Two retinal dystrophy-associated missense mutations in GUCA1A with distinct molecular properties result in a similar aberrant regulation of the retinal guanylate cyclase

Abstract

Two recently identified missense mutations (p. L84F and p. I107T) in GUCA1A, the gene coding for guanylate cyclase (GC)-activating protein 1 (GCAP1), lead to a phenotype ascribable to cone, cone-rod and macular dystrophies. Here, we present a thorough biochemical and biophysical characterization of the mutant proteins and their distinct molecular features. I107T-GCAP1 has nearly wild-type-like protein secondary and tertiary structures, and binds Ca²⁺ with a >10-fold lower affinity than the wild-type. On the contrary, L84F-GCAP1 displays altered tertiary structure in both GC-activating and inhibiting states, and a wild type-like apparent affinity for Ca²⁺. The latter mutant also shows a significantly high affinity for Mg²⁺, which might be important for stabilizing the GC-activating state and inducing a cooperative mechanism for the binding of Ca²⁺, so far not been observed in other GCAP1 variants. Moreover, the thermal stability of L84F-GCAP1 is particularly high in the Ca²⁺-bound, GC-inhibiting state. Molecular dynamics simulations suggest that such enhanced stability arises from a deeper burial of the myristoyl moiety within the EF1–EF2 domain. The simulations also support an allosteric mechanism connecting the myristoyl moiety to the highest-affinity Ca²⁺ binding site EF3. In spite of their remarkably distinct molecular features, both mutants cause constitutive activation of the target GC at physiological Ca²⁺. We conclude that the similar aberrant regulation of the target enzyme results from a similar perturbation of the GCAP1–GC interaction, which may eventually cause dysregulation of both Ca²⁺ and cyclic GMP homeostasis and result in retinal degeneration.

Introduction

Autosomal-dominant cone (COD) and cone-rod (CORD) dystrophies constitute particularly severe forms of inherited visual diseases. The usually rapid clinical course leads to an initial decrease in visual acuity, abnormal color vision, photophobia and decreased sensitivity in the central visual field, later followed by progressive loss in peripheral vision and night blindness (1,2).

Most inherited retinal dystrophies are very often caused by an imbalance of the homeostasis of the second messengers operating in rod and cone photoreceptor cells, namely Ca²⁺ and cyclic GMP (cGMP). These important molecules control the kinetics of photoreceptor light response and play a crucial role in light adaptation (3,4). Indeed, photoactivation of the visual pigment in photoreceptors leads to the light-induced hydrolysis of cGMP, thereby causing the closure of cyclic nucleotide-gated channels; these block the influx of cations including Ca²⁺, and cause the hyperpolarization of the cell membrane. A second consequence of illumination is a decrease in intracellular Ca²⁺ due to the continuous extrusion of the cation via a Na⁺/Ca²⁺, K⁺-exchanger.

Among other Ca²⁺-sensor proteins, guanylate cyclase-activating proteins (GCAPs) detect changes in Ca²⁺ concentration and respond by adopting conformations that control the activity of their target protein, the membrane-bound guanylate cyclases (GC). At high [Ca²⁺], GCAPs inhibit the GC activity, thus ensuring basal levels of cGMP; at low [Ca²⁺], after light activation of the phototransduction cascade, they switch to a GC-activating state, which triggers the synthesis of cGMP and rapidly restores dark-adapted conditions (5,6). As a result of the delicate interplay between Ca²⁺ and cGMP homeostasis, the concentration of Ca²⁺ during the phototransduction cascade is strictly controlled, and ranges between several hundred nanomolar Ca²⁺ in the dark down to below 100 nm in the light (3,4).

GCAPs are sensor proteins able to bind up to three divalent metal ions in their EF-hand motifs, namely EF2, EF3 and EF4. They have been proven to be both Ca²⁺ and Mg²⁺ sensors, and the Mg²⁺-bound form is implied in GC activation, while the Ca²⁺-saturated form is necessary to fully inhibit the GC (5–9). The presence of various isoforms of GCAPs in photoreceptor cell types seems to be justified by the fact that GCAPs can respond to incremental changes in [Ca²⁺] and work in a relay mode fashion (10–14).

To date, up to 12 missense mutations have been identified in GUCA1A, the gene coding GCAP1, in patients suffering from retinal disease, including COD, CORD and macular dystrophy (MACD) (15,16). Biochemical investigations revealed that most of such GCAP1 mutations exhibit a more or less severe disturbance in their Ca²⁺-sensing properties, specifically, they maintain the GC constitutively active over the physiological range of [Ca²⁺], in which they normally switch between activating and inhibiting regulatory modes (17–21). The final trigger of photoreceptor degeneration remains still unclear, although experiments performed on transgenic mice suggest apoptosis as a possible mechanism (22,23).

Among retinal dystrophy-related mutations in GUCA1A, two (c.250C>T and c.320T>C) were very recently discovered in unrelated Spanish families affected by autosomal-dominant retinal degeneration (16). The clinical phenotype was typically broad, showing features of COD, CORD and MACD (16). The missense mutations lead to amino acid substitutions that significantly modify the physico-chemical nature of the side chain, with possible consequent perturbation of the protein structure/function relationship. Indeed, the c.250C>T mutation leads to the Leu 84 to Phe substitution (L84F-GCAP1), while c.320T>C leads to Ile 107 to Thr substitution (I107T-GCAP1), in the expressed protein. The affected amino acid positions are in close vicinity at the level of the primary structure, but are located in significantly remote regions of the GCAP1 tertiary structure (Fig. 1A). L84 is located in the EF1–EF2 domain, at the end of the exiting helix (αF2) of the EF2 motif. The region includes several important hydrophobic residues that interact with the myristoyl moiety post-translationally bound at the N-terminal (Fig. 1B), and the neighbor residue K85 has been shown to be significantly involved in the binding interface with the GC (24,25). On the other hand, I107 is located in the EF3 Ca²⁺ binding loop, although the residue is not involved in direct Ca²⁺-coordination. The EF3–EF4 domain has been defined a ‘hot spot’ for retinal dystrophies (15), in that many of the disease-associated mutations in GUCA1A codify for residues located in this structural region (Fig. 1A, blue sticks).

Figure 1.

Three-dimensional homology model of human GCAP1. (A) Cartoon representation of the GCAP1 overall fold. EF1 is colored in yellow, EF2 in green, EF3 in orange and EF4 in light blue. N- and C-terminals are colored in pale grey and pale cyan, respectively, and the myristoyl group is represented by grey spheres. Calcium ions are represented by red spheres, while residue targets of cone dystrophy-associated mutations are shown by dark blue sticks. L84 and I107 are represented by magenta sticks. (B) Details of the region surrounding the myristoyl group. Hydrophobic residues showing persistent interactions with the myristoyl moiety are represented as sticks and labeled.

The particularly broad phenotypical variety of these two newly identified GUCA1A mutants includes MACD, which has so far only been associated with the Y99C (26,27) and P50L (28) substitutions. We therefore asked, whether L84F and I107T-GCAP1 share common molecular features with other retinal dystrophy variants of GCAP1, or rather exhibit novel properties. We performed a thorough biochemical and biophysical investigation of the two mutants and found that, surprisingly, very different molecular properties shown by the two mutants may result in a similarly aberrant regulation of the target enzyme.

Results

Ca²⁺- and Mg²⁺-dependent structural changes and protein thermal stability in I107T and L84F-GCAP1 monitored by circular dichroism spectroscopy

Circular dichroism (CD) spectroscopy was used to assess the potential effects of the point mutations on both the secondary and tertiary structure of I107T and L84F-GCAP1, in their GC-activating and inhibiting states, corresponding to the Mg²⁺-bound and Ca²⁺-bound conformations, respectively.

