| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
TRANSLATIONAL PHYSIOLOGY
Bioengineering Institute, The University of Auckland, Auckland, New Zealand
Submitted 18 June 2003 ; accepted in final form 27 April 2004
 |
ABSTRACT |
|---|
 TOP ABSTRACT METHODSRESULTSDISCUSSIONGRANTSREFERENCES |
|---|
fluorescence; cardiac electrical mapping
Despite these advantages, a number of problems are associated with optical measurement of cardiac transmembrane potential. Fluorescence emission depends on effective dye concentration, which is influenced by photobleaching. In the case of transmembrane voltage-sensitive dyes, progressive internalization of the dye at the inner surface of the cell membrane may also contribute to a decline in effective concentration (14). Most problematic, contraction and associated heart wall motion affect the fluorescence intensity detected. This phenomenon is generally referred to as "motion artifact," and we will follow that convention here. Surprisingly, the characteristics of this artifact are very similar for intramural measurements, where the optical probe moves with the contracting heart (13), and surface measurements, where it does not (14, 20). Although translation of the myocardium with respect to the optical system contributes to motion artifact, cell membrane deformation and redistribution of intramural vascular volume associated with contraction may also play a role. Moreover, absorption, scattering, and diffraction of excitation and fluorescent light by the tissue will vary throughout the cardiac cycle, and these processes are wavelength dependent (2).
Fast-response potentiometric probes such as di-4-ANEPPS and di-8-ANEPPS typically have a fractional fluorescence in the range ±0–0.1 per 100 mV (10, 17). Two approaches have been used to prevent motion artifact from masking this signal in optical mapping of cardiac electrical activation. Laurita et al. (15) stabilized the epicardial surface of the heart by pressing it gently against an imaging window. More commonly, mechanical activity is suppressed pharmacologically using 2,3-butanedione monoxime (2,3-BDM). Although this agent reversibly blocks cardiac contraction (12, 18), it also has significant effects on cardiac electrophysiology. 2,3-BDM reduces the transient outward current (Ito) and depresses the delayed rectifier (IK) and background (IK1) currents in a variety of different species (16). As a result, 2,3-BDM reduces action potential duration (APD) in a dose-dependent fashion from concentrations as low as 5 mM. Physical stabilization is difficult to employ and is probably less effective for intramural fluorometric measurements than for epicardial surface recordings. Therefore, methods that enable action potentials to be retrieved reliably in the absence of 2,3-BDM or at low concentrations of this agent may be of considerable utility in the study of electrical repolarization in the heart.
Ratio imaging has been used to minimize the effects of factors such as photobleaching and nonuniform dye distribution in numerous fluorometric applications (1, 2), including optical recording of membrane potential (4, 13, 14). The fluorochrome di-4-ANEPPS is classified as a ratiometric dye, because its emission spectrum shifts reversibly to shorter wavelengths with membrane depolarization. It is possible to identify a wavelength, i.e., the isosbestic wavelength, at which there is no change in emission with altered membrane potential. Fluorescent emission due to membrane depolarization is reduced at wavelengths longer than the isosbestic wavelength and increased at shorter wavelengths. Within this context, ratiometry is applied as follows: Fluorescent emission is simultaneously recorded in two wavelength windows (above and below the isosbestic wavelength), background fluorescence (recorded previously) is subtracted from both signals, and a ratio is formed. This procedure preserves signal components that differ across the two wavelengths but attenuates components that are similar in both. Ratiometry provides a simple and robust technique for minimizing the effects of photobleaching and also reduces the extent of motion artifact (2, 13, 14). However, it is often not possible to recover voltage-dependent signals using ratiometry when wall motion is substantial (13). Brandes and co-workers (2) demonstrated that signal components due to motion were not simply proportionate in short and long wavelengths and argued that motion artifact, therefore, could not be completely removed with ratiometry. An alternative, dual-wavelength approach was previously reported by Rohr and Kucera (20). In this case, the fractional fluorescence was estimated in long- and short-wavelength signals, and the former was subtracted from the latter. This approach does not correct for photobleaching-induced decay and will not fully remove motion artifact if it has different amplitudes in short and long wavelengths.
