Self-adaptive cardiac optogenetics device based on negative stretching-resistive strain sensor.

Precision medicine calls for high demand of continuous, closed-loop physiological monitoring and accurate control, especially for cardiovascular diseases. Cardiac optogenetics is promising for its superiority of cell selectivity and high time-space accuracy, but the efficacy of optogenetics relative to the input of light stimulus is detected and controlled separately by discrete instruments in vitro, which suffers from time retardation, energy consumption, and poor portability. Thus, a highly integrated system based on implantable sensors combining closed-loop self-monitoring with simultaneous treatment is highly desired. Here, we report a self-adaptive cardiac optogenetics system based on an original negative stretching-resistive strain sensor array for closed-loop heart rate recording and self-adaptive light intensity control. The strain sensor exhibits a dual and synchronous capability of precise monitor and physiological-electrical-optical regulation. In an in vivo ventricular tachycardia model, our system demonstrates the potential of a negative stretching-resistive device in controlling-in-sensor electronics for wearable/implantable autodiagnosis and telehealth applications.

INTRODUCTION

In the field of precision medicine, and particularly in the area of cardiovascular diseases, a topic of major interest is the development of implantable devices capable of prevention, diagnosis, and treatment of specific diseases. These devices should provide physiological parameters from biomarker levels to alert individuals with underlying health conditions and their doctors to the occurrence of abnormal disease changes (). To shorten the salvage time of arrhythmia and improve the treatment efficiency, implantable devices with the property of closed-loop self-monitoring and accurate controlling are vital for precision medicine. Closed-loop monitoring means that the self-diagnosis output regarding the health status of patient directly determines the input (often turned on or off) of the treatment, and the intensity, frequency, and rate of treatment are regulated by the output of monitoring. In general, the monitoring and treatment in electrical therapy are achieved with multiple pieces of physiological equipment, which introduces the problems of potential asynchronism and delay between different monitoring devices. With the development of medical technology, optogenetics emerged as an accurate treatment method for neurological disease. Subsequently to being applied in neurology, optogenetics has been applied in cardiology. Several studies have demonstrated the feasibility of peacemaking in mice heart in vivo (–). Cardiac optogenetics is promising for its excellent cell selectivity, high time-space accuracy, pace controllability, and painlessness. However, more researches on cardiac optogenetics are still needed, such as on the effects of different kinds of photosensitive proteins, parameters of light intensity, pulse time, and pulse frequency. Since optogenetics can be accurately controlled, a highly integrated implantable cardiac optogenetics device with the property of closed-loop monitor and control is desired, which would be indispensable for the immediate treatment of arrhythmia.

Heart rate monitoring by electrodes requires the application of multiple sensors, which are often invasive and coupled to bulky instruments, resulting in inconvenience and short-term detection. In contrast, flexible strain sensors conforming to the curved surface of the heart can actively monitor the real-time heart rate by the strain changes in the periods of systole and diastole, thus simplifying the complex monitoring operation. Over the past decades, flexible stretchable strain sensors have been widely explored and been of increasing demand with the development of wearable human-machine interaction (–), medical monitoring (–), and soft bioelectronics (, ). In general, in the process of stretching, the raising resistance of strain sensor with strain, namely, the positive gauge factor, is the result of the conductive path breakage (–). Unlike the traditional strain sensors with positive gauge factor, there are also rarely developed negative stretching-resistive strain sensors (the resistance decreases with the increase in strain), which can effectively regulate the power of electricity and heat of current-typed device. For example, they can be used to control resistance reduction in a stretchable state. A negative stretching-resistive strain sensor can bring about stretchable optical therapies based on light-emitting diode (LED), such as optogenetics, broadening the closed-loop sensing and low-power application with requirements of high stretchability (, ). In addition, resistor thermal noise in conductors is positively associated with the thermal motion of free electrons, so decreasing resistance induces lower resistor thermal noise (). Moreover, the decreasing resistance reduces the time constant of the resistor-capacitance circuit system, thus raising the applications’ throughput (). Briefly, a negative stretching-resistive strain sensor with reliable performance becomes the prerequisite condition to realize a cardiac optogenetics system together with continuous closed-loop monitoring and self-adaptive light stimulation in a highly integrated device.

