Morphologic fate of diencephalic prosomeres and their subdivisions revealed by mapping cadherin expression

Animals

Embryos from the following stages (according to Hamburger and Hamilton, 1951) were studied: embryonic day 11 (E11; stage 37), E15 (stage 41), and E19 (stage 45).

Antibodies

In addition to the antibodies against cadherins described in the accompanying paper (Yoon et al., 2000), mouse monoclonal antibody R5 was used to visualize radial glial cells and their processes. This antibody specifically reacts with a vimentin-associated protein (Dräger et al., 1984; Herman et al., 1993; kind gift of U. Dräger).

Histochemistry

Immunostaining procedures are as described in the accompanying paper by Yoon et al. (2000). Brains from chicken embryos at E11, E15, and E19 were fixed for 2–6 hours at 4°C. Several complete series of sections were cut through entire brains (E11: three transverse series, one horizontal series, and one sagittal series; E15: one transverse series, one horizontal series, and one sagittal series; E19: one transverse series and one sagittal series) and were stained with antibodies. For each antibody, sections within the set of immunostains were spaced 100–140 μm apart.

For embryos at E15 and E19, one of the parallel section series was reacted for acetylcholine esterase (AChE) activity, which was visualized with a modified Karnovsky-Roots technique, as described in detail previously (Gänzler and Redies, 1995). This material provided additional evidence for the identification of gray matter structures.

All stains were visualized with Axiophot or Ultraphot microscopes (Zeiss, Oberkochen, Germany). Photographic images of selected sections were processed as described elsewhere (Yoon et al., 2000). To display the staining results for two or three cadherins simultaneously (Fig. 11), photographic images were enhanced in contrast, color coded, and superimposed by using the Photoshop software program (Adobe Systems, Mountain View, CA).

Data analysis and terminology

Data analysis was based on the prosomeric model of the vertebrate brain by Puelles, Rubenstein, and coworkers (Bulfone et al., 1993; Puelles and Rubenstein, 1993; Puelles, 1995). In addition, the seminal studies by Rendahl (1924) and Vaage (1969) as well as more recent publications on the development of the chicken diencephalon (Martínez-de-la-Torre, 1985; Puelles et al., 1987; De Castro et al., 1998) were consulted. For identification of gray matter structures, we generally followed the terminology used in the atlases of the chicken brain by Kuenzel and Masson (1988) and of the pigeon brain by Karten and Hodos (1967). For some gray matter areas, additional studies served as a frame of reference for subdivision schemes and terminology, e.g., the studies by Martínez-de-la-Torre et al. (1990) and Puelles et al. (1991).

General observations

The four cadherins studied in the current report (cad6B, cad7, Rcad, and Ncad) each are expressed in a spatially restricted manner consistent with patterns of radial glial heterogeneity in the chicken diencephalon at E11 and E15 (E11: Figs. 2-12; E15: Fig. 13). In Part I below, evidence is described that supports the general usefulness of mapping cadherin expression and studying radial glial topology for the localization of prosomere boundaries and subdivisions in the chicken diencephalon at intermediate stages of development. In Part II, the pattern of cadherin-expressing gray matter structures in each embryonic division is described in more detail. Some novel aspects for specific diencephalic regions that have been studied in less detail previously also are described (Figs. 14-17).

Part I: Observations on prosomeric boundaries and subdivisions

Prosomere-derived nuclear domains in the diencephalon (Fig. 1) were identified in the past largely by analysis of cytoarchitectonic development. They were referred to as prosomeres and named the synencephalon (our p1), the posterior parencephalon (p2), and the anterior parencephalon (see Table 1; Rendahl, 1924; Puelles et al., 1991). Note that Rendahl's anterior parencephalon was subdivided later into p3–p6 by Puelles and Rubenstein (1993) and Puelles (1995). Alar derivatives of the prosomeres are the pretectal nuclei in p1, the dorsal thalamic and epithalamic nuclei in p2, and the ventral thalamic formations in p3 (Fig. 1A). The classical hypothalamus (mamillary, tuberal, and retrochiasmatic nuclei) largely falls across basal p4–p6, as postulated more recently on morphologic and molecular evidence (Puelles and Rubenstein, 1993; Puelles, 1995; Shimamura et al., 1995). In contrast, anterior, suprachiasmatic, and preoptic parts of the conventional hypothalamus fall within the alar plate of the p4–p6 domains together with the eminentia thalami (Fig. 1). Observed patterns of cadherin expression relative to cytoarchitectonic features (Figs. 2-6) can be interpreted readily in terms of this neuromeric morphologic model, as summarized in Figure 7.

Part II: Detailed description of cadherin expression in the diencephalon

The transverse sections shown in Figures 2-6 illustrate five levels of sectioning through the E11 diencephalon that were immunoreacted with either cad6B, cad7, or Rcad antibodies plus an adjacent section that was stained with thionine for Nissl substance. The levels are arranged in a rostral-to-caudal sequence. Figure 7 schematically identifies the prosomeric domains and main alar subdivisions for the section levels shown in Figures 3-6; for the rostralmost level (Fig. 2), the same information is drawn in Figure 2D.

Table 2 lists all gray matter structures identified in this study, grouped according to the subdivisions in which each structure is located, together with the profile of cadherin expression at E11. Table 2 also indicates the figure(s) in which each structure is shown. In Figure 7 and in the other schematic diagrams (Figs. 1, 2C, 8D, 9D, 10D, 12F, 14J–L, 18), the pretectum (PT or p1), the ventral thalamus (VT or p3), and p5 are shaded in gray, whereas the dorsal thalamus (DT or p2) and p4 and p6 are not shaded. Intermediate shading and striped areas represent alar subdivisions within the pretectum (PTj) and the dorsal thalamus (DTi).