This study outlines a modified subtraction method, in which the decay envelope in short- and long-wavelength channels is characterized nonparametrically and separately removed from both. Motion artifact is then identified as the time-varying signal component common to both decay-corrected signals and subtracted. Results obtained in this way are compared with those yielded by ratiometry for intramural fluorescence signals recorded at different 2,3-BDM concentrations and various levels of motion artifact. The dual-wavelength subtraction technique provides significantly better results than conventional ratiometry in the presence of motion artifact and may enable optical studies of cardiac electrical activation to be carried out at lower concentrations of 2,3-BDM than has previously been possible.
|
METHODS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONGRANTSREFERENCES |
|---|
|
All surgical procedures were approved by the Animal Ethics Committee of The University of Auckland. Four pigs weighing 20–30 kg were anesthetized initially with tiletamine-zolazepam (Zoletil, 10 mg/kg im), and anesthesia was maintained with halothane (2–5%) in oxygen. A catheter was inserted into the left internal carotid artery for arterial pressure monitoring, and the left external jugular vein was cannulated to provide intravenous access. The heart was exposed via a median sternotomy and a left lateral thoracotomy. After the pericardium was opened, the left anterior descending coronary artery (LAD) was dissected circumferentially close to its origin, and a snare was positioned around it. An epoxy needle probe (12 mm long, 650 µm OD) was introduced into the anterior free wall of the left ventricle (LV). The probe contained three bipolar pacing electrodes (70-µm silver wire, 1-mm separation) spaced 
3 mm apart along the shaft. A silver-plated reference electrode (1 cm2) was sutured to the adventitia of the pulmonary artery. Heparin (100 IU/kg) was administered, and 250–350 ml of cooled (4°C) physiological saline solution were infused rapidly via the jugular vein. The terminals of a 9-V battery were briefly touched on the epicardial surface of the ventricles to induce fibrillation, and the heart was excised and immersed in cooled saline solution.
The heart was then mounted in a Langendorff perfusion apparatus. A short cannula (12 mm OD) was inserted into the aorta, positioned above the coronary ostia, and secured, while the heart remained immersed in cold saline. An 18-gauge catheter, passed through the perfusion cannula, was introduced into the LAD via the left coronary ostium and advanced through the snare (see above). A modified Tyrode solution (119 mM NaCl, 4.0 mM KCl, 2.5 mM CaCl2, 1.2 mM MgCl2, 20 mM NaHCO3, 1.0 mM NaH2PO4, and 12 mM dextrose; osmolarity 291 mosmol/l) was supplied to the aortic cannula from a reservoir 2 m above it. The perfusate was maintained at 37°C, bubbled with 95% O2-5% CO2, and passed through a bubble trap. The coronary effluent could be returned to the reservoir via a micropore (0.8 µm) filter. Perfusate flow rate was monitored with an electromagnetic flowmeter (model SP2202, Gould), and perfusion pressure was measured with a strain gauge pressure transducer (model P23DB, Gould Statham).
The heart was initially perfused for several minutes in nonrecirculating mode to clear the coronary circulation (1 liter of Tyrode solution was exhausted to waste). Subsequently, the perfusate was recirculated, and the heart was defibrillated (15 J shock; model 640, Physio Control). As soon as vigorous contraction was initiated, an extracellular electrode was attached to the epicardial surface of the LV to measure the cardiac electrogram (CEG; with respect to the pulmonary artery reference). 2,3-BDM was added to the perfusate to bring it to an initial concentration of 7.5 mM. The snare was briefly tightened around the LAD catheter while 10 ml of 50 µM di-4-ANEPPS (Molecular Probes) were infused. The snare was then released, and the catheter was retracted. An optrode was positioned in the anterior free wall of the LV at the center of the dyed region adjacent to the pacing probe. A purpose-built, constant-current stimulator was used to pace the heart with typical parameters as follows: 2-ms duration, current at 1.5 times capture threshold, and rate of 1–1.8 Hz. In each study, 2,3-BDM concentration was progressively increased from 7.5 to 12.5 mM, and data were acquired in sinus rhythm and during LV pacing at different intramural sites.