The apo form (EGTA) of I107T-GCAP1 (Fig. 2, upper left) showed a near-UV CD spectrum similar to that of the wild-type (7), though with a more pronounced negative dichroic band in the Tyr-Trp region. The addition of Mg²⁺ lowered the intensity of the spectrum and led to regions of positive dichroism (Fig. 2, upper left). Following saturation with Ca²⁺, the spectrum yielded a typical shape already observed for wild-type GCAP1 (7) and for other COD-related mutants (29), with a mutant-specific variation in the amplitude of the positive Tyr band.

Figure 2.

Structural changes occurring in I107T and L84F GCAP1 upon binding of Ca²⁺ and Mg²⁺ monitored by CD spectroscopy. Near UV spectra were recorded at T = 37°C in 5 mm Tris–HCl pH 7.5, 150 mm KCl, 2 mm DTT buffer. Upper panel, left: 25.3 µm I107T-GCAP1 in the presence of 400 µm EGTA (black, solid line), 1 mm Mg²⁺ (gray, solid line) and after addition of 400 µm free Ca²⁺ (gray, dashed line); right: in the presence of 400 µm EGTA (black, solid line), 400 µm free Ca²⁺ (gray, solid line) and after addition of 1 mm Mg²⁺ (gray, dashed line). Lower panel, left: 26 µm L84F-GCAP1 in the presence of 400 µm EGTA (black, solid line), 1 mm Mg²⁺ (gray, solid line) and after addition of 400 µm free Ca²⁺ (gray, dashed line); right: in the presence of 400 µm EGTA (black, solid line), 400 µm free Ca²⁺ (gray, solid line) and after addition of 1 mm Mg²⁺ (gray, dashed line).

The binding of Ca²⁺ to apo I107T-GCAP1 caused significant variations in its tertiary structure, while the following addition of Mg²⁺ did not significantly perturb the near UV CD spectra (Fig. 2, upper right), substantially in line with what was observed for wild-type GCAP1 (7). The far UV CD spectra of I107T-GCAP1 (see Supplementary Material,Supplementary Data, upper panels) showed the typical all α-helix pattern, with a dampening of the signal of the Mg²⁺-bound form compared with the apo form, but no other alteration of the spectral shape as confirmed by the identical ratio between the ellipticity values at 208 and 222 nm (Θ₂₂₂/Θ₂₀₈ = 0.89; Table 1). A major increase in the signal after addition of Ca²⁺ (see Supplementary Material,Supplementary Data, upper left) and an increase in Θ₂₂₂/Θ₂₀₈ (0.91) was observed, indicative of changes in secondary structure and/or protein compactness (17). Ca²⁺-bound I107T showed an increase in both the signal intensity (see Supplementary Material,Supplementary Data, upper right) and in the Θ₂₂₂/Θ₂₀₈ ratio (0.91; Table 1) compared with the apo form, with no significant differences upon Mg²⁺ addition.

Table 1.

Results from far UV CD spectroscopy, thermal denaturation and fluorescence spectroscopy

	Θ222/Θ208a	Tmb (°C)	λmax fluoc (nm)
I107T
EGTA	0.89	42.0–91.5	341
Mg2+	0.89	43.4–91.8	341
Mg2+ Ca2+d	0.91	79.3	344
Ca2+	0.91	80.2	—
Ca2+ Mg2+d	0.91	—	—
L84F
EGTA	0.90	38.2 to >96	341
Mg2+	0.94	42.2 to >96	341
Mg2+ Ca2+	0.94	91.9	342
Ca2+	0.92	92.6	—
Ca2+ Mg2+	0.92	—	—

^aRatio between ellipticity signal at λ = 222 and λ = 208 nm.

^bMelting temperature obtained by fitting the experimental curve to a four-parameter Hill sigmoid.

^cWavelength of maximal fluorescence emission upon excitation at λ = 290 nm.

^dLines labeled ‘Mg²⁺ Ca²⁺’ refer to experiments where Ca²⁺ was added after saturating Mg²⁺, whereas cations were added in the opposite order in the ‘Ca²⁺ Mg²⁺’ cases.

Interestingly, a significantly different tertiary structure was observed for L84F-GCAP1 already in the apo form (Fig. 2, lower left). Peculiar features were the positive dichroism in all the spectral regions and the onset of a maximum in the Phe band, at 255 nm, which was previously observed only in ion-bound wild-type GCAP1 (7). Mg²⁺-binding caused a small but significant conformational change in the Tyr band, and further addition of Ca²⁺ changed the shape of the spectrum, particularly in the Phe and Tyr bands (Fig. 2, lower left). The same kind of changes were observed in Ca²⁺-bound L84F-GCAP1 (Fig. 2, lower right) compared with the apo form, and the addition of Mg²⁺ after Ca²⁺ did not significantly alter the aromatic amino acid microenvironment. As for the far UV spectra, Mg²⁺-binding slightly lowered the signal at 208 nm, as shown by the increase in Θ₂₂₂/Θ₂₀₈ from 0.90 to 0.94 (see Supplementary Material,Supplementary Data, lower left; Table 1), while addition of Ca²⁺ increased the signal without altering the shape of the spectrum (Table 1). Ca²⁺-binding to the apo protein caused instead an increase in both the intensity and the Θ₂₂₂/Θ₂₀₈ ratio to 0.92 (Table 1), while the addition of Mg²⁺ resulted only in a small dampening of the signal (see Supplementary Material,Supplementary Data, lower right).

By taking advantage of the typical all-α architecture of both GCAP1 variants, demonstrated by the far UV CD spectra in all tested conditions, their thermal stability could be conveniently monitored by following the ellipticity signal at 208 nm. Differently from what observed for other COD/CORD-associated variants (17), apo I107T-GCAP1 showed a three-phase thermal denaturation profile (Fig. 3A) with two apparent transitions at 42.0 and 91.5°C (Table 1). The binding of Mg²⁺ had a very mild effect on the two transition temperatures (43.4 and 91.8°C, respectively) while keeping the same unfolding pattern as the apo form. This behavior also differs from wild-type GCAP1, which showed a typical two-phase transition pattern both in the apo and in the Mg²⁺-bound forms, with melting temperatures of 50.1 and 51.5°C, respectively (7). Surprisingly, I107T-GCAP1 showed a higher thermal stability in the presence of Ca²⁺ compared with the wild-type (7) both in the absence (80.2 versus >70°C) and in the presence (79.3 versus >70°C) of Mg²⁺, which slightly lowered the T_m (79.3 versus 80.2°C).

Figure 3.

Thermal denaturation profiles of I107T and L84F GCAP1 mutants in the presence and in the absence of Ca²⁺ and Mg²⁺. (A) Thermal denaturation profile of ∼6 µm I107T-GCAP1 in the presence of 200 µm EGTA (black triangles), 1 mm Mg²⁺ (gray diamonds), 200 µm Ca²⁺ and 1 mm Mg²⁺ (black squares) and 200 µm Ca²⁺ (gray circles). (B) Thermal denaturation profile of ∼6 µm L84F-GCAP1 in the presence of 200 µm EGTA (black triangles), 1 mm Mg²⁺ (gray diamonds), 200 µm Ca²⁺ and 1 mm Mg²⁺ (black squares) and 200 µm Ca²⁺ (gray circles). Thermal denaturation was followed by monitoring the ellipticity signal at 208 nm over the 20–96°C temperature range. Quantitative data obtained by fitting the experimental values to a four-parameter Hill sigmoidal function are reported in Table 1.

Thermal denaturation profiles for L84F-GCAP1 also showed the same three-phase transitions for apo and Mg²⁺-bound forms (Fig. 3B), with a clear transition at 38.2 and 42.2°C (Table 1), respectively, and an only partly visible transition at >96°C. Ca²⁺ binding greatly increased the stability of L84F-GCAP1 both in the absence (92.6°C) and in the presence of Mg²⁺ (91.9°C).

