Clones in the Chick Diencephalon Contain Multiple Cell Types and Siblings Are Widely Dispersed

Development 122, 65-78 (1996) 65 Printed in Great Britain © The Company of Biologists Limited 1996 DEV8292

Clones in the chick diencephalon contain multiple cell types and siblings are widely dispersed

Jeffrey A. Golden1,2 and Constance L. Cepko1,3 1Department of Genetics, Harvard Medical School, 2Department of Pathology, Brigham and Women’s Hospital, and 3Howard Hughes Medical Institute, 200 Longwood Avenue, Boston, MA 02115, USA

SUMMARY

The thalamus, hypothalamus and epithalamus of the ver- clones dispersed in all directions, resulting in sibling cells tebrate central nervous system are derived from the populating multiple nuclei within the diencephalon. In embryonic diencephalon. These regions of the nervous addition, several distinctive patterns of dispersion were system function as major relays between the telencephalon observed. These included clones with siblings distributed and more caudal regions of the brain. Early in develop- bilaterally across the third ventricle, clones that originated ment, the diencephalon morphologically comprises distinct in the lateral ventricle, clones that crossed neuromeric units known as neuromeres or prosomeres. As development boundaries, and clones that crossed major boundaries of proceeds, multiple nuclei, the functional and anatomical the developing nervous system, such as the diencephalon units of the diencephalon, derive from the neuromeres. It and mesencephalon. These ﬁndings demonstrate that prog- was of interest to determine whether progenitors in the enitor cells in the diencephalon are multipotent and that diencephalon give rise to daughters that cross nuclear or their daughters can become widely dispersed. neuromeric boundaries. To this end, a highly complex retroviral library was used to infect diencephalic progenitors. Retrovirally marked clones were found to contain Key words: cell lineage, central nervous system, diencephalon, neurons, glia and occasionally radial glia. The majority of thalamus, chick, thalamus, hypothalamus

INTRODUCTION of the diencephalon resembles the more caudal mesencephalon (midbrain) and rhombencephalon (hindbrain), which are One of the hallmarks of the adult nervous system is the parceled into discrete nuclei with mostly discrete projections. exquisite complexity of its cell types and synaptic connections. How the intermediate structure of the diencephalon develops The mechanisms that generate complexity during development is not clear. The development of some of the more rostral and remain largely unknown. Early in development, the neural tube caudal regions of the brain have been relatively well described is thought to be parceled into unique units or segments such that we can now appreciate that these areas follow (Lumsden, 1990; Puelles and Rubenstein, 1993; Rubenstein et different rules in key aspects of their development. al., 1994; Rubenstein and Puelles, 1994). Distinct morphologi- The rhombencephalon is transiently parceled into 8 mor- cal units first appear shortly before neural tube closure when phologically identifiable units known as rhombomeres. The multiple vesicular outpouchings develop rostrally. The primary neurons in each rhombomere comprise groups of nuclei, each vesicles are termed the prosencephalon, the mesencephalon with defined functions and a stereotypical pattern of axonal and the rhombencephalon. The prosencephalon gives rise to projections (Lumsden and Keynes, 1989; Keynes and the telencephalon (cerebral hemispheres) and the dien- Lumsden, 1990; Lumsden, 1990). A molecular basis for the cephalon. The diencephalon is the embryonic precursor to the morphological and functional specification of rhombomeres hypothalamus, thalamus and epithalamus, and is anatomically has been established through studies of the expression and mis- situated between the cerebral hemispheres and more caudal expression of Hox genes and other transcription factors areas of the brain. The diencephalon appears to be conserved (Guthrie and Lumsden, 1991, 1992; Hunt et al., 1991; Guthrie structurally and functionally throughout evolution, although et al., 1992). Lineage analysis conducted in the hindbrain of there is some debate about whether the organization within the chicks has shown that once the boundaries of the rhombomeres diencephalon, and the thalamus in particular, is similar across are established, the majority of clones appear to be restricted species (Kappers et al., 1960; Jones, 1985). The embryologi- to a single rhombomere (Fraser et al., 1990; Birgbauer and cal origin of the diencephalon appears to be conserved in Fraser, 1994) during the next 48 hours of development. Clones disparate species (Kappers et al., 1960). also appear restricted in cell fate in that most clones comprise While functional aspects of the diencephalon appear to siblings that adopt the same or a related neuronal cell fate mimic the rostral telencephalon, the anatomical organization (Lumsden et al., 1994). 66 J. A. Golden and C. L. Cepko

The telencephalon, the most rostral part of the central neuromeric units (Bulfone et al., 1993a,b; Rubenstein et al., nervous system (CNS), is not separated into morphologically 1994; Rubenstein and Puelles, 1994), similar to the correlation defined, repeated units. Rather, the cerebral cortex, the largest noted in the hindbrain. Short-term lineal relationships in the component of the telencephalon, is organized into a diffuse diencephalon have been analyzed using the method of single laminated sheet of cells. Although the cortex can be parceled cell microinjection of a fluorescent dye (Figdor and Stern, into functional domains, there is no discrete nuclear organiz- 1993). The results indicated that the morphological neuromeres ation as is found in the hindbrain and diencephalon. Studies of the diencephalon are akin to the hindbrain rhombomeres in investigating the expression pattern of a variety of genes, that clones were restricted to a single neuromere up to 48 hours mostly transcription factors, have uncovered several genes after injection, the latest time point analyzed. However, this with nested (Simeone et al., 1992; Bulfone et al., 1993a,b) or technique precludes analysis of the final patterns of clonal dis- lamina-specific (Frantz et al., 1994a,b; Leifer et al., 1994) persion or the mature cell types within any one clone. patterns of expression. However, genes investigated thus far Chick/quail chimera studies have also been performed in the are generally not expressed within morphologically or physio- diencephalon (Martinez and Alvarado-Mallart, 1989). In these logically defined areas. Lineage analysis in the telencephalon studies, quail mesencephalon was transplanted into chick dien- has demonstrated that siblings spread over great distances to cephalon, and thus the potential of diencephalic progenitors give rise to neurons in functionally and anatomically unrelated was not tested. Nonetheless, these studies indicated that mes- parts of the cerebral cortex (Walsh and Cepko, 1992, 1993; encephalic progenitors dispersed to populate several, but not Reid et al., 1995). Lineage analysis in the telencephalon has all, nuclei in the diencephalon. One interesting pattern of shown that individual progenitor cells are capable of generat- spread for the mesencephalic progenitors in these studies was ing neurons and glia (Price and Thurlow, 1988; Walsh and that they selectively populated primary visual nuclei, suggest- Cepko, 1992; Levison et al., 1993; Reid et al., 1995). In the ing a relationship with the tectum, the normal derivative of the retina (an outgrowth of the diencephalon) and the chick tectum mesencephalon. (a derivative of the mesencephalon), lineage analysis has also Using a complex retroviral library comprising DNA tags as shown that neurons and glia arise from a common progenitor lineage markers (Golden et al., 1995), we have evaluated 275 cell (Turner and Cepko, 1987; Galileo et al., 1990; Turner et clones in the chick diencephalon. This technique allows al., 1990; Gray and Sanes, 1992; Fekete et al., 1995). analysis of clones in the mature diencephalon, after clonal dis- The diencephalon is organized into individual nuclei (col- persion is complete. We found that sibling cells can spread lections of neurons with a specific projection or defined set of extensively within the diencephalon, sometimes occupying projections) with some groups of nuclei having similar func- nuclei derived from more than one neuromere. Dispersion tional properties and other neighboring nuclei having distinct occasionally led to cells being located in both the diencephalon functions. Functionally, however, the diencephalon and, par- and the mesencephalon. Furthermore, siblings were found on ticularly, the thalamus closely resemble the telencephalon both the left and right sides of the third ventricle in approxi- (cerebral cortex and basal ganglia). The thalamus functions as mately 16% of the clones. We also found that progenitors of the major integration and projection region to the cerebral diencephalic cells were located in the ventricular zone of the cortex from more caudal structures, including the spinal cord, third ventricle, as well as in the ventricular zone of the lateral cerebellum, hindbrain and midbrain (Jones, 1985). The ventricle. Clones frequently contained neurons, glia and radial neurons of the thalamic nuclei project to different cortical areas glia, supporting the existence of multipotent progenitor cells. with many thalamic nuclei projecting to overlapping cortical regions. Furthermore, the thalamocortical projections to the cortex appear to define the functional cortical units (see Jones, MATERIALS AND METHODS 1985 for review). During development, the thalamocortical projections show growth into specific cortical areas and thala- Production of the retroviral library CHAPOL mocortical axons may provide cues that help define the course Construction and characterization of the retroviral library used in this of cortical neuronal projections (reviewed in O’Leary and study have been described (Golden et al., 1995). Briefly, the library Koester, 1993). of retroviral vectors was constructed from a replication defective Given that the telencephalon and rhombencephalon appear avian retrovirus, CHAP (Ryder and Cepko, 1994), that encodes the human placental alkaline phosphatase gene (PLAP). A pool of to use distinct mechanisms for patterning and perhaps gener- synthetic degenerate oligonucleotide tags with a theoretical complex- ating cellular diversity, it was of interest to determine clonal ity of >107 were cloned into the vector to create the vector library, relationships within the diencephalon. Classical embryologists CHAPOL. A large scale viral preparation yielded a stock with a con- have described the diencephalon as arising from four horizon- centration of 1.1×107 cfu/ml that was used for all experiments (Golden tal strips (His, 1893; Herrick, 1910; Khulenbeck, 1973) or neu- et al., 1995). Numerous experiments with this same stock of CHAPOL romeric units (Orr, 1887; Bergquist, 1952; Keyser, 1972) based have led to recovery of >350 unique inserts (current paper and on the presence of bulges and sulci along the medial walls of (Golden et al., 1995)). These data indicate that the CHAPOL library the third ventricle. These morphological units, labeled neu- has a complexity of at least 105 members (Walsh et al., 1992). Fur- romeres or prosomeres (we have adopted the neuromere thermore, each insert has been recovered from only one infection nomenclature in this paper), have been proposed to be event (i.e. only 1 brain or 1 well of an infected microtiter plate of tissue culture cells), indicating the library has an approximately equal analogous to the rhombomeres (Puelles et al., 1987; Figdor and distribution of inserts. Stern, 1993; Puelles and Rubenstein, 1993), although this analogy has not been completely tested (see Guthrie, 1995). In vivo infection Examination of the expression patterns of a variety of devel- The neural tube of fertilized virus-free White Leghorn chicken opmentally regulated genes has shown a correlation with the embryos was injected as previously described (Fekete and Cepko, Clones in the chick diencephalon 67

