Action Potential Monitoring Using Neuronanorobots: Neuroelectric Nanosensors

View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by Universidade do Minho: RepositoriUM International Journal of Nanomaterials and Nanostructures Vol. 1: Issue 2 www.journalspub.com

Nuno R. B. Martins1*, Wolfram Erlhagen1, Robert A. Freitas Jr2 1Center for Mathematics (CMAT), University of Minho, Campos de Azurém, Guimarães, 4800-058,Portugal. 2Institute for Molecular Manufacturing, 555 Bryant Street, Suite 354, Palo Alto, California 94301, USA

Abstract Neuronanorobotics, a key future medical technology that can enable the preservation of human brain information, requires appropriate nanosensors. Action potentials encode the most resource-intensive functional brain data. This paper presents a theoretical design for electrical nanosensors intended for use in neuronanorobots to provide non-destructive, in vivo, continuous, real-time, single-spike monitoring of action potentials initiated and processed within the ~86 x 109 neurons of the human brain as intermediated through the ~2.4 x 1014 human brain synapses. The proposed ~3375 nm3 FET-based neuroelectric nanosensors could detect action potentials with a temporal resolution of at least 0.1 ms, enough for waveform characterization even at the highest human neuron firing rates of 800 Hz.

Keywords: Human brain information, nanorobot, nanorobotics, neuronanorobot, neuronanorobotics, nanomedicine, nanosensor, nanotechnology, synaptobot, endoneurobot

*Author for Correspondence: Email ID: [email protected]

Abbreviations: AIS - axon initial segment, BOINC - barcoding of individual neuronal connections, CNT - carbon nanotube, DNA - deoxyribonucleic acid, DWCNT - double-wall carbon nanotube, FET - field-effect transistor, MRI - magnetic resonance imaging, RESCOP - redundant-skeleton consensus procedure, SNR - signal-to-noise ratio, SWCNT - single-wall carbon nanotube, VSNP - voltage sensing inorganic nanoparticle.

INTRODUCTION brain in real-time, in vivo, with adequate Information pertaining to brain neural temporal and spatial resolution. connectivity (e.g., the connectome) and the associated electrical action potential Technology capable of providing whole activity at the cellular and subcellular human brain, non-destructive, in vivo, real- level, together with other sources of brain time, functional information with adequate structural and functional information, temporal and spatial resolution will have underlies higher mental states and several specific requirements. Such individuality. This information can be lost technology would have to monitor, among as a result of physical trauma, pathogenic diseases, and a variety of degenerative other brain data, all action potential based disorders. Current medical technology for functional data traffic passing through brain information scanning, either (86.06 ± 8.2) x 109 human brain neurons[1] destructive or non-destructive in nature, and (2.42 ± 0.29) x 1014 human brain cannot monitor the structural and synapses[2], accurately recording functional information of a whole human synaptically-processed (4.31 ± 0.86) x 1015 spikes/sec[2]. Accomplishing this objective

IJNN (2015) 20–41 © JournalsPub 2015. All Rights Reserved Page 20 Action Potential Monitoring Using Neuronanorobots Martins et al. ______will require appropriate sensing, Alzheimer’s diseases, other brain-related communication and hardware neurodegenerative disorders, epilepsy, infrastructure to handle an estimated dementia, memory and sensory disorders, neuroelectric data rate of (5.52 ± 1.13) x spinal cord and neuromuscular disorders, 1016 bits/sec for the entire living human pain and toxic disorders, and a wide brain[2]. This data rate appears necessary variety of traumatic injuries to the brain. to capture even the fastest firing rates in Non-medical applications of this the 400–800 Hz range from fast spiking promising technology include the neurons[4, 5] and eventually to characterise possibility of becoming the virtually even the fastest voltage velocities at 20 perfect brain-machine interface technology mV/ms[6]. Another requirement is the necessary to finally bridge human brain ability to transmit this huge data flow into and machine[33]. an external supercomputer, possibly using an in vivo fiber network[7] capable of The advent of medical neuronanorobotics handling 1018 bits/sec of data traffic[7, 2]. requires the ability to build nanorobotic Such a fiber network may occupy 30 cm3 devices and to produce these devices in and generate 4–6 W of waste sufficient therapeutic quantities to treat heat[7]. Ideally the transit time from signal individual patients. The most advanced origination inside the human brain to the neuronanorobots will likely be fabricated external computer system through such a using diamondoid materials, because these network would have negligible signal materials provide the greatest strength, latency in comparison to the action durability, and reliability in the in vivo potential waveform temporal resolution[7]. environment and have good biocompatibility[9, 20]. Possible methods to Medical nanorobotics offers an ideal achieve massively parallel molecular technology for monitoring, recording, and manufacturing technologies, such as a even manipulating many of the different nanofactory, have been reviewed in the types of brain-related information, in literature[38, 20, 39], and methods for particular functional action potential based controlling individual and large numbers electrical information[8, 7, 9, 10, 11, 12, 13, 14, 15, of medical nanorobots are also the subject 16, 17, 18]. Medical nanorobotics has received of current research[34, 35, 36]. An ongoing preliminary technical exploration[7, 9, 19, 20] international collaboration is pursuing the and there are several detailed theoretical objective of constructing a nanofactory designs for a variety of medical capable of mass-manufacture medical nanorobots[8, 21, 22, 23, 24, 25, 20, 27]. diamondoid nanorobotics devices for medical treatments[37, 20, 39]. Neuronanorobots, a specific class of medical nanorobots, are expected to permit Neuronanorobotic sensors are a key in vivo, whole-brain, real-time monitoring technology for all subclasses of of single-neuron neuroelectric activity and neuronanorobots, especially for local neuropeptide traffic, permitting also endoneurobots (neuron-resident robots) the acquisition of all relevant structural and synaptobots (synapse-monitoring information including neuron surface robots). Appropriate monitoring of the features and connectome mapping [28, 29, 30, different types of functional human brain 31, 32]. Non-destructive whole-brain information requires nanosensors with monitoring would be enabled by the crucial performance characteristics, coordinated activities of large numbers of including: appropriate dynamic range to cooperating neuronanorobots. Medical capture the entire signal amplitude, high- neuronanorobotics might be the ultimate accuracy as an appropriate percentage of technology needed to treat Parkinson’s and full scale output, high-sensitivity, small

hysteresis permitting good discrimination highest spatial resolution being 2 µm, still between similar inputs, low output noise not enough to detect most synapses. The compared with the fluctuation in the state-of-the-art tomographic Nano-CT physical signal, good resolution for scanners (the Micro-CT successor) achieve measuring the minimum detectable signal structural resolutions between 50–500 nm. fluctuation with good margins of safety, Three Nano-CT scanners are commercially and appropriate bandwidth with fast available today: the Nanotom, the response time to a rapid change in physical SkyScan-2011, and the Xradia nanoXCT. signal. The Nanotom provides a resolution of ~500 nm pixels, and handles maximum After a brief survey of contemporary brain object size of 150 mm height and 120 mm scanning techniques (Section 2) and diameter[45] (roughly the size of a whole specific action potential measurement human brain). The SkyScan-2011 has a requirements (Section 3), we review slightly better resolution of ~400 nm appropriate sensor choices (Section 4) and pixels with similar object size constraints. then provide a preliminary design for a The Xradia nanoXCT claims to be capable specific nanorobot sensor (Section 5) that of providing a spatial resolution between is intended for use in endoneurobots and 50–300 nm[46]. Nano-CT scanner synaptobots performing real-time resolutions might permit extraction of monitoring of in vivo action potentials. some cellular detail and eventually Nanosensor biocompatibility is briefly identification of some synapses, but much addressed in Section 6. structural information remains uncaptured and, most problematically, the technology CONTEMPORARY BRAIN SCANNING does not permit in vivo brain scanning. Non-Destructive Techniques Non-destructive structural whole brain Current techniques for non-destructive monitoring techniques in the form of functional whole brain monitoring do computerized scanning-based imaging enable the creation of detailed system level modalities, such as positron emission maps of the brain functional connectome, tomography and magnetic resonance achieved using resting-state functional imaging, provide non-destructive three- magnetic resonance imaging[47] with voxel dimensional views of the brain down to ~1 size resolution of 2 mm x 2 mm x mm resolution, with typical clinical MRI 2.5 mm[48], but these are still incapable of scan voxel resolutions of 1 mm x 1 mm x cellular-level resolution. 3 mm[40, 41]. Such resolution permits regional analyses of brain structure but is Destructive Techniques clearly insufficient for investigation of Contemporary destructive structural whole structures underlying intercellular brain monitoring can provide resolutions communication at the level of individual down to the nanometric level. neurons or synapses[42]. High-definition Ultramicrotome scanning, for example, fiber tractography provides accurate provides near nanometric resolution after reconstruction of white matter fiber chemical preservation of the tissue. It is tracts[43, 44] but also with a resolution of being used to scan larger and larger brain only ~1 mm [43, 44]. volumes with nanometric detail permitting visualization of individual synapses and its Micro-CT scanners, with a typical scan components. Ultramicrotome sections 30– time between 10 min and 2 hr, allow high- 100 nm thick are scanned by either a resolution tomography of specimens up to transmission electron microscope, a serial a few centimeters in diameter, with the block-face scanning electron microscopy,