Table 2. Gray Matter Derivatives of Diencephalic Prosomeres and Their Secondary Subdivisions and Cadherin Expression Profile in Chicken Embryos at 11 Days of Incubation1

Structure	Immunoreactivity				Abbreviation2	Shown in Figure(s)
	Cad6B	Cad7	Rcad	Ncad
Pretectum (alar plate of prosomere 1)
Commissural pretectum (PTc)
Area pretectalis	−	−	p	p	AP	6, 7, 12, 14, 18
Area pretectalis diffusa	−	+	−	−	APd	6, 12, 14
Nucleus of posterior commissure	−	+	−	−	CP	—
Interstitial nucleus of the pretectosubpretectal tract	−	−	s	+	IPS	5–7, 10–12, 18
Principal pretectal nucleus	−	−	s	+	PPT	6, 7, 10–12, 14, 16, 18
Medial pretectal nucleus	−	(+)	−	−	PTM	10, 14
Subpretectal nucleus	−	p	−	p	SP	5–7, 10–12, 18
Juxtacommissural pretectum (PTj)
External pretectal nucleus	+	+	−	−	PE	4–7, 9–12, 16, 18
Lateral spiriform nucleus	(s)	−	+	−	SpL	6, 7, 10, 12–14, 18
Medial spiriform nucleus					SpM	14, 18
Caudolateral part	−	+	−	−	SpMcl	6, 7, 10, 12, 14
Rostromedial part	−	+	−	−	SpMrm	6, 7, 10, 14
Precommissural pretectum (PTp)
Nucleus of the basal optic root3	p	p	−	−	BOR	13
Nucleus dorsofrontalis	−	p	−	−	DF	5–7, 12–14, 18
Principal precommissural nucleus	−	+	−	−	PPC	3–6, 7, 9, 10, 12–14, 16, 18
Superficial synencephalic nucleus	−	+	−	−	SS	3–5, 7, 9, 10, 12, 16, 18
Dorsal thalamus (alar plate of prosomere 2)
Epithalamus (ET)
Lateral habenula	−	p	−	−	HL	4, 7–9, 15, 18
Medial habenula					HM	4–6, 9
Dorsal portion	+	p	p	−	HMd	4–7, 15
Ventral portion	−	+	+	(+)	HMv	4–9, 15
Pineal gland (epiphysis)	s	p	p	+	Pi	18
Dorsal tier (DTd)
Dorsal anterior nucleus	−	−	−	−	DA	3, 4, 7, 12
Dorsal posterior nucleus	−	(+)	p	(+)	DPo	5–7, 9, 12, 14
Dorsointermediate posterior nucleus	−	+	p	−	DIP	4–7, 9, 14
Dorsal intermediate ventral anterior nucleus	−	+	+	−	DIVA	3, 4, 7, 8, 12, 13
Dorsolateral anterior nucleus	−	+	p	−	DLA	3, 7, 8, 12, 13
Dorsolateral lateral nucleus4	−	+	−	−	DLL	3, 4, 7, 9, 13, 17, 18
Dorsolateral medial nucleus	−	+	−	−	DLM	—
Dorsolateral posterior nucleus	−	p	(+)	−	DLP	4–7, 15
Dorsomedial anterior nucleus	−	−	−	−	DMA	8
Dorsomedial posterior nucleus	−	−	−	−	DMP	4–7, 14
Microcellular superficial nucleus	−	−	(+)	+	SMC	1, 3–5, 7–9, 12, 13, 15, 17, 18
Parvocellular superficial nucleus4	−	p	−	−	SPC	1, 4, 5, 7–9, 12, 13, 17, 18
Subhabenular region, lateral	−	p	p	p	SHl	4, 6–8, 15
Subhabenular region, medial	−	−	−	−	SHm	4–8, 15
Ventrointermediate area	−	+	+	+	VIA	3, 7, 8
Intermediate tier (DTi)
Intermediomedial nucleus	−	−	−	−	IM	4–7, 9, 14
Intermedioposterior nucleus					IP	7
Lateral portion	+	p	p	−	IPl	5, 6, 14, 16
Medial portion	(p)	+	(+)	(+)	IPm	5, 6, 12, 14
Rotundus nucleus	p	p	p	p	R	1, 3, 5, 7, 9, 11–13, 18
Triangular nucleus	−	−	p	(+)	T	4–7, 9, 11, 12, 16, 18
Nucleus uvaeformis	−	−	−	−	Uva	10, 18
Ventral tier (DTv)
Nucleus ovoidalis	−	+	−	−	OV	1, 5, 7, 9, 11, 16, 18
Posterior nucleus of the ventral supraoptic    decussation	s	+	+	(+)	PDSV	5–7, 10, 13, 18
Posterointermediate nucleus	(+)	s	+	(+)	PI	6, 7, 13, 14, 18
Nucleus paramedianus internus	−	(+)	s	−	PM	4–7, 9, 14
Posterior nucleus	−	(+)	(+)	−	Po	5–7, 14
Nucleus periovoidalis	−	(+)	+	(+)	POV	4, 5, 7
Posteroventral nucleus	s	(+)	+	(+)	PV	5–7, 13
Retroovoidal area	−	(p)	+	(+)	ROV	4–7, 14
Nucleus semilunaris periovoidalis	−	p	+	+	SPO	4, 5, 7, 9
Nucleus zeta	−	+	(+)	+	Z	5–7, 13, 18
Anteroventral subdivision (DTav)
Perirotundic area	(+)	+	+	−	APR	1, 3, 4, 7, 9, 11, 12, 16, 18
Retrorotundic area	s	(+)	+	(+)	RR	5, 7, 11, 12, 16
Prospective interstitial nucleus of the optic tract	(+)	−	+	+	ITO	1, 3–5, 7, 9, 12, 13, 16, 18
Subrotundic nucleus	−	−	s	−	SRt	4, 5, 7, 9, 11–13, 16
DT/VT boundary region
Zona limitans intrathalamica	p	−	p	p	zl	3, 4, 7–9, 12
Ventral thalamus (alar plate of prosomere 3)
Area triangularis	−	−	−	−	AT	2, 18, 13
Ventral geniculate nucleus	p	+	−	−	GV	2–5, 7–9, 12, 13, 16, 18
Nucleus intercalatus	−	s	−	−	ICT	3, 7–9, 12
Lateroanterior nucleus	+	n, s	−	−	LA	2, 8, 12, 18
Superior reticular nucleus					RS	2
Dorsal portion	(s)	p	(s)	−	RSd	3, 7, 8
Ventral portion	−	(+)	−	−	RSv	3, 8
Ventrolateral nucleus	−	(p)	−	−	VL	2, 3, 7, 8, 18
Dorsal area of ventral thalamus	−	−	−	−	VTD	3, 7, 12, 18
Zona incerta	−	p	(s)	−	ZI	3, 7, 9
Basal plate of prosomeres 1–3
Retromamillary region	+	−	+	−	RM	18
Interstitial nucleus of Cajal, rostral portion	+	+	−	−	ICr	10
Nucleus of Darkschewitsch, rostral portion	−	+	−	(+)	Dr	10, 18
Anteromedial preoptic nucleus	−	s	p	−	AM	2, 18
Alar and roof plates of prosomeres 4–6
Paraphysis	−	p	p	+	Pa
Lateral preoptic area	−	+	−	−	LPA	2, 18
Medial (main) portion of preoptic nucleus	−	−	−	−	POM	2, 18
Lateral portion of preoptic nucleus	−	−	+	−	POMl	2
Medial suprachiasmatic nucleus	+	p	(+)	−	SCNm	2
Lateral suprachiasmatic nucleus	−	−	+	+	SCNl	2
Basal and floor plate of prosomere 4
Paraventricular nucleus	−	−	+	−	PVN	8, 18
Anterior nucleus of the ventral supraoptic   decussation	+	+	+	−	ADSV	3, 7, 9
Anterior nucleus of the ansa lenticularis	(s)	(s)	+	−	ALA	3, 4, 7, 9, 11, 16, 18
Lateral mamillary nucleus	−	−	(+)	+	ML	4, 7, 18
Nucleus of the occipitomesencephalic tract	(s)	s	+	−	OM	3, 4, 7, 18
Periventricular hypothalamic organ	−	−	−	−	PVO	4, 5, 7
Stratum cellulare externum	−	(s)	s	−	SCE	5–7, 18
Stratum cellulare internum	−	(s)	−	−	SCI	5–7, 18
Lateral hypothalamic region	−	−	p	p	LH	3, 5–7
Basal and floor plate of prosomere 5
Dorsomedial hypothalamic nucleus	−	+	+	+	DM	4, 5, 7
Inferior hypothalamic nucleus	+	−	+	−	IH	4, 5, 7, 18
Infundibular nucleus	−	−	+	+	IN	4, 5, 7, 18
Mamillary area	−	(+)	−	−	MA	6, 7, 18
Periventricular hypothalamic nucleus	s	−	−	−	PH	3, 18
Ventromedial hypothalamic nucleus	p	−	+	−	VM	3, 7, 8, 18
Median eminence	+	−	−	+	ME	4, 7, 18
Stalk of neurohypophysis	+	−	−	−	NH stalk	5
Basal and floor plate of prosomere 6
Anterobasal nucleus	−	−	(+)	+	AB	3, 7, 18
Juxtacommissural plate	+	−	+	p	JCP	—

