Calcium dynamics change in degenerating cone photoreceptors

Cone photoreceptors (cones) are essential for high-resolution daylight vision and colour perception. Loss of cones in hereditary retinal diseases has a dramatic impact on human vision. The mechanisms underlying cone death are poorly understood, and consequently, there are no treatments available. Previous studies suggest a central role for calcium (Ca²⁺) homeostasis deficits in photoreceptor degeneration; however, direct evidence for this is scarce and physiological measurements of Ca²⁺in degenerating mammalian cones are lacking.

Here, we took advantage of the transgenic HR2.1:TN-XL mouse line that expresses a genetically encoded Ca²⁺biosensor exclusively in cones. We cross-bred this line with mouse models for primary (“cone photoreceptor function loss-1”, cpfl1) and secondary (“retinal degeneration-1”, rd1) cone degeneration, respectively, and assessed resting Ca²⁺levels and light-evoked Ca²⁺responses in cones using two-photon imaging. We found that Ca²⁺dynamics were altered in cpfl1 cones, showing higher noise and variable Ca²⁺levels, with significantly wider distribution than for wild-type and rd1 cones. Unexpectedly, up to 21% of cpfl1 cones still displayed light-evoked Ca²⁺responses, which were larger and slower than wild-type responses. In contrast, genetically intact rd1 cones were characterized by lower noise and complete lack of visual function.

Our study demonstrates alterations in cone Ca²⁺dynamics in both primary and secondary cone degeneration. Our results are consistent with the view that higher (fluctuating) cone Ca²⁺levels are involved in photoreceptor cell death in primary (cpfl1) but not in secondary (rd1) cone degeneration. These findings may guide the future development of therapies targeting photoreceptor Ca²⁺homeostasis.

Introduction

Human vision depends primarily on cone photoreceptors, which mediate high-resolution, day-light colour vision. Rod photoreceptors (rods), on the other hand, are responsible for vision under dim-light conditions, i.e. at night. Cone loss can occur as a consequence of a mutation in the cone itself (primary cone degeneration) or as a “side-effect” of rod photoreceptor loss (secondary cone degeneration). A rapid cone loss as a consequence of mutations in cone-specific Pde6c (cGMP-specific phosphodiesterase) is present in the cpfl1 mouse, an animal model for primary cone degeneration (1–3). Secondary cone degeneration, on the other hand, is observed as a consequence of primary rod photoreceptor cell death in diseases like Retinitis Pigmentosa (4,5). The rd1 mouse carries a rod-specific Pde6b mutation and is a well-known model for secondary cone degeneration (6). We used cpfl1 and rd1 mouse lines to investigate the role of Ca²⁺homeostasis in primary and secondary cone degeneration, respectively.

In the intact photoreceptor, Ca²⁺has many important functions (Fig. 1A and B) , including modulating the transduction cascade in the outer segment (OS), and triggering the transmitter release from the axon terminals (7–9). In the dark, high cGMP levels in the OS activate cyclic nucleotide-gated (CNG) channels and allow for Ca²⁺influx (10–12). In the light, Ca²⁺levels drop due to cGMP hydrolysis by phosphodiesterase 6 (PDE6) and the subsequent closure of CNG channels. Low Ca²⁺levels allow dissociation of Ca²⁺from guanylate cyclase-activating proteins (GCAPs), disinhibition of guanylate cyclase (GC), restoring cGMP levels. This close coupling between cGMP and Ca²⁺levels, together with the finding that mutant photoreceptors show an abnormal cGMP accumulation (13–15), led to the hypothesis of cGMP-mediated Ca²⁺“overload” triggering degeneration (16).

Figure 1.

Photoreceptor Ca²⁺signalling and experimental model. (A) The phototransduction cascade tightly regulates outer segment (OS) Ca²⁺influx via cGMP-dependent cyclic nucleotide-gated (CNG) channels. Currents through CNG channels modulate the membrane potential and, thereby, voltage-gated Ca²⁺channels (VGCCs) in the photoreceptor axon terminal and glutamate release. (B) Ca²⁺is highly compartmentalized in healthy photoreceptors and diffusion is restricted between OS, inner segment (IS), soma, and axon terminal. (C) Vertical slice of a transgenic HR2.1:TN-XL mouse retina (outer retina shown), with cones selectively expressing the Ca²⁺biosensor TN-XL (green; cf. green areas in (B); grey, photoreceptor nuclei stained with DAPI). (D) Ca²⁺responses (bottom) to a 1-s light-flash recorded from a cone axon terminal (top; yellow circle) in a slice retinal preparation (mean, black trace; single trials, grey traces). Ca²⁺signals shown as the change in ratio (ΔR = F_A/F_D) of the acceptor fluorescence (F_A) to the donor fluorescence (F_D) (Methods). PDE6, phosphodiesterase 6; cGMP, 3’-5’-cyclic guanosine monophosphate; ONL, outer nuclear layer; OPL, outer plexiform layer. Scale bars: c, 20 µm; d, 5 µm.

Many studies support this “high Ca²⁺hypothesis” (15,17–20), however, there are also several contradicting results: For instance, treatments with Ca²⁺channel blockers, aimed at reducing Ca²⁺overload and, thus, preventing or at least delaying photoreceptor degeneration, produced inconsistent results for primary rod degeneration (21–24). Furthermore, data from CNGA3-deficient mice show that in the absence of cone CNG channels (and CNG channel-mediated Ca²⁺influx) cones still degenerate (25). Additionally, a functional study in Pde6c^w59 mutant zebrafish found no increase in cone Ca²⁺levels, but instead a suppression of (spontaneous) Ca²⁺transients (26). Finally, also an alternative hypothesis has been promoted, which is based on the idea that photoreceptor death might be triggered by too low Ca²⁺(i.e. the “low Ca²⁺hypothesis”) (27).

