Recombinant protein delivery enables modulation of the phototransduction cascade in mouse retina

Abstract

Inherited retinal dystrophies are often associated with mutations in the genes involved in the phototransduction cascade in photoreceptors, a paradigmatic signaling pathway mediated by G protein-coupled receptors. Photoreceptor viability is strictly dependent on the levels of the second messengers cGMP and Ca²⁺. Here we explored the possibility of modulating the phototransduction cascade in mouse rods using direct or liposome-mediated administration of a recombinant protein crucial for regulating the interplay of the second messengers in photoreceptor outer segments. The effects of administration of the free and liposome-encapsulated human guanylate cyclase-activating protein 1 (GCAP1) were compared in biological systems of increasing complexity (in cyto, ex vivo, and in vivo). The analysis of protein biodistribution and the direct measurement of functional alteration in rod photoresponses show that the exogenous GCAP1 protein is fully incorporated into the mouse retina and photoreceptor outer segments. Furthermore, only in the presence of a point mutation associated with cone-rod dystrophy in humans p.(E111V), protein delivery induces a disease-like electrophysiological phenotype, consistent with constitutive activation of the retinal guanylate cyclase. Our study demonstrates that both direct and liposome-mediated protein delivery are powerful complementary tools for targeting signaling cascades in neuronal cells, which could be particularly important for the treatment of autosomal dominant genetic diseases.

Introduction

The molecular processes underlying vision are triggered by the absorption of photons by opsins in retinal photoreceptors. Located in specific membranous compartments in the outer segments of rods and cones, opsins are G protein-coupled receptors (GPCRs) that activate the signaling cascade known as phototransduction. For many years, phototransduction has been considered paradigmatic for the largest class of GPCR-mediated signaling pathways (rhodopsin-like or class-A GPCRs), and the accumulated knowledge about the structural, biochemical, and physiological details of this cascade has enabled significant advances in drug design and pharmacological approaches for many other signaling pathways [1, 2].

The phototransduction cascade converts the light signal detected by the opsins into a chemical signal, culminating in the transient fall of vesicular glutamate release from the photoreceptor synaptic terminal, which is sensed by downstream neurons [3]. Rods and cones adapt to dramatic changes in ambient light by modifying the kinetics of phototransduction, a finely regulated process orchestrated by the second messengers Ca²⁺ and cyclic guanosine monophosphate (cGMP). Absorption of light by rhodopsin (or cone opsins) triggers the hydrolysis of cGMP by activating the phosphodiesterase 6, thereby causing the dissociation of cGMP from cyclic nucleotide-gated channels (CNG) and their closure. The ensuing decrease in the inflow of Na⁺ and Ca²⁺ hyperpolarizes the cell which, in turn, causes a reduction in neurotransmitter release. In parallel to these events, the light-independent extrusion of Ca²⁺ from the Na⁺/Ca²⁺, K⁺-exchanger leads to a drop of Ca²⁺ concentration in the outer segments (from ~ 600 nM in the dark to below 100 nM in bright light [4]).

These light-evoked alterations of second messenger levels in the photoreceptor outer segment trigger feedback mechanisms necessary for the timely shutoff of the cascade, as well as for the adaptation to specific light or dark conditions [3, 5]. Subtle changes in Ca²⁺ concentration are promptly detected by guanylate cyclase-activating proteins (GCAPs), members of the neuronal calcium sensors family [6]. Two isoforms (GCAP1 and GCAP2) are expressed in rods and cones, but in human only GCAP1 seems to be actively involved in the phototransduction cascade as a modulator of retinal guanylate cyclase (GC) activity, with the most prominent contribution arising from the GC1 isozyme [7]. In human photoreceptors GCAP2 is probably involved in biochemical processes other than phototransduction [8], although its role in mouse phototransduction has been demonstrated [9].

GCAP1, the main regulator of GC1, is a 23 kDa protein belonging to the EF-hand superfamily [10] that ensures rapid detection of Ca²⁺ oscillations in the submicromolar range with a nanomolar affinity for Ca²⁺ [11]. When Ca²⁺ concentration drops because of phototransduction activation, Ca²⁺ is replaced by Mg²⁺, which can bind in the same metal binding loops of motifs EF2, EF3 and EF4 (Fig. 1a) [12,13,14]. This mechanism allows GCAP1 to switch between different signaling states, namely Ca²⁺-bound (GC1-inhibitor) and Mg²⁺-bound (GC1-activator), regulated by specific allosteric mechanisms involving the protein, the metal cations and the myristoyl group bound at its N-terminal [13, 15]. The conformation adopted by Mg²⁺-GCAP1 stimulates the synthesis of cGMP by GC1, thus permitting rapid restoration of dark-adapted cell conditions by reopening of the CNG channels [12, 16]. The Ca²⁺–Mg²⁺ exchange results in relatively minor conformational changes for GCAP1 [13, 15] (Fig. 1a), which are nevertheless sufficient to trigger the GC1 inhibitor-to-activator transition over the narrow physiological range of Ca²⁺ variation in the photoreceptor outer segment.

Fig. 1

Biochemical and biophysical characterization of GCAP1 variants. a Three-dimensional structure of E111V-GCAP1 in its Ca²⁺-loaded (left) and Mg²⁺-bound (right) state after 200 ns Molecular Dynamics (MD) simulations (trajectories are from Ref [24]). Protein structure is represented as cartoons with EF1, EF2, EF3 and EF4 colored in green, pink, blue and orange, respectively; the myristoyl moiety is shown as teal sticks, Ca²⁺ and Mg²⁺ ions are depicted as red and green spheres, respectively. b Detail of the Ca²⁺-binding loop of EF3 in WT-GCAP1 (left) and E111V-GCAP1 (right) after 200 ns MD simulations. Protein structure is shown as cartoons colored in green for WT-GCAP1 and blue for E111V-GCAP1; the sidechains of Ca²⁺-coordinating residues are labelled and are represented as sticks with O atoms in red, N atoms in blue, S atoms in yellow and C atoms in the same color as cartoons; the C atoms of E111 and V111 residues are colored in orange and magenta, respectively. Ca²⁺-ions are represented as pink spheres and labelled, zero-order bonds with Ca²⁺-coordinating residues are shown as dashed red lines; the seventh O atom required for Ca²⁺-coordination is provided by a water molecule (not shown). c 15% SDS-PAGE of ~ 30 µM WT-GCAP1 and E111V-GCAP1 in the presence of 1 mM EGTA + 1.1 mM Mg²⁺ and 1 mM Mg²⁺  + 1 mM Ca²⁺. d Representative far UV (upper panels) and near UV (lower panels) CD spectra of WT-GCAP1 (left panels) and E111V-GCAP1 (right panels) recorded at 25 °C in PBS pH 7.4. Protein concentration for far and near UV was 10 and 33 µM, respectively. Far UV CD spectra were registered in the presence of 300 µM EGTA (black) and after serial additions of 1 mM Mg²⁺ (blue) and 600 µM Ca²⁺ (red), thus resulting in 300 µM free Ca²⁺. Near UV CD spectra were registered in the presence of 500 µM EGTA and after serial additions of 1 mM Mg²⁺ and 1 mM Ca²⁺ (500 µM free). e Hydrodynamic diameter estimation by Dynamic Light Scattering of ~ 40 µM WT-GCAP1 (black) and E111V-GCAP1 (red) at 25 °C in the presence of 1 mM Ca²⁺; standard errors are shown in grey and orange, respectively. f GC1 enzymatic activity as a function of Ca²⁺ concentration (< 19 nM to 1 mM) upon regulation by 5 µM WT-GCAP1 (black) or E111V-GCAP1 (red); cGMP synthesis was half maximal (IC₅₀) at (323.3 ± 15.1) nM and (20.2 ± 7.6) µM with Hill coefficients of 2.16 and 0.99, respectively. The Ca²⁺-concentration range in photoreceptor cells is represented by the grey box. Data are presented as average ± standard deviation of 3 technical replicates. g GC1 activity as a function of WT-GCAP1 (black) and E111V-GCAP1 (red) concentration (0–20 µM range); cGMP synthesis was half maximal (EC₅₀) at (1.88 ± 0.15) µM and (1.55 ± 0.24) µM, respectively. Data are presented as average ± standard deviation of 3 technical replicates after normalization on the amount of GC1 present in cell pellets. h Fluorescence emission spectrum of 5 µM GCAP1^CF640R upon excitation at 639 nm

GCAP1 has been associated with autosomal dominant cone (COD) or cone-rod dystrophies (CORD) [17,18,19,20,21,22,23,24,25,26,27,28,29,30,31], a class of severe inherited retinal dystrophies (IRD) characterized by central vision loss, impaired color vision, and photophobia, due to photoreceptor degeneration [32]. Indeed, more than twenty point-mutations in the gene encoding for GCAP1 (GUCA1A) have been found to be linked to COD or CORD. Recently, some of us identified a missense mutation in GUCA1A responsible for a particularly severe form of CORD. At the protein level, the mutation substitutes a glutamate residue in position 111 with a valine [24]. E111 is the twelfth residue of the Ca²⁺-binding loop of the EF3 motif and it is directly responsible for Ca²⁺-coordination by providing two negatively charged oxygen atoms from the carboxyl group (Fig. 1b); this bidentate ligation is fundamental to ensure the correct pentagonal bipyramidal geometry required for coordination of Ca²⁺-ions by seven oxygen atoms. The hydrophobic sidechain of V111, on the other hand, leads to a structural distortion of the EF3 loop, which becomes unable to coordinate Ca²⁺-ions (Fig. 1b), thus resulting in an 80-fold lower apparent affinity for Ca²⁺ [24].

