We have investigated the role of the homeodomain transcription factor genes En-1 and En-2, homologs of the Drosophila segment polarity geneengrailed, in regulating the development of the retinotopic map in the chick optic tectum. The En proteins are distributed in a gradient along the rostral–caudal axis of the developing tectum, with highest amounts found caudally. Previous evidence suggests that En-1 and En-2 may regulate the polarity of the rostral–caudal axis of the tectum and the subsequent topographic mapping of retinal axons. We have tested this hypothesis by using a recombinant replication-competent retrovirus to overexpress theEn-1 or En-2 genes in the developing tectum. Anterograde labeling with the axon tracer DiI was used to analyze the topographic mapping of retinal axons after the time that the retinotectal projection is normally topographically organized. Overexpression of either En-1 or En-2perturbed the topographic targeting of retinal axons. InEn-infected tecta, nasal retinal axons form an abnormally diffuse projection with numerous aberrant axons, branches, and arbors found at topographically incorrect locations, colocalized with domains of viral infection. In contrast, temporal axons did not form a diffuse projection or discrete aberrant arbors; however, many temporal axons were stunted and ended aberrantly rostral to their appropriate TZ, or in other cases either did not enter the tectum or formed a dense termination at its extreme rostral edge. These findings indicate that En-1 and En-2 are involved in regulating the development of the retinotopic map in the tectum. Furthermore, they support the hypothesis that En genes regulate the polarity of the rostral–caudal axis of the tectum, most likely by controlling the expression of retinal axon guidance molecules.
- axonal guidance
- gene transfer
- neural maps
- recombinant retrovirus
- transcription factors
- visual system development
The projection of retinal ganglion cells to the optic tectum in nonmammalian vertebrates, or to its homolog, the superior colliculus in mammals, has served as a model system for studies of topographic map development. The targeting of retinal axons to their topographically correct termination sites in the tectum is believed to be controlled by guidance molecules distributed in a graded or restricted manner across the tectum (Bonhoeffer and Gierer, 1984;Gierer, 1987; Holt and Harris, 1993; Kaprielian and Patterson, 1994;Roskies et al., 1995). The differential expression of retinal guidance molecules must be regulated by transcription factors (Sanes, 1993).
Several lines of evidence suggest that En-1 andEn-2, two closely related homeodomain transcription factor genes, which are vertebrate homologs of the Drosophilasegment polarity gene engrailed, may regulate the development of the retinotopic map in the tectum, and in particular the polarity of the rostral–caudal tectal axis. First, En-1 andEn-2 are expressed in a caudal-to-rostral gradient in the tectum well before the arrival of retinal axons (Gardner et al., 1988; Martinez and Avarado-Mallart, 1990; Martinez et al., 1991). Levels ofEn-1 and En-2 mRNA and protein are highest in caudal tectum, which is innervated by nasal retinal axons, and lowest in rostral tectum, which is innervated by temporal retinal axons (Gardner et al., 1988; Martinez et al., 1991). Second, the En proteins are the earliest identified molecules to be distributed in a gradient along the rostral–caudal axis of the tectum (Davis and Joyner, 1988;Davis et al., 1988; Gardner et al., 1988). Third, in ectopic tecta formed by mesencephalic transplants in which En is distributed in a normal or inverted gradient, the terminations of retinal axons along the rostral–caudal tectal axis correlate with the relative level of En protein: nasal axons terminate in regions of high En, whereas temporal axons either terminate in regions of low En or are not found in the ectopic tecta (Itasaki et al., 1991; Itasaki and Nakamura, 1992). Although these findings suggest that En may regulate tectal polarity, the correlation between the level of En and rostral–caudal polarity of the retinotopic map in the tectum may be coincidental, and the retinotectal map and the expression of En-1 andEn-2 may be regulated independently by other transcription factors.
One approach to study more directly the role of En in retinotopic map formation would be to analyze the projections of retinocollicular axons in the En-1 and En-2 null-mutant mice, which have been produced by homologous recombination (Joyner et al., 1991; Wurst et al., 1994). Several points, however, would confound such an analysis. En-1 homozygous mutant mice die around birth, days before topographic order develops in the retinocollicular projection. Furthermore, the deletion of En-1 results in a reduced superior colliculus (Wurst et al., 1994), which would complicate interpretations of any aberrancies in early retinal axonal targeting. An additional complication is that En-1 and En-2 seem to be functionally redundant and can compensate for the loss of each other (Joyner et al., 1991; Millen et al., 1994; Hanks et al., 1995).