• 1 −, Structure is not immunoreactive; +, structure is immunoreactive; n, only neuropil is immunoreactive; p, only parts of the structure are immunoreactive; s, only scattered cells are immunoreactive. Parenthesis denote weak immunoreactivity.
• 2 Abbreviation used in the current work.
• 3 This nucleus extends rostrally into p2 (Fig. 10) and p3 (Figs. 4-7).
• 4 Likely to originate in the anteroventral subregion (DTav; Puelles et al., 1999b).

Pretectum (alar p1).

The sets of nuclei derived from the three subdivisions of the pretectum display prominent differences in cadherin expression (Figs. 3-6, 10, 11C, 12, 14). Ncad is expressed by the migrated nuclei of the commissural pretectum (PPT, IPS, and SP; Figs. 11C, 12D; Redies et al., 1993) but not by any of the nuclei of the other two subdivisions. In the area pretectalis, which lies at the dorsal pial surface of the commissural pretectum, a diffuse (deep) region (APd) can be distinguished from a dense (superficial) part (AP) on the basis of differential cad7 expression (AP, APd; Figs. 14D–F; compare with Medina et al., 1994; Puelles and Medina, 1994).

In the juxtacommissural pretectum, Rcad expression is prominent in SpL (Figs. 6C, 12C), and there also are some scattered, Rcad-positive cells in PPT and IPS (Arndt and Redies, 1996). Cad7 expression is restricted to the medial spiriform nucleus, which appears to be divided into rostromedial and caudolateral parts (SpMrm and SpMcl in Figs. 10A, 12B, 14E,F); the rostralmost parts of SpM show less Cad7 expression (SpMrm, SpM in Figs. 6B, 14D). Cad7 immunoreactivity is conspicuously absent from pretectal nuclei that express Ncad and Rcad, except for SP, which shows cad7 immunoreactivity medially. Cad6B is expressed weakly in SpL (Figs. 6A, 12A, 14A–C) and in the external pretectal nucleus (PE in Figs. 5A, 6A, 12A).

The precommissural pretectum is dominated by cad7 immunoreactivity that is especially prominent in its principal precommissural and superficial synencephalic nuclei (PPC and SS in Figs. 3B, 4B, 5B, 6B, 10A, 12B, 14D–F). The location of the dorsofrontal nucleus (DF), which is a dorsal prolongation of PPC and weakly expresses cad7, is rostromedial to SpM (DF in Figs. 6, 7D, 12B,E,F, 14D,E,G,H). The caudalmost part of the nucleus of the basal optic root originates in p1 and shows partially overlapping expression of cad6B and cad7 (BOR in Fig. 13A,B; Wöhrn et al., 1998).

Epithalamus and dorsal thalamus (alar p2)

Epithalamus.