COD and CORD remain incurable diseases, and the nature of their transmission — often dominant — makes gene therapy-based approaches particularly challenging. Alternative approaches, for example based on protein delivery, are therefore needed to mitigate the effects of the mutations and re-establish the physiological functionality of the signaling cascade. In this work, we explored the possibility of using direct or liposome-mediated administration of recombinant human GCAP1 to modulate the phototransduction cascade in mouse rods. Our rationale in comparing these two approaches was that free protein could offer high intraocular mobility and acute action, while liposomes delayed but prolonged release. We initially used a model eukaryotic cell culture for protein delivery experiments, and then increased the complexity with in vivo and ex vivo experiments aimed at assessing the biodistribution of proteins in mouse retinas. Imaging experiments were complemented by functional ones, in which acute changes in flash responses were monitored while incubating the retinas ex vivo. The administration of the free and liposome-encapsulated protein was compared in each case. Our findings reveal that direct and liposome-mediated protein delivery are powerful complementary tools for targeting signaling cascades in retinal neurons and could be particularly important for the treatment of autosomal dominant genetic diseases.

Results

Perturbed Ca²⁺-sensing properties of E111V-GCAP1 lead to constitutive activation of GC1

The purity and functionality of recombinantly expressed GCAP1 variants were verified by a combination of biophysical and biochemical techniques, to exclude potential effects of protein delivery treatments due to impurities or structural/functional defects. Ca²⁺-sensor proteins, including GCAP1, are known to modify their electrophoretic mobility in SDS-PAGE experiments under denaturing conditions [11], depending on their Ca²⁺-loading state. Indeed, Ca²⁺-free proteins appear as single bands at their theoretical molecular weight, whereas their Ca²⁺-bound forms exhibit an electrophoretic shift to smaller apparent molecular weight proportional to their apparent Ca²⁺-affinity. Such peculiar behavior exhibited by many other Ca²⁺-sensor proteins can be attributed to the fact that Ca²⁺-ions significantly stabilize the structure of the protein and are retained even under denaturing conditions, resulting in a protein that is not fully unfolded, even in the presence of sodium dodecyl sulfate. This makes electrophoretic migration dependent on factors other than just the mass of the protein in the Ca²⁺-bound form, such as Ca²⁺-affinity, because the higher the affinity, the greater the probability of partial structure retention. We exploited this peculiar feature to assess both the purity of protein samples and the capability of wild type (WT)-GCAP1 and E111V-GCAP1 to function as Ca²⁺-sensors (Fig. 1c). In the absence of Ca²⁺, both purified GCAP1 variants showed a single band around their theoretical molecular weight (23 kDa), which shifted to ~ 17 kDa in the case of WT-GCAP1 and to ~ 20 kDa in the case of E111V-GCAP1 upon Ca²⁺-binding, thus implying a substantial reduction in Ca²⁺-affinity for the pathological variant, confirming previous results from some of us [24].

The structural response of GCAP1 variants to ion binding was monitored by circular dichroism (CD) spectroscopy, which allows monitoring changes in protein secondary and tertiary structure in solution at protein concentrations (~ 10 to 40 µM) that mimic physiological ones (~ 3 to 4 µM, Fig. 1d). Both variants exhibited a far ultraviolet (UV) (200–250 nm) CD spectrum compatible with an all α-helix protein, with minima at 208 and 222 nm (Fig. 1d, top panels) and negligible variations in shape and intensity upon ion binding. This behavior was partly in contrast to that previously shown [24], most likely attributable to the different buffer and temperature (PBS pH 7.4 and 25 °C vs 20 mM Tris, 150 KCl, 1 mM DTT, pH 7.5 and 37 °C). Concerning the tertiary structure (near UV CD spectrum, 250–320 nm), both variants displayed a significant rearrangement of aromatic residues upon Ca²⁺-binding in PBS pH 7.4 at 25 °C (Fig. 1d, bottom panels), indicative of a change in protein tertiary structure. Mg²⁺-binding instead resulted in a minor conformational change, which was more pronounced in the case of WT-GCAP1. These results were substantially in line with the spectra recorded by some of us at 37 °C, with minor differences attributable to the different buffer and temperature [24], which was found to affect also the hydrodynamic diameter of GCAP1 variants (Fig. 1e). As assessed by dynamic light scattering (DLS), both WT-GCAP1 and E111V-GCAP1 in PBS, pH 7.4 displayed a significantly larger hydrodynamic diameter ((8.68 ± 1.07) nm and (11.08 ± 0.07) nm, respectively), compared to their counterparts in 20 mM Tris, 150 mM KCl, 1 mM DTT, pH 7.5 buffer ((6.47 ± 0.03) nm and (6.08 ± 0.04) nm, respectively) [24], a difference ascribable to the absence of the reducing agent and of a centrifugation step in this study, which nevertheless did not trigger any aggregation process (results not shown). The slightly different experimental conditions of the present and the previous study [24] lead to the same conclusions that the E111V substitution significantly impairs the Ca²⁺-sensitivity of GCAP1 with minor structural repercussions.

The enzymatic activity of the GCAP1-GC1 complex and its Ca²⁺-dependence creates a tight interconnection between Ca²⁺ and cGMP levels, which is crucial for both light adaptation and photoreceptor viability. Regulation of the GC1 enzymatic activity by GCAP1 variants was assessed both in terms of Ca²⁺ sensitivity and dependence on the level of protein regulator by measuring, respectively, the Ca²⁺ concentration at which GC1 activation is half-maximal (IC₅₀) and the concentration of GCAP1 at which the synthesis of cGMP is half-maximal (EC₅₀). The activation profile of GC1 by WT-GCAP1 exhibited an IC₅₀ of (323.3 ± 15.1) nM (Fig. 1f), thus falling in the physiological intracellular Ca²⁺-range (< 100 to 600 nM) [5]. On the other hand, the pathological variant E111V significantly dysregulated the activity of GC1, with an IC₅₀ value ((20.2 ± 7.6) µM, p-value < 0.05) ~ 63-fold higher than that of the WT, indicative of constitutive cGMP synthesis under physiological Ca²⁺ levels. Nevertheless, both variants displayed comparable EC₅₀ values ((1.88 ± 0.14) µM for WT-GCAP1 and (1.55 ± 0.23) µM for E111V-GCAP1 (p-value > 0.1), Fig. 1g), suggesting a similar affinity for the target enzyme, in line with previous results from some of us [24].

Liposome-mediated GCAP1 delivery to HEK293 cells

To assess the potential of liposome (LP)-mediated delivery of GCAP1 in biological systems of increasing complexity (in cyto, ex vivo, and in vivo) and investigate its biodistribution by minimizing the contribution of tissue auto-fluorescence (see Methods), the far-red fluorescent dye CF640R was conjugated to the primary amines of solvent-exposed Lys residues of GCAP1 (namely, either of K8, K23, K24, K46, K87, K97, K142 or K162, Movie S1) to obtain the GCAP1^CF640R complex. SDS-PAGE confirmed the purity of protein samples and the success of the conjugation reaction (Fig. S1a). The effective removal of the unconjugated dye (Fig. S1b) and the number of CF640R molecules bound to each GCAP1 protein were then assessed by absorption spectroscopy (degree of labelling = 1.96, see Methods). Finally, the emission fluorescence spectrum of GCAP1^CF640R (Fig. 1h) upon excitation at 639 nm was recorded before imaging experiments to verify the compatibility of the conjugated dye with our optical setups. The unconjugated dye (CF640R), GCAP1^CF640R, WT-GCAP1 and E111V-GCAP1 were then encapsulated in LPs with a lipid composition corresponding to that of rod outer segment membranes (see Methods for details).

The suitability of LP as carriers of small molecules and proteins was assessed by evaluating the size and monodispersion of the liposome suspensions loaded with different molecules. Regardless of the type of encapsulated molecule, nanoparticle tracking analysis (NTA) measured a LP diameter between (149.1 ± 3.0) nm and (168.7 ± 0.7) nm (Fig. S2a–c, Table S1), with minor differences in both size and concentration (Table S1) up to 180 days (Fig. S2d–f), suggesting that the LP suspension is stable over time. Finally, effective encapsulation of fluorescent molecules was also visually confirmed by the display of point-like fluorescence emission when LPs were filled with unconjugated CF640R and immobilized in agarose gel (Fig. S2g).

The capability of LPs to deliver recombinant GCAP1^CF640R was assessed on two different HEK293 stable cell lines. The first was transfected with pIRES plasmid encoding for eGFP and GC1 under the same promoter, characterized by a cytosolic green fluorescence (from now on cGFP); the second was transfected with pcDNA3.1 encoding for the eGFP-GC1 fusion protein, thus showing membrane green fluorescence (mGFP).

