neural lobe confirm previous findings (Tbeodosis et al., 1991, Kiss et al., 1992). Surprisingly, however, the PSA-immunostaining was generally found to be localized at the periphery of these large axons located in the neural lobe, while it appeared homogeneously distributed within the thin PSA-IR axonal fibers detected throughout the whole neurohypophysis. A possible explanation for such a differential immunostaining pattern is that optical sections of about 1 #m in thickness, observed under the confocal microscope, include the complete thickness of the thin GABA-IR or tyrosine hydroxylase-IR fibers innervating the neurohypophysis, whereas they only contained sections of the large vasopressin- or oxytocin-axonal dilatations that are present in the neural lobe. It is known from previous neuroanatomical studies that the median eminence and neurointennediate lobe are densely innervated by fibers immunoreactive to both GABA and tyrosine hydroxyIase that originate from GABAergic/dopaminergic neurons located in the dorsomedial and periventricular subdivisions of the arcuate nucleus (Vuillez et al., 1987; Meister and H6kfelt, 1988; Schimchowitsch et al., 1991). Interestingly, the present data demonstrate that within the intermediate hypophysial lobe, PSA-NCAM is systematically associated with fibers immunoreactive to GABA or tyrosine hydroxylase, two neuronal markers which have been demonstrated to be constantly colocalized within GABAergic/dopaminergic fibers innervating this organ (Schimchowitsch et al., 1991). Therefore, it can be assumed that the large majority of fibers double-labeled for PSA and for GABA or tyrosine hydroxylase detected in the median eminence and the neural lobe actually corresponds to GABAergic/dopaminergic axons originating in neurons of the mediobasal hypothalamus. The present study also indicates that although the large majority of the GABA-IR fibers projecting to the external median eminence or the neural lobe also exhibit PSA-IR, some tyrosine hydroxylase-IR fibers innervating these regions appeared PSA-negative. Since it is known that tyrosine hydroxylase is not detected

G. Alonso /J. Chem. Neuroanat. 8 (1994) 33-45

within the totality of GABAergic neurons innervating the neurohypophysis (present observations and Vuillez et al. 1987; Meister and H6kfelt, 1988; Schimchowitsch et al., 1991) it is likely that a number of PSA-IR/GABAIR fibers innervating the neurohypophysis correspond to the axonal projections of non-catecholaminergic hypothalamic GABAergic neurons. Moreover, it is possible that some of the PSA-IR/tyrosine hydroxylaseIR fibers detected throughout the different neurohypophysial regions correspond to the noradrenergic projections that have been identified in the neurohypophysis (Bj6rklund et al., 1973). However, such fibers certainly represent a minority of the tyrosine hydroxylase-IR fibers detected here, since experiments of total hypothalamic deafferentation demonstrate that the contribution to the tyrosine hydroxylase content of the neurohypophysis by extrahypothalamic noradrenergic neurons is negligible (Brownstein et al., 1976). 4.3. Functional significance of the presence of PSA on neurohypophysial axons PSA was detected by means of an antibody that specifically recognizes ~-2,8-1inked PSA (Rougon et al., 1986). Although PSA residues can be contained in other neuronal molecules such as sodium channels (James and Agnew, 1987, Zuber et al., 1992), it is clear from a number of immunocytochemical and biochemical studies that NCAM is the major carrier of PSA in mammals' brains (Aaron and Chesselet, 1989; Theodosis et al., 1991; Bonfanti et al., 1992; Le Gal La Salle et al., 1992). The functional significance of the continued expression of high levels of PSA-NCAM by some specific neurons of the adult brain is still unknown. Since PSA modulates the binding of NCAM itself (Sadoul et al., 1983) and of other adhesion molecules (Rutishauser and Landmesser, 1991; Landmesser, 1992), the idea is generally accepted that PSA is a marker of plasticity. Concerning the neurohypophysial neurons studied here, a first noteworthy feature is that, in contrast to the majority of the other neuron types, their axon terminals do not establish synaptic connections with their targets. Instead, most of them terminate at the vascular spaces of the median eminence or the neural lobe, or in close apposition with the endocrine cells of the intermediate hypophysial lobe. It is now well documented that under physiological or experimental conditions, some neurohypophysial axons can modify their interactions with the surrounding glial environment. In the external layer of the median eminence or in the neural lobe, for instance, tanycyte or pituicyte processes, respectively, form a continuous barrier between axon terminals and the blood vessels. Under specific conditions these glial processes undergo extensive retractions resulting in direct apposition of these axons onto the perivascular space. Such morphological modifications that are assumed to facilitate the release of secretory products

