Soft, low-power, biomolecular memristors leverage similar composition, structure, and switching mechanisms of bio-synapses. Presented here is a protocol to assemble and characterize biomolecular memristors obtained from insulating lipid bilayers formed between water droplets in oil. The incorporation of voltage-activated alamethicin peptides results in memristive ionic conductance across the membrane.
The ability to recreate synaptic functionalities in synthetic circuit elements is essential for neuromorphic computing systems that seek to emulate the cognitive powers of the brain with comparable efficiency and density. To date, silicon-based three-terminal transistors and two-terminal memristors have been widely used in neuromorphic circuits, in large part due to their ability to co-locate information processing and memory. Yet these devices cannot achieve the interconnectivity and complexity of the brain because they are power-hungry, fail to mimic key synaptic functionalities, and suffer from high noise and high switching voltages. To overcome these limitations, we have developed and characterized a biomolecular memristor that mimics the composition, structure, and switching characteristics of biological synapses. Here, we describe the process of assembling and characterizing biomolecular memristors consisting of a 5 nm-thick lipid bilayer formed between lipid-functionalized water droplets in oil and doped with voltage-activated alamethicin peptides. While similar assembly protocols have been used to investigate biophysical properties of droplet-supported lipid membranes and membrane-bound ion channels, this article focuses on key modifications of the droplet interface bilayer method essential for achieving consistent memristor performance. Specifically, we describe the liposome preparation process and the incorporation of alamethicin peptides in lipid bilayer membranes, and the appropriate concentrations of each constituent as well as their impact on the overall response of the memristors. We also detail the characterization process of biomolecular memristors, including measurement and analysis of memristive current-voltage relationships obtained via cyclic voltammetry, as well as short-term plasticity and learning in response to step-wise voltage pulse trains.
It is widely recognized that biological synapses are responsible for the high efficiency and enormous parallelism of the brain due to their ability to learn and process information in highly adaptive ways. This coordinated functionality emerges from multiple, highly complex molecular mechanisms that drive both short-term and long-term synaptic plasticity1,2,3,4,5. Neuromorphic computing systems aim to emulate synaptic functionalities at levels approaching the density, complexity, and energy efficiency of the brain, which are needed for the next generation of brain-like computers6,7,8. However, reproducing synaptic features using traditional electronic circuit elements is virtually impossible9, instead requiring the design and fabrication of new hardware elements that can adapt to incoming signals and remember information histories9. These types of synapse-inspired hardware are known as mem-elements9,10,11 (short for memory elements), which, according to Di Ventra et al.9,11, are passive, two-terminal devices whose resistance, capacitance, or inductance can be reconfigured in response to external stimuli, and which can remember prior states11. To achieve energy consumption levels approaching those in the brain, these elements should employ similar materials and mechanisms for synaptic plasticity12.
To date, two-terminal memristors13,14,15 have predominantly been built using complementary metal-oxide-semiconductor (CMOS) technology, characterized by high-switching voltages and high noise. This technology does not scale well due to high power consumption and low density. To address these limitations, multiple organic and polymeric memristors have been recently built. However, these devices exhibit significantly slower switching dynamics due to time-consuming ion diffusion through a conductive polymer matrix16,17. As a result, the mechanisms by which both CMOS-based and organic memristive devices emulate synapse-inspired functionalities are highly phenomenological, encompassing only a few synaptic functionalities such as Spike Timing Dependent Plasticity (STDP)18, while overlooking other key features that also play essential roles in making the brain a powerful and efficient computer, such as pre-synaptic, short-term plasticity19.
Recently, we introduced a new class of memristive devices12 featuring voltage-activated peptides incorporated in biomimetic lipid membranes that mimics the biomolecular composition, membrane structure, and ion channel triggered switching mechanisms of biological synapses20. Here, we describe how to assemble and electrically interrogate these two-terminal devices, with specific focus on how to evaluate short-term plasticity for implementation in online learning applications12. Device assembly is based on the droplet interface bilayer (DIB)21 method, which has been used extensively in recent years to study the biophysics of model membranes21 and membrane-bound ion channels22,23,24, and as building blocks for the development of stimuli-responsive materials25,26. We describe the membrane assembly and interrogation process in detail for those interested in neuromorphic applications but have limited experience in biomaterials or membrane biology. The protocol also includes a full description of the characterization procedure, which is as important as the assembly process, given the dynamic and reconfigurable electrical properties of the device27. The procedure and representative results described here are foundations for a new class of low-cost, low-power, soft mem-elements based on lipid interfaces and other biomolecules for applications in neuromorphic computing, autonomous structures and systems, and even adaptive brain-computer interfaces.