To date, there are limited comprehensive reports about strain sensors with negative gauge factors and the existing researches mainly focus on conductive composite and liquid metal filled with iron particles (, ). However, these two reported materials can only induce conductivity with strain and the discontinuous changes (step response corresponding to stretched or unstretched state) cannot be precisely controlled. Besides, the continuous negative stretching-resistive effect has been explored with two-dimensional conductive fillers such as MoS2 nanosheets using their intrinsic piezoresistive property and with fabrics by increasing the contact area under strain (, ). Meanwhile, the negative stretching-resistive strain sensor is always accompanied by the “shoulder effect” in cyclic loading (, ). The shoulder effect exhibits uncontrollable characteristic and the resistance decreasing is faint, which is not applicable for strain sensors. Furthermore, in all proposed sensors, the small period of tensile strain (<15%) in the negative stretching-resistive effect limits their applicability in wider stretchable situations with real-time sensing and multi-integration. Accordingly, a strain sensor with a negative gauge factor and a continuous change of resistance over a wide period of strain is in high demand of cardiac optogenetics.

Here, we report a self-adaptive implantable cardiac optogenetics system based on an original negative stretching-resistive strain sensor array for closed-loop heart rate recording and self-adaptive light intensity control. The resistance of the negative stretching-resistive sensor, which is composed of carbon nanotubes (CNTs) and natural latex (NL), continuously changes with stretching and decreases by 75.3% with 86.6% strain. The mechanism behind the controllable and continuous negative strain gauge has been theoretically and experimentally demonstrated to verify the effectiveness. The orientation of CNT and NL particles enhances the conductive path constructed by CNT-wrapped NL particles, and a high CNT concentration and strong bonding between CNT and NL are two essential conditions. On the basis of the unique closed-loop monitoring and self-adaptive control functions of the negative stretching-resistive strain sensor, a highly integrated implantable cardiac optogenetics system achieves active heart rate monitoring and triggers self-adaptive light stimulation with low power consumption (<15% duty cycle) and efficient performance [58.9% increase in cycle length (CL) in 8 min] in an in vivo experiment, verifying its effectiveness and for curing ventricular tachycardia (VT).

RESULTS

Controllable negative stretching-resistive effect

Here, a self-adaptive implantable cardiac optogenetics system based on an original negative stretching-resistive strain sensor array is proposed (Fig. 1A). A negative stretching-resistive strain array is placed on a flexible and stretchable substrate with an LED array, enabling closed-loop monitor and control with low energy consumption. The negative stretching-resistive strain sensor is based on the CNTs and NL particles, and its peculiar properties are demonstrated. In general, the resistance of strain sensors increases greatly with stretching when the concentration of the conductive filler is below a certain concentration threshold (Cth), leading to an open circuit in series with the luminescent device (Fig. 1B, gray area). Once the concentration exceeds Cth, the resistance of the proposed strain sensor drops sharply across the strain range, which would not only maintain the electrical performance of the circuit application but also improve the light intensity in the scene of stretching (Fig. 1B, orange area). This special phenomenon of resistance decrease with stretching is called the negative stretching-resistive effect in this paper. A nanocomposite is fabricated by mechanically mixing well-dispersed CNT solution and NL solution and then heating at 80°C in a mold (Fig. 1C). After being activated by several cycles of stretching, a more stable internal conductive network is established in CNT-NL (CNL) membrane, showing a unique negative resistance variation with strain during stretching.

Fig. 1.

Negative stretching-resistive effect.

(A) A self-adaptive implantable cardiac optogenetics system based on an original negative stretching-resistive strain sensor array. (B) The characteristics of stretching resistance when the concentration of the conductive filler is below or above the concentration threshold (Cth) and their opposite effects in optoelectronic applications. (C) Fabrication process of the CNL membrane. Activation of the negative stretching-resistive membrane via prestretching after release from the mold. MWCNT, multiwalled CNT. (D) Resistance characteristics of CNL membranes with different CNT ratios. The resistance increases first at stage 1 and is dominated by resistance decrease during 100% strain. (E) Scanning electron microscope (SEM) of the surface of unstretched CNL membrane. (F) SEM of the surface of CNL membrane shows vertical cracks and wrinkles. (G) Light off in the series circuit of LED and unstretched CNL membrane. (H) Light on when it is stretched breadthwise. Photo credit: Chunpeng Jiang, Shanghai Jiao Tong University.