To address the potential direct membrane uptake of free-CF640R due to its small size (~ 832 Da) we monitored the cGFP cell line for 6 h (Movie S2) and visualized the mGFP cell line for 6 h and 24 h after incubation with free-CF640R (Fig. 2a). The absence of intracellular red fluorescence in both cell lines after 6 h and 24 h suggested that free-CF640R is per se unable to penetrate cell membranes.

Fig. 2

Live imaging of HEK293 cells incubated with free-CF640R, free-GCAP1^CF640R and GCAP1-encapsulating liposomes. a representative images at a fixed z-plane of the mGFP cell line incubated with 100 µl of 140 µM free-CF640R after 6 h (top) and 24 h (bottom); b representative images of the mGFP cell line after 24 h incubation with 100 µl of 104 µM free-GCAP1^CF640R, top panels show eGFP (left) and CF640R (right) fluorescence, bottom panels show the merged fluorescence at z = 200.7 µm (left) and z = 212.4 µm (right). c Live-cell imaging at the same z-plane of the mGFP cell line after 4 h, 7 h, 24 h, and 48 h incubation with 100 µl of 4.3 nM LP-GCAP1^CF640R (containing the same number of GCAP1^CF640R molecules in the aqueous core as compared to the free protein solution case). After 24 h the cell medium was replaced with FluoroBrite DMEM to avoid interference from phenol red, which gave rise to the red background fluorescence present in all but bottom right panel

Indeed, a diffused red fluorescence signal was observed only in the extracellular milieu, in the same time frames (Fig. 2a). This signal is likely attributable to the presence of phenol red in the medium, as in real time imaging no medium replacement was performed.

Similar experiments performed with fluorescently labelled GCAP1 (GCAP1^CF640R), showed punctuated fluorescence spots which tended to accumulate on the surface of mGFP cells after 6 h (Movie S3; punctuated red fluorescence is attributable to the interaction of GCAP1^CF640R with other molecules in the cell medium or with cell membrane, at odds with the diffused phenol red fluorescence). The intracellular space was not reached by the labelled protein even after 24 h (Fig. 2b, top panels). Further analysis at specific z-plane values (bottom panels in Fig. 2b) clearly confirmed that the accumulation of fluorescence attributed to GCAP1^CF640R is limited to the cell membrane, as no signal was detected at the intracellular space. To confirm this finding, we repeated the same experiment with cGFP cells, which would allow the detection of overlapped green and red fluorescence signals in case of GCAP1^CF640R entering the intracellular milieu. Indeed, this was not observed in 6 h (Movie S4). To dampen the contribution of diffused red fluorescence from the extracellular milieu, we replaced the cell medium with FluoroBrite DMEM, which does not contain phenol. The same results were confirmed (Fig. S3), indicating that GCAP1^CF640R did not enter HEK293 cells in the observed timeframe.

As both cGFP and mGFP cell lines exhibited the same impermeability to free-CF640R and free-GCAP1^CF640R, the capability of liposomes to deliver GCAP1^CF640R was tested only on the mGFP cell line. Live-cell imaging showed that 4 h after incubation (Fig. 2c, Movie S5), a punctuated red fluorescence, compatible with that emitted by LP-GCAP1^CF640R, was accumulating on the cell surface. At t = 7 h, the same punctuated red fluorescence was distinctly detected in the cytosol, suggesting whole liposome internalization by the cells. Only after 24 h the fluorescence pattern started to change. A more diffused red signal initially appeared at around 24 h, indicative of the release of GCAP1^CF640R from LPs. The observation of red and green colocalized fluorescence was more apparent at t = 48 h, indicative of a more unhindered diffusion of GCAP1^CF640R in the cytosol (Fig. 2c and Fig. S4), although the persistence of the punctuated pattern suggests the residual presence of internalized liposomes. Finally, 48 h after incubation with LP-GCAP1^CF640R the cell medium was replaced with FluoroBrite DMEM to improve the signal-to-noise ratio; while a more diffused colocalization of red and green fluorescence attributed to the intracellular release of GCAP1^CF640R was more clearly observed (Fig. S4d), residual quantal fluorescence was present.

Retinal distribution in live tissue of GCAP1^CF640R and LP-GCAP1^CF640R following ex vivo incubation and intravitreal injection

Ex vivo incubations

To move to a higher level of biological complexity, we first assessed LP-mediated delivery of molecules by ex vivo incubation of isolated retinas. The rationale here was to eliminate variability related to in vivo transport across tissues, focusing solely on intraretinal mechanisms. Far red fluorescence was chosen because in preliminary tests we found that the extremely low tissue autofluorescence in this band greatly improved signal-to-noise ratio. Retina pairs (n = 3) were incubated in parallel with LP-CF640R and LP-empty suspensions for 2 h at 37 °C. After incubation, these live retinas were rinsed with fresh medium and immediately viewed with a widefield fluorescent microscope either as wholemounts or 250 µm thick slices, with the two treatment partners placed adjacent in the dish. In all cases the LP-CF640R-treated retina showed fluorescence much above the control one. Unexpectedly, fluorescence was unevenly distributed across the thickness of the retina, being much stronger in the inner layers (Fig. 3a). Nonetheless, even the outer retina showed a significant signal. Control slices had a flat autofluorescence profile, at the level of the chamber background. Interestingly, in the LP-CF640R-treated retinas sparse cell bodies in the ganglion cell (GCL) and inner nuclear layer (INL) could be clearly distinguished. In the outer retina cones were also occasionally stained (Fig. 3b), albeit much more rarely than the aforementioned neurons. These data suggest enrichment of LPs in the inner retina, although it remained unclear whether they were internalized as intact nanovesicles or their fluorescent cargo released in the cell.

Fig. 3

Biodistribution of free GCAP1 and liposome-encapsulated GCAP1 in living mouse retinas following ex vivo incubation and in vivo intravitreal injections. a Distribution of fluorescence in 250 µm thick live slices obtained from a pair of retinas incubated ex vivo with 20 µl of 3.9 nM LP-CF640R and 5.1 nM LP-empty, respectively: the same image acquisition and display parameters were used (blue chain links symbol). Plot shows the average fluorescence along the vertical axis of the same images. Scale bars 25 µm. b Fluorescent cones in a slice from a retina incubated with 20 µl of 3.9 nM LP-CF640R. Scale bar 10 µm. c Distribution of fluorescence after ex vivo incubation with 20 µl of 88.4 µM free-GCAP1^CF640R and 4.3 nM LP-GCAP1^CF640R (containing the same number of GCAP1^CF640R molecules in the aqueous core as compared to the free protein solution case). In the example images the focal plane was intentionally set to display zones rich with stained neuronal somata. Scale bar 25 µm. Plots show radial fluorescence profiles across retina pairs in each experiment; blue chain links: the same acquisition parameters were used in each retina pair. d Distribution of fluorescence after ex vivo incubation with 88.4 µM free-GCAP1^CF640R and PBS. The white point of the images was adjusted to enhance the outer retina: same acquisition and display parameters. e Low magnification examples of retinal slices from the eyes of a mouse, both intravitreally injected with identical aliquots of 4.3 nM LP-GCAP1^CF640R: the same acquisition and display parameters were used (blue chain links). Thick white arrows: zones of accumulation of fluorescence in the vitreous humor near the inner limiting membrane. Inner: GCL + IPL + INL; outer: ONL + IS + OS. Scale bar 250 µm. f Examples of the distribution of fluorescence in retina slices after intravitreal injection with 2 µl of 88.4 µM free-GCAP1^CF640R and 4.3 nM LP-GCAP1^CF640R (containing the same number of GCAP1^CF640R molecules in the aqueous core as compared to the free protein solution case). Scale bar 25 µm. In all images of this figure the focal plane lies deep in the slice. In all experiments retinal layer boundaries were identified as shown in Fig. S9

To determine whether LP encapsulation affects the tissue access of large molecules, we incubated retina pairs (n = 10) with free-GCAP1^CF640R protein solution or LP-GCAP1^CF640R suspension for 3.5 h at 37 °C, followed by slicing and imaging with identical parameters. The average fluorescence of the inner retina (GCL + IPL + INL) and photoreceptors (ONL + IS + OS) was measured to quantitatively assess tissue distribution. As we had observed qualitatively for LP-CF640R, fluorescence in the inner retina was 168% (SD 50%) of that in photoreceptors following incubation with LP-GCAP1^CF640R (p < 0.01, n = 10; paired Wilcoxon test) and 160% (SD 55%) of photoreceptors with free-GCAP1^CF640R (p < 0.001, n = 12) (Fig. 3c). Furthermore, treatment with LP-encapsulated protein led to significantly higher fluorescence compared to free protein in both the inner retina (150% of free protein, SD 52%; p-value < 0.05, n = 10) and photoreceptors (127% of free protein, SD 21%; p-value < 0.01, n = 10) (Fig. 3c). In the GCL and INL distinctly stained neuronal somata were observed after both types of incubations (Fig. 3c, images). It must be noted that, even in the photoreceptor layer, signal from retinas incubated with the free protein was well above the level of tissue autofluorescence. This was confirmed by comparing incubation with free protein and PBS for 3.5 h at 37 °C (n = 2; 152% and 162%) (Fig. 3d). Interestingly, the peculiar distribution of both free and LP-encapsulated GCAP1^CF640R was also observed in retinas incubated with free CF640R fluorophore (Fig. S5a). We ruled out a significant contribution of the unbound fluorophore to the distribution observed with free-GCAP1^CF640R, as its concentration in our samples was estimated to be < 1% of the labelled protein (see Methods). Furthermore, during the preparation of LP-GCAP1^CF640R (encapsulation and subsequent washing) any residual unbound fluorophore would have fallen to even lower levels. Thus, had such traces of unconjugated fluorophore played a significant role in our experiments, we would have obtained results opposite to those shown in Fig. 3c (i.e. a higher signal in free-GCAP1^CF640R-incubated retinas).