IJNN (2015) 20–41 © JournalsPub 2015. All Rights Reserved Page 22 Action Potential Monitoring Using Neuronanorobots Martins et al. ______or a light-optical microscope, with reconstructed with histological (pixel size automation of the collection of ultrathin 0.95 μm2) and MRI data (voxel size serial sections for large volume 12.3μm3) [1, 60, 2]. At the structural cellular transmission electron microscope level, other methods such as CLARITY reconstructions[49, 50, 51]. Current scanned enable estimations of the joint volumes are far from a whole human brain morphological statistics of many neurons volume, but a process to achieve a whole in a tissue sample at the same time[62, 63]. human brain is envisioned[51]. Several For in vitro cellular approaches, scanning methods were developed to improve the light microscopy (e.g. confocal analysis of the ultra-thin microtome microscopy) provides three-dimensional images. After scanning, posterior software views of individual neurons but only down reconstruction uses specialized software to ~1 μm resolution. such as RESCOP or KNOSSOS to trace the connections between neurons[52]. To date, several smaller-than-human-brain Tagging individual neurons with connectomes have been scanned, including fluorescent proteins[53,54] facilitates the the C. elegans connectome[64, 65, 66], the analysis of neuronal circuitry and glial predatory nematode Pristionchus pacificus territory mapping on a large scale. A high- connectome[67], the connectomes of six throughput technique called BOINC interscutularis muscles[29], the partial (“barcoding of individual neuronal structural and functional connectome of connections”) for establishing circuit the mouse primary visual cortex (via connectivity at single-neuron and synaptic electron microscope and two-photon resolution was proposed using high- microscopy, 800 TB data set, 5nm ×5 nm throughput DNA sequencing[55]. ×50 nm spatial resolution, and the entire data set captures a tissue volume of 30 × Other strategies are also being pursued to 30 × 30 μm3)[68], and the inner plexiform avoid the laborious ultrastructural electron layer of the mammalian retina connectome microscopy based techniques. For (via automated transmission electron example, The X-ray nanotomography microscope imaging, 16.5 TB data set, microscope delivers a high-resolution 3-D ~2 nm resolution of a 0.25 mm diameter image of the entire cell in one step, an tissue column spanning the inner nuclear, advantage over electron microscopy in inner plexiform, and ganglion cell layers which a 3-D image is assembled out of of the rabbit)[30]. many thin sections which can take up to weeks for just one cell[56, 57]. Cell Structural 3-D destructive reconstruction ultrastructure has been imaged with X-rays of a whole human brain with 20 μm down to 30 nm resolution. resolution has been completed, preserving the first human whole-brain Partially destructive techniques have been cytoarchitectural anatomy[3]. A planned used to study, physiologically and future project is the creation of a ~1 μm anatomically, a group of neurons in the spatial resolution brain model, intended to mouse primary visual cortex[58]. Two- capture details of single cell morphology photon calcium imaging was used to and to integrate gene expression data from characterize functional properties, and the Allen Institute Brain Activity Map large-scale electron microscopy of serial Project. There are also efforts underway to thin sections were employed to trace a map the whole human brain at the synaptic portion of these neurons’ local network). level of resolution[70]. The remaining Other techniques can obtain whole brain challenges necessary to scale these structural gene expression information at destructive processes into a whole human the cellular level, as computationally

brain appear surmountable in the decades opening of ~20 Ca2+ channels per active ahead. zone, and consequent monitoring of ion fast release with a delay of 50–500 μs [78, Contemporary destructive approaches 79]. Also potentially inferable might be the provide near-nanometric structural resultant Ca2+ transient (lasting 400– resolution, but many inherent problems 500 μsec)[79]. By measuring synaptic remain. First, it is not clear if the different electrical activity, neuronanorobots can biomolecular machinery can be also monitor synaptic plasticity including distinguished by the next generation of synaptic based long-term potentiation, these techniques. Second, functional long-term depression, short-term plasticity, information is not captured by these metaplasticity, homeostatic plasticity, and techniques (although functional cross-talk. information might not always be necessary, in case atomic structural Action potentials may encode information resolution is achieved). Third, destructive in spike timing pattern and in the spike techniques are expected to face resistance waveform. While there is evidence that the for implementation in clinical practice, in action potential waveform encodes some part because the destruction is irreversible type of information, its relevance is not and in part because of the difficulty in clear. The rate of information transfer proving the continuity of consciousness. including action potentials waveforms may be significantly higher than the rate ACTION POTENTIAL MEASUREMENT assuming only stereotyped spike train REQUIREMENTS impulses[80]. In the interim, a conservative The human connectome sets the design criterion for action potential underlying structure for the synaptic- nanosensors would include the capacity to processed (4.31 ± 0.86) × 1015 spikes/sec measure individual action potential signal traffic processed in the whole waveforms. human brain, constituting the most crucial and data-intensive information channel For acquiring optimal spatial and temporal corresponding to (5.52 ± 1.13) x 1016 resolution the action potential nanosensors bits/sec [2]. Synapses, the structural sub- need to be positioned as close as possible cellular components responsible for to the action potential initiation site. In processing this data, play a crucial role in most cases, action potentials are initiated brain information processing[3], are at the axon initial segments (AIS)[81], but involved in learning and memory (either in some cases action potentials are long-term and short-term memory storage initiated at the axon hillock, and and deletion)[71, 72, 73, 74], participate in sometimes they are even initiated at the temporal processing of information[75], and first node of Ranvier[82, 83]. For example, are key elements for signal transduction the site of action potential initiation in and plasticity in the human brain[76, 77]. cortical layer 5 pyramidal neurons is ~35 µm from the axon hillock (in the AIS)[83]. The key functional-information In some other neuronal types, the action measurement task at synapses is potential may be initiated at the first nodes monitoring action potentials, capturing of Ranvier[84, 85, 83] which, in layer 5 even the fastest 400–800 Hz firing rates pyramidal neurons, is ~90 µm from the occurring at fast spiking neurons[4, 5] and axon Hillock – the first myelin process is the fastest voltage velocities at ~40 µm from soma and the length of the 20 mV/ms[6]. Inferable from the electrical first myelin process is ~50 µm[83]. Since data might be the action-potential-induced action potentials might be initiated in

