Neurogenic development of the visual areas in the Chinese softshell turtle (Pelodiscus sinensis) and evolutionary implications

Abstract

To characterize the neurogenic development of the visual areas of the turtle (Pelodiscus sinensis) during embryogenesis, a single dose of [³H]-thymidine (10 µCi) was injected into egg yolks from stages S11∼12 to S21. At hatching, localization of [³H]-thymidine incorporation was examined, and led to three main observations. (1) Neurogenesis occurred in the stratum griseum centrale of the tectum opticum from S11∼12 to S16 with a peak at S12. No obvious gradients of neurogenesis were observed. (2) Neurogenesis in the nucleus rotundus (Rot) and in the dorsal lateral geniculate nucleus (GLd) occurred from S11∼12 to S15. Gradients of neurogenesis were detected along ventral–dorsal and lateral–medial axes in the Rot, but only the latter neurogenic gradient occurred in the GLd. (3) In the visual region of the dorsal ventricular ridge, neurogenesis lasted from S11∼12 to S16. Similarly, neurogenesis occurred from S11∼12 to S16∼17 in the dorsal cortex, with a peak at S12 for both telencephalic visual regions. Neurogenesis followed a ventrolateral to dorsomedial gradient in the visual region of the dorsal ventricular ridge, and a superficial to deep gradient in the caudal dorsal cortex. A significant number of neurons in the rostral dorsal cortex followed a deep (earlier arising) to superficial (later arising) pattern of neurogenesis, similar to that in the avian Wulst or in the mammalian isocortex. Finally, we compared the timing and development of neurogenesis in the turtle with birds and mammals to understand the evolutionary implications of these processes.

Introduction

In amniotes, visual information reaches the telencephalon through two ascending pathways. The first, the thalamofugal pathway, connects the retina to the dorsal lateral geniculate nucleus (GLd) in the thalamus, then to the dorsal cortex (DC) in reptiles, the Wulst in birds, and to the primary visual cortex in mammals. The other ascending pathway, the tectofugal pathway, connects the retina with the tectum opticum, which projects to the nucleus rotundus (Rot) in sauropsids (reptiles and birds), the lateral posterior (Lp)/pulvinar nucleus (Pulv) in mammals, and finally to the anterior dorsal ventricular ridge (ADVR) of sauropsids, and the visual extfrastriate cortex in mammals (Hall et al. 1977).

An outstanding evolutionary question is to what degree are the visual areas, especially the telencephalic regions of the two visual pathways, homologous among the three sister taxa of amniotes: the reptiles, birds and mammals? On the basis of similarities in neural circuitry, the Rot is thought to be homologous to the Lp/Pulv, while the ADVR of sauropsids and the DC of reptiles are thought to be homologous to the extrastriate and primary visual cortices of mammals, respectively (Karten, 1969; Hall & Ebner, 1970; Hall, 1972; Shimizu & Karten, 1991; Major et al. 1998, 2000). However, several lines of evidence call for a re-evaluation of these comparisons.

First, embryological data have shown that the ADVR originates from the lateral pallium, whereas most of the isocortex arises from the dorsal pallium in birds (Striedter et al. 1998). Although some interneurons in the isocortex are descending from the lateral pallium via tangential migration, there is no evidence that this is the primary origin of the isocortex. Further support of this hypothesis is provided by analysis of several regulatory homeobox genes whose expression patterns are highly conserved in the embryonic forebrain throughout phyla. For example, Emx, Otx and Pax families of homeobox genes are expressed in the pallium, including the reptilian dorsal cortex, the avian Wulst and the mammalian isocortex, while Dlx1 and Dlx2 are expressed in the subpallium, including the striatum and the pallidum. In contrast, brain areas derived from an intermediate region located between the pallium and the subpallium, including the reptilian or avian ADVR and the mammalian claustroamygdala complex, lack detectable Emx1 or Dlx1 expression, but express Pax6 (Anderson et al. 1997; Fernandez et al. 1998; Puelles et al. 2000; Muzio & Mallamaci, 2003). These data suggest that the ADVR is similar to at least some parts of the claustroamygdala complex of mammals rather than to the extrastriate cortex. However, it has been noted that the aforementioned intermediate region does give rise to the ADVR of sauropsids, but it is generally dwarfed in mammals, and even obliterated from the neuroepithelial surface in later embryonic stages (Aboitiz et al. 2002). Thus, there might not be a strict homologous relationship between the mammalian telencephalon and late-arising components such as the ADVR of sauropsids. Second, a study of the neural circuitry has recently shown that, like the Rot of sauropsids, the posterior thalamic and intralaminar nuclei of mammals also receive collothalamic inputs (Butler, 1994), and project to the lateral amygdala, suggesting that the Rot is more similar to the posterior intralaminar complex than to the Lp/Pulv of mammals (Guirado et al. 2005).