1993) using CHAPOL. Injections were performed at either stage 10- on a Zeiss Axioskop. The 2× magnification Kodachromes were 12 or stage 16-17 (all staging according to Hamburger and Hamilton, scanned into a Macintosh Quadra 650 using a Kodak 2035 plus slide 1951). Approximately 0.3-0.5 µl of CHAPOL stock was injected into scanner and Adobe Photoshop software. The location and cell type of the neural tube at stage 10-12 and 1.0-1.5 µl at stage 16-17. Infected each alkaline phosphatase positive (AP+) cell and/or cluster of cells embryos were incubated in a humidified, 37°C chamber until was given a unique identification based upon the section number. embryonic day 18 (E18) at which time the brains were harvested in Once documented, each cell or cluster of cells (with a small group of PBS. After fixation overnight in 4% paraformaldehyde (in PBS, pH surrounding white AP− cells) was removed using heat-pulled glass 7.4) at 4¡C, brains were washed with three changes of PBS over 24 micropipettes and transferred to a 96-well PCR plate with 10 µl of hours and then cryoprotected in 30% sucrose in PBS. Brains were 400 µg/ml proteinase K solution (Golden et al., 1995). Approximately oriented for coronal sections, embedded in OCT medium and cut on 50-60 AP+ cells were picked from sections and analysed on each 96- a Reichart-Jung CM3000 cryostat at 60 µm. Each section was well PCR plate. One or two picks from a similar stage uninfected brain collected and mounted sequentially on Superfrost/Plus microscopic that had been processed in parallel with the experimental brain were slides (Fisher Scientific). On average, 3 sections/brain (approximately included on each plate as negative contrals. The cells were then 45 sections included the diencephalon) were lost and the ordered digested, amplified and sequenced as previously described (Golden et position of each missed section was documented. Infected cells were al., 1995). Once all sequences were collected and stored, all common identified histochemically by incubating the sections with X-phos sequences could be grouped and the position and cell type(s) for (Research Organics, Inc.) and NBT (Research Organics, Inc.) for 60 sibling cells revealed. Each AP+ cell or cluster or AP+ cells was then to 240 minutes according to previously published protocols (Ryder reidentified on the computer images to obtain a map of the distribu- and Cepko, 1994; Golden et al., 1995). Cells infected with the retro- tion of sibling cells in the diencephalon. The position of each clone virus were identified by the purple formazan precipitate. was also transposed onto an atlas of the chick diencephalon (Kuenzel Coronal sections that included the diencephalon, from the anterior and Masson, 1988). AP+ cells were anatomically positioned on atlas commissure (anterior limit of analysis) to the red nucleus (posterior sections by first defining their location within nuclei or white matter limit of analysis) were photographed at 2× magnification with a Nikon tracts on histological sections (Fig. 1). Since all nuclei and white SMZ-U stereomicroscope equipped with a Nikon DX FX-35WA matter tracts could not be defined on the AP-stained sections, cells Camera. Representative cells were photographed at higher magnifi- outside anatomically definable structures were localized according to cation on a Zeiss Axiophot microscope. In addition, camera-lucida their position relative to defined structures. However, the majority of drawings of selected cells were made from a drawing tube mounted cells could be definitively assigned based upon the tissue morphol-

Fig. 1. Examples of distinct cell types and histological identification of diencephalic nuclei. (A-F) Representative examples of cell types identified on the basis of morphology. Neurons were identified by the presence of long thin processes. Glia (astrocytes) were identified by intense staining and indistinct borders, with occasional thick short processes. Radial glia were identified by a cell body apposed to the ventricular surface and a long thin process radiating into the parenchyma. Cells that could not be identified were generally small and round with no clear morphological feature. (G,H) Two representive sections of the diencephalon stained for AP activity. The left half of each figure is a schematic outline of the nuclei and white matter tracts that could be defined as a result of the AP histochemistry and anatomical landmarks. Note the AP+ cells in the nucleus rotundus in G and the nucleus ovoidalis in H. Scale bar 100 µm for A, B and E; 50 µm for C; 267 µm for D; 200 µm for F; and 450 µm for G and H. AL, ansa lenticularis posterior; DLAl, nucleus dorsolateralis anterior thalami (lateral); DLAm, nucleus dorsolateralis anterior thalami (medial); FPL, fasciculus prossencephali lateralis (lateral forebrain bundle); GLv, nucleus geniculatus lateralis, pars ventralis; ICT, nucleus intercalatus thalami; OM, tractus occipitomesencephalicus; OT, optic tract; OV, nucleus ovoidalis; ROT, nucleus rotundus. 68 J. A. Golden and C. L. Cepko ogy. The following structures were identified on histological sections: into 275 clones (the same insert has never been recovered from ansa lenticularis, anterior commissure, nucleus anterior medialis more than one brain). Of these, 154 picks contained an insert hypothalami, posterior commissure, nucleus dorsointermedialis that was recovered only once and were therefore defined as posterior thalami, nucleus dorsolateralis anterior thalami (lateral), ‘single pick clones’. These 154 clones had from 1 to 8 cells nucleus dorsolateralis anterior thalami (medial), nucleus dorsolater- (i.e. each pick had from 1 to 8 cells), 118 having a single cell. alis posterior thalami, nucleus dorsomedialis anterior thalami, nucleus 121 insert sequences were obtained 2 or more times from the dorsomedialis posterior thalami, fasciculus prossencephali lateralis (lateral forebrain bundle), nucleus geniculatus lateralis, pars ventralis, 761 picks and these were defined as ‘multiple pick clones’. nucleus intercalatus thalami, optic chiasm, tractus occipitomesen- Each pick from the multiple pick clones also had from 1 to 8 cephalicus, optic tract, nucleus ovoidalis, nucleus paraventricularis cells. The single pick clones and multiple pick clones are hypothalami, nucleus rotundus, tractus tectothalamicus. Three-dimen- analyzed separately below. sional reconstruction of the diencephalon was performed by trans- forming the Photoshop two-dimensional images using Spyglass Dicer. Single pick clones The 154 single pick clones contained an average of 1.52 cells DiI labeling of which 0.4 were neurons, 0.4 were glia and 0.1 were radial Radial glia were labeled as previously described (Gray and Sanes, glia. An average of 0.6 cells in each clone was not classifiable 1992). Briefly, brains from uninfected E15-E18 chicks were dissected on morphological criteria alone (Fig. 1). Oligodendrocytes in PBS and fixed overnight in 4% paraformaldehyde. After washing in PBS, approximately 0.1 ml of a 2.5 mg/ml solution of DiI were rarely found and difficult to identify morphologically in (Molecular Probes, Inc.) in 100% ethanol was injected into the right this study. Since they represented such a small subset of cell lateral ventricle of the brain using a 30-gauge needle. The DiI solution types identified, they have been lumped with astrocytes in the was allowed to passively fill the entire ventricular system. The brains category of glial cells. The data derived from the single cell were then placed in a 60 mm Petri dish filled with sterile water and clones provide an approximation of the birthdates for different 0.03% sodium azide and incubated for two weeks at 37¡C. After incu- cell types (see Table 1 and Discussion). For example, a greater bation the brains were imbedded in 5% agar (in H2O) and sectioned percentage of the single pick clones with more than one cell, on a vibratome at 100-200 µm. The sections were mounted on Super- compared to single pick clones with one cell, comprised glia frost/Plus microscopic slides, coverslipped with gelvatol and viewed only (31% vs 5%). This suggests that progenitors that ulti- with rhodamine fluorescent filters on a Zeiss Axiophot microscope. mately gave rise only to glia were making some mitotic Photos were taken with Kodak Elite 100 film. daughters at the time that they were infected with the retrovirus. In contrast, the percentage of neurons from single pick clones with multiple cells, compared to those with a single cell, RESULTS was relatively similar (22% vs 27%). Thus neurons were being generated at the time of injection more often then glia and some Summary of brains analyzed and recovery rates of progenitors produced mitotic daughters that gave rise only to infected cells neurons. Furthermore, the percentage of single pick clones The clones described here were isolated from seven E18 with a single cell that was a neuron (27%) compared to the per- brains. Injection of the retroviral library was at stage 10 (n=1), centage of clones containing neurons in the entire data set stage 11 (n=1) or stage 16-17 (n=5). Small tissue samples con- (60%) was greater than the comparable numbers of clones containing a single alkaline phosphatase (AP+) cell or small taining glia (5% compared to 65%). Thus, neurons were far clusters of AP+ cells along with surrounding AP− tissue were more frequently represented in the single cell clone category. removed (each sample is referred to as a ‘pick’). The total number of picks was 1,537 from the diencephalon of the 7 brains. An additional 263 picks were collected from areas with Table 1. Cell type composition of single and multiple pick no visible AP+ cells. PCR amplification yielded the predicted clones 121 base pair product from 1,088 (71%) of the 1,537 picks. Single pick Sequencing of the PCR products revealed that 70% (761) of Multiple multiple single the AP+ regions isolated contained a single insert and 17% pick cells cell total (186) contained multiple inserts. (Those with multiple inserts (a) Number of inserts 121 36 118 154 were not analyzed further since identification of more than one recovered sequence implies that cells that are members of more than one (b) Number (percent) clones clone were present in the tissue sample analyzed. If one AP+ with cell types cell was present, it was impossible to define which sequence Include neurons 59(49) 16(44) − 48(31) belonged to the AP+ cell and which belonged to the AP− Neurons only 6(5) 8(22) 32(27) 40(26) cell(s). Similarly, if two or more AP+ cells were present, it was Include glia 74(61) 17(47) − 23(15) again impossible to determine which sequence belonged to Glia only 9(7) 11(31) 6(5) 17(11) which cell or cells. Furthermore, technical limitations have Include radial glia 20(17) 3(8) − 13(8) made it difficult to separate multiple PCR products, so that the Radial glia only 0 2(6) 10(8) 12(8) sequence of each can be defined (Golden et al., 1995).) No sequence was obtained in 141 (13%) of the cases. Efficiency The number (and percent) of clones that include the indicated cell type are grouped according to whether they are multiple pick or single pick clones. of recovery of PCR products and sequences was similar from Multiple pick clones are the clones that gave the same insert from multiple single cells and small clusters of cells (data not shown, and tissue picks in the same brain. Single pick clones are clones in which an insert Golden and Cepko, 1995). was recovered only once. The single pick clones are further divided into the The 761 picks that gave a single sequence were classified picks comprising a single cell or multiple cells. Clones in the chick diencephalon 69