The epithalamus, which usually is divided into medial and lateral habenular nuclei (HM and HL, respectively), is quite heterogeneous in our cadherin-stained material (Figs. 4-6; enlargement in Fig. 15). A dorsal part of the medial habenular nucleus (HMd) placed medial and under the stria medullaris is strongly cad6B positive (Figs. 4-6, 15A); a ventral subarea of HMd is cad7 positive (Fig. 15B). The larger ventral part of the medial habenular nucleus (HMv) is labeled strongly for Rcad and shows weaker expression of cad7 periventricularly (Figs. 4-6, 15B,C). The ventricular stratum associated medially with HM strongly expresses Rcad and Ncad (Fig. 15C,D) and does not express cad6B, which is in contrast to the remaining p2 alar plate (Fig. 15A). The expression of cad6B and cad7 also marks projection fibers of HMd and HMv entering the retroflex fascicle (fr in Figs. 6A, 15A,B). No Rcad-immunoreactive fibers could be followed from HMv into fr, but Rcad is expressed in the radial glial processes of the habenular region (Fig. 15C,E). A compact bundle of Rcad-positive nerve fibers appears within the stria medullaris tract (sme in Fig. 2-6, 15). Here, this bundle is identified tentatively as an olfactohabenular tract, because it can be traced all the way from the telencephalon (oh in Figs. 2C–6C, 8, 15). This tract also stands out in glia-immunostained and Nissl preparations (Fig. 15E,F). Its fibers weakly express cad7 and Ncad (oh in Fig. 15B,D). The lateral habenular nucleus (HL), placed interstitially to passing fibers largely ventral to the stria medullaris, shows only limited expression of cad7. Some Ncad-positive cells of the dorsal thalamic subpial cell lamina, which we call the superficial microcellular nucleus (SMC), seem to invade slightly the surface of the epithalamus (SMC in Fig. 15D,F).

Dorsal tier.

The dorsal tier of the dorsal thalamus predominantly expresses cad7, particularly by cell masses adjoining the boundary with the intermediate tier (see sagittal sections in Figs. 12B, 13B). This pattern is displayed in more detail in Figures 3B–6B. The entire DLA, including the contained DIVA and ventrointermediate area (VIA) cell groups (which selectively coexpress Rcad), strongly expresses cad7 and shows a straight boundary with the nucleus rotundus of the underlying intermediate tier (Figs. 3B,C, 12B, 13B). Medially, the cad7 expression gradually diminishes toward the periventricular dorsomedial anterior nucleus, which is largely cad7 negative (DMA in Figs. 8A, 11A). Caudally, the cad7-positive complex of the dorsal tier becomes reduced to the dorsal intermedioposterior nucleus (DIP), which also shows a gradual loss of signal toward the periventricular dorsomedial posterior nucleus (DMP; Figs. 4B, 5B, 6B, 9A, 11B, 14D–L). The dorsal tier neurons that dorsally and caudolaterally fill the space between the cad7-positive DLA/DIP complex and the epithalamus hardly express cad7. This cell group can be divided into a rostrodorsally placed dorsal anterior nucleus (DA in Figs. 3C, 4B,C, 12B) and a caudodorsally located dorsal posterior nucleus (DPo in Figs. 5B,C, 6B,D, 9A, 12B, 14D–L). Laterally and medially, both of these nuclei are covered by the subhabenular nuclei. In addition, lateral to DIP, the dorsolateral posterior nucleus is found that shows only weak cad7 and Rcad immunoreactivity (DLP in Figs. 4B–D, 5B–D). The DLP, DIP, and DMP nuclei are aligned with the dorsalmost, cell-poor stripe of the intermediate tier (Figs. 4D, 5D). The subhabenular region of the dorsal tier coincides topographically with a separate ventricular bulge and the dorsalmost ventricular expression of cad6B (Figs. 4A–6A); this region shows a rather heterogeneous cadherin immunoreactivity pattern that is distinct laterally but less prominent medially (SHm and SHl in Figs. 4-6, 15).

Some superficially placed cell populations in the dorsal tier of the dorsal thalamus strongly express cad7 and more weakly express cad6B. Apart from the nucleus dorsolateralis lateralis (DLL), which originates in the anteroventral compartment (Puelles et al., 1999), these populations are formed by the superficial parvicellular nucleus (SPC) and a previously scarcely known thin subpial layer of diminutive cells termed here the superficial microcellular nucleus (SMC; Guillén, 1992). Ncad is expressed by SMC but not by SPC, and vice versa for cad7 (Figs. 12B,D, 17A–F).

Intermediate tier.

The intermediate tier of the dorsal thalamus has a narrow periventricular portion occupied by the intermediomedial nucleus that does not express any of the four cadherins studied here (IM in Figs. 4D–6D, 9D). The ventricular surface of this tier is marked by particularly strong cad6B expression (Fig. 14A–C).

The largest nucleus of the intermediate tier, the nucleus rotundus, is located ventrolaterally and extends practically along the entire rostrocaudal dimension of the dorsal thalamus. The nucleus rotundus is separated from the zona limitans by the subrotundic nucleus and from the pretectum by perirotundic and retrorotundic areas. All of these surrounding regions are derivatives of the anteroventral region (see below). The nucleus rotundus itself (R) contains posterior, parafascicular, posterolateral, posteromedial, intermediate, anterolateral, and anteromedial subregions (Rp, Rpf, Rpl, Rpm, Ri, Ral, and Ram, respectively, in Figs. 3-5, 9A, 11B,D, 12, 13, 16; see Martínez-de-la-Torre et al., 1990) that differentially express all four cadherins. At its ventrolateral border, the nucleus rotundus is demarcated (and separated from the brain surface) by migrated anteroventral region derivatives (perirotundic area and ITO; see Fig. 1B). Medially, the nucleus rotundus contacts the nucleus triangularis (T in Figs. 4-6, 7B–D, 9A, 12A–F, 13, 16C,D), a deep domain of the intermediate tier that forms a connection to the periventricular IM nucleus. Both nuclei share a similar profile of cadherin expression.

The intermediate tier approaches the pial surface of the brain only dorsocaudally, behind the dorsal tier, where the SPC and SMC cell populations covering the intermediate tier become thinner. For consistency of terminology and to avoid confusion, we have renamed the intermediate tier area at this position the intermedioposterior nucleus (IP; see Discussion). This nucleus contains distinct medial and lateral subregions (IPm and IPl in Figs. 5, 6, 7B–D, 12, 14) that express several of the cadherins. Cad6B immunoreactivity predominates in IPl, whereas weak cad7 expression is found in IPm, and Rcad appears throughout IP. Some Rcad expression appears throughout IPl, whereas Ncad expression is weak.

Ventral tier.