Analysis of data from 131 species of primates, bats and insectivores has led to the proposal that the timing of brain development is constrained at a local as well as at a broader cerebral regions, and the order of neurogenesis has been proposed to be highly conserved across different species, relating to brain size (Finlay & Darlington, 1995). Using the equation R = developmental time at which a given nucleus arises × the ratio of the incubational or gestational periods of the two species examined, Wu et al. (2000) compared the tectorotundo-entopallium pathway in the chick with those of the colliculo-Lp/Pulv extrastriate cortex pathway in the monkey, and found that they were closely related and possibly homologous. Embryological studies on the areas in the two visual pathways may be helpful for understanding their organization and evolution. Yet, to our knowledge, detailed studies of embryonic neurogenesis in the tectofugal and thalamofugal pathways are still lacking in the reptile. To address this issue, we injected a small dose of [³H]-thymidine into turtle embryos at serial developmental stages, with the aim to identify the developmental timing (the period of neurogenesis) and neurogenic patterns of these two visual pathways. We then compared these results with those reported for birds and mammals.

The use of [³H]-thymidine autoradiography is based on the principle that proliferating cells that incorporate [³H]-thymidine during the DNA synthesis (S) phase of mitosis will retain label if no subsequent cell division occurs, whereas the label will become diluted if further cell division continues. Cells arising before [³H]-thymidine injection, however, will not be labelled as they will not incorporate [³H]-thymidine into their DNA.

Abbreviations used in the text and figures

A, anterior nucleus; ADVR, anterior dorsal ventricular ridge; ADVRV, visual region of the anterior dorsal ventricular ridge; Cb, cerebellum; DC, dorsal cortex; DLA, dorsolateral anterior nucleus; DMA, dorsomedial anterior nucleus; DVRd, dorsomedial part of the visual area of the anterior dorsal ventricular ridge; DVRv, ventrolateral part of the visual area of the anterior dorsal ventricular ridge; E, embryo; GLd, dorsal lateral geniculate nucleus; H, habenula; LC, lateral cortex; Lp, lteral posterior nucleus; MC, medial cortex; PHLN, proportion of heavily labelled neurons; Pulv, pulvinar nucleus; Rot, nucleus rotundus; Re, nucleus reuniens; S, stage; SGC, stratum griseum centrale of the TO; SGFS, stratum griseum et fibrosum superficiale of the TO; SGP, stratum griseum periventriculare of the TO; SO, stratum opticum of the TO; STR, striatum; TO, tectum opticum; TS, torus semicircularis; V, ventricle; VZ: ventricle zone.

Materials and methods

Animals

Chinese softshell turtle (Pelodiscus sinensis) eggs were incubated at a constant temperature (30 °C) and relative humidity (90%). Under these conditions, hatching generally occurred within 52 days of egg laying. The day of egg laying was defined as embryonic day 1 (E1). Eggs were classified into 13 age groups (Table 1), ranging from E8 to E20 [approximately corresponding to embryonic stages (S) 11∼12 to 19∼20 of Pelodiscus sinensis (Yntema, 1968)]. All experimental procedures adhered to the guidelines of the Beijing Animal Administration Committee.

Table 1.

Developmental stages of Chelydra serpentina

Age (days)	Corresponding developmental stage at 30 °C*
E8	11∼12
E9	12
E10	13
E11	13∼14
E12	14∼15
E13	15
E14	15∼16
E15	16
E16	16∼17
E17	17
E18	18
E19	19
E20	19∼20

^*Yntema (1968).

