The DNA coding sequences of Case12 and Case16 sensors were amplified by PCR from the corresponding pQE30-based expression vectors (Qiagen) as described in [7]. For amplification of cDNA inserts a “sticky-end” PCR [17] was performed using a pair of 5′-end primers complementary to M13-peptide (5′-CTCACGTCGTAAGTGGAA-3′; 5′-GATCCTCACGTCGTAAGTGGAA-3′) and a pair of 3′-end primers complementary to CaM (5′-GGCCGCATTATTTTGCAGTCATCATCT GTACG-3′; 5′-GCATTATTTTGCAGTCATCATCTGTACG-3′) containing Bam HI and Not I restriction sites, respectively. To obtain the final expression vectors the coding sequence of the corresponding sensor variant was cloned using BamHI/NotI restrictions sites into reengineered expression vector pET41a(+) (Merck Biosciences) where the GST-tag has been deleted and the thrombin site was replaced by N-terminal fusion partner containing a His₆-tag followed by a S-tag and a PreScission protease cleavage site [18].

Each of the plasmids coding either Case12 or Case16 sensor variant was transformed into BL21(DE3) Tuner E. coli cells. The bacteria were grown in TB mod cultivation medium containing 0.1 M MOPS buffer and 30 mg/L kanamycin to an OD₆₀₀ = 0.8. For the induction of protein expression 0.1 mM IPTG was added and cells were incubated overnight at 20 °C. The cell pellets were harvested by centrifugation, resuspended in IMAC buffer A (50 mM Na-phosphate, 300 mM NaCl, 20 mM imidazole, pH 8.0) and lysed with a high pressure homogenizer (Avestin). The filtered lysate containing soluble proteins was loaded on a 5 mL His-Trap IMAC column mounted on an AEKTA Explorer 100 chromatography system (GE Healthcare). After extensive washing of the column with IMAC buffer A, 500 U of PreScission protease (GE Healthcare) were loaded onto the His-Trap column for direct cleavage of the N-terminal His6 S-tag fragment from the fusion protein. After further incubation at 4 °C during 10 h the cleaved Case12 and Case16 sensor proteins were eluted from the column with IMAC buffer A. Pooled fractions were subjected to size-exclusion chromatography that was performed on a Superdex75 XK16/60 column (GE Healthcare) using TBS running buffer (50 mM Tris-HCl, 150 mM NaCl, pH 7.4). The fractions containing purified sensor proteins were pooled and aliquots were snap-frozen at −80 °C. LC-MS analysis confirmed the correct and expected mass of the proteins with the overhanging amino acids Gly-Pro-Gly-Ser at the N-terminus derived from the PreScission protease site and the BamHI restriction site as described earlier [18].

For “crystallization condition A” (low Ca²⁺ concentration) crystallization was performed in a hanging drop, vapor diffusion set-up (Case16 structure A). Protein stock solution: 4.1 mg/mL Case16, 50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 10 mM DTT. Reservoir solution: 50 mM imidazole, 1.9 M Na₂malonate, pH 6.4. Drop: 3 μL protein solution and 1 μL reservoir solution. The reservoir volume was 0.5 mL. Incubation was performed at 20 °C. For “crystallization condition B” (high Ca²⁺ concentration) crystallization was performed in a hanging drop, vapor diffusion set-up (Case12 structure B). Protein stock solution: 7.6 mg/mL Case12, 50 mM Tris-HCl, pH 7.4, 150 mM NaCl. Seed stock solution: 200 mM CaCl₂, 20% PEG-3350. Reservoir solution: 100 mM Tris-HCl, pH = 5.5, 100 mM (NH₄)₂SO₄, 21% PEG-3350. Drop: 1 μL protein solution, 1 μL reservoir solution and 1 μL seed stock solution. The reservoir volume was 0.5 mL. Incubation was performed at 20 °C.

The data set of Case16 crystals (“crystallization condition A”) was collected to a resolution of 2.35 Å at the PXII (X10SA) beam line at the Swiss Light Source (SLS) in Villigen, Switzerland. This beam line was equipped with a MARCCD detector and cryo-stream to keep the crystals at 100 K. Images were collected with 0.5° oscillation each and a crystal-to-detector distance of 200 mm. The dataset of the Case12 crystal (“crystallization condition B”) was collected to a resolution of 2.6 Å on a Rigaku FRE generator equipped with a MAR Image plate and an Oxford Cryo-system. Raw diffraction data were processed and scaled with XDS/Xscale [19] within the APRV program package [20]. The data collection statistics can be found in Table 1.

Case16 structure A (at low Ca²⁺ concentration) was solved using the molecular replacement program MOLREP [21] with GFP as a search model. The distorted CaM was traced by placing polyalanine α-helices into the residual electron density, followed by several cycles of refinement and gradually completing the model. For Case12 structure B (at high Ca²⁺ concentration) cpFP and CaM were localized independently by molecular replacement. Model building was performed using the COOT program [22]. The model was refined by the program REFMAC [23] which is part of the CCP4 suite (Collaborative Computational Project 1994). The refinement statistics is given in the Table 1. Figures were made using the program PYMOL (DeLano Scientific LLC, Palo Alto, CA, USA).

Case12 and Case16 sensors were analyzed at concentrations of 4 mg/mL or 16 mg/mL. For higher concentrations the protein samples were concentrated using an Amicon Ultra 4 10kDa MWCO membrane (Millipore). Solutions containing 100 mM EGTA or 100 mM CaCl₂ were added to a final concentration of 5 mM resulting in Ca²⁺-free or Ca²⁺-saturated states of the sensor, respectively. After further incubation for 1 h, 200 μL of the protein solution were loaded onto a Superdex 200 HR30/10 column mounted on an AEKTA Purifier 100 chromatography system. The running buffer contained 50 mM Tris-HCl, 300 mM NaCl, pH 7.4 and either 5 mM EGTA or 1 mM CaCl_2, depending on the sensor form to be analyzed. The size exclusion column was connected serially with a MiniDawn Tri-Star multi-angle light scattering detector (Wyatt Technologies) and RI-71 refractive index detector. This set-up allowed a direct on-line determination of the oligomerization state (i.e., monomer vs. dimer) of the eluted protein.

Each of the three proteins Case 12, Case 16 and bovine CaM (Calbiochem) was incubated in TBS buffer together with 10 mM EGTA and subsequently dialyzed extensively against pure TBS to obtain the Ca²⁺-free form of the proteins excluding the Ca²⁺-chelating agent EGTA. M13-peptide (Ac-SSRRKWQKTGHAVRAIGRLSS-NH₂; Biosyntan GmbH) was dissolved in PBS. The final concentrations for different components were 10 μM for both Ca²⁺ sensor proteins, 50 μM for CaM and 200 μM for M13-peptide. The solutions containing CaM or M13 peptide were diluted serially in 2-fold steps into a 384 low volume black multi-well plate (Greiner), 20 μL in each well. An equal volume of the samples, containing either Case 12 or Case 16 was then added. The final concentration of the sensors was fixed at 5 μM and ranged between 0.2–25 μM for CaM and 0.8–100 μM for M13-peptide, respectively. Ca²⁺ sensors diluted in pure TBS served as “zero value” and all samples were mixed in triplicates. The fluorescence intensity of the samples was measured using an Envision multilabel plate reader (Perkin Elmer) with the excitation wavelength set at 485 nm and the emission wavelength set at 535 nm. The first measurement cycle was performed with the proteins in absence of Ca²⁺, giving the “low value (L)”. The second measurement cycle was performed after addition of 4 μL 0.1 M CaCl₂ (final concentration 10 mM) and incubation for 15 min at room temperature resulting in the “high value (H)”.

Earlier we described circularly permuted GFP (cpGFP)-based Ca²⁺ sensors Case12 and Case16 with superior dynamic ranges of up to 12-fold and 16.5-fold increase in green fluorescence between Ca²⁺-free and Ca²⁺-saturated forms [7]. The overall structure of these sensors is a monomer consisting of cpGFP “core” in its typical β-barrel shape inserted between an M13-peptide and CaM-domains.

In the present study Case16 sensor protein was crystallized at a low Ca²⁺ concentration (see Experimental Section, “crystallization condition A” and Table 1). No Ca²⁺ was added in the protein stock solution and in the reservoir solution. At the same time, no special Ca²⁺-free reagents or Ca²⁺ purification steps were used. Therefore the starting solution for crystallization contained background levels of Ca²⁺ coming from Na₂malonate, NaCl and other reagents. It was reported earlier that adding an excess of a Ca²⁺ chelator (EGTA) did not yield a Ca²⁺-free structure and still produced crystals at dimeric Ca²⁺-bound form [15]. However, under “crystallization condition A” we were able to freeze the Case16 sensor in its “half Ca²⁺-bound state”. The crystal structure of Case16 (Case16 structure A, refined to 2.35 Å resolution) formed a monomer in the asymmetric unit (Figure 1a).

It was reported earlier that CaM possesses 4 EF-hand Ca²⁺-binding sites with differing Ca²⁺-binding affinities [24,25]. While the high-affinity C-terminal pair of sites is occupied at relatively low Ca²⁺ concentrations (below 500 nM), the N-terminal pair is occupied only at higher Ca²⁺ concentrations (above 500 nM). Therefore the initial Ca²⁺-dependent fluorescent response of the sensor in 100–500 nM range of Ca²⁺ can be determined by the structural rearrangements in response to filling of the first two high-affinity Ca²⁺-binding pockets of the C-terminal part of CaM. In agreement with this data, the structure A of Case16 reveals that only the two C-terminal Ca²⁺-binding sites of CaM are occupied with Ca²⁺ whereas both of its N-terminal sites are Ca²⁺-free. It is assumed that this structure of Case16 corresponds to the first stage of the fluorescent response of the sensor to low (below 500 nM) Ca²⁺ concentrations.

It is also important to note that Case16 structure A demonstrates an unexpected binding mode of CaM and M13-peptide. The CaM domain is in its open extended conformation and the M13-peptide was bound only to the C-terminal lobe of CaM (Figure 1b). Most likely, this particular CaM-M13 binding mode is mediated by the two complexed Ca²⁺ ions and an aromatic residue of the M13-peptide (Trp⁹) penetrating into a deep pocket of the C-terminal lobe of CaM. The N- and C-terminal CaM lobes of the previously reported structure [26] (PDB Access code = 1 CDL) align with an rmsd (root mean square deviation) of 4.60 Å (72 C-alpha atoms) and 1.00 Å (67 C-alpha atoms), respectively. This shows that the structure of the C-teminal lobe aligns well, while the N-terminal lobe of CaM shows a completely different arrangement, not only due to the open conformation, but also due to conformational changes induced by the missing Ca²⁺ ions.

In the wild type Aequorea victoria GFP (wtGFP) amino acid residues 143–152 form the 7th β-strand of the β-barrel. In Case16 the GFP moiety is circularly permuted leading to a break between positions 145 and 148. This break shortens the 7th β-strand in comparison to wtGFP and the new N- and C-termini of cpGFP are connected to M13- and CaM-domains of the sensor, respectively (a.a. numbering corresponds to wtGFP, see Supplementary Figure 1). Local rearrangements that are caused by the binding of the M13 peptide to CaM can influence the stability of this β-strand and thus may alter the fluorescent properties of the sensor molecule. Based on the resolved structure of Case16 we propose that binding of the M13-peptide to the C-terminal lobe of CaM in the context of the sensor stabilizes the arrangement of the 7th β-strand of GFP in its natural position. In other words, amino acid residues corresponding to positions 148–150 of the GFP should protrude out of the β-barrel as was predicted for the reconstructed Ca²⁺-free structure of GCaMP2 (PDB Access code = 3EKJ) [15]. The N-terminal lobe of CaM is in close contact with cpGFP and should further “clasp” and stabilize the 7th strand of GFP β-barrel (Figure 1). Therefore, the most likely explanation for the markedly increased fluorescence response of Case16 to low Ca²⁺ concentrations is that interactions between the CaM and the M13-peptide trigger chromophore conversion to the deprotonated high-fluorescent state. This hypothesis is in good agreement with conclusions from the Ca²⁺-saturated structure of GCaMP2 [15], which suggests that the key role of Ca²⁺-bound CaM is to reduce the solvent access to the chromophore. The structural changes that occur between the half Ca²⁺-saturated and Ca²⁺-saturated forms likely induce further changes in the chromophore environment that result in the fluorescent response to high Ca²⁺ concentrations. We believe that the obtained data can be extrapolated to all cpGFP-based FCIPs since they share high sequence homology (Supplementary Figure 1).

Case16 structure A demonstrates that Glu¹⁴⁸ contacts the cpGFP chromophore directly (Figure 2), which is consistent with our previous conclusions based on site-specific mutagenesis [7]. Therefore it can influence significantly the fluorescent properties of the sensor. In contrast, Ser¹⁴⁵ is located rather distant from the chromophore (Figure 2, Supplementary Figure 2), refuting our previous hypothesis that both residues (Ser¹⁴⁵ and Glu¹⁴⁸) are in direct contact with the chromophore [7]. The present results demonstrate that Ser¹⁴⁵ is instead a part of the peptide linker joining cpGFP with CaM-domain. To confirm this we performed site-directed mutagenesis to generate the mutant variant Case16-Ser145Ala. In vitro tests demonstrated that Case16-Ser145Ala was also characterized by high-contrast Ca²⁺-dependent fluorescent response, (up to 10-fold increase of green fluorescence upon Ca²⁺-saturation) which is close to the response of Case12 (data not shown). These results indicate that the amino acid residue 145 does not contact cpGFP chromophore within cpGFP-based Ca²⁺ sensor constructs directly. This is consistent with the previously reported Ca²⁺-saturated structures of Ca²⁺ sensor GCaMP2, demonstrating that Thr¹⁴⁵ is not in close proximity to the chromophore [14,15].

Also of note is that Tyr¹⁴³ is the last residue of the sensor construct which keeps its original location in the β-barrel as compared to GFP, while the NSRDQL stretch acts as a linker between cpGFP and CaM-domains. The side chain of Tyr¹⁴³ protrudes out of the beta-barrel and thus can not interact with the chromophore of cpGFP in this structure.

Crystallization condition B (see Experimental Section) was used to generate a Ca²⁺-saturated crystal of another high-contrast Ca²⁺ sensor variant, Case12 [7]. The 3D-structure of Case12 at high Ca²⁺ concentration (Case12 structure B) was refined to a 2.6 Å resolution (Table 1) and revealed that the binding mode of the M13-peptide to the CaM-domain was similar to the recently reported Ca²⁺-saturated dimeric crystal structures of GCaMP2 [14,15]. Case12 structure B is a dimer in the asymmetric unit where the M13-peptide is bound to the CaM-moiety of the neighboring molecule (Figure 3). In contrast to the Case16 structure A, both lobes of CaM are bound to M13-peptide of a neighboring sensor molecule. As was expected for the Ca²⁺-saturated form of the sensor, all four Ca²⁺-binding pockets of CaM are occupied by Ca²⁺ ions. The binding mode between CaM and the M13- peptide is similar to that reported in [26] (1CDL)). The alignment of both structures has an rmsd of 0.91 for the whole CaM-domain (127 Cα positions).

To estimate the rate of homodimerization of Case12 and Case16 we performed in vitro measurements using static multi-angle light scattering in conjunction with gel filtration at moderate and high sensor protein concentrations in the presence/absence of Ca²⁺ (Figure 4 and Table 2). For the first set of experiments we used Case12 and Case16 at a concentration of 4 mg/mL. At this concentration, in the presence of 1 mM Ca²⁺ or EGTA, both sensors were characterized by a homogeneous molecular mass of about 45 kDa, corresponding to the monomeric form. Similar results were obtained for Case12 and Case16 at high concentration (16 mg/mL) in presence of EGTA. In contrast, at high concentration (16 mg/mL) in presence of 1 mM Ca²⁺ a second distinct molecular mass of about 90 kDa was observed indicating dimerization for both sensors. These light-scattering experiments demonstrate that in the presence of Ca²⁺ Case12 and Case16 can form dimers at high sensor protein concentrations (Figure 4 and Table 2).

Case12 and Case16 did not show any dimerization taken at 4 mg/mL or 16 mg/mL in the absence of Ca²⁺. However both indicators tend to dimerize at concentration of 16 mg/mL in presence of 1 mM Ca²⁺. At concentration of 4 mg/mL both sensors formed only low rate of the dimeric form in presence of 1 mM Ca²⁺.

Importantly the maximum fluorescent response of Case12 and Case16 in vitro does not require high concentrations of the sensor (to test the Ca²⁺ response in vitro we usually use sensor protein concentrations 0.01–0.1 mg/mL) while dimerization only becomes significant at much higher protein concentrations (16 mg/mL). Therefore the detected fluorescent response of the sensor within the whole range of the monitored Ca²⁺ concentrations is induced by its monomeric form rather than its homodimer. The obtained Case12 structure B most probably resembles the non-functional homodimer and should not be interpreted as an example of the active Ca²⁺-saturated state of the sensor. Similar data were reported earlier for GCaMP2 Ca²⁺-saturated dimeric form [14,15].

In general, high expression levels of FCIPs are desirable for monitoring of Ca²⁺ changes in vivo. It should be noted that initial attempts to generate functional transgenic mouse lines with constitutive expression of FCIPs were largely unsuccessful. The possible explanation is that at low expression levels typically achieved in transgenic mice (40–190 nM) a substantial fraction of the genetically encoded sensor becomes immobilized and unresponsive. Such low FCIP expression levels were insufficient to monitor single-cell activity and to perform single-spike detection in neurons [27]. High levels of expression are often more easily attainable using virus-mediated gene transfer because of the presence of multiple copies of the viral genome. Recombinant adeno-associated virus (rAVV) gene transfer was used to drive robust (tens of μM) expression of an improved fluorescence resonance energy transfer-based FCIP D3cpv in neurons that allowed single-spike detection [28]. The rAVV expression system permits gene delivery in a wide range of animals and should be useful for targeting most neuronal cell types in the brain. For example, using rAAV delivery strategy of Case12 it was demonstrated that astroglia is the main cellular substrate of angiotensin-(1-7) action in rat rostral ventro-lateral medulla, indicating that high expression level of FCIP is crucial for its efficient functioning [29].

On the other hand, our data indicate that homodimer formation is favored at high concentrations of GCaMP-like sensors. Although such homodimerization is negligible at moderate sensor concentration (4 mg/mL or approximately 85 μM, which corresponds to the level of FCIP expression in neurons by means of rAVV delivery), it becomes significant at 4-fold higher concentration, that can be also potentially achieved locally in living cells at high expression levels, especially for the FCIP versions targeted to the specific cellular compartments. This effect should be taken into account for the future design of improved Ca²⁺ sensors.

Apart from homodimer formation, CaM/M13-based FCIPs can also interact with free CaM and CaM-binding proteins that may interfere with their functions. In order to decrease interactions with putative intracellular targets, Palmer and coworkers successfully reengineered the binding interface of CaM and M13-peptide by computational design and generated a mutant CaM-M13 pair that was unaffected by large concentrations of excess CaM and was used to obtain improved FCIP variants [30]. However, no titration was performed using excess M13-peptide, thus it remained unclear whether this design also reduced the sensitivity to numerous CaM-binding partners.

We performed experiments where Case12 or Case16 were mixed with either free CaM or M13-peptide at various concentrations in order to check whether they may disrupt the intramolecular binding of M13-peptide to CaM within the sensor and thus decrease its dynamic range (for the experimental setup see Experimental Section). For the titration experiments, free M13-peptide and CaM were incubated with the sensor proteins at a final concentration of 5 μM (corresponding to 0.25 mg/mL). Purified Case12 and Case16 proteins were titrated in vitro with 5-fold molar excess of CaM or 20-fold molar excess of M13-peptide, respectively. Under our conditions CaM did not inhibit Ca²⁺-dependent fluorescent response of both sensor variants. In contrast, addition of M13-peptide significantly inhibited the sensor response even at low concentrations (Figure 5). Based on this data we hypothesize that interaction of the sensor construct with free M13-peptide is spatially and energetically more favorable than the intramolecular binding between M13-peptide and CaM. It remains unclear why the analogous effect was not observed with free CaM. A possible explanation is that the larger size and/or electrostatic repulsion of CaM in the sensor construct interferes with the binding of added CaM molecule to the M13-peptide of the sensor construct.