IJNN (2015) 20–41 © JournalsPub 2015. All Rights Reserved Page 24 Action Potential Monitoring Using Neuronanorobots Martins et al. ______different cellular subcompartments, 800 Hz range[4, 5]; duration is usually 0.4– endoneurobots will be parked at the AIS 0.6 ms. (the most likely spot for action potential initiation) where they will monitor the Bursting neurons (which generate clusters large majority of action potentials. In of spikes either singly or repetitively) neurons where some action potentials are respond to a depolarizing stimulus of 0.3 initiated at the first nodes of Ranvier or the nA with a repetitive burst discharge, with axon hillock, two synaptobots placed at the an intraburst frequency of 300 Hz (the ﬁrst first node of Ranvier and at the axon burst might reach 600 Hz) and an hillock can ensure proper action potential interburst frequency of 40 Hz[88]. Bursting waveform detection of all initiated action neuron high-frequency (300–600 Hz) potentials. spike bursts recur at fast rates (30–50 Hz) within a certain range of membrane Once in the right position, nanosensors potentials[89, 90, 4, 88]. Bursting neurons will detect individual action potentials have two main subtypes: intrinsically with proper waveform temporal resolution. bursting (IB) and fast-repetitive bursting Estimating the necessary waveform (FRB)). During sustained depolarization, temporal resolution requires an overview IB neurons fire a short burst of 3–5 action of a neuron’s firing frequency variability potentials at ~200 Hz which becomes and theoretical maximum firing frequency. repetitive usually at frequencies around 5– It is well-known that the “typical” ~20 μm 15 Hz[5]. During sustained depolarization, human neuron discharges 5-100 sec-1, FRB neurons fire bursts containing 25 moving from ˗60 mV potential to +30 mV spikes at frequencies from 200–600 Hz, potential in ~1 ms. However, the with short spikes of ~0.6 ms bursts variability of action potential frequencies repeating regularly at 20–80 Hz[5]. is large and depends largely on the electrophysiological class of the neuron[2]. The maximum firing frequency reported in There are three main electrophysiological all human electrophysiological neuron classes of neurons in the human brain[86, types is 800 Hz, although other vertebrates 87]. employ somewhat higher maximum firing frequencies, e.g., 2000 Hz for chicken[91, 92, Regular Spiking neurons (which fire at 93, 94]. For non-vertebrates, the maximum low rates and adapt to continuous stimuli) firing frequency for mechanosensory respond to a “typical” depolarizing neurons in copepod antennules with single stimulus of 0.3 nA with initial frequencies neurons firing was a maximum frequency of 100 Hz in the first 2 ms, then of 5000 Hz and sustaining frequencies of accommodate during the following 50 ms 3000–4000 Hz for up to 4 ms[95]. to steady frequencies of about 30 Hz Comprehensive electrophysiological (usually range 20–50 Hz)[4, 5]. Regular studies may be necessary to guaranty that Spiking firing frequencies can rise to 200– such high frequencies do not occur 300 Hz [5] with each spike lasting for anywhere in human brains. ~1 ms [5]. A detailed recording of human action Fast Spiking neurons (which sustain very potential waveforms might conceivably high firing frequencies with little or no require a 0.05 ms temporal resolution (the adaptation) respond to a depolarizing fastest voltage velocities at 20 mV/ms stimulus of 0.3 nA with a sustained high would require 0.05 ms resolution for frequency of 250–350 Hz, though having mV resolution). However, a discharges can sometimes reach the 400– somewhat lower temporal resolution is expected to be necessary because each

spike is estimated to carry 7–10 bits of potential as a result of entering Na+ ions information[2], motivating the choice and exiting K+ ions, the respective 0.1 ms (i.e., 10,000 Hz) temporal concentrations inside and outside of the resolution in the present work. This axoplasm change relatively little when resolution should permit detailed compared to the total number of Na+ and monitoring of a very large majority of K+ ions present. This change in membrane neuroelectrical waveforms and will ensure potential is associated with a certain total the detection of even the fastest action number of charges that move across the potentials. plasma membrane per unit area, creating a charge differential across the membrane of 2 CHOICE OF NANOSENSOR FOR Qaction = Cm Vm / qe ≈ 6250 ions/μm , ACTION POTENTIAL MONITORING taking membrane capacitance Cm ~ At least two basic nanosensor types could 1 μF/cm2 for biological lipid bilayers[97] potentially be used on an intracellular and Vm ~ 100 mV across the biological basis to detect individual action potentials: membrane, with each monovalent ion concentration-based nanosensors (Section carrying one elementary charge qe = 1.6 x 4.1) and electrical field-based nanosensors 10˗19 coul. (Section 4.2). The choice is largely driven by the need for a nanosensor that provides Considering the neuron soma (cell body) the required signal resolution necessary to as a whole, and assuming a “typical” 10 characterize the action potential waveform. μm diameter neuron with a total soma 2 Thermal sensors, a third type, appear surface area of Asoma ~ 314 μm and 3 marginal because the thermal time volume Vsoma ~ 524 μm , the neuron + constant across a distance Ln ~ 20 μm (the cytoplasm contains NNa+ = [Na ]inVsoma ~ 9 + + maximum diameter of human nerve axons) 5.8 × 10 Na ions and NK+ = [K ]inVsoma ~ 10 + for neurons having thermal conductivity Kt 4.4 × 10 K ions. During an action = 0.6 W/mK (~water at K) and heat potential, nNa+ (=nK+) ~ AsomaQaction~ 2 x 6 3 6 + capacity CV = 4 x 10 J/m K (brain tissue) 10 Na ions enter the cell and an equal 2 + is τeq = Ln CV / Kt ~ 3 ms, much longer number of K ions exit the cell. Such a than the minimum ~0.1 ms temporal small ion current represents an increase of resolution that is probably required to only ΔCNa+ = nNa+ / NNa+≈ 0.03% in the sufficiently characterize fast spike sodium ion concentration of the entire waveforms. neuron soma, and an increase of only ΔCK+ = nK+ / NK+≈ 0.005% for potassium Concentration-Based Nanosensors ions during the rise time of the action One approach to monitor action potential potential. (Of course, overall Na+ and K+ waveforms is to use nanosensors that can concentrations can depart significantly directly measure the Na+ and K+ ion from these values in axons having small concentration changes intracellularly, near cytoplasmic volumes when firing the axon membrane in the interior of the sustainedly at high frequencies.) axon hillock during an electrical event. Typically, the resting ion concentrations in A more complete analysis would include + the cytosolic axoplasm are [Na ]in = the diffusion rates of ions and the number 7 3 + + + 18.0 mM (1.1 x 10 ions/μm ) and [K ]in = density of Na and K channels and pumps 140.0 mM (8.4 x 107 ions/μm3), while in order to account for the much slower + extracellular concentrations are [Na ]out = pump rate per square micron by which the + [96] 144 mM and [K ]out = 4 mM . While an ions are returned to their original side of action potential event causes a Vm ≈ 100 the membrane. Since the turnover rate of mV change in the neuron membrane Na+/K+ pumps (~500 ions/sec) is so much

IJNN (2015) 20–41 © JournalsPub 2015. All Rights Reserved Page 26 Action Potential Monitoring Using Neuronanorobots Martins et al. ______slower than the Na+/K+ channels (~106– sensors for real-time action potential 107 ions/sec), some increase in ion monitoring inside living human neurons. concentrations in the near-membrane volume might also be expected during fast Electrical Field-Based Nanosensors spiking rates, consequently reducing the Another approach to monitoring action available concentration change for later potential waveforms is to use nanosensors action potentials. Ionic species can become that can measure the change in local highly hydrated when dissolved in water. electric field strength during the action For instance, a naked proton (H+) is potential event. The electric field E + + usually present as H5O2 or H7O3 , or even (volts/m) surrounding a single monovalent + [98] as H9O4 in strong acid solutions , and ion is given by Coulomb’s law as: E = qe / + 2 ˗19 some ions such as Li and I are found in 4 π ε0 κe r , where qe = 1.60 x 10 coul ˗12 large solvent cages coordinated to as many (one charge), ε0 = 8.85 x 10 F/m [99, 100] as 46 water molecules . Another (permittivity constant), κe = dielectric complicating factor is that little is known constant (relative permittivity) of the about the compartmentalization and matter traversed by the electric field (e.g., dynamics of sodium and potassium fluxes taking κe = 74.31 for pure water at 310 K; [104] in neuron cells with complex κe decreases slightly with salinity ), and cytoarchitectures[101]. r = distance from the charge, in meters. In an aqueous medium such as the interior of An ideal concentration sensor, limited only an axon, the field at a distance of 10– by diffusion constraints and drawing 100 nm from the singly-charged ion is through a spherical boundary surface of 200000–2000 V/m. radius Rs, provides a minimum detectable concentration differential of Δc/c = (1.61 Patch clamp is an existing laboratory ˗1/2 Δt D0 cion Rs) , where Δt = measurement technique that allows the study of single or time, D0 = aqueous diffusion coefficient of multiple ion channels in neurons. The the hydrated ion at infinite dilution, and method combines scanning ion [102] cion = ion concentration . Requiring Δt conductance microscopy, which is used to ≤ 0.1 ms to ensure minimally adequate scan the exterior surface and identify the action potential waveform resolution, then positions of ion channels on the neuron Rs/Na+ ≥ 3.9 μm to detect a Δc/c = ΔCNa+ = membrane, with patch-clamp recording 0.03% change in Na+ ion concentration through a single glass nanopipette [26] from the cytosolic baseline of cion = 1.1 x probe . The blunt-ended nanopipette, 7 3 + 10 ions/μm for Na , and Rs/K+ ≥ 12.3 μm which has an inside diameter of 100– to detect a Δc/c = ΔCK+ = 0.005% change 200 nm and can be positioned with in K+ ion concentration from the cytosolic nanometer precision, is first scanned over 7 3 2 baseline of cion = 8.4 x 10 ions/μm for the neuron membrane area of ∼0.03 μm , + ˗9 2 + K , taking D0 ~ 1.6 x 10 m /sec for Na using current feedback to obtain a high- ˗9 2 + and D0 ~ 2.4 x 10 m /sec for K , in water resolution topographic image. The tip is [103] at 310 K . These values for Rs are then sealed onto the membrane by already unfeasibly large but are only lower applying suction to develop a tight high- limits because the indicated Δc/c occurs resistance seal, guaranteeing that all ions over a ~1 ms rise time, not over the fluxing the membrane patch flow into the shortest measurement interval Δt ~ 0.1 ms, pipette to be recorded by a chlorided silver hence the required Δc/c detection electrode connected to a highly sensitive threshold may be as much as tenfold electronic amplifier. This method, also lower. These considerations appear to rule called nanopatch-clamp[105, 106, 107, 108], in out the use of chemical concentration principle allows the counting of each ion passing through a selected individual