Like the other tiers, the ventral tier has a distinct periventricular stratum that contains a small cell population, the so-called nucleus paramedianus internus of classic literature (Rendahl, 1924; PM in Figs. 4-6, 7B–D, 9A, 14E). This nucleus weakly expresses cad7 and Rcad. Superficial to it, the nucleus ovoidalis (OV) expresses cad7 strongly but is distinctly negative for Rcad and Ncad (OV in Figs. 5B,C, 9A, 11B). It is surrounded by the following Rcad-positive regions: a rostrally located, dense periovoidal nucleus (POV); a caudal retroovoidal area (ROV); and a lateroventrally adjacent nucleus semilunaris periovoidalis (SPO; Figs. 4-6, 9, 14). Extending radially in an apparently oblique plane from the ovoidal complex toward the brain surface, there is a group of ventral tier nuclei that differentially express all four cadherins: In a deeper position, there is a less well known, faintly Rcad-positive region that we call the posterior nucleus (Po in Figs. 5C, 6C, 7C,D, 14A,D). This nucleus appears to be divided into two superposed strata: a deep stratum that contains larger neurons and an intermediate stratum that contains small cells and is traversed by some fiber fascicles of the occipitomesencephalic tract (om in Figs. 5C, 6C,D, 14D–F).

Near the brain surface, the ventral tier derivatives form an enlarged, ovoid complex that may be confused with the nucleus rotundus due to similarities in general shape and position in transverse sections. In sagittal sections (Fig. 13), this nuclear complex is found clearly ventral to the nucleus rotundus. It consists of the posterior nucleus of the ventral supraoptic decussation (PDSV), the posterointermediate nucleus (PI; or posterior nucleus of the ansa lenticularis), the posteroventral nucleus (PV), and the nucleus zeta (Z; Figs. 5, 6, 7, 13). This nuclear complex expresses Rcad, which distinguishes it from adjacent p1 and p3 regions (Fig. 6C, arrowheads). The PDSV nucleus is traversed longitudinally by the tectothalamic tract. This nucleus distinctly expresses both cad6B and Rcad and is marked by intense AChE activity (Figs. 5A,C, 6A,C, 13A,C,D). It has a thin, tail-like, superficial extension that reaches the brain surface under the optic tract after sorting between the SS, the GV, and the BOR nuclei (not shown; best seen in AChE material, see Martínez-de-la-Torre et al., 1990). The PV nucleus weakly expresses cad7 and Rcad and lies just under the nucleus rotundus, whereas PI protrudes somewhat into the pretectum and weakly expresses cad6B, with some interspersed cad7-positive cells (Fig. 13A,B). An anteromedial subregion of this complex, which is interstitial to the cad7-positive tectoovoidal tract, is called here the nucleus zeta (z in Figs. 5B, 6B, 13B). This nucleus lies at a topologic position similar to that of the nucleus zeta (ζ) in reptiles (Beccari, 1923), warranting the use of the same name in the avian brain.

Anteroventral subregion.

In the chicken, cells originating in the boomerang-shaped anteroventral histogenetic zone of the dorsal thalamus (see above) migrate to the surface, partly filling the area of origin in depth and partly dispersing tangentially caudalward at the surface, to finally largely cover the dorsal, intermediate, and ventral tiers (Fig. 1A,B; Rendahl, 1924; Puelles et al., 1991, Puelles et al., 1999; Yoon et al., 2000). The best known derivatives of this region (nucleus subrotundus, perirotundic area, prospective interstitial nucleus of the optic tract; see Table 2) differentiate in close spatial relation to the nucleus rotundus of the intermediate tier. They are characterized mainly by their expression of Rcad but also show differential expression of the other cadherins (SRt, APR, and ITO in Figs. 3, 4, 12). For example, the prospective interstitial nucleus of the optic tract (ITO) is the only nucleus in this region that shows strong Ncad expression (Fig. 13D). This finding supports its distinction from the perirotundic area (APR), as proposed by Puelles et al. (1991) based on other evidence. Caudodorsally, the anteroventral derivatives extend behind the nucleus rotundus as a nucleus that we call the retrorotundic nucleus (RR in Figs. 5, 7C, 12, 16; see also Yoon et al., 2000).

Ventral thalamus (alar p3).

Rostrally, parts of the ventral thalamus protrude into our rostralmost section levels (p3 in Fig. 2D). The ventrolateral nucleus (VL) and the overlying area triangularis (AT) show only weak or no cad7 expression (Figs. 2B, 3B, 13B). In contrast, the lateroanterior nucleus (LA) is the only structure in the ventral thalamus that strongly expresses cad6B (Figs. 2A, 12A; see also Wöhrn et al., 1998); it is only moderately cad7-positive (Figs. 2B, 12B). There also is practically no Rcad or Ncad expression in the ventral thalamus except for some weak, dispersed Rcad signals at the periventricular zona incerta (ZI in Fig. 3C,D). The ventral geniculate nucleus (GV) strongly expresses cad7, particularly in its superficial neuropil (Figs. 2B, 3B, 4B, 12B, 13B; see also Wöhrn et al., 1998). Medial to GV, the p3 (rostral) portion of the nucleus of the basal optic root is found (BOR in Figs. 4-7). Other cad7-positive populations in p3 include the dorsal part of the superior reticular nucleus (RSd in Fig. 3B) and a deeper reticular subpopulation at the limit of the zona incerta (ZI in Fig. 3B). The intercalate nucleus (ICT) contains some isolated, strongly cad7-immunoreactive neurons (Figs. 3B, 9A,B). On the whole, cad7 is the predominant cadherin expressed in the ventral thalamus (Fig. 11B).

In the rostral section levels, there is a periventricular area situated immediately below the diencephalic roof. This area does not express any of the four cadherins and is separated from the dorsal thalamus by the Rcad-positive glial raphe of the zona limitans intrathalamica (p2/p3 border). We conclude that this area represents the topologically dorsalmost part of ventral thalamus. Consequently, we call it the ventral thalamic dorsal area (VTD in Figs. 3, 7A, 12C,F). Caudal to VTD, the dorsalmost parts of the ventricular surface are occupied by the zona limitans (Fig. 3C, inset) and, even more caudally, by the dorsal thalamus (Figs. 4-7), as expected.

Other divisions.