Intrinsic fluorescence properties of I107T and L84F-GCAP1

The fluorescence properties of the two retinal dystrophy-related GCAP1 mutants were compared with those of the wild-type in conditions mimicking dark-adapted (high [Ca²⁺]) and illuminated (low [Ca²⁺]) photoreceptors. Intrinsic fluorescence spectra of wild-type GCAP1 (Fig. 4A) showed a major decrease (∼45%) of the intensity upon Mg²⁺ binding to the apo form, together with a small red shift in the wavelength of maximal intensity (342 versus 341 nm). Significant increases in fluorescence and red shift (344 nm) were observed upon saturation with Ca²⁺. Apo I107-GCAP1 also showed (Fig. 4B) quenching of the intrinsic fluorescence upon Mg²⁺ binding, but to a lesser extent (∼10%), without any shift in the maximum (341 nm, Table 1). Major quenching was instead observed upon Ca²⁺ addition (∼50% with respect to the apo form), which shifted the maximum of fluorescence toward higher wavelength (344 nm).

Figure 4.

Fluorescence spectra of GCAP1 variants. Fluorescence emission spectra (λ_ex = 290 nm) were recorded at T = 25°C in 5 mm Tris–HCl pH 7.5, 150 mm KCl, 1 mm DTT buffer. (A) Intrinsic Trp fluorescence emission spectrum of ∼8 µm wild-type GCAP1 in the presence of 1 mm EGTA (black, solid line), after addition of 1 mm Mg²⁺ (gray, solid line), after further addition of 1 mm free Ca²⁺ (gray, dashed line). (B) Intrinsic Trp fluorescence emission spectrum of ∼8 µm I107T-GCAP1 in the presence of 1 mm EGTA (black, solid line), after addition of 1 mm Mg²⁺ (gray, solid line), after further addition of 1 mm free Ca²⁺ (gray, dashed line). (C) Intrinsic Trp fluorescence emission spectrum of ∼8 µm L84F-GCAP1 in the presence of 1 mm EGTA (black, solid line), after addition of 1 mm Mg²⁺ (gray, solid line), after further addition of 1 mm free Ca²⁺ (gray, dashed line). The wavelengths of maximal emission are reported in Table 1.

L84F-GCAP1 exhibited a small quenching of fluorescence upon Mg²⁺ binding (∼10%) (Fig. 4C) with no shift of the maximum (341 nm, Table 1), in a similar fashion to the I107T variant. Interestingly, after Ca²⁺ addition, this mutant showed a smaller red shift with respect to both wild-type and I107T-GCAP1 (342 nm), accompanied by an increase in fluorescence intensity.

Monitoring Ca²⁺ and Mg²⁺ binding to I107T and L84F-GCAP1

To investigate the origin of the distinct fluorescence properties observed for the two GCAP1 mutants in their different regulatory states, we monitored the binding of Ca²⁺ and Mg²⁺ to each variant by two titration experiments. In a first approach, Ca²⁺ titrations were studied by monitoring spectroscopically the competition with the 5,5′Br₂-BAPTA chelator, both in the absence and in the presence of physiological 1 mm Mg²⁺. The working conditions under which Ca²⁺ titrations were performed ensured robust and reliable determinations in terms of protein stability and Ca²⁺ saturation. Examples of titration curves are reported in Figure 5A and B, and show very different patterns for I107T and L84F-GCAP1. Data could be nicely fitted by a two or three binding sites model and the macroscopic binding constants were hence determined (Table 2). Both variants bound Ca²⁺ with higher affinity with respect to the chelator (Fig. 5A and B) and the presence of Mg²⁺ slightly lowered the apparent affinity (Table 2). In detail, for the I107T-GCAP1 variant, which carries a mutation in the EF3 Ca²⁺ binding motif, only the higher affinity individual macroscopic binding constants K₁ and K₂ were reasonably detectable, being comparable to that of the chelator (log K_chelator = 5.6, K_Dchelator = 2.3 µm), while K₃ was below the experimental detection limit. Overall, this resulted in an apparent 2.5 µm affinity in the absence of Mg²⁺ and 2.8 µm in the presence of Mg²⁺, thus more than 10-fold lower than that of the wild-type (31).

Table 2.

Results of Ca²⁺ binding and guanylate cyclase assay experiments

	log K1a	log K2a	log K3a	KDapp (µm)b	IC50 (µm)c	EC50 (nm)d	X-folde
Wt	7.1 ± 0.9f	6.5 ± 0.8f	6.1 ± 0.5f	0.16f	0.24 ± 0.01	61 ± 32	33x
Wt Mg2+	6.7 ± 0.8f	5.9 ± 0.1f	6.1 ± 1.0f	0.50f	—	—	—
I107T	5.82 ± 0.14 (5.79 ± 0.11)	5.4 ± 0.3 (5.4 ± 0.3)	1.94 ± 1.91	2.5 (2.6)	7.5 ± 3.8	39 ± 26	17.5x
I107T Mg2+	5.987 ± 0.002 (5.98 ± 0.04)	5.1 ± 0.5 (5.2 ± 0.4)	1.5 ± 1.1	2.8 (2.7)	—	—	—
L84F	6.98 ± 0.04	6.68 ± 0.14	4.72 ± 0.17	0.15	1.8 ± 0.5	49 ± 17	10.5x
L84F Mg2+	6.24 ± 0.08 (6.2 ± 0.3)	6.58 ± 0.03 (6.5 ± 0.3)	4.90 ± 0.15	0.39 (0.43)	—	—	—

^aDecimal logarithm of macroscopic binding constants for each Ca²⁺ binding site obtained by fitting experimental data to a model with three (or two, text in italics, in paraentheses) binding sites with the software CaLigator (30); errors represent standard deviation of at least three repetitions.

^bApparent affinity obtained by averaging the significant macroscopic binding constants, see the Materials and methods section.

^cIC₅₀ is the Ca²⁺ concentration at which the GC activity is half maximal; standard errors estimated from fitting.

^dEC₅₀ is the concentration of GCAP1 at which the GC activity is half maximal; standard errors estimated from fitting.

^eX-fold activation is expressed as GC_max − GC_min divided by GC_min.

^fData from (31).

Figure 5.

Ca²⁺ and Mg²⁺ titrations of I107T and L84F-GCAP1 mutants obtained by absorption and fluorescence spectroscopies. All experiments were performed at T = 25°C in 5 mm Tris–HCl pH 7.5, 150 mm KCl, 1 mm DTT buffer. (A) Ca²⁺ titration curves obtained by absorption spectroscopy in the absence of Mg²⁺. Experimental points for I107T-GCAP1 (black circles) and for L84F-GCAP1 (gray diamonds) are shown together with the optimal curve by computer fitting (black and gray solid lines, respectively) and the theoretical (simulated) curve for the titration in the sole presence of the chelator (gray dashed line). (B) Ca²⁺ titration curves obtained by absorption spectroscopy in the presence of 1 mm Mg²⁺, experimental points for I107T-GCAP1 (black circles) and for L84F-GCAP1 (gray diamonds) are shown together with the optimal curve by computer fitting (black and gray solid lines, respectively) and the theoretical (simulated) curve for the titration in the sole presence of the chelator (gray dashed line). The curves include dilution effects upon titration. The axes have been normalized as elucidated in the Materials and Methods section. (C) Ca²⁺ titration monitored by fluorescence spectroscopy of ∼0.7 µm I107T-GCAP1 (black line and circles) and of L84F-GCAP1 (gray line and diamonds) performed in the 2 nm–100 µm range. (D) Mg²⁺ titration monitored by fluorescence spectroscopy of ∼0.7 µm I107T-GCAP1 (black line and circles) and of L84F-GCAP1 (gray line and diamonds) performed in the 9 µm–10 mm range. Titration experiments were performed by exciting Trp residues (λ_ex = 290 nm) and monitoring the intrinsic Trp fluorescence emission at λ_em = 340 nm. Three to four replicates were accumulated and results are shown as average ± standard deviation.

Surprisingly, L84F-GCAP1 showed a wild-type-like Ca²⁺ affinity in the absence of Mg²⁺ (K_D^app = 0.15 versus 0.16 µm) and an even higher affinity in the presence of 1 mm Mg²⁺ (K_D^app = 0.39 versus 0.50 µm). The shape of the titration curve in the presence of Mg²⁺ (Fig. 5B) showed a change in convexity typically observed in protein systems that bind Ca²⁺ with a positive cooperativity mechanism (30,32). Indeed, by assuming that one of the EF-hand binding motifs was occupied by a Mg²⁺ ion, fitting the experimental data to a two-binding site mechanism (Table 2) led to substantially unaltered values for the macroscopic association constants (logK₁ and logK₂) (Table 2), suggesting that the Ca²⁺ affinity for the second binding site increases as a consequence of the binding of a Ca²⁺ ion to the first binding site. It is worthwhile to notice that both wild-type and all other cone-dystrophy-related GCAP1 variants investigated to date with this method showed a sequential Ca²⁺ binding mechanism, without apparent cooperativity (17).

In a different titration experiment, we monitored the intrinsic Trp fluorescence of each variant upon titration with Ca²⁺ and Mg²⁺. Ca²⁺ fluorescence titration measurements for L84F-GCAP1 (Fig. 5C) showed a wild-type-like biphasic behavior in the physiological range of Ca²⁺ (0.1–0.6 µm) and in the whole tested range, with a minimum around 100 nm, in line with a previous work (8). Moreover, consistent with the fluorescence spectra and at odds with the wild-type, L84F-GCAP1 showed an increase in fluorescence emission intensity in Ca²⁺-saturating conditions with respect to the apo form. Interestingly, also I107T-GCAP1 showed an overall decrease in fluorescence intensity when switching between apo and Ca²⁺-saturated forms and a biphasic conformational change, yet limited to the physiological range—a feature shown also by G159V-GCAP1 (19).

Mg²⁺ fluorescence titration for L84F-GCAP1 (Fig. 5D) showed an overall decreasing pattern, in line with fluorescence spectra, but also a minimum at around 80 µm, which might possibly imply that a Mg²⁺ ion is bound to the protein with significant affinity in the absence of Ca²⁺. I107T-GCAP1 exhibited a monophasic decreasing fluorescence behavior, in line with fluorescence spectra. Experimental points could also be fitted to a four-parameter Hill sigmoid, with estimated transition concentrations of 3.6 mm for I107T-GCAP1 and 1.1 mm for L84F-GCAP1.

Structural insights from molecular dynamics simulation

The biophysical data presented above could not completely clarify some important mechanistic aspects as to the specific structural features of I107T and L84F-GCAP1 variants, which confer distinct molecular hallmarks to each mutant. In particular, the following aspects remained somewhat unclear: (i) I107T-GCAP1 may bind only two, and not three, Ca²⁺ ions in physiological conditions corresponding to dark-adapted photoreceptors, i.e. high intracellular Ca²⁺; (ii) Mg²⁺ could bind to L84F-GCAP1, presumably to EF2 and most likely to EF3 as well, with higher affinity compared with the wild-type. On the other hand, the data collected in this study indicate that the Mg²⁺-EF2 form of I107T-GCAP1 (EF2^Mg2+ I107T-GCAP1) corresponding to the GC-activating state, as well as the Ca²⁺-saturated form of L84F-GCAP1 (EF2^Ca2+ EF3^Ca2+ EF4^Ca2+ L84F-GCAP1), corresponding to its GC-inhibiting state, may indeed be of physiological relevance. We therefore ran 150 ns all-atom molecular dynamics (MD) simulations for each variant only in these latter states, and compared the results with those of the same states for the wild-type, investigated in a previous study (7).

EF2^Mg2+ I107T-GCAP1 was overall less structurally stable and more flexible than the wild-type, as clearly demonstrated by the higher α-carbon root mean-square deviation (RMSD) plot (see Supplementary Material,Supplementary Data) and root-mean square fluctuation (RMSF) per residue (see Supplementary Material,Supplementary Data). Significantly, higher flexibility was particularly apparent in the 95–105 and 115–135 regions, corresponding to the EF3 entering helix (αE3) and Ca²⁺ binding loop, and the linker region between EF3 and EF4, including portions of αF3 and αE4 (see Supplementary Material,Supplementary Data). On the contrary, the structure of EF2^Ca2+ EF3^Ca2+ EF4^Ca2+ L84F-GCAP1 was significantly more stable (see Supplementary Material,Supplementary Data) and less flexible (see Supplementary Material,Supplementary Data) than that of the wild-type in the same time frame, which is in line with the high thermal stability of this variant.

Interestingly, in the time course of 150 ns, the two GCAP1 mutants also showed distinct dynamical features in the structural region constituting the highest-affinity Ca²⁺ binding site EF3, fundamental for keeping the GC inhibiting state (9,33). Indeed, by monitoring for each mutant and for the wild-type, the RMSD of the EF3 Ca²⁺-coordinating residues D100, N104 and E111 (see Supplementary Material,Supplementary Data, left panel), we found: (i) a significantly higher flexibility of the three residues for EF2^Mg2+ I107T-GCAP1 when compared with the EF2^Mg2+ wild-type (see Supplementary Material,Supplementary Data, left column); (ii) a significantly lower flexibility of the same residues for EF2^Ca2+EF3^Ca2+EF4^Ca2+ L84F-GCAP1 with respect to the Ca²⁺-saturated wild type (see Supplementary Material, Fig.Supplementary Data, right column). While such effects were expected for the I107T mutation that is located in EF3, the decrease in flexibility of the Ca²⁺-saturated L84F mutant came as a surprise and can be interpreted as long-range structural effects originating in EF3.

In order to get some further insight into the high stability of L84F-GCAP1, we analyzed the persistency of hydrophobic interactions with the myristoyl group by computing the percentage of frames along the 150 ns simulation, in which the distance between the centers of mass of hydrophobic residues and the myristoyl group was smaller than 5.5 Å (see Fig. 1B for a snapshot of such hydrophobic residues). Results for EF2^Ca2+EF3^Ca2+EF4^Ca2+ L84F-GCAP1 showed a significantly increased persistence of hydrophobic interactions with the myristoyl group (Fig. 6C) compared with the wild-type protein for most of the residues involved, namely L13, F43, L45, L176 and I179. M59 and L183 did not show significant variations and V180 was the only residue with decreased persistence of hydrophobic interactions. The significantly tighter packing of hydrophobic residues around the myristoyl group appears to be of particular importance for the sake of keeping the myristoyl moiety buried within the EF1–EF2 domain (see Supplementary Material, Movies S1 and S2). In particular, the interactions involving F43 (∼80% of persistence) and L13 (∼28% of persistence) became stronger as the two residues dynamically moved significantly closer to one another compared with the wild-type case (see Supplementary Material,Supplementary Data, right panel, representing the distance between the C_β of the 43 and 84 residues). Overall, this reflects a remarkable decrease in solvent accessible surface for the myristoyl group (Fig. 6D) compared with the wild-type case, along the whole simulated time frame.

Figure 6.

Intramolecular interactions and solvent accessibility of the myristoyl moiety monitored by MD simulations. (A) Histograms representing the persistence of intramolecular hydrophobic interactions with the myristoyl group for EF2^Mg2+ wild-type GCAP1 (black) and EF2^Mg2+ I107T-GCAP1 (gray) over the 150 ns simulation time frame. (B) Time evolution over 150 ns of the SASA for EF2^Mg2+ wild-type GCAP1 (black) and EF2^Mg2+ I107T-GCAP1 (gray). (C) Histograms representing the persistence of intramolecular hydrophobic interactions with the myristoyl group for EF2^Ca2+EF3^Ca2+EF4^Ca2+ wild-type GCAP1 (black) and EF2^Ca2+EF3^Ca2+EF4^Ca2+L84F-GCAP1 (gray). (D) Time evolution over 150 ns of the SASA for EF2^Ca2+EF3^Ca2+EF4^Ca2+ wild-type GCAP1 (black) and EF2^Ca2+EF3^Ca2+EF4^Ca2+ L84F-GCAP1 (gray). Persistence has been calculated as the percentage of frames, for which the distance between the center of mass of the myristoyl group and the hydrophobic side chains was lower than 5.5 Å.

At variance with this behavior, in EF2^Mg2+ I107T-GCAP1 (Fig. 6A), L13, L45, I179, V180 and L183 had considerably less persistent hydrophobic interactions with the myristoyl moiety compared with the wild-type, while F43, M59 and L176 had a more persistent interaction with the acyl group. A more loosely packed myristoyl group with respect to the wild-type case was also shown by a small increase in solvent accessibility (Fig. 6B) and high structural fluctuations (see Supplementary Material,Supplementary Data).

Hydrodynamic properties and oligomeric state of I107T and L84F-GCAP1

Dynamic light scattering (DLS) measurements were performed to assess the hydrodynamic properties of the two GCAP1 variants in their specific GC-activating and inhibiting states. I107T-GCAP1 (see Supplementary Material,Supplementary Data) showed a slight increase in the hydrodynamic diameter upon binding of Mg²⁺ [d = (7.94 ± 0.11) versus (7.65 ± 0.06) nm, see Supplementary Material,Supplementary Data], while Ca²⁺ binding decreased the diameter to 6.79 ± 0.01 nm. No time-dependent aggregation was observed in all the investigated conditions as shown by the mean count rate stably ranging between 210 and 230 kcps. The reduction in the size of the mutant is in line with the behavior of the wild-type and of other COD-related GCAP1 mutants (34). Size exclusion chromatography (SEC) measurements for the same mutant (see Supplementary Material,Supplementary Data) resulted in a hydrodynamic diameter of 6.6 nm for the apo form (see Supplementary Material,Supplementary Data), which decreased to 6.0 nm upon Mg²⁺ binding and underwent a further shrinking upon Ca²⁺ binding to 5.8 nm, with a Ca²⁺-dependent increase in dimerization from 2 ± 1% to 13 ± 2% with respect to the apo form (see Supplementary Material,Supplementary Data).

DLS measurements of L84F-GCAP1 (see Supplementary Material,Supplementary Data) could not provide a definitive estimate for the hydrodynamic size in the absence of Ca²⁺ due to the presence of low-order oligomers, which affect the correlation signal. This GCAP1 variant exhibited, in the apo form, a clear time-dependent aggregation over the course of 350 min (see Supplementary Material,Supplementary Data), with an increasing count rate from 160 to 275 kcps. The aggregation propensity was somewhat impaired by Mg²⁺ binding, with a count rate between 160 and 180, and even reversed by Ca²⁺ binding, which decreased the count rate from 160–200 to 95–100 kcps. The presence of Ca²⁺ allowed also a more realistic estimation of the L84F-GCAP1 monomer hydrodynamic diameter, which resulted 8.41 ± 0.04 nm when 1 mm Mg²⁺ was already present in solution, or 8.53 ± 0.04 nm when Mg²⁺ was added to Ca²⁺-bound L84F-GCAP1. SEC measurements showed a different behavior for L84F-GCAP1 (see Supplementary Material,Supplementary Data). Indeed, the estimated size for the monomer in the apo form was 6.4 nm (see Supplementary Material,Supplementary Data), which decreased upon Mg²⁺ binding to 5.8 nm but increased again to 6.4 nm upon Ca²⁺ binding. All L84F samples though showed a higher percentage of aggregates (∼14–19%) than I107T-GCAP1, and 6–7% dimers in solution, which may explain the partial discrepancy with the DLS determinations. In summary, both DLS and SEC experiments confirmed that in the physiologically relevant states (Mg²⁺-bound and Ca²⁺-bound), both mutants are essentially monomers.

Regulation of the guanylate cyclase activity by I107T and L84F-GCAP1

GCAP1 is the physiological Ca²⁺-sensor regulating GC activity in mammalian rod and cone cells. Its regulating properties are characterized by two values: the EC₅₀ is the concentration of GCAP1 at which the activation of the target GC is half-maximal; the IC₅₀-value is the free Ca²⁺-concentration at which the GC activity is half-maximal in the presence of saturating GCAP1 concentration. We determined these values for the GCAP1 variants that were investigated in this study. The EC₅₀-values did not differ significantly among all tested GCAPs revealing 49 ± 17 nm for L84F, 39 ± 26 nm for I107T and 61 ± 32 nm for wild type (Fig. 7B and Table 2). These parameters indicate that the apparent affinity of GCAP1 for the target GC was not disturbed or changed by the point mutations. However, large differences were found for the IC₅₀-values (Fig. 7A; Table 2): while the wild-type activated GC with an IC₅₀ of 0.24 μM (X-fold activation was 33-fold), activation by the mutants was shifted to higher free Ca²⁺ (IC₅₀ = 1.8 μM for L84F and IC₅₀ = 7.5 µm for I107T, Table 2). The X-fold activation was also lower, reaching only 10.5-fold for L84F and 17.5-fold for I107T.

Figure 7.

Photoreceptor guanylate cyclase 1 (ROS-GC1) regulation by wild-type and mutant GCAP1. (A) Washed bovine photoreceptor membranes containing ROS-GC1 were reconstituted with 5 µm GCAP1 wild-type or mutants between 0.0017 and 1000 µm free Ca²⁺. Activation was half maximal (IC₅₀) at 0.24 µm (wt), 1.8 µm (L84F) or 7.5 µm (I107T) free Ca²⁺. (B) ROS-GC1 activity as a function of the GCAP1 concentration at 0.0017 µm free Ca²⁺. Activation was half maximal (EC₅₀) at 61 nm (wt), 49 nm (L84F) or 39 nm (I107T). In both cases, ROS-GC1 activity was estimated at 30°C in nmol cGMP min⁻¹ mg⁻¹ rhodopsin from the data obtained from at least three independent experiments. Error bars indicate the standard deviation.

Discussion

The analysis of biochemical and biophysical properties of COD- and CORD-related GCAP1 mutants previously performed by some of us (17,19,29) highlighted common hallmarks, which allowed a clear distinction between the wild-type and the pathologic variants. These include minor structural changes (17,29), a generally lower thermal stability (17), significantly reduced affinity for Ca²⁺ (17,29), alteration in the hydrodynamic properties (34) and in fluorescence emission (19). From a functional viewpoint, significantly increased IC₅₀ values leading to constitutive activation of the GC at physiological Ca²⁺ were observed for all the mutants investigated so far by us and by others, with less significant—if present at all—alterations in the EC₅₀ values (15).

The study presented here extends the molecular characterization of disease-associated missense mutations found in GUCA1A to the biochemical and biophysical analysis of the proteins expressed by two gene mutations recently found in Spanish families (16). The newly identified L84F and I107T-GCAP1 variants showed a rather heterogeneous clinical phenotype (16) ascribable to COD, CORD and MACD, similar to those previously observed for Y99C and P50L-GCAP1 (26,28). We asked whether the expressed proteins showed molecular features in common with the previously characterized GCAP1 mutants. Surprisingly, we found that this is not the case for both the variants.

CD spectroscopy showed that, while both GCAP1 variants are clearly folded in a typical predominantly α-helix architecture (see Supplementary Material,Supplementary Data), I107T-GCAP1 is characterized by a tertiary structure and a conformational transition pattern (Fig. 2) similar to that recently observed for D100E, L151F and G159V (29). Instead, L84F-GCAP1 showed slight but significant alterations in its tertiary structure in both the GC-inhibiting and activating states (Fig. 2). Moreover, the secondary structure and the protein compactness for L84F-GCAP1 appeared to be related to the sequence of cation binding events, as reflected by the different Θ₂₂₂/Θ₂₀₈ ratios obtained in the Mg²⁺ and Ca²⁺-saturating conditions, depending on the order of cation saturation (Table 1 and see Supplementary Material,Supplementary Data). A peculiar feature of the newly investigated mutants is that, while they both showed a 10–12°C lower thermal stability than the wild-type in the Mg²⁺ bound/GC-activating state (Fig. 3 and Table 1) (7,17), they displayed higher stability in the presence of saturating Ca²⁺ (GC-inhibiting state), a phenomenon that was particularly apparent for L84F, resulting in the most stable form of GCAP1 observed so far.