sodium or potassium ion channel. A single ~KHz frequencies are reported for the [119] ion channel conducts between 1–10 perovskite-related oxide CaCu3Ti4O12 ). million ions/sec, a current of 1–10 pA [109]. A well-designed “passive” electric For nanorobots operating inside a neuron, nanosensor of this size and configuration, permanently sealing a sensor around the when pressed near the axonal internal cytosolic aperture of a large number of ion membrane surface, should readily detect channels would be logistically challenging the passage of one or a small number of and would likely interfere with normal ions and thus the variation in electric field neuron function, e.g., ion transport and caused by each action potential discharge, protein recycling. If we seek instead to without resort to patch-clamping. measure ion current without sealed clamping of the sensor to the cell The density of Na+ ion channels in the AIS membrane, it is useful to examine the of the axon is estimated as 100– changes caused by the different types of 200 µm˗2[81], so channels are spaced noise in neurons that might affect sensor ~100 nm apart across the membrane accuracy[110]. Sources of response surface. An R = 25 nm nanosensor placed variability in neurons and neural networks 10 nm directly beneath an ion channel in may include thermal noise, ionic the membrane would reliably detect the conductance noise, ion pump noise, ion initial 200000 V/m field from the entry of channel shot noise, synaptic release noise, single Na+ ions through the local channel, synaptic bombardment, chaos, whereas Na+ ions entering through the connectivity noise, and environmental nearest adjacent channel 100 nm away and stimuli[111, 112, 113, 114, 115, 116, 117]. flowing around a second R = 25 nm nanosensor will be at closest 50 nm from After considering these potential sources the first nanosensor, generating a of noise, the main conclusion is that 8000 V/m field, just 4% of the local signal. thermal noise is the only source of noise relevant to evaluate if the cytoplasmic- Exiting K+ ions can also be detected if the resident nanosensors must be capable of potassium ion channel locations are distinguishing the entrance of each single known. Kv1 potassium channels [120] ion on the nearest ion channel without control axonal action potential waveform having a nanopatch-clamp sealed around and synaptic efficacy, shaping the the ion channel aperture. waveform in the AIS of layer 5 pyramidal neurons independent of the soma[121]. The theoretical thermal noise limit for electric field detection using a “passive” The first 50 μm of the AIS has a 10-fold cylindrical sensor of radius R = 25 nm, increase in Kv1 channel density[121]. length L = 250 nm, and wall electrical Considering the lifetime of a “typical” thickness dwall = 10 nm has been estimated sodium or potassium ion channel, the [118] 1/2 as: Elimit = 2√2 (kBT dwall / 4 π ε0 κe) Kv1.3 potassium channel has an estimated 1/2 3/2 1/2 [1 / (R L (ν tmeas) )] ~ 2000 V/m, turnover rate (half-life) of 3.8 ± 1.4 hr, taking Boltzmann’s constant kB = 1.38 x rising to ~6.3 hr in the presence of 10˗23 J/K, T = 310 K, electric field TrkB[122]. frequency ν = 10 KHz, measurement time tmeas = 0.1 ms, and relative permittivity κe This is consistent with reported half-lives ~ 2000 for the wall material (cf., values on the order of hours for ion channels on [123] approaching κe ~ 100,000 at 310 K and the cell membranes of cardiomyocytes , and suggests that our cytosolic-resident

IJNN (2015) 20–41 © JournalsPub 2015. All Rights Reserved Page 28 Action Potential Monitoring Using Neuronanorobots Martins et al. ______nanosensors may need to be repositioned interface and permitting nanosensor probe and retargeted several times a day to miniaturization to smaller sizes[125]. maintain proper signal. Decreasing sensor size is beneficial because capacitance decreases and A small array of sensors will also be resistance terms are relatively improved or required because target ion channel not limiting, so the RC time constant proteins will be rapidly drifting in and out remains very small. The smallest "active" of range of individual cytosolic FET-based nanosensor that has been built nanosensors during long monitoring times and tested has a probe of 40 nm diameter – lateral diffusion time in cell membrane and 50 nm length, and has demonstrated for transmembrane ion channel proteins is good SNR[126]. 2 of order tD ~ X / 2 DL ~ 0.1–1 sec between submembrane nanosensors located X = PROPOSED FET-BASED 40 nm apart, taking lateral diffusion NEUROELECTRIC NANOSENSOR 2 coefficient DL ~ 0.001–0.01 µm /sec for FET-based nanosensors can record electric Na+ ion channels at the axon hillock and potentials intracellularly in living neuritic terminal[124]. neurons[127, 128] using kinked nanowire structures (Fig. 1), providing high signal- However, “passive” electrodes have a to-noise ratio (SNR) and high temporal theoretical minimal size due to impedance resolution[125]. because the sensing process in such detectors requires electrochemical ionic The voltage rise/fall time-frame ranged exchange. In “active” electrodes, such as from 0.1–50 ms, with the FET nanosensors the two-terminal transistors found in Field demonstrating capacity to detect 0.1 ms Effect Transistor (FET) based action potentials pulse rise/fall without nanosensors, there is no similar exchange detectable delay[129]. and the device/electrolyte interface has effectively “infinite” impedance. Thus, SWCNT or DWCNT FET-based nanosensors seem to be a promising In such sensors, impedance is not relevant technology for nanorobotic monitoring of to recording bandwidth or noise, allowing action-potential based electrical FETs to detect action potentials information. independently on the device/electrolyte

A B C Fig. 1: (A) Experimental 3-D Free-Standing, Kinked Nanowire FET Bent Probe; the Yellow Arrow and Pink Star Mark the Nanoscale FET. Scale Bar, 5 μm (Reprinted with Permission)[100,102]. (B) Differential Interference Contrast Microscopy Images of an HL-1 Cell and 60° Kinked Nanowire Probe Whose V-Shaped Apex is Visible Inside the Cell (Reprinted with Permission)[100]. (C) Experimentally-Recorded Intracellular Action Potential Peak Using FET Sensors with Kinked Nano-Wire Gate, from Cells Cultured on Polydimethylsiloxane Substrate, with Intracellular Cytosolic Resting Potential Indicated by the Dashed Line (Reprinted with Permission)[102].

However, common FET-based sensor When immersed in the electrolytic components such as metal electrodes, if environment of the neuron cytosol, only operated inside the neuron cell, would the nanotube gate is physically exposed to disrupt the normal functioning of the cytosolic fluid and the dependence of the cell[125]. conductance on gate voltage makes our nanosensor an electrically-based voltage Existing design proposals reflect the nanosensor (Figure 3). present state of fabrication technologies and assume that nanosensor components must function extracellularly.

In vivo intracellular action potential monitoring using FET-based nanosensors would also require miniaturization of all the nanosensor components.

For our intracellular FET-based neuroelectric nanosensor (Figure 2), a carbon nanotube (“Gate”) connects the source (“S”) and drain (“D”) electrodes.

The sensor also includes a voltmeter and \ ammeter, and receives power from a Fig. 2: Basic Schematic of a FET-based battery connected by nanowires. Neuroelectricnanosensor.

Fig. 3: Vertical Cross-Section (Side View, at Left) and Horizontal Cross-Section (Top View, at Right) of the Fet-Based Neuroelectricnanosensor.