Analysis of the composition of small clones reveals that appeared to disperse both laterally and ventrally (e.g. Fig. 2, there are more radial glia than would be predicted based upon clones 2 and 3; Fig. 3, clone 9; and Fig. 4, clones 12 and 14). the overall frequency of radial glia in the entire data set. 10 of Some clones showed a more limited dorsal to ventral disper- the 33 (30%) clones that contained radial glial cells were single sion with extensive lateral dispersion (e.g. Fig. 2, clone 12 and cell clones. An additional 2 multiple cell clones from the single Fig. 3 clone 5), but few showed extensive dorsal-ventral dis- pick clones were composed of radial glia only, making a total persion without accompanying lateral dispersion. While the of 12 of 33 (36%) clones with only radial glia. These two vast majority of clones showed lateral and ventral spread, clones that contained radial glial cells as their only members several clones violated this general trend. Two clones were indicate that there may be progenitor cells that can divide to found to spread a significantly greater percentage along the give rise to 2 or more radial glia (also see Gray and Sanes, dorsal to ventral axis compared to the medial to lateral axis 1992). However, since there were no multiple pick clones (Table 2, asterisked clones). composed entirely of radial glia, such a progenitor cell, if it The dispersion of clones in three planes indicates that cells exists, does not produce many radial glia. from an individual clone occupy >1 nucleus of the dien- Analysis of the distribution of the 118 single cell clones in cephalon (e.g. Fig. 2, clones 1 and 10; Fig. 3, clones 1 and 14; three dimensions suggested no particular bias for an anterior- and Fig. 4, clone 3). Clonally related neurons and glia could to-posterior, medial-to-lateral, or dorsal-to-ventral birth order. be found in nuclei that are both adjacent to and distant from However, the limited number of clones and injection times each other. Some clones remained localized to a single nucleus preclude definitive assessment of birthdating gradients in the or cluster of nuclei (e.g. Fig. 2, clone 6; and Fig. 4, clone 13). diencephalon. Since sibling cells populated multiple nuclei, we attempted to define whether clones crossed neuromeric boundaries. Multiple pick clones However, detailed maps delineating which mature structures 121 inserts were recovered from two or more picks in the same are derived from the developmentally defined neuromeres are brain. The average number of cells in these clones was 7.8. The not available for any species. Thus, in order to provide an average numbers of neurons, glia and unidentified cell types in estimate of the number of clones that occupy derivatives of >1 these clones were 1.4, 3.6 and 2.8 respectively. In contrast to neuromere, we extrapolated from the data of Figdor and Stern the single pick clones, multiple pick clones frequently had (1993), identifying only those clones that clearly crossed neu- multiple cell types (Table 1). It remains possible that some of romeric boundaries (Table 3). Cells that were in the region of the clones that contained unidentifiable cells would have been a proposed boundary (Puelles and Zabala, 1982) were placed classified as clones with only one cell type. Nonetheless, 43 of in the ‘could not determine’ category of Table 3. Based on this the 121 (36%) clones had phenotypically distinguishable cells analysis, at least 8% of all clones crossed neuromeric bound- (Table 2), either neurons and glia, neurons and radial glia, glia aries. Because the precise boundaries have not been defined for and radial glia, or all three. These data indicate that multipo- the mature diencephalon, this represents the most conservative tent progenitor cells were frequently infected. estimate possible. It is likely that some of the 47% of the clones 101 of the clones were found exclusively on either the left that could not be definitively assessed did cross a neuromeric or right side of the lateral ventricle. Table 2 is a list of all 121 boundary. Furthermore, the fact that many clones were found multiple pick clones and includes the composition of cell types within a single neuromere does not mean that they were and the extent of dispersion in the anterior-posterior, medial- restricted by the boundaries of that neuromere. lateral, and dorsal-ventral planes. Figs 2-4 are composite maps showing the distribution of cells within clones from a subset Unusual clone types (44/121) of the multiple pick clones and Fig. 5 shows the Several clones were found with unusual distributions of cells anterior to posterior spread (also see Fig. 2, clone 3 and Fig. or an unexpected site of genesis. One type of clone had cells 3, clone 7). on both sides of the midline surrounding the 3rd ventricle. 20 Clones were observed to spread from 2% (60 µm) to 54% clones (16%) in the data set were found with a bilateral distri- (1140 µm) of the anterior to posterior extent of the entire dien- bution (Table 2; Fig. 3, clones 3, 6, 9, 11, 13 and Fig. 4, clone cephalon, with the average clone extending 11% of the 2). These clones contained both glia and neurons, including anterior-posterior distance. Thus cells were separated by neurons on both the right and left sides of the third ventricle approximately 350 µm on average along the anterior-posterior in individual cases. Overall, the distribution of cell types was axis. Spread of sibling cells was most pronounced in the similar to that of the unilateral clones. Several presumptive medial-lateral axis, where the average extent of spread was clones in which cells appeared to be migrating across the 57.5% (between 1500 and 2000 µm; table 2), and spread anterior or posterior commissure have been found in brains ranged from none (cells located along the third ventricle) to infected at stage 16-17 with CHAP or CHAPOL and harvested spread across virtually the entire diencephalon from the third at E8-E10 (Szele F, Golden J, and Cepko C, unpublished data ventricle to the lateral margin (in Fig. 3G compare clone 8 to and Fig. 6A). Analysis of the anterior-to-posterior distribution clone 5). of bilateral clones showed no specific localization. Cells also spread along the dorsal-ventral axis across an Although most studies have focused upon the germinal zone average of 19% of the entire diencephalon, corresponding to of the third ventricle as the source of cells that populate the approximately 1000 µm. The range of spread was from 1% (5 diencephalon, another rare clone type appeared to originate in µm) to 74% (2900 µm; Table 2). The spread of cells in the the lateral ventricle. Several of these clones included radial dorsal-ventral plane showed a common bias toward the ventral glial cells projecting from the lateral ventricle medially and side with progressively more laterally displaced cells. Siblings ventrally into the diencephalon (Fig. 4 clone 10; note that the leaving the third ventricle, the most common site of birth, cerebral hemispheres and lateral ventricles, located along the 70 J. A. Golden and C. L. Cepko 3A. 2A. Clones in the chick diencephalon 71 Three-dimensional Fig. 5. reconstruction of clones within the diencephalon. Representative clones have been transposed to the surface of a reconstructed diencephalon. Each cell was localized according to its anterior-posterior and dorsal-ventral position; medial-lateral position is not shown. Several clones spread widely in the anterior-posterior plane, while others remain m sections were entered onto µ Schematic representation of 44 clones. Diagrams the diencephalon and Figs 2-4. associated structures were traced from an atlas of the chick brain (Kuenzel and Masson, 1988). Cells from approximately 7 serial 60 each of seven coronal images, (A-G) oriented from anterior to posterior, respectively, based on the location of cells in histological sections (see Methods). Each color represents a separate clone that has been randomly numbered. All of the cells within clone are represented regardless of cell type. For illustrative purposes, the clones from several brains, regardless of their location in the brain, have been placed on left side of each figure except bilateral clones which are shown on both sides. White matter tracts are colored light grey and nuclei clear. AC, anterior commissure; ALP, nucleus ansae lenticularis posterior; AM, anterior medialis hypothalami; CP, posterior commissure; CPa, commissura pallii; DIP, nucleus dorsointermedialis posterior thalami; DLAl, nucleus dorsolateralis anterior thalami (lateral); DLAm, nucleus dorsolateralis anterior thalami (medial); DLP, posterior thalami; DMA, nucleus dorsomedialis anterior DMP, posterior thalami; FPL, fasciculus prossencephali lateralis (lateral forebrain bundle); GCt, substantia grisea centralis; GLdp, nucleus geniculatus lateralis, pars dorsalis; GLv, nucleus geniculatus lateralis, pars ventralis; ICT, intercalatus thalami; IH, nucleus inferioris hypothalami; LA, lateralis anterior thalami; LHy, lateral hypothalamic nucleus; OC, optic chiasm; OM, tractus occipitomesencephalicus; OT, optic tract; OV, nucleus ovoidalis; PHN, nuclus periventricularis hypothalami; PVN, nucleus paraventricularis hypothalami; ROT, rotundus; SCE, stratum cellulare externum; SL, nucleus septalis lateralis; SM, medialis; T, triangularis; Te, tectum; TeO, optic TT, tractus tectothalamicus; VLT, nucleus ventrolateralis thalami. relatively constricted. Dispersion along the dorsal-ventral axis also varied. The overall extent of dispersion a clone did not always reflect the number cells within clone, as small clones showed similar dispersion patterns to large clones. 4A. 72 J. A. Golden and C. L. Cepko