Here, we set out to study Ca²⁺dynamics in degenerating cones using a transgenic mouse line (HR2.1:TN-XL) that expresses a genetically encoded ratiometric Ca²⁺biosensor selectively in cones and allows monitoring of light-evoked Ca²⁺signals in individual cone axon terminals (28–31). In healthy photoreceptors, opening of CNG channels in the dark depolarizes the cone membrane. Voltage-gated Ca²⁺channels (VGCCs) in the synaptic terminal translate this depolarization into a Ca²⁺influx, which in turn triggers synaptic vesicle fusion with the membrane and glutamate release (32–35). Thus, the presynaptic Ca²⁺level in photoreceptors can serve as a proxy for upstream CNG channel activation in the OS (28).

To study cone Ca²⁺in primary and secondary cone degeneration, we crossbred the HR2.1:TN-XL mouse line with the cpfl1 and rd1 PDE6 mutant lines, respectively. Ca²⁺signals were compared between these cpfl1 and rd1 mutant crossbreds and the “wild-type” retina of HR2.1:TN-XL mice. We observed differential changes in Ca²⁺in the two mutant lines: In primary cone degeneration, Ca²⁺levels were more variable, noisy, and tended to be higher than in wild-type (wt). Notably, a substantial fraction of these cones showed robust light-evoked Ca²⁺responses. On the other hand, in secondary cone degeneration, cone Ca²⁺levels appeared to be lower and less variable compared to wt and did not show light-evoked Ca²⁺responses. Our data are consistent with the idea of higher Ca²⁺levels being involved in primary (cpfl1), but not in secondary (rd1) cone degeneration.

Materials and Methods

Animals

The transgenic mouse line HR2.1:TN-XL (C57BL/6J background) expresses the Ca²⁺biosensor TN-XL (36) under the control of the human red opsin promoter (HR2.1) selectively in cone photoreceptors (28) (Fig. 1C). To study primary and secondary cone degeneration, we crossbred the biosensor line with the original rd1 and cpfl1 mutant animals to generate the HR2.1:TN-XL x cpfl1 (C57BL/6J background) and HR2.1:TN-XL x rd1 (C57BL/6J x C3H background) lines. Using PCR amplification and NdeI digestion, we verified that these lines were free of the rd8 (Crb1) mutation (37). While a previous study on C3H rd1 and congenic C3H wt animals, the latter of which show normal electroretinographic responses, had in principle ruled out the presence of the Nob5 (Gpr179) mutation (38), we additionally performed a PCR-based analysis employed recently in a survey on C3H mouse lines (39). As shown in Supplementary Material, Figure S1, the Nob5 mutation is absent in C3H rd1 mice used in our study. For simplicity, in the later parts of the manuscript, we refer to the biosensor lines as wt, cpfl1, and rd1. For Ca²⁺imaging experiments (see below), we used mice in two time windows (postnatal days 18-20 (P18-20) and P30-33), to which we refer to as P18+ and P30+, respectively. See results for a justification for the use of these time-frames. In general, we used mice irrespective of gender.

Prior to Ca²⁺imaging, the mice were dark-adapted for approx. 2 hours and then anaesthetized using isoflurane (CP Pharma, Burgdorf, Germany). For immunostainings, we used P20 and P30 as analysis time points and the animals were anaesthetized with CO₂. In both cases, anaesthetized mice were killed by decapitation. All procedures were performed in accordance with the law on animal protection issued by the German Federal Government (Tierschutzgesetz) and approved by the institutional animal welfare committee of the University of Tübingen.

Immunohistochemistry and fluorescence microscopy

Eyes were marked on the nasal side prior to enucleation. They were fixed in 4% PFA in 0.1 M phosphate buffer saline (PBS, pH 7.4) (∼1 hour) and cryo-protected in 30% sucrose in PBS at 4^oC overnight. Then, eyes were embedded in Tissue-Tek OCT compound (Sakura Finetek Europe, Alphen aan Den Rijn, Netherlands) and stored at −20 ^oC until cryo-sectioning into 12 μm thick vertical sections. Sections were rehydrated with PBS, permeabilized in 0.1% Triton X-100 in PBS containing blocking solution (10% goat or donkey serum, 1% BSA). As primary antibodies, we used rabbit anti-GFP (1:600; Millipore, Darmstadt, Germany) or mouse anti-GFP (1:500; Abcam, Cambridge, UK); sheep anti-cGMP (1:500, from Harry W.M. Steinbusch, Maastricht University, Maastricht, Netherlands) and guinea pig anti-glycogen phosphorylase (1:1000, kind gift from B. Pfeiffer-Guglielmi, University of Tübingen, Tübingen, Germany). As secondary antibodies, we used Alexa Fluor 568 donkey anti-sheep (1:500; Invitrogen, Carlsbad, USA), Alexa Fluor 488 goat anti-rabbit (1:500; Molecular Probes, Eugene, USA), Alexa Fluor 488 goat anti-mouse (1:500; Molecular Probes) and FITC goat anti-guinea-pig (1:300; Sigma, Saint Louis, USA). The terminal deoxynucleotidyl transferase dUTP nick end labelling (TUNEL) assay (Roche Diagnostics, Mannheim, Germany) was performed using the TMR red kit as per manufacturer’s instruction. Z-stack and mosaic images were captured on an Imager Z1 ApoTome microscope using a 20x air objective (0.8 NA; cf. Fig. 2A, wt and cpfl1) or a 63x oil immersion objective (1.4 NA; cf. Fig. 2A, rd1), and AxioVision (v.4.8.1.0) software (Carl Zeiss Microscopy GmbH, Oberkochen, Germany). Figures were assembled either in Adobe Photoshop CS5 (Adobe Systems, San Jose, USA) or Canvas 11 (ACD Systems International Inc., Seattle, USA).