The CNL composites exhibit a concentration-controllable original resistance, and the resistivity minimally reaches 20 ohm·mm. The CNL composites show three stages of resistance change across the strain range of 0 to 100% (Fig. 1D). The resistance first slightly increases at very small strain, then decreases greatly, and lastly slowly rises again. The first and third resistive change stages are caused by the reconstruction of the CNT. Inflections A and B of resistance drift with the conductive filler ratio. Inflections appear at larger strain with higher CNT ratio, while inflection B on the red curve [5 weight % (wt %)] occurs at 86.6% strain. Thus, the CNL composites with a concentration of 5 wt % exhibit a wider negative resistive range, and their resistance decreases by 75.3% of the original resistance (under zero strain) with 86.6% strain. Therefore, all of the characteristics tests are performed with 5 wt % CNL composites because of the high sensitivity, unless specifically stated.

Different ratios of CNL composites are investigated, and it is found that the negative stretching-resistive effect no longer exists when the CNT concentration is below 2.9 wt % (fig. S1). In addition, the concentration limitation of CNT for film formation is 8 wt %, over which the nanocomposite will appear almost in a powdered and nonstretchable form. The serial weight ratio of CNT solution to NL solution is presented in table S1. In addition, the proposed CNL membrane can be stretched to a maximum of 640% before total fracture (fig. S2).

The morphology of the CNL membrane can be observed with a scanning electron microscope (SEM). Compared with the SEM image of the unstretched CNL membrane surface (Fig. 1E), the CNL membrane subjected to transverse stretching shows the feature of crosswise wrinkles and vertical cracks (Fig. 1F). It can be strongly inferred that shrinkage perpendicular to the direction of stretching occurs, which is a non-negligible factor on the negative stretching-resistive effect. In addition, the dynamic process of cracks and wrinkles formation during stretching is presented in movie S1.

To demonstrate the negative stretching-resistive effect of the CNL composites, we connect an LED, a CNL membrane, and power supply in series. With the stretching performed breadthwise on the CNL membrane, the LED is turned on (Fig. 1, G and H, and movie S2). The light intensity change verifies the unique function of current regulation induced by the negative stretching-resistive effect of the CNL composites.

Mechanism and characteristics of negative stretching-resistive effect

To fully explore the mechanism behind this special negative stretching-resistive effect, we carry out experimental and theoretical investigations. First, to eliminate the effect of single component, we perform many groups of experiments including NL with carbon black, silver nanowires, and graphene and CNTs with thermoplastic polyurethanes, hydrogels, and polydimethylsiloxane, but none exhibits the negative stretching-resistive effect (fig. S3). Consequently, it is inferred that the negative stretching-resistive effect is caused by the combined action of CNTs and NL. NL particles are a kind of natural polymer material structured with a cis-1,4-polyisoprene–based core and a protein-phospholipid–based shell (Fig. 2A) () with a diameter of approximately 2 μm (Fig. 2B). The CNT has strong connections with phospholipids at the end of cis-1,4-polyisoprene by cation-π bonding and with the chain itself by van der Waal bonding (Fig. 2C) (, ). To account for the high concentration of CNT, we fully enclose the NL particle surface by CNTs, which means that there are reliable connections between the polymer matrix and the conductive fillers.

Fig. 2.

Mechanism of negative stretching-resistive effect.

(A and B) Schematic and SEM of core-shell structure of NL particle. (C) Bonding between CNT and cis-1,4-polyisoprene. (D) Schematic of conductive paths in CNL membrane before stretching. (E) Schematic of conductive paths when stretching. CNT and NL are squeezed together because of the orientation of CNT and NL. (F) Schematic of conductive path breakage when overstretching. (G) Neighboring NLs connected by CNT bridge. (H) NL particles trapped in the adjacent cells that are segregated by CNT. (I) NL aggregation with low CNT concentration. (J) Fourier transform infrared spectroscopy spectra of pure NL and its composites with multiwalled CNT, silver nanoparticles (AgNPs), and graphene. a.u., arbitrary units. (K) NL wrapped by abundant CNT. (L) No bonding between CNT and lipid-free deproteinized natural rubber (P-DPNR) particles is found. (M) Positive stretching-resistance response of CNT–P-DPNR with strain. (N) CNT distribution before stretching. (O) Orientation of CNT along the direction of stretching. Photo credit: Chunpeng Jiang, Shanghai Jiao Tong University.