Glass-coated hemispherical nanoelectrodes current-limiting resistor followed by a with dimensions as small as ~1 nm have small resistor, plus a galvanometer. been fabricated and exhibit reproducible Another option is to use voltage sensing and stable voltammograms without inorganic nanoparticles (vsNPs) which are hysteresis, withstanding 6 hours of currently employed to self-insert into the continuous use and 15 hours of iterative cell membrane and optically record, non- processes of heat and voltammetry[134, 135]. invasively, action potentials at multiple These types of nanoelectrodes offer sites and in a large field-of-view[138]. several advantages: (1) Mass-transport rate Alternatively, we might use an analog of increase, allowing steady-state the 30 nm “photonic voltmeter” which is voltammetric responses to be readily one thousand times smaller than existing achieved,(2) Smaller RC constants and (3) voltmeters and is claimed to enable The ability to make measurements in complete 3-D electric field profiling solutions of high resistance because of the throughout the entire volume of living lower influence of solution resistance. The cells[139]. size and shape of the nanoelectrode is crucial because its electrochemical The high conductivity of metallic properties are often exceedingly sensitive nanotubes makes CNTs interesting to even small variations in its geometry. building blocks for future advanced molecular electronic circuits[140]. SWCNTs Electrical measurement devices with tens are the most conductive carbon fibers of nanometer size are necessary to serve as known, with resistivity on the order of 10–4 ammeters and voltage detectors[125]. The ohm/cm at 27°C and current density of smallest ammeters are expected to be ~10˗7 A/cm2, though in theory SWCNTs electron ammeters that show the real-time may be able to sustain stable current dynamics of single electron tunneling[136] densities up to ~10˗13 A/cm2. In DWCNTs and provide high-sensitivity high- the difference between the radius of the bandwidth single electron detection, inner tube and the outer tube is ~3.6 Å, measuring currents in the attoampere range independently of the DWCNT (10-18 A)[137]. The charge detector circumference, with the lattice structures employed might be a single-electron of inner and outer tubes having no transistor or a double quantum dot to allow translational symmetry. Thus, the intertube monitoring the direction of the flow of transfer is negligibly small and has no electrons[136, 137]. The double quantum dot effect on transport properties of can act as its own electrometer[136]. The DWCNT[141]. By managing the electronic ammeter might also be designed with the properties of CNTs (dependant on the use of a small resistor and a sensitive orientation of the honeycomb lattice with current detector, such as a galvanometer, respect to the tube axis, known as helicity), that converts electricity into a mechanical the neuroelectric nanosensor wires can be movement, possibly constructed within a produced using DWCNTs[142]. Combining 1000 nm3 volume using the techniques of an internal CNT having metallic or molecular manufacturing. semiconductor properties with an external nanotube having insulation properties In the FET-based neuroelectric gives a nanosensor wire that is a molecular nanosensor, a voltmeter is placed in analog of coaxial cable[142]. parallel with a circuit element to measure the voltage and must not appreciably Some of the most important performance change the circuit it is measuring. The metrics on nanosensors are sensitivity, nanosensor nano-voltmeter could employ SNR, limit-of-detection, cross- traditional voltmeter concepts, using a sensitivity/selectivity, signal rise/fall time

(speed), repeatability, offset/sensitivity, SENSOR BIOCOMPATIBILITY drift, hysteresis, and Our CNT-based FET nanosensor design lifetime/robustness[143]. These sensor must be carefully analysed to anticipate characteristics must be futher evaluated potential biocompatibility problems during experimentally in the context of the intracellular neuron action potential proposed neuroelectric nanosensor. For monitoring. Neurons in general seem to example, in extracellular recordings of accept carbon nanotubes e.g., CNT-based cardiac myocytes an SNR of 2030 was substrates have proven to be biocompatible measured for CNT based nanosensors with neural cells and even stimulate neural (tenfold higher than competition)[145, 144]. cell growth, improving the cell’s ability to An SNR of ~257 was determined for extend processes and improving neuronal another CNT electrode during in vitro networks electrical performance[148, 149]. recording of neural signals in crayfish nerve cord[146]. A better SNR allows As the sensor gate is the only sensor smaller signals to be detected, improving component exposed to the neuron cytosol, the sensor’s limit of detection[147]. one primary biocompatibility concern is the effect of carbon nanotubes on neural In the medical nanorobot implementation cells[150]. If a CNT gate is not physically envisioned here, each endoneurobot will disrupted, biocompatibility problems seem incorporate ~100 FET-based neuroelectric very unlikely because CNTs have nanosensors in its outer hull (Figure 4). demonstrated electrochemical and Each nanosensor has ~3375 nm3 volume biological stability[151], resistance to bio- (including power, housing and mechanical fouling and mechanical compatibility with control, but not including control, brain tissue[133]. The unlikely disruption, communication and computational detachment and release into the neuron processing machinery). The endoneurobot cytosol of a small number of CNT gates nanosensors are organised in groups of ten seems unlikely to cause problems on the nanosensors, distributed along the cell. Only if CNTs are released in large endoneurobot perimeter. While monitoring quantities inside the neuron might cellular action potentials, at least one group of ten cytotoxic effects be induced. nanosensors should be near the axon membrane with nanosensor gates Depending upon shape and concentration, separated by ~40 nm. these effects could potentially include: (1) stronger than normal metabolic activity, (2) elevated lactate dehydrogenase, (3) generation of reactive oxygen species in a concentration- and time-dependent manner, indicating an oxidative stress mechanism, (4) activation of time- dependent caspase 3 showing evidence of apoptosis, (5) decrease of mitochondrial membrane potential, (6) increased level of lipid peroxide, and (7) decrease the activities of superoxide dismutase, Fig. 4: A Multitude of Neuroelectric glutathione peroxidase, catalase and the Nanosensors, Incorporated into the content of glutathione[148]. The effects on Surface of a Single Endoneurobot, are cell viability will vary in a concentration Positioned Near the AIS Membrane. dependent manner (Figure 5).

nanosensor surface to remove attached proteins.

The Young’s modulus of SWCNTs depends on their size and chirality, but averages 1.09 TPa for a generic nanotube, hence CNTs are stiffer than steel and very resistant to damage from physical forces, implying a long nanosensor lifetime and robustness. CNTs have strong in-plane graphitic carbon-carbon bonds which

Fig. 5: Effect of Short SWCNTs make them exceptionally strong and stiff (SSWCNTs) on PC12 Cell Viability, When against axial strains. Tenfold redundancy Cells are Treated with 500–2000 nm Long per group of sensors should preserve SWCNTs for 24 and 48 hr at 37°C mission-long functionality. Protocols for (Reprinted with Permission)[80] removal and replacement of the endoneurobot should also be in place to Short SWCNTs are defined as nanotubes handle unanticipated problems. with 1–2 nm diameter and 0.5–2 μm length, or more than 100 times longer than CONCLUSION the proposed FET-based nanosensor gate. Comprehensive preservation of human Consequently higher concentrations are brain information requires proper scanning expectably necessary to produce an of functional connectome data using equivalent effect on cell viability. A 10% appropriate nanosensors. Neuronanorobots (both endoneurobots and synaptobots) reduction on cell viability corresponds to a 3 5 μg/ml concentration of short SWCNT. equipped with the proposed ~3375 nm The release of all hundred 15 nm long FET-based neuroelectric nanosensors FET-based nanosensor gates on one might provide adequate temporal endoneurobot into the cytosol of a resolution for preserving action potential (20 μm)3 volume neuron would amount to waveform information and could detect a ~0.001 μg/ml dose, causing a likely even the fastest human action potential undetectable 0.002% reduction in cell firing rates of 800 Hz while presenting viability. Another biocompatibility minimal biocompatibility problems. A set problem might be hysteresis. The FET of such neuroelectric nanosensors installed on a sufficient number of well-placed nanotube gate, lying on a SiO2 surface, is very likely to exhibit hysteresis in its endoneurobots and synaptobots will enable electrical characteristics due to charge these robots to non-destructively and trapping by water molecules around the continuously monitor virtually all action nanotube, regardless of CNT potentials arising throughout a living hydrophobicity[152]. A protocol for closing human whole brain. the sensors and removing the water, perhaps using a shutter-like system or a set ACKNOWLEDGMENTS of molecular pumps, should reduce The principal author (NRBM) thanks the transistor hysteresis. Another solution is to “Fundação para a Ciência e Tecnologia” create a virtually hysteresis-free transistor (FCT) for their financial support of this by passivating the nanotube with polymers work (grant SFRH/BD/69660/2010). that hydrogen bond with silanol groups on REFERENCES SiO2 (e.g., with polymethyl methacrylate)[152]. The shutter-system is 1. Azevedo F.A., et al. Equal numbers also useful for periodically cleaning the of neuronal and nonneuronal cells