Intravitreal injections

We went on to examine retinal delivery in vivo by injecting intravitreally free-GCAP1^CF640R solution and LP-GCAP1^CF640R suspension in the two eyes. After 20–24 h in darkness the animals were sacrificed, their retinas isolated, sliced 250 µm thick and viewed as live tissue with a widefield fluorescence microscope. Both retinas were imaged in the same session using identical acquisition parameters. We found that most injections led to some degree of retinal fluorescence (LP-GCAP1^CF640R: n = 9 out of 12; free-GCAP1^CF640R: n = 6 out of 11). Notably, while fluorescence was similar in different slices from the same retina, it varied greatly from eye to eye despite our utmost care in performing reproducible injections. This intrinsic variability was confirmed in a subset of animals in which both eyes were injected with identical solutions (n = 3 pairs; Fig. 3e). Importantly, it hindered our ability to detect any significant differences in the delivery of LP-GCAP1^CF640R and free-GCAP1^CF640R to the retina (n = 9 pairs). It is worth noting that residues of vitreous humor still adhering to the inner limiting membrane were often strongly fluorescent (Fig. 3e thick white arrows). Similarly to ex vivo incubated retinas, average fluorescence was quantified in the inner retina and photoreceptors. In those retinas displaying fluorescence after intravitreal injection, the inner layers were significantly brighter than the photoreceptors, both in the case of LP-GCAP1^CF640R (120% of photoreceptors, SD 22%; p-value < 0.05, n = 8) and free-GCAP1^CF640R (121% of photoreceptors, SD 19%; p-value < 0.05, n = 6) (Fig. 3f). This mimicked what was observed in ex vivo incubations. However, following intravitreal injections individual cell bodies did not stand out (Fig. 3f): perhaps in these experiments there was sufficient time for uniform uptake by all neurons. Diffuse retinal fluorescence could be similarly observed after intravitreal injection of free CF640R fluorophore (Fig. S5b). In summary, we found that intravitreal injections are a viable, albeit rather inconsistent, means of delivery of free or encapsulated proteins to the retina.

Immunofluorescence reveals different timing for free and LP-encapsulated GCAP1 internalization

Despite some discrepancies in the observed biodistribution, partly due to its inherent variability, the ex vivo incubation and intravitreal injection experiments suggested that fluorescently labeled GCAP1 can be internalized by all retinal layers. To study biodistribution by a complementary methodology, we repeated the ex vivo incubation experiments using a variant of GCAP1 with a His-tag at the C-terminus (GCAP1^His), which allowed direct detection by immunofluorescence. Incubation was performed for both free GCAP1^His and LP-GCAP1^His using the same protocol used for CF640R-conjugated counterparts, at three different time intervals, namely, (i) 30 min; (ii) 4 h and 30 min; and (iii) 24 h (Fig. 4).

Fig. 4

Distribution of delivered recombinant GCAP1^His in mouse retinas following ex vivo incubation as detected by immunofluorescence. Representative central plane of Z-stack of retinal cryosections after a 24 h of incubation with 180 µl PBS; b 30 min, 4 h 30 min and 24 h incubation with 180 µl of 99.6 µM free-GCAP1^His. c 30 min, 4 h 30 min and 24 h incubation with 180 µl of 4.5 nM LP-GCAP1^His (containing the same number of GCAP1^His molecules in the aqueous core as compared to the free protein solution case). Sections were stained with an anti-His antibody (red) and DAPI (light blue). The same image acquisition and display parameters were used in all samples

Incubation of the untreated tissue with a primary antibody against the His-tag resulted in an almost nonexistent background signal (Fig. 4a), which allowed high-resolution detection of the internalized GCAP1^His protein by immunofluorescence.

A substantially different time-course of protein internalization was observed when comparing the free and LP-encapsulated protein. Delivered free-GCAP1^His (Fig. 4b) was observed in both inner and outer retina even after 30 min, with a detectable signal at the photoreceptor layer (yellow arrow). Fluorescence increased with incubation time, resulting in a clear accumulation of protein in the photoreceptor layer, with specific signals both in the IS and OS layers. Interestingly, 24 h after incubation the delivered protein was clearly visible in the somata of several ganglion cells (yellow arrow in Fig. 4b, right panel), but a much stronger signal was present in the photoreceptor layer even within individual cones (Fig. 5a, yellow arrows and insets). In the case of LP-GCAP1^His, the intra-retinal signal became detectable only after a few hours, and at t = 4 h 30 min a speckled signal was observed at the level of the OPL (yellow arrow in Fig. 4c), suggesting that LPs were only partially internalized and did not release their content. However, at t = 24 h, a strong signal was observed in both inner and outer retina and individual ganglion cell somata became visible, along with the photoreceptor layer region. At this time, an overlap of the signal with that of the individual cones was observed (Fig. 5b and inset), indicating that the liposomes released their contents within the individual photoreceptors.

Fig. 5

Cellular distribution of delivered recombinant GCAP1^His in mouse retinas following ex vivo incubation as detected by immunofluorescence. Representative central plane of Z-stack of cryosections of retinas after a 30 min and 24 h incubation with 180 µl of 99.6 µM free-GCAP1^His; or b 24 h incubation with 180 µl of 4.5 nM LP-GCAP1^His (containing the same number of GCAP1^His molecules in the aqueous core as compared to the free protein solution case). Sections stained with an anti-His antibody (red), DAPI (light blue) and Peanut Agglutinin (PNA, green). The same image acquisition and display parameters were used in all samples. Insets show individual cones with the red point of the images adjusted to enhance the signal for GCAP1^His. Note the signal overlap from the two channels, resulting in yellow pixels, indicating intracellular location.

In summary, our comparative experiments show a substantially different timing of retinal internalization for free and LP-encapsulated GCAP1, which reflects in a slightly different biodistribution of exogenous protein across the retinal layers, as shown by the alternative presence of more diffused and speckled signals (Fig. S6).

Delivery of E111V-GCAP1 induces a CORD-like phenotype in WT mouse retinas

If a protein such as GCAP1 was able to gain access to retinal neurons in sufficient concentration, it could potentially be used to modulate biochemical processes [33]. As a proof of concept, we tested the functional effects on photoreceptors of E111V-GCAP1, known to be associated with CORD [24]. Ex vivo ERG recordings were made in a novel purpose-built chamber (see Methods), which enabled prolonged incubation of retinas with relatively high concentrations of expensive test substances (i.e., using tiny overall amounts). ERG recordings were made at 35 °C (except when stated otherwise) under pharmacological blockade of synaptic transmission to ON-bipolars (40 µM AP4) [34]. Except in one case, we did not remove the slow glial component with BaCl₂ in order to avoid any direct or secondary effects on photoreceptor physiology, which could affect unpredictably liposome and/or protein uptake [35]. The above conditions were associated to stable recordings of scotopic flash responses for over 4 h. Two parameters were extracted: (i) light sensitivity measured by the flash intensity required to obtain a 50% response (i₅₀); (ii) time to peak of the 50% response (TTP@i₅₀) (Fig. S7). These were normalized to their pre-treatment levels and processed to remove any trends also present in the control retina, leaving us with (ideally) the net effect of treatment (see Methods).

We first compared the incubation with WT-GCAP1 to PBS (n = 14 animals; Fig. 6a). Over three hours of incubation no significant and systematic effects were detected either on sensitivity or kinetics (Fig. 6a). However, the incubation with free-E111V-GCAP1 slowed response kinetics when compared to PBS (n = 14; Fig. 6b), an effect highly significant already from the first minutes after delivery through the entire three hours of incubation (p < 0.01). We confirmed this surprising result by comparing the same concentration of free-E111V-GCAP1 and free-WT-GCAP1, again observing a highly significant slowing of kinetics at most time points (n = 8; Fig. 6c), which indicated that the effect is attributable solely to the E111V point mutation. It should be noted that, even 30 min after incubation, the exogenous protein was observed to be internalized by the retina and was detected in the photoreceptor layer (Fig. 4b). The phenotypic effect was therefore induced very rapidly.