1. General Instructions and Precautions
2. Preparation of Aqueous Buffer Solution
3. Preparation of Liposomes
NOTE: Step 3.1 only applies if phospholipids are acquired as lyophilized powders, and therefore, may be skipped if the phospholipids are purchased in chloroform.
4. Reconstitution of Alamethicin Peptides
NOTE: This procedure describes the process of alamethicin reconstitution in liposomes to a final concentration of 1 μM. This concentration is sufficient to induce nA-level currents similar to those previously published12. Increasing the peptide concentration will reduce the switching threshold and increase the amplitudes of currents induced by applied voltage29.
5. Preparation of Agarose gel
6. Fabrication of the Oil Reservoir
NOTE: The procedure described below is just one of many ways that an oil reservoir can be fabricated. The reader is encouraged to design and fabricate a reservoir based on available materials, machining capabilities, and specific needs.
7. Preparation of Electrodes
8. Setting Up the Experiment
9. Proper Grounding to Reduce Electrical Noise
10. Feedback-Controlled Heating
11. Setting Up the Software and Equipment
12. Pipette Offset
NOTE: The procedure described below applies only to current amplifier mentioned in Table of Materials.
13. Formation of the Lipid Bilayer
14. Electrical Characterization of the Biomolecular Memristor
Figure 1 displays the experimental setup used to assemble and characterize the biomolecular memristor. Lowering the free ends of the electrodes to the bottom of the oil reservoir, as shown in Figure 1b, was found helpful to minimize vibrations of the electrodes and droplets that can result in variations in measured current and bilayer area, especially in cases where heating the oil can generate convective flow in the oil. Figure 2 shows the procedure and result of assembling the Ag/AgCl wires, class capillaries, and electrode and micropipette holders. The setup is housed within a properly grounded Faraday cage (Figure 3) to minimize electromagnetic interference.
It is imperative to form a stable, insulating lipid bilayer for this study. In this protocol, a lipid monolayer assembles at the oil/water interfaces of the aqueous droplets immersed in oil. Upon contact between droplets, excess oil is excluded, and the opposing lipid monolayers thin to a 5-nm thick lipid bilayer. The most common technique used in bilayer electrophysiology is voltage-clamp, where the voltage across the bilayer is controlled and the induced current is measured. Figure 4a portrays the capacitive square-wave current induced by a 10 mV, 10Hz voltage during bilayer formation. While the amplitude increases upon the start bilayer thinning and subsequent radial expansion of the thinned membrane, the waveform remains square. Using the steady-state amplitude of square wave current, the nominal area of the lipid bilayer can be calculated using a predetermined value of specific membrane capacitance for a DPhPC bilayer21. Also, the bilayer area can be visually assessed by measurement of the bilayer diameter from an image taken with the microscope Figure 4b. For accurate lipid bilayer area calculations, the reader should refer to Taylor, et al.21. The area of the lipid bilayer can be adjusted by changing the relative positions of the droplets21,31.
Upon application of a voltage bias to an alamethicin-free lipid bilayer, the current response will vary based on the frequency of the input voltage. At low frequencies (<10-50 mHz), where the resistance of the bilayer dominates the complex impedance, the ohmic current response is negligible because the nominal membrane resistance is typically greater than 10 GΩ. As the input frequency increases, membrane capacitance contributes more to the impedance of the system, resulting in the non-zero current response displayed in the plot of current versus voltage in Figure 5a. When the same input voltage waveform (150 mV) is applied to a biomolecular response consisting of an alamethicin-doped lipid membrane, and when the voltage amplitude surpasses a critical insertion threshold (~ 100 mV for a DPhPC membrane at room temperature), alamethicin peptides residing at the surface of the lipid bilayer insert into the membrane and aggregate to form conductive pores. The threshold-dependent formation of ion channels results in a nonlinear macroscopic current response, with exponentially increasing currents at voltages higher than the insertion threshold (Figure 5b). While alamethicin peptides are known to form rectifying ion channels only at sufficiently positive voltages, the symmetric nature of these current responses at both polarities is due to the insertion and aggregation of separate populations of peptides, each from opposite sides of the membrane. Depending on the frequency of the applied voltage, the induced current response may also contain contributions from the capacitive current. Therefore, the capacitive current in Figure 5a must be subtracted from the total current displayed in Figure 5b to obtain only the memristive pinched hysteresis current-voltage response, displayed in Figure 5c, d.