Figure 2.

cGMP accumulation in photoreceptors of Pde6 mutant mice expressing the HR2.1:TN-XL Ca²⁺-biosensor. (A) Cross-sections of wt (left), cpfl1 (centre), and rd1 (right) outer retinas immunostained for cGMP (magenta) at P20 (top) and P30 (bottom). The biosensor TN-XL is shown in green. (B) Cone density (in # cells/10 μm) in rd1 animals is significantly reduced at P20, while at P30 both cpfl1 and rd1 retinas show cone loss. (C) The percentage of cGMP-positive cones is strongly increased in cpfl1 retina at P20 and P30, while wt and rd1 cones are negative for cGMP. Scale bars: a, 10 µm. P20 data obtained from 9 sections from 3 wt mice (9/3); cpfl1, 11/3; rd1, 9/3; P30: wt, 14/4; cpfl1, 11/3; rd1, 9/3); *** =P < 0.001. OPL, outer plexiform layer, ONL, outer nuclear layer.

Analysis of immunodata

The data were obtained from at least three different animals and for each animal at least three immunostained vertical sections were quantified, using mosaic images acquired at 20× magnification. For analysis, the area of the outer nuclear layer (ONL) and the length of the retinal sections were determined using the Zeiss Axiovision software. cGMP positive cells were counted in both biosensor and non-biosensor lines and were expressed as percent positive cells in the ONL (rd1 retina), or as percent positive cones (wt, rd1 and cpfl1). To estimate cone density (number of somata per 10 μm) for P20, P24 and P30 in both biosensor and non-biosensor lines, cones were labelled using glycogen phosphorylase (40) and counted in defined areas. Cones were only included if at least the inner segment (IS) and the soma could be clearly identified. To quantify dying cells, the in situ TUNEL assay was also performed on vertical sections (Supplementary Material, Fig. S2). We measured the area of the ONL, divided it by the total number of cells (obtained from DAPI staining) in that area and determined the fraction of TUNEL-positive nuclei (Supplementary Material, Fig. S3). Statistical comparisons were made using the Wilcoxon rank-sum test using IGOR Pro (Wavemetrics, Lake Oswego, USA).

Two-photon Ca²⁺imaging

The preparation of retinal slices for Ca²⁺imaging was performed as described previously (28,31). In brief, eyes were enucleated and dissected in carboxygenated (95% O₂, 5% CO₂) artificial cerebrospinal fluid (ACSF) solution, which contained (in mM): 125 NaCl, 2.5 KCl, 2 CaCl₂, 1 MgCl₂, 1.25 NaH₂PO₄, 26 NaHCO₃, 0.5 l-glutamine, and 20 glucose; maintained at pH 7.4 using carboxygen. Consecutive vertical slices (∼200 µm thick) were cut using a tissue chopper (41) and then transferred to the recording chamber of a two-photon microscope, where they were constantly perfused with warmed (∼37 °C) ACSF. The microscope (for details, see (42)), consisted of a customized MOM (movable objective microscope; designed by W. Denk, MPI of Neurobiology, Martinsried; purchased from Science Products/Sutter Instruments, Novato, USA), equipped with a mode-locked Ti:Sapphire laser (MaiTai-HP DeepSee; Newport Spectra-Physics, Darmstadt, Germany) tuned to 860 nm, two fluorescence detection channels (483 band-pass (BP) 32; 535 BP 50; AHF, Tübingen, Germany), and a 20x water-immersion objective (XLUMPlanFL, 0.95 NA; Olympus, Hamburg, Germany). The system was controlled by the image acquisition software ScanM (by M. Müller, MPI of Neurobiology, Martinsried, Germany, and T.E.) running under IGOR Pro 6.3, or CfNT (M. Müller, MPI, Martinsried, Germany). A custom-built dichromatic light stimulator (29), equipped with two band-pass filtered LEDs (UV filter: 360 BP 12; green: 578 BP 10) and mounted below the recording chamber, was used to present temporally-modulated, full-field light stimuli (approx. 2 mm in diameter) to the retinal slices. Ca²⁺recordings were performed with a constant background illumination of 10⁴ P*s ^− 1 (photo-isomerization rate) for at least 15 seconds. Light stimuli consisted of a series of 1-second bright flashes with 4-second intervals. Flashes evoked similar photo-isomerization rates in both medium (M-) and short (S-) wavelength-sensitive cones (green LED: 6.7; UV LED: 6.5). At these light levels, rods are expected to be saturated and, therefore, light-evoked responses should originate in cones and not in electrically coupled rods. To record Ca²⁺levels and light-evoked Ca²⁺responses in cone axon terminals, we captured image series (128 x 16 pixels at 31.25 Hz), aligned with the outer plexiform layer (OPL) and covering an area of ∼75 x 10 µm on the slice (Fig. 1D). We recorded both fluorescence channels, as TN-XL allows ratiometric measurements via FRET between its fluorophores enhanced cyan fluorescent protein (eCFP, donor) and citrine (acceptor).

In a subset of experiments, absolute Ca²⁺concentrations were measured using an “ex-vivo slice calibration” procedure (31). We recorded the fluorescence ratio (see Analysis of Ca²⁺imaging data) first in a standard ACSF solution (R_base, see below and Fig. 1D), then in Ca²⁺-free ACSF supplemented with ionomycin (5 μM) and EGTA (10 mM) to determine the minimum ratio (R_min), and finally in ACSF supplemented with ionomycin (5 μM) and high Ca²⁺(2.5 mM) to determine the maximum (saturating) ratio (R_max). The absolute Ca²⁺concentration was estimated using R_min, R_max, the donor fluorescence with a saturating Ca²⁺level (F_{D, Ca-bound}) and under Ca²⁺-free conditions (F_{D, Ca-free}), as well as the in-vivo dissociation constant (K_d=0.77, see (43)) for TN-XL:

$$\left[{Ca}^{2+}\right]=K_d\frac{R-R_{min}}{R_{max}-R}·\frac{F_{D,Ca-free}}{F_{D,Ca-bound}}$$

Analysis of Ca²⁺imaging data

All Ca²⁺data were analysed using custom scripts for IGOR Pro 6.3 (Wavemetrics). To extract cone Ca²⁺signals, regions of interest (ROIs) were drawn manually around the cone terminals (Fig. 1D) and the pixels within ROIs were averaged for each image frame. An ROI placed outside the OPL was used to determine the baseline fluorescence in both detection channels for background subtraction. The relative Ca²⁺level was then determined as the ratio (R) between acceptor and donor fluorescence (F_A/F_D) signals after background subtraction. For each ROI, stimulus trials were averaged and the mean trace was filtered (box-car, 320 ms filter-width).