In general, the divisions of the prosomeric basal and floor plates are less well demarcated by cytoarchitectonic features or by the expression of the four cadherins than those of the alar plate. The basal and floor plate nuclei that can be characterized by their cadherin expression profiles are listed in Table 2. Many of these nuclei, especially those of the hypothalamus (basal p4–p6), are shown in Figures 3-5 and Figures 8-10. In the p4–p6 roof plates, the paraphysis shows regionalized expression of cad7, Rcad, and Ncad (data not shown).

Cadherins as markers for developing gray matter structures

The current study focuses on gray matter architecture at a stage of development when most gray matter structures already have formed and assumed their final topologically invariant position. Taking into account cytoarchitectonic features, radial glial topology, and previously published data, the results from the cadherin-expression mapping allow us to study developing gray matter structures, including the prosomeres and their secondary subdivisions, as well as diencephalic brain nuclei and their regional differentiation with considerable spatial resolution and detail.

In general, the diencephalic regions expressing a given cadherin maintain their expression from E11 to E15. Results from prehatching stages (E19; data not shown) were found to be very similar to those at E15. Thus, cadherins can be used as expression markers for specific gray matter structures. The immunostaining data displayed in the current paper are predominantly from sections through E11 diencephalon, because, at these stages, neurons still are packed more densely, and the borders of the main embryonic subdivisions can be related more easily to radial glial fiber density. At E15 (Figs. 13, 16) and E19 (data not shown), cadherin immunoreactivity has become more restricted to individual diencephalic brain nuclei, as described previously for Rcad (Arndt and Redies, 1996). Consequently, the embryonic subdivisions at E19 are not as clearly visible as they are at E11.

General similarities in spatial extent and overlap of cadherin expression

The expression of many cadherins by neurons in the living brain is combinatorial, both at the regional level (Fushimi et al., 1997; Suzuki et al., 1997) and at the cellular level (Kohmura et al., 1998; Wöhrn et al., 1999). In the current study, many brain regions showed immunoreactivity for more than one of the four cadherins, in some cases also at the cellular level (data not shown). Generally speaking, the number of cad6B-expressing gray matter areas is relatively low, whereas cad7, Rcad, and Ncad are expressed more widely. Many cad6B-positive regions coexpress cad7. There is relatively little overlap between Ncad and Rcad, and the overlap between Ncad and cad7 is only moderate. Similar observations on the extent and overlap of expression were made for the same four cadherins in the visual and cerebellar systems (Arndt and Redies, 1998; Arndt et al., 1998; Wöhrn et al., 1998, Wöhrn, et al., 1999) and in the telencephalon and hindbrain (our unpublished data). It is therefore conceivable that the genetic patterning processes regulating the expression of the four cadherins and the spatial relation of their expression domains are similar in different parts of the brain.

Persistence of neuromeric subdivisions in mature diencephalic gray matter

The importance of early embryonic regionalization and segmentation for understanding vertebrate brain development has been supported by numerous studies during the last decade. What is less well known is how this type of pattern formation relates to the appearance of mature brain architecture. Although seminal work studying this issue was published a long time ago (Rendahl, 1924; Coggeshall, 1964; Vaage, 1969; Keyser, 1972; Vaage, 1973), there have been relatively few modern studies on this topic (Puelles et al., 1991, Puelles et al., 1992; Figdor and Stern, 1993; Puelles, 1995). In the hindbrain, migration of early neurons is predominantly radial after rhombomere formation has taken place (Fraser et al., 1990; Birgbauer and Fraser, 1994). For the hindbrain, it has been shown that, as a consequence of the restricted tangential migration, segmental organization persists and provides the basis for nucleus formation (Vaage, 1973; Puelles and Martínez-de-la-Torre, 1987; Marin and Puelles, 1995; Wingate and Lumsden, 1996).

To our knowledge, the current study is the first comprehensive description of the diencephalic prosomeres and their gray matter derivatives in the chicken. Previous studies in the diencephalon have shown that, like in the hindbrain, some of the major divisional borders restrict tangential migration (Figdor and Stern, 1993; Golden et al., 1997). In addition, morphologic and histochemical studies have revealed that the neuromeric divisions persist as neuromere-derived domains of the mantle layer in the mature diencephalic gray matter of the chicken (Rendahl, 1924; Puelles et al., 1991), like in other vertebrate species (Martínez-de-la-Torre, 1985; Puelles et al., 1992, Puelles et al., 1996; Pombal and Puelles, 1999).

Previous studies have demonstrated that, in chicken, mouse, frog, and zebrafish, the borders of embryonic divisions often coincide with changes in cadherin expression at early stages of brain development. The regional changes in cadherin expression are subserved mostly by neuroepithelial cells and radial glial cells that extend processes all the way to the pial surface (Espeseth et al., 1995; Gänzler and Redies, 1995; Matsunami and Takeichi, 1995; Kimura et al., 1996; Redies and Takeichi, 1996; Inoue et al., 1997; Korematsu and Redies, 1997; Yoon et al., 2000). Thus, at early stages, many cadherins are expressed across the entire thickness of the neural tube wall. As the mantle layer of the diencephalic neural tube increases in thickness and approaches its mature appearance, cadherin-immunoreactive radial glial processes are no longer a prominent feature, although some radial glial fibers still are present in specific regions, as demonstrated in the current study. At the same time, regional differences in cadherin expression appear in the mantle layer (Gänzler and Redies, 1995; Yoon et al., 2000). During subsequent development, the mantle layer increases in thickness and differentiates into the anlagen of brain nuclei. In parallel, differential cadherin expression by brain nuclei begins to be a prominent feature in the mantle layer. Despite these overall changes in the predominant cell types expressing cadherins, the divisional borders remain clearly detectable in the immunostained sections (see, e.g., Figs. 11B,C; 12). However, in contrast to the earlier stages, the divisional borders now coincide with the borders of cadherin-expressing brain nuclei situated adjacent to the borders. Thus, at different levels of mantle layer stratification, several different cadherins may demarcate a given divisional border, thereby increasing the complexity of cadherin expression as morphologic and functional differentiation of diencephalic gray matter takes place (for a schematic diagram of this model of brain development, see Fig. 10 in Redies, 2000).