43

into the blood stream have been described in the median eminence after adrenalectomy, which stimulates the release of corticotropin releasing hormone (Wittowski and Scheuer, 1974), and in the neural lobe during dehydration or lactation, which stimulate the release of vasopressin or oxytocin (Tweedle and Hatton, 1982, 1987 ). As already proposed (Theodosis et al., 1991), the high levels of PSA detected on vasopressin or oxytocin axons may play a major role in such structural reorganizations occuring as the result of intense functional stimulation. However, such an interpretation does not fit in with various other types of peptidergic endocrine axons terminating in the median eminence, and, namely, with the corticotropin releasing hormoneIR axons which were found to contain very little if any PSA. Another remarkable capacity of morphological plasticity of vasopressin and oxytocin neurohypophysial axons concerns their ability to regenerate after a surgical lesion (DeUmann, 1973). Since membranes containing large amounts of PSA-NCAM are thought to favor the moving of cellular structures, it can be assumed that the remarkable capacity for regeneration exhibited by vasopressin and oxytocin axons is related to the presence of PSA on their membrane. Similarly, it can be postulated that PSA associated with GABAergic/ dopaminergic axons innervating the median eminence and the neurointermediate lobe is also involved in such structural plasticity. Although some information is available on the morphological plasticity of these specific axons, this idea may be supported by the present finding that high levels of PSA were constantly associated with those GABAergic/dopaminergic axons that project outside the brain through the ventral layers of the hypothalamus, to innervate the intermediate hypophysial lobe. Ongoing studies aimed at characterizing the different neuronal systems of the adult rat brain that continue to express high levels of PSA will certainly help in better understanding the functions of this molecule within mature neurons.

Acknowledgements This work was supported by IRME. The author wishes to thank Dr. G. Rougon (Lumigny, France) for her generous gift of the antibody against PSA. He also thanks A. Legrand for her excellent technical assistance and Dr. C. Faivre-Sarailh for valuable advice and discussion on this manuscript.

References Aaron, L.I. and Chesselet, M.F. (1989) Heterogenous distribution of polysialylated neuronal-cell adhesion molecule during post-natal development and in adult: an immunohistochemical study in the rat brain. Neuroscience 28, 701-710. Alonso, G. (1988) Effects of colchicine on the intraneuronal transport of secretory material prior to the axon: a morphofunctional study

