9

SUPPORTING MATERIAL

Cardiac cell culture

Primary cultures of neonatal rat ventricular myocytes were obtained by enzymatic digestion from two day old Sprague-Dawley rats based on our previously described methods (1-2). Briefly, diced ventricles were incubated overnight in cold trypsin solution (US Biological) and further enzymatically treated by a series of collagenase type II digestion steps at 37˚C. Dissociated cells were resuspended in seeding cardiac media (DMEM/F12, 10% calf serum 10% horse serum, 50 U/mL penicillin, and 50 µg/mL streptomycin) and preplated in two 45 min steps to reduce the fraction of cardiac fibroblasts. Isotropic cultures were created by seeding cell suspension on 21 mm diameter Aclar coverslips. Anisotropic cultures were created by seeding cell suspension onto grooved (6 µm width and spacing, 3 µm depth) PDMS coated 22 mm glass coverslips. Cells were maintained in Medium 199 (Gibco) supplemented with 2% v/v fetal bovine serum (Hyclone), 10 mM HEPES, non-essential amino acids, 2 mM L-glutamine, 3.5 mg/mL glucose, 4 µg/mL cobalamin, and 100 U/mL penicillin G. On culture day 5-7, confluent cardiac monolayers were used for optical mapping studies.

Data acquisition and LED illumination

Red fluorescence signals from cardiomyocytes were first converted to voltage, band-pass filtered (0.5-950 Hz) and two-stage amplified in the PDA system (RedShirt Imaging), and then sampled and acquired using a high performance A/D board (DAP 5400a, ±10 Volts, 14-bit, Microstar Laboratories) inside a host PC running Windows XP. Digital outputs of a second A/D card (DAP 840, Microstar), synchronized with the DAP 5400a, were used to control stimulus sequence, master shutter, and timing of the LED illumination (Fig. S1). The acquisition process was managed by the user via a custom designed LabVIEW 7.1 (National Instruments) graphical user interface. To ensure consistent synchronization of illumination and data sampling, the acquisition clock of the DAP 5400a was coupled to the input clock of the DAP 840, which controlled digital outputs. During each complete sampling period of the DAP 5400a, 64 clock cycles were output to the DAP 840. Each of the 64 cycles in DAP 5400a served to simultaneously acquire 8 PDA outputs (for the total of 512 channels of which 504 were optically recorded signals).

In constant illumination mode, recorded signals were sampled by DAP 5400a with a period of 409.6 μs (Fs ≈ 2441.4 Hz), low-pass filtered with a balanced 9-tap Hanning FIR at 400 Hz, down-sampled to a final rate of 1220.7 Hz, and streamed to the computer hard disk. As the 5400a A/D card acquires 8 channels of data at a time, data was sampled non-simultaneously with an interval of 12.8 μsec between groups of 8 channels. After low-pass filtering, linear temporal shift correction was performed to effectively yield simultaneous sampling.

For dual flash illumination mode, the DAP 5400a sampling period was set to 102.4 μs (Fs ≈ 9765.6 Hz) yielding a DAP 840 clock cycle of 102.4/64 = 1.6 μs. The LED driver was triggered by a digital pulse from the DAP 840 on the last acquisition clock of a DAP 5400a sampling period. Temporal shift correction was not necessary at this data rate (1.6 μs between groups of 8 channels), and digital filtering was left to post-processing in MATLAB (The MathWorks, Natick, MA) to reduce the load on the DAP 5400a processor.

The PP600 LED Driver (Gardasoft, UK) was used to drive the LED matrix. This driver can supply up to 4 A of continuous or 10 A of pulsed current for two individual LED channels, and can be switched between constant illumination and flash mode using push button control. The duration of the illumination pulse can be set in 20 μs increments, with ±1 µs variability. Activation of one LED channel can be precisely delayed relative to the activation of another channel, allowing for accurate dual pulse illumination by use of only a single external trigger pulse (supplied by DAP 840). A bench-top 15 amp DC power supply (model 22-508, Radio Shack) was used to power the LED driver. Flash illumination and data acquisition was achieved with careful synchronization of the LED driver and the DAP 5400a acquisition clock. The maximum rate of LED flashing was dictated by the exponential-like response of the PDA’s low-pass filter to a step LED pulse, the clock frequencies of A/D boards, and the resolution of LED driver, as described and shown in figure 5 and the main text.

Determination of driving currents for green and blue LED excitation

The difference in X-Rhod-1 vs. Di-4-ANEPPS baseline fluorescence level of stained cells determined the useable dynamic range of the PDA (within the total linear range of ±5V). The significantly higher baseline fluorescence of Di-4-ANEPPS vs. X-Rhod-1 and absence of Di-4-ANNEPS signal in the green channel (Fig. 6A) yielded considerably higher baseline fluorescence of blue (Fblue) vs. green (Fgreen) channel. Therefore, the difference between blue and green LED current amplitudes had to be adjusted to minimize the Fblue vs. Fgreen difference to avoid signal saturation, while amplitudes of driving currents needed to be maximized to obtain the highest S/N ratios in Vm and Cai signals. We found that coating a 0.1 mm glass coverslip with five layers of PSCRed photoresist resulted in an emission filter with the desired blue and green absorption and red emission characteristics. In order to optimize the S/N ratio and limit detector saturation, the optimal driving currents for green and blue excitation for a given emission filter were determined empirically. We found that ratios of blue vs. green driving currents that generated Fblue/Fgreen ratio of ~0.3 in the absence of cells yielded a similar magnitude of Fblue and Fgreen in the presence of stained cells. Typically, driving currents of 3500-3800 mA for green illumination and 1700-1900 mA for blue illumination resulted in both high S/N ratio and Fblue and Fgreen magnitudes similar enough to avoid detector saturation.