[³H]-thymidine autoradiography

[³H]-thymidine (6.7 Ci/mm; Amersham Pharmacia Biotech, UK Ltd) was dissolved in an aqueous solution to a final dosage of 1.0 mCi mL⁻¹. The [³H]-thymidine solution was injected into egg yolks for a dose of 10 µCi between 10:00 and 11:00 h on the scheduled day. Each group included 5–6 eggs. Our [³H]-thymidine injection schedule covered the primary time when neurons generated in the visual areas of telencephalon (Goffinet et al. 1986; our unpublished data).

A small hole was drilled in the eggshell before [³H]-thymidine injection. After injection, the hole was sealed with a piece of sterile parafilm. Eggs were incubated until hatching, and the hatch rate was 92%, which is slightly lower than untreated eggs (∼95%), probably because of an increased rate of infection. After hatching, heads or dissected brains were fixed in Bouin's solution for 24 h, and then immersed in 30% sucrose at 4 °C overnight. Samples were then embedded in Jung compound (Leica) and sectioned on a freezing microtome (Leica) into 10-µm sections. Every tenth frontal section was mounted on a gelatin-coated slide. For each brain, a total of six sets of sections were collected, one of which was processed for [³H]-thymidine autoradiography.

Detailed autoradiographic procedures were as described elsewhere (Alvarez-Buylla et al. 1988). In short, emulsion-covered slides were exposed for about 5 weeks in lightproof boxes at 4 °C (Kodak NTB-2 emulsion). Slides were developed in D19, and then counterstained with cresyl violet.

Nomenclature, cell counting and statistics

The nomenclature used in this study was established by Belekhova et al. (2003) for the tectum opticum and thalamic nuclei, and by Balaban & Ulinski (1981) for the visual areas in the dorsal ventricular ridge and in the visual dorsal cortex. In these reports, visual areas were defined based mainly on cytoarchitecture and neural circuitry.

Only cells with clear neuronal morphology, such as darkly cresyl violet-stained nucleoli, and relatively large cytoplasms, were regarded as neurons. Background grain counts were determined for each specimen. A neuron was considered labelled if it had more than 10 grains overlying its nucleus. This cutoff was more than 20 times background levels used previously (Tsai et al. 1981a,b; Yurkewicz et al. 1981). [³H]-thymidine labelled cells could be divided into three groups: heavily labelled cells (grains too dense to be distinguished, > 50 grains), medium-labelled cells (distinguishable grains, 20–50 grains) and lightly labelled cells (sparse grains, 11–20 grains). Heavily labelled cells are generally believed to result from a lack of cell division following [³H]-thymidine labelling. Thus, the birth date of a cell can be estimated to be concurrent with the injection time that results in its heavy labelling. We thus restricted our analyses to heavily labelled cells.

To detect gradients of neurogenesis, for example along rostral–caudal, lateral–medial or ventral–dorsal axes, each visual area was first divided into three similarly sized regions through two coronal brain levels (rostral and caudal). At each coronal brain level, regions were further divided into lateral and medial or superficial (dorsal) and deep (ventral) regions (Fig. 1). For each region, [³H]-thymidine labelling was examined in three adjacent sections (70-µm intervals) at each coronal brain level. Thus, six sections were analysed in a brain hemisphere, comprising more than half of the total sections collected (60–90%). In each region, labelled, unlabelled, heavily labelled and total neurons were counted for each hemisphere. These values were tabulated as a proportion of heavily labelled neurons (PHLN, the number of heavily labelled neurons/the total number of neurons × 100%). For each animal, PHLN values for three adjacent sections at rostral or caudal levels were averaged to calculate a mean value (mPHLN) ± SEM. Damaged sections or failed autoradiographs were excluded from the statistical analysis. Three to six turtle brains were analysed in each of the studied groups. The percentages of labelled neurons were compared among the different brain parts (first factor), i.e. rostral/caudal, lateral/medial or superficial/deep (dorsal/ventral) regions, and among embryonic stages (second factor) by two-way anovausing SPSS software for Windows 11.0. Data analysed by two-way anovawere confirmed to be homogeneously distributed. An α level of 0.05 was considered statistically significant.

Fig. 1.