The CNTs wrapping NL particles agglomerate into differently shaped clusters, and some floating CNTs exist. When the CNL membrane is unstretched, there are three types of conductive situations: conductive path formed by CNT connections (S1), minor gap to transit according to the tunneling effect (S2) (–), and nonconductive path due to the large gap being unsuitable for electrons to pass through (S3) (Fig. 2D). When the CNL membrane is stretched, the conductive situations are reinforced because of the extrusion perpendicular to the stretching direction with the orientation of the CNT and NL particles (). For S1, the original conductive path is strongly compressed by NL particles and becomes more compact. Similarly, for S2, the gap tends to become smaller with stretching, leading to the formation of CNT connections or smaller gaps for electron transit. The large gap in S3 becomes minor, providing the possibility of tunneling current. Consequently, the aforementioned three situations contribute to the negative stretching-resistive effect of CNL composites (Fig. 2E). However, as stage 3 in Fig. 1C, the orientation of the NL particles and CNTs gradually recedes during overstretching, leading to the breakage and large gap of CNT conductive networks and the CNT line structure (Fig. 2F) (). If the orientation effect cannot resist the breakage, then the resistance appears to rise in the macro view. As for CNL composites with low CNT concentration, except for the breakage of the conductive path, the conductive path is blocked when stretched (fig. S4) because the NL particles are not sufficiently wrapped by CNTs. As a result, electrons are prevented from transferring to the NL particle aggregations. Accordingly, the resistance of the membrane with low CNT weight ratio rises exhibiting a positive stretching-resistive effect (Fig. 1B, gray dashed curve). The theoretical analysis of the decreasing resistance is performed according to the electron conduction mechanism (see Supplementary Text and figs. S5 and S6).

According to the proposed hypothesis, the negative stretching-resistive effect is based on combination of the following three influences: (i) high concentration of CNT, (ii) strong bonding between CNTs and NL particles, and (iii) orientation of CNT and NL particles. Transmission electron microscope (TEM) images further demonstrate the proposed hypothesis. First, in CNL dilute nanocomposites with high CNT concentration, the NL particles are wrapped by CNTs, and the CNTs between neighboring NL particles and NL islands serve as bridges for electron transfer (Fig. 2G and fig. S7). When the NL particles assemble, CNTs still exist between natural rubber (NR) particles, preventing the NL particles from directly contacting each other (Fig. 2H). Thus, the obstacle of the conductive paths, caused by the complete electrical insulation in NL clusters, will not appear. In contrast, TEM images of the 1.1 wt % CNL solution show that the NL particles are not wrapped because of insufficient CNTs and the NL particles tend to agglomerate with each other (Fig. 2I). As a result, the conductive paths and connections between islands disappear.

Second, Fourier transform infrared spectroscopy confirms the strong connection between phospholipids and CNTs (Fig. 2J) (). Comparing the pure NL with CNL composite, the absorption peaks at 2960, 2913, and 2952 cm−1 shift to 2958, 2911, and 2950 cm−1, respectively, while the spectra of latex mixed with silver nanoparticles and graphene show no obvious changes within the range. These shifts indicate a strong interaction between CNTs and NL particle surroundings, enhancing the connections between phospholipids and CNTs (). CNTs wrapping NR particles are shown in Fig. 2K. To determine the effect of CNL bonding on the negative stretching-resistive effect, we subject NL to phospholipid loss and deproteinization to acquire the lipid-free deproteinized NR (P-DPNR; fig. S8). The P-DPNR signature is defined by the process cycle number. Control groups are fabricated by three types of P-DPNR with CNTs at the ratio of 5 wt %. In the TEM image (Fig. 2L), it can be seen that proteins are exfoliated from NL and CNTs are no longer attached to NL particles. As for the resistance variation of CNT–P-DPNR during strain, as predicted, the stretching-resistive effect switches from negative to positive after phospholipid loss and deproteinization due to the nonexistence of phospholipid-CNT bonding (Fig. 2M). The inset of Fig. 2M shows P-DPNR after centrifugation. Beside the two conditions, the orientation of CNTs is observed. Observed with TEM, the orientation of CNTs in the unstretched CNL membrane (Fig. 2N) is compared with that in the stretched CNL membrane (Fig. 2O).

Electrical and mechanical characteristics of the CNL membrane are measured with a tensile tester (Fig. 3A). The stress of the CNL membrane sample during strain of 100% is measured, and it can be found that, with the increasing concentration of CNT, Young’s modulus of the material tends to become larger (Fig. 3B). The stress growth with the increase in strain gradually becomes slow, indicating that the regular structure breaks severely at small strain. However, at large strain, the orientation of the NL and CNT dominates the stress change instead of regular structure breakage. The decrease in Young’s modulus conforms to the Payne effect (). Moreover, x-ray diffraction (XRD) patterns show that the crystallinity decreases from 11.4 to 9.2% after stretching, further verifying the decrease in Young’s modulus (fig. S9). For the CNL membrane with 30, 50, and 70% strain for continuous several cycles, the resistance response stays almost uniform. In addition, stage 1 of resistance increase at small strain is gradually weakened because the membrane has been further activated by decades of cycles in advance to mitigate the undesirable stage (Fig. 3C). The effects of parameters such as strain rate on the resistance response are ignorable (fig. S10).

