Generation of Thy1-YFP-NL2 transgenic mice
YFP-NL2 expressed here was modified from Graf et al.  and consists of the signal sequence of mouse Neuroligin-1, hexahistidine (HHHHHH), Flag (GGDYKDDDDK), and EYFP tags followed by the mature coding sequence of mouse NL2. The Thy1 promoter  was used to drive expression of YFP-NL2. The transgene was generated by cloning the Thy1 promoter fragment into shuttle vector LNL, which contained the needed restriction sites for transgene release. YFP-NL2 was digested with HindIII and Afl II and blunted into XhoI cut and blunted thy1-LNL. The Thy1-YFP-NL2 transgene was released from the vector backbone sequence by restriction digestion with AscI and PmeI and injected into B6/CBA F1 hybrid pronuclei to generate founder mice.
C57BL/6 (P21) and Thy1-YFP-NL2 (P7, P12, P15, P21) mice were deeply anesthetized with 5% isofluorane and decapitated. Eyes were removed and placed in ice cold mouse artificial cerebrospinal fluid (mACSF; 119 mM NaCl, 2.5 mM KCl, 2.5 mM CaCl2, 1.3 mM MgCl2, 1 mM NaH2PO4, 11 mM glucose (20), 20 mM HEPES, pH = 7.4). After removing the lens and vitreous, the eye cup was fixed in 4% paraformaldehyde for 15 to 30 minutes. After fixation, the eye cups were rinsed in 0.1 M PBS. The retina was removed from the eye cup, embedded in 4% low-melting point agarose and cut into 60 μm thick sections using a vibratome. Sections were mounted and used for imaging or processed for immunostaining as follows: blocked in 10% NGS in PBS for 1 hour followed by overnight incubation in 5% NGS, 0.5% Triton-X100, with the corresponding primary antibodies. The primary antibodies used were: rabbit anti-NL2 antibody (1:8,000; generous gift of F Varoqueaux and N Brose) [35, 40], guinea-pig anti-γ2 antibody (1:1,000; generous gift of JM Fritschy), a monoclonal mouse antibody (mAb4a (P21) against all glycine receptor subunits,1:400; or mAb2b (P10), against the α1 subunit of the glycine receptor, 1:400; Synaptic Systems, Goettingen, Germany), anti-gephyrin mAb7a (1:500; Synaptic Systems), anti-VGAT antibody (1:1,000; Millipore, Temecula, CA, USA), anti-CtBP2 antibody (1:1000; BD Transduction, Franklin Lakes, NJ, USA). Sections were then washed and incubated for 1 hour with the corresponding secondary antibody conjugated to either Alexa-488 or Alexa-568 (1:1,000; Invitrogen, Carlsbad, CA, USA).
Thy1-YFP-NL2 mice were deeply anesthetized with 5% isofluorane and decapitated. Eyes were removed and placed in ice cold mouse mACSF. Retinas were removed from the eye cup and mounted RGC side up on black nitrocellulose filter paper (HABP013, Millipore, Bedford, MA, USA). Gold particles were coated with 20 μg CMV:tdtomato DNA (gift of R Tsien) and delivered using a Bio-Rad Helios gene-gun as previously described . Retinas were incubated for 18 to 24 hours at 33°C, fixed for 30 minutes in 4% parafolmaldehyde in mACSF, washed in PBS and mounted in Vectashield (Vector Labs, Burlingame, CA, USA). The data presented in this study were obtained from six P8, five P12, five P15, ten P21 and five P30 RGCs.