Therefore, to assess the role of En in the development of the retinotectal map, we have chosen the approach of using a replication-competent recombinant retrovirus to misexpressEn-1 or En-2 in the developing chick tectum, combined with axonal labeling techniques to assay the effect of misexpression on the topographic mapping of retinal axons. The retroviral construct that we have used is based on the RCASBP vector (Hughes et al., 1987), which infects only replicating cells and has been used previously to express the alkaline phosphatase gene in chick tectum (Fekete and Cepko, 1993) and Hox 4.6 in the chick limb bud (Morgan et al., 1992). We find that misexpression of En-1 orEn-2 results in topographically aberrant terminations of nasal and temporal retinal axons. These results indicate that En-1 and En-2 are involved in the regulation of the retinotopic map in the tectum.
This work was presented previously as an abstract at the Annual Society for Neuroscience meeting (Friedman and O’Leary, 1995).
MATERIALS AND METHODS
Viral construction, production, and infection
The mouse En-1 and En-2 cDNAs, with the ATG start sites modified to an NcoI restriction site and subcloned into pClaNco12 shuttle vector, were gifts from A. Joyner and C. Logan. These cDNAs were subcloned into the NcoI andHindIII site of the SLAX shuttle vector (Morgan and Fekete, 1996). A partial ClaI fragment was then cloned into theClaI site of the RCASBP A or B retroviral vector (Hughes et al., 1987; Homburger and Fekete, 1996). The completed expression construct is diagrammed in Figure 1. Subtype 0 chick embryo fibroblasts (CEFs) were transfected with the retroviral constructs by the CaPO4 precipitation and glycerol shock. The transfected cells were propagated and expanded over 10 d in CEF growth medium [DMEM, 10% fetal bovine serum (FBS), 2% chick serum (CS), penicillin/streptomycin (P/S), and glutamine], by which time 100% of the cells were infected. The viral particles were collected in DMEM, 1% FBS, 0.2% CS, P/S, and glutamine over two 24 hr time periods. The viral-conditioned medium was centrifuged for 3 hr in a Beckman SW28 rotor at 21,000 rpm, the viral pellet was resuspended in 0.1 of the original volume, and the viral titer was determined (see Results).
Fertile White Leghorn eggs, subtype 0 (SPAFAS), were incubated at 37.5°C in a 100%-humidified incubator. After 36 hr (approximately stage 10–12), the eggs were windowed, and ∼80 nl of viral concentrate containing 0.025% fast green and 80 μl/ml polybrene (for B envelope virus) was injected into the embryonic mesencephalic vesicle with a forced air picospritzer attached to a micromanipulator. An additional injection was made 12–20 hr later (approximately stage 14). The eggs were sealed with tape and returned to the incubator.
Cells and cryostat sections were fixed in 4% paraformaldehyde (PF) in PMF buffer (0.2 m PIPES, pH 6.95, 2 mmMgSO4, 4 mm EGTA) for 15 min, washed 3 × 5 min in PBS, incubated in 1 m glycine, pH 7.5, followed by 3% H2O2 for 2 min, and washed in PBS. DiI-injected animals were perfused, postfixed overnight in 4% PF, and sectioned on a vibratome. Sections were counterstained with neutral red, dehydrated, and mounted in DPX.
Gag detection. Cells and sections were blocked in 10% FBS in PBT (PBS, 0.1% Triton X-100) for 30 min and incubated for 2 hr at room temperature for cells and overnight at 4°C for sections, with the AMV 3C2 antibody (which recognizes the gag matrix protein P19) (Potts et al., 1987). Cells and sections were washed in PBT and incubated with goat anti-mouse peroxidase-conjugated or fluorescein-conjugated antibodies (Jackson ImmunoResearch Labs, West Grove, PA) for 1 hr at room temperature and then washed 4 × 10 min in PBT. The peroxidase antibody was detected by reacting in 0.03% DAB, 0.03% H2O2 in PBT.
En detection. Cells and sections were blocked in PBSMT (PBS, 2% dry milk, 0.1% Triton X-100) for 30 min and incubated with a 1:1000 dilution αEnhb1 antibody (Davis et al., 1991) (a gift from A. Joyner), which recognizes the homeodomain of En-1 and En-2, in PBSMT overnight at 4°C. The sections were washed 4 × 10 min in PBSMT, incubated 2 hr at room temperature with a goat anti-rabbit antibody (Jackson ImmunoResearch Labs), washed 2 × 10 min in PBSMT and 2 ×10 min in PBT, and reacted in DAB as described above.