Fig. 3.

Characteristics of CNL membrane.

(A) CNL membrane tested on the tensile tester. (B) Tensile properties of the CNL membrane with different concentrations. (C) Nine-cycle loop tests of CNL membrane under different strains with a consistent stretching velocity. (D) Combinations of different stretching directions and testing electrodes. Negative stretching-resistive effects are found in both isotropic tests, such as stretching and electrodes, performed (E) in the same direction and in the (F) vertical direction. (G) Resistance of 100-cycle stretching test of CNL membrane. (H) CNL membrane attached onto the knuckle. (I) Relative resistance change of CNL membrane with 30-cycle knuckle bending. Photo credit: Chunpeng Jiang, Shanghai Jiao Tong University.

Considering the influence of strain direction on the negative stretching-resistive effect, tests with combinations of different directions of stretching and electrodes are performed (Fig. 3D). Regardless of whether the stretching force and the testing electrodes are applied in the same direction or in the vertical direction, the membrane exhibits an isotropic negative stretching-resistive effect (Fig. 3, E and F). Consequently, a luminous device in series with the negative effect composite can be lighted by stretching in any direction, allowing a universal direction-insensitive application. The 100-cycle stretching test shows that the negative resistance variation tends to be stable (Fig. 3G). For the first several cycles, the resistance decreases sharply due to the plastic deformation of the membrane and the unrecoverable orientation and extruding of the CNT. The shoulder effect appears as the inset shows, which can be explained by the competition of the CNT breakage and reconstruction (, ). In addition, a long-term test of the CNL membrane with 2000 cycles is performed, verifying its reliable long-term performance (fig. S11). Afterward, the membrane is attached to a knuckle for 30-cycle bending (Fig. 3H). Obviously, the resistance reduces when bending the finger and the original value, which is in a reasonable and acceptable range, fluctuates due to wrinkling when straightening the finger (Fig. 3I). These results show the sensitivity and reliability of the negative stretching-resistive CNL composite, displaying high electrical conductivity during strain and mechanical functionality for practical applications.

Applications to cardiac optogenetics

The CNL membrane with negative stretching-resistive effect can effectively maintain electrical conductivity when stretched, so it is ideally suited for both strain measuring and controlling within a photoelectrical circuit, such as in the application of cardiac optogenetics. VT is a common complication of acute myocardial infarction (AMI). The most used method to treat VT, implantable cardioverter defibrillator (ICD), has the disadvantage of eliciting pain in patients due to electric shock and being nonselective for all tissues or nerves and causing irreversible damage in nontarget tissues and nerves (, ). To overcome these disadvantages, optogenetics is a promising method with its characteristics of cell selectivity, accuracy, controllability, and painlessness.

On the basis of the negative stretching-resistive CNL composites, a self-adaptive implantable cardiac optogenetics system with closed-loop heart rate recording and self-adaptive light intensity control is proposed for curing VT (Fig. 4A). The fan-shaped device with CNL membrane and LEDs wrapping around the heart by its stretchable structure and secured in place with biomedical glue and three buckles designed at the edge achieves good connection and effective sensing and control. It is worth mentioning that the CNL membrane has dual identities, one of which is to serve as a strain sensor to provide continuous closed-loop heart rate monitoring persistently. Moreover, the current change caused by the CNL membrane will regulate the light intensity around the AMI area, which is directly capable of inducing hyperpolarizing currents and silencing myocardial cells, thus restraining VT by light cardioversion.

Fig. 4.

Design and fabrication of closed-loop self-adaptive optogenetics system based on CNL membrane.

(A) Flexible and stretchable device based on negative stretching-resistive CNL composites and LEDs wrapped around the heart to cure VT by means of optogenetics. (B) Systematic design of the optogenetics system with a key function of closed-loop heart rate sensing and self-adaptive light stimuli array (red). (C) Schematic diagram of the flexible and stretchable structure, including three layers of polyimide (PI) and two layers of metal lines. (D) Photo of the fabricated system, including the flexible device and the processing circuit. (E) Enlarged photo showing the LED array and negative stretching-resistive sensors based on CNL membrane. Photo credit: Wen Hong, Shanghai Jiao Tong University.