Fig. 6

Functional effects on isolated retinas of incubation with free or LP-encapsulated recombinant GCAP1. The ex vivo ERGs of retina pairs were obtained in control conditions (Time < 0) and during parallel incubation with test and reference solutions for up to 3 h or 18 h. Changes in light sensitivity (i₅₀) and response kinetics (TTP@i₅₀) were monitored by normalizing for pre-treatment control and reference solution. a–i Red lines represent individual experiments, each involving both retinas from an animal. Red (blue) shaded areas cover the 95% (99%) confidence interval. Loci above unity indicate a decrease in sensitivity or a slowing in kinetics. Red stars: p-value < 0.05; double blue stars: p-value < 0.01

We then went on to examine the effect of LP encapsulation. According to immunofluorescence data (Fig. 4c), LPs were observed after 4 h 30 min incubation in the OPL and, to a lesser extent also in the photoreceptor layer, although the fluorescence pattern suggested that their cargo may not have been released during this time. When LPs-empty were compared to PBS no significant effects were detected (n = 13; Fig. 6d), suggesting that LPs by themselves do not perturb phototransduction. Interestingly, LP-E111V-GCAP1 compared to PBS did not replicate the effects seen with the free mutant protein (n = 9; Fig. 6e). Based on our experience with LPs holding fluorescent molecules we hypothesized that LPs might be scavenged by the filter paper supporting the retina in the chamber. To exclude this possibility, we modified our approach to hold the retinas in place during the recordings thereby dispensing with the filter paper. Furthermore, to promote LP fusion/internalization in cells we raised the incubation temperature to 37℃. Despite these efforts no significant effects were detected over the course of 3 h (n = 21; Fig. 6f). We also compared the incubation of free-E111V-GCAP1 with LP-E111V-GCAP1 (n = 7) and, given previous results, we were not surprised to find a significant slowing of kinetics throughout incubation (Fig. 6g). A weakly significant increase in light sensitivity, not seen in previous comparisons was observed. Furthermore, we confirmed that free-E111V-GCAP1 slow response kinetics also when 50 µM BaCl₂ is present in the bath solution (n = 5; Fig. 6h; Fig. S8), although these incubations were terminated after about 1 h. Taking into account the results from immunofluorescence, we reasoned that while LPs might be rapidly internalized, their contents could be released over much longer time scales, also in line with previous findings [36, 37]. To begin examining this hypothesis we exploited recent advances by some of us in long duration ex vivo ERG [38] and succeeded to prolong incubations of LP-encapsulated protein to 18 h. When comparing LP-E111V-GCAP1 to PBS, we detected a significant slowing of the response kinetics only in the final few time bins (n = 18; Fig. 6i), thus approaching 24 h after incubation. This result is fully in line with immunofluorescence experiments, which clearly show that efficient release of the protein following LP encapsulation takes longer than delivery of the free protein, particularly to reach adequate levels in the photoreceptor layer and specifically in the outer segment. While our recordings are the state of the art in terms of duration, only future technical developments will allow to monitor the effects of slow drug release by liposomes over several days.

Discussion

In recent years, delivery of proteins and peptides to the eye have emerged as promising avenues for the treatment of a variety of ocular diseases [39], although significant physiological and anatomical challenges remain [40], especially when the goal is to modify biochemical processes occurring in the outer retina. In particular, to assist the development of effective therapies, basic knowledge is needed on whether and how different proteins/peptides move across the ocular compartments [41]. Our investigation represents a proof of concept that protein delivery to the retina may indeed be an effective strategy to modify the phototransduction cascade, which could be relevant to the treatment of IRDs.

Our experiments were based on the delivery of recombinant GCAP1 variants, whose in vitro characterization showed that the CORD-associated E111V-GCAP1 mutant constitutively activates GC1 as a consequence of its impaired Ca²⁺ sensing, although without altering its affinity (EC₅₀) for the target. Furthermore, the Ca²⁺ sensitivity of the GC1-GCAP1 system reconstituted in vitro was perfectly in line with the intracellular Ca²⁺ changes that occur in photoreceptors during phototransduction activation, thus demonstrating the functionality of recombinant proteins.

The simplest system used in this work for testing the potential of protein delivery was eukaryotic cell lines. Experiments with two different lines of HEK293 cells expressing GC1 clearly showed that fluorescently labelled GCAP1 (GCAP1^CF640R) tended to accumulate near the membrane but did not cross it. In contrast, 4 h after incubation, the same LP-encapsulated protein (LP-GCAP1^CF640R) started to enter the cell and was clearly observed in the cytoplasm 24 and even 48 h later. Considering that HEK293 cells were impermeable to the unconjugated dye and that GCAP1^CF640R failed to cross cell membrane in a 24 h timeframe, this suggests that liposomes are indeed required to transport GCAP1 inside these cells.

Experiments performed with mouse retinas unveiled a completely different scenario. Both free and LP-encapsulated GCAP1^CF640R were found to enter retinal neurons in the short time span of our ex vivo incubations (4.5 h), although with different timing and efficacy, as highlighted by immunofluorescence experiments. This also occurred 20–24 h after intravitreal injection, although in this case the fluorescence was locally uniform in the tissue. This could be due to the relatively short time following injection and lower effective protein concentration in intravitreal injections, as immunofluorescence suggests that only prolonged incubation results in cell-specific distribution. The mechanisms underlying retinal biodistribution of endogenous proteins deserves dedicated attention in future studies, using ad hoc model systems, such as organotypic retina cultures that permit high-resolution monitoring of the biodistribution of delivered protein. Besides the efficient delivery to the outer retina, in both ex vivo incubation and intravitreal experiments we detected particularly efficient delivery to the inner retina, suggesting that ocular diseases affecting retinal ganglion cells may be particularly well suited to protein therapy approaches relying on the delivery of recombinant proteins, either with or without the use of LPs as vectors. On the other hand, our functional studies demonstrate that the extent of delivery of exogenous GCAP1 into the photoreceptor outer segments is sufficient to modify the phototransduction cascade.

Taken together, the contrasting data obtained when comparing cell cultures and mouse retinas suggest that cell membrane composition plays an important role in determining the fate of free extracellular GCAP1. The lipid composition of HEK293 membranes [42] significantly differs from that of photoreceptors, which is known to change during retinal development [43] as well as between cone- or rod-dominant retinas [44], and in pathological conditions [43, 45]. Moreover, photoreceptors possess a host of complex and only partially understood molecular mechanisms of communication with the extracellular environment, including a high rate of synaptic membrane turnover due to synaptic vesicles exo/endocytosis [46] or disk membrane and nutrient recycling [47]. Thus, the complexity of retinal lipid composition and metabolism could partly explain the differences observed in the two cell types. Indeed, liposomes with a lipid composition that mimics the photoreceptor membrane are apparently able to enter both HEK293 and retinal neurons, although the process takes at least 24 h in the former, while being significantly faster in the latter. Our results indicate that exogenous GCAP1 can, in the absence of lipid carrier, cross retinal cell membranes and quickly achieve a detectable concentration in both the inner and outer layers, at odds with HEK293 cells, where no protein internalization was observed even after 24 h. While we did not anticipate such behavior, comparably challenging feats by exogenously applied proteins are not unprecedented. Indeed, even without clarifying the inherent mechanisms, several studies performed both in murine and human systems have previously shown that nerve growth factors can potentially cross several barriers in the visual system upon topical (eyedrops) [48] or intravitreal administration [49], leading to tangible clinical outcomes. Nevertheless, the mechanism by which intracellular distribution of exogenous GCAP1 can be observed across different neuronal layers remains currently unknown. Perhaps the protein distribution among photoreceptors is somehow related to the recently discovered nanotube-like connections [50, 51] that allow the exchange of intracellular material [52] including whole proteins. As for the inner-to-outer retina protein exchange, it could be mediated by glial transcytosis operated by Müller cells [53]. These cells generate the inner limiting membrane between the vitreous humor and the retina and span the whole retina longitudinally. Uptake of proteins and liposomes by Müller cells following intravitreal injections could therefore explain the broad retina biodistribution observed in our in vivo experiments. These hypotheses, which have tremendous implications for protein targeted therapy of retinal diseases, need further investigation.

In a comprehensive set of electrophysiological experiments, we found that free human E111V-GCAP1 rapidly induces a significant slowing of the photoresponse. Crucially, the WT protein did not evoke this effect, thereby pointing to a key role of the E111V point mutation. On the broader level this finding provides strong independent confirmation that free GCAP1 is taken up by retinal neurons and reaches the OS, where the phototransduction machinery is located. On the specific level of phototransduction it is striking, considering the compensating effect played by GCAP2 in murine photoreceptors [9, 33, 54]. In stark contrast, when the mutant protein was delivered encapsulated in LPs no such effects on kinetics were observed in 3 h long incubations and recordings. However, electrophysiological recordings performed over a longer time scale of up to 18 h, suggest that liposomes could be initially internalized intact, while only over longer times scales release their cargo, in line with immunofluorescence in this study and previous results [36, 37]. In fact, the latter property can be considered a major benefit in terms of sustained release and drug pharmacokinetics, be it the case of encapsulated small molecules [55] or, as shown by our study, recombinant proteins.

