Cardiac pacing using transmural multi-LED probes in channelrhodopsin-expressing mouse hearts

Abstract

Optogenetics enables cell-type specific monitoring and actuation via light-activated proteins. In cardiac research, expressing light-activated depolarising ion channels in cardiomyocytes allows optical pacing and defibrillation. Previous studies largely relied on epicardial illumination. Light penetration through the myocardium is however problematic when moving to larger animals and humans. To overcome this limitation, we assessed the utility of an implantable multi light-emitting diode (LED) optical probe (IMLOP) for intramural pacing of mouse hearts expressing cardiac-specific channelrhodopsin-2 (ChR2).

Here we demonstrated that IMLOP insertion needs approximately 20 mN of force, limiting possible damage from excessive loads applied during implantation. Histological sections confirmed the confined nature of tissue damage during acute use. The temperature change of the surrounding tissue was below 1 K during LED operation, rendering the probe safe for use in situ. This was confirmed in control experiments where no effect on cardiac action potential conduction was observed even when using stimulation parameters twenty-fold greater than required for pacing.

In situ experiments on ChR2-expressing mouse hearts demonstrated that optical stimulation is possible with light intensities as low as 700 μW/mm²; although stable pacing requires higher intensities. When pacing with a single LED, rheobase and chronaxie values were 13.3 mW/mm2 ± 0.9 mW/mm² and 3 ms ± 0.6 ms, respectively. When doubling the stimulated volume the rheobase decreased significantly (6.5 mW/mm2 ± 0.9 mW/mm²).

We have demonstrated IMLOP-based intramural optical pacing of the heart. Probes cause locally constrained tissue damage in the acute setting and require low light intensities for pacing. Further development is necessary to assess effects of chronic implantation.

1. Introduction

Optogenetics is a powerful technique that allows one to optically investigate and actuate genetically-targeted cell types in native tissue. Most electrophysiological actuation studies are based on the use of channelrhodopsin-2 (ChR2), a blue light-gated non-selective cation channel that can be used for light-mediated depolarization of cells (Nagel et al., 2003). While initially focused on neural applications, ChR2 also has been applied to optically trigger excitable cells in the heart (Bruegmann et al., 2010). Optogenetic studies range from basic science to translational applications with findings from optogenetic-based research now used to form clinical hypotheses (Ferenczi and Deisseroth, 2016; Keppeler et al., 2018). First clinical trials of optogenetic treatments to restore vision are currently underway (GenSight Biologics, 2017).

In the cardiac field, a major focus of optogenetic research is optical pacing and defibrillation (Quiñonez Uribe et al., 2018; Bruegmann et al., 2017; Nyns et al., 2017; Ayub et al., 2018; Crocini et al., 2016). These approaches would benefit from implantable illumination devices, in combination with light-activated ion channel expression in the cell-type of interest, to implement pain-free modulation of cardiac activity in vivo. Conceptually, this is a very attractive alternative to conventional cardiac rhythm management devices, such as cardiac pacemakers. Optical pacing of ChR2-expressing cardiomyocytes results in channel opening in the cell membrane, primarily leading to an influx of protons and sodium, which depolarises the membrane, thereby activating voltage-gated Na⁺ channels and initiating an action potential (AP). Optically stimulated AP closely mimic native excitable cell activation, with the effect of ChR2 opening resembling gap-junction mediated membrane depolarization. Therefore, optical pacing provides a more physiological stimulus than classical electrical pacing, and it enables lower energy consumption as it directly acts on cells rather than altering the electrical field of the extracellular space for cellular excitation.

Both ex vivo and in situ studies on mouse and rat hearts expressing ChR2 have demonstrated that optical pacing is feasible (Bruegmann et al., 2010; Nussinovitch and Gepstein, 2015; Zaglia et al., 2015; Johnston et al., 2017). Appropriate optical stimulation becomes significantly more complex when moving from mice to humans, not only because of challenges in targeted expression of optogenetic actuators, but also due to the difference in cardiac dimensions (the external volume including chambers occupied by a mouse heart is approximately 0.1 cm³ compared to 270 cm³ for a human heart). ChR2 is most sensitive to blue light, which, when illuminating cardiac tissue, drops to 50% of the incident light intensity within 0.5 mm (Quinn et al., 2012). This low tissue penetration means that, even in the mouse model, fully transmural activation of ChR2 is unlikely to occur upon epicardial illumination. While conventional pacing only requires illumination of a small tissue volume (in the healthy mouse heart only approximately 1300 cardiomyocytes need to be activated to trigger contraction, and this drops to a few hundred if targeting cells in the conductive pathway) (Zaglia et al., 2015). It is likely that in the diseased or fibrotic heart a greater tissue volume must be illuminated due to both a decreased cell density and increased number of non-myocytes coupled to cardiomyocytes reducing excitability. As transmural heterogeneity and/or re-entry form a common basis for arrhythmias, fully transmural illumination would be preferred to limit the risk of arrhythmia development (Antzelevitch and Burashnikov, 2011; Valderrábano et al., 2001).

A possible approach for transmural light delivery, at least in the experimental setting, is through the use of light sources that physically penetrate the tissue. Invasive delivery of light is typically achieved using optical glass fibres coupled to light sources such as lasers or high-power light-emitting diode (LED) chips. This results in a single illumination spot per fibre, although approaches using conically shaped optical fibre tips with micro-patterned optical windows enabled multiple illumination spots for optical brain stimulation (Pisanello et al., 2014). The mechanical stiffness of such fibres is a potential limitation when moving from neuroscience towards cardiac applications, where cardiac tissue stiffness, rhythmic contraction, and mechano-sensitivity pose additional challenges.