In its application to cardiac optogenetics, the CNL membrane achieves the function of real-time detection and low energy consumption by closed-loop control (Fig. 4B). The closed-loop heart rate sensing module and self-adaptive LED stimuli array are the two key components of the system. Analog signals read out from the negative stretching-resistive strain sensors are firstly filtered and amplified. Then, the regular voltage variation caused by diastole (V_sense) in each heartbeat is converted to digital signal (Digital_pulse) to calculate the estimated heart rate. A signal processing circuit continually monitors the heart rate to raise an alarm once the heart rate exceeds the threshold of VT. According to the above closed-loop heart rate sensing mechanism, a self-adaptive light stimuli array will be enabled at the first time. The self-adaptive light intensity depends on both the degree and frequency of heartbeat. The negative stretching-resistive sensors functioning as an adjustable component could directly add a current increment to the driving current in every cycle of diastole, which can be modeled as a second driving module in series with the original module. The proposed self-adaptive system including negative stretching-resistive strain sensors is compared with the traditional cardiac optogenetics device (fig. S12), illustrating its advantages of integrated and multifunctional detection, lower energy consumption, the improved cure efficiency of VT, and a stable performance immune to the electromagnetic interference (table S2).

The flexible and stretchable self-adaptive LED stimuli array is based on a wandering substrate, which includes three layers of polyimide (PI) and two layers of metal lines (Fig. 4C). The CNL membrane is riveted and the LEDs are welded to the flexible substrate (fig. S13). All these connecting positions are encapsulated by Ecoflex for isolation and biocompatibility (fig. S14). The flexible device based on the negative stretching-resistive strain sensor and LEDs with high light intensity is then connected to the processing circuit by flexible printed circuit (FPC) (Fig. 4, D and E).

A light-sensitive heart is injected with AAV2/9-CAG-ArchT-GFP virus in advance (see Materials and Methods and fig. S15). During the in vivo test (fig. S16), four-channel negative-strain sensors are set on the location of the largest heart expansion to monitor the normal rhythm of diastole or VT in every cycle (figs. S17 and S18). The measured heart rate tested is close to 120 beats/min (Fig. 5A), which is much higher than the standard values (fig. S19), so the self-adaptive LED stimuli array is activated immediately to realize the optogenetics therapy (Fig. 5B and movie S3). The partial enlarged image of the lighted LED and negative stretching-resistive strain sensor show that the device fits the heart well by the structural design and biomedical glue (Fig. 5, C and D). Meanwhile, because the resistance variation of the CNL membrane worked as a stimulation of the current-type LED component, the light intensity is accommodated by heart diastole control. An adjustable duty cycle method is adopted in different periods of heart beat: 13% at the stage of diastole and 40% at the stage of systole, rather than a uniform and high duty cycle of 40% all the time (Fig. 5E) (). Therefore, the radiance of the proposed device obviously increases from 74.94 to 89.41 mW/mm2 in the period of diastole due to the presence of negative stretching-resistive sensor (Fig. 5F). Last, with the treatment of the self-adaptive LED stimuli, VT was restrained, and the CL increased. The average CLs of 10 heart beats were separately 118, 129, 130, and 132 ms originally. After about only an 8-min treatment, the average CL raised to 181, 185, and 187.5 ms, showing a shorter treatment time than the previous report of 30 min and confirming the effectiveness of the closed-loop control by CNL strain sensor (Fig. 5G) (). In addition, a control in vivo experiment on animals without the ArchT has been performed to exclude effects of heating and electrical pulses, in which the heart rate turned to be faster (CL decreases from 118 to 98 ms) in 2.5 min even if the LEDs are illuminated with the same frequency and duty cycle (fig. S20).

Fig. 5.

In vivo experiment of optogenetics system based on negative stretching-resistive strain sensor array.