In a previous study we incubated mouse retinas for 2 h at 37 °C with LPs containing either the protein recoverin (homologous to GCAP1) or an antibody against the same protein [56]. In that case, we observed a significant difference in saturating response kinetics between the two treatments. The lipid composition of those liposomes was somewhat different (phosphatidylcholine/cholesterol at various molar ratios) than that used in the present work. Also different was the electrophysiological recording technique (loose seal patch clamp from single rods), which did not require a pharmacological blockade of synaptic transmission nor involve the presence of a slow glial ERG response component. Aside from these relatively minor differences, a key factor could be the relative ability of the different proteins to perturb phototransduction. If, as postulated above, liposomes slowly release their cargo once inside the photoreceptors, a functional effect after 2–3 h may only be detectable when delivering a strongly impacting protein. Recoverin antibody could, in principle, possess such an effect, considering the crucial role of the recoverin-mediated Ca²⁺-feedback on rhodopsin kinase in accelerating the shutoff [57]. In contrast, in the present study the recombinant GCAP1 mutant had to outcompete the endogenous WT protein and the compensating effect of GCAP2, which would overall occur when a sufficiently high amount of exogenous GCAP1 has reached the photoreceptor outer segment.

An unexpected finding of this study is that intravitreal injections in mice show extreme trial-to-trial variability in the translocation of delivered molecules to the retina. It must be noted that while mouse eyeballs are approximately tenfold smaller than human eyeballs (3 mm vs 24 mm diameter), their respective vitreous chambers exhibit a 1000-fold difference in volume (4.4 µl vs 4.3 ml [58]). This, together with the high intraocular pressure after injection, renders the entire procedure much more difficult to reproduce in mice than in human, for which several approved eye therapies are administered via intravitreal injections, and could result in lower effective concentration of delivered protein. We intentionally injected an excess volume of 2 µl to ensure that, despite some inevitable backward reflux through the injection hole, a significant amount of test solution always remained in the eye (utmost care was taken in this respect). We would thus tend to attribute a significant part of the variability observed to complex flow dynamics or inhomogeneities in the vitreous. Whatever the mechanism, studies employing intravitreal injections in mice should carefully consider whether variability in their observed therapeutic effects may have the same origin. Clearly, demonstrating significant effects of a drug candidate (or conversely excluding any medically relevant effects) may require a high number of test subjects.

Our study shows that direct and liposome-mediated protein delivery, while acting over different time scales, are powerful complementary tools for targeting signaling cascades in neuronal cells and could be particularly important for the treatment of retinal diseases. While genome editing represents the most promising therapeutic approach for the treatment of IRDs [59], a number of issues remain to be addressed, such as the risk of integrating viral DNA into the host genome increasing the likelihood of oncogenesis in the case of widely used viral vectors. Moreover, persistent expression of the editing machinery could give rise to antiviral immune responses in the long term; finally, the viral vectors currently in use pose severe limitations for the delivery of large genes needed for most gene therapies [60]. On the one hand, our study shows that LPs may represent ideal nonviral vectors for even large gene delivery, or they could be used to deliver ribonucleoproteins by eliciting a low immunogenic response, thus representing a promising strategy for genome editing in the eye. On the other hand, we have clearly shown that administration of recombinant proteins that mimic endogenous ones can induce a specific phenotype in retinal neurons, and this could have therapeutic relevance, especially in cases of autosomal dominant transmission, where a pool of mutated protein is responsible for the disease-phenotype. In the specific case of COD-CORD associated with missense mutations in GCAP1, a possible mutation-independent approach to therapy could be the delivery of WT-GCAP1 to overcome the effect of the mutant protein. In the case of the E111V variant, this strategy relies on two robust findings. First, WT- and E111V-GCAP1 show very similar apparent affinity for GC1 (EC₅₀), therefore could stoichiometrically compete for the same GC1 target. Second, our previous findings [33] demonstrated in vitro that the prolonged administration of exogenous GCAP1 could attenuate the pathological phenotype by: (i) shifting the IC₅₀ towards physiological values, with an increase in cooperativity of cGMP synthesis; (ii) reshaping the photoresponses towards a wild-type like kinetics; (iii) re-establish a wild-type-like homeostasis of second messengers (Ca²⁺ and cGMP) in dark-adapted cells. The proof-of-concept study presented here has therefore high therapeutic potential.

The implications of our findings could extend to a broader scale. Indeed, the GPCR-mediated molecular machinery building up the phototransduction cascade is shared by other signal transduction processes, including chemotaxis, neurotransmission, cell communication, activation of olfaction and taste, and many others [61]. Understanding the mechanisms that influence this signaling cascade and achieving its controlled modulation is critical for drug discovery, since about one-third of all drugs on the market target members of class A GPCRs [2]. More specifically, considering that GCAP1 is the major regulator of GC1 in human photoreceptors and that an increasing number of point mutations in its gene are associated with autosomal dominant COD or CORD, the development of novel biological therapies targeting this protein may help to restore the dysregulation of second messenger homeostasis in IRDs, ultimately slowing or blocking cell death.

Materials and methods

Materials

Tris(hydroxymethyl)aminomethane (Tris), Guanidine-HCl, NaCl, KCl, CaCl₂, MgCl₂, DTT, EGTA, β-mercaptoethanol, NH₄HCO₃, 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid) (HEPES), Ames’ medium, ethanolamine, phosphatidylethanolamine, phosphatidylcholine, phosphatidylserine, cholesterol, acrylamide, Coomassie blue, cGMP, polyethyleneimine, sucrose, OCT, NH₄Cl, citric acid, Triton X-100, Tween 20, Bovine Serum Albumin, chloramphenicol, cOmplete EDTA-free Protease Inhibitor Cocktail, paraformaldehyde, ketamine, xylazine, atropine, hydrocortisone and BaCl₂ were purchased from Merck (Darmstadt, Germany).

DMEM, OptiMEM, penicillin, streptomycin, 2-(4-amidinophenyl)-1H-indole-6-carboxamidine (DAPI), Phosphate Saline Buffer (PBS), Fetal Bovine Serum (FBS), HBSS, glutamine, Normal Goat Serum, Normal Donkey Serum were purchased from ThermoFisher Scientific (Waltham, MA, USA).

Cloning, expression, and purification of GCAP1 variants

Human myristoylated WT-GCAP1 was expressed in E. coli BL21 (DE3) after co-transformation with pBB131 containing the cDNA of S. cerevisiae N-myristoyl transferase (yNMT) [62]. The cDNA for E111V variant was obtained by PCR using QuikChange II Site-Directed Mutagenesis kit (Agilent, Milan, Italy) as described in Ref [24], while the cDNA for His-tagged WT-GCAP1 (GCAP1^His) was purchased from Genscript. Both variants were expressed and purified following the same protocol as for the WT [33], briefly consisting of: (i) denaturation of inclusion bodies with 6 M Guanidine-HCl; (ii) refolding by dialysis against 20 mM Tris–HCl pH 7.5, 150 mM NaCl, 7.2 mM β-mercaptoethanol, and a combination of (iii) Size Exclusion Chromatography (SEC, HiPrep 26/60 Sephacryl S-200 HR, GE Healthcare, Chicago, IL, USA) and (iv) Anionic Exchange Chromatography (AEC, HiPrep Q HP 16/10, GE Healthcare, Chicago, IL, USA). The purity of GCAP1 variants was assessed by 15% acrylamide SDS-PAGE, samples were either exchanged against PBS, aliquoted and frozen with liquid nitrogen, or exchanged against NH₄HCO₃, aliquoted and lyophilized. Protein samples were finally stored at -80 °C.

The three-dimensional structure of human GCAP1 was obtained by homology modeling using the structure of Ca²⁺-loaded chicken GCAP1 [63] following the procedure illustrated in Ref [18]. In silico mutagenesis of E111V variant was obtained according to the protocol detailed in Ref [24]. The structures presented in Fig. 1a and b were extracted from the last frame of 200 ns Molecular Dynamics simulations from Ref [24], whose settings and protocols for energy minimization, equilibration and production phases were elucidated in Refs [13, 15].

Electrophoretic mobility shift assay

WT-GCAP1 and E111V-GCAP1 were dissolved in 20 mM Tris–HCl pH 7.5, 150 mM KCl, 1 mM DTT at a concentration of 30 µM, incubated for 5 min at 25 °C with either 1 mM EGTA + 1.1 mM Mg²⁺ or 1 mM Mg²⁺ + 1 mM Ca²⁺, boiled, and run for 50 min at 200 V on a 15% acrylamide gel under denaturing conditions. Finally, protein bands were visualized by Coomassie blue staining.

Circular dichroism (CD) spectroscopy

The effects of ion binding and of the E111V substitution on the secondary and tertiary structure of GCAP1 were evaluated by CD spectroscopy using a J-710 spectropolarimeter (Jasco, Cremella, Italy) thermostated by a Peltier-type cell holder. Lyophilized proteins were dissolved in PBS pH 7.4 buffer at a concentration of 35 and 10 µM for near UV and far UV spectra, respectively. Five accumulations of each spectrum were recorded at 25 °C in the absence of ions (500 µM EGTA for near UV, 300 µM for far UV) and after serial additions of 1 mM Mg²⁺ and Ca²⁺ (1 mM for near UV, 600 µM for far UV, leading to a free Ca²⁺ concentration of 500 and 300 µM, respectively). All spectra were subtracted with that of the buffer, near UV spectra were also zeroed by subtracting the average ellipticity between 310 and 320 nm, where no signal was expected.