Advanced microfabrication techniques overcome these limitations by enabling the creation of significantly more compact and flexible optical tools for delivering light to targeted tissue sites. One of the first devices in this category utilised integrated waveguides, coupled to laser diode chips at the base of the probes, resulting in a highly localised illumination capability (Kampasi et al., 2016; Schwaerzle et al., 2017). More recently, LED chips have been integrated directly onto implantable probes, initially as a single LED at the probe tip (Cao et al., 2013), before multiple LED chips were implemented along a probe shank (Schwaerzle et al., 2016). Furthermore, thin-film μLED were integrated on silicon (Si) (Wu et al., 2015; Scharf et al., 2016; Ayub et al., 2016) or polymer substrates (Gossler et al., 2014; Klein et al., 2018; Jeong et al., 2015) used for brain research (Jeong et al., 2015) as well as the optical stimulation of the cochlea (Gossler et al., 2014; Klein et al., 2018). The latter probe type raises the possibility of pacing the heart transmurally while minimally affecting cardiac mechanics and electrophysiology due to the low mechanical rigidity of the material.

Here, we provide a first test of an implantable multi-LED optical probe (IMLOP) for modulating cardiac electrical activity. We measure the force required to implant the probes, evaluate the tissue damage resulting from this process, and assess how the probes alter underlying cardiac electrophysiology. We investigate the temperature increase in the IMLOP environment when operating the probes in bench tests using a tissue phantom based on agarose gel as well as an isolated mouse heart. We further examine the utility of the IMLOP to pace the heart while assessing the radiant fluxes and light intensities needed. These results demonstrate the ability to use penetrating probes, as already utilised in neural research, in cardiac optogenetic studies. This serves as an initial feasibility study for potential applications of transmural optical pacing, by assessing how implantable probes alter cardiac function, both by their presence and by light-induced alterations of cardiac electrophysiology.

2. Materials and methods

2.1. Probe design

IMLOP microfabrication has been described previously and is presented here as a brief overview (Ayub et al, 2017, 2018). The optical probes incorporate commercially available 270 × 220 μm LED chips (TR2227, Cree, Durham, USA) that are flip-chip bonded to flexible polyimide (PI) substrates. Up to ten LED chips were arranged as a linear array on the probe, at a centre-to-centre distance of 300 μm. A 10-μm-thin flexible substrate has been chosen in order to minimise the device cross-section (though this is defined, to a large part, by the thickness of 50 μm of the LED chips). A custom-designed micro-machined Si stiffening structure surrounds the LED chips and imparts the rigidity needed for successful tissue penetration to the otherwise highly flexible LED probe. This stiffening structure has been updated from previous work to include apertures confining the emitted light. This study uses circular apertures with diameters of 50 μm, 75 μm, 100 μm, or 175 μm, and probes with lengths of 5 mm and 8 mm. The semi-transparent PI substrate allows light emission towards the rear of the probe. This was reduced by embedding an optical shielding layer, made of platinum, in the PI substrate. In a preliminary report, the possibility of pacing ChR2-expressing murine hearts with a single-sided IMLOP prototype was established (Ayub et al., 2018).

To enable tissue illumination on both sides of the IMLOP, one can alternate apertures at the front with openings in the rear optical shielding layer (Ayub et al, 2017, 2018). This type of bi-directional illumination results in an increased inter-LED distance on each side. Also, given that light exits at the rear through the semi-transparent PI substrate with metal lines partially covering this area, the same aperture size would result in a reduced photon flux. This can either be compensated by increasing the aperture size at the rear, or the input power of the LED.

To avoid these constraints, we assembled two probes back-to-back, which results in identical illumination options on both sides of the double-sided IMLOP. This is linked to an increase in probe thickness from 70 μm to 165 μm. The two single-sided probes were glued together using the fluoropolymer CYTOP (AGC Chemicals, Tokyo, Japan) and a silicone mould to facilitate alignment. After CYTOP curing, the assembled double-sided IMLOP was encapsulated in silicone (RT 604, Wacker Chemie AG, Munich, Germany). The assembled device is shown in Fig. 1a. The schematic cross-section in Fig. 1b illustrates the layer stack of the PI substrates and the apertures integrated in the Si stiffener. The IMLOP is electrically interfaced to an external circuitry using highly flexible PI cables (Ayub et al., 2017), minimizing the mechanical force imparted on the beating heart.

Fig. 1.

Assembled double-sided IMLOP. (a) Photograph illustrating a 5-mm-long double-sided IMLOP with circular apertures (50 μm diameter). The total thickness and width of the optical probe are 165 μm and 275 μm, respectively. LED chips are interfaced via flexible PI cables shown on the left. Scale bar 500 μm. (b) Schematic cross-section along the probe shank illustrating two probes glued back-to-back to allow bi-directional illumination (not to scale).

2.2. Technical measurement setups

Optical characterization: The optical spectrum and radiant flux were measured using an integrating sphere combined with a spectrometer (Ocean Optics ISP-50-IUSB, Dunedin, USA). Each LED on the probe was operated by applying a defined current, pulsed at a frequency of 30 kHz and 10% duty cycle. In order to assess the absorption of light by cardiac tissue, we covered an IMLOP with tissue slices of different thickness t_s. Slices of murine left ventricle (100 μm, 200 μm, 300 μm, 500 μm, 700 μm, 1000 μm) were cut using a vibratome (Ci 7000 smz, Campden Instruments, Loughborough, UK), as described previously (Wang et al., 2015). For these tests, the IMLOP was assembled onto a printed circuit board (PCB) for electrical interconnection to the external LED controller and the transfer of samples into the integrating sphere, as schematically illustrated in Fig. 2a. Tissue slices were positioned on the IMLOP to completely cover at least four LED apertures each with a diameter of 100 μm.