(A–D) Schematic diagrams illustrating representative transverse sections containing visual regions in the turtle brain (Pelodiscus sinensis), including the tectum opticum (A), the diencephalic visual nuclei (B), and the visual area in the anterior dorsal ventricular ridge and the dorsal cortex (B and C). (D) Parasagittal view, approximately 500 mm lateral to the midline, showing the stereotaxic locations of the transverse planes in A–C (shown by vertical lines). (A) Dashed line divides the stratum griseum centrale (SGC) into lateral and medial halves. (B) Dashed lines divide the nucleus rotundus (Rot) into dorsal and ventral portions and the dorsal lateral geniculate nucleus (GLd) into lateral and medial parts; broken linesdivides the Rot into medial and lateral halves. (C) The dorsal cortex (DC) is divided into superficial and deep parts by dashed lines. The visual region of the anterior dorsal ventricular ridge (ADVRV) is divided into dorsomedial and ventrolateral portions by dashed lines. The orientation for A–D is shown in A. (E–H) Nissl staining of the visual nuclei and areas that we detected. The boxed areas in E–H are used in subsequent figures (Figs 2, 3, 4, 5 and 6). Scale bar represents 0.5 mm for A–C, 0.2 mm for E and F, and 1.0 mm for D. D: dorsal. L: lateral. For abbreviations, see text.

Results

For each of studied visual areas, we first examined the embryonic period during which neurogenesis occurs, as well as the period when a peak in neurogenesis occurs. Next, we examined whether gradients in neurogenesis could be detected. The results for the studied visual areas appear below in the order in which they appear from the mesencephalon to the telencephalon.

Tectum opticum

Heavily labelled cells appeared in the s tratum griseum centrale of the tectum opticum (SGC) of the S11∼12 to S16 groups, with a peak detected in the S12 group at both rostral and caudal brain levels. Although some labelled cells could be seen after the S16 stage, they appeared to be glial cells (not shown), as they were characterized by small size, irregular shape, scant cytoplasm and random distribution (Figs 2, 7A). Two-way anovaindicated a significant change in mPHLN in the rostral/caudal (F = 133.21, P < 0.001), or lateral/medial SGC (F = 153.19, P < 0.001) from S11∼12 to S20. There were no significant changes in mPHLN between the rostral/caudal (F = 4.114, P = 0.054), or the lateral/medial portions of the SGC (F = 0.093, P = 0.911) from S11∼12 to S20.

Fig. 2.

(A–D) [³H]-thymidine labelling in the boxed area of stratum griseum centrale (SGC) shown in Fig. 1E, following a single injection of [³H]-thymidine into embryos at S12 (A: medial and B: lateral), and S15 (C: rostral and D: caudal). A and B are at the same transverse level. Some heavily labelled cells are marked with arrowheads and some moderately or lightly labelled cells are indicated by arrows. Asterisks indicate unlabelled cells. Lateral is right and dorsal is up. Scale bar, 50 µm.

Fig. 7.

(A–E) Proportions of heavily labelled neurons during embryogenesis in the SGC (A), Rot (B), GLd (C), visual area in the ADVR (D), and rostral (E) and caudal (F) DC. SGCr: the rostral part of the SGC. SGCc: the caudal potion of the SGC. Rotv: the ventral half of the Rot. Rotd: the dorsal half of the Rot. Rotl: the lateral part of the Rot. Rotm: the medial part of the Rot. GLdl: the lateral potion of the GLd. GLdm: the medial potion of the GLd. DVRd: the dorsal part of the visual area in the ADVR. DVRv: the ventral part of the visual area in the ADVR. DCrs: the superficial layer of the rostral DC. DCrd: the deep layer of the rostral DC. DCcs: the superficial layer of the caudal DC. DCcd: the deep layer of the caudal DC.

Nucleus rotundus

At rostral and caudal brain levels, heavily labelled cells were observed among the S11∼12 to S15 groups, with a peak observed in the S12 group. In the S13 group, cell division had ceased in the ventral and lateral regions, but continued in the dorsal and medial portions. After the S15 stage, no heavily labelled cells were observed in any region of the Rot (Figs 3, 7B). The mPHLN changed significantly over time in the rostral and caudal (F = 272.78, P < 0.001), ventral/dorsal (F = 158.98, P < 0.001) and lateral/medial (F = 112.41, P < 0.001) regions. Significant differences in mPHLN were not detected between the rostral and caudal brain levels (F = 0.608, P = 0.445), but were detected between ventral and dorsal (F = 4.305, P = 0.047) and between lateral and medial (F = 26.967, P < 0.001) regions of the Rot (Fig. 8A and B).