Table 2. Cell type composition and dispersion of all multiple pick clones

N G(RG) U APµm(%) DVµm(%) MLµm(%) N G(RG) U APµm(%) DVµm(%) MLµm(%) 0 0 2 60(2) 1515(42) 3030(74) 0 0 2 2 0 0 120(4) 61(3) 3152(83) 2 2 0 60(2) 520(12) 2881(75) 2 18(1) 4 540(16) 667(16) 1667(70) 1 0 1 60(2) 273(7) 2515(75) 8 5 1 660(20) 1152(19) 2273(71) 0 7(1) 3 * 0 0(2) 2 60(2) 2273(54) 1758(48) 0 5 0 120(4) 182(5) 636(28) 0 1 2 120(4) 152(3) 2667(80) 0 0 3 0 8(1) 0 360(11) 667(15) 1212(36) 0 3 3 120(2) 394(8) 455(14) 0 0 2 60(2) 212(5) 1970(52) 0 0 1 0 1 1 0 8 0 300(11) 394(10) 1576(63) 3 0(2) 0 60(2) 455(10) 364(10) 1 0 2 60(2) 515(12) 1606(76) 0 4 1 300(9) 2121(56) 2879(63) 0 4 1 120(5) 152(3) 2515(73) 1 0 2 360(11) 91(2) 1000(30) 8 0 7 240(9) 303(7) 1061(78) 1 1 1 1 0 1 120(5) 2394(53) 1455(55) 1 1 0 60(2) 91(2) 1121(30) 0 0 2 60(2) 61(1) 485(19) 1 4 0 840(25) 1758(39) 2000(73) 0 0 2 60(2) 91(2) 545(29) 0 3 1 120(4) 1152(25) 1424(39) 1 0(1) 0 60(2) 0 0 3 60(2) 242(8) 1364(50) 0 3 1 180(7) 667(16) 2939(94) 1 10 0 120(4) 697(14) 2697(70) 1 0(1) 0 300(11) 1303(31) 91(8) * 1 10 0 300(9) 2667(70) 2727(45) * 1 2 0 840(32) 2939(69) 1606(65) 0 1(4) 0 0 0 3 360(14) 2061(74) 2455(79) 3 28 0 480(15) 1515(36) 2636(82) 1 0 3 300(11) 212(5) 1303(41) 1 0(1) 0 60(2) 2212(46) 2182(64) 1 0 1 60(2) 273(6) 667(22) 1 5 0 300(9) 303(6) 2273(63) 1 2 3 360(14) 2455(54) 2606(82) 6 2 2 1020(31) 2273(47) 2576(79) 0 4 0 120(6) 50(1) 93(26) 5 4 0 60(2) 667(16) 1364(59) 0 2 0 60(3) 250(5) 355(53) 0 2 0 60(2) 60(1) 1667(46) Many Many 180(9) 550(13) 2930(76) 0 13 2 300(9) 1485(32) 1606(57) 0 2 4 1140(54) 930(20) 2470(67) 1 0 1 480(15) 1303(28) 2848(76) 11 14 24 480(23) 930(20) 2900(82) 0 5(3) 2 420(13) 1091(22) 1606(27) 2 15 0 240(10) 580(15) 2090(77) 0 15 2 1140(35) 1727(38) 1970(57) 0 2 1 60(2) 3 0 0 60(2) 394(10) 2152(62) 0 3 1 60(2) 600(22) 1900(72) 2 0 0 420(13) 1394(30) 3333(92) 2 20(2) 13 660(24) 500(12) 2350(80) 0 2 2 60(2) 154(4) 3364(74) 0 0 2 210(7) 300(7) 1850(63) 3 21 1 300(9) 3030(68) 2667(97) 3 0(2) 1 60(2) 200(7) 500(37) 1 17 0 1020(31) 1000(21) 2242(62) 0 4 0 240(9) 300(7) 800(26) 0 2(1) 0 120(4) 8 1 5 120(4) 650(15) 600(24) 0 0 5 420(13) 636(13) 818(23) 2 0 0 60(2) 0 0 8 0 2 300(9) 303(7) 1788(87) 1 0 2 360(13) 200(7) 1700(58) 0 1 1 360(11) 394(8) 1455(44) 1 0(1) 1 240(9) 300(8) 500(19) 0 1 2 360(11) 1364(31) 1970(54) 2 3(1) 4 900(33) 1900(48) 1900(62) 0 2 0 60(2) 364(8) 2636(76) 0 3 1 240(9) 80(3) 1950(70) 0 6 1 360(11) 0 2 3 840(31) 200(5) 1900(70) 0 3(2) 2 720(22) 1000(22) 3061(76) 0 2 0 60(2) 150(4) 2350(84) 1 13(1) 0 360(11) 455(11) 1152(32) 0 0 3 3 3 23 900(40) 1300(30) 2200(90) 2 0 0 60(2) 180(4) 1750(66) 5 0 5 900(40) 185(4) 750(27) 0 9 3 1020(38) 1000(25) 1470(54) 18 0 42 60(3) 460(18) 1850(95) 0 2 3 60(2) 150(5) 0 0 0 7 480(20) 185(4) 400(15) 3 0 0 60(2) 410(11) 400(14) 0 0 4 60(3) 185(4) 0 0 4 60(2) 1000(24) 1320(69) 0 0 14 540(25) 850(30) 2000(98) 0 0 2 60(2) 300(7) 1320(41) 4 2 7 180(8) 370(15) 1750(67) B 1 6 2 B 1 13(1) 2 B 0 5 0 B 1 0(1) 5 B1 60 B032 B0 31 B003 B4 70 B021 B7 122 B083 B0 310 B142 B0 04 B3511 B 0 1 1 B 1 0(1) 4 B2 01 B001

n= 164 405(30) 340 Clones: n=121(44) Range µm 60-1140 5-2939 0-3550 % 2%-54% 1%-74% 0-98% ave% 11.10% 18.90% 57.50% The table is a complete listing of the cell types and the dispersion of all multiple pick clones (n=121). Dispersion is given in µm and in parenthesis is the percent of the distance from anterior to posterior (AP), dorsal to ventral (DV), and medial to lateral (ML) at the level of the brain where the clone was located. If dispersion distances are not listed, this indicates that either these clones arose from the lateral ventricle, or includes cells from AP− areas that also gave the same insert and thus the full extent of dispersion could not be accurately determined. An asterisk identiﬁes the clones that showed a greater dorsal-to-ventral spread compared to the medial to lateral. B indicates that the clone is bilateral. N, neurons; G, glia; RG, radial glia; U, unidentiﬁed cell. Clones in the chick diencephalon 73

