Animals
All procedures were in compliance with the Association for Research in Vision and Ophthalmology Statement for the use of Animals. The protocols approved by the National Eye Institute Animal Care and Use Committee. Animals, mixed gender were housed under a 12 hours light/12 hours dark cycle and given ad-libitum access to food and water. The two congenic C3H lines [14] (C3Hrd1/rd1 and C3Hwt/wt) were re-derived and maintained at Charles River. The rd7 colony was obtained from Jackson Laboratories, (USA). The homozygous knockout mice Nxnl1 and their controls are described in [9]. The Nxnl2 line is described in Jaillard et al. (manuscript in preparation). All three lines have a pure BALB/c background. Day of birth was designated postnatal day 0 (PN0).
Tissue collection
Retinal tissues were obtained from C3Hwt/wt and C3Hrd1/rd1 eyes aged PN15 to 90 days. All the right eyes of each genotype (n = 7) per age (PN15, 35, 43, 60 and 90 days) were analyzed by automated counting and the left eyes by stereological counting. We then replicated the experiment by inverting eye polarity. Animals were sacrificed by decapitation. The orientation of the eyes was marked on the limbus at 12 O'Clock position before enucleating. Neural retinas were dissected in phosphate-buffered saline (PBS) at room temperature from the posterior eyecup followed by immersion in cold 4% paraformaldehyde (PFA) in PBS at pH 7.4. For cryosectioning, eyes were fixed in cold PFA 4% in PBS overnight after puncturing the cornea then left 1 hour in 10%, 11/2 hour in 20% and 3 hours in 30% sucrose then quickly embedded in 4% gelatin with liquid nitrogen. The section plane was extended along the vertical axis from the optic nerve head in the posterior retina to the cornea.
Production of polyclonal antibodies against S-Opsin
A peptide corresponding to the second extracellular loop of mouse short wave opsin (CGPDWYTVGTKYRSE) was synthesized and coupled to ovalbumin and injected into two New-Zealand rabbits. The specificity of the polyclonal antibodies was confirmed by immunohistochemistry and compared with the pre-immune serum on rd7 retina (not shown).
Immunohistochemistry on flat-mounted retina
Fixed retinas were rinsed with PBS three times, before permeabilization (30 seconds for all genotypes and 5 minutes for C3Hrd1/rd1) in PBS containing 0.1% Triton X-100, followed by incubation 1 hour in blocking buffer PBS containing 1% of bovine serum albumin (BSA), 0.1% Tween-20 and 10% normal goat serum (NGS). After washing, retinas were labeled for 3 hours at room temperature or overnight at 4°C with Alexa Fluor 594-coupled peanut agglutinin lectin from arachis hypogae (PNA) [15] (1:40, Invitrogen, USA). For S-Opsin, the retinas were incubated with anti-S-Opsin (1:400) 4 hours at 37°C followed by incubation overnight at 4°C in the same blocking buffer. The retinas were washed six times with blocking buffer without NGS and incubated with secondary antibody, goat anti-rabbit IgG conjugated to Alexa Fluor 488 (1:400, Invitrogen, USA). We then performed four incisions of the whole retina to obtain flat-mounted retina with the photoreceptor layer facing up. Four retinas were placed on a vertical line. For automated acquisition, the first and fourth retinas were kept at a distance of 5 mm from the upper and lower edges of the slide respectively.
Immunohistochemistry on frozen eye sections
Ten μm thick frozen eye sections were dried at room temperature for immunohistochemistry as previously described [9] with few modifications. No permeability was applied and PNA was added for 3 hours at room temperature in the blocking buffer before immerging slides into blocking buffer with Dulbecco's Modified Eagle Medium (DMEM), 4.5 g/l of glucose, 10% of NGS and 10% of fetal calf serum, 1 hour at room temperature followed by S-Opsin incubation in the same buffer during 3 hours at room temperature. After washing, the nuclear marker 4'-6-diamidino-2-phenylindole (1:1000, DAPI, Sigma) was added with the secondary antibody. Finally the sections were mounted on glass slides with fade-resistant mounting media (Biomeda, Forster city, CA) topped with coverslip and imaged with a fluorescent microscope (Leica, USA).
Automated image acquisition of labeled cones on flat-mounted retina
An inverse microscope (Nikon, Eclipse TE200, and 2000) was equipped with a mercury super high pressure lamp, a computer driven motorized stage (Multicontrol 2000, Martzauzer, Wetzlar), an optical filter switch (Lambda 10-2, Shutter Instrument company) for two excitation filters (485 and 520 nm) and two emission filters (520 and 635 nm), a shutter driver (JML Direct Optics), two objectives 4X (0.10, infini/- , WD 30) and 40X (Plan Fluor 40X/0.75 Dic M, infini/0.17, WD 0.72), and a CCD camera (CoolSNAP FX and HQ, Photometrics). The pieces of equipment were connected (Figure 1) and controlled by algorithm (Additional file 1) developed with Metamorph (Universal Imaging Corporation, Sunnyvale, CA, USA). A maximum of 5 slides with 4 flat-mounted retinas were placed on the motorized stage platform and the position of the centre of the retina were set-up manually using a 4X objective to record the coordinates (X, Y). The objective was switched to 40X and the platform stage moves back to the first retina and the focus manually set-up on labeled cones and recorded (Z). The program offers two options for acquisition: the entire retina (grid) or a draughtboard grid. The acquisition grid was designed from nine assembled images (4X objective) and used for the acquisition (X40 objective) of ~300 fields (Figure 1) of identical surface area (0.0376 mm2), covering the entire surface of the retina (CoolSNAP FX: 217.75 × 172.53 μm, CoolSNAP HQ: 224.46 × 167.70 μm). Two incremented stages of autofocus were executed to recover the focal plane for each individual plane (~300) on the overall retinal surface. The first stage scans six different planes within the depth of the retina with an increment of 60 μm approaching the focal plane. The second stage is a fine adjustment from this position using 8 planes with a lower increment of 15 μm. Centered on this focal position, a stack of nine images per field distanced by 0.7 μm were acquired in the Z-direction (Figure 1). All the stacks were registered in Tagged Image File Format (TIFF) in a file named by the date of the day followed by an increment numbers. All images had a resolution of 650 × 515 pixels in 16-bits images (1.72 Gigabytes for the entire retina).
