Reelin Is Required for Class-Specific Retinogeniculate Targeting

Identification of nuclei-specific candidate targeting cues

To address how functionally distinct classes of RGC axons target distinct domains within retino-recipient nuclei, we sought to identify the mechanisms regulating class-specific targeting of the mouse LGN—a thalamic nucleus in which classes of image-forming and non-image-forming retinal inputs are segregated into distinct subnuclei. The segregation of class-specific retinal inputs appeared to occur as retinal axons first enter the LGN (supplemental Fig. S1, available at www.jneurosci.org as supplemental material) (Kim et al., 2010), suggesting that molecular cues within LGN subnuclei pattern class-specific retinogeniculate axon targeting. Since there is a paucity of information regarding the factors regulating class-specific targeting of axons in LGN subnuclei, we screened differentially expressed guidance and targeting molecules in either vLGN and IGL (vLGN/IGL) or dLGN. vLGN and IGL were pooled since they receive innervation from similar classes of RGCs (Hattar et al., 2006). RNA was isolated from LGN subnuclei at ages corresponding to the arrival of retinal axons (i.e., P3) (Godement et al., 1984; Jaubert-Miazza et al., 2005) and differentially expressed genes were identified by Agilent microarrays. Numerous genes coding for extracellular matrix molecules and soluble morphogens known to guide axonal growth and targeting were identified as being more than twofold enriched in either vLGN/IGL or dLGN (Fig. 1C). Furthermore, microarray analysis confirmed the differential expression of known transcription factors in either vLGN or dLGN and identified a novel set of LGN subnuclei-specific transcription factors (supplemental Fig. S2, available at www.jneurosci.org as supplemental material) (Nakagawa and O'Leary, 2001; Jones and Rubenstein, 2004).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Reelin is selectively expressed in vLGN and IGL. A, B, Schematic representation of non-image-forming RGC (A) and image-forming RGC (B) axons in LGN subnuclei. C, Differentially expressed guidance and targeting cues in either vLGN/IGL or dLGN, displayed in a heat plot and normalized to the mean intensity of each gene. Color scale depicts fold difference between vLGN/IGL and dLGN. RNA was purified from three sets of vLGN/IGL and dLGN samples, each of which was analyzed by Agilent microarray and is represented by a column in the heat plot. Each sample contained tissue pooled from at least five mice. D, Developmental regulation of differential gene expression in vLGN/IGL compared with dLGN (±SD) assessed by qPCR. RNA was isolated from five ages. In contrast to other genes, the highest levels of reln mRNA enrichment in vLGN/IGL coincided with the arrival of RGC axons. E–G, IHC for reelin on coronal sections of mouse thalamus. IHC confirmed the developmental regulation of reelin in vLGN and IGL during periods of retinogeniculate targeting. The dotted green and white lines outline vLGN and dLGN, respectively. Scale bar: (in G) E–G, 300 μm. d, dLGN; i, IGL; v, vLGN; OT, optic tract.

In the present study, we focus on molecular cues enriched in and necessary for retinal innervation of vLGN/IGL since all retinogeniculate axons encounter these subnuclei (Fig. 1A,B). To narrow our list of vLGN/IGL-enriched targeting genes to those likely to contribute specifically to retinogeniculate targeting, we examined the developmental expression of candidate genes at ages corresponding to RGC axon arrival (i.e., P3), arborization, and refinement (P6–P14) (Jaubert-Miazza et al., 2005). The peak expression of one gene—reln—coincided with the arrival of RGC axons (Fig. 1D). Reln encodes the large glycoprotein reelin, which has previously been shown to contribute to both the regulation of neuronal migration (D'Arcangelo et al., 1995) and axonal growth and targeting (Borrell et al., 1999, 2007; Wu et al., 2008). It is noteworthy that microarray analysis not only revealed an approximately sevenfold enrichment in reln mRNA in vLGN/IGL (compared with dLGN), but also identified an alternatively spliced form of reln mRNA that was more than sixfold enriched in the more ventral LGN subnuclei (data not shown). At P3, reln expression (at both the mRNA and protein level) was highly enriched in vLGN/IGL compared with the dLGN and other retino-recipient nuclei, including the olivary pretectal nucleus (OPN) and suprachiasmatic nucleus (SCN) (Fig. 1D–F; supplemental Fig. S3, available at www.jneurosci.org as supplemental material). The specificity of reelin IHC at these ages was confirmed by demonstrating a lack of immunoreactivity in LGN of mutant mice lacking reelin (i.e., reln^rl/rl mutants) (D'Arcangelo et al., 1995) (supplemental Fig. S3, available at www.jneurosci.org as supplemental material). In contrast to its high level of expression from P0 to P3, reelin expression in the vLGN/IGL was markedly decreased when RGC axons were no longer entering the LGN (Fig. 1D,G). Based on its developmental expression and previously documented functions, reelin was a prime candidate to contribute to class-specific targeting of RGC axons in the LGN.