Table 3. Clones crossing neuromeric boundaries zone of the lateral ventricle, supporting the finding that pro- Could not genitors located in the lateral ventricle can contribute to the Clone type n Cross Don’t cross determine diencephalon. Unilateral 90 5 (6%) 47 (52%) 38 (42%) − Bilateral 18 4 (22%) 1 (6%) 13 (72%) Analysis of alkaline phosphatase negative (AP ) areas Total 108 9 (8%) 48 (44%) 51 (47%) A potential problem in retroviral mediated lineage analysis is The number (and percent) of clones that were observed in >1 neuromere the failure to detect infected cells. For example, transcription (crossed) was determined by comparing the location of all members of a from the LTR promoter could be weak or absent or the AP clone to the boundaries proposed by Figdor and Stern (1993). Only clones protein could be mutant or unstable. To investigate this issue, with siblings clearly located in more than one neuromere were included as crossing. All clones that had cells near borders were classified as “could not the entire diencephalon from one heavily infected brain was determine”. systematically analyzed for the presence of non-expressing cells. 263 regions were isolated from histochemically negative areas (i.e. no purple cells). These picks were defined as either dorsal border, are not included on the illustration). Other cells small (n=52), approximately the size of the picks from single from these clones included neurons and glia that were within AP+ cells, or large (n=211). The large picks involved removing the body of the diencephalon. A total of 5 of the 275 (2%) from 1/8 to 1/4 of a section of diencephalon from a slide and clones appeared to have arisen from the lateral ventricles, subjecting the tissue to PCR, as was done with the AP+ areas. making the contribution of progenitors from the lateral ven- For the small AP− picks, an insert was amplified from 19 tricles to the final cell numbers in the diencephalon relatively (37%) picks. Sequence analysis of the inserts from the small small during the time period studied. No specific composition picks revealed a single insert in 17 of the 19 amplified of cell types within these clones was found (Table 2). products, and multiple inserts in 2 of the 19. The large picks Another unusual clone type had siblings in both the dien- had a higher recovery rate for inserts after amplification cephalon and the mesencephalon (tectum) (Fig. 3, clone 12 and (178/211, 84%). Sequence analysis revealed 133 of the 178 Fig. 4, clone 10). Since only a few AP+ cells were analyzed inserts were actually multiple sequences and only 25 of the 178 from the mesencephalon, and no cells from the telencephalon, inserts were single sequences. 20 amplified products yielded an estimation of the frequency of clones that cross from the no sequencing product. The higher recovery rate and presence diencephalon to adjacent brain regions cannot be made. of multiple inserts in a greater proportion of the large picks However, the presence of siblings in both the diencephalon and was attributed to the large tissue volumes collected. In some mesencephalon indicate that the earliest subdivisions of the cases, the sequences from the histochemically negative regions CNS do not form absolute lineage boundaries, as all infections were unique indicating that entire clones either did not express were made after the prosencephalon and mesencephalon active PLAP or turned off the expression at some time prior to boundaries were established. One of the clones with cells in E18. Among other clones (5/93, 5.4%), a subset of cells the diencephalon and mesencephalon appeared to have its exhibited alkaline phosphatase activity while other cells did not origin in the lateral ventricle. Thus it is possible that cells at the time of analysis. In addition, 25 samples from uninfected generated in either the lateral ventricle or third ventricle can brains were analyzed and an insert was never recovered from migrate between these two brain regions. Examination of these brains. Several explanations for the AP− cells exist. Since whole-mount E8 chick brains showed a relatively simple and we completed this study, we observed that some cells, partic- direct possible pathway for the migration of cells into these ularly some neurons, require a staining time longer than used two structures (diagramed in Fig. 6, cell C). here to become positive by light microscopy (J. Lin, F. Szele, J. Zitz, J. Golden and C. Cepko, unpublished data). Studies The pattern of radial glia in the diencephalon from other brain regions have also shown that some cells can The most common pattern of clonal dispersion in the dien- turn off AP activity (Halliday and Cepko, 1992) while other cephalon was from medial to lateral with a progressively more cells appear never to exhibit AP activity. Although we cannot ventral location as members of the clone migrated more determine the contribution of any one mechanism in our data, laterally. One common pathway for migrating newborn it remains likely that a combination of these factors play a role neurons in the CNS is along radial glial fibers. To investigate in recovering viral genomes from AP− areas. whether radial glia could account for the most common pattern of dispersion seen here, DiI was used to label radial glial cells. Fig. 7 shows the pattern of fibers revealed by the DiI staining DISCUSSION in the diencephalon. From the third ventricle, radial glial fibers course laterally and ventrally in a pattern paralleling the dis- We have analyzed lineal relationships in the chick dien- tribution of sibling cells. Clones that did not disperse in this cephalon using a highly complex retroviral library. Although common pattern may be using other guides that were not iden- the rodent has served as a model organism for lineage analysis tified in this study. in the cerebral cortex (Price et al., 1987; Walsh and Cepko, Radial glial cells were also noted in the lateral ventricle 1988, 1992; Austin and Cepko, 1990; Parnavelas et al., 1991; extending into the diencephalon (Fig. 7). The processes of Levison et al., 1993) and retina (Turner and Cepko, 1987; these cells showed a general downward trend as they moved Turner et al., 1990), technical limitations prevent the use of from the lateral ventricle into the substance of the dien- rodents for studying structures such as the diencephalon that cephalon. This pattern reflects the distribution of cells found develop before E12-E13, the earliest age that the neuroepi- in a few of the clones that appeared to originate in the germinal thelium is accessable for viral injections. The timing of 74 J. A. Golden and C. L. Cepko

no characteristics that distinguished them from the clones of the animals injected later, the data from all 7 animals were pooled. Neuromeres and patterning of the diencephalon The developmental mechanisms governing segregation of the diencephalon into nuclei remain largely unknown. One hypothesis is that the diencephalon is parceled into segments, each giving rise to a set of functionally related nuclei, analogous to the rhombomeres in the hindbrain. Two models have been proposed for parcelling the diencephalon into segments. The first model divides the diencephalon into three units (prosomeres) and has been most extensively studied in the mouse (see Rubenstein and Puelles, 1994 and Rubenstein et al., 1994 for reviews). In the second model, four units (neuromeres) have been identified, with the most caudal two in this second model corresponding to the first prosomere in the first model (Figdor and Stern, 1993). Since the second model of four neuromeres has been proposed for the chick and the first primarily established in the mouse, we have chosen to use the neuromeric model in analysis of the data. Figdor and Stern (1993), utilizing single cell injection and DiI and DiO labeling techniques, found that clones did not cross the morphological boundaries between the neuromeres of the diencephalon by stage 25 in development of the chick. These findings are similar to those of lineage analysis in the hindbrain rhombomeres. In contrast, however, using the boundaries of neuromeres defined by Figdor and Stern in the stage 40 chick (see Fig. 2 in Figdor and Stern, 1993) at least 8% (a minimum defined by conservative criteria) (see Table 3) of the clones in the current study cross the boundaries of neu- Fig. 6. Generation of bilateral clones and clones that span the romeres. Furthermore, in the prosomeric model (Rubenstein diencephalon and mesencephalon. (A) A 60 µm coronal section at and Puelles, 1994; Rubenstein et al., 1994), the hypothalamus the level of the anterior commissure (AC) of an E9.5 chick brain and thalamus are derived from distinct prosomeres, thus the injected with CHAP at stage 17 is shown. A nearly continuous clones spanning these two regions identified in this paper population of cells appears to trace back to an origin in the would also have to be considered boundary crossers (see ventricular zone of the third ventricle. Migrating cells, with long below). In the calculations performed for Table 3, hypothala- leading processes, appear to move away from the ventricle, then mic-thalamic clones were not included as crossing boundaries. appear to turn and enter the AC. A bundle of fibers crossing the midline clearly defines the dorsal and ventral limits of the AC. (B) A Several explanations exist for the difference in dispersion whole-mount of an E8 chick brain viewed from the dorsal side with between the two studies. The first, and most likely, is that the the midline opened to display the ventricular surfaces. The left side tracer labeling studies were analyzed after only 48-72 hours, of the brain was traced and the major subdivisions colored and whereas in the current study, embryos were harvested 16 days labeled. The position of the anterior and posterior commissures (AC beyond the time of infection. Thus it seems likely that pro- and PC, respectively) is shown on the diagram (arrows) as possible genitor cells give rise to daughters that are initially restricted routes for cells to cross the midline. Cell A depicts one potential to within a single neuromere, but that later some siblings route in which a cell could migrate around one end of the third escape restrictions and move according to other mechanisms ventricle. Although this is only shown at the anterior end, there is a or cues. Alternatively, a relatively small subset of clones cross continuous surface from the anterior end to the posterior limit of the neuromeric boundaries early in development; such a subset of diencephalon, along the ventral surface. Cell B illustrates a model in which a midline progenitor cell is the source of cells that migrate clones may have been missed by the techniques used in the into each side of the diencephalon. Cell C shows where a clone previous study. This scenario is similar to the recognition of a might originate that spans the diencephalon and the mesencephalon. small percentage of clones that violate rhombomere boundaries Scale bar 200 µm. early in development (Birgbauer and Fraser, 1994). We have also analyzed retrovirally marked clones in the ventral forebrain, focusing on the hypothalamus, in brains injection of the retrovirus was selected according to the time harvested at E8-E10 (Arnold-Aldea and Cepko, 1995). when neuromeric boundaries were established in the dien- Approximately 96% of the clones were simple radial columns, cephalon (Figdor and Stern, 1993). 5 of the 7 chicks were which appeared to respect neuromeric boundaries. However, injected at stage 17 after the major boundaries were established approximately 4% of the retrovirally marked clones were and the other two were injected prior to the formation of dien- widely dispersed, with distinctive patterns of dispersion at E8- cephalic boundaries. Since the two brains infected at stage 10- E10, including bilaterally symmetric clones in the hypothala- 11 contained a total of only 9 clones, and these clones showed mus. Ongoing experiments in our laboratory are seeking to Clones in the chick diencephalon 75