Relative resting Ca²⁺level (R_base) was defined as the mean prior to the light flash (cf. Fig. 3A). Further, we quantified the area between the base line and light response (“response area”, R_A), response amplitude (ΔR), response rise (t_20–80). The latter was determined by fitting a sigmoid to the response rise to determine the duration between the time points when 20% and 80% ΔR were reached. As a measure for the “slowness” of a light response, we determined a “response lag” (t_lag), which we defined as the duration between t_20% and t_decay, the time point when the Ca²⁺level started to recover (= to increase). In addition, we quantified response noise (R_noise), defined as s.d. of the traces (trials) after subtracting the low pass-filtered mean response (cf. Fig. 6A).

Figure 3.

Cone resting Ca²⁺levels. (A) Resting Ca²⁺level (R_base) was determined as the mean trace prior to light flash. (B) Distribution of relative cone R_base in the three mouse lines at P18+ (left; wt, 853 cells from 3 mice (853/3); cpfl1, 944/4; rd1, 878/5) and P30+ (right; wt, 1,825/8; cpfl1, 1,048/8; rd1, 818/5; thick outline, mean distribution across mice; grey outlines, distributions from individual mice). Distributions were statistically compared using the K-S test (5% alpha level; **P < 0.01; *** P < 0.001; n.s.  0.37). (C) Box-and-whisker plot of estimated absolute cone Ca²⁺concentrations at P30+ for the three lines (wt, 16/2; cpfl1, 16/4; rd1, 16/3). Data points ≤ 10 nM were considered to represent unrealistically low Ca²⁺levels (possibly due to a limitation of TN-XL, as discussed in (28)) and were thus excluded from the analysis.

Previous studies showed that during slicing, cells close to the surface might be mechanically damaged and show abnormal Ca²⁺ratios (28). We tried to address this issue by recording from cones at least 20-50 µm in the slice and by excluding cones with unstable baseline and/or extreme noise level. We used the noise distribution of the wt P30+ dataset and determined from the histogram the noise level (R_noise,max) that excluded the 5% cones with the highest noise (95^th percentile). Then, from all datasets we excluded cones with R_noise>R_noise,max unless they displayed clear light responses (R_A > 10 R_A,noise, with R_A,noise defined as the s.d. over the whole 5-s trace times the trial duration).

To compare responsiveness to light stimuli between the different data sets, we categorized the cones into “responsive” and “non-responsive”. To include only cones with robust light responses into the “responsive” category, we again used the wt P30+ dataset: We determined the value (R_A,50%) that divided the R_A histogram (cf.Fig. 4C) into halves and defined the 50% cones with the larger R_A values as responsive. This R_A,50% was then applied to the other datasets. Note that cones for which the response curve could not be fitted (to determine timing parameters, see above) were not classified as “responsive”, even if they fulfilled the R_A>R_A,50% criterion.

Figure 4.

Light-evoked cone Ca²⁺responses. (A) Determining area under the response curve (R_A) and relative resting Ca²⁺level (R_base). (B) Exemplary Ca²⁺signals (mean of ≥ 10 trials) in response to a light flash recorded from individual wt (black), cpfl1 (blue), and rd1 (red) cones at P30+. (C) R_A as function of R_base at P18+ (left) and P30+ (right). Each dot represents a single cone (colours as in B; P18+: wt, 853 from 3 mice (853/3); cpfl1, 944/4; rd1, 878/5; P30+: wt, 1,825/8; cpfl1, 1,048/8; rd1, 818/5). (D) Distribution of R_A in “responsive” cones (for definition, see Methods); only data points between the two dashed lines in (C) were included (colours as in b). (E) Box-and-whisker plot of normalized R_A at P18+ and P30+. At P18+, wt: 0.65 (median) (black, 160/3), cpfl1: 0.62 (blue, 32/4). At P30+, wt: 0.84, (730/8), cpfl1: 1.13, (99/8). R_A values were statistically compared using the Wilcoxon rank-sum test (5% alpha level; ***P < 0.001; n.s. P = 0.47).

Statistics

In total, we analysed n = 2,678 (89% of 3,007 cells) wt, n = 1,992 (81% of 2,469 cells) cpfl1, and n = 1,696 (92% of 1,845 cells) rd1 cones. To compare Ca²⁺level distributions statistically, we fitted the distribution of R_base for each line and time point with a Gaussian that was convoluted with an exponential (ExpModGauss in IGOR Pro) and used R_base at the distribution peak to centre the histograms for comparison between data sets (Fig. 3B). R_base distributions were compared using the Kolmogorov-Smirnov test (K-S test, IGOR Pro) for an alpha level of 5%. To compare the light response sizes statistically, we first normalized R_A of the responsive cells for all the datasets to the mean R_A of wt cones at P30+. To compare noise levels, we normalized R_noise measured in the mutant lines and at other time windows to R_noise,max (see above). The Wilcoxon rank-sum test (as implemented in IGOR Pro; 5% alpha level) was used to compare R_A, t_lag, and R_noise between mouse lines and time windows. These data are shown as box-and-whisker plots (top whisker: 90^th percentile; bottom whisker: 10^th percentile).