In situ hybridization
cRNA probes were synthesized from NcoI-linearizedEn-1 or En-2 SLAX constructs. One microgram of linearized plasmid was mixed with T7 RNA polymerase (10 U) (New England Biolabs, Beverly, MA) in 1× transcription buffer (40 mmTris HCl, pH 8.25, 6 mm MgCl2, 2 mmspermidine, 10 mm dithiothreitol), ribonuclease inhibitor (Promega, Madison, WI), and nucleotides [1 mm each of GTP, ATP, and CTP, 0.65 mm UTP, and 0.35 mmdigoxygenin-UTP (Boehringer Mannheim, Indianapolis, IN)]. The transcription reaction was incubated at 37°C for 2 hr. The RNA transcript was purified by incubation with 2 μl DNaseI (RNase-free, Promega) at 37°C for 15 min, followed by precipitation with 0.4 m LiCl in 70% EtOH at −20°C for 1 hr. The precipitate was resuspended in 5 mm Tris HCl, pH 8.0, 0.5 mm EDTA, and 50% formamide.
Twenty micrometer cryostat sections were thaw-mounted and air-dried on 3-aminopropyltriethoxysilane (Sigma, St. Louis, MO) subbed slides, fixed for 10 min in 4% PF in PBS, washed 3 × 3 min in PBS, acetylated in 0.25% (v/v) acetic anhydride in 0.1 mtriethanolamine, permeabilized in 1% Triton X-100/PBS, and washed 3 × 5 min in PBS. The sections were then blocked for 6 hr in hybridization buffer [50% formamide, 4 × SSC (0.6 mNaCl, 60 mm sodium citrate, pH 7.0, 4 mm EDTA] and 1× Denhardt’s solution [0.02% Ficoll, 0.02% polyvinylpyrrolidone, 0.02% BSA (Fraction V), 20% dextran sulfate, 500 μg/ml tRNA]. The sections were hybridized with 1 mg/ml digoxygenin-labeled cRNA in hybridization buffer for 16 hr at 65°C. The sections were then washed in 0.2× SSC for 1 hr at 72°C, washed 2 × 10 min in PBS, blocked with 10% heat-inactivated lamb serum in PBT, and then incubated with a 1:2000 dilution of Dig-specific antibody (Boehringer Mannheim) for 2 hr. The sections were then washed 3 × 10 min in PBS, equilibrated in AP buffer (0.1 mTris, pH 9.5, 0.1 m NaCl, 0.05 mMgCl2, 0.1% Tween 20, 2 mm Levamisol) for 5 min, and then reacted with 0.45 mg/ml 4-Nitroblue tetrazolium chloride (Boehringer Mannheim), 0.175 mg/ml 5-Bromo-4-chloro-3-indolyl-phosphate (X-phosphate, Boehringer Mannheim) in AP buffer for 16 hr at room temperature.
Axonal labeling and analysis
DiI was used as an anterograde axon tracer (Honig and Hume, 1986, 1989) to label retinal axons projecting to the tectum. Injections and analysis of labeling were carried out as described in detail byNakamura and O’Leary (1989). Briefly, focal injections of a 10% solution of DiI (Molecular Probes, Eugene, OR) in dimethylformamide (Sigma) were made into the peripheral nasal or temporal retina on the 15th day of incubation. The embryos were incubated for an additional 2–3 d, perfused with 4% PF in 0.1 m phosphate buffer, and staged according to the criteria of Hamburger and Hamilton (1951). Whole mounts of the retina and the contralateral optic tectum were examined and photographed on a fluorescence microscope under rhodamine illumination to document the location of the DiI injection site and the distribution of DiI-labeled retinal axons. In addition, the location of the retinal injection sites was mapped onto drawings of the retina, and the trajectories and arborizations of labeled retinal axons in the tectum were plotted on tracings of the tecta. The tecta were then processed for the expression of the virally expressed genes in one of two ways. Whole mounts either were directly immunostained with the AMV 3C2 antibody for gag protein or were embedded in agar, vibratome-sectioned parallel to the rostral–caudal tectal axis, and then immunostained for gag protein. In some cases, sections through the tectal whole mounts were photographed under rhodamine fluorescence before immunostaining to document the distribution of DiI-labeled axons and arbors. Aberrant mapping of retinal axons in tectal whole mounts and sections was correlated with distribution of gag protein and the retinal location of the DiI injection site. Photomontages and figures were constructed with Adobe Photoshop.
Expression of virally introduced genes in vitro and in vivo
Figure 1 shows schematics of the replication-competent recombinant retroviruses used and their putative transcripts. Viral titer was determined by immunocytochemistry with either AMV 3C2 (Potts et al., 1987), which recognizes a viral gag protein, or αEnhb1 (Davis et al., 1991), which recognizes the homeodomain of the En-1 and En-2 protein, or by in situhybridization with En-1 or En-2digoxygenin-labeled cRNAs. Essentially identical titers were detected by immunohistochemistry and in situ hybridization for the recombinant retroviruses used. For the En-1 RCASBP envB construct, a titer of 2.7 × 108 IU/ml was detected with both the gag and En antibodies, and 2.5 × 108IU/ml with cRNA probes. For the En-2 RCASBP envA construct, a titer of 7.7 × 107 IU/ml was detected with both antibodies, and 5.0 × 107 IU/ml with cRNA probes. In addition, as a control for the effect of viral infection, we used both an RCASBP envB construct and an RCASBP envA construct, each of which contained the human placental alkaline phosphatase (hPLAP) cDNA in place of an En cDNA. The PLAP RCASBP envB and envA viruses had titers of 1 × 108 IU/ml and 1 × 109 IU/ml, respectively, as detected by alkaline phosphatase histochemistry. These findings demonstrate the effectiveness of retroviral infection and indicate that the retrovirally introduced genes are coexpressed in cells.