Fig. 7. DiI labeling of radial glia. (A) Overview of a DiI-labeled, 200 µm section through the diencephalon of E15 brain. The third (3 V) and lateral (LV) ventricles of a fixed embryo were filled with a DiI solution; periventricular regions show bright labeling. Single fibers are seen radiating into the body of the diencephalon from both ventricular locations. (B) A schematic diagram of (A) with a tracing of several of the radial glia arising along the third and lateral ventricles. The two boxes indicate the location of both C and D. (C) A view at higher magnification from the region where the radial glial fibers decussated. Fibers are seen coming from the lateral ventricle (upper left) and third ventricle (upper right). (D) The radial glial fibers located along the third ventricle showed a slight ventral slope as they projected from medial to lateral. Scale bar 1000 µm for A; 125 µm for C; and 250 µm for D. identify the timing of the dispersion that occurs in the majority cells disperse. To examine one possible mechanism, the archi- of clones reported in the current study, which should clarify tecture of the radial glia in the diencephalon was explored the relationships of the patterns reported in all of the afore- using DiI labeling. The labeling of radial glial fibers arising mentioned studies from the third ventricle revealed that fibers projected into the diencephalon from the medial-to-lateral direction, with a Clonal dispersion and the generation of dorsal-to-ventral slope, paralleling the most common patterns diencephalic nuclei of distribution within clones. It is worth noting that not all Data from birthdating studies using [3H]thymidine have been clones showed this pattern of dispersion. Several clones used as the basis for a proposal of gradients of cell generation showed marked displacement in the dorsal-ventral plane with and other properties of the germinal zone of the third ventricle. relatively little dispersion from medial to lateral. This suggests A study in Xenopus led to a proposal of an overall gradient that other mechanisms for clonal dispersion in the dorsal- from ventral-lateral to dorsal-medial in the generation of ventral plane exist in the diencephalon. In addition, a neurons in the diencephalon (Tay and Straznicky, 1982). In mechanism that could explain the marked anterior-to-posterior mammals, where the majority of work on the diencephalon has dispersion of clonally related cells is currently unknown. been conducted, a detailed analysis of the birthdates of each nucleus or small cluster of nuclei has been reported (Angevine, Generation of cell types within clones 1970; Altman and Bayer, 1978a-c, 1979a-c, 1988a-c, 1989a- The majority of large clones and approximately 50% of small c). Together these studies have indicated that some nuclei clones contained more than one cell type. This indicates that within the diencephalon have fairly short periods of genesis progenitor cells are frequently multipotent, capable of while the neurons of other nuclei are born over a longer period producing both neurons and glia, similar to progenitors within of time. Furthermore, the birthdays of particular nuclei did not the rodent retina (Turner and Cepko, 1987; Turner et al., 1990) precisely correlate with the anatomical location of the nuclei. and telencephalon (Price and Thurlow, 1988; Walsh and These birthdating studies led to the hypothesis that the Cepko, 1992; Levison et al., 1993; Reid et al., 1995) and the germinal epithelium is a mosaic of patches, each giving rise to chick tectum (Galileo et al., 1990; Gray and Sanes, 1992). The neurons in a nucleus or specific region of the diencephalon distribution of cell types indicates that most progenitors (Altman and Bayer, 1988a). The clonal distributions reported generate relatively small numbers of neurons and large here included clones where the sibling cells tended to cluster numbers of glia (see Table 2). This was not always the case, in one or a small group of nuclei of the diencephalon. These as several clones had large numbers of neurons with relatively clones would be consistent with the [3H]thymidine data, sup- few, or no, glia. The large number of glia in many clones, along porting the idea that progenitors within the ventricular zone of with the small numbers of neurons in some of these same the third ventricle generate specific regions in the dien- clones, is consistent with a multipotent progenitor that divides cephalon. However, other members of some of these clones as and first gives rise to one neuron at each cell division. Multiple well as many other clones were widely dispersed, indicating cell divisions with one neuronal daughter at each division that most progenitors are not dedicated to producing cells for would account for the clones with multiple neurons. One or one nucleus or a small group of nuclei. more mitotic daughters from these same progenitors could Since siblings did not strictly populate one region of the continue to proliferate and later produce one or many glial diencephalon, we wanted to investigate how clonally related cells. Analysis of single cell clones (see below) supports this 76 J. A. Golden and C. L. Cepko order of genesis, as does [3H]thymidine birthdating (Angevine, identification of cells within each of these commissures. 1970). The observation of two cell clones with a radial glial However, as yet, clonally related cells on each side of the dien- cell and a neuron might suggest that progenitor cells remain cephalon and within one of these commissures have not been multipotent up to the last cell division, although cell death observed within the infected brains in the current data set. A and/or inefficiency in amplification or sequencing could result second pathway to the generation of bilateral clones is for cells in missing cells from such clones. to migrate around the anterior, inferior or posterior limits of Analysis of the single cell clones also provides insights into the diencephalon (Fig. 6B, cell A). We have not seen this type the timing of genesis of different cell types in the diencephalon. of migration in the material that we have examined, but a more The 118 single cell clones could be derived by one of several extensive examination of younger brains would be required to mechanisms, which are not mutually exclusive. As retroviruses exclude these routes. A third possibility is that a population of integrate during the M phase of the cell cycle, integration into progenitor cells exists along the midline that is capable of gen- a single chromosome during M-phase means that only one erating siblings that can migrate to populate both sides of the daughter cell from the first cell division will be marked (Roe diencephalon (Fig. 6B, cell B). Similar midline cells are et al., 1993). If this daughter cell does not re-enter the cell present in invertebrates (Crews et al., 1988; Nambu et al., cycle, it will result in a 1-cell clone. Alternatively, clones of 1990, 1991; Crews et al., 1992), and in vertebrates such as the greater than 1 cell would appear as single cell clones if some zebrafish (Hatta et al., 1991), and have been proposed as the siblings were AP−, died, did not amplify with PCR, or did not source of bilaterally symmetrical clones in the chick dien- sequence. However, it appears that no particular cell type cephalon seen at E8 (Arnold-Aldea and Cepko, 1995). We amplified or sequenced preferentially and thus these potential have no direct evidence from the current data set for this problems would not bias the data set. An estimation of the mechanism of generating bilateral clones. number of clones that contained AP+ cells and AP− cells is One rare subset of clones had their origins in the lateral approximately 5% (see Results and unpublished data). ventricle and descendants in the diencephalon. Radial glia with Therefore, single cell clones can be interpreted as an approxi- their cell bodies in the wall of the lateral ventricle and mation of cell birthdays. processes radiating medially into the diencephalon were 27% of all single cell clones were neurons, indicating that observed to have sibling cells, both glial and neuronal, within neurons were being born near the time of injection of the retro- the body of the diencephalon. Several other studies have virus at stage 16-17 in development. No single cell clones of suggested that progenitors from the ventricular zone of the neurons were found in the two brains analyzed from the stage lateral ventricle could provide cells to specific nuclei in the 10-12 injections, but only 9 clones were analyzed, making the diencephalon (Rakic and Sidman, 1969; Altman and Bayer, data set too small to permit a definite conclusion as to whether 1978b) Although no specific nuclei were populated by the neurons were being born at this time. In contrast, glial cells clones originating in the lateral ventricles, the few clones were rarely born at the stages of development that injections observed in this study preclude a definitive analysis. were performed. Despite their high frequency in the total data Surprisingly, clonally related cells in the diencephalon and set, only 6% of the single cell clones comprised glial cells the mesencephalon were observed. Since the first cerebral (other than radial glia). This is consistent with classical vesicles that form are the prosencephalon, the mesencephalon [3H]thymidine birthdating studies in the diencephalon of many and the rhombencephalon, the presence of clonally related cells other species (Altman and Bayer, 1979a-c, 1988b,c, 1989a-c; in the diencephalon and mesencephalon indicates that even Tay and Straznicky, 1982). The fact that there were any single these very early segments of the nervous system do not form cell glial clones is of some interest. The siblings of these cells absolute lineage boundaries. Such clones are undoubtably may have been missed due to the inefficiency of the PCR more frequent than reported here as AP+ cells in all regions of and/or sequencing, or death. Alternatively, the single cell glial the brain were not systematically analyzed. However, if they clones may have arisen from radial glia. Several lines of are rare, it is not clear what meaning one can ascribe to them. evidence have indicated that radial glia transform into astro- They could simply be due to progenitor cells that are situated cytes