Results

Normal wt cones and Ca²⁺biosensor-expressing wt cones were previously found to be functionally equivalent, showing no differences in electroretinographic (ERG) recordings and opsin expression (28–30). To ascertain that the expression of the genetically encoded Ca²⁺sensor TN-XL in mutant photoreceptors would not alter the degeneration phenotype, we first assessed degeneration markers in HR2.1:TN-XL crossbred mutants. For our analysis, we focussed on two time-windows, postnatal days (P) 18-20 and P30-33, to which we refer to as P18+ and P30+, respectively. In both rd1 and cpfl1 genotypes, these two periods correspond to times just beyond the onset of cone degeneration (P18+) and when roughly 50% of all cones are gone (P30+) (3,44).

Presence of biosensor does not alter Pde6 mutant phenotype

Cone morphology in the wt HR2.1:TN-XL retina (control) appeared normal, whereas HR2.1:TN-XL x cpfl1 cones were characterized by misplaced somata (Fig. 2A), as described before for non-biosensor cpfl1 cones (3). Due to primary rod loss, HR2.1:TN-XL x rd1 retina showed a marked thinning of the outer nuclear layer (ONL) (45), accompanied by dramatic changes in cone morphology (rounded somata with strongly altered neurite morphology; Fig. 2A), virtually identical to what has been described in the (non-biosensor) rd1 mouse (5). We counted cones in the three mouse lines and found cone densities to be significantly reduced over time in HR2.1:TN-XL x cpfl1 and HR2.1:TN-XL x rd1 (Fig. 2B), indicating severe cone degeneration compared to wt HR2.1:TN-XL. Cone densities were within the ranges reported for non-biosensor cpfl1 and rd1 (Table 1). In addition, we performed TUNEL assays to evaluate the number of dying cells at P20 and P30 (Supplementary Material, Fig. S2) and found the fraction of TUNEL-positive cells in HR2.1:TN-XL x cpfl1 and HR2.1:TN-XL x rd1 lines to be similar to that reported for cpfl1 and rd1 before, respectively (Table 1) (3,38).

Table 1.

Cone density and percentage of TUNEL-positive cells in the studied mouse lines

	cone density[(10 µm)−1]		TUNEL positive cells[%]
	P20	P30	P20	P30
“wild-type” HR2.1:TN-XL	1.28 (1.19)	1.31 (1.35)	0.17 (0.14)	0.03 (0.03)
HR2.1:TN-XL x cpfl1	1.09 (1.14)	0.63 (0.69)	0.32 (0.38)	0.17 (0.17)
HR2.1:TN-XL x rd1	0.81 (0.80)	0.64 (0.63)	2.51 (2.77)	0.76 (0.84)
wt	1.4 [P21] (61)	1.4 (61)	0.08 (15)	0.04 (15)
cpfl1	1.12 (1.13) [P24]	0.7 (†, 61)	0.15 (15)	0.02 (15)
rd1	0.75 (62)	0.60 (61)	1.64 (15)	0.55 [P28] (15)

For calculation of cone density and % TUNEL positive cells, see Methods. To allow for a comparison with earlier literature, values shown in this table are expressed as mean, with median (where available) in parenthesis. In cases where P20/P30 values were not available for comparison, the actual postnatal age is given in square brackets. cpfl1 at P24: median 1.13 ± 0.18 SD, 14 vertical sections from 5 mice.

†

value given (“50% of wt”) was estimated using data given in ⁽⁶¹⁾. References in superscript brackets.

Next, we assessed cGMP accumulation in the outer retina (3,15): In wild-type retina, whether expressing the TN-XL biosensor or not, essentially no cGMP accumulation could be observed in the ONL, at P20 and P30 (Fig. 2A and C) (3). In rd1 retina, the overall percentage of cGMP positive cells in the ONL at P20 was similar for biosensor and non-biosensor expressing retina (rd1: 7.44 ± 1.09% SD, n = 9 vertical sections from 3 mice; HR2.1:TN-XL x rd1: 6.91 ± 0.91% SD, P =  0.5, n = 9 vertical sections from 3 mice) and all available evidence suggests that this cGMP accumulation is restricted to rods exclusively. Since degenerating cones may lose the expression of conventional markers during the final stages of cell death, in non-biosensor rd1 retina there is a degree of uncertainty as to whether rd1 cones do show cGMP accumulation or not. However, in HR2.1:TN-XL x rd1 retina, no cGMP positive cones could be detected, indicating that rd1 cGMP accumulation was indeed restricted to rods only (Fig. 2A and C). In non-biosensor cpfl1 retina, approximately 50% of the cones were positive for cGMP (3,15), similar to what was found in HR2.1:TN-XL x cpfl1 cones (Fig. 2C).

In summary, we did not find any significant differences between the cones of the TN-XL biosensor-expressing and “normal” rd1 and cpfl1 mutants, suggesting that the mutant lines crossbred with HR2.1:TN-XL strain are suitable models to study cone function in health and disease. In addition, we showed that cGMP levels are strongly elevated in cpfl1 but not in rd1 cones. In the following, we will be addressing results obtained from the TN-XL expressing animals and, for ease of use, refer to these simply as wt (wild-type), rd1, and cpfl1.

Resting Ca²⁺is very heterogeneous within cone populations and differs between mutants

To evaluate potential differences in Ca²⁺homeostasis between mutant and wt lines, we recorded Ca²⁺levels from TN-XL-expressing cones in acute vertical retinal slices using two-photon Ca²⁺imaging (Fig. 1C and D; see Methods). We first assessed relative resting Ca²⁺(R_base; Fig. 3A) and found that at the beginning of cone degeneration, at P18+, resting Ca²⁺level distributions were comparable in wt and cpfl1, whereas the distribution for rd1 was more narrow (Fig. 3B, left column). At P30+, when cone degeneration had progressed, the Ca²⁺level distribution in rd1 remained narrow, whereas the distributions for both wt and cpfl1 broadened (Fig. 3B, right column). As some broadening was also observed in the wt, it may partially reflect developmental changes. Despite this, the effect was more pronounced in cpfl1. Taken together, these data suggest that the mutations in cpfl1 and rd1 have opposite effects on cone Ca²⁺resting levels: As the degeneration progressed, cpfl1 cone Ca²⁺levels became increasingly heterogeneous, whereas rd1 cone Ca²⁺levels were (and remained) more homogenous across the population.