Changes in radial glial fiber density at divisional borders

Our results demonstrate the presence of radial glial fibers at relatively late stages of diencephalic development (E11 and E15), when the mantle layer has attained to considerable size and approaches its final thickness and morphology. We could not follow individual fibers from the ventricular layer to the pial surface, probably because of the gross morphogenetic distortions that have taken place during development or because we did not cut sections in the exact plane of individual radial glial fibers. Nevertheless, the presence of radial glial fibers at the pial surface suggests that some (if not most) fibers span the entire thickness of the neural tube wall also at this relatively late stage. Radial glial fibers and divisional borders can be expected to be distorted roughly in parallel as they run from the ventricular surface to the pial surface.

Changes in radial glial fiber density at the boundaries of embryonic divisions have been described before, including at the corticostriatal border (Misson et al., 1988; Silver et al., 1993; Stoykova et al., 1997; Mai et al., 1998) and at the rhombomere boundaries (Heyman et al., 1995). Here, we show that such changes are associated more systematically with divisional boundaries than was believed previously, at least in the diencephalon of the chicken. Previous reports demonstrated raphes of radial glial fibers at the borders of some diencephalic brain nuclei (Geisert and Bidanset, 1993; Steindler, 1993; Kalman et al., 1998); however, it is unclear how these correlate with the divisional borders demonstrated in the current study. Some diencephalic glial boundaries shown by Silver et al. (1993) and Mai et al. (1998) are related more clearly to the interprosomeric boundaries underlined in our material.

How can the regional differences in radial glial density be explained? It has been shown that many of the divisional borders constitute borders of morphogenetic fields that display different spatial and temporal profiles of cell proliferation and migration (Rendahl, 1924; Bergquist, 1957; Puelles et al., 1987; Figdor and Stern, 1993). In the hindbrain, the rhombomeric boundaries contain molecularly distinct cell populations with relatively low mitotic activity compared with their surroundings (Guthrie et al., 1991). The presence of radial glial raphe at some of the divisional borders in the diencephalon suggests that these border regions have proliferative properties similar to those of the hindbrain. One way to explain differences in radial glial density is to assume that the overall number of radial glial cells becomes fixed early and changes little during further diencephalic morphogenesis; areas with lower proliferative activity (e.g., border regions) would then show a higher density of radial glial fibers. Alternatively, one could assume that differences in radial glial proliferation and differentiation between the embryonic divisions or between the divisions and their boundary regions account for the differences in radial glial density.

For some subdivisions, we also found considerable differences in the density of radial glial fibers at different levels of stratification. These differences may reflect the number of radially migrated neurons and the extent of their radial migration (Rakic, 1972; for hindbrain, see Marin and Puelles, 1995; Wingate and Lumsden, 1996). For example, in the dorsal thalamus, the density of radial glial fibers is high at the position of the superficially located prospective interstitial nucleus of the optic tract, low within the nucleus rotundus, and intermediate in the periventricular region (Fig. 9C). These changes are related inversely to the thickness of the dorsal thalamus at these different levels of radial stratification. Similarly, the areas in the pretectum containing densely packed neurons (e.g., PPT, SpL, and SpMcl in Fig. 10) display lower densities of radial glial fibers than the gray matter areas of lower cell density immediately surrounding these nuclei.

In conclusion, the current analysis supports the notion that the embryonic divisions of the diencephalon persist in the architecture of the mature diencephalon. In contrast to the hindbrain, in which cells originating in different neuromeres often merge into morphologically relatively homogeneous brain nuclei (Marin and Puelles, 1995), most of the boundaries in the diencephalon remain distinct histologically. With the exception of the nucleus of the basal optic root which seems to extend across several prosomeres (p1–p3; see footnote 3 in Table 2), all other diencephalic nuclei can be assigned to a single diencephalic (sub)division (Table 2, Fig. 18). The two-dimensional scheme of transverse and longitudinal divisions of the neural tube, thus, is transformed rather directly into the three-dimensional architecture of mature diencephalic gray matter. The third dimension emerges during mantle layer development and is represented by the radial stratifications of the neural tube wall.

Application of the prosomeric model to the developing diencephalon

The prosomeric model has been employed successfully as a paradigm in studies on brain development in several vertebrate species (Puelles and Rubenstein, 1993; Ekstrom and Ohlin, 1995; Puelles, 1995; Puelles et al., 1996; Milán and Puelles, 2000; Pombal and Puelles, 1999; Wullimann and Puelles, 1999). In the current work, this model is used to define the primary embryonic divisions in the chicken diencephalon in conjunction with earlier observations on the segmentation of the chicken diencephalon (Rendahl, 1924; Vaage, 1969; Martínez-de-la-Torre, 1985; Puelles et al., 1987, Puelles et al., 1991; Martínez et al., 1991; Puelles and Medina, 1994).

The following predictive properties of the prosomeric model also are applicable to the developing chicken diencephalon: 1) Each of the six prosomeres extends in a radial-glia-coherent fashion from the ventricular surface to the pial surface of the brain; 2) the alar domains forming the pretectum, dorsal thalamus/epithalamus, and ventral thalamus (p1–p3) extend in a coherent fashion from the basal plate to the roof of the diencephalon; in the ventral thalamus, the dorsalmost area (VTD) is relatively small and represents a previously unrecognized region of this diencephalic division [note that the three rostral prosomeres, p4–p6, extend from the basal midline (hypothalamus) into the telencephalic vesicle]; and 3) despite the morphogenetic distortions taking place during diencephalic development, the topologic relations between the different prosomeres remain invariant. For example, the area rostral to the pretectum is always dorsal thalamic territory that, in turn, is always followed rostrally by ventral thalamic territory.

Secondary subdivisions in the pretectum and dorsal thalamus

Rendahl (1924) suggested that the pretectal nuclei of chicken derive from two morphogenetically independent alar regions that were defined by their relation to the posterior commissure (commissural and precommissural pretectal regions). His pioneering data on the development of the spiriform nuclei suggested the existence of a third independent pretectal region, the juxtacommissural pretectum, found intercalated between the other two (see also Martínez-de-la-Torre, 1985; Martínez, 1987; Puelles et al., 1996; De Castro et al., 1998). The current study supports the usefulness of this general scheme for defining subdivisions in the chicken pretectum by mapping cadherin expression and radial glial topology. Evidence for the existence of transverse subdivisions within the pretectum also stems from studies on the early restriction of diencephalic cell intercalation (Figdor and Stern, 1993) and from mapping the expression of axonin-1, an adhesion molecule of the Ig superfamily (Redies et al., 1997), or other molecular markers (Sax-1, Pax-6, Ebf-1; Schubert et al., 1995; Garel et al., 1997). The same tripartite scheme also has been used in previous work on the mammalian, reptilian, amphibian, and lamprey pretectum (Martínez-de-la-Torre, 1985; Caballero-Bleda et al., 1992; Puelles and Medina, 1994; Puelles et al., 1996; Pombal and Puelles, 1999).