later in development (Schmechel and Rakic, 1979; Levitt exactly on the border between the two areas. Alternatively, and Rakic, 1980; Pixley and De Vellis, 1984; Voigt, 1989; they could play a more meaningful role in patterning or con- Cullican et al., 1990; Gray and Sanes, 1992). A large percent- nections between brain areas. AP+ cells were not sampled from age of radial glial clones were in fact single cell clones, indi- the telencephalon and therefore no clones would have been cating that they were born near the time of the retroviral identified that crossed from the diencephalon into the telen- infection, consistent with birthdating of radial glia in other cephalon. However, clones were found in the hypothalamus studies (Levitt and Rakic, 1980; Misson et al., 1988). and thalamus. In the model where prosomeres form the segments of the brain, the thalamus and hypothalamus are in Unusual clone types distinct prosomeres, including the hypothalamus arising from Several unusual clone types were identified in this study. One the ventral telencephalon (Puelles and Rubenstein, 1993; type was bilateral, with sibling cells on either side of the third Rubenstein et al., 1994; Rubenstein and Puelles 1994). Clones ventricle, but unlike the bilateral clones observed in the dien- that occupy both the thalamus and hypothalamus thus would cephalon from another study (Arnold-Aldea and Cepko, 1995), cross the diencephalon-telencephalon boundary if the model the bilateral clones in this study were not symmetric. At least proposed for prosomeres held for the chick. three possible migration pathways could result in bilateral In summary, lineage analysis has been conducted in the clones. A cell could cross through one of the major com- diencephalon of the chick using a complex retroviral library. missures (see Fig. 6), which include the anterior, posterior and We have characterized the patterns of dispersion of clones and supraoptic decussation dorsalis. Analysis at E18 has led to found several novel distributions. Clones were found to spread Clones in the chick diencephalon 77 in all directions. Dispersion of clones from medial to lateral related cells during development of the ventral forebrain. Dev. Biol. in and from dorsal to ventral was found to parallel the pattern of press. radial glial fibers, suggesting that migration along radial glial Austin, C. P. and Cepko, C. L. (1990). Cellular migration patterns in the developing mouse cerebral cortex. Development 110, 713-732. fibers is one mechanism for clonal dispersion. Furthermore, Bergquist, H. (1952). Studies on the cerebral tube in vertebrates. The siblings were found to occupy multiple nuclei, including nuclei neuromeres. Acta Zool. 33, 117-187. derived from more than one diencephalic neuromere. Charac- Birgbauer, E. and Fraser, S. (1994). Violation of cell lineage restriction terization of the cell types within clones demonstrated a high compartments in the chick hindbrain. Development 120, 1347-1356. frequency of clones containing neurons and glia, supporting Bulfone, A., Kim, H. J., Puelles, L., Porteus, M. H., Grippo, J. F. and Rubenstein, J. L. (1993a). The mouse Dlx-2 (Tes-1) gene is expressed in the hypothesis that progenitor cells in the chick diencephalon spatially restricted domains of the forebrain, face and limbs in midgestation are multipotential. In addition to the common clones, unusual mouse embryos [published erratum appears in Mech Dev 1993 clones showed bilateral dispersion, origins in the lateral Aug;42(3):187]. Mech. Dev. 40, 129-140. ventricle and/or siblings in both the diencephalon and mesen- Bulfone, A., Puelles, L., Porteus, M., Frohman, M., Martin, G. and Rubenstein, J. (1993b). Spatially restricted expression of Dlx-1, Dlx-2 (Tes- cephalon. The significance of these rare patterns of dispersion 1), Gbx-2, and Wnt-3 in the embryonic day 12.5 mouse forebrain defines remains unknown. potential transvere and longitudinal segmentation boundaries. J Neuroscience 13, 3156-3172. We would like to thank Suzanne Bruhn, John Lin and Francis Szele Crews, S., Franks, R., Hu, S., Matthews, B. and Nambu, J. (1992). for critically reviewing the manuscript and for making many helpful Drosophila single-minded gene and the molecular genetics of CNS midline suggestions. We are grateful to Julie Zitz for technical assistance. development. J. Exp. Zool. 261, 234-244. Supported by grants from the NIH (NS-01664 to J. A. G.; NS-23021 Crews, S. T., Thomas, J. B. and Goodman, C. S. (1988). The drosophila to C. L. C.) and the Howard Hughes Medical Institute. single-minded gene encodes a nuclear protein with sequence similarity to the per gene product. Cell 52, 143-151. Cullican, S., Baumrind, N., Yamamoto, M. and Pearlman, A. (1990). Cortical radial glia: identification in tissue culture and evidence for their REFERENCES transformation into astrocytes. J Neurosci. 10, 684-692. Fekete, D. and Cepko, C. (1993). Replication-competent retroviral vectors Altman, J. and Bayer, S. A. (1978a). Development of the diencephalon in the encoding alkaline phosphatase reveal spatial restriction of viral gene rat. I. Autoradiographic study of the time of origin and settling patterns of expression/transduction in the chick embryo. Mol Cell Biol. 13, 2604- neurons of the hypothalamus. J. Comp. Neurol. 182, 945-971. 2613. Altman, J. and Bayer, S. A. (1978b). Development of the diencephalon in the Fekete, D., Perez-Miguelsanz, J., Ryder, E. and Cepko, C. (1995). Clonal rat. II. Correlation of the embryonic development of the hypothalamus with analysis in the chicken retina reveals tangential dispersion of clonally related the time of origin of its neurons. J. Comp. Neurol. 182, 973-993. cells. Dev. Biol. In press. Altman, J. and Bayer, S. A. (1978c). Development of the diencephalon in the Figdor, M. and Stern, C. (1993). Segmental organization of embryonic rat. III. Ontogeny of the specialized ventricular linings of the hypothalamic diencephalon. Nature 363, 630-634. third ventricle. J. Comp. Neurol. 182, 995-1015. Frank, E. and Sanes, J. R. (1991) Lineage of neurons and glia in chick dorsal Altman, J. and Bayer, S. A. (1979a). Development of the diencephalon in the root ganglia: analysis in vivo with a recombinant retrovirus. Development rat. IV. Quantitative study of the time of origin of neurons and the 111, 895-908. internuclear chronological gradients in the thalamus. J. Comp. Neurol. 188, Frantz, G., Bohner, A., Akers, R. and McConnell, S. (1994a). Regulation of 455-471. the POU domain gene SCIP during cerebral cortical development. J. Altman, J. and Bayer, S. A. (1979b). Development of the diencephalon Neurosci. 14, 472-485. in the rat. V. Thymidine-radiographic observations on internuclear Frantz, G., Weimann, J., Levin, M. and McConnell, S. (1994b). Otx1 and and intranuclear gradients in the thalamus. J. Comp. Neurol. 188, 473- Otx2 define layers and regions in developing cerebral cortex and cerebellum. 499. J. Neurosci. 14, 5725-5740. Altman, J. and Bayer, S. A. (1979c). Development of the diencephalon in Fraser, S., Keynes, R. and Lumsden, A. (1990). Segmentation in the chick the rat. VI. Re-evaluation of the embryonic development of the thalamus on embryo hindbrain is defined by cell lineage restrictions. Nature 344, 431- the basis of thymidine-radiographic datings. J. Comp. Neurol. 188, 501- 435. 524. Galileo, D., Gray, G., Owens, G., Majors, J. and Sanes, J. (1990). Neurons Altman, J. and Bayer, S. A. (1988a). Development of the rat thalamus: I. and glia arise from a common progenitor in chicken optic tectum: Mosaic organization of the thalamic Neuroepithelium. J. Comp. Neurol. 275, demonstration with two retroviruses and cell type-specific antibodies. Proc. 346-377. Natl Acad. Sci. USA 87, 458-462. Altman, J. and Bayer, S. A. (1988b). Development of the rat thalamus: II. Golden, J. and Cepko, C. (1995). Lineage analysis using retroviral vectors. Time and site of origin and settling pattern of neurons derived from the American Zoologist In press. anterior lobule of the thalamic neuroepithelium. J. Comp. Neurol. 275, 378- Golden, J., Fields-Berry, S. and Cepko, C. (1995). Construction and 405. charaterization of a highly complex retroviral library for lineage analysis. Altman, J. and Bayer, S. A. (1988c). Development of the rat thalamus: III. Proc. Natl Acad. Sci. USA 92, 5704-5708. Time and site of origin and settling pattern of neurons of the reticular nucleus. Gray, G. and Sanes, J. (1992). Lineage of radial glia in the chicken optic J. Comp. Neurol. 275, 406-428. tectum. Development 114, 271-283. Altman, J. and Bayer, S. A. (1989a). Development of the rat thalamus: IV. The Guthrie, S. (1995). The status of the neural segment. Trends in Neurosciences intermediate lobule of the thalamic neuroepithelium, and the time and site of 18, 74-79. origin and settling pattern of neurons of the ventral nuclear complex. J. Guthrie, S. and Lumsden, A. (1991). Formation and regeneration of Comp. Neurol. 284, 534-566. rhombomere boundaries in the developing chick hindbrain. Development Altman, J. and Bayer, S. A. (1989b). Development of the rat thalamus: V. The 112, 221-230. posterior lobule of the thalamic neuroepithelium and the time and site of Guthrie, S. and Lumsden, A. (1992). Motor neuron pathfinding following origin and settling pattern of neurons of the medial geniculate body. J. Comp. rhombomere reversals in the chick embryo hindbrain. Development 114, Neurol. 284, 567-580. 663-673. Altman, J. and Bayer, S. A. (1989c). Development of the rat thalamus: VI. The Guthrie, S., Muchamore, I., Kuroiwa, A., Marshall, H., Krumlauf, R. and posterior lobule of the thalamic origin and settling pattern of neurons of Lumsden, A. (1992). Neuroectodermal autonomy of Hox-2.9 expression the lateral geniculate and lateral posterior nuclei. J. Comp. Neurol. 284, 581- revealed by rhombomere transpositions. Nature 356, 157-159. 601. Halliday, A. L. and Cepko, C. L. (1992). Generation and migration of cells in Angevine, J. B. (1970). Time of neruon origin in the diencephalon of the the developing striatum. Neuron 9, 15-26. mouse: an autoradiographic study. J. Comp. Neurol. 139, 129-188. Hamburger, V. and Hamilton, H. (1951). A series of normal stages in the Arnold-Aldea, S. and Cepko, C. L. (1995) Dispersion patterns of clonally development of the chick embryo. J Morphol. 88, 49-91. 78 J. A. Golden and C. L. Cepko