make the human brain an 12. Patel G.M., et al. Nanorobot: A isometrically scaled-up primate versatile tool in nanomedicine. brain. J Comp Neurol. 2009; Journal of Drug Targeting. 2006; 513(5):532–41p. 14(2): 63–7p. 2. Martins N.R.B., Erlhagen, W., 13. Park H.H., et al. Rise of the Freitas, R.A. Jr. Non-destructive nanomachine: The evolution of a whole-brain monitoring using revolution in medicine. nanorobots: Neural electrical data Nanomedicine 2007; 2(4): 425–39p. rate requirements. Intl. J. Machine 14. Popov A.M., et al. Biocompatibility Consciousness. 2012; 4:109–40p. and applications of carbon nanotubes 3. IBM (2008) IBM Seeks to Build the in medical nanorobots. International Computer of the Future Based on Journal of Nanomedicine. 2007; Insights from the Brain. IBM News 2(3): 361–72p. Room. 20 Nov 2008. 15. Martel S., et al. Flagellated Website: https://www- magnetotactic bacteria as controlled 03.ibm.com/press/us/en/pressrelease/ MRI-trackable propulsion and 26123.wss steering systems for medical 4. Wilson H.R. Simplifed Dynamics of nanorobots operating in the human Human and Mammalian Neocortical microvasculature. International Neurons. Journal of Theoretical Journal of Robotics Research. 2009; Biology. 1999; 200 (4): 375–88p. 28(4): 571–582p. 5. Contreras, D. (2004) 16. Mallouk T.E., Sen, A. Powering Electrophysiological classes of nanorobots. Scientific American neocortical neurons. Neural 2009; 300(5): 72–77p. Networks 2004; 17(5–6) 633–46p. 17. Kostarelos K.Nanorobots for 6. Naundorf B., et al. Unique features medicine: How close are we? of action potential initiation in Nanomedicine (Lond). 2010; cortical neurons. Nature 2006; 440: 5(3):341–2p. 1060–3p. 18. Mavroides D., Ferreira A. 7. Freitas R.A. Jr. Nanomedicine, NanoRobotics: Current Approaches Volume I: Basic Capabilities. Landes and Techniques. Springer, New York. Bioscience, Georgetown, TX. 1999. 2011. 8. Freitas R.A. Jr. Exploratory Design 19. Freitas R.A. Progress in in Medical Nanotechnology: A Nanomedicine and Medical Mechanical Artificial Red Cell. Artif. Nanorobotics. Rieth M, Schommers Cells, Blood Subst., and Immobil. W, eds., Handbook of Theoretical Biotech. 1998; 26:411–30p. and Computational Nanotechnology, 9. Freitas R.A. Jr. Nanomedicine, 6 (Bioinformatics, Nanomedicine, Volume IIA: Biocompatibility. and Drug Design), American Landes Bioscience, Georgetown, TX. Scientific Publishers, Stevenson 2003. Ranch, CA; (2006a); 619–72p. 10. Morris K. Macrodoctor, come meet 20. Freitas R.A. Jr. Comprehensive the nanodoctors. Lancet. 2001; 357: Nanorobotic Control of Human 778p. Morbidity and Aging. in Fahy, G.M., 11. Astier Y., et al. Protein components et al. The Future of Aging: Pathways for nanodevices. Current Opinion in to Human Life Extension, Springer, Chemical Biology. 2005; 9(6): 576– New York. 2010; 685–805p. 84p. 21. Freitas R.A. Jr. Clottocytes: artificial mechanical platelets. Institute for

Molecular Manufacturing; Report 32. Seung H. S. Neuroscience: Towards No. 18, June 2000. functional connectomics. Nature. 22. Freitas R.A. Jr. Nanodentistry. J. 2011; 471, 170–172p. Amer. Dent. Assoc. 2000b; 131: 33. Seo D., et al. Neural Dust: An 1559–66p. Ultrasonic, Low Power Solution for 23. Freitas R.A. Jr. Microbivores: Chronic Brain-Machine Interfaces. Artificial Mechanical Phagocytes arXiv, 2013, using Digest and Discharge Protocol. arxiv.org/abs/1307.2196. J. Evol. Technol. 2005; 14(Apr):1– 34. Hogg T., Sretavan D.W. Controlling 52p. tiny multi-scale robots for nerve 24. Freitas R.A. Jr. Pharmacytes: An repair. In Proc. of the 20th Natl. Ideal Vehicle for Targeted Drug Conf. on Artificial Intelligence (AAAI Delivery. J. Nanosci. Nanotechnol. 2005), 2005; 1286–91p. 2006: 2769–75p. 35. Hogg T., Coordinating Microscopic 25. Freitas R.A. Jr. The Ideal Gene Robots in Viscous Fluids. Auton Delivery Vector: Chromallocytes, Agents Multi-Agent Syst. 2007; 14: Cell Repair Nanorobots for 271–305p. Chromosome Replacement Therapy. 36. Freitas R.A. Jr. Computational Tasks J. Evol. Technol. 2007; 16: 1–97p. in Medical Nanorobotics. Eshaghian- 26. Gorelik J. Ion channels in small cells Wilner MM, ed., Bio-inspired and and subcellular structures can be Nano-scale Integrated Computing. studied with a smart patch-clamp John Wiley & Sons, NY, 2009; 391– system. Biophys J. 2002; 428p. 83(6):3296–303p. 37. Freitas R.A. Jr., Merkle R.C. 27. Freitas R.A. Jr., Phoenix, C.J. Nanofactory Collaboration. June (2002). Vasculoid: A personal 2006. nanomedical appliance to replace 38. Freitas R.A. Jr., Merkle R.C. human blood. J. Evol. Technol. Kinematic Self-Replicating 11(April):1–139p. Machines. Landes Bioscience. 28. Sporns O., et al. The Human Georgetown, TX. 2004. Connectome: A Structural 39. Freitas R.A., Mavroidis, C., Ferreira Description of the Human Brain. A. NanoRobotics: Current PLoS Comput Biol 2005; 1(4). Approaches and Techniques. 29. Lu J.Q., (2009) 3-D hyperintegration Springer. 2013. and packaging technologies for 40. Kandel E.R., et al. Principles of micro-nano systems. Proc. IEEE. Neural Science. New York. 4 Edn 2009; 97(1): 18–30p. McGraw Hill. 2000. 30. Anderson J.R., et al. The Viking 41. Johnson, G.A., et al. Waxholm viewer for connectomics: scalable Space: An image-based reference for multi-user annotation and coordinating mouse brain research. summarization of large volume data NeuroImage, 2010; 53(2): 365p. sets. J Microscopy: 2011; 241: 13– 42. Fiala J.C., Harris K.M. Extending 28p. unbiased stereology of brain 31. Kleinfeld D., et al. Large-scale ultrastructure to three-dimensional automated histology in the pursuit of volumes. J Am Med Inform Assoc. connectomes. Journal of 2001 Jan-Feb; 8(1):1–16p. Neuroscience 2011; 31(45), 16,125– 43. Wolters Kluwer Health: Lippincott 38p. Williams & Wilkins (2012, August 20). High-definition fiber tractography is major advance in