Fig. 3.

(A–H) [³H]-thymidine labelling in the boxed area of the nucleus rotundus (Rot) shown in Fig. 1F following a single injection of [³H]-thymidine into embryos at S12 (A: medial, B: lateral, C and D: rostral and caudal), and S13 (E and F: dorsal and ventral, G and H: medial and lateral). Broken lines indicate the boundary of the Rot. Some heavily labelled cells are marked with arrowheads and some moderately or lightly labelled cells are indicated by arrows. Asterisks indicate unlabelled cells. A, B, E, F, G and H are approximately at the same transverse level. Lateral is right and dorsal is up. Scale bar represents 50 µm. For abbreviations, see text.

Fig. 8.

Summary figures of the proposed neurogenetic gradients in the Rot (A,B), the GLd (A,B), the caudal DC (C), the rostral DC (D) and the visual area in the ADVR (E). A–E are higher magnifications of the corresponding regions in Fig. 1B and C. A and B: arrows indicate the lateral to medial and ventral to dorsal neurogenetic gradients in the Rot, and show the lateral to medial gradient of neurogenesis in the GLd. (C–E) Arrows indicate the superficial to deep neurogenetic gradients in the caudal DC (C), the deep to superficial neurogenetic gradient in the rostral DC (D), and the ventrolateral-to-dorsomedial neurogenetic gradient in the visual area of the ADVR.

Dorsal lateral geniculate nucleus

Heavily labelled cells appeared in the S11∼12 to S15 groups, with a peak in the S12 group. In the S13∼14 groups, almost no labelled cells were detected in the lateral GLd, although a significant number of heavily labelled cells were detected in the medial GLd (Figs 4 and 7C). The mPHLN changed significantly over time in rostral and caudal (F = 34.16, P < 0.001), and lateral and medial (F = 34.164, P < 0.001) portions of the GLd. The mPHLN observed at the level of the rostral brain showed no significant difference from that observed at the caudal brain level (F = 0.385, P = 0.544). However, a significant difference in mPHLN was evident between the lateral and medial halves of the GLd (F = 6.501, P = 0.017; Fig. 8B).

Fig. 4.

(A–D) [³H]-thymidine labelling in the boxed area of the dorsal lateral geniculate nucleus (GLd) shown in Fig. 1F, following a single injection of [³H]-thymidine into embryos at S11∼12 (A), S12 (B), S13 (C) and S15 (D). Broken lines mark the boundary between the Rot and the GLd. Some heavily labelled cells are marked with arrowheads and some moderately or lightly labelled cells are indicated by arrows. Asterisks indicate unlabelled cells. A, B, C and D are approximately at the same transverse level. Lateral is right and dorsal is up. Scale bar represents 50 µm. For abbreviations, see text.

Dorsal ventricular ridge

Heavily labelled cells were detected in ventrolateral regions of the ADVR in the S11∼12 to S16 groups, with a peak in the S12 group. A peak was also observed in the dorsomedial region of the ADVR in the S13 group. Two-way anovaconfirmed that changes in mPHLN in the rostral and caudal (F = 28.16, P < 0.001), and ventral and dorsal (F = 58.01, P < 0.001) regions were significant. A rostral to caudal neurogenic gradient was not observed (F = 0.254, P = 0.619), but a ventrolateral to dorsomedial neurogenic gradient of mPHLN was evident (F = 4.449, P = 0.039; Figs 5, 7D and 8E).

Fig. 5.

(A–D) [³H]-thymidine labelling in the boxed area of the anterior dorsal ventricular ridge (ADVR) shown in Fig. 1G following a single injection of [³H]-thymidine into embryos at S11∼12 (A: dorsal, B: ventral), and S13 (C: dorsal and D: ventral). Broken lines mark the boundary of the DVRV. Arrowheads mark some heavily labelled cells and arrows indicate some moderately or lightly labelled cells. Unlabelled cells indicated by asterisks. A–D are approximately at the same transverse level. Lateral is right and dorsal is up. Scale bar represents 50 µm. For abbreviations, see text.