Distinct behaviors were also observed by analyzing the fluorescence emission of the two GCAP1 variants, which reflects different conformational and hydrophobicity properties following cation binding (Figs 4 and 5). Both forms showed a dynamically changing, Ca²⁺-dependent emission pattern in the 100–600 nm physiological range of Ca²⁺, indicating the operation as Ca²⁺ sensors. However, the two mutants showed opposite trends, resulting in a maximum for L84F and a minimum for I107T at ∼200 nm Ca²⁺ (Fig. 5D). It is worth noticing that the Ca²⁺-titration fluorescence pattern shown by I107T-GCAP1 lacked the typical biphasic behavior that is observed for the wild-type. However, this lack was observed earlier in other COD mutants (19), while L84F-GCAP1 displayed a unique and opposite trend, with enhanced intensity of fluorescence in conditions of saturating Ca²⁺. Results from titrations with Mg²⁺ also significantly differed for the two mutants. While I107T-GCAP1 showed a pattern compatible with a wild-type like, relatively low affinity for Mg²⁺ (8,35), L84F-GCAP1 proved to switch its conformation in the physiological range of Mg²⁺ (1 mm), hence suggesting a more prominent role for this cation in regulating its function (Fig. 5D).

Ca²⁺-titration experiments in competition with a chromophoric chelator also highlighted remarkable differences among the two mutants. I107T-GCAP1 showed, with respect to the wild-type (17,31), decreased macroscopic association constants for each EF-hand (Table 2), both in the presence and in the absence of physiological Mg²⁺, consistent with a ∼10-fold reduction in apparent affinity for Ca²⁺. Data fitting to a two-site binding model further suggest that the binding of a Ca²⁺ ion might be compromised for this mutant. This trend is substantially in line with what was observed previously with the COD/CORD-associated mutants E89K, D100E, L151F and G159V (17). On the contrary, titrations for L84F-GCAP1 showed a very different pattern, with a wild-type-like distribution of macroscopic association constants, at least for the higher affinity binding sites, and an overall comparable apparent affinity for Ca²⁺ (Table 2). Interestingly, titration experiments performed in the presence of physiological Mg²⁺ resulted in a less prominent decrease in apparent affinity compared with the wild-type [see Table 2 and (31)]. Data fitting to a three-site binding model points to a positive cooperativity mechanism, in which the binding of a Ca²⁺ ion to a first binding site increased the affinity for the second (Table 2). Notably, this proposed mechanism is consistent also with a two-site Ca²⁺-binding model (Table 2), i.e. if one of the EF-hands were occupied by Mg²⁺ under physiological conditions, the binding of Ca²⁺ could still occur with positive cooperativity.

MD simulations were run to unveil mutant-specific structural features and helped explaining their distinct properties. For instance, they highlighted a significantly higher flexibility for I107T-GCAP1 when compared with the wild-type, in the EF2^Mg2+-bound form (see Supplementary Material,Supplementary Data), which is particularly apparent in the EF3–EF4 domain. On the contrary, in the Ca²⁺-saturated state, L84F was significantly stable in all structural regions (see Supplementary Material,Supplementary Data) including the 120–135 region connecting EF3 and EF4, which has been shown to be highly dynamic in wild-type GCAP1 (7).

Structural differences between the two mutants in the vicinity of the myristoyl group became also visible by MD simulations. In the L84F mutant, the myristoyl moiety was found to be deeply buried in the cleft formed by the EF2 and EF3 motives (Fig. 6D), and most of the hydrophobic residues that interact with the acyl group, in particular, F43, L13 and L45 (Fig. 1B), were found to maintain significantly more persistent interactions compared with the wild-type, in the simulated time frame (Fig. 6C). This was not the case for the Mg²⁺-bound I107T-GCAP1, in which the myristoyl moiety appeared more loosely packed than in the wild-type (Fig. 6B), as further demonstrated by the generally less persistent hydrophobic interactions (Fig. 6A).

Finally, MD simulations permitted us to explain possible functional consequence for the pathological mutants. The EF3 Ca²⁺-coordinating residues D100, N104 and E111 were found to be more dynamical in the Mg²⁺-bound I107T-GCAP1 than in the wild-type (see Supplementary Material,Supplementary Data, left column), which is not surprising considering the location of the substitution in the EF3 loop (Fig. 1A). However, the same residues were significantly less flexible in the L84F-GCAP1 variant than in the wild-type in the presence of saturating Ca²⁺ (see Supplementary Material,Supplementary Data, right column). The remote distance of L84F to the EF3 Ca²⁺-binding loop implies an intra-molecular communication pathway that connects the region of the myristoyl group and the EF3 binding motif, which occurs via an allosteric mechanism. Such intra-molecular communication may be related to the one observed between the same region and EF4 in myristoylated wild-type GCAP1, and is involved in the recently proposed calcium-myristoyl ‘tug’-mechanism for tuning Ca²⁺ sensitivity and interaction with the GC target (36).

In spite of their remarkably distinct biochemical and biophysical properties, I107T and L84F-GCAP1 showed rather similar functional properties, in terms of aberrant regulation of the GC target. They both led to constitutive activation of the target enzyme at physiological Ca²⁺ (Fig. 7 and Table 2), reflected by the 7.5-fold higher IC₅₀ value for L84F and the ∼31-fold higher IC₅₀ for I107T. In line with other COD/CORD-associated GCAP1 mutants (17), no significant correlation exists between the apparent affinity for Ca²⁺ of the GCAP1 variant and the corresponding IC₅₀ values (Table 2). This apparently surprising evidence might be explained by the fact that the IC₅₀ values summarize a rather complex series of events, including the replacement of cations in the GC-bound GCAP1 and the concerted conformational change transferred to the target, as well as the rearrangement of the GC dimer interface. Interestingly, the concerted action of these sequential steps is mirrored in the extremely high cooperativity of GC activation (37), although Ca²⁺ binding to isolated GCAP1 per se occurs generally without cooperativity (17). Indeed, GCAP1 might display different Ca²⁺ affinities when GCAP1 is bound to the GC, pointing to a very complex regulation of the GC–GCAP1 complex. As proven recently by some of us (38), when GC variants carrying mutations in its dimerization domain are incubated with wild-type GCAP1, a significant shift in Ca²⁺-sensitivity is observed. Although GC is per se not a Ca²⁺-sensing unit, it seems to integrate and process information regarding Ca²⁺ sensing by GCAP1. Similar phenomena are observed in other Ca²⁺-sensor systems, for example, for the ubiquitous protein calmodulin, which changes its affinity for Ca²⁺ when bound to the several physiological targets (39).

In terms of target functionality, both GCAP1 mutants showed remarkably decreased activation of the target GC (10.5× in the case of L84F, 17.5× in the case of I107T, versus 33× in the case of the wild-type). However, the EC₅₀ values for the two mutants were relatively similar to those of the wild-type (Table 2), suggesting that the interaction between GCAP1 and GC may occur for the mutants with similar affinity as for the wild-type under physiological conditions.