(A) Analog strain signal (V_sense, left axis) obtained from the negative stretching-resistive strain sensor array and the corresponding digital pulses (right axis) related with every heartbeat, showing an estimated heart rate close to 120 beats/min. (B) Activated self-adaptive LED stimuli array. (C) Enlarged photo of luminous LEDs. (D) Enlarged photo of negative stretching-resistive sensor and LEDs wrapping tightly on the heart. (E) The duty cycle settings of the pulse width modulation (PWM) wave of the optogenetics device with single LED and LED in series with negative or positive stretching-resistive sensor, respectively. (F) Radiance of the LED without and with negative stretching-resistive strain sensor in different periods of systole and diastole. (G) Average CL before and after optogenetics treatment. (H) Device wrapped on the heart with strain sensor. (I) Radiance of single LED and LED with series connection of the unstretched and stretched CNL membrane. (J) Immunofluorescent analysis of the green fluorescent protein (GFP)/ArchT transfection. Photo credit: Wen Hong, Shanghai Jiao Tong University.

In the postanalysis of the animal experiments, it is observed that the optogenetics device provides good conformality to the heart (Fig. 5H). The long-term stability of the proposed device is validated by an accelerated aging experiment at 57°C in 10× phosphate-buffered saline (PBS) for 3 weeks, showing a small decreased relative resistance variation of 4.68% within the acceptable range and the same performance of illumination as desired (fig. S21). The biocompatibility has been further explored by a cell viability test, showing good biocompatibility of the CNL film sealed by Ecoflex and the Ecoflex itself (fig. S22). Meanwhile, radiance tests of a single LED and LED in series with unstretched and stretched CNL membranes are performed (Fig. 5I). Radiance at the stimuli wavelength of 565 nm indicates sufficient luminous power per unit area to activate the transfected cells. Moreover, the tissues injected with virus, which have been whitened a little bit (fig. S23), are used for immunofluorescent analysis. The slice shows that green fluorescent protein (GFP)/ArchT virus is mainly expressed in the cardiac muscle, verifying the successful GFP/ArchT transfection to the tissue (Fig. 5J).

DISCUSSION

We have introduced a self-adaptive implantable cardiac optogenetics system based on an original negative stretching-resistive strain sensor array for closed-loop heart rate recording and self-adaptive light intensity control. The controllable negative resistance variation over a wide strain range makes the sensor suitable for advanced circuits requiring simultaneous precise strain sensing and physiological-electrical-optical regulation. Systematic in vivo validation of cardiac optogenetics with negative stretching-resistive CNL membrane provides valuable insights on closed-loop, self-adaptive, and low-power light cardioverters. The robust mechanical performance of the negative stretching-resistive sensor with enhanced conductivity offers a promising mode for electrical communication and light conversion within arbitrary controlling-in-sensor stretchable bioelectronics.

MATERIALS AND METHODS

Fabrication of the membrane

First, the CNT dispersion solution (XFZ29, Nanjing XFNANO Materials Tech Co. Ltd., China) is processed with a 195-W ultrasound probe for 10 min in ice bath. Then, the processed CNT solution and NL (001a, Maoming Zhengmao Petrochemical Co. Ltd., China) are mixed by magnetic stirring in ice bath for 6 hours. The weight ratio of CNT solution and NL solution is set as 1:0.25, 1:0.5, and 1:1, respectively. Last, the mixed solution is disposed into fabricated Teflon molds and heated under 80°C for 12 hours to fabricate the strain sensitive membrane. Zeta potential diagram of solution indicates the homogeneity of the mixed solution (fig. S24).

Electrical and mechanical characterization

The membrane is stretched by dynamic thermomechanical analysis (Q850, TA Instruments) for stretching-resistive property. Loop test is performed by tensile tester (AGS-X 50N, SHIMADZU). Meanwhile, copper conductive tape (3M) is applied onto each ends of the membrane. The resistance of the membrane under different strains is measured by LCR meter (4100, Wayne Kerr). Necessary isolation is taken to avoid conductance to the environment.

Surface, components, and structural analysis

To analyze the special performance of the CNL membrane, SEM (ZEISS Ultra Plus), TEM (TALOS F200X, FEI), XRD (Bruker D8 ADVANCE Da Vinci), and infrared spectrum (Nicolet 6700, Thermo Fisher Scientific) are performed. The sample for SEM photo of single NL particle is prepared by 100 times dilution of NL solution and then heated in 80°C for 12 hours. The ultrathin TEM membrane sample (thickness of 60 μm) of CNL film is prepared by liquid nitrogen and then be put on the copper grid. The TEM sample of CNL solution is prepared by 100 times dilution and phosphotungstic acid dyeing on carbon grid.

