Graphene cardiac tattoos
The graphene biointerface technology reported in this work is derived from the previously introduced GETs40–42, additionally supported with two layers of ultrathin elastomer. The top layer of the elastomer features an electrode opening (see Fig. 1) with a diameter varying between 1 and 3 mm. An ultrathin (10 µm) adhesive, elastic, and conductive gold tape was leveraged as the hybrid interconnect: soft enough for contacting graphene yet hard and robust enough for external connection with measuring and stimulating electronics. Furthermore, to ensure that graphene does not affect biological processes in vivo or ex vivo, the GETs were modified with micrometer-sized holes, allowing for efficient water transport43. Besides, the holey GETs feature a unique tendency toward forming an electrically robust interface with the conductive material on both sides/planes of the GET (graphene plane and holey PMMA plane), which has given us the freedom to create the hybrid cardiac tattoo samples. In this work, we used 3-layer-GETs as they provide the most reproducible qualities, including lowest sheet resistance and lower interface impedance.
A simple, easy-to-replicate method (Fig. 1a) was developed to produce GET-electrodes (Fig. 1b) for a direct electrical interface with the heart (Fig. 1c) that only requires two steps. First – spin coating of poly(vinyl alcohol) (PVA) and silicone acted as the temporary supporting layer and flexible encapsulating layer, respectively. Second – defining different electrode window patterns with biopsy punches of various sizes (diameter 1-3mm). A GET-electrode consists of three primary layers with different functional purposes (Fig. 1a-b) involving (i) transparent and flexible encapsulating films, (ii) GETs that could come with various shapes based on anatomical and device requirements, and (iii) a flexible and ultrathin conductive gold tape that acted as the interconnect between GET-electrodes and data acquisition and stimulation hardware. The lightweight and soft GET-electrode could easily conform to the curvature of small anatomical structures of a mouse heart (Fig. 1c).
Electrical and electrochemical characterization of graphene electrodes
To ensure reliability ex vivo and in vivo, the GET cardiac patches were first tested for their electrochemical performance. Large surface area GETs have been made for initial tests, typically 15 to 18 mm2. The large GETs were made in three configurations: monolayer graphene (1L), bilayer graphene (2L), and trilayer graphene (3L, see Methods for fabrication details). As one can see from Fig. 2a, there is a clear trend in decreasing of the electrochemical impedance of the devices with a larger number of graphene layers in the stack. On average, the 3L-GETs have an interface impedance of 2.5±0.7 kOhm at 1kHz (see Fig. S1) or area-normalized impedance of 345±151 Ohm×cm2 (see Fig. 2b). Interestingly, the small devices with electrode openings of 1mm, 1.5mm, 2mm, and 3mm in diameter show a downwards trend of reducing normalized impedance with smaller electrode openings, suggesting that smaller electrodes are more efficient (see Fig. 2b and Table S2). This efficiency is likely to come from the larger edge/area ratio of the smaller electrode, and potentially the charge transfer is more efficient through the edges rather than that middle part of the electrodes.
The average value of area-normalized impedance at 1 kHz for 1mm opening 3L GETs is 40±13 Ohm×cm2, which is on par and even exceeds that value of highly conducting PtTe2 and gold e-tattoos44. The GETs exhibited outstanding stability and no degradation of the properties over a few days and a week of measurements (see Fig. S2). To move forward with the electrochemical characterization of the cardiac graphene tattoos, we characterized their electrochemical stability and determined their so-called water window. As well-known from the previous works, graphene has a very wide water window, which we corroborated in this work. We have been able to safely perform cyclic voltammetry (CV, see Fig. S3) measurements in the range of applied potential between -0.9V and +1.2V, which defines the graphene water window37,45. Furthermore, we have found no significant dependency of the water window per electrode opening area or per number of graphene layers (see Figure S3). The CV measurements, performed at 5 mV/sec rate, have been used to estimate another essential figure-of-merit of microelectrodes, their charge storage capacity (CSC). Similar to the case of interface impedance, the CSC follows a similar trend (see Fig. 2c), where area-normalized values are much larger for the smaller electrodes. For large devices, the average CSC is estimated in the range of 2-5 mC/cm2. At the same time, the value is almost an order of magnitude higher, 63.7±14.6 mC/cm2, for smaller electrodes, with ~1mm diameter opening (see Table S3 for details). Finally, the essential value for effective stimulation of electrogenic tissue is the charge injection capacity (CIC) of the electrodes. For this purpose, the charge-balanced measurements with constant-current pulses were applied to the electrodes. The resulting voltage transient for various pulse amplitudes and pulse widths are shown in Fig. 2d. Extracting the CIC capacity is giving us a somewhat similar performance for 1L, 2L, and 3L large GETs, in the range of 60-90 µC/cm2, with the values drastically increasing for smaller devices (see Fig. 3e), reaching up to 704±144 µC/cm2 for electrodes with ~1mm diameter opening (see Table S4 for details). One can note a slight upwards trend with increasing the number of layers, suggesting that multilayer graphene helps in out-of-plane charge injection. The recorded properties for large devices are in good accordance with other works46, while the superior performance measured for small GET-arrays is unprecedented. When put in perspective (see Table S1), the graphene electrodes are the thinnest made biointerfaces with the lowest (for graphene) interface impedance. Only a handful of works, mainly utilizing PEDOT and porous metals, can compete with the GETs. The same trend can be found for CSC and CIC, with the exception of composite multi-coated electrodes (e.g., CNTs+porous Pt).