Fig. 2.

Light absorption in cardiac tissue slices. (a) Schematic of IMLOP interfaced on a printed circuit board (PCB), and introduced into an integrating sphere. The tested LED sites are covered by a cardiac tissue slice of thickness t_s (not to scale). (b) Averaged spectral distribution of emitted light for different tissues thicknesses t_s. (c) Transmittance of cardiac tissue as a function of t_s for LED currents between 7.5 and 10 mA and exponential fit.

Thermal characterization: The thermal analysis of the IMLOP uses a test device comprising a resistive temperature sensor integrated in the PI substrate directly underneath an LED. The probe stiffener and PI substrate are otherwise identical in size and material composition. The thermistor is implemented as a platinum meander using the metal of the optical shielding layer. Four-point measurement is used for precise temperature recording. The thermistor is located within 10 μm of the LED, as illustrated in Fig. 3a, which enables precise registration of short-lived temperature changes. The test probe used for temperature evaluation had 100 μm diameter apertures. The thermistor was calibrated in a temperature chamber at reference temperatures determined using a standard PT100 temperature sensor positioned next to the probe. For temperature evaluation, the test probe was inserted into agarose gel (0.6 wt.-% agarose), kept on a hotplate with a set temperature of 37 °C. The LED chip with the subjacent thermistor was operated at 10 kHz and 50% duty cycle, applying ten pulses of 10 mA with a duration of 10 ms at frequencies of 6 Hz, 10 Hz, and 15 Hz. The temperature was measured every 1 ms, for a total duration of 4 s. In order to increase the signal-to-noise ratio of the temperature evaluation tests, each measurement was repeated 100 times and averaged. The results from this agarose gel-based phantom test were validated by identical temperature measurements with an IMLOP implanted in a Langendorff-perfused heart. In addition, extended duration light pulses were applied in the isolated heart to evaluate the long-term temperature increase of the surrounding tissue. We tested in addition the potential effect of cardiac coronary vessel perfusion on the temperature evaluation; a first set of current pulses were applied with the heart being perfused, as described below, while the second set of ten pulses was applied with cardiac perfusion stopped. Data were analysed in Mathematica (Wolfram Research, Champaign, Illinois, USA) by fitting model curves to the averaged data sets.

Fig. 3.

Investigation of temperature increase during IMLOP operation. (a) Schematic illustration of the thermal test probe in agarose. (b) Averaged measurement data and fit curves of ΔT while operating the LED with ten 10-ms-long pulses at 6 Hz repetition rate (drive current 10 mA, frequency 10 kHz, duty cycle 50%). The fit curves are extracted from a model representing combined ΔT of the LED and the surrounding medium. (c) Fit curves at three pulse repetition rates (pulse number, duration, drive current, frequency and duty cycle as above) showing the calculated ΔT of the surrounding medium, i.e. the agarose gel.

Implantation force analysis: IMLOP insertion into cardiac tissue was monitored using a custom-designed setup with a force sensor (Model 31, range code: AJ, Honeywell, Columbus, OH, USA) underneath the tissue-containing dish. A motorized linear stage (M-235, Physik Instrumente, Karlsruhe, Germany) is used to drive the 8-mm-long IMLOP into the myocardium (Fig. 5a). The isolated heart was placed on a cradle restricting its lateral movement to avoid slippage during the vertical insertion of the probe. A perfusion cannula provided physiological saline solution to the heart via its coronary vasculature (containing [in mM]: 140 NaCl, 6 KCl, 1 MgCl₂, 1.8 CaCl₂, 10 glucose, 10 HEPES; pH adjusted to 7.4 at 37 °C using NaOH; see whole-heart pacing section for more detail). Following calibration of the force sensor, probes were attached to the linear stage and driven into the tissue at an insertion speed of 100 μm/s. Depending on the position relative to the heart, travel distance varied between 2 mm and 7 mm. Measurements were conducted in air at room temperature.

Fig. 5.

Force measurement during IMLOP implantation into cardiac tissue. (a) Photograph of the insertion force measurement setup, showing the perfused heart in a dish on top of a force sensor platform. A linear stage above the heart moves the 8-mm-long probe vertically into the cardiac tissue. Scale bar 10 mm. (b–d) Implantation force (black, left axis) and vertical probe displacement (orange, right axis) as a function of time; implantation depths (b) 1.7 and 3 mm, (c) 4.7 mm, and (d) 6.7 mm for (b) an arrested heart, and hearts beating at (c) 1.5 Hz, and (d) 0.5 Hz. The implantation in (b, second trial) and (d) progressed through the ventricular free wall, across the chamber, and then into the septum, while the detail graph (d’) shows individual heart beats during the force relaxation phase that occur once IMLOP movement is halted.