Preparation of lipid-free deproteinized natural rubber

The NL is named as solution A. 0.2 wt % of primary alcohol ethoxylate, 0.04 wt % of basic proteinase, and 0.02 wt % of phospholipase are mixed and named as solution B. Solutions A and B are then mixed for thermostatic mixing for 6 hours and are separated into three parts evenly before centrifugation. Next, ironized water is added into the dry NL from the first part to get P-DPNR-1; Solution B is added into the dry NL from the second part to get P-DPNR-2; 10 wt % of solution B is added into the third part for centrifugation to get the dry NL and then solution B is added again to get P-DPNR-3. The CNT-P-DPNR membrane is fabricated with the same methods as CNL membrane.

Fabrication of optogenetics device

First, the flexible PI-Au–based substrate includes five layers, which, from bottom to the top, are PI substrate, metal-1, PI isolation, metal-2, and top PI isolation. This substrate is fabricated by lithography, wetting etching, and laser beam cutting. On the substrate, the LEDs with the area of 2.2 mm2 are welded by stencil printing, and the radiance is tested by integrating sphere (FOIS-1, Ocean Optics). The CNL membranes are riveted. Then, a specially designed wide-line FPC is connected to the substrate by anisotropic conductive film. Last, all these connecting points are sealed by Ecoflex, which is mixed by part A and part B with ratio of 1:1, and heated for 1 hour at 80°C.

Design of self-adaptive circuit

The closed-loop heart rate sensing module and self-adaptive LED stimuli array are integrated into a single-printed circuit board with voltage plane clearance. Every CNL-based strain sensor with a magnitude-controllable pull-up resistor (AD5227, Analog Devices) offers a sensitive input signal. Several low noise complementary metal-oxide semiconductor operational amplifiers (AD8692, Analog Devices) and Schmitt-Trigger inverters (SN74HC14, Texas Instrument) have been used to achieve analog signals to digital conversion. The processing unit (ESP32-PICO-D4, Espressif Systems Co. Ltd) is used in this study. To obtain sufficient and stable light intensity, commercial LED driver chips (TPS92200, Texas Instruments) provide the fixed reference current of LEDs.

Performance of in vivo experiments

A male beagle, number of quality certification 2200006000187, weighing 15 kg, is provided by Shanghai Jiagan Biotechnology Co. Ltd. The animal experiment is performed under the ethical review number of GJDS20201208-1. First, at the process of thoracotomy, the left coronary artery anterior descending branch ligated to establish the AMI model. After that, the virus AAV2/9-CAG-ArchT-GFP [OBiO Technology (Shanghai) Co. Ltd] is injected by 50-μl microneedle, micropump (model: LSP02-1B, LongerPump), capillary, and adjustable microneedle. All viruses are injected at 11 sites; among them, 8 sites distribute around the predicted AMI area, 2 sites are near the blood vessel, and 1 near the tip of the heart. At each site, 5 μl of virus is injected at the speed of 1 μl/min. Then, the chest is stitched. After 4 weeks of virus expression, the self-controlled optogenetics device is immersed into the ethyl alcohol solution before wrapped around the heart by surgical suture for testing. VT model is determined from left ventricular apex, left ventricular base, and the mid left ventricle by programmed stimulation. The electrocardio signal is recorded by electrophysiology recording system (LEAD-EP, JJET, China). The control experiment was performed as the same method by the virus AAV2/9-CAG-GFP [OBiO Technology (Shanghai) Co. Ltd].

Immunofluorescent analysis by frozen slides

The potential expression tissues are cut from the heart and immersed into the formalin solution after the in vivo experiment. The whole tissue is cut into pieces of around 10 mm by 10 mm. Then, slice up each piece from the top to the bottom with each slice thickness of 8 μm by freezing microtome section (Cryotome E, Thermo Fisher Scientific) and take one slice for 4′,6-diamidino-2-phenylindole (G1012, Servicebio) and observation (NIKON ECLIPSE C1, NIKON DS-U3) every five slices.

Accelerated aging experiment

The accelerated aging experiment is designed as the following process: immersing the whole device into the 10× PBS at 57° ± 5°C for 3 weeks.

Biocompatible test

The CNL film encapsulated by the Ecoflex and the Ecoflex film itself were incubated with the neuronal HT-22 cells in a six-well plate of Dulbecco’s modified Eagle’s medium (DMEM) solution at 37°C and 5% CO2 for 36 hours, respectively. The cell viability test is grouped as follows: the group with the CNL membrane sealed by Ecoflex (group 1), the group with bare Ecoflex film (group 2), and the control group that cells are incubated with DMEM only. After washing thoroughly with PBS buffer, the morphology of the cells was visualized by the microscopy (CKX 53, Olympus).