En is first evident in the tectum on E2 (Gardner et al., 1988; Gardner and Barald, 1991), and tectal neurogenesis occurs from E4 to E9 (La Vail and Cowan, 1971). Therefore, for the in vivo analysis of the influence of En on retinotectal mapping, embryos were infected at E2 (stage 10–12) with recombinant retrovirus containingEn-1 or En-2 cDNA. To enhance the level of infection, about half of these embryos were reinfected at E3 (stage 14). In an initial set of in vivo experiments, we assessed the coexpression of the En-1 (n = 4) andEn-2 (n = 4) genes and the gaggene at E4/E5 (stages 24–27), ∼2 d before retinal axons first enter the tectum (Crossland et al., 1975; McLoon, 1985). At E4, the caudal to rostral gradient of endogenous En protein is readily demonstrable (Fig.2 A,B). Coexpression was determined by detecting En mRNA by in situ hybridization, and the En protein and a gag protein were determined by immunohistochemistry. Retrovirally infected cells and their progeny form dense columns of cells spanning from the ventricular to pial surfaces of the tectum, with a scattering of infected cells outside of these columns (Figs. 2 C–E; also see Fig. 5). These columnar patterns were anticipated, because the replication-competent retrovirus infects only proliferating cells and the retrovirally introduced genes are passed on to their progeny. This labeling pattern is similar to that reported by others using replication-incompetent, recombinant retroviruses to mark tectal progenitor cells and their progeny (Gray et al., 1988; Galileo et al., 1990; Gray and Sanes, 1991). We find thatEn mRNA, En protein, and gag protein are coexpressed at the same infection site. An example of coexpression in rostral tectum is shown in Figure 2 C–E for an embryo infected with a retrovirus containing En-1 cDNA. Similar findings were obtained for En-2 (data not shown). Thus, infected tectal cells express exogenous En mRNA and its protein product as well as the viral gag protein.
The coexpression of En and gag retrovirally introduced genes allowed us to use immunohistochemical localization of gag protein as a marker for En overexpression in subsequent experiments. One advantage of using gag as a marker for infection is that it is present only in infected cells, whereas both En-1 and En-2 endogenous proteins normally present in the chick tectum are recognized by the αEnhb1 antibody (Fig. 2). Although staining for the endogenous En proteins is low in rostral tectum, caudally the staining is heavy enough to obscure detection of the retrovirally introducedEn.
Effect of overexpression of En genes on the retinotectal map
In additional sets of embryos, we assessed the effects of misexpression of En-1 (using an RCASBP envB retroviral construct; n = 8) and En-2 (using an RCASBP envA retroviral construct; n = 21) on the topographic organization of the retinotectal projection. About half of these embryos were infected at E2 (stage 10–12), and the remaining embryos were infected at both E2 and E3 (stage 14 or 15), with recombinant retrovirus containing En-1 or En-2 cDNA or with virus containing the hPLAP cDNA in place of an EncDNA to control for nonspecific influences of retroviral infection. On the 14th or 15th day of incubation, by which time the retinotectal projection has taken on its mature-like topographical organization (Crossland et al., 1975; Thanos and Bonhoeffer, 1983, 1987; Nakamura and O’Leary, 1989), small injections of DiI were made into peripheral nasal or temporal retina to label retinal axons projecting to the contralateral tectum. On the 17th or 18th day of incubation, the embryos were fixed and staged, and the retinotectal projection and viral infection patterns were analyzed. In addition, the size and location of DiI injections in the retinas, and the projections of labeled retinal axons to the optic fissure, were analyzed. The projection of retinal axons within the retina appeared normal in all cases. Cases were not included if the injection site was not well localized, failed to label axons, or was located at a position not consistent with those in other cases.