44

G. Alonso /J. Chem. Neuroanat. 8 (1994) 33-45

in hypothalamic neurosecretory neurons of the rat. Brain Res. 453, 191-203. Alonso, G. and Privat, A. (1992) Reactive astrocytes involved in the formation of lesional scars differ in the mediobasal hypothalamus and in other forebrain regions. J. Neurosci. Res. 34, 523-538. Alonso, G., Siaud, P. and Assenmacher I. (1988) Immunocytochemical ultrastructural study of hypothalamic neurons containing CRF in normal and adrenalectomized rats. Nearoscience 24, 553-565. Arluison, M., Diet, M. and Thibault, J. (1984) Ultrastructural morphology of dopaminergic synapses in the striatum of the rat using tyrosine hydroxylase immunocytochemistry: a topographical study. Brain Res. Bull. 13, 269-285. Bjrrklund, A., Moore, R.Y., Nobin, A. and Steveni, U. (1973) The organization of tubero-hypophyseal and reticulo-infundibular catecholamine neuron systems in the rat brain. Brain Res. 117, 287-295. Bonfanti, L., Olive, S., Poulain, D.A. and Tbeodosis, D.T. (1992) Mapping of the distribution of polysialylated neural cell adhesion molecule throughout the central nervous system of the adult rat: an immunohistochemical study. Neuroscience 49, 419-436. Brownstein, J.M., Palkovits, M., Tappaz, M.L., Saavedra, M. and Kizer, J.S. (1976) Effect of surgical isolation of the hypothalamus on its neurotransmitter content. Brain Res. 117, 287-295. Chuong, C.M. and Edeiman, G.M. (1984) Alterations in neural cell adhesion molecules during development of different regions of the nervous system. J. Neurosei. 4, 2354-2368. Dcllmann, H.D. (1973) Degeneration and regeneration of neurosecretory systems. Int. Rev. Cyt. 36, 215-315. Eddman, G.M. and Crossin, K.L. (1991) Cell adhesion molecules in neural morphogenesis. In Volume Transmission in the Brain (eds Fuxe, K. and Agnati, L.F.), pp. 25-47. Raven Press, New York. Finne, J., Finne, U., Deagostini-Bazin, H. and Goridis, C. (1983) Occurence of alpha 2-8 linked polysialosyl units in a neural cell adhesion molecule. Biochem. Biophys. Res. Commun. 112, 482-487. Hoffman, S. and Edelman, G.M. (1983) Kinetics of homophilic binding by embryonic and adult forms of the neural cell adhesion molecule. Proc. NatL Acad. Sci. USA 80, 5762-5766. Hfkfelt, T., Holets, V.R., Staines, W., Meister, B., Melander, T., Sehalling, M., Sehultzberg, M., Freedman, J., Bj6rklund, H., Olson, L., Lindh, B., Elfin, L., Lundberg, J.M., Lindgren, J.A., Samnelsson, B., Pernow, B., Terenius, L., Post, C., Everitt, B. and Goldstein, M. (1986) Coexistence of neuronal messengers - - an overview. Prog. Brain Res. 68, 33-78. James, W.M. and Agnew, W.S. (1987) Multiple oligosaccharid¢ chains in the voltage sensitive Na channel from Electrophorus electricus: evidence for c~-2,8-1inked polysialic acid. Biochem. Biophys. Res. Commun. 148, 817-826. Kiss, J.Z., Wang, C. and Rougon, G. (1992) Nerve dependent expression of high PSA NCAM in neurohypophysial astrocytes of adult rats. Neuroscience 53, 213-221. Landmesser, L. (1992) The relationship of intramuscular nerve branching and synaptogenesis to motoneuron survival. J. Neurobiol. 23, i 13 I - 1139. L¢Gal La Salle, G., Rougon, G. and Valin, A. (1992) The embryonic form of neural cell surface molecule (E-NCAM) in the rat hippocampus and its reexpression on glial cells following kainic acidinduced status epilepticus. J. Neurosci. 12, 872-882. Meister, B. and Hrkfelt, T. (1988) Peptide- and transmitter-containing neurons in the mediobasal hypothalamus and their relation to GABAergic systems: possible roles in control of prolactin and growth hormone secretion. Synapse 2, 585-605. Miller, P.D., Chung, W.W., Lagenaur, C.F. and De Kosky, S.T. (1993) Regional distribution of neural cell adhesion molecule (NCAM) and LI in human and rodent hippocampus. J. Comp. Neurol. 327, 341-347.