2.3. Isolated heart experiments

Animal model: Transgenic mice (C57BL/6J) expressing Cre-recombinase under the control of the αMHC-promotor (B6.FVB-Tg(Myh6-cre)2182 Md/J (Agah et al., 1997), αMHC-Cre) were crossed with mice transgenic for floxed Cop4 H134R-eYFP (B6; 129S-Gt(ROSA)26Sortm32(CAG-COP4*H134R/EYFP)Hze/J (Madisen et al., 2012)), giving 25% of offspring with ChR2 targeted to cardiomyocytes (αMHC-ChR2). Offspring were genotyped by PCR of genomic DNA from ear tissue samples. Experiments were performed on double-positive offspring; age-matched litter mates not expressing ChR2 formed a control group.

All animal experiments were carried out according to the guidelines stated in the Directive 2010/63/EU of the European Parliament on the protection of animals used for scientific purposes and were approved by the local authorities in Baden-Württemberg, Germany.

Whole-heart pacing: Mice were euthanised by cervical dislocation, their chest opened, the heart removed and placed in warm (37 °C) followed by cold (4 °C) heparin-containing (10 units/mL) physiological saline solution. The aorta was cannulated (within 8 min) for Langendorff-perfusion with oxygenated saline solution at 2 mL/min to 5 mL/min, with the saline flow adjusted to ensure a coronary perfusion input pressure of approximately 80 mmHg. Mice with an age between 7 months and 15 months were used for determination of maximum pacing rates with additional 2 month-old mice used for further confirmation of light emittance effects on the maximum pacing rate. Mice with an age between 4 and 6 months were used for the rheobase-chronaxie analysis.

The hearts were positioned in a water-jacketed bath (37 °C) and instrumented with two spring-loaded Ag/AgCl pellet electrodes (73–0200, Harvard Apparatus, Holliston, MA, USA) to provide a surface electrocardiogram (ECG). The LED chips were controlled using a system described previously, which allows ECG-timed triggering of the optical probe (Lee et al., 2011). ECG, perfusion pressure, LED-trigger signal and camera exposure were recorded using AcqKnowledge with a BIOPAC recording system (BIOPAC MP150, BIOPAC Systems Inc, Goleta, CA, USA). Subsequent data analysis was performed in Matlab (Mathworks, Natick, MA, USA).

A subset of hearts were loaded with 5 μL–20 μL of 1.4 mM Di-4-ANBDQPQ (dissolved in ethanol; dye courtesy of Leslie Loew, University of Connecticut Health Centre, Farmington, CT, USA) (Matiukas et al., 2007) by direct injection into the aortic cannula (added in 0.4 μL increments over 8 min, i.e. diluted in 20 mL–40 mL of saline) to enable optical mapping. The heart was illuminated with band-pass filtered (DS640/20X, Chroma Technology Corp, Bellow Falls, VT, USA) red LED (CBT-90R, Luminus Devices Inc, Billerica, MA, USA) and imaged with an electron multiplying charge-coupled device (Cascade 128+, Photometrics, Tucson, AZ, USA) via a macroscope (MVX10, Olympus, Tokyo, Japan). The camera was controlled and signals acquired using MultiRecorder (courtesy of Stefan Luther and Johannes Schröder-Schetelig, Max Planck Institute for Dynamics and Self-Organisation, Göttingen, Germany; http://www.bmp.ds.mpg.de/multirecorder.html). Optical mapping data was analysed using a custom-designed Matlab programme (courtesy of Alexander Quinn, Dalhousie University, Halifax, NS, Canada).

Once the hearts were positioned in the bath, initial baseline recordings were made prior to implanting the IMLOP. Both single- and double-sided probes were used to assess for differences in ability to pace and in tissue damage. Probes were inserted by hand through the ventricular wall (both left and right ventricular walls) and into the septum, and recordings were taken to assess for changes in heart rate following implantation (N = 18, 8 single-sided and 10 double-sided probes, 15 αMHC-ChR2 positive hearts and 3 non-ChR2-expressing control hearts). If spontaneous beating rate dropped below 4 Hz, hearts were discarded (one heart implanted with a single-sided probe).

The minimum light emittance required to stimulate the heart was determined to assess the dependency on aperture size. Furthermore, the maximum pacing frequency was found with different aperture sizes, radiant fluxes, and light emittances (N = 11, 6 single-sided and 5 double-sided probes).

Emittance-duration curves were found by measuring the emittance required for stable pacing (defined as responding to 70 light pulses applied at 7 Hz) at different pulse durations (1, 2, 3, 5, 10, 20, 30 and 50 ms) when using one LED or two back-to-back LED chips (N = 4). The resulting curves were subsequently fitted with the Hill-Lapicque equation to find rheobase and chronaxie values (Fozzard and Schoenberg, 1972; Williams and Entcheva, 2015).

Histology: Following the experiments, hearts were perfused with 1% triphenyltetrazolium chloride (TTC) in physiological saline solution at 37 °C for 30 min to visualise the puncture site. The hearts were then coronary perfusion-fixed with 4% paraformaldehyde, processed, and embedded in paraffin according to standard protocols. Serial sections (10 μm) were cut using a microtome (RM2255, Leica Microsystems, Vienna, Austria), de-paraffinised, rehydrated, stained with haematoxylin and eosin, and visualised using an automated slide scanner Axio Scan.Z1 (Zeiss, Wetzlar, Germany).