To control for the effect of retroviral infection per se on the targeting of retinal axons, embryos were infected with a replication-competent retrovirus (RCASBP envA or envB) containing thehPLAP cDNA, and either nasal retina (n = 4 for envA subtype; n = 4 for envB subtype) or temporal retina (n = 4 for envA subtype) was injected with DiI. In each case, the pattern of retinal axonal terminations is indistinguishable from that seen in normal, aged-matched embryos labeled in the same manner (nasal, n = 3; temporal,n = 2). Examples of control-infected nasal and temporal cases are shown in Figure 3 A,B. Labeled axons arising from either nasal or temporal retina extend rostrocaudally across the tectum, and all terminate in a small, well localized zone of dense arborizations in either caudal or rostral tectum, respectively. Only rarely were arbors located aberrantly outside of the appropriate termination zone (TZ). Histochemical treatment of tissue sections verified that clusters of cells in the infected tecta contain enzymatically functional, exogenous alkaline phosphatase (Fig.3 C,D). These findings indicate that infection with the replication-competent retroviruses does not alter the targeting of retinal axons.
In contrast, the misexpression of En-1 or En-2perturbs the targeting of nasal retinal axons. In these cases, we are able to define a zone of dense arborizations, which likely corresponds to the normal TZ of the labeled axons based on the location of the DiI injection site in the retina. This TZ, however, often appears to be less focused than that observed in normal chicks, with many labeled axons, branches, and arbors found around it or radiating from it (Fig.4 A,E). What is most striking, though, are the numerous mistargeted axons and aberrant arbors. The degree of aberrant targeting and arborization varies among cases. For example, in theEn-2-infected tectum shown in Figure 4 A, several well defined arbors are present at a distance from the dense TZ (Fig.4 B–D), whereas in the En-2-infected tectum shown in Figure 4 E, a substantially greater number of aberrant arbors are present, giving the impression of a much more diffuse projection. Occasionally, we observe small, dense ectopic TZs outside of the normal TZ, as illustrated for the En-2-infected tectum in Figure 4 E. Immunostaining for gag protein, which correlates with domains of exogenous En protein, confirms the effective viral infection in these cases (Fig. 4 F,G). A topographically disorganized projection was found in 8 out of 8 En-1-infected tecta and in 7 out of 11 En-2-infected tecta. Variations in the extent of infection, and the relationship between the domains of Enoverexpression and the trajectories and terminations of the labeled retinal axons, likely account for the variations in the degree of aberrancy in the retinotopic map observed between cases. These findings indicate that overexpression of En leads to an aberrant topographic mapping of nasal retinal axons.
In a subset of cases, we correlated aberrant arbors of nasal axons with domains of En-1 (n = 4) or En-2(n = 4) retroviral infection. This was carried out in one of two ways: one, by photographing the distribution of labeled retinal axons and arbors in tectal whole mounts and then dissecting small regions of the whole mounts containing aberrant projections and processing them for gag immunostaining; or two, by photographing the distribution of labeled retinal axons and arbors in sections of tectum and then processing the sections for gag immunostaining. Figure5 shows an example of a case processed as a whole mount, and Figure 6 shows a case processed as sections. In the whole mount of an En-1-infected tectum illustrated in Figure5, nasal axons form a dense TZ but also show a very diffuse projection with many aberrant arbors present at ectopic locations (Fig.5 A). A collection of aberrant arbors coalescing at a topographically incorrect site (Fig. 5 B) is coincident with a cluster of gag-positive cells indicative of a domain of ectopicEn expression (Fig. 5 C). In the sections of an En-1-infected tectum illustrated in Figure 6, the dense primary TZ is nestled in a site of low gag immunostaining (Fig. 6 A–C), whereas isolated topographically aberrant arbors of nasal axons colocalize with domains of dense gag immunostaining (Fig.6 D–G). In contrast to the aberrant topographic distribution of the nasal axons and arbors, their laminar distribution appears normal. These colocalizations of gag protein and aberrant arbors are evidence of a direct correlation between a site of infection and the presence of aberrant arborizations.
In another set of experiments, we analyzed the effect ofEn-2 overexpression on the targeting of temporal retinal axons. Normally, by E15 temporal axons project to a well defined TZ in the rostral tectum (Nakamura and O’Leary, 1989). In contrast, the projection of temporal retinal axons is aberrant in 7 of the 10 En-2-infected cases examined at E15 or later. In two of theEn-2-infected tecta, we observed labeled temporal axons that are stunted and end at various positions rostral to their appropriate TZ (Fig. 7 A). Such stunted temporal axons are not seen in normal tecta at this age. In five of the seven abnormal cases, labeled temporal axons extend through the optic tract and arborize in pretectal nuclei, and they either form terminations that are restricted to the extreme rostral edge of the tectum abutting the optic tract (Fig. 7 B) or the axons do not extend into the tectum (Fig. 7 C). Again, immunostaining for the expression of the gag gene confirms effective retroviral infection (Fig. 7 D–F) and, in particular, dense domains of infection in rostral tectum. The intercase differences in the targeting and terminations of the labeled temporal axons may be attributable to observed variations in infection between cases and to differences between cases in the probability that labeled axons will encounter ectopic domains of En expression. These findings indicate thatEn overexpression disrupts the projection of temporal retinal axons, but in a manner distinct from nasal retinal axons.