Dynamic light scattering (DLS)

The hydrodynamic diameter of Ca²⁺-loaded WT-GCAP1 and E111V-GCAP1 was estimated by DLS using a Zetasizer Nano-S (Malvern Instruments, Malvern, UK). Proteins were dissolved in PBS pH 7.4 at 42 µM concentration and filtered with a Whatman Anotop 10 filter (20 nm cutoff, GE Healthcare, Chicago, IL, USA) before starting the measurements. Samples were equilibrated for 2 min at 25 °C and for each variant at least 100 measurements were collected, each consisting of 13 runs.

Guanylate cyclase activity assay

GC1 enzymatic activity as a function of Ca²⁺ and GCAP1 concentration was measured after reconstituting WT-GCAP1 and E111V-GCAP1 with cell membranes of mGFP-GC1 cells (see below) previously extracted by lysis (10 mM HEPES pH 7.4, Protease Inhibitor Cocktail 1×, 1 mM DTT buffer) and 20 min centrifugation at 18,000×g, as previously described [30, 64, 65]. Cell membranes were resuspended in 50 mM HEPES pH 7.4, 50 mM KCl, 20 mM NaCl, 1 mM DTT and incubated with 5 µM GCAP1 variants at increasing [Ca²⁺] (< 19 nM to 1 mM, controlled by Ca²⁺-EGTA buffer solutions [66]) to estimate the Ca²⁺ concentration at which cGMP synthesis by GC1 was half-maximal (IC₅₀). To estimate the GCAP1 concentration at which GC1 activation was half-maximal (EC₅₀), cell membranes were reconstituted with increasing amounts of each GCAP1 variant (0–20 µM) at low Ca²⁺ (< 19 nM). Reported IC₅₀ and EC₅₀ values are represented as average ± standard deviation of 3 technical replicates. The statistical significance of the differences in IC₅₀ and EC₅₀ between WT-GCAP1 and E111V-GCAP1 was evaluated by means of two-tailed t tests (p value = 0.05).

Conjugation of CF640R-N-hydroxysuccinimide (NHS) ester with WT-GCAP1

Far-red fluorescent dye CF640R (Biotium, Fremont, CA, USA) was conjugated via NHS to WT-GCAP1 primary amines (Lys residues, Movie S1) according to the manufacturer protocol. Briefly, GCAP1 was diluted in PBS pH 7.4 and 1 mM DTT to a final concentration of 76 µM in a final volume of 900 µl; then the solution was added with 100 µl sodium bicarbonate 1 M pH 8.3 and 2 CF640R-NHS aliquots previously resuspended in 50 µl total DMSO. The mixture was then wrapped in aluminum and incubated in rotation at RT for 1 h. Unconjugated dye was removed by washing 4 times the protein solution (see Fig. S1b for representative spectra of the 4 flowthrough) with PBS pH 7.4 for 10 min at 4400 × g and 4 °C using an Amicon Ultra-4 concentrator with 3 kDa cutoff (Merck Millipore, Burlington, MA, USA). The degree of labelling (DOL = 1.96) was calculated as the ratio between the concentration of dye in the protein solution measured based on the absorbance at 642 nm (ε = 105.000 cm⁻¹ M⁻¹), and the concentration of protein calculated by considering the dilution factor and the retention of Amicon concentrators (95%, according to manufacturer instructions). The concentration of free-CF640R in the protein solution was calculated by measuring the absorbance at 642 nm of wash 4, which was < 1% with respect to protein concentration in all conjugation experiments. Unconjugated dye was blocked with 50 µl ethanolamine 1 M.

Fluorescence spectroscopy

The emission fluorescence spectrum of 2 µM GCAP1^CF640R (645–680 nm) dissolved in PBS pH 7.4 was collected at 25 °C on a FP-750 spectrofluorometer (Jasco, Cremella, Italy) after excitation at 639 nm; the spectrum reported in Fig. 1h is an average of 3 accumulations after subtraction of the emission spectrum of the buffer in the same range.

Liposome preparation

LPs were prepared by hydrating a thin lipid film of the same composition as photoreceptors rod outer segment membranes [67] (phosphatidylethanolamine, phosphatidylcholine, phosphatidylserine, and cholesterol at a molar ratio of 40:40:15:5) previously mixed in chloroform and dried in a speed-vac concentrator. Four mg of lipid film were hydrated with 1 ml PBS pH 7.4, vortexed for 30 min at room temperature, sonicated for 15 min in a water bath on ice and extruded 20 times through a 200 nm polycarbonate filter (Whatman, Maidstone, UK). The encapsulation of CF640R, WT-GCAP1, E111V-GCAP1, GCAP1^His, or GCAP1^CF640R in LPs was achieved by dissolving the molecule to be loaded in PBS before lipid film hydration. Unencapsulated molecules were removed by washing at least 4 times the LPs suspensions with PBS pH 7.4 for 20 min at 4 °C and 5000×g using an Amicon Ultra-4 concentrator with 100 kDa cutoff (Merck Millipore, Burlington, MA, USA). The degree of encapsulation was calculated by subtracting from the total mass of the molecule to be encapsulated that present in the flow-through and was found to be higher than 75% in all LP preparations. The efficient separation of non-encapsulated proteins was assessed by measuring protein concentration of the flowthrough of the 4 washing steps, similarly to what was done for CF640R. The concentration of non-encapsulated protein in LP suspensions was estimated from the concentration of protein in the last washing step and was found to be < 7% of the encapsulated protein.

Nanoparticle tracking analysis (NTA)

The concentration and size of LP suspensions were measured at 25 °C by means of NTA on a NanoSight (Malvern Instruments, Malvern, UK) by recording 3 videos of 1 min each at 25 fps by setting 20 µl/min flow rate; camera level and detection threshold were automatically optimized for each measurement to maximize the signal-to-noise ratio. LP size reported in Fig. S2 and LP concentration reported in Table S1 represent the average ± standard error of 3 technical replicates.

Fluorescence imaging of gel-immobilized liposomes

Stock suspensions of LPs, either filled with free-CF640R or empty, were diluted 1:400 v/v in 0.5% low gelling temperature agarose in Ames’ medium at 37 °C. A thin film was polymerized over a pure agarose meniscus in a Petri dish and covered with Ames’ medium. 3D image stacks were acquired with a 63x/0.9NA water immersion objective and a CCD camera (DFC350 FX, Leica Microsystems, Milan, Italy) in an upright widefield fluorescence microscope (DM LFSA, Leica Microsystems, Milan, Italy) using a Cy5 filterset (49,006; Chroma, Olching, Germany). Stacks were deconvolved and max projected along the z-axis using Fiji/ImageJ as detailed in Ref [68].

Generation of cGFP-GC1 and mGFP-GC1 stable HEK293 cell lines

HEK293 cells were cultured in DMEM medium supplemented with fetal bovine serum (10%, v/v), penicillin (100 units/ml) and streptomycin (100 μg/ml) at 37 °C in humidified atmosphere with 5% CO₂. Cells (6.25 × 10⁵) were seeded in 6-well plates in DMEM medium and grown overnight; the next day cell medium was replaced with OptiMEM reduced serum medium and cells were transfected using polyethyleneimine (PEI) as transfection reagent and 2 different vectors to obtain eGFP-expressing stable cell lines: (i) pIRES encoding for eGFP and human GC1 under the same promoter, thus resulting in a cytosolic fluorescence (cGFP), and (ii) pcDNA3.1 + N-eGFP encoding for GC1-eGFP fusion protein for localizing fluorescence on the membrane (mGFP). DNA (2.5 µg) was mixed dropwise to 10 µl PEI solution at a concentration of 1 µg/µl (DNA:PEI ratio of 1:5 w/w), added dropwise to 500 µl of pre-warmed OptiMEM, mixed and incubated 30 min at room temperature to allow DNA-PEI polyplex formation. Polyplexes were finally added dropwise to each well and the plate was incubated overnight at 37 °C and 5% CO₂. The next day, OptiMEM medium was replaced with DMEM and 48 h after transfection eGFP positive cells were selected using geneticin (500 µg/ml).

Live-cell imaging

Cells (8 × 10⁴) were seeded in 4-well chambers (Ibidi, Graefelfing, Germany) in DMEM medium; two days later the medium was replaced with OptiMEM reduced serum medium, then cells were incubated with 100 µl LP suspension per well (containing each ~ 0.4 mg lipid) and monitored in live-cell imaging. Experiments with fluorescently labelled GCAP1^CF640R were performed taking care of incubating the cells with the same nominal concentration of protein encapsulated in the LP aqueous core.

