Imaging cellular signals in the heart in vivo: Cardiac expression of the high-signal Ca²⁺ indicator GCaMP2

Development of GCaMP2.

Designed and random alterations in the previously described, circularly permutated eGFP-based, Ca²⁺-sensing molecule, which we now term GCaMP1 (11), were undertaken to improve brightness and stability. Ca²⁺-dependent fluorescence of GCaMP arises from the interaction between Ca²⁺/calmodulin at the C terminus of cpGFP and an N-terminal myosin light chain kinase (MLCK) fragment. GCaMP1 has only a fraction of the brightness of the parent eGFP molecule and is unstable at physiological temperatures. To address these deficiencies, three mutations were first introduced into GCaMP1: V163A, S175G, and A206K (Fig. 1a); V163A and S175G are previously described mutations that improved the temperature stability of GFP (24), whereas the A206K was introduced to prevent GFP dimerization (25). V163A and S175G mutations in GCaMP1 were previously reported as GCaMP1.6 (26), result in an intermediate improvement in the brightness of GCaMP1, but neither these mutations or A206K correct the instability above 30°C. A plasmid leader sequence (RSET) was attached to the N terminus for purification, GCaMP1.6 cDNA was subjected to PCR-based random mutagenesis, and bacterial colonies displaying the brightest fluorescence at 37°C were picked and sequenced (see Materials and Methods). This process identified two new mutations (D180Y and V93I) in separate β-sheets that improved the brightness of the cpGFP molecule. The addition of the N-terminal 35-residue polyHis RSET linked to the MLCK sequence was critical for thermal stability, because removal of the leader sequence led to loss of fluorescence at 37°C. The RSET-M13-cpGFP-calmodulin molecule GCaMP2 (GenBank accession no. DQ381402) is, respectively, ≈200 times and 6 times brighter than GCaMP1 and GCaMP1.6, and displays a similar 4- to 5-fold increase in signal between 0 and saturating Ca²⁺ (Fig. 1b; Table 1). Most importantly, the thermal stability of GCaMP2 results in retained high brightness at 37°C (Fig. 1c). Titrations of purified GCaMP2 protein with fixed Ca²⁺ solutions indicated highly cooperative Ca²⁺ binding over the physiological range of cellular free Ca²⁺ (K_D = 146 nM; Hill coefficient = 3.8), with no significant Mg²⁺ dependence (data not shown). The high brightness, thermal stability, and large dynamic range of GCaMP2 constitute significant attributes relative to previously reported genetic Ca²⁺ sensors (refs. 11, 12, and 22 and Table 1), suggesting that the molecule may enable the detection of physiological Ca²⁺ signals in mice in vivo. A previous report (27) has described the use of GCaMP2 for in vitro measurements.

Transgenic Mice Conditionally Expressing GCaMP2 in Heart Cells.

We conditionally expressed GCaMP2 in heart cells using a tet-Off system, to avoid potential pathology associated with high expression of a Ca²⁺-binding protein in myocytes. GCaMP2 was placed downstream of a weakened α myosin heavy chain (ΔαMHC) promoter and seven tetO enhancer sequences (28) (Fig. 1d). Six founder lines (ccGCaMP2 mice) were crossed with hemizygous αMHC-tetracycline transactivator allele (tTA) transgenic mice expressing the hybrid transactivator protein (28) and screened for transgene expression in the absence of doxycycline (dox). Three lines of ccGCaMP2 were selected as having strong, homogenous transgene expression in all areas of the heart and minimal or no leak expression of GCaMP2 in the absence of the tetracycline transactivator allele (tTA-mice) (Fig. 1e). As expected, transgene expression in the hearts of double transgenic [ccGCaMP2⁺/tTA⁺ (ccGC2)] mice was restricted to myocytes, with no staining observed in endocardial, coronary vascular smooth muscle, or non-muscle tissues; expression was observed in smooth muscle in the large pulmonary arteries, likely reflecting the derivation of these tissues from myocardium (29). Administration of dox for 2–5 weeks resulted in a time-dependent decrease of transgene expression, with almost complete suppression by 3–4 weeks (Fig. 1e). Unregulated expression of GCaMP2 from birth resulted in significant cardiomegaly in adult double transgenic mice, similar to that reported in mice with targeted overexpression of calmodulin (30). However, heart size in mice maintained on dox from birth and hearts from GCaMP2⁺ or tTA⁺ single transgenic mice were not enlarged relative to strain-matched WT controls (Fig. 1g), further demonstrating effective regulation of transgene expression by dox. Because GCaMP2 expression is effectively regulated, ccGC2 mice should be useful for the study of cardiac hypertrophy during controlled induction. To demonstrate that this phenotype can be effectively avoided, mice were maintained on dox starting in utero until 13–15 weeks and then removed for up to 6 weeks, and transgene expression was monitored. No cardiomegaly was detected up to 6 weeks after withdrawal (Fig. 1f), whereas transgene expression was detected by 2 weeks and robust functional signals were recorded beginning 4 weeks after withdrawal (see below). We confirmed that expression of GCaMP2 does not itself alter excitation–contraction coupling characteristics of heart cells by transiently loading control and transgenic hearts with the Ca²⁺-sensitive dye Rhod2. In four transgenic and age-matched control hearts, the amplitude and duration of Rhod2 ventricular Ca²⁺ transients were not significantly different (Fig. 1g).

Optical Measurements of Ca²⁺ Transients in the Heart in Vivo.

A prominent increase in eGFP fluorescence was observed in vivo in open-chested, ventilated, ccGC2 mice in all heart regions with every systole. As shown in Fig. 2a and Movie 1, which is published as supporting information on the PNAS web site, Ca²⁺-dependent fluorescence was restricted to cardiac muscle, resulting in the highlighting of the vasculature pattern by the surrounding fluorescent cardiac myocytes. The increase in fluorescence observed between end diastole and peak systole varied between ≈0.5 and 1.5 (ΔF/F_diastole) in spontaneously beating, adult hearts, and high signal/noise Ca²⁺ transients could be measured in all areas of the surface without constraining heart motion. Similar results were obtained in neonatal mice as early as 1 day after birth (Fig. 2b), although the level of transgene expression was more modest at this stage. Inotropic stimulation of the heart with isoproterenol (iso) resulted in a marked augmentation in end-diastolic and peak-systolic GCaMP2 fluorescence, and in the amplitude and rate of rise of the Ca²⁺-dependent fluorescent transients, which were obtained from individual heart regions as small as 30 μm² beginning at postnatal day 1 (Fig. 2 b and c). In vivo recordings were also obtained from phenotypically normal mice withdrawn from dox for 4–6 weeks (Fig. 2d), indicating the ability to experimentally circumvent cardiac hypertrophy. To our knowledge, there are no previously reported measurements of cellular Ca²⁺ transients in the mammalian heart in vivo, or while the heart is pumping against a physiological load.

Optical Measurements of Ca²⁺ Transients in the Isolated, Perfused Heart.

To more quantitatively examine GCaMP2 signaling in the intact heart, cellular Ca²⁺ transients from the heart surface were recorded in spontaneously beating, isolated, perfused hearts by using photodiode arrays or a high-speed complementary metal oxide semiconductor (CMOS) camera (see Materials and Methods). The high brightness of GCaMP2 enabled the recording of Ca²⁺ signals with excellent signal/noise characteristics and spatial resolution over the entire heart surface (Fig. 3a and Movie 2, which is published as supporting information on the PNAS web site). Fluorescent transients in paced hearts were not altered by inhibition of contraction with 2 μM cytochalasin D, indicating a lack of significant motion artifact in this system (data not shown). Transients recorded at frame rates as high as 5 kHz from a single pixel comprising a spatial dimension of ≈40 μm², or <10 myocytes, were highly stable, providing reproducible Ca²⁺ signals from all heart regions for up to 5 h without signal/noise deterioration or loss of signal amplitude (Fig. 3 b and c). In a series of experiments, the conduction velocity of the Ca²⁺ transient was 24.1 ± 3.7 cm/s and increased 8–20% after perfusion with 1 μM iso. The spectral separation between GCaMP2 and commonly used potentiometric dyes (31) also allowed the simultaneous optical recordings of action potential and Ca²⁺ transients using spatially registered photodiode arrays (32). GCaMP2-expressing hearts were briefly loaded with RH237, illuminated at wavelengths that excited both indicators, and imaged with spatially registered photodiode arrays (Fig. 3d). Action potentials and Ca²⁺ signals were recorded from the same heart regions in spontaneously beating and paced hearts; the presence of the Ca²⁺ indicator in cardiac cytosol did not alter the characteristics of cardiac action potentials.

Biophysical Characteristics of GCaMP2 in Vivo.

When expressed as transgenes in animals, genetically encoded sensors undergo unpredictable changes in photochemical properties, most likely because of interactions between the indicator and endogenous proteins, resulting in a marked degradation of indicator signaling properties relative to those observed in transiently transfected cells in tissue culture (20, 21, 23). To determine the behavior of GCaMP2 in this context, we compared Ca²⁺-dependent fluorescence of heart cytosol from ccGC2 mice with GCaMP2 protein purified from bacteria. The Ca²⁺ affinity of GCaMP2 in cardiac cytosol was somewhat lower than the purified protein (K_D 188 versus 146 nM), but the Hill slope (3.8) was equivalent and the high indicator dynamic range was maintained (data not shown). An equivalent shift was detected by using cardiomyocytes isolated from transgenic hearts and fixed Ca²⁺ concentrations in the presence of ionomycin (data not shown). We estimated the concentration of GCaMP2 expressed in ccGC2 mouse hearts at 1.6 μM, by comparing the magnitude of fluorescence of diluted cytosol (488/508) under saturating Ca²⁺ conditions with that of the purified protein.

Fusion protein indicators such as GCaMP2 undergo complex transitions to and from the photoactive state, potentially resulting in kinetic discrepancies between the underlying signal and emitted fluorescence. In the context of rapid Ca²⁺ signaling in the heart, these transition delays would be expected to result in a significant distortion of the underlying rapid Ca²⁺ transient kinetics and might be affected by interaction with endogenous proteins. We examined the fluorescence association and dissociation kinetics of GCaMP2 in ccGC2 heart cytosol in stop-flow experiments and compared these with the purified protein. As shown in Fig. 4a and b and Table 2, the association time constant was ≈14 ms and did not vary significantly as a function of cytosol concentration; addition of purified GCaMP2 protein to cardiac cytosol also did not alter the association kinetics (data not shown). The fluorescence association rate constant of the purified protein was equivalent to that of cardiac cytosol, indicating that neither competing Ca²⁺ buffers within the cytosol nor interactions between GCaMP2 and cytosolic proteins significantly alter the kinetic performance of GCaMP2. The dissociation time constant was ≈75 ms for cardiac cytosol, which was equivalent to purified protein and unaffected by cytosol dilution or addition of recombinant GCaMP2 (Table 2). Thus, the association and dissociation time constants for GCaMP2 are markedly faster than reported for GCaMP1 (11), almost certainly because of evaluation of the less stable GCaMP1 kinetics at room temperature; not surprisingly decreases in temperature similarly delayed fluorescence transitions in GCaMP2 (Fig. 4 a and b and Table 2). Of interest was the absence of the prominent Ca²⁺ dependence of the association time constant of GCaMP1, which increases steeply below ≈400 nM (11); fluorescence on and off rates for GCaMP2 were equivalent over a 200-fold range of free Ca²⁺. The more rapid and simpler kinetics of GCaMP2 at 37°C constitute a marked advantage for the measurement of rapid Ca²⁺ signals although, in the context of the rapidly beating mouse heart, fluorescence transitions in the ms range would be expected to produce predictable distortions of the underlying Ca²⁺ transients at higher rates.

To determine the effect of slower fluorescence transitions of GCaMP2 on rapid heart Ca²⁺ transients, we measured Ca²⁺ signals in paced hearts from ccGC2 mice transiently loaded with Rhod2, a fast Ca²⁺ indicator. GCaMP2 and Rhod2 fluorescent signals simultaneously recorded from the same array subregions provided a direct comparison of the raw signals. The dynamic range (ΔF/F_o) of GCaMP2 was almost three times greater than observed for Rhod2 during cardiac cycles (data not shown); however, GCaMP2 fluorescence signals displayed a consistent, rate-dependent delay relative to the fast dye (Fig. 4c). In a series of experiments in hearts paced at 200-ms cycle length, the rise of the GCaMP2 raw fluorescent signal was 45% slower than the Rhod2 signal (38.2 ± 4.2 vs. 26.3 ± 2.8 ms), and the diastolic decay similarly prolonged. The lagging GCaMP2 signals were well explained by the slower fluorescence transition kinetics, because the converted Ca²⁺ signals matched when corrected by using fluorescent association and dissociation rate constants obtained experimentally (Fig. 4c Inset; Rhod 2 transition kinetics were submillisecond and could not be measured by stop-flow). At cycle lengths lower than ≈200 ms, a rise in basal fluorescence was observed proportional to the stimulation rate, indicating that complete decay of fluorescence did not occur during diastole. Ca²⁺ sparks were not observed in single, voltage-clamped myocytes enzymatically prepared from the hearts of double transgenic mice, a limitation shared by other small molecule Ca²⁺ indicators such as Fura2; bulk cytosolic transients were readily recorded in single myocytes (data not shown).

Conduction of Ca²⁺ Waves in the Embryonic Heart.

Genetically encoded cell sensors are particularly useful in the dissection of the complex cellular interactions that occur during development. We examined GCaMP2 signals in spontaneously beating embryonic hearts by imaging open-chested embryos in vivo and ex vivo. Signals were augmented in embryos by mating mice homozygous for GCaMP2, with one parent also expressing tTA, resulting in embryos with two functional GCaMP2 alleles and markedly augmented transgene expression. Ca²⁺-dependent fluorescent signals were robust across the surface of the tubular, unseptated heart beginning at embryonic day (e.d.) 10, shortly after regular contractions are initiated, a period in which only limited ex vivo optical measurements have been attempted (33, 34). We were particularly interested in understanding the mechanism of atrioventricular (AV) delay and the timing of development of a functional AV node. As shown in Fig. 5a, prominent Ca²⁺ waveforms initiated in the atria of spontaneously beating embryonic hearts were rapidly conducted across the atrial surface. A dramatic slowing of the conducted Ca²⁺ wave occurred in the region of the AV canal, which separates the common atrial and ventricular chambers (Fig. 5 a and b and Movie 3, which is published as supporting information on the PNAS web site). Ca²⁺ waves exiting from the AV canal region continued along the ventricular surface; rapid conduction occurred through the ventricle, indicating a functional connection of the AV conduction pathway to the His/Purkinje system, consistent with histological (35) and electrophysiological (33, 34) data. The rate of conduction through the AV canal was ≈10-fold slower than through the atrial and ventricular chambers, effectively delaying connection to the ventricular conduction pathway, because ventricular breakthrough occurred after transmission of the signal through the AV canal (Fig. 5c), providing a sufficient delay for effective mechanical pumping. At e.d. 10.5, the atria could be efficiently paced from the ventricles, and bidirectional conduction was observed through the AV canal (data not shown). A distinct surface pathway has been anatomically identified in the embryonic mouse heart and hypothesized to mediate AV conduction before formation of a functional AV node (35); and a similar region of slowed cardiac conduction has been documented in the chick heart (36) but has not previously been demonstrated in mammals. Cells in this tissue layer display a distinct genetic pattern (37) and are eliminated by programmed cell death by e.d. 13.5 in the mouse (35). Consistent with the timing of elimination of these cells, we observed a loss of surface conduction through the AV canal at e.d. 13.5 (Fig. 5d); surface conduction of the Ca²⁺ waves ends at a discrete ring and does not travel continuously along the ventricular wall (see also Movie 4, which is published as supporting information on the PNAS web site) but emerges at the apical ventricular surface after a slight distinct delay. Similarly, pacing of embryonic hearts from the ventricles did not result in retrograde conduction into the atria (data not shown), indicating development of effective unidirectional conduction. These data indicate the presence of an AV conduction pathway that connects to both ventricular myocardium and the developing conduction system, demonstrating the important functional role of AV canal conduction in the preseptated heart and lending functional support to the importance of AV canal developmental abnormalities in fetal and adult arrhythmias. Given the difficulty in loading the heart without perfusion and the relatively slow spontaneous embryonic heart rate, ccGC2 mice will be particularly useful for the study of cardiac organogenesis.

Generation of Tet-Off αMHC-CaMP2 Mice.

The αMHC fragment containing seven tetO sequences (28) was ligated to a GCaMP2 BglII/NotI digestion product to αMHC-GCamP2. The 7,831-bp transgene was excised with NotI, gel purified, and injected into oocytes using standard techniques. Animals were genotyped by using primers for eGFP (18). Three lines were bred to tTA-αMHC mice to produce double transgenic mice. Transgene suppression was examined by addition of dox (1 mg/ml) in 2% sucrose in drinking water. Immunocytochemical methods were as described (38).

Ca²⁺-GCaMP2-Binding Kinetics.

Fluorescence association and dissociation was monitored in a stop-flow spectrofluorometer (Applied PhotoPhysics, Surrey, UK). Mouse hearts (2 ml volume) were suspended in 8 ml of buffer (200 mM sucrose/20 mM histidine, pH 7.2), homogenized with a Teflon glass homogenizer, and centrifuged at 2,000 × g for 20 min. Aliquots of Ca-EGTA-buffered solutions were added to the supernatant to clamp free Ca²⁺ at the desired level (39).

Optical Mapping.

Hearts were perfused and suspended in a custom-designed optical chamber (32). Di-4-ANNEPS and Rhod2 loading, photodiode array mapping, restitution protocols, and analysis were as described (40). GCaMP2 was excited at 488 nm, and di-4-ANEPPS and Rhod-2 were excited at 530 nm. Emission was collected at >520 nm for eGFP, >610 nm for di-4-ANEPPS, and 585 nm for Rhod-2. Activation maps that included both the atria and ventricles (100 × 100 pixels) were captured with a CMOS camera (Micam Ultima; SciMedia, Irvine, CA) at ≥1 kHz. In vivo images were obtained by using a cooled charge-coupled device (CCD) camera (iXon; Andor Technology, South Windsor, CT) and acquired at 66 or 128 frames per s.

In Vivo Imaging.

Mice were time-mated, and embryos removed at the designated period of development by dissection from the uterus. Embryos were kept on ice until studied at 37°C in Tyrode’s solution (136 mM NaCl/5.4 mM KCl/1 mM MgCl₂/0.33 mM NaH₂PO₄/2.5 mM CaCl₂/10 mM Hepes/10 mM glucose, pH 7.4) and imaged using a dissecting microscope (MSFLIII; Leica Microsystems, Heerbrug, Switzerland) and intensified charge-coupled device camera (iXon 860-BI; Andor Technology). Data were analyzed and processed by using customized matlab, imagej, or idl routines. Color-coded images were normalized on a pixel-by-pixel basis throughout the series, setting the highest value to the maximum value, the lowest value to 0, and linearly scaling other values. Neonatal and adult mice were anesthetized with isofluorane and cannulated via tracheotomy. Positive pressure ventilation was initiated, and the chest was opened to expose the heart. Animals were maintained at 37°C with a heating pad, and ventilation frequency and tidal volume were adjusted to eliminate spontaneous breathing. The ventilator was briefly halted during image capture.