We have used recombinant replication-competent retroviruses to overexpress the En-1 and En-2 homeobox transcription factors in the developing chick tectum to assess their role in regulating the development of the retinotopic map. Overexpression of either En-1 or En-2 perturbed the topographic targeting of nasal and temporal retinal axons. The disruption of nasal retinal axons was evident by a diffuse TZ and numerous axons, branches, and arbors at aberrant topographic locations. The disruption of temporal axon targeting was evident by the presence of stunted axons that failed to reach their appropriate TZ, ending rostral to it, or by the finding in other cases that temporal axons either did not enter the tectum or formed a TZ at its extreme rostral margin. These findings indicate that En-1 and En-2 are involved in regulating the development of the retinotopic map in the tectum. Because En-1 and En-2 are transcription factors, they must act indirectly on axon mapping, most likely by controlling, either directly or through intermediaries, genes that encode retinal axon guidance molecules.
In the developing chick, most retinal axons initially grow past the appropriate rostral–caudal position of their TZ. The majority of arbors are formed by collateral branches that extend from the primary retinal axons (Nakamura and O’Leary, 1989; Yates et al., in press). During normal development, collateral branching and arborization occur at topographically appropriate and inappropriate sites but are densest near the appropriate location. In vitro experiments suggest that branching is controlled by molecules anchored to the membrane of tectal cells by a phosphatidylinositol (PI)-linkage and is distributed differentially along the rostral–caudal tectal axis (Roskies and O’Leary, 1994). The final stage in the development of the topographically ordered projection is the elimination of aberrant axons, branches, and arbors and the consolidation of arborizations at the appropriate TZ (Nakamura and O’Leary, 1989).
We have examined the topographic organization of retinal projections in the En-infected tecta at ages when the projection has taken on its mature topographic order. Before preceding, however, we should consider whether our findings of a diffuse nasal retinal projection in the En-infected tecta might be explained by a slowing of overall development and consequently a delay in the normal elimination of aberrantly located arbors. Several lines of evidence seem to rule this out. First, two controls were carried out to take into account a possible slowing of normal development in the viral-infected chicks. All of the cases were staged according to the Hamburger and Hamilton (1951) criteria, which controls for a slowing of overall development, and the projections in the En-infected cases were extremely diffuse compared with those in age-matched control cases infected with the same parental virus that contained the PLAP gene instead of an En gene. Second, the aberrant arbors formed by nasal axons colocalize with domains of viral infection, which are coincident with domains of En overexpression. And third, although both temporal and nasal axons normally form transient aberrant arbors at an earlier stage of development, we find that temporal axons do not have aberrant arbors in the En-infected cases, whereas the nasal axons do. This distinct effect of En viral infection on temporal and nasal populations cannot be accounted for by a slowing of development.
The aberrant retinal projections in the En-1- andEn-2-infected embryos may be attributable to a stabilization of normally transient aberrant connections or to the promotion of aberrant connections coupled with their stabilization or both. In any case, our findings suggest that En-1 and En-2 regulate the expression of axonal guidance molecules that identify positional addresses in the tectum, and they control the targeting, branching, and arborization of retinal axons.
In Figure 8, we illustrate two possible scenarios that could account for our findings. One scenario is that En-1 or En-2 regulates the expression of a molecule(s) that operates as an attractant, a growth promoter, or a trophic activity specific for nasal retinal axons. In this scenario, misexpression of En-1 orEn-2 would result in ectopic domains of molecules that positively influence the growth of nasal retinal axons resulting in the maintenance of aberrant nasal projections. These aberrant projections could result from a failure of nasal axons to remodel and prune incorrectly positioned axons, branches, and arbors, which normally occurs through competitive interactions between correctly and incorrectly targeted retinal axons (Kobayashi et al., 1990; Simon et al., 1992, 1994; Goodman and Shatz, 1993), or to the promotion and retention of additional aberrant connections. In this scenario, correctly targeted temporal axons may be excluded from domains of En overexpression by competitive interactions with aberrantly terminated nasal retinal axons. Thus, the stunted temporal axons observed in theEn-infected tecta could be in the process of retraction and perhaps elimination. An alternative scenario is that En-1 and En-2 regulate the expression of molecules that specifically inhibit or repulse temporal retinal axons. Temporal axons encountering these negative domains would be inhibited or repulsed and fail to establish appropriate connections within them. This could also account for the aberrancies in the mapping of nasal axons. The exclusion of temporal axons would remove the competition within these domains and allow nasal axons to stabilize normally transient aberrant arbors that project to them. Of course, En-1 and En-2 may regulate both sets of putative molecules (one set positive for nasal axons and one set negative for temporal axons), and the two scenarios presented in Figure 8 may work in concert.