Statistics: Data from whole-heart pacing experiments are presented as mean ± standard error of the mean, with n giving the number of observations and N the number of hearts (each heart only had one probe inserted). Paired Student t-tests were used to assess changes in electrophysiology following implantation. Unpaired Student t-tests were used to assess whether the rheobase and chronaxie values differed with increased area of illumination. ANOVA of linear mixed-effects models were used to assess whether there were significant differences between threshold and maximum pacing rates for different hearts. Subsequently, linear mixed-effects models were used to assess whether radiant flux and emittance affected the threshold and maximum pacing rate. Calculations were performed using Excel (Microsoft, WA, USA) or Matlab (Mathworks, MA, USA). A value of p < 0.05 was considered to indicate a statistically-significant difference between means.

3. Results

3.1. Probe characteristics

An assembled double-sided IMLOP with an inter-LED distance of 300 μm and an aperture diameter of 50 μm is presented in Fig. 1a. Operating the LED with a current of 10 mA and a 10% duty cycle produces an average optical radiant flux of 40 μW. This corresponds to an emittance of 5.1 mW/mm². The corresponding average emittance of the LED under conditions used in the biological experiments (10-ms-long stimulation pulses, 7.8 mA current, 30% duty cycle on average) is 12.4 mW/mm². Both single and double-sided IMLOP assemblies show stray light emission to the sides through the small gap between PI substrate and Si stiffener after probe assembly (Ayub et al., 2017). The stray light intensity was found to amount to 34% of the light exiting from a 50 μm diameter aperture of the double-sided IMLOP.

3.2. Light absorption in cardiac tissue

The light absorption of cardiac tissue was investigated using an IMLOP interfaced via a PCB and introduced into the integrating sphere (Fig. 2a). Changes in the optical spectrum and transmittance when covering LED apertures with cardiac tissue slices of different thickness t_s are shown in Figs. 2b and c, respectively. As it is obvious from Fig. 2b, wavelengths below the peak wavelength of 456 nm experience a stronger absorption than longer wavelengths. The transmittance (T) vs. tissue thickness (t_s) shows a decay which is here modeled using a single exponential, i.e., T(t_s) = exp(-t_s/t_a), with an absorption length of t_a = 446 μm at 456 nm, as indicated by the fit curve in Fig. 2c. The observed deviation from an exponential decay is likely due to the experimental approach using a light source that is small compared to the layer thicknesses and emits light with a Lambertian characteristic. This is similar to literature, where the light intensity of 470 nm light was reported to decay to 50% after around 350 μm (Quinn et al., 2012), we reach this intensity attenuation after about 300 μm.

3.3. Thermal characterization

Temperature was measured using a test probe with thermistor embedded in the PI substrate underneath the LED chip (Fig. 3a). A representative temporal temperature profile of the probe is shown in Fig. 3b for ten consecutive current pulses at a repetition rate of 6 Hz. The thermal behaviour over time can be described by two thermal relaxation mechanisms with different time constants. One is attributed to the rapid heating/cooling of the LED chip while the second mechanism is related to the more gradual heat transfer to the surrounding medium (here agarose gel). The time constant corresponding to the rapid temperature increase ΔT can be extracted from the rising part at the beginning of individual pulses using an appropriate exponential fit function. After switching the LED off, the temperature decay allows the calculation of both fast and slow components, which are 8.35 ms and 210 ms, respectively. The applied LED parameters result in a peak temperature increase of 2.85 K directly underneath the LED, calculated from the fit curve.

Fig. 3c illustrates the temperature dynamics of the surrounding medium near the LED for pulse repetition frequencies of 6 Hz, 10 Hz, and 15 Hz, with the LED operated for 10 pulses each. For the given LED pulse duration, the maximum ΔT of the medium depends on the repetition rate, i.e. ΔT equals 0.35 K, 0.54 K, and 0.73 K at repetition frequencies of 6 Hz, 10 Hz, and 15 Hz, respectively.

Figs. 4a and b show experiments equivalent to Figs. 3b and c, performed in a Langendorffperfused heart. Again, the fast ΔT by up to 2.6 K, 2.8 K, and 3.0 K at repetition rates of 6 Hz, 10 Hz, and 15 Hz, respectively, describe mainly the heating of the LED chip. In contrast, the ΔT of the surrounding tissue is estimated as 0.36 K, 0.57 K, and 0.75 K at the given repetition rates. These ΔT values are in good agreement with those extracted using the agarose gel-based phantom.

Fig. 4.

Investigation of temperature increase in cardiac tissue: (a) Raw data and fit curve of temperature increase at 6 Hz repetition rate of ten 10-ms-long LED current pulses. (b) Fit curves of temperature increase of the surrounding tissue at repetition rates of 6 Hz, 10 Hz, and 15 Hz similar to Fig. 3c. (c) Fitted temperature increase data for switching on the LED for 1.67 s for a coronary-perfused and a non-perfused heart (inset: raw and fitted data of the non-perfused heart). Fit curves in panels a–c are extracted similar to Figs. 3b and c. Data in panels a and b are averaged using 100 subsequent sets of 10 pulses, while data in panel c is based on ten long pulses. In all cases a 10 mA LED current at 10 kHz with 50% duty cycle was used during pulses.

Fig. 4c shows the temperature increase measured with the integrated temperature sensor when applying 1.67-s-long current pulses with a duty cycle of 50%. These LED activation parameters exceed by far those typically applied during the pacing experiments of isolated hearts. The temperature increase in this setting reaches values of 7.46 K towards the end of the long light pulse. In the case of non-perfused heart, we observed a slightly higher ΔT (by 4%, i.e. 7.76 K). The inset in Fig. 4c shows averaged data of ten consecutive temperature measurements (blue circles) and the fit curve (red line) in the non-perfused heart. The time constants for fast and slow heating of LED and surrounding tissue are τ₁ = 16 ms and τ₂ = 400 ms, respectively.