Data acquisition. Data were recorded and stored with a multichannel acquisition system (UnEmap, Auckland UniServices). Perfusate flow, perfusion pressure, and optrode and laser intensity signals were low-pass filtered (500 Hz) and digitized with 16-bit resolution. The CEG was band limited (0.05–500 Hz) and digitized with 12-bit resolution. The sampling rate was 1 kHz per channel. Background signals were collected over 2–10 s for all optical channels at the start of each study, with the optrode in air and in a darkened environment.
Simple model of motion artifact.
For most fluorescence imaging systems, the light collected by the photodetectors is integrated across a wavelength window determined by the optical system (dichroic mirrors and filters). The light detected over a particular wavelength window i, Si(t), consists of two components: one due to excitation of the fluorescent dye and the other to a background signal associated with the autofluorescence of the optical system. Both are assumed to be linearly dependent on the intensity of the excitation source I(t). Thus
 | (1) |
Membrane potential-sensitive fluorescent dye emission in the heart is modulated by the cardiac action potential but is also affected by cardiac wall motion. Moreover, there is a slow reduction in fluorescence intensity due to, for example, dye photobleaching. In Eq. 2, each of these factors is lumped as a separate time-dependent function
 | (2) |
 | (3) |
 | (4) |
Ci(t) describes the slow, wavelength-dependent reduction of fluorescence emission with maintained excitation.
Analysis of ratiometry.
We use the simple phenomenological model described above to analyze the process of ratiometry. If we combine Eqs. 2–4, the recorded signal can be expressed as
 | (5) |
 | (6) |
If the numerator and denominator of this expression are expanded and a Taylor series approximation is used to further expand the denominator it can be shown that
 | (7) |
If Hi(t) = kHj(t) and if bi = bj or if the mechanical term M(t) is very small, the ratioed signal reduces to
 | (8) |
A novel subtraction technique.
We use the simple phenomenological model to analyze an alternate technique developed to overcome these deficiencies. The first step is to characterize the decay envelope Hi(t) in the fluorescence signal Si so that it can be removed. The corrected fluorescence signal SI*
 | (9) |
 | (10) |
This analysis suggests that it should be possible to remove motion artifact if a common contraction-dependent component can be identified and matched in decay-corrected short- and long-wavelength signals. Removal of the decay envelope is intrinsic to the approach. The analysis above indicates that a weighted subtraction method should be more effective than direct ratioing in removing motion artifact.
Data analysis. Short- and long-wavelength signals for all optrode channels were corrected by subtracting the average value of the background recorded for each at the beginning of the study (see Data acquisition). Optical and electronic noise in these signals was concentrated at 50 Hz and its harmonics, and indirect filtering was used to minimize each noise spike across a 1-Hz frequency window.
Ratio signals for each channel were then obtained from corrected short- and long-wavelength data and averaged over 8–16 cycles. Signals were synchronized across successive cycles by aligning the maximum first derivative.
Subtraction was also carried out for each channel, and the first step in this process was to characterize the decay envelopes for corrected long- and short-wavelength signals. Characteristic data sets were constructed by identifying a single time series across all signals with one point per cardiac cycle, each phase locked to a common reference time within the cycle. Signal means were then evaluated over 20 ms at each point. Sequential fitting of quadratic functions to these data provided piecewise continuous estimates of signal decay envelopes, which were used to linearize corrected signals as outlined in Eq. 9. Linearized long- and short-wavelength signals for each channel were averaged over 8–16 cycles with individual cycles synchronized with respect to the time points identified previously. The relative magnitudes of motion components in averaged long- and short-wavelength signals were then estimated for all channels. Two complementary approaches were employed here. Where modulation due to cardiac electrical activation could be distinguished, the variation of long- and short-wavelength signals was matched across the diastolic interval before the upstroke of the action potential. Alternately, when modulation due to cardiac electrical activation was masked by motion artifact, long- and short-wavelength signals were matched across the complete cardiac cycle. This step provided estimates of the ratio b2/b1 (see Eq. 10) and enabled the subtraction 
(t) – 
(t) x b2/b1 to be evaluated across all channels.