For clarity, these scenarios describe the behaviors of temporal and nasal retinal axons as two uniform populations interacting with each other as well as with guidance molecules; however, these scenarios are an oversimplification, because the temporal-nasal axis of the retina maps smoothly along the rostral–caudal axis of the tectum. It is likely that this mapping continuum is controlled by a graded distribution of guidance molecules—regulated by the graded distribution of En—and differential sensitivities of retinal axons to these molecules, which depends on the point of axon origin along the temporal-nasal axis of the retina. For example, the mapping of nasal retina onto caudal tectum is attributable to differential responses of nasal axons to guidance molecules, depending on the central-to-peripheral point of axon origin within nasal retina. In anEn-infected tectum, variability in the levels of exogenous En would be expected from one overexpression domain to another. This, combined with the graded distribution of endogenous En protein, would likely result in variability in the overall levels of En protein, and in turn retinal guidance molecules, from one overexpression domain to another. This interpretation could account for the aberrant mapping of nasal axons within the caudal tectum of En-infected embryos.
Our results showing that the overexpression of En-1 orEn-2 perturbs the mapping of retinal axons, and the interpretation that this perturbation is attributable to En-induced misexpression of guidance molecules, are consistent with studies in the quail/chick chimera system, which suggest that the En gradient regulates tectal polarity and subsequently retinal axon targeting. When a graft of mesencephalon/metencephalon from an E2 quail is rotated 180° around its rostral–caudal axis and transplanted homotopically into an E2 chick, the En gradient in the graft regulates to coincide with the host En gradient (Martinez and Alvarado-Mallart, 1990). Similarly, the rostral–caudal polarity of the retinotectal projection within such a graft inverts to parallel that of the host (Ichijo et al., 1990). In other studies, the transplantation of E2 or E3 quail mesencephalon into E2 or E3 chick diencephalon results in the formation of an ectopic tectum rostral to the diencephalon/mesencephalon border, which is innervated by retinal axons entering the graft from its original caudal pole (retinal axons normally enter the tectum from its rostral edge) (Itasaki et al., 1991; Itasaki and Nakamura, 1992). The rostral–caudal polarity of the retinotectal projection to the ectopic tectum correlates with its En gradient. Nasal axons project to the region expressing high levels of En. Temporal retinal axons fail to enter the ectopic tectum when the high levels of En are expressed in the caudal pole (the site of axonal entry in the ectopic tectum). These transplantation experiments indicate a correlation between the level of En and the topographic terminations of nasal and temporal retinal axons: nasal axons terminate in regions of high levels of En, and temporal axons terminate in regions of low levels of En.
The correlative evidence provided by the quail/chick chimera transplant experiments and the more direct evidence provided by our recombinant retroviral overexpression experiments lead to the conclusion that En-1 and En-2 regulate the topographic targeting of nasal and temporal retinal axons along the rostral–caudal tectal axis. Furthermore, taken together, these studies suggest that the graded levels of En-1 and En-2 regulate the graded expression of at least a subset of molecules that guide the topographic targeting of retinal axons. For example, an activity preferentially associated with caudal tectum seems to promote or attract the growth of nasal retinal axons in vitro as well as have a trophic effect specific for nasal neurites (von Boxberg et al., 1993). In particular, however, three PI-anchored molecules that are expressed in a caudal-to-rostral gradient in the developing tectum, with highest levels caudally, may have a role in the topographic mapping of retinal axons. One is a 33 kDa protein, termed Repulsive Guidance Molecule, which acts as a repellent to growing temporal retinal axons in vitro (Walter et al., 1987a,b, 1990; Stahl et al., 1990; Müller and Bonhoeffer, 1994; Müller et al., 1995). The other two, Repulsive Axon Guidance Signal (RAGS) (Drescher et al., 1995) and Elf1 (Cheng and Flanagan, 1994; Cheng et al., 1995), are 25 kDa ligands for receptor protein tyrosine kinases of the Eph subfamily. In vitro, the RAGS protein is equally repulsive for both temporal and nasal retinal axons (Drescher et al., 1995), whereas Elf1 is a repellent for temporal axons but not nasal axons (Friedman et al., in press). Interestingly, the overexpression ofEn-1 in rostral tectum results in an increased expression of Elf1 and RAGS (Logan et al., in press). That En regulates the expression of these two molecules which repel temporal axons is consistent with our finding that temporal axons aberrantly end rostral to their normal TZ in En-infected tecta.