In a subset of experiments (n = 9 slices from 9 mice), we estimated the absolute resting Ca²⁺concentrations in cones using an ex-vivo calibration approach in retinal slices (31,46). Because the difference in relative Ca²⁺level distribution between wt and the two mutants was more prominent at P30+, we performed the calibration experiments at this age window. In wt cones, we determined an absolute cone Ca²⁺concentration of 200 nM (median; n = 15 cones in 2 slices from 2 mice), which is in agreement with the literature (35,47). In cpfl1 cones, the Ca²⁺concentration was 339 nM (n = 14 cones in 4 slices from 4 mice), with a higher variability compared to wt (Fig. 3C). While these differences were not statistically significant, the higher variability in cpfl1 Ca²⁺concentration was consistent with the results from the relative Ca²⁺distributions (Fig. 3B). Because of the low number of remaining cones in the rd1 retina at P30+, we could only get a rough estimate for the absolute Ca²⁺level in rd1 cones (180 nM; n = 4 cones in 3 slices from 3 mice; Fig. 3C). However, since most Ca²⁺values in rd1 cones were below the detection limit of the TN-XL biosensor, the actual intracellular Ca²⁺levels in rd1 cones are likely to be much lower, similar to those measured under Ca²⁺-free conditions.

Unexpected, robust light evoked Ca²⁺responses in cpfl1 cones

Previous ERG studies have reported very weak responses of cpfl1 retina to bright-field light flashes but could not pinpoint the origin of these responses (2). To test whether cpfl1 cones respond to light, we recorded cone Ca²⁺signals and presented 1-s full-field light flashes on a constant background (cf. Fig. 1D). In wt cones, a light-evoked a decrease in the fluorescence ratio (F_A/F_D) indicates a decrease in Ca²⁺level due to a light-induced hyperpolarization of the cone (Fig. 4A). To quantify Ca²⁺response sizes, we determined the response area (R_A). Light-evoked Ca²⁺responses were detected in wt, and to our surprise, also in cpfl1 cones but not in rd1 cones (Fig. 4B). In general, response size tended to increase as resting Ca²⁺levels rose.

As expected, we found the highest percentage of cones that met our conservative “responsiveness” criterion (Methods) in wt retina. The percentage of responsive wt cones increased from ∼36% at P18+ to ∼50% at P30+, (Fig. 4C and D), possibly due to postnatal maturation of mouse retina (48,49). In cpfl1 retina, we found a substantial percentage of cones that displayed light responses comparable to those measured in wt. At P18+, ∼8% of the recorded cpfl1 cones were responsive; later this fraction increased to ∼21%. Moreover, the response size dramatically increased in cpfl1 cones at P30+ (Fig. 4E). Remarkably, at this stage of cpfl1 degeneration, about half of the cones were already lost. In comparison, in the secondary cone degeneration model rd1, where cones are genetically intact, less than 1% of the recorded cones were responsive.

Next, we wanted to see if the light-evoked Ca²⁺response recorded in cpfl1 cones differed from those in wt cones. Here, we focused on response kinetics, as responses appeared “slower” in cpfl1 compared to wt (cf. Fig. 4B). First, we determined the “response lag” (Δt_lag) defined as the time period between the time when 20% of the response amplitude was reached (t_20%) and the time when the Ca²⁺level started to recover for a 1-s stimulus (t_decay; Fig. 5A). We found that Δt_lag was significantly longer in cpfl1 than in wt cones at both ages (Fig. 5B). Moreover, Δt_lag in cpfl1 cone responses increased significantly between P18+ and P30+. Taken together, our data suggest that even at P30+, a substantial fraction of cpfl1 cones generate large, although sluggish light-evoked responses.

Figure 5.

Kinetics of light-evoked Ca²⁺responses in wt and cpfl1 cones. (A) Exemplary light-evoked Ca²⁺responses (means of ≥ 10 trials; smoothed using box-car, 320 ms filter-width) at P18+ and P30+ from wt (black) and cpfl1 (blue). Response lag time (Δt_lag) was determined as the time between 20% of response rise (t_20%) and beginning of response recovery (t_decay). (B) Box-and-whisker plot for Δt_lag for 1-s light flashes. At P18+, wt: 0.91 (median), (160 cones from 3 mice (160/3)); cpfl1: 0.99, (32/4). At P30+, wt: 0.93, (730/8), cpfl1: 1.03, (99/8). Δt_lag were statistically compared using the Wilcoxon rank-sum test (5% alpha level; *** P < 0.001; n.s. P = 0.051).

Cpfl1 cones are noisier than wt and rd1 cones

We noticed that Ca²⁺traces recorded in cpfl1 cones appeared noisier compared to those in wt cones (Fig. 6A and B). This noise may reflect higher (spontaneous) activity caused by fluctuations in cGMP levels, possibly related to the cGMP accumulation we detected in cpfl1 cones (50) (cf. Fig. 2A). To include rd1 cones in this comparison, we analysed light-responsive and non-responsive cones separately (Fig. 6C and D). At P18+, noise levels were not significantly different between responsive cpfl1 and wt cones, whereas non-responsive cpfl1 cones were noisier than their wt counterparts. At P30+, both responsive and non-responsive wt cones became noisier, but much less so than cpfl1. The dramatic increase in Ca²⁺signal noise recorded in cpfl1 cones may indicate that cGMP fluctuations progressively increased in responsive cones. In contrast, rd1 cones showed significantly lower noise than wt in both time windows. The small increase in rd1 cone noise at P30+ may relate to the remodelling of the outer retina connectivity (5). In fact, this notion is supported by a study showing that the oscillatory activity that arises in the outer rd1 retina (P30 and later) at least partially depends on voltage-gated Ca²⁺channel activity in cone terminals (45).