A novel, three-tiered subdivision scheme also was adopted for the dorsal thalamus, and it is presented here in detail for the first time, although precursor ideas and additional evidence can be found in Martínez-de-la-Torre's (1985) and Guillén's (1992) doctoral dissertations as well as in recent work on the reptilian dorsal thalamus (Diaz et al., 1994). In an analogy to the secondary subdivisions of the pretectum, the dorsal thalamic subdivisions (dorsal, intermediate, and ventral tiers plus an anteroventral subregion) are defined based on cytoarchitectonic evidence, cadherin expression data, and radial glial topology as well as on previously published indications for such subdivisions (see Rendahl, 1924; Kuhlenbeck, 1937; and other references cited above).

To understand possible developmental implications of our scheme, note that the rostral limit of the dorsal thalamus is formed by the zona limitans intrathalamica. Molecular expression data indicate that this structure coincides with a local dorsal extension of the basal/alar plate boundary into the alar plate (Macdonald et al., 1994; Puelles, 1995; Shimamura et al., 1995). The zona limitans possibly has inductive properties on neighboring diencephalic tissue, perhaps serving as an organizer for local secondary regionalization (for review, see Rubenstein et al., 1998). Note that the postulated anteroventral subregion of the dorsal thalamus represents a narrow stripe of tissue that is immediately adjacent to both the alar/basal plate boundary and the zona limitans (Fig. 18). On the molecular level, this domain clearly correlates with the early expression of the gene Nkx-2.2 at the caudal slope of the zona limitans, and all derivatives of the anteroventral region in the mantle continue to express this gene (Puelles et al., 1999). Anteroventral derivatives of alar p2 also are characterized by their selective RNA expression of the Sox family genes, Sox14 and Sox21, during and after their migration (Uchikawa et al., 1999), by their cad6B expression (Yoon et al., 2000), and by the absence of Gbx-2 and Sox2 expression that is found in the rest of the dorsal thalamus (Redies et al., 1997; Uchikawa et al., 1999).

In sagittal views (Figs. 12, 13, 18), the borders between the three tiers of the dorsal thalamus have an oblique orientation, intermediate between the longitudinal orientation of the alar/basal plate boundary in p2 and the transverse orientation of the pretectothalamic and zona limitans boundaries (p1/p2 and p2/p3 borders, respectively). It remains unclear to which extent this obliquity may be a consequence of the slanting topology of the caudal slope of the zona limitans. A peculiarity related to this obliquity may be that the epithalamic habenular complex is excluded from the caudalmost roof area of p2 (Fig. 18; Martínez-de-la-Torre, 1985; Guillén, 1992). Curiously, the boundaries between the three subregions of the pretectum also are slightly oblique relative to the prosomeric boundaries of p1, as reflected by the occupancy of the whole roof area by the commissural subregion (Fig. 18). The obliquity is one of the reasons why we do not regard these subdivisions as prosomeres by themselves, as proposed by Figdor and Stern (1993). Note that the secondary subdivisions in the ventral thalamus (i.e., see boundary between LA and GV) also seem to be oblique. Pending further analysis, they are not described here in detail.

Together, these data strongly suggest that oblique secondary limits appear at later stages of forebrain regionalization without the disappearance of previously established transverse interprosomeric (primary) boundaries. The oblique secondary limits introduce additional complexity in the dorsoventral and anteroposterior patterning of the neural tube wall. They establish immature intermediate primordia within the thalamic mantle zone. The cells in these primordia seem to share characteristic cellular and molecular properties (migration patterns, radial glial processes, cadherin expression, etc.). Like the primary divisions, the secondary subdivisions of the pretectum and dorsal thalamus each encompasses a complete radial unit. In the terms of Bergquist (1932) and Bergquist and Källén (1954), these radial units represent the structural product of one histogenetic migration area. The subdivision scheme proposed here is to be understood as a model that is a first approximation and will require changes and modifications in the future if additional independent areas are discovered.

The biologic significance of the secondary subdivisions, e.g., in partially restricting tangential cell migration among them (see, e.g., Figdor and Stern, 1993), remains to be established experimentally. Here and in the accompanying paper (Yoon et al., 2000), we have interpreted this aspect in terms of the best available information, suggesting that most tangential migration originates in the anteroventral region and takes place at the surface of the dorsal thalamus, largely respecting its segmental boundaries (Fig. 1A,B; Puelles et al., 1991, Puelles et al., 1999).

Later in development, additional molecular differentiation and probably connectivity-related specification divide the secondary primordia into tertiary subdivisions, leading to the development of brain nuclei, as proposed by Rose (1942) for his nuclear anlagen. Given nuclei still may undergo quaternary subdivisions among their constituent neuronal populations, as demonstrated here in an exemplary fashion by the complex cadherin expression profile of the nucleus rotundus (Fig. 16) and the ovoidal complex (Figs. 4, 5, 9). These subdivisions may correlate with functional specializations within the nuclei, as shown for cadherin expression in other brain regions, e.g., the parasagittal domains of the cerebellar cortex (Arndt et al., 1998).

Novel gray matter subdivisions of dorsal thalamus

By mapping cadherin expression, a number of gray matter areas were studied in more detail than had been done previously, leading to some novel terminology. We concentrate most of our comments below on the dorsal thalamus, in which the nuclear divisions in birds have been studied by Rendahl (1924), Huber and Crosby (1929), Kuhlenbeck (1937; for review, see Kuhlenbeck, 1973), and Puelles et al. (1991). Other more recent additions to our knowledge of distinct dorsal thalamic nuclei stem from hodologic analyses (for review, see Dubbeldam, 1998).