Miragall, F., Kadmon, G., Husmann, M. and Schachner, M. (1988) Expression of cell adhesion molecules in the olfactory system of adult mouse: presence of embryonic form of NCAM. Dev. Biol. 129, 516-531. Muijen, G.N.P., Van Ruiter, D.J., Leeuven, C., Van Prins, F.A., Rietsema, K. and Warnaar, S.O. (1984) Cytokeratin and neurofilament in lung carcinomas. Am. J. Pathol. 116, 363-369. Perera, A.D., Lagenaur, C.F. and Plant, T.M. (t993) Postnatal expression of polysialic acid neural cell adhesion molecule in the hypothalamus of the male rhesus monkey (Maeaca mulatta). Endocrinology 133, 2729-2735. Piekut, D.K. and Casey, M. (1983) Penetration of immunoreagents in vibratome-sectioned brain. A light and electron microscope study. J. Histochem. Cytochem. 31,669-674. Rothbard, J.B., Brackenbury, R., Cunningham, B.A. and Edelman, G.M. (1982) Differences in the carbohydrate structures of neuralcell-adhesion molecules from the adult and embryonic chicken brains. J. Biol. Chem. 257, 11064-11069. Rougon, G., Deagostini-Bazin, H., Him, M. and Goridis, C. (1982) Tissue and developmental stage-specific forms of a neuronal celt surface antigen linked to differences in glycosilation of a common polypeptide. Fur. ,91ol. Biol. Org. J. 1, 1239-1244. Rougon, G., Dubois, C., Buckley, N., Magnani, J.L. and Zollinger, W. (1986) A monoclonal antibody against Meningococcus group B polysaccharides distinguishes embryonic from adult N-CAM. J. Cell. Biol. 103, 2429-2437. Rutishauser, U. and Jessel, T.M. (1988) Cell adhesion molecules in vertebrate neural development. Physiol. Rev. 68, 819-857. Rutishauser, U., Acheson, A., Hall, A., Mann, D. and Sunshine, J. (1988) The neural cell adhesion molecule (N-CAM) as a regulator of cell-cell interactions. Science 240, 53-57. Rutishauser, U. and Landmesser, U (1991) Polysialic acid on the surface of axons regulates patterns of normal and activity-dependent innervation. Trends Neurosci. 14, 528-532. Sadoul, R., Him, M., Deagostini-Bazin, H., Rougon, G. and Goridis, C. (1983) Adult and embryonic neural cell adhesion molecules have different binding properties. Nature 304, 347-349~ Schimchowitsch, S., Vuillez, P., Tappaz, M.L., Klein, M.J. and Stoeckel, M.E. (1991) Systematic presence of GABA-immunoreactivity in the tubero-infundihuiar and tubero-hypophyseal dopaminergic axonal systems: an ultrastructural immunogold study on several mammals. Exp. Brain Res. 83, 575-586. Seki, T. and Arai, Y. (1993) Highly polysialilated neural cell adhesion molecule (NCAM-H) is expressed by newly generated granule cells in the dentate gyrus of the adult rat. J. Neurosci. 13, 2351-2358. Szabat, E., Soinila, S., Hhppl61/i, O., Linnala, A. and' Virtanen, I. (1992) A new monoclonal antibody against the GABA'protein conjugate shows immunoreactivity in sensory neurons of the rat. Neuroscience 47, 409-420. Theodosis, D.T., Rougon, G. and Poulain, D. (1991) Retention of embryonic features by an adult neuronal system capable of plasticity: polysialylated neural cell adhesion molecule in the hypothalamo-neurohypophysial system. Proc. Nat/. Acad Sci. USA 88, 5494-5498. Tweedle, C.D. and Hatton, G.I. (1982) Magnocellular neuropeptidergic terminals in neurohypophysis: rapid glial release of enclosed axons during parturition. Brain Res. Bull. 8, 205-209. Tweedle, CD. and Hatton, G.1. (1987) Morphological adaptability at neurosecretory axonal endings on the neurovascular contacts zone of the rat neurohypophysis. Neuroscience 20, 241-246. Viale G., Cambacorta, M., Goggi, G., Dell'Orto, P., Milani, M. and Dogiioni, C. (I 991) Glial fibrillary acidic protein immuno~ctivity in normal and diseased human breast. Virchows Arch. A. Pathol. 418, 339-348. Vuillez, P., Carbajo-Perez, S. and Stoeckel, M.E. (1987) Colocalization of GABA and tyrosine hydroxylase immunoreactivities in the

G. Alonso/J. Chem. Neuroanat. 8 (1994) 33-45

axons innervating the neurointermediate lobe of the rat pituitary: an ultrastructural immunogold study. Neurosci. Lett. 79, 53-58. Wall~n, P. (1991) Confocal microscopy in chemical neuroanatomy. J. Chem. Neuroanat. 4, 387-395. Wittowski, W. and Scheuer, A. (1974) Functional changes of the neuronal and glial elements at the surface of the external layer of the median eminence. Z. Anat. Entw, Gesch. 143, 255-262.

45

Zuber, C., Lackie, P.M., Catterall, W.A. and Roth, J. (1992) Polysialic acid is associated with sodium channels and the neural cell adhesion molecule N-CAM in adult rat brain. J. Biol. Chem. 267, 9965-9971.