Blue and green channel separation from raw composite signal

Since the LED excitation source in the dual flash mode oscillates between green and blue light, this results in an oscillation of the baseline fluorescence levels in the recorded composite signal between the Fgreen and Fblue values (as evident in the amplitude of the raw signal in Fig. S2A, left). The ΔFgreen change in green fluorescence level (representing Cai signal) and the ΔFblue change in blue fluorescence level (representing Vm + scaled Cai signal) are superimposed on their respective F values and are recognized in the top and bottom envelopes of the composite dual flash signal (Fig. S2A,right). To obtain final Cai and Vm signals, the ΔFgreen and ΔFblue traces at each recording site were first extracted from the raw composite signal (Fig. S2B) and then corrected for the optical crosstalk using the algorithm presented in the main text.

Stability of ΔFbluexrhod/ΔFgreenxrhod ratio

The correction algorithm used to remove optical crosstalk from the blue channel assumes that the ratio of ΔFblue to ΔFgreen for X-Rhod-1 stained cells remains constant throughout the experiment. This was validated by applying driving currents of 1700-1900 mA and 3500-3800 mA for the blue and green LEDs, respectively, in X-Rhod-1 stained isotropic monolayers over 10 consecutive recordings within a 14 min period (Fig. S3). The mean ΔFbluexrhod/ΔFgreenxrhod ratio was found to remain constant and deviate from the initial value by ~8.9% (Fig. S3A). These deviations also remained relatively uniform among different recording sites in the monolayer (with an average site-site variation of ~6.7%, Fig. S3B). The ΔFbluexrhod/ΔFgreenxrhod ratio also scaled linearly with changes in the ratio of current amplitudes used to drive blue and green LED excitation (not shown).

Effect of crosstalk correction on the shape of the action potential

Subtraction of crosstalk (scaled Cai signal) from the biphasic ΔFblue trace yielded both increased depolarization and decreased hyperpolarization phase of the resulting Vm signal. For the ideally removed crosstalk, the resulting action potential trace had no hyperpolarization phase while its depolarization phase (representing action potential amplitude) was increased by 69.8% relative to that before the crosstalk correction. In a portion of recording sites where crosstalk removal was inadequate, the tail of the action potential showed small hyperpolarization phase with an amplitude that averaged 16.7% of the action potential amplitude. The crosstalk correction factor (ΔFbluexrhod/ΔFgreenxrhod ratio) at these recording sites could be adjusted in the analysis algorithm to remove the hyperpolarization phase from the final Vm signal.

Figure S1. Hardware architecture for the dual mapping system. A graphical user interface running in a host PC was used to control two acquisition boards, which were synchronized to simultaneously control data acquisition by the PDA and illumination using the LED matrix.

Figure S2. Illustration of blue and green channel separation. A) Representative example of a 9.7 kHz raw signal recorded in one PDA channel when monolayers were paced at 2 Hz and illuminated with green and blue LEDs pulsed at opposite phase. The green channel is visible at the top envelope of the raw signal and the blue channel is visible at the bottom envelope. Right, the close-ups in green and blue borders show raw signal envelopes recorded during one propagated pulse. B) Assigning the data points from the top and bottom of the raw signal to the green and blue channel, respectively, produced the separated green and blue channels at the sampling rate of 488.3 Hz. Subsequently, crosstalk correction was applied as described in the main text to obtain the Vm signal from the blue channel, which resulted in the removal of the artificial hyperpolarization phase at the end of the action potential.

Figure S3. Stability of the ΔFblue/ΔFgreen ratio for X-Rhod-1 dye used in removal of optical crosstalk from the Vm signal. Signals were recorded in X-Rhod-1 stained monolayers (n = 3) paced at 2 Hz every 1.5 minutes over a period of 14 minutes. The amplitude of the response to blue (ΔFblue) or green (ΔFgreen) constant LED illumination and the absolute percent change in ΔFblue/ΔFgreen ratio from that in the first recording (t = 0 min) was determined for each recording site. The mean ratio change in the monolayer was determined by averaging ratio changes from all monolayer recording sites. The site-site variability in the ratio change in the monolayer was determined by finding the standard deviation of ratio changes from all monolayer recording sites. Panel A shows mean percent ratio change averaged over 3 monolayers. Panel B shows site-site variability in the percent ratio change averaged over 3 monolayers.

References

1. McSpadden, L. C., R. D. Kirkton, and N. Bursac. 2009. Electrotonic loading of anisotropic cardiac monolayers by unexcitable cells depends on connexin type and expression level. Am J Physiol Cell Physiol 297:C339-351.

2. Pedrotty, D. M., R. Y. Klinger, N. Badie, S. Hinds, A. Kardashian, and N. Bursac. 2008. Structural coupling of cardiomyocytes and noncardiomyocytes: quantitative comparisons using a novel micropatterned cell pair assay. Am J Physiol Heart Circ Physiol 295:H390-400.