brain imaging. Science Daily. 52. Helmstaedter M., et al. High- Retrieved July 4, 2013, from accuracy neurite reconstruction for http://www.sciencedaily.com high-throughput neuroanatomy. Nat /releases/2012/08/120820121050.ht Neurosci 2011; 14:1081–8p. m. 53. Livet J., et al. Transgenic strategies 44. Fernandez-Miranda, J.C., et al. High- for combinatorial expression of definition fiber tractography of the fluorescent proteins in the nervous human brain: neuroanatomical system. Nature 2007; 450 (7166): validation and neurosurgical 56–62p. applications. Neurosurgery. 2012 54. Lichtman J., et al. A technicolour Aug; 71(2):430–53p. approach to the connectome. Nature 45. General Electric Phoenix Nanoatom Reviews Neuroscience. 2008; 9 (6): S. 2013. 417–422p. Website:http://www.ge- 55. Zador A.M., et al. Sequencing the mcs.com/en/radiography-x-ray/ct- Connectome. PLoS Biol. 2012; computed-tomography/nanotom- 10(10): e1001411p. s.html 56. Withers P.J. X-ray nanotomography, 46. Tkachuk A, et al. X-ray computed Materials today. 2007; 10(12): 26– tomography in Zernike phase 34p. contrast mode at 8 keV with 50 nm 57. Helmholtz Association New resolution using Cu rotation anode microscope reveals ultrastructure of X-ray source. Zeitschrift für cells. Public Release: 19 Nov 2010 Kristallographie. 2007; 222(11) http://www.eurekalert.org/pub_releas 650–5p. es/2010-11/haog-nmr111910.php 47. Biswal, B.B., et al. Toward 58. Bock D.D., et al. Network anatomy discovery science of human brain and in vivo physiology of visual function. Proceedings of the cortical neurons. Nature. National Academy of Sciences 2010; 2011;471:177–82p. 107(10): 4734–40p. 59. Allen Brain Atlas. Website: ©2014 48. Nishimoto S., et al. Reconstructing Allen Institute for Brain Science. Visual Experiences from Brain Allen Human Brain Atlas. Activity Evoked by Natural Movies. Available from: http://human.brain- Current Biology. 2011; 21: 1641–6p. map.org/. 49. Hayworth K.J., et al. Automating 60. Hawrylycz M.J. An anatomically the collection of ultrathin serial comprehensive atlas of the adult sections for large volume TEM human brain transcriptome. Nature. reconstructions. Microscopy and 2012; 489, 391–9p. Microanalysis 2006; 12 (Suppl 2): 61. Allen Mouse Atlas. Website: ©2013 86–7p. Allen Institute for Brain Science. 50. Anderson J. M., et al. Physiology Allen Developing Mouse Brain and Function of the Tight Junction. Atlas. Available from: http://developi Cold Spring Harbor Perspectives in ngmouse.brain-map.org. Biology. 2009; 1(2), 2584p. 62. Chung K., Deisseroth, K. CLARITY 51. Hayworth K.J. Electron imaging for mapping the nervous system. technology for whole brain neural Nature Methods.2013; 10(6), 508– circuit mapping; International 513p. Journal of Machine Consciousness. 63. Chung K., et al. Structural and 2012; 4(1):87–108p. molecular interrogation of intact

biological systems; Nature 2013; fear memory recall. Nature. 2012; 497: 332–337p. 484, 381–385. 64. White, J.G., et al. The structure of 75. Fuhrmann G., et al. Coding of the nervous system of the nematode Temporal Information by Activity- Caenorhabditis elegans. Phil Trans R Dependent Synapses The Journal of Soc Lond B. 1986; 314: 1–340p. Neurophysiology. January 2002; 65. Cheng D., et al. Relative and 87(1): 140–8p. Absolute Quantification of 76. Rollenhagen A., Lübke J.H.R. The Postsynaptic Density Proteome morphology of excitatory central Isolated from Rat Forebrain and synapses: from structure to function, Cerebellum. Molecular & Cellular Cell Tissue Res. 2006; 326: 221–37p. Proteomics. 2006; 5:1158–70p. 77. Rollenhagen A., et al. Structural 66. Varshney L.R., et al. Structural determinants of transmission at large Properties of the Caenorhabditis hippocampal mossy fiber synapses. elegans Neuronal Network, PLoS J. Neurosci. 2007; 26; 27(39): Comput Biol. 2011 February; 7(2). 10434–44p. 67. Bumbarger D.J., et al. System-wide 78. Sabatini B.L., Regehr W.G. Timing Rewiring Underlies Behavioral of neurotransmission at fast synapses Differences in Predatory and in the mammalian brain. Nature. Bacterial-Feeding Nematodes. Cell. 1996; 384:170–2p. 2013; 152(1): 109–19p. 79. Südhof T.C. The synaptic vesicle 68. Kaynig V., et al. Large-Scale cycle. Annual Review of Automatic Reconstruction of Neuroscience 2004; 27: 509–47p. Neuronal Processes from Electron 80. Polavieja G. G., et al. Stimulus Microscopy. Medical image analysis. History Reliably Shapes Action 2013; 16(1). Potential Waveforms of Cortical 69. Amunts K., et al. BigBrain: An Neurons. J Neurosci. 2005 Jun 8; Ultrahigh-Resolution 3D Human 25(23):5657–65p. Brain. Model Science. June 2013; 81. Kole M.H., et al. Action potential 340(6139): 1472–5p. generation requires a high sodium 70. OCP -Open Connectome Project. channel density in the axon initial 2014. Website: segment. Nat Neurosci. 2008; 178– http://www.openconnectomeproject. 86p. org 82. Sanford L. He axon hillock and the 71. Black J.E., et al. Learning causes initial segment; The Journal of Cell synaptogenesis, whereas motor Biology, 1968; 38, 193–201p. activity causes angiogenesis, in 83. Palmer, L.M., Stuart, G.J. Site of cerebellar cortex of adult rats. Proc. Action Potential Initiation in Layer 5 Nati. Acad. Sci. USA, 87: 5568–72p. Pyramidal Neurons. The Journal of 72. Bliss, T.V., et al. Collingridge A Neuroscience, February 8, 2006, synaptic model of memory: long- 26(6):1854–63p. term potentiation in the 84. Colbert, C.M., et al. Axonal Action- hippocampus. Nature 1993; 361: 31– Potential Initiation and Na+ Channel 39p. Densities in the Soma and Axon 73. Kandel E.R. The molecular biology Initial Segment of Subicular of memory storage: A dialogue Pyramidal Neurons. J Neurosci. between gene and synapses. Science. 1996; 16(21): 6676–86p. 2001; 294(5544): 1030–8p. 85. Clark B.A., et al. The site of action 74. Liu X., et al. Optogenetic stimulation potential initiation in cerebellar of a hippocampal engram activates

purkinje neurone. Nature neurons in copepod antennules. J Neuroscience. 2005; 8: 137–9p. Comp Physiol A. 2004; 190: 877– 86. Connors B.W., Gutnick, M.J. (). 82p. DOI 10.1007/s00359-004-0543- Intrinsic firing patterns of diverse 2. neocortical neurons. Trends in 96. Cressman J.R. Jr, et al. The influence Neuroscience. 1990; 13(3), 99–104p. of sodium and potassium dynamics 87. Tateno T., et al. Threshold Firing on excitability, seizures, and the Frequency–Current Relationships of stability of persistent states: I. Single Neurons in Rat Somatosensory neuron dynamics. J Comput Cortex: Type 1 and Type 2 Neurosci. 2009; 26(2):159–70p. Dynamics; JN Physiol. 2004; 92(4): 97. Fettiplace R. et al. The thickness, 2283–94p. composition and structure of some 88. Steriade M. Neocortical Cell Classes lipid bilayers and natural Are Flexible Entities. Nature membranes. The Journal of Reviews Neuroscience. 2004; 5: 121– Membrane Biology. 1971; 5(3): 277– 134p. 296p. 89. Steriade M., et al. Dynamic 98. Clever H.L. The Hydrated properties of corticothalamic neurons Hydronium Ion. J. Chem. Ed. 1963; and local cortical interneurons 637–41p. generating fast rhythmic (30-40 Hz) 99. Israelachvili J.N. Intermolecular and spike burst. Journal of Surface Forces. Academic Press, NY. Neurophysiology. 1998; 79(1), 483– 2 Edn. 1992. 490p. 100. Lehr L., et al. Electron Solvation in 90. Wang X. J. Fast burst firing and Finite Systems: Femtosecond short-term synaptic plasticity: A Dynamics of Iodide-(Water). Anion model of neocortical chattering Clusters. Science. 23 April 1999; neurons. Neuroscience. 1999; 284: 635–8p. 89(2):347–62p. 101. Dubach J. et al. Abstract 1879: 91. Rubel E.W., Parks, T.N. Novel Intracellular Sodium Organization and development of Nanosensors to Study Sodium brain stem auditory nuclei of the Dynamics in Cardiomyocytes. chicken: tonotopic organization of n. Circulation. 2008; 118: 399p. magnocellularis and n. laminaris. J 102. Berg H.C., et al. Physics of Comp Neurol. 1975; 164:411–33p. Chemoreception. Biophysical 92. Berry M.J., et al. Refractoriness and Journal 1977; (20):193–219p. neural precision. J Neurosci; 1998; 103. Li Y., Gregory S. Diffusion of ions 18 (6): 2200–11p. in sea water and in deep-sea 93. VanRullen R., Thorpe S.J. Rate sediments. Geochimica et Coding vs Temporal Order Coding: Cosmochimica Acta. 1974 (38): 703– What the Retinal Ganglion Cells tell 14p. the Visual Cortex. Neural Comput. 104. Gadani, D.H., et al. Effect of salinity 2001; 13(6): 1255–83p. on the dielectric properties of water. 94. Kuznetsova M.S., et al. Adaptation Indian J. Pure Appl. Phys. 2012; of Firing Rate and Spike Timing (50): 405–10p. Precision in the Avian Cochlear 105. Nano Med Labs Nano Patch-Clamp Nucleus; J Neurosci. 2008; 28(46): (NPC): Development of a High- 1190615p. Throughput System Based On Planar 95. Fields D.M., Weissburg M.J. Rapid Patch-Clamp. 15 July 2010. firing rates from mechanosensory