Hatta, K., Kimmel, C., Ho, R. and Walker, C. (1991). The cyclops mutation and mature astrocytes studied with a monoclonal antibody to vimentin. Dev. blocks specification of the floor plate of the zebrafish central nervous system. Brain Research 15, 201-209. Nature 350, 339-341. Price, J. and Thurlow, L. (1988). Cell lineage in the rat cerebral cortex: Herrick, C. (1910). The morphology of the forebrain in amphibia and reptilia. a study using retroviral-mediated gene transfer. Development 104, 473- J. Comp. Neurol. 20, 215-348. 482. His, W. (1893). Vorschlage zur einteilung des gehirms. Arch. Anat. Price, J., Turner, D. and Cepko, C. (1987). Lineage analysis in the vertebrate Entwicklungsgesch. 17, 172-179. nervous system by retrovirus-mediated gene transfer. Proc. Natl Acad. Sci. Hunt, P., Gulisano, M., Cook, M., Sham, M.-H., Faiella, A., Wilkinson, D., USA 84, 156-160. Boncinelli, E. and Krumlauf, R. (1991). A distinct Hox code for the Puelles, L., Amat, J. and Martinez-de-la-Torre, M. (1987). Segment-related, branchial region of the vertebrate head. Nature 353, 861-864. mosaic neurogenetic pattern in the forebrain and mesencephalon of early Jones, E. G. (1985). The Thalamus. New York: Plenum Press. chick embryos: I. Topography of AChE-positive neuroblasts up to stage Kappers, C. U. A., Huber, G. C. and Crosby, E. C. (1960). The Comparitive HH18. J. Comp. Neurology 266, 247-268. Anatomy of the Nervous System in Vertebrates. New York: MacMillan Co. Puelles, L. and Rubenstein, J. L. (1993). Expression patterns of homeobox Keynes, R. and Lumsden, A. (1990). Segmentation and the origin of regional and other putative regulatory genes in the embryonic mouse forebrain diversity in the vertebrate central nervous system. Neuron 4, 1-9. suggest a neuromeric organization. Trends in Neurosciences 16, 472-479. Keyser, A. (1972). The development of the diencephalon of the Chinese Puelles, L. and Zabala, C. (1982). Evidence for a concealed neuromeric hamster. Acta Morphologica Neerlando-Scandinavica 9, 379. pattern in the disposition of retinorecipient grisea in the avain diencephalon. Khulenbeck, H. (1973). Overall morphologic pattern. In The Central Nervous Neurosci. Lett. Suppl. 10, 396-397. System of Vertebrates. Basel: Karger. Rakic, P. and Sidman, R. (1969). Telencephalic origin of pulvinar neurons in Kuenzel, W. and Masson, M. (1988). A stereotaxic atlas of the brain of the the fetal human brain. Zeitschrift fur Anatomie und Entwicklungsgeschichte chick (Gallus Domesticus). Baltimore, MD: The Johns Hopkins University 129, 53-82. Press. Reid, C. B., Liang, I., and Walsh, C. (1995) Systematic widespread clonal Leifer, D., Golden, J. and Kowall, N. (1994). Myocyte-specific enhancer organization in cerebral cortex. Neuron 15, 299-310. binding factor 2C expression in human brain development. Neuroscience 63, Roe, T., Reynolds, T., Yu, G. and Brown, P. (1993). Integration of murine 1067-1079. leukemia virus DNA depends on mitosis. EMBO J. 12, 2099-2108. Levison, S., Chuang, C., Abramson, B. and Goldman, J. (1993). The Rubenstein, J. L., Martinez, S., Shimamura, K. and Puelles, L. (1994). The migrational patterns and developmental fates of glial precursors in the rat embryonic vertebrate forebrain: the prosomeric model. Science 266, 578- subventricular zone are temporally regulated. Development 119, 611-622. 580. Levitt, P. and Rakic, P. (1980). Immunoperoxidase localization of glial Rubenstein, J. L. and Puelles, L. (1994). Homeobox gene expression during fibrillary acidic protein in radial glial cells and astrocytes of the developing development of the vertebrate brain. Current Topics in Developmental rhesus monkey brain. J. Comp. Neurol. 193, 815-840. Biology 29, 1-63. Lumsden, A. (1990). The cellular basis of segmentation in the developing Ryder, E. F. and Cepko, C. L. (1994). Migration patterns of clonally related hindbrain. Trends in Neurosciences 13, 329-335. granule cells and their progenitors in the developing chick cerebellum. Lumsden, A., Clarke, J., Keynes, R. and Fraser, S. (1994). Early phenotypic Neuron 12, 1011-28. choices by neuronal precursors, revealed by clonal analysis of the chick Schmechel, D. and Rakic, P. (1979). A Golgi study of radial glial cells in embryo hindbrain. Development 120, 1581-1589. developing monkey telencephalon: morphogenesis and transformation into Lumsden, A. and Keynes, R. (1989). Segmental patterns of neuronal astrocytes. Anat. Embryol. 156, 115-152. development in the chick hindbrain. Nature 337, 424-428. Simeone, A., Acampora, D., Gulisano, M., Stornaiuolo, A. and Boncinelli, Martinez, S. and Alvarado-Mallart, R-M. (1989). Transplanted E. (1992). Nested expression domains of four homeobox genes in developing mesencephalic quail cells colonize selectively all primary visual nuclei of rostral brain. Nature 358, 687-690. chick diencephalon: a study using heterotopic transplants. Dev. Brain Tay, D. and Straznicky, C. (1982). The development of the diencephalon Research 47, 263-274. in Xenopus. An autoradiographic study. Anatomy & Embryology 163, 371- Misson, J. P., Edwards, M. A., Yamamoto, M. and Caviness, V. S. J. (1988). 388. Mitotic cycling of radial glial cells of the fetal murine cerebral wall: a Turner, D. L. and Cepko, C. L. (1987). A common progenitor for neurons and combined autoradiographic and immunohistochemical study. Brain glia persists in rat retina late in development. Nature 328, 131-136. Research 466, 183-190. Turner, D. L., Snyder, E. Y. and Cepko, C. L. (1990). Lineage-independent Nambu, J. R., Franks, R. G., Hu, S. and Crews, S. T. (1990). The single- determination of cell type in the embryonic mouse retina. Neuron 4, 833-845. minded gene of Drosophila is required for the expressionof genes important Voigt, T. (1989). Development of glial cells in the cerebral wall of ferrets: for the development of CNS midline cells. Cell 63, 63-75. direct tracing of their transformation from radial glia into astrocytes. J. Nambu, J. R., Lewis, J. O., Wharton, K. A. and Crews, S. T. (1991). The Comp. Neurol. 289, 74-88. Drosophilia single-minded gene encodes a helix-loop-helix protein that acts Walsh, C. and Cepko, C. L. (1988). Clonally related cortical cells show as a master regulator of CNS midline development. Cell 67, 1157-1167. several migration patterns. Science 241, 1342-1345. O’Leary, D. and Koester, S. (1993). Development of projection neuron types, Walsh, C. and Cepko, C. L. (1992). Widespread dispersion of neuronal clones axon pathways, and patterned connections of the mammalian cortex. Neuron across fuctional regions of the cerebral cortex. Science 255, 434-440. 10, 991-1006. Walsh, C. and Cepko, C. L. (1993). Clonal dispersion in proliferative layers of Orr, H. (1887). Contributions to the embryology of the lizard. J. Morphol. 1, developing cerebral cortex. Nature 362, 632-635. 311-372. Walsh, C., Cepko, C. L., Ryder, E. F., Church, G. M. and Tabin, C. (1992). Parnavelas, J. G., Barfield, J. A., Franke, E. and Luskin, M. B. (1991). The dispersion of neuronal clones across the cerebral cortex (letter). Science Separate progenitor cells give rise to pyramidal and nonpyramidal neurons in 258, 317-320. the rat telencephalon. Cerebral Cortex 1, 463-468. Pixley, S. and De Vellis, J. (1984). Transition between immature radial glia (Accepted 23 October 1995)