Global automated cone quantification
The stacks of images were treated with an algorithm using Metamorph to quantify cones (Additional file 1). A projection though the nine plans of each file was applied in order to reveal best focus plane to which was applied an auto-threshold to reduce the background. The algorithm is not perturbed by the overall orientation and twist of the cones due to the use of a morphological filter. In addition, the projection through the 9 planes that is operated before counting is creating a virtual image in which the cones are preferentially seen as a transversal section though the cell body. This threshold was used to detect dark areas resulting from retina pigmented epithelium residue and to exclude the images that do not fulfill the preset criteria. When this phenomenon is predominant, the field is excluded using the function dark_max (Additional file 1b, line 11 in Editor\variables). The image resulting from one field is segmented and the threshold is applied locally to overcome difference in brightness within an image (Additional file 1b, line 12-14 in FINDSPOT\3D Measurements). An adjustment of digital contrast was operated to discriminate cells among the cluster from background. At this stage eleven ordered variables empirically determined were called [best focus average intensity, autothreshold area %, number of cells detected on best focus before treatment, minimal intensity, number of minimal cells, max background, cluster, dark max, spot cut-off, spot size, spot surface, Additional file 1] to filter the images to be counted. The variables are: lane 1) Best Focus average intensity, average intensity of the projection of the nine images; 2) Percentage Best Focus. Threshold autothreshold area percentage, percentage of cells within the projected image over the threshold; 3) Number of cells detected with Best Focus before treatment, before the treatment with the algorithm findpot 4) Spotcutoff, the cutoff that is used to exclude objects with aberrant measures; 5) Spotsize, the size above which objects are excluded; 6) SurfSpot, the surface above which objects are excluded; 7) IntMin, the minimal intensity for an object to be retained; 8) NBObjectsMin, the minimal number of objects that should contain a projected image; 9) FondMax, the maximal intensity of the total space between the objects; 10) Cluster, The size of an object that is rejected since it is considered as a cluster of cells; 11) Dark_max, the minimal intensity of a projected image. The values above the threshold were arbitrarily equaled to zero in order to enhance the contrast. The function find-spot was used on each image filtered with the eight fits variables, to dissociate neighboring cells and by calling three variables: the spot cut-off, the spot size and the spot surface resulting in an accurate counting (Additional file 2). Finally, the number of counted cells in the selected image was mentioned with all parameters described above before closing it. The density of cones are locally associated with their coordinates (Figure 1), and used to calculate the density of cones over the surface of the retina after having removed the excluded fields.
Stereological cone Quantification
The stereological counting was achieved as described previously [5, 7]. Briefly, the cones were counted on 50-80, non-overlapping, 1,225 μm2 fields selected with a systematic random sampling procedure applied to the retinal surface from the centre of the optic nerve head over a radius of 2 mm.
Stereological automated cone quantification
The process was adapted from [5]. The area comprising the optic nerve is indicated by the operator for its exclusion. The number of fields was reduced from the global method by using a draughtboard grid. Within each field, a frame (30 μm × 30 μm) was drawn with exclusion of object in contact with one of both axes X and Y. Within these fields, the cones were counted with the parameters used for the global methods. The results are expressed as the average cone density over the counted fields.
Spatial distribution
The spatial distribution of cones was inferred from the local densities with their coordinates X, Y recorded in an Excel file and the orientation (Ventro-Dorsal, Naso-Temporal). A virtual eye fundus representation was designed to visualize the data. The value X and Y were used to calculate the position of the optic nerve and the local density of cones in 9 rows taken from this centre are projected on the disk annulus by annulus. To represent the amount of cones in each segment, a color scale was applied using multiples of 45 (45, 90,......,405).
Statistical Analysis
Statistical analyses were performed with GraphPad Prism5.0, with p < 0.05 considered significant. Unpaired Student's t-test was used to compare density of cones levels in rd1 at different ages with a Welch's correction. In figure 4, we compared the value between three groups, by a two-way anova where one factor stands for the days and the second for the technique. For the normalized test using the quantile-quantile normalization [16], the quantiles of the set of measures obtained by each technique are identical. To apply this normalization while avoiding any bias due to overrepresentation of some dataset, curves have been made first comparable by resizing each set in such a manner that for a dataset the number of measurements is equal using each method. As this can be done in several ways by taking out 4 values, (one for PN60 for global automated method, one for PN60 and for PN90 for stereological method and one for PN35 for automated stereological method), all 2,353 possible combinations were tested. For each combination the two-way Anova led to a p-value of one after normalization indicating that the there is no difference between the three methods. To test for cone regionalization in figures 7C and 7E, a paired t-test was used to compare the means cone density (PNA) from the same retina of the dorsal and ventral region for Nxnl+/+ (n = 10), Nxnl1-/- (n = 7) and Nxnl2-/- (n = 8) mouse. To compare the means S-cone density (S-opsin) from the same retina of the dorsal and ventral region for Nxnl+/+ (n = 5) and Nxnl2-/- (n = 5) mouse.