It is presently unclear whether the En proteins act directly to regulate the genes encoding retinal axon guidance molecules or indirectly through intermediaries; however, the expression of theEn genes in the tectum seems to be regulated by two distinct domains, one caudally at the mesencephalic/metencephalic border and one rostrally at the diencephalic/mesencephalic border. TheWnt-1 proto-oncogene (Wilkinson et al., 1987; McMahon and Bradley, 1990; Thomas and Capecchi, 1990; Gardner and Barald, 1991;Molven et al., 1991; Bally-Cuif et al., 1992; McMahon et al., 1992;Parr et al., 1993; Bally-Cuif and Wassef, 1994) and thePax-2 paired-box transcription factor (Krauss et al., 1992) expressed at the mesencephalic/metencephalic border seem to be positive regulators of En expression. In contrast, an unidentified molecule(s) present at the diencephalic/mesencephalic border has been implicated as a negative regulator of En (Itasaki and Nakamura, 1992). These regulatory molecules may act in concert to establish the gradient of En across the tectum.
The structure of En genes is conserved across evolutionarily distant species. En-1 and En-2 are homologs of the Drosophila segment polarity genesengrailed and invected (Nusslein-Volhard and Wieschaus, 1980; Poole et al., 1985). Homologs of engrailedhave been cloned at least partially and are expressed in similar patterns in zebrafish (Pattel et al., 1989; Ekker et al., 1992; Fjose et al., 1992), Xenopus (Pattel et al., 1989; Davis et al., 1991; Hemmati-Brivanlou et al., 1991), chick (Gardner et al., 1988; Pattel et al., 1989; Davis et al., 1991; Logan et al., 1992), mouse (Joyner and Martin, 1987; Davis et al., 1991; Logan et al., 1992), and human (Logan et al., 1992). This structural conservation amongengrailed genes is underscored by the finding that the phenotype of midbrain and hindbrain structural deficits seen inEn-1 knockout mice (Wurst et al., 1994) can be partially rescued by the Drosophila engrailed gene placed under the control of the mouse En-1 promoter (M. Hanks and A. Joyner, unpublished observations).
In addition to conservation of structure, the function of En proteins in defining caudal (i.e., posterior) fate of certain tissues seems to be conserved across species as well as by different En proteins within the same species. For example, the caudal midbrain is lacking in mice in which the En-1 gene is deleted by homologous recombination (Wurst et al., 1994). Together, the experimental evidence presented here and that derived from quail/chick transplant experiments (see above) indicate that En proteins regulate the polarity of the rostral–caudal (i.e., anterior–posterior) tectal axis and confer caudal (i.e., posterior) tectal fate. Drosophila engrailedand invected are expressed in the posterior part of each segment of the Drosophila embryo (Kornberg, 1981; Di Nardo et al., 1985). Although a function for invected has not been determined, engrailed is involved in conferring posterior identity to each segment of the Drosophila embryo (Morata and Lawrence, 1975; Nusslein-Volhard and Wieschaus, 1980; Kornberg, 1981; Lawrence and Struhl, 1982; Di Nardo et al., 1985; Brower, 1986). In addition, En-1 and En-2 seem to have a similar role in regulating the retinotopic map (present study). This functional similarity is predicted by gene “knock-in” experiments in mice. For example, the structural deletions in the colliculus and cerebellum inEn-1 knockout mice are rescued when an additional copy of the En-2 gene is placed under the control of theEn-1 promoter (Hanks et al., 1995). This finding supports the suggestion that the lack of an obvious structural defect in the superior colliculus of En-2 homozygous mutants is attributable to compensation by En-1 (Joyner et al., 1991; Millen et al., 1994). In the homozygous mutants, En-1 may be more effective at compensating for the loss of En-2 because En-1 expression begins earlier in development than En-2 (Davis and Joyner, 1988; Davis et al., 1988; McMahon et al., 1992). Taken together, these findings suggest that engrailed-like proteins may be general posterior-defining molecules in the developing tectum of vertebrates and in the segments of Drosophila.
Note added in proof: After the submission of this manuscript to The Journal of Neuroscience, a similar study was published in Neuron by N. Itasaki and H. Nakamura (A role for gradient En expression in positional specification on the optic tectum. Neuron 16:55–62). These investigators also misexpressed En-1 andEn-2 in the developing chicken optic tectum using an RCASBP retrovirus. Although the details differ between our study and theirs, both groups have obtained comparable results.
This work was supported by National Institutes of Health Grant R01 EY07025 and Fellowship F32 EY06550. We thank Martyn Goulding for much valuable advice on vector construction, Alexandra Joyner and Cairine Logan for pClaNco12 shuttle vectors containing murineEn-1 and En-2 cDNAs and theαEn1hb antibody, Connie Cepko and Donna Fekete for RCASBP retrovirus with and without the human PLAP gene, Donna Fekete for valuable advice on their use, and Horst Simon for comments on this manuscript.
Correspondence should be addressed to Dennis D. M. O’Leary, MNL-O, The Salk Institute, 10010 North Torrey Pines Road, La Jolla, CA 92037.