3.4. Implantation force analysis

Figs. 5b–d show IMLOP insertion experiments with the insertion force and probe displacement given as a function of time. As indicated, probes are inserted and retracted at a constant speed of 100 μm/s; at the intended insertion depth, the position is held for up to 75 s (horizontal section of orange line). Fig. 5b shows an exemplary measurement of two consecutive insertions carried out at the same location on an arrested heart. Measurements performed while the heart was beating at 1.5 Hz and 0.5 Hz are shown in Figs. 5c and d, respectively. For all measurements, perfusion was stopped during IMLOP insertion in order to avoid possible changes in weight due to accumulation of the perfusion liquid in the dish supporting the heart. Upon contact with the cardiac tissue, the force increases non-linearly while the probe is being driven against and eventually into the tissue. Once the linear stage stops at the intended implantation depth, the force relaxes over time, with a residual force on the probe in the range of 10 mN–15 mN. Retracting the probe from the tissue decreases the force either to the baseline (Fig. 5b, first trial; Fig. 5c) or below baseline (Fig. 5b, second trial; Fig. 5d) which is attributed to frictional forces and fluid coupling between probe and tissue. Depending on the probe insertion site and orientation, the probe either introduces an indentation in the tissue (Fig. 5b,(first trial; Fig. 5c), or crosses the ventricular wall passing through one of the heart’s chambers before the tip reaches the cardiac septum (Fig. 5b, second trial; Fig. 5d).

In case of the first insertion trial (arrested heart; Fig. 5b), the probe induces an indentation in the heart surface without tissue penetration. During the resting period and probe retraction, the force relaxes to 6.6 mN and reaches the base line, respectively. In a second attempt at the same location with a larger displacement, a peak force of 21.4 mN is obtained, before the probe abruptly penetrates the tissue surface and enters the ventricular wall and chamber, detectable by the sharp dip in the recorded force. The force curve shows a local minimum and a second peak as the probe progresses through the ventricular cavity and into the septum before reaching the targeted travel distance. The resting period results in a force relaxation to 14.9 mN. During retraction, the probe experiences a negative force down to −3.9 mN before returning to the base line.

The two perfused, beating heart measurements (Figs. 5c and d) show clearly visible force oscillations, which are due to cardiac contractions, as further highlighted by the magnified presentation of five force peaks given in Fig. 5d’. For the lower beating rate of ca. 0.5 Hz (Figs. 5d and d’), the peak force resulting from the IMLOP implantation is 23.6 mN (experienced during contraction when myocardial tissue stiffness is highest). Although the probe travels more than 4 mm in case of Fig. 5c, we observed no penetration into the cardiac chamber as reflected by the lack of a sharp force peak before reaching the intended travel distance and no negative forces during probe retraction similar to Fig. 5b (first trial). Furthermore, a noticeably high amplitude of the force oscillations due to the heart beats is present in both cases during the relaxation phase.

In our probe insertion experiments, carried out in order to assess IMLOP penetration capability prior to in situ pacing experiments, peak forces of 22.41 mN ± 1.73 mN (n = 5, N = 1) were found to be necessary to penetrate cardiac tissue (when using an arrested heart). Penetration experiments on a perfused heart indicate that similar forces are needed. Further experiments are required to investigate whether a cardiac contraction cycle synchronised probe insertion, targeting diastole, may allow one to minimise penetration force in the more slowly beating hearts of larger animals.

3.5. Gross electrophysiological effects of IMLOP application in tissue

IMLOP implantation and presence in the ventricular tissue did not change sinus rhythm, whether using single-sided (p = 0.67, N = 8) or double-sided probes (p = 0.48, N = 9, Fig. 6a). Implantation did not cause any detectable changes in the speed of ventricular activation (no significant difference in QRS duration; p = 0.31, N = 6 for single-sided, and p = 0.26, N = 9 for double-sided probe, Fig. 6b; in two single-sided cases the ECG was too noisy to allow accurate assessment of QRS duration).

Fig. 6.

Effect on electrophysiology of IMLOP implantation. (a) Sinus rate and (b) QRS duration before (filled bars) and after (open bars) IMLOP implantation using single- or double-sided IMLOP.

Tissue damage by IMLOP insertion was assessed histologically, as shown in Fig. 7. Tissue disruption is evident along the IMLOP penetration channel, but no other gross morphological changes outside the immediate vicinity of the penetration site were observed. Furthermore, optical pacing was possible in αMHC-ChR2 hearts for periods of up to 150 min.

Fig. 7.

Representative serial histology sections stained with haematoxilin and eosin showing an IMLOP implantation site. Cross-sections of a double-sided IMLOP implantation site at various depths within the left ventricular free wall, at the epicardial surface (left), and at 120 μm (middle) and 200 μm (right) depths. Scale bars 200 μm.

The absence of thermal damage caused by LED activation is supported by functional measurements in control hearts where all LED chips on an IMLOP were operated for sustained periods (>1 s) at 10 mA and 50% duty cycle (radiant flux of 875 μW, compared to the average threshold radiant flux of 40.6 μW for pacing). No major alteration in heart rate or ventricular activation patterns observed from optical mapping was detected (Fig. 8).

Fig. 8.