Most functions of reelin can be attributed to its binding of very low density lipoprotein receptor (VLDLR) or apolipoprotein E receptor 2 (ApoER2) (Benhayon et al., 2003). For reelin to directly influence retinogeniculate targeting, we hypothesized that classes of RGCs must express reelin receptors. RGCs reside within the innermost layer of the retina, the ganglion cell layer (GCL) (Fig. 2A). In situ hybridization revealed that a small subset of cells within this layer expressed mRNAs encoding either VLDLR or ApoER2 (Fig. 2B,C).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Reelin-signaling components are expressed in the GCL of mouse retina. A, ISH for synaptotagmin 1 (syt1) mRNA in P13 wild-type retinal cross-sections demonstrated the distribution of neurons within the GCL. B, C, In contrast to the wide distribution of syt1 mRNA, ISH revealed that apoer2 (B) and vldlr (C) mRNAs were expressed by small subsets of cells within the GCL (see arrowheads in B, C). D, Dab1 IHC in P13 wild-type retinal cross-sections confirmed expression of Dab1 by amacrine cells in the INL (arrows) (Rice and Curran, 2000). Sparse labeling of Dab1 was also observed in a subset of cells in the GCL (arrowhead). E, Colabeling of Meln and Dab1 in P13 wild-type retinal cross-section. The arrowhead depicts Dab1-expressing ipRGC. The arrows indicate Dab1-expressing amacrine cells. The asterisk denotes nonspecific immunoreactivity. F, Reelin IHC in P13 wild-type retinal cross-sections. Many cells within the GCL express reelin; however, the arrowheads indicate cells with no significant reelin expression. G, H, Colabeling of P13 wild-type retinal cross-sections with Meln and reelin antibodies revealed that most ipRGCs do not express reelin (see arrowheads). Meln-expressing RGCs in G and H presumably represent different classes of ipRGCs since there dendrites stratify in different sublaminae of the IPL (see arrows). Both classes of ipRGCs appear to lack reelin expression. I, Reelin IHC was performed on P13 retinal cross-sections from opn4-tau-LacZ transgenic mice. M1 ipRGCs were labeled by LacZ IHC. The arrowhead highlights a M1 ipRGC that lacks reelin expression. For all images, nuclei were labeled by DAPI staining. Scale bar: A–C, E, 15 μm; D, F–I, 20 μm.

After binding to VLDLR or ApoER2, reelin induces the activation of Dab1, a cytoplasmic molecule essential for reelin function (Howell et al., 1997; Sheldon et al., 1997). To assess whether Dab1 was also expressed in subsets of RGCs, IHC was performed on P13 wild-type retinal cross-sections. It is noteworthy that previous reports have documented Dab1 expression in AII amacrine cells within mouse retina (Rice and Curran, 2000; Rice et al., 2001). Our analysis confirmed expression of Dab1 in amacrine cells within the inner nuclear layer (INL) (Fig. 2D,E′, arrows) but also revealed a small subset of cells that expressed Dab1 within the GCL (Fig. 2D,E′, arrowhead). To determine whether these Dab1-expressing cells were RGCs whose axons target vLGN and IGL (i.e., ipRGCs), we colabeled retinal sections for Dab1 and Meln, the functional photopigment expressed by ipRGCs. Indeed, these analyses showed that some ipRGCs expressed Dab1 (Fig. 2E).

As previous reports also suggested that Dab1 and reelin were expressed by mutually exclusive cell types within the retina (Rice and Curran, 2000; Rice et al., 2001), we next examined reelin expression in the GCL of P13 wild-type retina. Reelin expression was detected in many, but not all, cells within the GCL (Fig. 2F). IHC for both reelin and Meln demonstrated that most ipRGCs lacked significant reelin expression [87.8 ± 3.2% (SD) ipRGCs lacked reelin; N = 3 mice, n = 452 ipRGCs] (Fig. 2G,H). Melanopsin IHC labeled several classes of ipRGCs (Fig. 2G,H) (Ecker et al., 2010). Classes of ipRGCs whose dendrites stratified in the OFF-sublaminae of the inner plexiform layer (IPL) (Fig. 2G) and those whose dendrites stratified in the ON-sublaminae of the IPL (Fig. 2H) both lacked reelin expression. To specifically address reelin expression in M1 ipRGCs, a class of ipRGC that projects to the vLGN and IGL (Hattar et al., 2006), reelin IHC was performed on retinal sections from transgenic mice in which M1 ipRGCs were selectively labeled with LacZ (opn4-tau-LacZ mice) (Hattar et al., 2002, 2006). Nearly all M1 ipRGCs lacked significant reelin expression [97.9 ± 0.2% (SD) M1 ipRGCs lacked reelin expression; N = 2 mice; n = 194 RGCs] (Fig. 2I). Therefore, our data show that classes of ipRGCs express Dab1 but not reelin. Together, these studies reveal that some RGCs have the machinery to respond to LGN-derived reelin.

Reelin and its canonical signaling pathway are necessary for RGC targeting of the vLGN and IGL

To assess the role of reelin in retinogeniculate targeting, we analyzed retinal projections in reln^rl/rl mutants. Intraocular injections of fluorescently conjugated CTB were used to label all RGC axons (Fig. 3A) (Muscat et al., 2003; Jaubert-Miazza et al., 2005). Several striking defects were observed in RGC projections in reln^rl/rl mutants. First, the spatial extent of retinal innervation to IGL and vLGN, but not dLGN, was markedly reduced in mutants [45.1 ± 6.5% (SEM) reduced extent of innervation in reln^rl/rl IGL, n = 6 mice, p < 0.05 by Student's t test; 16.4 ± 3.4% (SEM) reduced extent of innervation in reln^rl/rl vLGN, n = 6 mice, p < 0.05 by Student's t test] (Fig. 3A,B; supplemental Fig. S4, available at www.jneurosci.org as supplemental material). Second, numerous misrouted RGC axons were observed in non-retino-recipient thalamic nuclei adjacent to LGN (Fig. 3B,D). These misrouted axons originated from both contralateral and ipsilateral eyes and exited mutant LGN at either the IGL–dLGN border (Fig. 3B,D) or from the vLGN (data not shown). Reduced innervation and misrouting of retinal axons were observed as early as P1 (Fig. 3E,F), suggesting that reelin was necessary for the initial targeting of retinal axons to the vLGN and IGL.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Reelin is necessary for retinogeniculate targeting. A, B, Retinogeniculate projections in P14 control (A) or reln^rl/rl mutant mice (B) assessed by labeling of RGC axons with CTB. The left eyes were injected with Alexa Fluor 594-CTB and right eyes with Alexa Fluor 488-CTB. LGN from right hemispheres are shown. “Contra” denotes retinal arbors originating from cells in the contralateral eye, and “ipsi” denotes those originating from ipsilateral eye. In mutant LGN (B), note the near absence of retinal projections in IGL (arrows), the reduced pattern of vLGN innervation, and the presence of mistargeted axons (asterisk). C and D show high-magnification images of areas depicted by asterisks in A and B, respectively. Projections from both eyes have been merged into a grayscale image for improved visualization [CTB(contra+ipsi)]. E, F, Retinogeniculate projections in P1 control (E) or reln^rl/rl mutant mice (F). RGC axons from both eyes were labeled with the same fluorescently conjugated form of CTB [CTB(b)]. The arrowhead in F highlights mistargeted RGC axons. The arrows indicate IGL. d, dLGN; v, vLGN. Scale bars: (in A) A, B, 400 μm; (in C) C, D, 200 μm; (in E) E, F, 300 μm.

To determine whether reelin required Dab1 function for its role in retinogeniculate targeting, we examined retinal innervation of the LGN in scrambler mutants (dab1^scm/scm), which lack functional Dab1 (Howell et al., 1997; Sheldon et al., 1997). In all dab1^scm/scm mutants examined (n = 10), misrouted RGC axons were observed exiting the LGN medially to enter non-retino-recipient thalamic nuclei (Fig. 4A–C). In contrast to reln^rl/rl mutants, however, only 50% of dab1^scm/scm mutants displayed sparse retinal arbors in the IGL (Fig. 4B). In mutants in which a striking avoidance of the IGL was not observed, retinal arbors in the mutant IGL appeared disorganized compared with controls (Fig. 4A,C). As was observed in reln^rl/rl mutants, defects in the pattern of IGL innervation were observed as early as P2 (Fig. 4D–F). Together, these findings suggest that reelin signals, at least in part, through Dab1 to pattern the initial formation of retinogeniculate circuits.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Dab1 is necessary for retinogeniculate targeting. A, B, Retinogeniculate projections in P14 control (A) or dab1^scm/scm mutant mice (B, C) assessed by labeling of RGC axons with CTB. The left eyes were injected with Alexa Fluor 594-CTB and the right eyes with Alexa Fluor 488-CTB. LGNs from right hemispheres are shown. “Contra” denotes retinal arbors originating from cells in the contralateral eye, and “ipsi” denotes those originating from ipsilateral eye. In mutants, two different retinogeniculate phenotypes were observed. In one set of dab1^scm/scm mutants (B), a near absence of retinal projections was observed in IGL (arrows), whereas in another set, retinal projections were observed in the IGL (C). In both cases (B, C), mutants exhibited a reduced pattern of vLGN innervation and mistargeted axons were observed exiting the LGN (asterisks). A′, B′, and C′ show high-magnification images of areas depicted by asterisks in A–C, respectively. D–F, Retinogeniculate projections in P2 control (D) or dab1^scm/scm mutant mice (E, F). RGC axons from both eyes were labeled with the same fluorescently conjugated form of CTB (CTB[b]). Similar to the analysis at P14, two distinct phenotypes were observed in the pattern of IGL innervation by RGC axons in P2 dab1^scm/scm mutants (E, F, arrows). In both cases, misrouted axons were observed exiting mutant LGN and entering non-retino-recipient thalamic regions (E′, F′). D′, E′, and F′ show high-magnification images of areas depicted by the asterisks in D–F, respectively. The arrows indicate IGL. d, dLGN; v, vLGN. Scale bar, 150 μm.

In addition to its role in regulating axonal growth and targeting, the reelin–Dab1 signaling pathway has well documented roles in neuronal migration and lamination (D'Arcangelo et al., 1995; Howell et al., 1997; Sheldon et al., 1997). In contrast to other regions of the reln^rl/rl mutant brain, few significant alterations in the cytoarchitecture of mutant LGN have been documented. Our analysis confirmed a relatively normal distribution of neurons (by NeuN IHC) in mutant LGN (Fig. 5A,B). A reduction in the size of the LGN in reln^rl/rl mutants was observed, but this difference was not restricted to the LGN and was presumably attributable to mutants being considerably smaller than littermate controls (Falconer, 1951). One defect that has been reported in reln^rl/rl mutant LGN is the presence of aberrant axonal tracts coursing through mutant dLGN (Frost et al., 1986). Axon-tracing studies revealed that such misrouted axon tracts did not arise from the retinal axons but instead were formed by corticothalamic axons originating from layer VI cortical neurons (supplemental Fig. S5, available at www.jneurosci.org as supplemental material). The location of these corticothalamic axons in dLGN and the non-retino-recipient portion of vLGN suggest they are unlikely to contribute to the mistargeting of retinal axons in reln^rl/rl mutant LGN.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

LGN subnuclei are present, remain distinct, and contain appropriate neural populations in the absence of functional reelin. IHC and ISH were performed on coronal sections of P14–P21 reln^rl/rl mutant or littermate control brains with subnuclei- or neuron-specific markers. A, B, NeuN and NPY IHC in P14 reln^rl/rl mutant or control LGN. NPY remained properly restricted to the IGL in the absence of reelin. C, D, GFAP IHC demonstrated the presence and confinement of astrocytes in P21 control and mutant IGL. E, F, IHC revealed a normal distribution of Gad65 in P21 reln^rl/rl mutant vLGN and IGL. G, H, ISH revealed a confinement of syt2-expressing neurons in vLGN of P21 mutants and controls. I, J, SMI32 immunoreactivity was appropriately enriched in vLGN and dLGN, but absent from IGL in P21 mutants and controls. K, L, Adamts15 mRNA, assessed by ISH, was correctly localized in P14 mutant and control dLGN. Sections were costained with NPY antibodies to demonstrate the location of the IGL. Note that NPY-positive cells did not invade dLGN in the absence of functional reelin. In C–J, retinal axons were anterogradely labeled with CTB (CTB[b]). In each image, the IGL is encircled by white dots. d, dLGN; v, vLGN; v_r, retino-recipient portion of vLGN; v_nr, non-retino-recipient portion of vLGN. The asterisks denote corticothalamic axon tracts. Single-color panels for each merged image can be found in supplemental Figure 6 (available at www.jneurosci.org as supplemental material). Scale bar, 400 μm.

We next explored whether defects in retinogeniculate targeting resulted indirectly from an abnormal formation of the three LGN subnuclei in absence of reelin. LGN subnuclei are each derived from different embryonic origins (Altman and Bayer, 1989; Botchkina and Morin, 1995); therefore, to examine whether each were properly generated in the absence of functional reelin, we assayed the distribution of several known subnuclei-specific markers in reln^rl/rl mutant LGN. The IGL, but not the adjacent vLGN or dLGN, contain an abundance of NPY-expressing relay neurons, as well as glial fibrillary acidic protein (GFAP)-expressing astrocytes (Botchkina and Morin, 1995) (Fig. 5A,C). Many interneurons and projection neurons within the vLGN and IGL express Gad65 (Morin et al., 1992; Moore and Speh, 1993) (Fig. 5E). Relay neurons within the vLGN and dLGN, but not IGL, contain an isoform of neurofilament recognized by the antibody SMI32 (Fig. 5I) (Jaubert-Miazza et al., 2005). In addition to these known markers, we identified two novel subnuclei-specific markers. Neurons within the vLGN, but not the IGL or dLGN, express synaptotagmin 2 (syt2) mRNA (Fig. 5G), whereas only neurons within the dLGN express “a disintegrin and metalloproteinase with a thrombospondin type 1 motif member 15” (adamts15) mRNA (Fig. 5K). Immunostaining and in situ hybridization for each of these markers revealed that the three LGN subnuclei form appropriately and remain distinct in reln^rl/rl mutants (Fig. 5). Importantly, NPY-expressing IGL relay neurons, which are the targets of RGC axons (Thankachan and Rusak, 2005), remain confined to the IGL and were not dispersed into inappropriate thalamic regions (Fig. 5A,B,K,L). Together, these data suggest that misrouted axons in reln^rl/rl mutants are in fact mistargeted rather than correctly targeted to neurons that have migrated into inappropriate regions of thalamus.

Reelin is required for the targeting of ipRGCs in the LGN

Defects in the pattern of retinal innervation to reln^rl/rl vLGN and IGL led us to hypothesize that reelin was essential for retinogeniculate targeting by ipRGCs. To ensure that defects in vLGN and IGL targeting were not consequences of intraretinal defects, we examined the distribution and morphology of ipRGCs in reln^rl/rl mutant retina. Whole-mount Meln IHC revealed that the distribution and morphology of ipRGCs were indistinguishable between mutant or control retinas (Fig. 6A,B; supplemental Fig. S5A, available at www.jneurosci.org as supplemental material). Additionally, numbers of Meln-immunoreactive ipRGCs appeared similar in whole-mount preparations of P12 mutant and control retinas [controls contained 42.4 ± 7.9 (SD) ipRGCs per field, N = 11 retinas, n = 86 fields; mutants contained 41.3 ± 9.6 (SD) ipRGCs per field, N = 12 retinas, n = 101 fields]. Importantly, the observed defects in retinogeniculate targeting were also not attributable to an inability of mutant ipRGCs to endocytose or anterogradely transport CTB (Fig. 6A′,B′,C–F). Thus, defects in RGC innervation of vLGN and IGL did not appear to arise from intraretinal defects in ipRGCs.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

Reelin is not required for the generation or development of non-image-forming RGCs, nor their projections to non-LGN retino-recipient nuclei. A, B, Non-image-forming ipRGCs were labeled by Meln IHC in P12 control and reln^rl/rl mutant retinal whole mounts. The distribution and morphology of ipRGCs appeared similar in controls and mutants. A′, B′, Dendritic stratification of melanopsin-immunolabeled ipRGCs (green) in OFF- and ON-sublaminae of the IPL (arrows) appeared similar in cross-sections of P12 reln^rl/rl mutant and littermate control retinas. Moreover, mutant and control ipRGCs (green) demonstrated equal capacities to endocytose CTB (red). C, D, Retinal projection to other retino-recipient nuclei innervated by ipRGC appear normal in the absence of functional reelin. RGC projections to the OPN (C, D) and SCN (E, F) were assessed by injections of different fluorescently conjugated CTB into each eye of P12 controls and mutants. The patterns of innervation of these nuclei were indistinguishable in mutants or controls. G, H, PLRs remained present in reln^rl/rl mutants. P14 mutant and control mice (n = 3 each) were dark adapted for >1 h and then exposed to a 30 s 1.7 mW/cm² light. Pupil constriction was captured on video, and images were used to measure pupil size (G, white circles) before the onset of light and after 30 s of light. In H, the percentage pupil constriction after light exposure was calculated by comparing to pupil size during dark adaptation. Scale bars: (in A) A, B, 100 μm; (in A′) A′, B′, 25 μm; (in C) C–F, 200 μm.

We also assessed whether ipRGC axons correctly targeted other appropriate retino-recipient nuclei in reln^rl/rl mutants. Classes of ipRGCs innervate the OPN and SCN (Muscat et al., 2003; Hattar et al., 2006; Ecker et al., 2010). CTB labeling in reln^rl/rl mutants revealed normal patterns of retinal innervation to these nuclei (Fig. 6C–F). Furthermore, reln^rl/rl mutants retained rudimentary PLRs, behaviors that depend on ipRGC innervation of the OPN (Fig. 6G,H) (Güler et al., 2008). Therefore, in the absence of reelin ipRGCs were not only morphologically normal but generated axons capable of targeting some appropriate retino-recipient nuclei and retained rudimentary function.

To directly assess whether ipRGC axons were misrouted in the reln^rl/rl LGN, mutants were crossed with opn4-tau-LacZ mice, since M1 ipRGCs are labeled in this line and M1 ipRGCs densely innervate vLGN and IGL (Fig. 7A) (Hattar et al., 2006; Ecker et al., 2010). In P11 reln^rl/rl;opn4-tau-LacZ mutants, several defects were observed in the pattern of ipRGC projections to vLGN and IGL. First, M1 ipRGC projections avoided central regions of mutant IGL and instead formed a dense, narrow network of fibers adjacent to the ventral aspect of dLGN (Fig. 7B). Second, a large cohort of ipRGC axons invaded ventromedial regions of dLGN (Fig. 7B′,D′, arrows). The area of dLGN occupied by misrouted M1 ipRGC axons was measured in P21 mutants and controls. Although few M1 ipRGC arbors were observed in control dLGN, nearly 20% of the area of the reln^rl/rl mutant dLGN contained LacZ-expressing ipRGC arbors [2.9 ± 1.1% (SEM) of the area of control dLGN contained M1 ipRGC arbors versus 19.2 ± 2.9% (SEM) in reln^rl/rl mutant dLGN; N = 6; p < 0.05 by Student's t test]. Last, the majority of retinal axons misrouted out of the IGL or vLGN contained LacZ, revealing they originated from M1 ipRGCs (Fig. 7B,C).

• Download figure
• Open in new tab
• Download powerpoint

Figure 7.

Non-image-forming ipRGCs require reelin and Dab1 for retinogeniculate targeting. A–E, IpRGC projections in P11 controls (A), reln^rl/rl;opn4-tau-LacZ^+/+ mutants (B), and dab1^scm/scm;opn4-tau-LacZ^+/− mutants (D). M1 ipRGCs were labeled by LacZ IHC (green). All RGC axons were labeled by binocular injection of CTB (CTB[b]; red). The arrowheads in B–E indicate mistargeted RGC axons containing LacZ. The arrow in B′ and D′ highlight misrouted LacZ-containing M1 ipRGC axons invading dLGN in mutants (compare with arrow in A′). C–C″ and E–E″ are high-magnification images of mistargeted axons depicted by asterisks in B and D. Note that LacZ immunoreactivity is reduced in dab1^scm/scm;opn4-tau-LacZ^+/− mutants (D) compared with images in A and B, since opn4-tau-LacZ heterozygotes were analyzed. F–H, By P21 mistargeted ipRGC axons have extensive arbors within dLGN in reln^rl/rl;opn4-tau-LacZ and dab1^scm/scm;opn4-tau-LacZ mutants, but not controls (compare arrows in F–H). Scale bars: (in A) A, B, D, F–H, 300 μm; (in C) C, E, 25 μm.

We next examined the role of Dab1 in the correct targeting of M1 ipRGC axons. Scrambler mutants (dab1^scm/scm) were crossed to the opn4-tau-LacZ mice described above. In the absence of Dab1, M1 ipRGCs invaded non-retino-recipient thalamic regions and occupied a statistically significant area of mutant dLGN [13.0 ± 1.5% (SEM) of dab1^scm/scm mutant dLGN contained M1 ipRGC arbors; N = 6; p < 0.05 by Student's t test]. Analyses in both reln^rl/rl;opn4-tau-LacZ and in dab1^scm/scm;opn4-tau-LacZ mice also demonstrated that M1 ipRGC projections to SCN and OPN were unaffected by the absence of functional reelin (supplemental Fig. S6, available at www.jneurosci.org as supplemental material) (data not shown) (Hattar et al., 2006; Ecker et al., 2010). These findings identify, for the first time, that at least one class of non-image-forming retinal axons require reelin and Dab1 for targeting appropriate LGN subnuclei.

Reelin is not required for subnuclei-specific targeting of dLGN-projecting retinal axons

We next tested whether image-forming classes of RGCs require reelin for appropriately targeting dLGN. For this, we used three approaches—labeling retinal terminals, labeling dLGN-projecting axons, and colabeling misrouted retinal axons with CTB and cell-specific markers of dLGN-projecting axons. To label axon terminals from image-forming RGCs, we examined the distribution of VGLUT2, a synaptic component enriched in dLGN-projecting retinal axons (Fig. 8A) (Land et al., 2004). In reln^rl/rl mutant LGN, the density and distribution of VGLUT2 immunoreactivity appeared remarkably similar to controls, suggesting that image-forming, dLGN-projecting RGC axons were not misrouted into vLGN or IGL in the absence of functional reelin (Fig. 8B). Second, calretinin-expressing dLGN-projecting retinal axons (Lüth et al., 1993; Huberman et al., 2008b) were labeled in mutants and controls by IHC. As in control LGN, calretinin-containing retinal axons were excluded from vLGN and IGL in reln^rl/rl mutant LGN (Fig. 8C,D). VGLUT2 or calretinin immunoreactivity alone could not reveal whether axons misrouted into non-retino-recipient thalamic nuclei in reln^rl/rl mutants originated from image-forming classes of RGC. To test whether any mistargeted RGC axons belonged to image-forming classes of RGCs, axons in reln^rl/rl mutant thalamus were colabeled by calretinin IHC and CTB. No misrouted CTB-labeled RGC axons were observed to contain calretinin in mutant thalamus (Fig. 8E,F). Together, these experiments demonstrate that classes of image-forming RGC axons correctly target dLGN in the absence of functional reelin.

• Download figure
• Open in new tab
• Download powerpoint

Figure 8.

Reelin is dispensable for dLGN targeting by axons derived from image-forming RGCs. A, B, VGLUT2-immunolabeled dLGN-projecting retinal terminals appear similar in P12 controls (A) and reln^rl/rl mutants (B). C, D, Calretinin (Calr)-immunolabeled RGC axons appeared similarly confined to dLGN in P12 control (C) and reln^rl/rl mutant (D) LGN. The arrows indicate abrupt border of Calr-containing arbors at the dLGN–IGL border. E, F, Misrouted CTB-labeled RGC axons in adjacent non-retino-recipient thalamic nuclei do not contain Calr in P14 reln^rl/rl mutants (see arrowheads). Scale bars: (in A) A–D, 300 μm; (in E) E, F, 15 μm.

Reelin is not required for subnuclei-specific targeting of layer VI corticothalamic axons

In addition to retinal axons, cortical axons differentially target LGN subnuclei (Jacobs et al., 2007). Cortical neurons in layer VI project axons that course through the internal capsule and around vLGN and IGL to selectively innervate dLGN (Fig. 9A). Despite coursing around vLGN and IGL, these corticothalamic axons do not enter or arborize in ventral LGN subnuclei. To address whether cortical axons required reelin for normal LGN targeting, reln^rl/rl mutants were crossed to golli-tau-GFP transgenic mice, in which layer VI neurons were selectively labeled with GFP (Jacobs et al., 2007). Corticothalamic axons from layer VI neurons avoided vLGN and IGL and selectively innervate dLGN in controls (Fig. 9B,D). As expected, disruptions in cortical layering were observed in reln^rl/rl;golli-tau-GFP mutant mice (Fig. 9C). As described above, axons from layer VI cortical neurons also displayed fasciculation defects and took altered courses through the internal division of the vLGN, a region that does not receive retinal input (Harrington, 1997). Yet despite defects in neuronal position and axonal trajectory, layer VI corticothalamic axons correctly avoided the IGL and retino-recipient portion of the vLGN and arborized only in the dLGN of reln^rl/rl;golli-tau-GFP mutants (Fig. 9E). We draw two important conclusions from these findings: neither reelin nor neuronal position are critical determinants for layer VI corticothalamic axons to correctly avoid vLGN and IGL and specifically target dLGN.

• Download figure
• Open in new tab
• Download powerpoint

Figure 9.

Reelin is dispensable for dLGN targeting by corticothalamic axons. A, Schematic demonstrating layer VI cortical axons pass through the internal capsule (IC), bypass vLGN/IGL, and selectively innervate dLGN. B, C, Layer VI cortical neurons were labeled by GFP IHC in P12 reln^+/+;golli-tau-GFP controls (B) or reln^rl/rl;golli-tau-GFP mutants (C). Normal cortical layering and the location of white matter tracts (WM) are depicted in B. Layer VI cortical neurons and their projections in transgenic mice were labeled by GFP IHC. D, E, Although corticothalamic axons have an altered course in reln^rl/rl;golli-tau-GFP mutants (see asterisks in E), they selectively target dLGN as in reln^+/+;golli-tau-GFP controls. d, dLGN; v, vLGN; i, IGL. Scale bar, 300 μm.