To evaluate APD, the onset and termination of activation were estimated from maximum first and second derivatives of the transmembrane potential during CEG R and T waves, respectively (8). This index is identified as APDderiv. As an alternate measure, APD50 was estimated as the time that transmembrane potential exceeded 50% of the maximum deviation from resting levels during activation.
All data analysis procedures were carried out using automated applications written in the LabVIEW programming language (National Instruments).
|
RESULTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONGRANTSREFERENCES |
|---|
|
|
|
Typical examples of the effectiveness of ratiometry in removing motion artifact are given in Fig. 5 for different levels of mechanical activity in the same experiment. Ratiometry successfully recovered the action potential in Fig. 5B when mechanical artifact was moderate. It retrieved the onset of activation in Fig. 5C when mechanical artifact was substantial but did not preserve action potential morphology in the plateau or during repolarization. On the other hand, the subtraction recovered the action potential in all cases. As stated above, we grouped records in which mechanical activity was evident into three distinct categories: 1) negligible, 2) moderate, and 3) substantial. The ratio b2/b1 that best matched the mechanical components in short- and long-wavelength signals (t) and (t) ranged from 0.2 to 7 across all three categories. We found that ratiometry and subtraction were equally effective in correcting for negligible motion artifact. On the other hand, subtraction was more effective in recovering action potentials in 58 of 187 of the records in which motion artifact was judged to be moderate. Finally, subtraction was markedly better at retrieving action potentials in 31 of 38 of the records in the final category, where motion artifact substantially masked the modulation due to electrical activity.
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONGRANTSREFERENCES |
|---|
Characteristics of fluorescence emission signals. The results presented here clearly show the separate contributions to fluorescence emission in hearts stained with membrane potential-sensitive dyes of signal decay, membrane potential changes, and mechanical activity. The behavior of each component has been reported elsewhere (2, 14, 17), and our results are qualitatively consistent with those observations. Quantitative discrepancies are most likely due to differences in the wavelength windows over which fluorescence signals are integrated and the fact that our recordings are made with intramural probes fixed at a particular location in the LV free wall, rather than from the epicardial surface of the heart.
The decay in fluorescence emission of membrane-bound dyes such as di-4-ANEPPS is predominantly due to light-induced destruction of fluorophore molecules but may also be influenced by sequestration of the dye at the internal surface of the cell membrane and by dye washout. However, it appears that this process cannot be fully described as a simple reduction in effective fluorophore concentration. If this were the case, the fractional decrease in emission would be the same for all wavelengths, but we have found that the decay in fluorescence emission occurs at a greater rate for short than for long wavelengths. Although the difference was relatively small (1–4% over 10 s), it was highly consistent. Matching findings have been reported by Knisley and co-workers (14), who used techniques very similar to ours: di-4-ANEPPS excited by the 488-nm line of an argon ion laser. They argued that photobleaching could affect the kinetics of the remaining fluorophore by altering the environment to which its extracellular or intracellular domains are exposed or by increasing internalization of the dye. One consequence of the different decay rates for long and short wavelengths is that there is a component of photobleaching-induced drift that cannot be removed by ratiometry. On the other hand, the difference is quite small, and it has negligible impact on our subtraction technique, because the decay in short and long wavelengths is independently characterized and removed. Our results confirm that decay can be represented accurately across short- and long-wavelength windows using nonparametric techniques. Piecewise quadratic fitting was successfully employed in this study, but a range of other approaches could have been used. Although the decay envelope is not well fit as a single exponential, there seems little doubt that it could be characterized by an appropriate multiexponential function.