Assessment of potential effects of heating on cardiac conduction. (a) Fluorescence image of a heart with implanted IMLOP. (b–d) Activation maps showing the spread of sinus rhythm excitation across the ventricles, either with (b) all LED switched off, (c) or all LED switched on for 1 s at an optical radiant flux of 875 μW, i.e. emittance 22.3 mWmm⁻², and (d) difference of excitation timing between (b) and (c). The IMLOP is implanted surface-perpendicular through the left ventricular free wall of a control ChR2-negative heart (as seen in panel a). The grey area corresponds to the location of the probe and its PI cable. Activation timing in the tissue surrounding the probe is not affected by continuous illumination (d). Scale bar 1 mm.

3.6. Pacing

Control hearts could not be paced, and no effects of IMLOP presence or LED activation on cardiac electrophysiology were observed. Pacing of αMHC-ChR2 hearts was possible using a single LED with an aperture of 50 μm. This enabled pacing at different locations throughout the tissue, resulting in altered cardiac activation patterns (Figs. 9a and b). Hearts could be optically paced at radiant fluxes as low as 4.2 μW and emittance as low as 700 μW/mm². This pacing threshold varied, both between experiments and between LED chips along the probe in a single heart. The average radiant flux needed to trigger ventricular excitation was 42.2 μW ± 4.3 μW (emittance 10.1 mW/mm² ± 1.1 mW/mm²; n = 66, N = 9). There was no significant effect of whether single- or double-sided probes were used on threshold radiant fluxes and emittances (p = 0.86 and p = 0.36, respectively).

Fig. 9.

Cardiac pacing with an IMLOP. (a,b) Optical activation maps showing the spread of conduction across the ventricles when pacing with an LED either (a) close to the epicardial surface or (b) 1.2 mm deeper into the tissue. The grey area corresponds to the location of the probe and its PI cable. Scale bar 1 mm. (c, d) ECG traces showing an example of elevated maximum pacing rate with increased radiant flux and emittance. (c) Maximum pacing rate of 9 Hz with a radiant flux of 105 μW (emittance of 13.4 mW/mm²) and (d) pacing at 10 Hz with a radiant flux of 158 μW (emittance of 20.1 mW/mm²).

Emittance-duration curves were measured using a double-sided probe when illuminating either a single LED or both LED chips at the same location along the shaft (Fig. 10a). Rheobase values were significantly lower when using two LED chips compared to one LED (6.5 mW/mm2 ± 0.9 mW/mm² vs 13.3 mW/mm2 ± 0.9 mW/mm², respectively, p < 0.001; Fig. 10b and Supplementary Fig. 1). There was no significant difference in chronaxie values (2.1 ms ± 0.2 ms and 3.0 ms ± 0.6 ms, for one and two LED chips, respectively; p = 0.17; Fig. 10c).

Fig. 10.

Emittance-duration curves for stable optical pacing. (a) Individual values, means ± SEM, and Lapicque-Hill fit of (a) emittance and (b) radiant flux when illuminating with one (blue; n = 67, N = 4) or two (orange; n = 37, N = 4) LED chips. An IMLOP with 100 μm aperture was used, corresponding to 0.008 mm² and 0.016 mm² for one and two LED, respectively. (b) Rheobase and (c) chronaxie values for emittance taken from the Lapicque-Hill fits from (a). (e) Rheobase and (f) chronaxie values for radiant flux taken from the Lapicque-Hill fits from (d). *** indicates p < 0.001.

The absorption length in cardiac tissue is similar to the size of the light source, therefore examining the radiant flux rather than the light emittance may be more meaningful. This is supported by there being no significant difference in either rheobase or chronaxie values when using one LED compared to two (rheobase: 58.7 μW ± 4.1 μW vs 57.3 μW ± 8.4 μW, respectively, p = 0.88; chronaxie: 2.1 ms ± 0.2 ms and 3.0 ms ± 0.6 ms, respectively; p = 0.17; Fig. 10d–f).

The maximum pacing rates were higher with greater radiant flux or emittance (p < 0.001 for both, n = 147, N = 7; Fig. 9c and d, and Supplementary Fig. 2). Radiant flux can be raised by increasing drive current, number of LED chips used or aperture size, while emittance can be raised by increasing the drive current. Increasing the drive current for single LED chips we observe that the maximum pacing rate could be elevated by increasing emittance and radiant flux (an increase in maximum pacing rate by up to 9 Hz, from 5 Hz to 14 Hz was observed when increasing the radiant flux and light intensity by one order of magnitude, from 16 μW to 160 μW and from 3.5 mW/mm² to 36 mW/mm², respectively; Supplementary Fig. 3).

4. Discussion

In order to overcome the challenge of limited light penetration when illuminating hearts from their surfaces, we developed implantable optical probes utilising multiple LED chips to allow intramural illumination while minimally altering cardiac structural integrity and function. The IMLOP implantation requires low forces (on average 22.5 mN), due to the sharp tip geometry and smooth probe surface. Implantation through the apex was tested in some initial trials. However, the smooth surface of the probe also means that an implanted IMLOP is prone to movement and rotation, resulting in the probes rapidly being pushed out of the tissue during regular contractions of the heart due to longitudinal length changes. Further development is required to improve positional stability for trans-apical/intra-septal IMLOP placement.

Histological analysis showed that cell disruption was limited to the IMLOP insertion channel, with no other gross morphological changes. No changes in spontaneous beating rate and cardiac conduction parameters were detected upon IMLOP insertion.