Ex vivo cardiac electrophysiology sensing and actuating characterization
Upon completion of electrochemical characterization of the GET-electrodes, they were applied for monitoring cardiac electrical activity in an ex vivo Langendorff-perfused mouse heart model where the electrogram recorded by graphene (gEG) was compared with the traditional far-field ECG simultaneously recorded in the perfusion bath (Fig. 3a). The ECG was recorded from three sensing electrodes in the ECG lead II position. The gEG was recorded from a 2-sensing electrodes setup. The GET-electrodes came with both unipolar (1-, 1.5-, and 2 mm electrode windows) and bipolar (1 mm electrode window, window pitch distance 2 mm) setups. In the unipolar sensing and actuating mode, the same GET-electrode was used as the positive (+) electrode paired with another negative (-) electrode. In the bipolar mode, the two graphene electrodes were used as (+) and (-) electrodes, respectively. The simultaneously recorded ECG and gEG (Fig. 3b-c) showed a good temporal correlation between R waves and elapsed time between two successive R waves (RR interval, inversely related to heart rate) (Fig. 3d). Because the GET-electrode recorded local electrogram, the P waves were absent from gEG signals (Fig. 3c), present in ECG, which characterizes whole heart electrophysiology. The signal-to-noise ratio (SNR) was compared (Fig. 3e) for gEG recorded from unipolar GET-electrodes with various window sizes. We found that 1 mm samples show comparable performance to the control Pt electrodes. At the same time, there was a statistically significant decrease in the SNR of gEG recordings between the 1 mm and 2 mm group. Only in the 2 mm group, the gEG SNR was significantly lower than the control ECG SNR. This experimental data correlates with the data shown in Fig. 2b-c, where smaller diameter electrode arrays have featured superior interface impedance, charge storage, and charge injection capacities. Cardiac actuating was also achieved by connecting the GET-electrode to the cardiac stimulator (PowerLab 26T, ADInstruments). Cardiac pacing resulted in a faster heart rate, and pacing artifacts were followed by a wide QRS complex with increased amplitude, as commonly observed in patients with an implantable pacemaker (Fig. 3f). The ex vivo pacing strength-duration curve was characterized for various GET-electrodes along with a custom bipolar platinum electrode that served as the reference since it is always the electrode our lab uses for optical mapping studies (Fig. 3g, and Table S5 for details). The choices of stimulating pulses duration included both animal research related value (e.g., 2 ms for optical mapping studies) and clinically relevant pacing duration (i.e., from 0.1 to 0.5 ms)47–49.
Validation of GET-electrodes with optical mapping studies
Graphene is an optically transparent material28 and has been reported to make transparent electrodes for applications such as photovoltaics50. Cardiac optical mapping is a fluorescent imaging technique frequently used in cardiac physiology to study the excitation-contraction coupling between transmembrane potential and calcium handling using optical dyes that emit fluorescence upon excitation illumination51,52. Considering the almost complete transparency of graphene and the previous characterization of the sensing and actuating properties of GET-electrodes, it is natural to incorporate GET-electrodes into optical mapping studies where the transmission of light is critically important. It is common to use metal electrodes such as those made of platinum to stimulate the heart during experiments. However, this pacing electrode optically blocks the partial area of the tissue preventing optical recordings. Depending on the size of the electrode, it may be difficult (e.g., low optical signal amplitude, distorted signal morphology) or even impossible to analyze signals from pixels in the region of interest. By using the transparent GET-electrode, those issues are now resolved.