The response of di-4-ANEPPS to changes in membrane potential has been well characterized (17). An increase in membrane potential produces a leftward shift of fluorescence spectrum with respect to the isosbestic point (600 nm). Thus the emission detected in the short-wavelength range (520–600 nm) increases, whereas that in the long-wavelength range (>600 nm) decreases. At rest, the fraction of fluorescent emission power in the short-wavelength range is less than that in the long-wavelength range. The modulation of fluorescence emission by the cardiac action potential in long- and short-wavelength signals observed in this study is consistent with these expectations and with previously published results (14).
Motion artifact. At least four research groups have attempted to reduce motion artifact in cardiac fluorescence studies by recording fluorescence emission in two wavelength bands (2, 13, 14, 20). Brandes et al. (2) stained isolated perfused rat hearts with the reference fluorophore diphenyl hexatrine and recorded the changes in fluorescence emission associated with contraction over 10-nm wavelength bands centered at 395 and 454 nm, respectively. The remaining groups used the membrane potential-sensitive dyes di-4-ANEPPS and di-8-ANEPPS and measured emission in different short-wavelength (510–600 nm) and long-wavelength (>600 nm) bands. Rohr and Kucera (20) studied patterned monolayer cultures of ventricular myocytes stained with di-8-ANEPPS during electrical stimulation, and Knisley et al. (14) and Hooks et al. (13) collected dual-wavelength fluorescent emission from isolated beating rabbit hearts. The imaging systems employed addressed fixed spatial locations (14, 20) or employed fiber-optic probes that move with the heart surface (2, 13). Widely different mechanical constraints were imposed on the preparations between and, in some cases, within studies.
These investigations demonstrate that there is a component of cardiac fluorescent emission that is graded with respect to the extent of contraction and that the temporal variation of this component is near identical in long and short wavelengths. The results of the present study are wholly consistent with these findings. However, although the data presented by Rohr and Kucera (20) and Knisley et al. (14) indicate that the fractional changes in fluorescence due to contraction are the same in long and short wavelengths, Brandes et al. (2) and, by inference, Hooks et al. (14) showed that there were differences in the normalized motion amplitude at different wavelengths.
A variety of factors could give rise to motion artifact. For surface mapping studies, where the fluorescence imaging system is fixed in space, the myocardial region addressed by the detection system will change with heart wall motion, and the intensity of the signal detected will reflect the effective dye concentration in the region "seen" at any instant. Thus inhomogeneity of staining and spatial variation in the extent of photobleaching will give rise to motion artifact. Similar motion-related effects must occur for intramural optical probes, which shift within the ventricular wall as it thickens during systole. Local myocardial deformation will also directly affect fluorescence measurements. During contraction, the heart wall shortens circumferentially and longitudinally and thickens transmurally, while blood is redistributed in the intramural coronary circulation. As a result, the conformation of the cell membranes in the tissue volume to which excitation light is delivered will inevitably change throughout the cardiac cycle, giving rise to changes in the distribution and history of the dye molecules from which fluorescence emission is collected. Finally, the scattering of light by myocardium and its absorption by tissue and dye are frequency dependent and are influenced by the dimensional changes in the contractile lattice that occur throughout the cardiac cycle. The relative contribution of the different factors listed above to motion artifact for intramural optical probes or surface mapping is by no means clear.
Simple model of motion artifact. We have presented a simple phenomenological "model," in which the effects of decay, electrical activation, and mechanical activity on the fluorescence emission of fast potential-sensitive dyes are represented as independent components. The fluorescence emission at a given wavelength due to electrical activation is generally described as a linear function of the membrane potential (17), and we adopt that convention here. The component due to mechanical activity is similarly represented as a linear function of a mechanical waveform that is the same for all wavelengths. This approach was previously used by Brandes and colleagues (2) to analyze motion artifact and accommodates the fact that although the fractional fluorescence changes due to mechanical activity may be different in short- and long-wavelength signals, the shape of the associated waveform appears to be identical across wavelengths. Although these descriptions are empirical and certainly simplistic, the model provides a basis for analysis and comparison of ratiometry and subtraction methods. For instance, we have demonstrated that ratiometry cannot fully remove decay and motion artifact, unless the normalized temporal variation of the emission associated with each is identical in long and short wavelengths. Moreover, if contractile activity is substantial, residual motion artifact will be exacerbated by the nonlinearity inherent in ratiometry. It is also evident that simply subtracting the fractional fluorescence in long and short wavelengths does not alter the fluorescence decay and will not fully remove motion artifact if its amplitude is not identical in both wavelength bands. Our results and those of others (2) indicate that there are differences in normalized motion amplitude at different wavelengths. Even if this were not the case, effective application of ratiometry relies on precise estimation of background fluorescence in both wavelengths, and this information is also required for accurate determination of fractional fluorescence. The weighted subtraction approach employed here removes decay independently and matches the common-mode motion component in both wavelength bands. This process is not dependent on estimates of background fluorescence and does not require the motion component to be the same for all frequencies. The utility of the approach is reflected by its effectiveness in recovering action potentials in the presence of substantial motion.
Comparison of dual-wavelength ratiometry and subtraction methods. Our results indicate that ratiometry and subtraction are equally effective in removing motion artifact when the extent of contractile activity is relatively small. This is consistent with the analysis discussed immediately above, which demonstrates that ratiometry reduces to simple (nonweighted) subtraction under these circumstances.
When contractile activity was more extensive, ratiometry identified the onset of activation, but the ratioed signals often still contained large residual motion artifacts (13). In general, substantial further reduction of motion artifact was possible using weighted subtraction. Comparison of APD estimates obtained with the two techniques provides a valid, if indirect, means of establishing the reliability of the action potentials recovered. Indexes such as APD50, where activation and repolarization times are identified by the signal crossing a threshold that is a fraction of the deviation during depolarization, are highly sensitive to baseline drift and motion artifact. On the other hand, Efimov and co-workers (8) showed that the maximum second derivative of the action potential provides a unique index of repolarization that is readily recovered from action potentials containing substantial drift and motion artifact. The close correspondence of estimates of APDderiv for dual-wavelength records corrected using ratiometry and subtraction indicates that characteristic features of activation and repolarization are preserved with both techniques. However, the greater scatter of APD50 is consistent with the fact that residual decay and motion artifact are less effectively removed by ratiometry than by subtraction.
Limitations of the study. It is important to note that the weighted subtraction method presented here has been tested using dual-wavelength emission signals obtained with an intramural fiber-optic probe that moves with the heart wall. However, we see no reason why it should not work equally well for dual-wavelength surface imaging. As noted above, the characteristics of dual-wavelength motion artifact recorded with imaging systems that address fixed spatial locations and fiber-optic probes that move with the heart surface are remarkably similar. We have digitized short- and long-wavelength optical signals recorded from the surface of the heart and published by Knisley et al. (14) and have shown that our technique removes motion artifact in these cases. Finally, it must be acknowledged that although dual-wavelength correction techniques may eliminate signal components associated with contraction, they do not account for the fact that translational motion will "cause different phases of an action potential to be recorded from different groups of cells" (14).
We conclude that ratiometry provides an efficient and robust means of removing motion artifact when mechanical activity is moderate. On the other hand, our modified subtraction technique recovers the action potential faithfully in the presence of significant mechanical activity. The approach should make it possible to employ fluorescence imaging to map electrical activity in the absence of pharmacological uncoupling agents and, hence, obtain more reliable information about cardiac repolarization.
|
GRANTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONGRANTSREFERENCES |
|---|
|
FOOTNOTES |
|---|
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
|
REFERENCES |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONGRANTSREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  N. H. Brown, H. M. Dobrovolny, D. J. Gauthier, and P. D. Wolf A Fiber-Based Ratiometric Optical Cardiac Mapping Channel Using a Diffraction Grating and Split Detector Biophys. J., July 1, 2007; 93(1): 254 - 263. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. Nishimura, Y. Kawai, T. Nakajima, Y. Hosoya, H. Fujita, M. Katoh, H. Yamashita, R. Nagai, and S. Sugiura Membrane potential of rat ventricular myocytes responds to axial stretch in phase, amplitude and speed-dependent manners Cardiovasc Res, December 1, 2006; 72(3): 403 - 411. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Visit Other APS Journals Online |