Figure 6.

Ca²⁺noise levels in cones. (A) Exemplary of Ca²⁺response to a 1-s light flash (top, black trace). Noise was determined from traces with the low-pass filtered response (top, green trace; box-car, 320 ms filter-width) subtracted (bottom). (B) Exemplary noise traces at P18+ (top) and P30+ (bottom) from wt (black) and cpfl1 (blue) cones. (C) Box-and-whisker plot of normalized noise (R_noise) for responsive wt and cpfl1 cones at P18+ and P30+. At P18+, wt: 0.31 (median) (160 cones from three mice, 160/3), cpfl1: 0.36 (32/4). At P30+, wt: 0.47 (730/8), cpfl1: 0.59 (99/8). (D) Box-and-whisker plot of normalized R_noise for non-responsive wt, cpfl1, and rd1 cones at P18+ and P30+. At P18+, wt: 0.33 (597/3), cpfl1: 0.40 (870/4), rd1: 0.26 (877/5). At P30+, wt: 0.41 (894/8), cpfl1: 0.51 (841/8), rd1: 0.29 (816/5). R_noise values were statistically compared using the Wilcoxon rank-sum test (5% alpha level; ***P ≤ 0.001; n.s. P = 0.10).

Discussion

In many inherited retinal diseases, mutations that affect the photoreceptor OS lead to cell death, and the pathways associated with this are often linked to aberrant Ca²⁺signalling. Here, we monitored Ca²⁺dynamics directly during both primary and secondary cone degeneration, to show that Ca²⁺signalling was indeed altered. Remarkably, although cpfl1 cones suffer from primary Pde6c mutations, a subpopulation was found to be functional, while genetically intact rd1 cones were entirely dysfunctional. Our results raise new questions as to the possible involvement of Ca²⁺noise and random fluctuations, which could explain the stochastic nature of cone cell death (51–53). In addition, our study highlights the importance of using in vivo single-cell imaging techniques to improve our understanding of the intracellular processes leading to retinal degeneration.

Cone Ca²⁺dynamics are altered in primary cone degeneration

Cpfl1 cones carry Pde6c mutations (1), exhibit cGMP accumulation, and are expected to show altered Ca²⁺homeostasis (3,16). In line with previous studies, we found that the percentage of cones showing cGMP accumulation remained constant between P20 and P30. High cGMP is predicted to cause changes in Ca²⁺homeostasis, as evidenced indirectly by the increased activity of Ca²⁺-dependent calpain-type proteases (3,15). Our study confirms this prediction: Compared to wt, we found that cpfl1 cones showed higher relative resting Ca²⁺levels. In line with this, we found in our calibration experiments that cpfl1 cones tended to have elevated absolute Ca²⁺concentrations. In addition, cpfl1 cones had larger light responses, their resting Ca²⁺was much more heterogeneous, and they displayed increased Ca²⁺noise levels. Overall, our results on primary cpfl1 cone degeneration support the high Ca²⁺hypothesis.

These results are interesting in the context of previous studies aimed at developing treatments for hereditary retinal degeneration. The use of inhibitors of the synaptic L-type VGCCs induced a delay in rd1 photoreceptor degeneration (21), which implied that photoreceptor Ca²⁺overload (16) would occur primarily via the synaptic axon terminal. We show that Ca²⁺levels in the axon terminals of PDE6 mutant cones may indeed be increased; however, these changes are tightly linked to primary alterations of Ca²⁺in the cone OS, as evidenced by our light stimulation experiments. Thus, when therapeutic strategies aimed at blocking either Ca²⁺-permeable channels in the synapse (i.e. VGCCs) or in the OS (i.e. CNG-gated channels) (38,54) are compared, the latter appear to be a more suitable target.

Cone Ca²⁺dynamics in secondary cone degeneration

In contrast to cpfl1 cones, rd1 cones are genetically intact and did not show cGMP accumulation. Yet, major cone loss was evident already at P20, progressing rapidly at P30. In rd1, cone OS and overall morphology changed dramatically in the absence of supporting rods as the photoreceptor layer thinned out. In addition, virtually no rd1 cone responded to light, in line with previous ERG studies (23,38,54–56).

In this scenario, a modulation of cone terminal Ca²⁺via the OS seems questionable. The resting Ca²⁺distribution in the rd1 cone population was significantly different at both P18+ and P30+ compared to wt, and rd1 cone noise levels were significantly lower at both time points. When Ca²⁺was washed out of rd1 cones using ionomycin, the resting Ca²⁺levels hardly changed at all. These results altogether point at a very low resting Ca²⁺in cones upon loss of rods and structural disruption of the outer retina, indicating that rd1 secondary cone cell death may be caused by exceedingly low Ca²⁺levels ((57); discussed in (11,27)). This conclusion would have important implications for therapeutic developments, which often aim at lowering intracellular Ca²⁺. While such an approach may be viable for primary cone degeneration (see above) it could be detrimental in secondary cone degeneration. To better understand Ca²⁺dynamics in secondary cone degeneration, it may be necessary to study further animal models that exhibit a slower cone loss, not overlapping with postnatal retinal development, such as the Pde6b-mutant rd10 mouse (4,15).