Here we measured and compared the cardiac restitution properties recorded (Fig. 4a) from optical signals (i.e., action potentials and intracellular calcium transients) during pacing by a unipolar GET-electrode (1 mm window size) and a custom bipolar platinum electrode. The GET-electrode, which was placed on the anterior side of the left ventricular surface of the heart, could barely be seen in a bright-field image captured by the camera (Fig. 4b). The activation map showed the expected anisotropic propagation of the transmembrane potential originating from the site of cardiac pacing by GET-electrode throughout the ventricular myocardium (Fig. 4c). Clear representative optical signals from different heart locations, including where the GET-electrode was aligned well with the simultaneously recorded ECG (Fig. 4d). The wide and high-amplitude QRS complex immediately after each pacing artifact indicated successful ventricular capture by the GET-electrode. No statistical difference (see Fig. 4e) was found between the platinum and graphene group of action potential duration 80 (APD80), calcium transient decay constant (Ca Tau), as well as longitudinal and transverse conduction velocity (CVL, CVT). Taken together, these tests indicate that the flexible and transparent GET-electrode is readily applicable to optical mapping studies with high efficacy in cardiac pacing and no harm to the light transmission that allows precise measurement of cardiac restitution properties. Besides, we tested the GET-electrodes for potential cross-connection and insulation performance. As one can see from the control experiments (see Fig. S4), samples without graphene or passivation opening yield no ECG recording, ensuring that the signal originates specifically from graphene.
GET-electrode-array for cardiac electrical mapping
Electrical mapping with the high spatial resolution is critical during cardiac EP studies required to guide ablation therapy of arrhythmias. For example, unsatisfactory spatial resolution can result in the incorrect location of drivers of atrial fibrillation (AF), which is the most common type of cardiac arrhythmia affecting millions of patients53. To prove the potential high spatial resolution cardiac electrical mapping ability of electrodes based on GETs, we fabricated a novel array of micropatterned GETs (mGETs) and showcased its sensing ability on an in vivo beating rat heart (Fig. 5a-c). Unlike earlier version GETs with a whole graphene layer supported by a layer of ultrathin (200 nm) PMMA, the mGETs (Fig. 5a) had two unique features. First, the layer of graphene was patterned into multiple feedlines and individually addressed, all being supported by a single PMMA substrate. Secondly, the mGET-arrays were additionally passivated on top with another layer of PMMA with openings only at the contact and electrode opening sites (like classic microelectrode arrays). However, this PMMA-graphene-PMMA stack resulted in thinnest even microelectrode array, with a thickness below 500 nm.
Moreover, the mm-scale of the devices allowed us to fabricate the devices using off-the-shelf, low-cost additive fabrication components (see Methods). The 2 × 2 mGET-arrays were placed on a rat heart to record electrical signals from both right and left ventricles (Fig. 5a-c). In vivo ECG was recorded throughout the period of the experiments from sensing electrodes positioned in the lead II configuration. One can see prominent PQRST phases of a healthy cardiac wave. Because one graphene electrode (Fig. 5c, gEG-4) was placed close to the atrium, a clear P wave was observed. The simultaneously recorded ECG and gEG showed a good temporal correlation of R waves. Figure S5 shows the four gEGs individually.
Treating atrioventricular block (AV block) in an in vivo rat model
Fig. 5d shows the in vivo pacing strength-duration curve of bipolar GET-electrodes. Stimulus pulse duration covered the clinically used values such as from 0.2 to 0.5 ms47–49. Fig. 5e-g shows the application of the bipolar GET-electrode to the treatment of AV block. AV block was induced by fast pacing (cycle length 100ms) at the left ventricle after intraperitoneal injection of 120 mg/kg caffeine and 60 mg/kg dobutamine. When no pacing treatment was applied, the rat heart exhibited continuous 4:1 AV block (Fig. 5e-f). Upon ventricular pacing using mGET-arrays (pacing heart rate at 300 BPM), successful capture with clear pacing spikes and widen morphologies of the QRS complex was observed (Fig. 5g). The ventricular beating was also apparent during pacing compared with almost silent contraction during AV block (photographs not shown here). When pacing was stopped, the heart rhythm returned to AV block (Fig .5e, brown box). Overall, these in vivo tests demonstrate that the flexible mGET-arrays can effectively sense and capture a beating heart that has the potential to be applied to clinically relevant scenarios like ventricular rhythm normalization during common arrhythmias such as AV block. This is the first time graphene electrodes have been used to successfully treat a life-threatening heart rhythm disorder.