Volume 97, Issue 5, Pages 1049-1062.e6 (March 2018) An Attractive Reelin Gradient Establishes Synaptic Lamination in the Vertebrate Visual System  Vincenzo Di Donato, Flavia De Santis, Shahad Albadri, Thomas Oliver Auer, Karine Duroure, Marine Charpentier, Jean-Paul Concordet, Christoph Gebhardt, Filippo Del Bene  Neuron  Volume 97, Issue 5, Pages 1049-1062.e6 (March 2018) DOI: 10.1016/j.neuron.2018.01.030 Copyright © 2018 Elsevier Inc. Terms and Conditions

Figure 1 Genetic Loss of Function of Reelin Pathway Components Disrupts Synaptic Lamina Targeting of Retinal Ganglion Cell Axons in the Tectal Neuropil (A) Schematic lateral view of a 5 days post-fertilization (dpf) larva showing the anatomical localization of the optic tectum (inset). (B) Schematic lateral view of the optic tectum showing the relative positions of the major tectal cell populations. SINs, superficial inhibitory neurons (in purple); RGC axons, retinal ganglion cell axons (in green); PVNs, periventricular neurons (white ellipses, example monostratified PVN shown as black cell). The tectal neuropil, consisting predominantly of axons and dendrites, is shown in orange and can be further subdivided into four main layers: stratum opticum (SO), stratum fibrosum et griseum superficiale (SFGS), stratum griseum centrale (SGC), and stratum album centrale (SAC). (C) Lateral view of a confocal z-projection showing a 5 dpf Tg(s1156tEt; UAS:RFP) × Tg(Brn3C:mGFP) embryo displaying SINs in magenta (labeled by RFP) and RGC axons in green (labeled by mGFP). Scale bar, 20 μm. (D and E) Confocal reconstruction (D) and corresponding schematics (E) of one transiently isl2b:GFPCaax-labeled RGC axonal arbor in the tectum of a representative wild-type larva at 5 dpf shown from the side, parallel to the skin (dotted line), to highlight the RGCs’ laminar morphologies. The labeled RGC axon targets only a single neuropil lamina. Arbor thickness was measured using the fluorescence profile along a line perpendicular to the skin across the RGC arbor (solid white line, see STAR Methods). (F and G) Confocal reconstruction (F) and corresponding schematics (G) of a transiently GFP-labeled RGC axonal arbor in the tectum of a representative 5 dpf reln−/− larva highlighting the disruption of single lamina targeting. Arbor thickness was measured using the fluorescence profile along a line perpendicular to the skin across the RGC arbor (solid white line, see STAR Methods). (H and I) Confocal reconstruction (H) and corresponding schematics (I) of a transiently GFP-labeled RGC axonal arbor in the tectum of a representative 5 dpf vldlr−/− larva showing aberrant laminar targeting. (J and K) Confocal reconstruction (J) and corresponding schematics (K) of a transiently GFP-labeled RGC axonal arbor in the tectum of a representative 5 dpf dab1a−/− mutant larva spanning multiple laminae in the neuropil. (A–K) SO, stratum opticum; SFGS, stratum fibrosum et griseum superficiale; SGC, stratum griseum centrale; SAC, stratum album centrale; D, dorsal; A, anterior. Scale bars, 20 μm. (L) Average RGC arbor thickness in wild-type (n = 18 larvae, 37 RGCs total, mean arbor thickness per larva = 7.3 μm ± 0.3 μm SEM), reln−/− (n = 10 larvae, 22 RGCs in total, mean arbor thickness per larva = 20.7 μm ± 3.7 μm SEM), vldlr−/− (n = 21 larvae, 48 RGCs in total, average arbor thickness per larva = 16.0 μm ± 1.5 μm SEM), and dab1a−/− larva (n = 10 larvae, 28 RGCs in total, average RGC arbor thickness per larva = 10.2 μm ± 1.3 μm SEM). The average arbor thickness per larva was significantly larger in reln−/−, vldlr−/−, and dab1a−/− larvae compared to wild-type larvae, respectively (one-tailed two-sample t tests, wild-type < reln−/− p = 2.5e−05, wild-type < vldlr−/− p = 1.9e−06, wild-type < dab1a−/− p = 0.006). This suggests that genetic loss of function of components of the Reelin pathway disrupts the ability of retinal ganglion cell arbors to find their corresponding synaptic lamina in the tectum. Error bars represent mean ± SEM across larvae. ∗∗p < 0.01, ∗∗∗p < 0.001, one-tailed two-sample t test. See also Figures S1 and S2. Neuron 2018 97, 1049-1062.e6DOI: (10.1016/j.neuron.2018.01.030) Copyright © 2018 Elsevier Inc. Terms and Conditions

Figure 2 Genetic Loss of Function of Components of the Reelin Pathway Disrupts Lamination of Periventricular Neuron Arbors in the Tectal Neuropil (A and B) Confocal reconstruction (A) and corresponding schematics (B) of a transiently GFP-labeled bistratified PVN in the tectum of a 5 dpf wild-type larva shown from the side parallel to the skin (dotted line) to highlight the PVNs’ laminar morphology. Stratified PVNs did not show any lamination defects. A PVN was classified as displaying lamination defects if we could detect several arbor branches crossing multiple laminae per PVN. (C and D) Confocal reconstruction (C) and corresponding schematics (D) of a transiently GFP-labeled bistratified PVN in the tectum of a 5 dpf reln−/− larva highlighting aberrant PVN morphology. The white arrowheads indicate multiple branches atypically crossing distinct neuropil laminae. (E and F) Confocal reconstruction (E) and corresponding schematics (F) of a transiently GFP-labeled bistratified PVN in the tectum of a 5 dpf vldlr−/− larva showing aberrant PVN morphology. The white arrowheads indicate branches crossing different neuropil laminae. (G and H) Confocal reconstruction (G) and corresponding schematics (H) of a transiently GFP-labeled monostratified PVN in the tectum of a 5 dpf dab1a−/− larva showing aberrant PVN morphology. The white arrowheads indicate branches meandering across different neuropil laminae. (A–H) SO, stratum opticum; SFGS, stratum fibrosum et griseum superficiale; SGC, stratum griseum centrale; SAC, stratum album centrale; D, dorsal; A, anterior. Scale bars, 20 μm. (I) Frequency distribution of non-stratified PVNs, stratified PVNs without, or stratified PVNs with obvious lamination defects in wild-type (n = 14 PVNs from 13 larvae), reln−/− (n = 10 PVNs from 9 larvae), vldlr−/− (n = 9 PVNs from 8 larvae), and dab1a−/− (n = 13 PVNs from 13 larvae) at 5 dpf. Lamination defects of stratified PVNs were defined as the occurrence of several arbor branches originating from the cell body crossing multiple neuropil laminae. PVN frequency distributions were significantly different in reln−/− and dab1a−/− larvae compared to wild-type larvae, respectively (k × 2 chi-square test after Brandt-Snedecor, wild-type versus reln−/− p = 0.0082, wild-type versus dab1a−/− p = 0.0036), whereas there was no significant difference observed between wild-type and vldlr−/− (p = 0.12). These data suggest that genetic loss of function of components of the Reelin pathway also disrupt the ability of PVNs to target single synaptic lamina in the tectum. See also Figure S3. Neuron 2018 97, 1049-1062.e6DOI: (10.1016/j.neuron.2018.01.030) Copyright © 2018 Elsevier Inc. Terms and Conditions

Figure 3 Tectum-Derived Reelin and RGC-Derived VLDLR/Dab1a Regulate the Establishment of RGC Synaptic Laminae in the Tectal Neuropil (A) Schematics of blastula transplants from RGC-specific Tg(Brn3c:mGFP) donors supplying acceptor embryos of varying genotypes with green-labeled RGCs of donor genotype. (B) Side view of RGC axons at 5 dpf after blastula stage transplantations from a wild-type donor into a reelin mutant acceptor. Aberrant laminations (white arrowhead) were detected in all larvae analyzed (n = 8/8). (C) Side view of a RGC axon in the tectal neuropil at 5 dpf after blastula stage cell transplantation from a reelin mutant donor into a wild-type acceptor. No synaptic lamination defects were detected in all larvae analyzed (n = 0/12), suggesting that tectum-derived Reelin and not Reelin expressed by RGCs controls single synaptic lamina targeting. (D) Side view of a RGC axon in the tectal neuropil at 5 dpf after blastula stage cell transplantation from a wild-type donor into a vldlr−/− acceptor. No synaptic lamination defects were observed in analyzed larvae (n = 0/6). (E) Side view of RGC axons in the tectal neuropil at 5 dpf after blastula stage cell transplantation from a vldlr−/− mutant donor into a wild-type acceptor. Aberrant laminations (white arrowhead) were detected in most larvae (n = 5/10), suggesting that vldlr expressed by RGCs influences single lamina targeting by RGCs. In (A)–(E), scale bars, 30 μm. D, dorsal; A, anterior. (F) Side view of a RGC axon in the tectal neuropil at 5 dpf after blastula stage cell transplantation from a wild-type donor into a dab1a−/− acceptor. No synaptic lamination defects were observed in analyzed larvae (n = 0/7). (G) Side view of RGC axons in the tectal neuropil at 5 dpf after blastula stage cell transplantation from a dab1a−/− mutant donor into a wild-type acceptor. Aberrant laminations (white arrowheads) were detected in most larvae (n = 5/9), suggesting that dab1a expressed by RGCs is involved in single lamina targeting by RGCs. Neuron 2018 97, 1049-1062.e6DOI: (10.1016/j.neuron.2018.01.030) Copyright © 2018 Elsevier Inc. Terms and Conditions

Figure 4 A Superficial-to-Deep Reelin Gradient Is Present in the Tectal Neuropil and Requires Heparan Sulfate Proteoglycans for Its Stabilization in the ECM (A) Immunostaining of anti-reelin (green) and DAPI (blue) on a horizontal cryo-section through a 5 dpf larval tectum. Reelin protein accumulates at the basement membrane (i.e., the tectal surface), and its concentration decreases perpendicular to the laminae in the neuropil toward the periventricular zone, where the cell bodies of PVNs reside. Scale bar, 30 μm. (B) Grayscale image of anti-reelin staining shown in (A) and the rectangle in yellow along which the Reelin gradient was measured. Scale bar, 30 μm. (C) Densitometric plot of normalized anti-reelin fluorescence in 5 dpf wild-type tecta showing a superficial-to-deep Reelin gradient. Data are represented as mean ± SEM, calculated from four samples out of four independent experiments. Reelin was also expressed in a few PVNs (see A and B), but this did not seem to cause any detectable disturbance of the Reelin gradient in the deep neuropil laminae. (D–F) Spatial distribution of Reelin protein in cryo-sectioned tecta of 5 dpf wild-type (D), drgs510 (type IV collagen LOF) (E), and dakt0273b (exostosin-2 LOF, glycosyltransferase involved in the heparan sulfate proteoglycans synthesis) (F) larvae. Anti-reelin staining is shown in green and DAPI counterstaining in blue. The yellow rectangles indicate along which direction the Reelin gradients for (G) and (H) were measured. Scale bars, 30 μm. (G) Densitometric plots of normalized anti-reelin fluorescence intensity taken from wild-type and drgs510 tecta. Data are represented as mean ± SEM, calculated from five samples per genotype. No difference in Reelin gradient distribution was seen between wild-type and drgs510 larvae. (H) Densitometric plots of normalized anti-reelin fluorescence intensity taken from wild-type and dakt0273b tecta (see D and F). Data are represented as mean ± SEM, calculated from five samples per genotype. A difference in Reelin gradient distribution was seen between wild-type and dakt0273b larvae, suggesting that heparan sulfate proteoglycans are required for Reelin gradient stabilization in the tectum. See also Figure S4. Neuron 2018 97, 1049-1062.e6DOI: (10.1016/j.neuron.2018.01.030) Copyright © 2018 Elsevier Inc. Terms and Conditions

Figure 5 Ectopic Reelin Expression in Periventricular Neurons and Radial Glia Causes Perturbations of Single Laminae Targeting of RGCs in the Tectal Neuropil (A) Lateral projection of a representative confocal stack showing a single RGC axon genetically labeled with isl2b:RFP (magenta) in the tectum of a larva expressing the rpl5b:Gal4;UAS:GFP transgene (control), which is active in a subset of PVNs and radial glia cells (RGs) (green). All analyzed RGCs targeted a single lamina (n = 17/17 fish). Scale bar, 35 μm. (B) Schematics of a single RGC axon targeting a distinct lamina in the tectal neuropil of a 5 dpf larva expressing GFP (control) in a subset of PVNs and RGs (green), thus spatially opposite of the superficial Reelin-enriched neuropil zone (as in A). PVNs, periventricular neurons; RG, radial glia cells; D, dorsal; A, anterior. (C) Lateral projection of a representative confocal stack showing multiple isl2b:RFP-labeled RGC axons (magenta) in the tectum of a Tg(rlp5b:Gal4;UAS:relnT2AlynGFP) larva allowing the co-expression of mouse reelin and membrane-GFP in a subset of PVNs and RGs (green). RGC axons meander between multiple laminae (white arrowheads, altered lamination pattern observed in n = 11/19 larvae), thus suggesting that ectopic reeln expression is interfering with the endogenous graded Reelin distribution. Scale bar, 35 μm. (D) Schematics of multiple RGC axons in the tectal neuropil of a 5 dpf larva expressing relnT2AlynGFP in a subset of PVNs and RGs (green), thus spatially opposite of the superficial Reelin-enriched neuropil zone (as in C). PVNs, periventricular neurons; RGs, radial glia cells; D, dorsal; A, anterior. See also Figure S5. Neuron 2018 97, 1049-1062.e6DOI: (10.1016/j.neuron.2018.01.030) Copyright © 2018 Elsevier Inc. Terms and Conditions

Figure 6 The Reelin Gradient Is a Chemoattractant for Retinal Afferents during Laminar Targeting (A) Side view of the retinotectal projection of a 5 dpf wild-type larva expressing the Tg(Brn3c:mGFP) transgene that labels exclusively RGC axonal arbors targeting the two most superficial layers (SO and SFGS), but not the deep layers (SAC and SGC), of the tectal neuropil. As the most superficial SO layer is usually hard to distinguish from the more prominent SFGS, the thickness of the projection was measured by taking the distance between the most superficial and the deepest visible axonal branch (white dotted line). The average thickness of the RGC projection in wild-type larvae was 25.8 μm ± 0.6 μm SEM (n = 17 larvae). Scale bar, 30 μm. (B) Side view of the retinotectal projection of a 5 dpf reln−/− larva labeled by the Tg(Brn3c:mGFP) transgene. RGC branches were shifted to deep layers in the tectal neuropil, which is reflected by an increase in overall RGC projection thickness in reln−/− (32.3 μm ± 1.5 μm SEM, n = 8 larvae) compared to wild-type larvae (one-tailed two-sample t tests, p = 2.9e−05). Scale bar, 30 μm. (C) Side view of the retinotectal projection of a 5 dpf vldlr−/− larva labeled by the Tg(Brn3c:mGFP) transgene. RGC branches unusually projected to deep layers in the tectal neuropil as indicated by an increase in RGC projection thickness in vldlr−/− (32.3 μm ± 1.5 μm SEM, n = 7 larvae) compared to wild-type larvae (one-tailed two-sample t tests, p = 2.2e−05). Scale bar, 30 μm. (D) Side view of the retinotectal projection of a 5 dpf dab1a−/− larva labeled by the Tg(Brn3c:mGFP) transgene. RGC branches extended to deep layers in the tectal neuropil as indicated by an increase in RGC projection thickness in dab1a−/− (32.8 μm ± 1.4 μm SEM, n = 7 larvae) compared to wild-type larvae (one-tailed two-sample t tests, p = 4.5e−06). Scale bar, 30 μm. (E) Side view of the retinotectal projection of a 5 dpf larva expressing the Tg(isl2b:Gal4;UAS:RFP) transgene that labels the entire RGC population targeting all four main layers (SO, SFGS, SAC, and SGC) of the tectal neuropil (magenta). Membrane-targeted lyn-GFP is transiently expressed as a control in a mosaic subset of these RGCs. The inset highlights the neuropil area, where the deep neuropil layers, SGC and SAC, can be easily identified. Higher magnification of this inset shows axonal targeting within the deep neuropil layers by GFP-positive control RGCs. D, dorsal; A, anterior. Scale bar, 35 μm. (F) Side view of the retinotectal projection showing the entire RGC population (magenta) in 5 dpf larva expressing Tg(isl2b:Gal4;UAS:RFP). In addition vldlr-GFP is transiently expressed in a mosaic subset of RGCs. We expected these GFP-positive RGCs to be more sensitive to endogenous Reelin. Higher magnification of the inset shows the absence of vldlr-overexpressing RGCs in the deep layers of the tectal neuropil. D, dorsal; A, anterior. Scale bar, 35 μm. (G) Quantification (see E and F) showing the percentage of larvae displaying one or more GFP-labeled RGC axons in the main layers of the tectal neuropil (n = 20 UAS:lyn-GFP-expressing larvae; n = 20 UAS:vldlr-GFP-expressing larvae). The superficial layers SO and SFGS were merged and scored together, as their close proximity makes their discrimination in isl2b:Gal4 larvae difficult. RGCs overexpressing vldlr-GFP target the deep neuropil layers at a much lower frequency compared to RGCs expressing lyn-GFP (ctrl), suggesting that the former preferentially targets Reelin-enriched superficial neuropil layers. (A–G) SO, stratum opticum; SFGS, stratum fibrosum et griseum superficiale; SGC, stratum griseum centrale; SAC, stratum album centrale; SPV, stratum periventriculare. Neuron 2018 97, 1049-1062.e6DOI: (10.1016/j.neuron.2018.01.030) Copyright © 2018 Elsevier Inc. Terms and Conditions

Figure 7 Model of Axonal Lamination Regulated by Counteracting Target-Derived Gradients of Reelin and Slit1 (A) Reelin is likely secreted by SINs from the most superficial layer of the tectal neuropil. The enrichment of Reelin protein in the basement membrane (BM), depending on HSPGs, leads to the stable formation of a superficial-to-deep concentration gradient. The distribution in a gradient allows Reelin to act across multiple synaptic laminae as a short-range attractive cue for ingrowing vldlr-expressing RGC axons. (B) Slit1 accumulates in the BM through binding to collagen IV. Slit1 is presumably distributed in a superficial-to-deep concentration gradient in the tectal neuropil and acts as a short-range repulsive cue for robo2-expressing RGC axons. A RGC axon endowed with specific VLDLR and Robo2 concentrations would be able to identify its tectal target lamina, where attractive Reelin and repulsive Slit1 signaling are in balance. See also Figure S6. Neuron 2018 97, 1049-1062.e6DOI: (10.1016/j.neuron.2018.01.030) Copyright © 2018 Elsevier Inc. Terms and Conditions