Dorsal cortex

Heavily labelled cells were detected from the S11∼12 to S16∼17 groups. A peak in neurogenesis in different portions of the DC was observed among the S12 group. After S16∼17, almost no heavily labelled cells were detected. The mPHLN was found to change significantly in the rostral and caudal (F = 38.19, P < 0.001) and the superficial and deep (F = 60.97, P < 0.001) regions of the DC over time. In the rostral DC, the mPHLN in the superficial layer was lower than that in the deep layer at S11∼12 and S12. After the S12 stage, however, the mPHLN in the superficial layer was higher than that in the deep layer (Figs 6 and 7E). A deep to superficial neurogenic gradient of mPHLN was evident (F = 4.223, P = 0.04) in the rostral DC (Fig. 8D). In contrast, in the caudal DC, the mPHLN in the superficial layer was higher than that in the deep layer in the S12 and S13 groups, but lower than that in the deep layer after the S13∼14 stages. A significant change in mPHLN between the superficial and deep parts of the caudal DC was also noted (F = 8.418, P = 0.012; Figs 6, 7F and 8C).

Fig. 6.

(A–H) [³H]-thymidine labelling in the boxed area of the dorsal cortex (DC) shown in Fig. 1H, following a single injection of [³H]-thymidine into embryos at S12 [rostral: A (superficial) and B (deep); caudal: E (superficial) and F (deep)] and S15 [rostral: C (superficial) and D (deep); caudal: G (superficial) and H (deep)]. Arrowheads indicate some heavily labelled cells and arrows indicate some moderately or lightly labelled cells. Asterisks indicate unlabelled cells. A and B, C and D, E and F, and G and H are at the same transverse level. Lateral is right and dorsal is up. Scale bar represents 50 µm. For abbreviations, see text.

Discussion

Comparison with previous reports

Tectum opticum

Our observations suggested that SGC cells are generated between S11∼12 and S16. As for the chick, SGC neurons in the turtle were the earliest born cells in the optic tectum (Lavail & Cowan, 1971). However, unlike the chick (Lavail & Cowan, 1971; Wu et al, 2000), we did not find evidence of a gradient of neurogenesis within the SGC layer along the superficial to deep (outside to inside) axis, and neurogenesis in the SGC and the stratum griseum et fibrosum superficiale (SGFS) in the turtle lasted the same period of time (unpublished observation). Similarly, using cresyl violet staining, Hergueta et al. (1993) showed that SGC and SGFS both arise during tectal migration I, which is earlier than other layers of the tectum opticum that arise during migration II. Our observations are consistent with the lack of clear laminar gradients of neurogenesis observed in the rhesus monkey (Cooper & Rakic, 1981), mouse (Delong & Sidman, 1962) and rat (Altman & Bayer, 1981). These differences could be ascribed to differences in the optic tectum among different amniote species. For example, the optic tectum in the chick includes at least 14 easily identified laminae, whereas laminae are more difficult to identify in the superior colliculus of the rat, but may contain around half the number of that in the chick. Although the turtle and chick are both sauropsids, they show obvious differences in the organization of tectothalamic projections. In the chick, tectorotundal projection neurons are restricted to the SGC. By contrast, besides SGC neurons, some neurons in the stratum griseum periventriculare (SGP) and the SGFS also project to the Rot in the turtle (Belekhova et al. 2003). In addition, the SGC in the turtle seems to contain some types of somata generally observed in the avian stratum griseum et fibrosum superficiale (Baez et al. 2003). Thus, compared with the avian SGC, the reptilian SGC is considered to be incompletely differentiated from the SGFS. These differences may be due to divergence during amniote evolution.

Diencephalic visual areas

Our observations suggest that there is a ventrolateral to dorsomedial gradient of neurogenesis in the Rot of the turtle, consistent with previous reports from studies in the chick (Wu et al. 2000). Although no reports are available concerning neurogenic gradients in the GLd in birds, our observation that neurogenesis in the GLd is followed by an obvious lateral to medial gradient of neurogenesis is consistent with reports from studies on the rat (Altman & Bayer, 1989). In the rat Lp, neurogenesis follows a lateral to medial gradient (Altman & Bayer, 1989). Thus, neurogenic patterns are not common between the GLd and the posterior nucleus or between the GLd and the intralamina nuclei (Altman & Bayer, 1989).