Figure 6 displays the dynamic switching response of a biomolecular memristor induced by a voltage pulse train (130 mV (HIGH), 20 mV (LOW), 100 ms (ON), 20 ms (OFF)). The OFF voltage is chosen to be 20 mV to differentiate the return of the device to an insulating state as alamethicin channels leave the bilayer rather from current simply vanishing at zero-voltage input. The cumulative increase in ON-state current during successive voltage pulses represents paired-pulsed facilitation, a plasticity that volatile biomolecular memristors are capable of exhibiting12.
Figure 1: Experimental setup and main parts. (a) The standard workstation for assembling and characterizing the biomolecular memristor includes an inverted microscope, 3-axis micromanipulators, a digital camera, a vibration isolation table, an electrode holder, a glass micropipette holder, a current amplifier, a function generator, and an oil reservoir. The setup is assembled on the stage of the microscope as described in Steps 11-13. (b) A zoomed-in photograph of the setup showing the tips of the Ag/AgCl wires touching the bottom of the oil reservoir. Please click here to view a larger version of this figure.
Figure 2: Electrode preparation procedure. Photographs showing: (a) the silver wires soaking in bleach; (b) an electrode holder; (c) a 5 cm long glass capillary connected to the electrode holder; (d) a Ag/AgCl electrode fed through the glass capillary; (e) a glass micropipette holder; and (f) the fully assembled electrodes and holders. Please click here to view a larger version of this figure.
Figure 3: Grounding procedure. Photographs showing: (a) a screw threaded into the vibration isolation table surface to create a Ground bus when connected to earth ground; and (b) a lab-made Faraday cage covering the oil reservoir and electrode setup to shield the measurement from electromagnetic interference. Both the cage and microscope stage are tied to the Ground bus via cables I and II. Please click here to view a larger version of this figure.
Figure 4: Real-time current measurements show initial bilayer thinning and areal growth. (a) The current measured (top) during spontaneous bilayer formation between lipid-coated droplets in response to a triangular waveform voltage. The magnitude of the measured current is directly proportional to the capacitance of the interface and hence, the area of the bilayer. The area of the interface can be varied by changing the distance between the two droplet-bearing electrodes. (b) An image acquired through the inverted microscope shows a bottom view and dimensions of a typical membrane-based biomolecular memristor. Please click here to view a larger version of this figure.
Figure 5: Current-voltage relationship and pinched hysteresis. (a) The current-voltage responses of an alamethicin-free DPhPC lipid bilayer. A lipid-only membrane is highly insulating (~10 GΩ), which explains the low ohmic current response at 0.017 Hz, a frequency where the impedance is dominated by membrane resistance. At higher frequencies, membrane capacitance contributes more significantly to the total impedance of the interface, resulting in a non-zero induced capacitive current. (b) The dynamic current-voltage relationships versus frequency of a lipid bilayer formed between two droplets containing alamethicin peptides (obtained with a triangular input wave). (c) The memristive, pinched hysteretic current response of the device is obtained by subtracting the capacitive current displayed in a from the total current displayed in b. (d) Zooming-in to highlight the differences between the total and the memristive currents. Please click here to view a larger version of this figure.
Figure 6: Response of the biomolecular memristors to rectangular voltage pulses and plasticity. The device responds to subsequent voltage pulses with an increase in conductance during the ON time, despite intermittently restoring an insulating state during each OFF time. The increase in current from pulse to pulse shows that the instantaneous conductance of the device is a function of both the present stimulus and prior stimuli, analogous to short term plasticity in bio-synapses. Please click here to view a larger version of this figure.
This paper presents a protocol for assembling and characterizing biomolecular memristors based on ion channel-doped synthetic biomembranes formed between two droplets of water in oil. The soft-matter, two-terminal device is designed and studied to: 1) overcome constraints that are associated with solid-state technology, such as high noise, high energy consumption, and high switching voltages, 2) more closely mimic the composition, structure, switching mechanisms of biological synapses, and 3) explore the mechanisms and features of bio-synapse plasticity that are not exhibited by solid-state devices.
The droplet interface bilayer technique21, which represents the building block of the present technology12, is a simple, modular approach for membrane assembly that has been extensively used to study membrane biophysics21, proteins22, ion channels29, and other biomolecules32. It offers specific advantages for precisely controlling and interrogating model membranes, and represents a building block for stimuli-responsive and autonomous materials26. Multiple methods have been developed to assemble droplet interface bilayers, including the hanging drop21 method which was adapted as the main method to develop and characterize the biomolecular memristor. Even though this membrane assembly technique was used in previous studies, here we present a thorough protocol that allows researchers to recreate and study memristive droplet interface bilayers in their own laboratories. The protocol is specifically written in a way to allow researchers in non-membrane biology fields, such as the neuromorphic community, to understand and recreate these procedures.