Live-cell imaging was performed using TCS-SP5 Inverted Confocal Microscope (Leica Microsystems, Milan, Italy) equipped with temperature and CO₂ controller and motorized stage that provides precise and automated acquisition of multiple fields of view. Images were collected simultaneously on different points of the sample immediately after cell-LP incubation and at 30 min interval for 24 h or 48 h total acquisition time. Images were captured after 488 nm and 633 nm laser excitation with a 40× objective (1.2 NA oil immersion) and further analyzed by Imaris 9.8 software (Oxford Instruments, Abingdon-on-Thames, UK). The fluorescence intensity profiles of mGFP and LP-GCAP1^CF640R reported in Fig. S4 were collected along the line across the cell shown in the insets using ImageJ.

Fluorescence microscopy of mouse retinas following ex vivo incubation

All animal experiments made use of adult C57Bl/6 J mice of both sexes. These were reared at around 22 °C in small groups with the addition of environmental enrichment items, a 12 h day/12 h night cycle, ad libitum food and water. As in previous studies by our group, and in accordance with authorized protocols, dark adapted mice were deeply anesthetized with ketamine (80 mg/kg) + xylazine (5 mg/kg) and their retinas extracted through a corneal incision in room temperature Ames' medium under dim red light. This approach avoided even brief exposure of the tissue to anoxic conditions, which could affect protein and/or liposome uptake. Animals were then immediately sacrificed with an overdose of anesthetic. After removing the vitreous each retina was placed, freely floating, in a plastic well containing incubation solution (1–2 ml depending on the experiment), and the wells inserted in an airtight box with a water layer at the bottom and a 95%O₂/5%CO₂ atmosphere. Incubation solutions consisted in the test suspension/solution diluted in Ames’ medium, taking care of reaching virtually the same final concentration for each suspension. The box was left floating in a water bath at 37 °C. After the prescribed time the retinas were returned to room temperature Ames' medium, made to adhere to black filter paper (AABP02500; Merck, Burlington, MA, USA) with gentle suction and, optionally, sliced at 250 µm thickness with a manual tissue chopper. Image stacks were acquired as described for the imaging of gel-immobilized LPs, with 4x/0.1NA air, 20x/0.5NA and 40x/0.8NA water immersion objectives. Excitation was provided by an Hg lamp preheated to achieve stable output. Stacks were lightly deconvolved (Richardson-Lucy algorithm, 10 iterations) and a single image obtained by averaging along the z-axis a few adjacent slices of the stack, in all cases chosen well below the cut surface. The borders of retinal layers were identified by imaging the same tissue volume in the near IR (Fig. S9). Cones were identified based on their characteristic location and morphology, leveraging our experience with their intracellular staining. Identical acquisition parameters were used when comparing retinas treated with different incubation solutions.

Immunofluorescence experiments with mouse retinas following ex vivo incubation

Mice were anesthetized with isoflurane, euthanized via cervical dislocation and their retinas extracted through a corneal incision in room temperature DMEM medium supplemented with FBS (25%, v/v), HBSS (25% v/v), glutamine (1% v/v), penicillin (100 units/ml) and streptomycin (100 μg/ml). After 30 min incubation at 37 °C and 5% CO₂, tissues were incubated with 180 µl PBS, 180 µl of 100 µM free-GCAP1^His or 180 µl of 4.5 nM LP-GCAP1^His (containing the same number of GCAP1^His molecules in the aqueous core as compared to the free protein solution case) for 30 min, 4 h 30 min, and 24 h, and finally washed 3 times with PBS.

Retina sections were then fixed for 40 min in 10% formalin in PBS buffer, washed 3 times with PBS, incubated with 10%, 20% and 30% sucrose for 1 h each at RT, and kept overnight at 4 °C. The next day samples were incubated at RT for 1 h with OCT compound: 30% sucrose at a 1:1 ratio and processed for cryo-sectioning at – 14 °C.

Sections (14 µm thickness) were fixed for 5 min with paraformaldehyde, washed 3 times with PBS, incubated with 0.1% Triton X-100 in PBS for 1 h at RT, washed 3 times with PBS, incubated with ammonium chloride for 20 min, and washed 5 times with PBS.

Sections were incubated overnight at RT with mouse anti-His primary antibody (1:1000 dilution, SouthernBiotech, Birmingham, AL, USA) and PNA (1:250 dilution, Molecular Probes, Eugene, OR, USA) in blocking solution (5% Normal Goat Serum, 1% Bovine Serum Albumin, 0.3% Triton X-100 in PBS). The following day samples were washed 3 times with PBS and incubated with an Alexa Fluor 647-conjugated goat anti-mouse secondary antibody (1:1000 dilution, Invitrogen, Waltham, MA, USA). Cell nuclei were stained with a 1:1000 DAPI dilution in PBS; slides were coverslipped with Dako fluorescence mounting medium (Agilent, Milan, Italy). Sections were visualized using TCS-SP5 Inverted Confocal Microscope (Leica Microsystems, Milan, Italy), images were captured after 405 nm and 633 nm laser excitation with a 63× objective (1.2 NA oil immersion) and further analyzed by Imaris 9.8 software (Oxford Instruments, Abingdon-on-Thames, UK).

Intravitreal injections

Mice were first anesthetized with ketamine (80 mg/kg) + xylazine (5 mg/kg), followed by application of eyedrops containing atropine and chloramphenicol (1%) + hydrocortisone (0.5%). Intravitreal injections were made under a stereomicroscope and dim blue light as follows: (i) a hole was made in the cornea near the ora serrata with the tip of a 31G insulin needle; (ii) glass micropipettes with a broken tip, connected to a 25 µl syringe (Hamilton, Reno, NV, USA) via PE tubing filled with mineral oil (330,779; Merck, Burlington, MA, USA), were front loaded with 2 µl of solution; (iii) the micropipette was inserted in the hole and the entire volume slowly injected in the vitreous. Mice were returned to their cages and allowed to recover in a paper blanket. After 20–24 h we performed retinal dissection, slicing and imaging as described for ex vivo incubations.

Long duration ex vivo ERG recordings

ERG experiments were made in a custom designed incubation and recording chamber [38]. Retina pairs were isolated as described for ex vivo incubations, made to adhere to white filter paper (SMWP02500; Merck, Burlington, MA, USA) and placed at the bottom of two adjacent plastic wells, containing 2 ml/retina of 40 µM AP4 (0101; Tocris, Milan, Italy) in Ames' medium. Retinas were centered on a hole leading to the anode, while the cathode was in the chamber itself. In some experiments we dispensed with the filter paper and used instead small transparent cups to immobilize the retinas (Fig. 1b in Ref [38]). Both electrodes were silver chloride wires inserted in an agar bridge. The well assembly was placed on an aluminum platform covered with a layer of water, inside a sealed incubation chamber purged with 95% O₂/5% CO₂. The temperature of the platform was actively controlled with a custom apparatus [69]. Small diameter PTFE tubing, leading from inside the wells to syringes residing outside the chamber, allowed injection and mixing of test solutions (100 µl/retina) into the wells during the recordings with minimal perturbation. Immediately above the wells, attached to the lid of the chamber, a LED (505 nm; ND filters) delivered the same flash sequence every 15 or 30 min: (ph/µm²|no. of flash repetitions) 3.98|12, 8.27|10, 18.9|8, 50.5|6, 151|6, 510|4, 1660|3. Transretinal potentials were amplified by 5000, filtered in the band DC-100 Hz, digitized at 5 kHz and acquired with pClamp 9 (Molecular Devices, San Jose, CA, USA). Electrophysiological records were analyzed in Axograph X with automated custom scripts. i₅₀ was determined by fitting a Hill function to a plot of response amplitudes measured 90–130 ms after the flash (Fig. S7). This range minimized the contribution of the very slow glial response and gave parameter values close to those in BaCl₂ [38]. TTP@i₅₀ was estimated as the weighted average of the TTPs of the two flash responses straddling i₅₀ (10 Hz Gaussian filtered records). Two rounds of normalization were applied to these raw data, as follows. We assumed that the two retinas, being from the same animal, behaved identically except for (i) an initial stabilization phase due to slight variations in their isolation and manipulation, and (ii) a scaling factor in their steady state light sensitivity due to small differences in their orientation in the recording chamber. We first normalized the two sets of raw values over their respective pre-treatment levels. We then removed any trends common to both retinas by dividing the normalized values of the treated retina by those of the control. We were thus left with a single time series that shows the net effect of the tested drug (Fig. 6, red lines). Treated and control retina positions in the wells were alternated from animal to animal to cancel out any environmental biases. In a limited number of tests (Fig. 6h) BaCl₂ was injected in the wells using the syringe system after an initial stabilization period and stirred to obtain a final concentration of 50 µM. In a subset of the experiments, we included HyClone PenStrep (SV30010; Cytiva, Breisgau, Germany) in the incubation medium at 1% vol/vol, which enabled us to prolong our recordings up to 18 h.

Statistics

Statistical analyses were performed with the open-source software JASP 0.16 (jasp-stats.org; RRID:SCR_015823), Kaleidagraph 5 and Excel (Microsoft, Redmond, WA, USA). For ERG recordings visual representation of the population effect of a tested drug (Fig. 6) were given by the confidence interval of the Hodges-Lehmann estimator [38]. Statistical significance was estimated by the following parametric and non-parametric tests as mentioned in the text: two tailed t-test, paired Wilcoxon signed-rank, one sample Wilcoxon signed-rank.