Telencephalic visual areas

Our study suggests that in the visual region of the ADVR, neurogenesis follows a neurogenic gradient along the superficial to deep (outside to inside) axis, consistent with reports from other reptiles (Goffinet et al. 1986) and birds (Tsai et al. 1981a,b). As shown in the present study, neurogenesis in the visual region of the ADVR also differed between the ventral and dorsal portions in these studies. According to an electrophysiological study of the iguana, the ventral and dorsal portions of the visual areas of the ADVR also differed in physiological function (Manger et al. 2002), implying that the reptilian ADVR is not a homologous area.

Goffinet et al. (1986) showed that neurogenesis in the lateral, medial and dorsal cortices of turtle all follow a gradient along the outside to inside axis, which differs markedly from the mammalian isocortex, in which they follow an inside–out pattern of neurogenesis. In contrast, however, our study suggests that, although neurogenesis in lateral and medial cortices occurs in an outside–in direction, this was not the case in the dorsal cortex. In the dorsal cortex, many later arising neurons in the rostral dorsal cortex were identified in the superficial sublayer, suggesting that later arising cells migrated to their destinations through the older ones. Our study thus suggests that, as in the mammalian isocortex, a significant number of neurons in the reptilian rostral dorsal cortex also follow an inside–out gradient of neurogenesis. This proposal is consistent with a study by Striedter & Keefer (2000) on the chick. According to their report, the avian Wulst, an area homologous to the reptilian dorsal cortex (Fernandez et al. 1998), was also not generated in a strict outside–in fashion, and some Wulst neurons followed an inside–out pattern of neurogenesis. Taken together, these observations suggest that an inside–out mode of neurogenesis is not unique to the mammalian isocortex. Importantly, although neurogenesis in the mammalian cortical plate (layers II and VI) follows an inside–out, earlier to later programme, neurogenesis in the mammalian preplate (layers I and VII or subplate) follows an outside–in, earlier to later programme (Bayer & Altman, 1990). These data are interesting from a comparative embryological viewpoint and are discussed further below.

Evolutionary implications

Subtelencephalon

It is believed that the optic tectum of sauropsids and the superior colliculus of mammals, the Rot of sauropsids and the Lp/Pulv of mammals, and the GLd across amniotes are homologous structures (Foster & Hall, 1975). Based on neural circuitry, Guirado et al. (2005) recently showed that, like the Rot of sauropsids, the posterior thalamic and intralaminar nuclei of mammals also receive collothalamic input and project to the lateral amygdala. The Rot has been proposed to be more similar to the posterior intralaminar complex than to the Lp/Pulv of mammals (Guirado et al. 2005). Our observations suggest that the Rot or the Gld of turtle exhibit a similar timing of neurogenesis. In the rat and monkey, the Lp/ Pulv and the GLd are generated during similar developmental stages (Ogren & Rakic, 1981; Altman & Bayer, 1989). However, neurons in intralaminar nuclei of the rat are generated later than those in the GLd or the Lp (Altman & Bayer, 1979). In addition, the Rot in sauropsids and the Lp/Pulv in mammals are thought to arise from the collothalamic matrix, whereas the GLd and the intralaminar nuclear group are thought to derive from the lemnothalamic matrix (Butler, 1994). The Rot in sauropsids is thus more homologous to the Lp/Pulv than to the posterior intralaminar complex in mammals.

Telencephalon

The sauropsid DVR has been regarded as the homologue of some regions of the mammalian isocortex (Reiner, 1993; Butler, 1994; Aboitiz, 1995, 1999), and several non-cortical structures, such as the lateral amygdala (Bruce & Neary, 1995), the claustrum (Filimonoff, 1964) and the endopiriform nucleus (Striedter, 1997). The mammalian neocortex was previously reported to follow an inside–out, earlier–later gradient of neurogenesis (Rakic, 1975; Bruckner et al. 1976; Luskin & Shatz, 1985; Bayer & Altman, 1990). However, among structures arising from the mammalian pallio-subpallial boundary, some of these, including parts of the amygdala, follow an outside–in, earlier–later neurogenic programme, while others, including the piriform cortex and endopiriform claustrum, follow an inside–out, earlier–later programme (Ten Donkelaar et al. 1979; McConnell & Angevine, 1983; Kordower et al. 1992; Sanderson & Wilson, 1997). Interestingly, areas with different neurogenic programmes also differ in the expression of conserved homeobox genes (Anderson et al. 1997; Smith-Fernández et al. 1998; Puelles et al. 1999, 2000; Muzio & Mallamaci, 2003).