Website: http://www.nanomed.unige. functional connectivity. Neuroimage. it/Projects/NPC.htm 2012; 62(2), 864–870p.doi: 106. Electronic Bio Sciences. High 10.1016/j. neuroimage. 2012.01.016. performance single ion channel 118. Weaver J.C., Astumian R.D. The recordings at a low cost. 2013. Response of Living Cells to Very Website: Weak Electric Fields: The Thermal http://www.electronicbio.com. Noise Limit. Science. 1990; 107. Novak P., et al. Nanoscale-Targeted 247:459–62p. Patch-Clamp Recordings of 119. Homes C. C., et al. Optical Response Functional Presynaptic Ion Channels. of High-Dielectric-Constant Neuron. 2013; 79: 1067–77p. Perovskite-Related Oxide. Science. 108. Electronic BioSciences. 2014. 27 July 2001; 293: 673–6p. Website: 120. Gutman G.A., et al. Nomenclature www.electronicbio.com/Nanopatch/ and Molecular Relationships of 109. Veitinger S. The Patch-Clamp Voltage-Gated Potassium Channels. Technique: An Introduction. Pharmacol Rev. 2005; 57: 473–508p. Website:http://www.leica- 121. Kole M.H., et al. Axon Initial microsystems.com/science-lab/the- Segment Kv1 Channels Control patch-clamp-technique2011/ Axonal Action Potential Waveform 110. DeFelice L.J. Introduction to and Synaptic Efficacy. Neuron. 2007 Membrane Noise. Plenum Press, Aug 16; 55(4):633–47p. New York. 1981. 122. Colley B.S. et al. TRKB increases 111. Lauger P. Current noise generated by Kv1.3 ion channel half-life and electrogenic ion pumps. Eur Biophys surface expression; Neuroscience. J. 1984; 11(2): 117–28p. 2007; 144(2): 531–46. 112. Hô N., Destexhe A. Synaptic 123. Smyth J.W., Shaw R.M. Forward background activity enhances the trafficking of ion channels: What the responsiveness of neocortical clinician needs to know. Heart pyramidal neurons. J. Neurophysiol. Rhythm. 2010; (8):1135–40p. 2000; 84, 1488p. 124. Angelides K.J. et al. Distribution and 113. Li C., et al. Chaos in a three-neuron lateral mobility of voltage-dependent hysteresis hopfield-type neural sodium channels in neurons. J. Cell networks. Physics Letters A, 2001; Biol. 1988; 106: 1911–25p. 285: 368–72p. 125. De Vittorio M. et al. 114. Gerstner W., Kistler W.M. Spiking Nanotechnology and Neuroscience: Neuron Models: Single Neurons, Nano-electronic-13 Photonic and Populations, Plasticity, Cambridge Mechanical Neuronal Interfacing. University Press. August 2002. Springer. 2014. 115. Brunetti, R., et al. Shot noise in 126. Cohen-Karni T., et al. Synthetically single open ion channels: A Encoded Ultrashort-Channel computational approach based on Nanowire Transistors for Fast, atomistic simulations. Journal of Pointlike Cellular Signal Detection. Computational Electronics. 2007; Nano Lett. 2012; 12; 2639−44p. 6(1): 391–394p. 127. Tian B., et al. Three-dimensional, 116. Serletis D., et al. Complexity in flexible nanoscale field-effect neuronal noise depends on network transistors as localized bioprobes. interconnectivity. Ann Biomed Eng. Science. 2010; 329(5993): 830–4p. 2011; 39(6): 1768–78p. 128. Chen K., et al. Silicon nanowire 117. Birn, R.M., () The role of field-effect transistor-based physiological noise in resting-state biosensors for biomedical diagnosis

and cellular recording investigation. Electric Field Mapping. Biophysical Nano Today. 2011; 6(2): 131–54p. Journal. 2007; 93(4), 1163–74p. 129. Duan X., et al. Nanoelectronics 140. Dergan A. Electronic and transport Meets Biology: From New properties of carbon nanotubes. Nanoscale Devices for Live-Cell University of Ljubljana. Faculty of Recording to 3D Innervated Tissues; Mathematics and Physics. 2010. Chem. Asian J. 2013; 8: 2304–4p. 141. Uryu S. Electronic states and 130. Shimada T., et al. Double-wall quantum transport in double-wall carbon nanotube field-effect carbon nanotubes. Physical Review transistors: Ambipolar transport B. 2004; 69: 075402p. characteristics. Appl. Phys. Lett. 142. Bacsa R.R. Narrow diameter double- 2004; 84: 2412p. wall carbon nanotubes: synthesis, 131. Duan X., et al. Intracellular electron microscopy and inelastic recordings of action potentials by an light scattering. New Journal of extracellular nanoscale field-effect Physics. 2003; 5: 131.1–131.9p transistor. Nat Nanotechnol. 2011; 7: 143. Roman C., et al. Single-Walled 174–9p. Carbon Nanotube Sensor Concepts. 132. Cao D. et al. Electronic sensitivity of Springer Handbook of a single-walled carbon nanotube to Nanotechnology. 2010; 403–25p. internal electrolyte composition. 144. Nick C. et al. Three-Dimensional Nanotechnology. 2012; 23(8): Carbon Nanotube Electrodes for 85203p. Extracellular Recording of Cardiac 133. Yoon I., et al. Intracellular neural Myocytes. Biointerphases. 2012; recording with pure carbon nanotube 7(1–4): 58p. probes, PLoS One. 2013;8(6). 145. Liu S., Guo X. Carbon nanomaterials 134. Aoki K., et al. Fabrication of glass- field-effect-transistor-based coated electrodes with nano and biosensors. NPG Asia Materials. micrometer size by means of NPG Asia Materials. 2012; 4: e23p. dissolution with HF. Electrochimica 146. Lin C.M. Flexible carbon nanotubes Acta. 2010; 55(24): 7328–33p. electrode for neural recording. 135. Cox J.T., et al. Nanoelectrodes: Biosens Bioelectron. 2009; 15; 24(9): Recent Advances and New 2791–7p. Directions. Annu. Rev. Anal. Chem. 147. Mehrabani, S., et al. Hybrid 2012; 5: 253–72p. Integrated Label-Free Chemical and 136. Lambert N.J., et al. A charge parity Biological Sensors. Sensors. 2014 ; ammeter, 2014 American Chemical 14 : 5890–928p. Society, Nano Lett. 2014; 14, 148. Zhang Y. et al. Cytotoxicity Effects 1148−52p. of Graphene and Single-Wall Carbon 137. Fujisawa T. Bidirectional Counting Nanotubes in Neural of Single Electrons. Science. 2006; Phaeochromocytoma-Derived PC12 312(5780): 1634–6p. Cells. ACS Nano. 2010; 4(6): 3181– 138. Park K., et al. Single Molecule 6p. Quantum Confined Stark Effect 149. Bosi S., et al. Carbon nanotubes: a Measurements of Semiconductor promise for nerve tissue engineering? Nanoparticles at Room Temperature. Nanotechnol Rev. 2013; 2(1): 47– ACS Nano. 2012; 6: 10013–23p. 57p. 139. Tyner K. M., et al. Nanosized 150. Silva G.A. Shorting neurons with Voltmeter” Enables Cellular-Wide nanotubes. Nature Nanotechnology. 2009; 4, 82–3p.

151. Voge C.M., Stegemann J.P. Carbon 152. Kim W., et al. Hysteresis Caused by nanotubes in neural interfacing Water Molecules in Carbon applications. J Neural Eng. 2011; 8p. Nanotube Field-Effect Transistors, Nano Letters. 2003; 3(2):193–8p.