Cpfl1 cones show light evoked responses

To our surprise, a substantial number of cpfl1 cones responded to light stimulation. These Ca²⁺responses were larger and slower compared with wt cones, possibly due to low remnant PDE activity. Interestingly, cones treated with the PDE6 inhibitor zaprinast exhibited similar changes in response time course (28). Zaprinast application also caused a (reversible) increase in resting Ca²⁺, consistent with our finding of higher absolute Ca²⁺in cpfl1 cones. The cpfl1 allele carries two different mutations, one 116bp intronic insertion between exons 4 and 5 resulting in a splice defect, and one 1bp deletion resulting in a frameshift in exon 7 (1). These two mutations together are expected to abolish PDE6C function completely. Nevertheless, a small proportion of correctly spliced Pde6c transcript carrying only the 1bp deletion appears to be present in cpfl1 cones (1), suggesting that they express lower amounts of PDE6C protein. Due to the severity of the genetic damage, it is unlikely, however, that this protein is functional. On the other hand, cpfl1 mice do show weak responses in photopic ERG recordings (2). Here, a possible explanation is that cpfl1 cones recruit low levels of rod Pde6a, which would be expected to slow-down cone response kinetics (58), similar to what we observed. Irrespective of the underlying mechanism, our results show that cpfl1 mice do have a residual cone function at least until one month postnatal.

Degenerating cones show altered noise levels

In wt cones, Ca²⁺noise levels became significantly higher from P18+ to P30+, suggesting changes in PDE/cGMP/Ca²⁺signalling with time, as the development and maturation of cones and retinal circuitry progresses (48). Previous studies have found that cone noise depends on variations in CNG channel activity and fluctuations in cGMP levels in the OS (50,59). Moreover, cone noise persists in the downstream circuitry (60).

Our data show altered Ca²⁺noise levels in both primary and secondary cone degeneration. At P30+, in both responsive and non-responsive cpfl1 cones, Ca²⁺noise was significantly higher than in their wt counterparts, possibly due to the higher overall Ca²⁺levels in cpfl1 cones. It is tempting to speculate that such variations in Ca²⁺noise (or Ca²⁺fluctuations above or below a critical threshold) over time could contribute to the stochastic nature of photoreceptor cell death. Although at the tissue level, photoreceptor degeneration appears as a well-timed process, the time point when cell death is initiated at the level of the individual photoreceptor is random; it is inherently impossible to predict when a specific photoreceptor is going to die (51–53).

The precise processes that may be triggered by Ca²⁺ fluctuations remain unknown, but could potentially be linked to activation of Ca²⁺-dependent kinases (61,62) with bearings on gene expression (63) and/or Ca²⁺-dependent calpain-type proteases (3,19). On the other hand, a recent in vivo Ca²⁺imaging study followed degenerating cones over a time-course of several hours in intact zebrafish larvae carrying the cpfl1-like Pde6c^w59 mutation and reported suppressed Ca²⁺fluctuations when compared to wt (26). At present, it is unclear where these differences between mutant cones in zebrafish and mouse arise, but they could be related to developmental stage, experimental conditions, and perhaps species differences. Ca²⁺fluctuations may have either beneficial or detrimental consequences. In rd1 cones, on the other hand, Ca²⁺noise was significantly lower than in wt, most likely because of the lower Ca²⁺baseline levels. An exciting topic for future studies will be to understand if and how increased Ca²⁺noise or overall changes in Ca²⁺signalling contribute to these degenerative events.

Concluding remarks and future aspects

Several investigations were based on the assumption that altered Ca²⁺signalling is involved in or serves as a trigger for rod and cone degeneration (e.g. (21,54,64)). Our results indeed show alterations in the dynamics of Ca²⁺signalling in both primary and secondary cone degeneration. For primary cone degeneration, our data is in line with the “high Ca²⁺hypothesis”, suggesting that cell death may be triggered by Ca²⁺overload, possibly via higher random fluctuations in Ca²⁺levels. This may have a bearing on the stochastics of cone degeneration (as discussed above) but still leaves open the question whether changes in Ca²⁺and/or Ca²⁺-dependent enzymes are cause or consequence of primary cone degeneration. Secondary cone degeneration, on the other hand, is very likely associated with low Ca²⁺levels and is, thus, more consistent with the “low Ca²⁺hypothesis”. This may also be true for primary photoreceptor degeneration triggered by mutations that cause low Ca²⁺(e.g. CNG channel mutations). Taken together, these findings may have important ramifications for therapeutic approaches focusing on blocking Ca²⁺influx in hereditary retinal degeneration. Here, our study may help to reconcile both the high and low Ca²⁺hypotheses (16,27) and contribute to resolving a long-lasting controversy on the use of Ca²⁺channel antagonists for the treatment of inherited retinal degeneration (21–23).

To gain insight into the precise timing and the role of Ca²⁺signalling in the execution of cell death, Ca²⁺imaging experiments will need to be combined with cellular activity assays for Ca²⁺targets such as kinases, proteases, or histone deacetylases (65,66). On a different note, our study also showed that even genetically impaired cones may show considerable functional activity, highlighting the importance of Ca²⁺measurements at single-cell resolution to obtain direct readouts on the functionality and health of cones.

Supplementary Material

Supplementary Material is available at HMG online.

Acknowledgements

We thank Gordon Eske, Norman Rieger, and Klaudija Masarini for technical assistance; Simone De Giovanni, Christian Behrens, and Philipp Berens for helpful discussions; Ayse Sahaboglu-Tekgoez, and Robin Kemmler for technical advice and discussions.

Conflict of Interest statement. None declared.

Funding

This work was supported by the Deutsche Forschungsgemeinschaft (DFG) (Werner Reichardt Centre for Integrative Neuroscience Tübingen, EXC 307 to TE; EU 42/8-1 to TE; KFO 134 and PA 1751/7-1 to FPD; TR 1238/4-1 to DT), fortüne to TE, Neuro-Ophthalmologische Gesellschaft to TE, and the European Union (DRUGSFORD; HEALTH-F2-2012-304963 to FPD).