Like the mammalian neocortex, the reptilian dorsal cortex and avian Wulst have been reported to follow an inside–out gradient of neurogenesis at least partly (Striedter & Keefer, 2000; our unpublished data), whereas the sauropsid DVR, including areas receiving visual, auditory and somato-sensory projections, like some parts of mammalian amygdala, follows an outside–in gradient of neurogenesis (Tsai et al. 1981a; Goffinet et al. 1986; Zeng et al. 2007; our unpublished data). Thus, from a comparative embryology viewpoint, neurogenic modes support the proposal of homology between the reptilian dorsal cortex, avian Wulst and the mammalian neocortex, as well as the proposal of homology between the sauropsid DVR and some parts of the mammalian amygdala exhibiting outside–in gradients of neurogenesis. The latter proposal is consistent with detailed studies of the neural circuits (Turner & Herkenham, 1991;Bruce & Neary, 1995; Doron & Ledoux, 1999).

Although the telencephalic pallium of mammals follows an inside–out neurogenic gradient, there are prominent exceptions. For example, like the reptilian DC, the mammalian dentate gyrus of the hippocampal complex and layer I and subplate (layer VII or VIb) zones of the neocortex follow an outside–in neurogenic gradient (Bayer, 1980; Bayer & Altman, 1990). These data suggest that the mammalian archicortex, palaeocortex or layer I, and subplate (layer VII) zones of the neocortex can be regarded as primitive regions similar to the reptilian cerebral cortex (Marin-Padilla, 1998; Super et al. 1998), supporting the ‘dual-origin’ theory of the mammalian cerebral cortex (Marin-Padilla, 1971; Reiner, 1993).

Phylogenic comparison of developmental timing of visual areas across amniotes

Using the equation, R = developmental timing of a given nucleus × the ratio of the gestational periods of the two compared species, Wu et al. (2000) found evidence of similarities in the development of the tectofugal system between the chick and monkey. We propose that, if the two compared areas are homologous, their relative developmental timing should be close, and R should approach 1.0. We calculated R values in the visual areas of turtle, chick and rhesus monkey (Table 2).

Table 2.

Ratio (R) of development time of the visual regions in several amniote species in comparison with the chick

	Chick	Turtle		Monkey
Region	Development time	Development time	1/R	Development time	1/R
SGC	3–5.5*	8–15	0.96–0.93	SC: 30–56†	0.83–0.81
Rot	3.5–5‡	8–13	1.1–0.98	Pulv: 36–45§	2.9–2.8
GLd	3.5–5	8–13	1.1–0.98	GLd: 36–45¶	0.83–0.92
DVR	4–5**	8–15	1.1–0.8	VC: 45–102††	0.73–0.4
				VC(IV): 70–80††	0.47–0.52
				Amyg: 33–50‡‡	1.0–0.83
DC/Wulst	5–10**	8–16	1.4–1.5	VC: 45–102	1.03–0.81

1/R = actual developmental time for a compared nucleus/presumed time in relation to the chick (= birthdate of a given nucleus in the chick × the proportion of incubation period between a compared species and the chick). The incubation period of the chick is 20 days. The incubation period of the turtle is 52 days. The gestation period of the monkey is 165 days.

^*Lavail & Cowan (1971);

^†Delong & Sidman (1962);

^‡Wu et al. (2000);

^§Ogren & Rakic (1981);

^¶Rakic (1977);

^**Tsai et al. (1981a);

^††Rakic (1975).

Although this analysis is an approximation, it leads to two intriguing observations. (1) R values for the compared visual areas (except for the Pulv of the rhesus monkey) were close to 1.0, especially for the turtle, suggesting that the timing of development of the visual areas is conserved among turtle, chick and rhesus monkey, while the Pulv might be evolutionarily distant from the Rot. (2) The R value for the amygdala complex in the rhesus monkey was closer to 1.0 than that of the visual isocortex as a whole or than that of layer IV of the visual isocortex, and the DVR of reptiles or birds might be more closely related to the amygdala complex of mammals than is the visual cortex.