Imaging and image analysis
Images were obtained using a 1.35 NA 60× oil objective (Olympus). Images were acquired at 0.069 × 0.069 × 0.3 μm for double labeling immunohistochemistry (Figures 1, 3 and 4), and 0.103 × 0.103 × 0.3 μm voxel sizes for vertical slices of YFP-NL2 retinas (Figure 2) and retinal whole-mounts (Figures 2 and 6). Images were processed using Metamorph (Molecular Devices, Sunnyvale, CA, USA), Image J (NCBI), Amira (Mercury Computer Systems Inc., Chelmsford, MA, USA) and Matlab (Math Works, Natick, MA, USA). Images were median-filtered to reduce noise. The contrast and gamma of the images were adjusted to increase visualization of dim objects. Using the 'label-field' function of the AMIRA software program, a threshold was applied, plane by plane, to capture pixels representing the dendrites of the RGCs . This procedure generated a binary mask of the dendrites that was then used to isolate puncta from the YFP-NL2 channel that resided within the mask (YFP-NL2-labeled voxels outside the mask were then discarded). In the same process, cell somas were removed before further analysis. Custom Matlab programs were used to generate dendritic skeletons and to identify puncta, dendritic lengths and dendritic areas as previously described . Dendritic density represents the total dendritic length divided by the total area of the dendritic territory.
In order to assess whether NL2 signal significantly overlaps with other synapse markers, a custom Matlab program was used to calculate the two-dimensional cross-correlation coefficients of the signals from both channels (Josh Morgan and Daniel Kerschensteiner). This approach does not require identification of puncta, yet allows us to determine whether the fluorescent signals in the two channels are spatially correlated, or are randomly associated (random association determined by rotating one image 180° relative to the other image).
Within an optical plane, the signal intensity of a pixel in the green channel was compared with the signal intensity of the corresponding pixel in the red channel (0,0 location). To obtain the two-dimensional correlation plot, the intensity of a green pixel was compared with the intensity of red pixels (or vice versa) to the right, left, top and bottom, displaced from the reference pixel up to 2.8 μm. A pixel intensity threshold of 20 to 35, defining the background, was used while analyzing images obtained from wild-type mouse retina. Thresholding was not used when analyzing images obtained from the YFP-NL2 transgenic mouse line due to the lower immunohistochemistry background level in the transgenic retina.
Representative examples shown in the figures are cross-correlation coefficients calculated based on 20 to 45 image planes (z-step size = 0.3 μm) of an imaged field of view. The equation used to calculate the correlation coefficient is:
where R(r, g) is the correlation coefficient of the red and green channel and C is the covariance of the corresponding channels.
For two identical images, the correlogram peak is 1. The value of the positive peak for two identical images is only weakly influenced by the absolute intensities of the pixels within an image but strongly corresponds to how spatially 'similar' the two images are. For overlapping green and red puncta, the peak of the two-dimensional correlation plot falls off symmetrically on all sides. The correlograms with positive peaks have a full-width at half maximum of less than 1 μm, suggesting that there exist structures in the two images within less than 1 μm overlap in space.
Retinal flat mounts were prepared as described above for cell transfection, but mounted on white filter paper (Anodisc 13, Whatman Inc., Piscataway, NJ, USA) for better visualization. Recordings were performed at room temperature and retinas were maintained in bicarbonate-buffered mACSF containing 125 mM NaCl, 2.5 mM KCl, 2 mM CaCl2, 1 mM MgCl2, 1.25 mM NaH2PO4, 11 mM glucose and 26 mM NaHCO3 (equilibrated with 95% O2 and 5% CO2). Whole-cell recordings were performed with electrodes (4 to 8 MΩ) filled with 120 mM Cs-gluconate, 1 mM CaCl2, 1 mM MgCl2, 10 mM Na-HEPES, 11 mM EGTA, and 10 mM TEA-Cl (pH 7.2 adjusted with CsOH). For some experiments 2 mM QX314 was also included in the patch pipette. A liquid junction potential of 15 mV was corrected before the cell was attached, and series resistance was not compensated. Data were acquired using an Axopatch 200 B amplifier (Molecular Devices), low-pass filtered at 2 kHz and digitized at 5 kHz. sIPSCs were recorded at -0 mV, the reversal potential of cation currents in our recording conditions. Area and amplitude thresholds (Mini Analysis, Synaptosoft, Decatur, GA, USA) were optimized to detect > 90% of the events identified by eye for the entirety of recordings analyzed. For overlapping events, the baseline for amplitude measurement of each event was estimated from exponential decay extrapolation of the previous event.