A previous bioinformatic analysis (16) led to a proposal about the pathological mechanisms underlying the altered function of the two GCAP1 mutants. For I107T, perturbation of protein secondary (and eventually tertiary) structure was predicted, together with similar effects predicted for previously characterized COD/CORD-associated mutants (15). Since the same bioinformatic approach led to the prediction of no structural alteration for L84F, the authors hypothesized that the nucleotidic mutation c.250C>T may disrupt a splicing enhancer and/or create a splicing silencer (16), so that the final cause for disease would in that case be haploinsufficiency. Our data however suggest a different scenario. Indeed, the thorough biochemical and biophysical characterization performed here highlighted for I107T-GCAP1 very minor structural alterations with respect to the wild-type, in line with what we previously observed for the D100E, L151F and G159V COD/CORD-associated mutants (17,29,34). On the contrary, and at odds with the result of previous bioinformatic predictions (16), significant structural effects among retinal-disease-related GCAP1 mutants were indeed observed for L84F-GCAP1. Although our data cannot completely exclude the haploinsufficiency hypothesis, they rather strongly point to the autosomal-dominant nature of the disease for both of the expressed mutants. The observation that both I107T and L84F-GCAP1 lead to similar perturbation of target regulation, causing the constitutive activation of GC, suggests a similarly altered inter-molecular communication mechanism, which is described for nearly all the retinal-dystrophy GCAP1 mutants investigated so far (15). In light of the calcium-relay model (11), GCAPs that are unaffected by mutations, like GCAP2 in rods, may partly compensate for the dysfunction of the mutated GCAP1, thereby still allowing almost normal visual capabilities. In the long run, however, the photoreceptor would undergo a detrimental alteration of Ca²⁺ and cGMP homeostasis, thus triggering cell death. An important goal of future studies will be to understand in molecular detail, how remarkably distinct molecular features of GCAP1 variants exert a common regulatory effect on the target enzyme, thus leading to its aberrant regulation.

Materials and Methods

Materials

Tris (hydroxymethyl) aminomethane (Tris), 3-(N-morpholino) propanesulfonic acid (Mops), CaCl₂, MgCl₂, NaCl, KCl, EGTA and DTT were purchased from Sigma Aldrich or Carl Roth and were of the highest grade available.

Methods

Cloning of L84F and I107-GCAP1 mutants

The novel sequence variants c.320T>C (p.Ile107Thr) and c.250C>T (p.Leu84Phe) were introduced by PCR site-directed in vitro mutagenesis (primers available upon request). The amplified PCR product was cut with NdeI and NheI and ligated into the expression vector pET11-bGCAP1-D6S (40). All sequences were verified by DNA sequencing (LGC Genomics, Germany).

Expression and purification of I107T-GCAP1 and L84F-GCAP1

Wild-type, I107T and L84F-GCAP1 were overexpressed in BL21-CodonPlus Escherichia coli cells and subsequently purified to homogeneity as described before (10). In order to myristoylate GCAP1 variants, cells were co-transformed with the plasmid pBB131 containing a gene for the yeast (Saccharomyces cerevisiae) NMT. After cell lysis, GCAP1 wild-type and mutants were isolated from the insoluble fraction. Briefly, GCAP1 variants were extracted from inclusion bodies by homogenization in 6 m guanidinium hydrochloride and dialyzed against 3 l Tris-buffer (20 mm Tris–HCl, 150 mm NaCl, 1 mm DTT pH 7.5). Subsequently, the protein was applied onto a size exclusion column (Superdex 75, GE Healthcare, Germany) equilibrated in Tris-buffer with 2 mm EGTA. Fractions containing GCAP1 were further purified with an anion exchange column (HiLoad 26/10 Q Sepharose; GE Healthcare) equilibrated in Tris-buffer with 2 mm EGTA (40). Chromatography was performed with a gradient of 200–550 mm NaCl in 40 ml. The purity of the obtained GCAP1 forms was verified by sodium dodecyl sulfate–polyacrylamide gel electrophoresis. Proteins were decalcified, aliquoted and lyophilized until use.

CD spectroscopy and thermal denaturation profiles

CD spectroscopy studies were performed with a Jasco J-710 spectropolarimeter equipped with a Peltier type thermostated cell holder. Both near UV (250–320 nm) and far UV (200–250 nm) spectra were recorded at 37° C at a scan rate of 50 nm min⁻¹, a bandwidth of 1 nm and an integration time of 4 s. Five accumulations were averaged for each sample, the spectrum of the buffer (5 mm Tris–HCl pH 7.5, 150 mm KCl, 2 mm DTT) was recorded before each set of measurements and subtracted. Protein concentration for far UV spectra was ∼6 µm in 0.1 cm cuvettes, while for near UV spectra, it was ∼26 µm in 1.0 cm cuvettes. Both far UV and near UV spectra were recorded: (i) in the presence of >15-fold excess EGTA with respect to protein concentration; (ii) after addition of 1 mm MgCl₂ or 2-fold excess CaCl₂ with respect to the concentration of EGTA; and (iii) after the addition of the other salt, not added in the previous step. Thermal denaturation was monitored between 20°C and 96°C in the same conditions as for far UV spectra, as far as protein, EGTA, MgCl₂ and CaCl₂ concentrations are concerned. The ellipticity signal at λ = 208 nm (θ₂₀₈) was recorded at a scan rate of 1°C min⁻¹ and a response time of 4 s. The thermal denaturation analysis was performed assuming a two-state transition process, where the θ₂₀₈ signal recorded as a function of temperature represents the fraction of folded and unfolded protein. When possible, thermal denaturation curves were fitted to a four-parameter Hill sigmoid and a value for the melting temperature was estimated.

The equation for the fitting curve was:

$$y=b_{\mathrm{n}}+\frac{|b_{\mathrm{n}}-b_{\mathrm{u}}|T^H}{T_{\mathrm{m}}^H+T^H}$$

where b_n is the baseline value (θ₂₀₈) of the native protein, b_u the baseline value of the partly unfolded protein, T the temperature, H the Hill coefficient and T_m is the melting temperature.

DLS experiments

DLS measurements were performed with a Zetasizer Nano-S (Malvern Instruments) using a polystyrene low-volume disposable sizing cuvette (ZEN0112). Viscosity and refractive index of the aqueous solvent were set at 0.8872 cP and 1.330, refractive index of the protein was set to 1.45, protein absorption was set to 0.001 and the temperature was set to 25° C with 2 min equilibration time. The measurement angle was 173° backscatter and the analysis model was set to multiple narrow modes. For each measurement, a minimum of 24 determinations were performed, each consisting of at least 10 repetitions. All experiments were performed in 5 mm Tris–HCl, 150 mm KCl, pH = 7.5 buffer previously filtered through a Jet Biofilm 0.22 µm membrane; protein solutions were filtered through a 0.02 µm Anotop 10 filter (Whatman). The time evolution of the mean count rate over 350 min was monitored for L84F-GCAP1 in the same conditions as above, where each point was an accumulation of 10 measurements performed during 1min time frame.

High performance liquid chromatography

Purified protein samples (100 µg) were analyzed for the presence of monomeric, dimeric and/or aggregated GCAP1 forms by employing SEC on a BioSep-SEC S2000 column (Phenomenex, Aschaffenburg, Germany). All experiments were performed in 30 mm Mops-KOH pH 7.2, 50 mm KCl, 4 mm NaCl, 1 mm DTT. One millimolar EGTA, 0.3 mm CaCl₂ and/or 3.5 mm MgCl₂ were added, respectively.

Fluorescence spectroscopy and Mg²⁺ and Ca²⁺ titrations

The emission fluorescence spectra after excitation at λ = 290 nm were recorded using a Jasco FP-750 spectrofluorimeter between 300 and 380 nm at 25°C, in 1 cm quartz cuvette and at a scan rate of 60 nm min⁻¹. The excitation and emission bandwidth were set to 5 nm. All experiments were performed in 5 mm Tris–HCl pH 7.5, 150 mm KCl, 1 mm DTT buffer. The collected data were obtained by subtracting the signal from the buffer to an average of three accumulations. The protein concentration for all fluorescence spectra measurements was ∼8 µm in the presence of 1 mm EGTA, after addition of 1 mm MgCl₂ and after addition of 1 mm free CaCl₂.