In its simplest form, the protocol we have described herein for assessing memristive functionalities of a biomembrane can be replicated with basic laboratory equipment such as a function generator, a microscope, and a current measuring system. The assembled device is electrically equivalent to a resistor (~10 GΩ) and a capacitor wired in parallel. In the presence of peptides, such as alamethicin, that are capable of forming voltage-dependent pores in the membrane, the membrane resistance significantly drops, and resistive current can be detected in response to input voltage signals (DC or AC). However, the large membrane resistance and frequency-dependent electrical impedance of the device mean that: 1) induced currents are small (pA-nA), and subject to electromagnetic interference; and 2) care must be taken to accurately induce and measure the desired memristive properties separate from capacitive membrane responses, respectively. In response to an AC voltage, and depending on the frequency of the signal, the recorded current will contain both capacitive and resistive components. To achieve the pinched hysteresis, which is a signature of memristive device, one must follow the protocol described in Step 14. The hanging wires are susceptible to vibrations, which can result in artefactual responses such as oscillations mistakenly attributed to the actual dynamics of the device. Positioning the wires at the bottom of the oil reservoir ameliorates this behavior.
The biomolecular memristor with its current structure and design emulates the short-term synaptic plasticity that occurs in the presynaptic terminal. It also mimics some of the mechanisms that cause presynaptic paired pulsed facilitation in the brain due to the accumulation and depletion of neurotransmitter vesicles in the presynaptic neuron. This methodology for assembling synaptic mimics enables the study and validation of biomimetic processes responsible for many types of short-term plasticity, and the optimization of modularity and scalability not possible with other technologies33. Unforeseen functionality may be discovered by either modifying the membrane composition, the types of ion channels that are incorporated into the membrane, and even the number of connected droplets and interfacial membranes constituting each two-terminal device. As an example, we have recently demonstrated the online learning capabilities of the biomolecular memristor by interfacing it with a solid-state neuron34.
The authors have nothing to disclose.
Financial support was provided by the National Science Foundation Grant NSF ECCS-1631472. Research for G.J.T., C.D.S., A.B., and C.P.C. was partially sponsored by the Laboratory Directed Research and Development Program of Oak Ridge National Laboratory, managed by UT-Battelle, LLC, for the U.S. Department of Energy. A portion of this research was conducted at the Center for Nanophase Materials Sciences, which is a DOE Office of Science User Facility.
1,2-diphytanoy-sn-glycero-3-phosphocholine (DPhPC) | Avanti Polar Lipids | 850356P/850356C | Purchased as lyophilized powder (P) or in chloroform (C) |
Agarose | Sigma-Aldrich | A9539 | |
Agarose (0.5g Agarose Tablets) | Benchmark | A2501 | You can either use the powder form or the tablets |
Alamethicin | AG Scientific | A-1286 | |
Analytical balance | Mettler Toledo | ME204TE/00 | |
Axopatch 200B Amplifier | Molecular Devices | – | |
BK Precision 4017B 10 MHz DDs Sweep/Function Generator | Digi-Key | BK4017B-ND | |
Borosilicate Glass Capillaries | World Precision Instruments | 1B100F-4 | |
Brain Total Lipid Extracts (Porcine) | Avanti Polar Lipids | 131101 | |
DigiData 1440A system | Molecular Devices | – | |
Extruder Set With Holder/Heating Block | Avanti Polar Lipids | 610000 | This includes a mini-extruder, 2 syringes, 100 PC membranes, 100 filter supports, and 1 holder/heating block |
Freezer (-20 °C) | VWR International | SCUCBI0420AD | |
Glassware | VWR International | – | |
Hexadecane, 99% | Sigma-Aldrich | 544-76-3 | |
Isopropyl Alcohol | VWR International | BDH1133-4LP | |
Microelectrode Holder | World Precision Instruments | MEH1S | |
MOPS | Sigma-Aldrich | M1254 | |
Nitrogen (N2) Gas | Airgas | UN1066 | |
Parafilm M All-Purpose Laboratory Film | Parafilm | PM999 | |
Powder Free Soft Nitrile Examination Gloves | VWR International | CA89-38-272 | |
Precleaned Microscope Sildes | Fisher Scientific | 22-267-013 | |
Refrigirator (4 °C) | VWR International | SCUCFS-0504G | |
Silver wire | GoodFellow | 147-346-94 | Different diameters could be used depending on the application |
Sodium Chloride (KCl) | Sigma-Aldrich | P3911 | |
Stirring Hot Plate | Thermo Scientific | SP131325 | |
VWR Light-Duty Tissue Wipers | VWR International | 82003-820 | |
VWR Scientific 50D Ultrasonic Cleaner | VWR International | 13089 |