Once implanted, further tissue damage could conceivably result from local heating while operating the IMLOP. Thermal characterization using an agarose gel-based phantom showed that the overall temperature increase of the surrounding tissue should stay below 1 K when operating the IMLOP using typical experimental parameters. The bench-test data in the agarose gel phantom are comparable to the temperature measurements in an isolated heart. Furthermore, operation of the LED uninterrupted for a longer duration, i.e. 1.67 s at 10 mA and 50% duty cycle, results in ΔT > 7 K in the heart. However, the energy input for cardiac pacing (typically 10 stimuli at 6 Hz; 7.8 mA, 30% duty cycle, 10 ms duration) is only approximately 3% of that input for sustained illumination, therefore the corresponding ΔT values for pacing are regarded as safe for in vivo applications (tissue necrosis of skeletal muscle is estimated to require the temperature to stay at 1 K above normal, i.e. at 38 °C, for over 300 h in mice, or 16,000 h in human) (Dewhirst et al., 2003). No effects on cardiac conduction and heart rate were observed in control hearts, even when LED chips were continuously operated for periods of up to 5 s, with radiant fluxes more than an order of magnitude greater than those used for pacing.

These results indicate that tissue damage resulting from the use of IMLOP is minimal. However, available data are restricted to the acute setting. Future work will need to assess the utility of these probes in the chronic setting, which will strongly depend on IMLOP encapsulation, tissue fixation, and interfacing with the control circuitry. Initial measurements of our group on test devices in saline solution indicate that LED-based probes can be continuously operated for more than two months. However, for translational application, long-term stability needs to be assessed over years. Furthermore, in chronic settings, tissue remodelling is likely to occur around the probes. As an example, the Micra pacemaker system (Medtronic, USA) is an implantable leadless pacemaker in clinical use that uses tines to hold the pacemaker in place within the ventricle. The system has shown that the remodelling around these tines is relatively small, that it is not more damaging to the heart than the underlying pathophysiology, and that they are stable in the long-term. This indicates that, given an appropriate encapsulation of the IMLOP, long-term stability with minimal damage to the heart may be possible (Chen et al., 2016). Of course, tissue remodelling may alter the ability to optically pace. However, the optical absorption of cardiac scars tends to be lower than that of healthy myocardium (Lagarto et al., 2019). This may partially off-set the expected reduction in cardiomyocyte presence in the immediate proximity of an implanted device.

It was possible to pace hearts using IMLOP with aperture diameters as small as 50 μm, suggesting that a switch to smaller thin-film μLED emitters may be possible for future device refinement, to further minimise tissue damage (Klein et al., 2018). However, the reduction in radiant flux and/or emittance resulting from a smaller LED must be balanced against the observed ability to pace at faster rates with increases in these parameters. This effect may be explained by cardiomyocyte repolarization becoming incomplete at faster pacing rates, which – due to slowed recovery from inactivation of fast sodium channels – may raise the excitation threshold potential and reduce AP upstroke velocity and amplitude. Increasing radiant flux or light intensity increases the peak pacing rate attainable, presumably due to increasing the effective depolarising stimulus to a level sufficiently large to overcome the shift in threshold.

The emittances required for stable pacing (13.4 mW/mm² and 8.9 mW/mm² for one and two LED, respectively) were higher than those reported in literature, with intensities typically between 0.5 mW/mm² and 2 mW/mm² when pacing whole hearts with 10-ms-long pulses (Bruegmann et al., 2010; Nussinovitch and Gepstein, 2015; Vogt et al., 2015). Rheobase values (13.3 mW/mm² and 6.5 mW/mm² for one and two LED, respectively) were also higher than those reported on cell cultures (≈0.04 mW/mm²) and whole hearts (≈0.3 mW/mm²), however observed chronaxie values (2.1 ms and 3.0 ms for one and two LED, respectively) were similar to findings for cell cultures and whole hearts (≈4 ms and ≈7 ms, respectively) (Vogt et al., 2015; Yu et al., 2015). The increased emittances required when using IMLOP may be explained by the small tissue volume exposed to optical stimulation. When decreasing the stimulated volume, an increase in both rheobase (Vogt et al., 2015) and emittance required for stable pacing (Bruegmann et al., 2010) has been observed previously. The light intensities required here for stable pacing with two LED (corresponding to an area of 0.02 mm²; 8.9 mW/mm²) are similar to published results with a small illumination area (0.05 mm²; ≈9 mW/mm²) (Bruegmann et al., 2010). Furthermore, when converting reported rheobase values from emittances to radiant fluxes we see that IMLOP devices require lower energies to pace (58.7 μW and 57.3 μW for 1 and 2 LED with 100 μm apertures, respectively corresponding to 0.008 mm² and 0.016 mm²) compared to reported literature data (≈4.1 mW and ≈6.3 mW for 10.2 mm² and 25.3 mm², respectively) (Vogt et al., 2015). These results indicate that the IMLOP approach is suitable for pacing the heart and, given that much lower light energies are required, it may be useful for studies where long-term optical stimulation is necessary. Further research into long-term pacing efficiency is needed, as well as on the stability of the probes and their effect on the heart during chronic presence.

Our study highlights the utility of implantable multi-site optical stimulation probes to modulate cardiac electrophysiology in ChR2-expressing hearts. IMLOP enable multi-site transmural excitation and they are compatible with cardiac electrical function for the typical duration of ex vivo experimental studies. Further studies should additionally explore the possibility of multi-site intramural triggering of excitation and of light-induced pacing in larger hearts, where this may offer benefits for uniform transmural stimulation, avoiding intramural re-entrant waves that may occur as a result of surface excitation.

Appendix A. Supplementary data

The following is the Supplementary data to this article: