Figure 8. Hevin is critical for the resolution of VGlut1/VGlut2-innervated SMECs.

(A) Representative Imaris 3D reconstructions of GFP-labeled dendrites (blue) in the S/Z of P21 WT and hevin KO V1. Presynaptic puncta are rendered as elliptical ‘spots’ (green = VGlut1; red = VGlut2). Numerous unisynaptic spines can be seen in the WT (arrow = VGlut1 spine; arrowheads = VGlut2 spines). In hevin KO, a SMEC can be seen contacting both a VGlut1 and VGlut2 ‘spot’ (asterisk). A unisynaptic VGlut1 spine (arrow) is also present. Scale bar, 0.5 µm. (B) Quantification of SMEC density at P21 shows that the increase in total SMECs in the hevin KO is driven by the VGlut1/VGlut2 SMEC subtype (3 animals/genotype, n = 15 dendrites per condition; p < 0.01, Student's t test). (C) Model for astrocytic control of thalamocortical connectivity by hevin. Left: In early V1 synaptic development, intracortical (primarily VGlut1, or VG1) axons compete with thalamocortical (primarily VGlut2, or VG2) axons for synapses, occasionally forming synapses on the same dendritic spine (resulting in a SMEC). In the WT, astrocytes secrete hevin which stabilizes VGlut2+ synapses, resulting in discrete populations of VGlut1+ and VGlut2+ unisynaptic spines. In the hevin KO, VGlut2+ synapses cannot be properly stabilized. These sites either remain in competition with VGlut1, explaining the persistence of SMECs in hevin KO, or become lost, resulting in more VGlut1+ synapses overall.

DOI: http://dx.doi.org/10.7554/eLife.04047.017

Figure 8—figure supplement 1. Overlap of VGlut1 and VGlut2 in light microscopy is due to close proximity of different presynaptic terminals.

(A) Apparent overlap between VGlut1 (green) and VGlut2 (red) by light microscopy appears as yellow puncta (arrows). Representative images show VGlut puncta at P15 and P25 in both WT and hevin KO. Scale bar, 5 µm. (B) Quantification of the apparent overlap in VGlut1/VGlut2 puncta, which is decreased between P15 and P25 in the WT. In the hevin KO, the VGlut1/VGlut2 overlap is slightly decreased at P15 but is unchanged as development continues to P25, at which point it is significantly higher than WT (n = 3 z-stacks per animal, 3–4 animals per genotype; *p < 0.01, **p < 0.05, nested ANOVA). (C) To show that the apparent overlap in VGlut1 and VGlut2 puncta was not due to random chance in densely stained tissue, confocal Z-stacks were split into two channels (VGlut1 and VGlut2), the VGlut2 channel was rotated 90° out of alignment, then the two channels were re-merged and analyzed for apparent co-localization of presynaptic puncta. For all conditions analyzed, overlap frequency was significantly decreased in the rotated image (red) compared to the original unrotated image (black) (n = 3 z-stacks per animal, 3–4 animals per genotype; p < 0.01, Student's t test). (D) Representative SIM image of VGlut1 (green)/VGlut2 (red)-stained S/Z in P15 WT cortex. At the high resolution afforded by SIM (∼100 nm), VGlut1 and VGlut2 presynaptic puncta do not appear co-localized. Scale bar, 5 µm. (E) Scatterplot showing the apparent overlap in VGlut1/VGlut2 puncta at differing co-localization distance thresholds. When the limit of co-localization is 100 nm between puncta, approaching the resolution of SIM, virtually no co-localization is detected. By increasing the allowable distance between co-localized puncta (closer to the resolution of confocal microscopy), VGlut1/VGlut2 overlap frequency eventually approaches the levels previously detected by confocal imaging (n = 3–4 Z-stacks each from 3 animals).

DOI: http://dx.doi.org/10.7554/eLife.04047.018

Figure 8—figure supplement 2. Imaging presynaptic terminals in proximity to dendritic spines.

(A) Schematic for the IUE technique. Anesthetized dams have their uterine horns exposed at E15.5. DNA plasmid containing GFP with loading dye is injected into the lateral ventricle of each pup, followed by electric pulses to facilitate uptake of the plasmid. The horns are then placed back into the dam, the incision sutured, and the dam is allowed to recover and give birth to the electroporated litter. (B) 10× magnification image taken in V1 at P15 showing that IUE resulted in specific GFP labeling of neurons in cortical Layer II/III. Secondary and tertiary dendrites from these neurons, which project to the S/Z, are then imaged by confocal microscopy. (C) Left: A single optical section taken from a confocal Z-stack shows a GFP-labeled dendrite (blue) in the S/Z along with surrounding presynaptic puncta (VGlut1-green; VGlut2-red). Middle: Imaris surface rendering was used to image the dendrite's structure, including spines, in 3D. Right: The ‘Spots’ function of Imaris allowed for the resolution of individual presynaptic puncta in three-dimensional space. Scale bar, 1 µm. (D) After applying the Matlab ‘Spots close to surface’ algorithm in Imaris, in order to isolate spots within 0.2 µm of the dendrite, spines with closely-associated presynaptic puncta can be quantified. An asterisk indicates the location of a SMEC which is contacted by both a VGlut1 and VGlut2 terminal. (E) Quantification in P15 WT V1 showing the percentage of spines that contact various presynaptic puncta, including the various subtypes of SMECs (3 animals, n = 9 dendrites).

DOI: http://dx.doi.org/10.7554/eLife.04047.019