Ca²⁺ buffers and Mg²⁺ buffers used for fluorescence titration experiments were prepared using 5 mm Tris–HCl pH 7.5, 150 mm KCl, 100 µm EGTA and CaCl₂ or MgCl₂ at variable concentrations, everything was dissolved in or diluted with the aforementioned buffer. Free Ca²⁺ and Mg²⁺ concentration for each titration point was calculated according to the Ca-Mg-ATP-EGTA Calculator software using NIST database (http://maxchelator.stanford.edu/CaMgATPEGTA-NIST.htm) by fixing T = 25°C, pH = 7.5 and ionic strength =0.15 M. By mixing these solutions, the final free Ca²⁺ concentration was varied in the 2 nm–100 µm range, while the free Mg²⁺ concentration was varied in the 9 µm–10 mm range. Protein concentration was ∼0.7 µm for each titration point. After recording three accumulations of the spectrum in the presence of 100 µm EGTA, only the intensity at the wavelength corresponding to the maximum fluorescence emission intensity was monitored for the other titration points, which were expected to change according to Ca²⁺ or Mg²⁺ binding. For the data reported in Figure 6, Y-axis has been normalized in the following way: Y = (F₃₄₀−F_min)/(F_max−F_min), where F₃₄₀ is the fluorescence intensity of each titration point, F_min the minimal fluorescence intensity recorded and F_max the maximal fluorescence intensity recorded.

Guanylate cyclase assay

Guanylate cyclase activities were determined as described in (33,41) with buffer-washed ROS membranes as a source for native ROS-GC1 [∼96% of the total cyclase activity in the membranes from bovine ROS is provided by ROS-GC1 (41)] with the following modifications: for the determination of EC₅₀-values, the GCAP1 concentration was varied between 0.05 and 10 µm. Ca²⁺-dependent GC activities were obtained by keeping GCAP1 concentrations fixed at 5 µm and varying the free [Ca²⁺] using a K₂H₂EGTA/ CaH₂EGTA buffer system as described previously (33,42). The free Mg²⁺-concentration was 1 mm. All incubation steps were performed under very dim red light; the reactions were stopped and further processed for analysis as described.

Determination of Ca²⁺-binding constants using a chromophoric chelator

Calcium binding to GCAP1 variants was monitored by a precise titration method described previously (30,32) based on the competition for Ca²⁺ between the protein and a chromophoric chelator, 5,5′Br₂-BAPTA, whose absorbance changes upon Ca²⁺ concentration. Briefly, a Ca²⁺-free solution of 5,5′Br₂-BAPTA was prepared by dissolving the chelator in the previously decalcified 5 mm Tris–HCl pH 7.5, 150 mm KCl buffer. Chelator concentration ranged between 17 and 26 µm, while initial Ca²⁺ concentration was 0.9–2.1 µm. For each titration experiment, Ca²⁺-free lyophilized proteins were dissolved in the chelator solution, and the final concentration (13–16 µm) was measured with the Bradford assay (43). In each experiment, 0.8–1 ml of chelator-protein solution was inserted in 1.0 cm cuvette and the absorbance at 263 nm (λ_max for the Ca²⁺ free chelator) was monitored upon titration with fixed amounts (0.8–1 µl, respectively) of 3 mm CaCl₂. Titrations were performed until no significant change in absorbance was detected. Three or four replications of each titration experiment were performed, data (Fig. 5) were fitted using a Newton–Raphson direct least-square fitting procedure implemented in the CaLigator software (30) to obtain individual macroscopic binding constants (log K_A, Table 2), which were further used to determine apparent affinity values (K_D^app, Table 2) allowing for quantitative comparisons between the Ca²⁺ affinity of the wild-type and the mutants. With the same argument as in previous works (17,31), we defined the apparent affinity of each mutant from the average of the logarithms of the first two macroscopic binding constants [K_D^app = 10^{−(logK1 + logK2)/2}]. Given the much lower affinity for the third site, this apparent affinity is not affected by the poorly accurate information on the low affinity binding site (i.e. log K₃ for I107T-GCAP1, Table 2).

All Ca²⁺ titrations were performed at room temperature. Axes in Figure 5 were normalized as follows: Y = (A₂₆₃ − A_Ca)/(A₀ − A_Ca), where A₂₆₃ is the absorbance at λ = 263 nm, A_Ca the absorbance at saturating [Ca²⁺] and A₀ the absorbance at the lowest [Ca²⁺]; X = [Ca²⁺]_tot/([Q]+3[P]), where [Ca²⁺]_tot is the total Ca²⁺ concentration, [Q] the concentration of the chelator (µm), 3 is the maximum number of binding sites per protein and [P] the measured concentration of the protein (µm).

MD simulations and analyses

A homology model of human myristoylated GCAP1 was built using the three-dimensional structure of chicken GCAP1 in its Ca²⁺-bound form as a template (44) according to a procedure elucidated in a previous work (17). Simulated states were modeled either by removing Ca²⁺ ions from each respective EF-hand binding site or by substitution with an Mg²⁺ ion as elucidated in (7). MD simulations of EF2^Mg2+ wild-type GCAP1 and of EF2^Ca2+ EF3^Ca2+ EF4^Ca2+ wild-type GCAP1 are the first 150 ns of the trajectories produced in (7). EF2^Mg2+ I107T-GCAP1 and EF2^Ca2+EF3^Ca2+EF4^Ca2+ L84F-GCAP1 mutants were built using the structures of the states simulated in (7) as a template, by replacing the mutated residue with the highest-scored rotamer proposed by the ‘Mutate Residue’ function implemented in Maestro (Schrodinger) software.

MD simulations of the two mutants were performed using GROMACS 4.6.3 simulation package (45), with the CHARMM27 all-atom force field (46). CHARMM27 parameters for describing the post-translational myristoylation were generated manually, and are available upon request for academic use. All settings for MD simulations were the same as in (7), in which details are provided. The system was equilibrated at 310 K for 2 ns of backbone position-restrained MD simulations and then at 310 K for 2 ns of unrestricted MD simulations, and the resulting structure was used as a reference in RMSD and RMSF calculations. After equilibration, the system underwent 150 ns unrestrained isothermal–isobaric (NPT ensemble; T = 310 K, P = 1 atm) MD simulation.

Generated MD trajectories were subjected to different analyses. The solvent accessible surface area (SASA) and the α-carbon RMSD were calculated over the 150 ns trajectory, the latter being referred to each structure after the 4 ns equilibration. The intrinsic flexibility of the system was evaluated by computing the RMSF for each amino acid, in the time frame of the simulation, defined as time averaged-RMSD with respect to the reference structure:

$$\text{RMS}{\text{F}}_i=\sqrt{\frac{1}{T}\sum _{t_j=1}^T{|r_i(t_j)-r_i^{\mathrm{ref}}|}^2}$$

where T is the 150 ns MD time frame and r_i(t_j) and r_i^ref represent the coordinates of the C_α atom of each residue at the time t_j and after 4 ns equilibration, respectively.

RMSF analysis was applied only on all α-carbons, RMSD analysis was applied on α-carbons of both the entire protein and selected residues, while the SASA was calculated on the atoms forming up the myristoyl group. RMSD, RMSF and SASA analyses were performed using Wordom software (47,48).

Persistence over the 150 ns time frame of hydrophobic interactions involving the myristoyl group was calculated as the ratio of frames, in which hydrophobic interaction could be detected between hydrophobic residues and the myristoyl group, over the total number of frames. Hydrophobic interactions were present if the distance between the center of mass of the side chain of hydrophobic residues and the myristoyl group was smaller than 5.5 Å. This analysis was performed using the PyInteraph software (49).

Supplementary material

Supplementary Material is available at HMG online.

Funding

This work was supported by funding from the Italian Ministry for Research and Education (MIUR) via Departmental funding (FUR2013 to D.D.O.), from CINECA through the Italian Super Computing Resource Allocation project (ISCRA Grant HP10CB7L79 to D.D.O.) and from the Deutsche Forschungsgemeinschaft (DFG grant KO948/10-2 to K.-W.K.). V.M. was the recipient of a travel grant provided by Consorzio Interuniversitario per le Biotecnologie (CIB).

Acknowledgements

Invaluable technical assistance by Stefan Sulmann and Matteo Tiberti is gratefully acknowledged. Part of this work resulted from a Fellowship at the Hanse-Wissenschaftskolleg Delmenhorst (Germany) to D.D.O.

Conflict of Interest statement. None declared.